Routledge Library Editions: Semantics and Semiology, 14-Volume Set 9781138697508, 9781315520292, 9781138690769, 9781315536712, 0815317425

Table of contents :
Cover
Volume1
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Preface
Acknowledgments
I: Introduction
1 Overview
2 The Lewis/Heim/Kamp Theory
3 Summary
Notes
II: An Overview of Semantic Analyses of Questions
1 Introduction
2 Categorial Approaches
3 Embedding Approaches
4 Propositional Approaches
5 Summary
Notes
III: A Nonquantificational Analysis of Wh-Phrases, I: Parallels Between Wh-Phrases and Indefinites
1 Introduction
2 The Quantificational Variability of Wh-phrases
3 Wh-Clauses as Restrictive Terms
4 Deriving the Restrictive Term: Presupposition
5 The Presuppositional Vagueness of Clause-embedding Predicates
6 Recapitulation
Notes
IV: Exhaustiveness
1 Introduction
2 (Weak) Exhaustiveness
3 The Domain of Quantification
4 Summary
Notes
V: A Nonquantificational Analysis of Wh-Phrases, II: Asymmetries Between Wh-Phrases and Indefinites
1 Introduction
2 Data and Generalizations
3 Wh-Movement as a Determinant of Wh-Phrase Quantifiability
4 Nonquantifiable Wh-phrases
5 Summary
Notes
Appendices
A: Wh-Phrases as Variables Over Functions
Notes
B: Definiteness and Global Accommodation
Notes
C: Complement If- and Whether-Claues
Bibliography
Index
Volume2
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Table of Contents
Foreword
Part I: Understanding and Misunderstanding
I: What are Words For?
II: The Key to Understanding
III: Some Causes of Misunderstanding
IV: Vague Words and Exact Words
V: Let us Look at Some Words
VI: Words That have Changed
VII: International Semantics
Part II: The Misuse of Language
VIII: Advertising
IX: What is Sentimentality?
X: What I Tell You Three Times is False
XI: Whitewash Language
XII: Language of Quarrels
XIII: Lying to Ourselves
Part III: Literary Semantics
XIV: Rhetoric
XV: Metaphor and Association
Index
Volume3
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Table of Contents
Preface
1: Introduction
1.1 Logical analysis
1.2 Ordinary language description
1.3 The role of semiotic
2: History of semiotic
2.1 The Classical tradition
2.2 Augustine and his successors
2.3 Peirce and Saussure
2.4 Behavioral semiotic
2.5 Semiotic's critics
3: Natural signs
3.1 Signs and evidence
3.2 Images
3.3 Natsigns: some basic features
3.4 Dynamic interpretation
4: Communication
4.1 Communicative intent
4.2 Conventional signs
4.3 Signals
4.4 Features of communicative systems
5: Language
5.1 The role of subjects
5.2 Denotation and reference
5.3 Meaning, truth and illocutionary force
5.4 Addresses
5.5 Discourse
Postscript
Notes
Name index
Subject index
Volume4
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Preface
1: Introduction
2: Bare Plurals and Genericity
1 Introduction
2 Genericity and the Readings of Bare Plurals
2.1 Genericity
2.2 Bare Plurals
3 The Kind Analysis
3.1 Basic Outline
3.2 Arguments for a Unified Treatment
3.2.1 Bare Plurals Are Unambiguous
3.2.2 Bare Plurals and Indefinite NP's
3.2.3 A Generic Operator and Bare Plurals
3.3 Limits and Limitations of Uniformity
3.4 Summary
4 The Indefiniteness Analysis
4.1 Basic Outline
4.2 Individual-Level and Stage-Level Predicates
4.3 The Generic Operator
4.3.1 Implicit Domain Restrictions
4.3.2 Modal Dimension
4.4 Conclusion
Appendix
Notes
3: Functional Reading of Bare Plurals
1 Introduction
2 The Functional Reading of Bare Plurals
2.1 Initial Observations
2.2 Functional Reading with Individual-Level Predicates
2.2.1 Genericity and the Functional Reading
2.2.2 Contextual Restrictions
2.3 Functional Reading with Stage-Level Predicates
2.4 Other Indefinites
2.5 Excluding a Purely Pragmatic Account
2.5.1 The Implicature Approach
2.5.2 The Referentiality Approach
2.6 Functional Reading in Quantified Contexts
2.6.1 Dependent Functional Reading
2.6.2 Quantificational and Modal Subordination
2.7 The Presupposition of Existence
2.7.1 Simple Cases
2.7.2 Projection of the Existential Presupposition
2.7.3 Functional Reading with Adverbs of Quantity
2.8 Overview
3 An Operator Analysis
3.1 Degenerate Genericity
3.2 Degenerate Genericity and the Functional Reading
3.2.1 Implicit Contextual Restrictions
3.2.2 Implication of Existence
3.2.3 Scopal Interaction
3.3 Stage-Level Predicates
3.4 Positive Contextual Sensitivity
3.5 Summary
4 Conclusion
Notes
4: Strong and Weak Novelty
1 Introduction
2 The Dynamic View on Meaning
2.1 Assertions and Contextual Update
2.2 Presuppositions and Contextual Admittance
3 The Novelty-Familiarity Theory of Definiteness and Indefiniteness
3.1 Files as Information States
3.2 Informativeness of Files
3.3 The Felicity Conditions of Definites and Indefinites
3.4 The Truth Conditions for Definites and Indefinites
4 The Functional Reading
4.1 Weakly and Strongly Novel NP's
4.2 Contextually Salient Functions
4.3 Negative Contextual Sensitivity
4.4 Strong and Weak Novelty and NP Strength
4.5 Existential Force and Strong vs. Weak Novelty
4.6 Maximality
4.7 Consequences of the Existential Presupposition
4.7.1 Positive Contextual Sensitivity
4.7.2 Dependent Functional Reading
5 Conclusion
Notes
Bibliography
Index
Volume5
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Acknowledgments
1: The philosophical context
2: Structuralism
3: Semiology as a science of signs
4: S/Z
5: Marxism, language, and ideology
6: On the subject of Lacan
7: The critique of the sign
8: Conclusion
Bibliography
Index
Volume6
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Prefece
Acknowledgements
Table of Contents
1: Prepositions and Verbs
1.1 Introduction: Methodological Preliminaries
1.1.1 Motivations for Lexical Representations
1.1.2 Lexical Semantics and the Grammar
1.2 Terminological Preliminaries
1.3 Prepositions
1.4 The Problem for Logical Form: A First Solution
1.5 Some more Preposition Meanings
1.5.1 Individuating Preposition Meanings
1.6 A Proposal Concerning English Prepositions
1.7 Lexical Representations
1.8 The logic of Frames
1.9 Core Participant and Adjuncts
1.9.1 Preposition Meanings
1.9.2 Direct Inheritance
1.9.3 The Right Theory
1.10 Revising Semantic Compatibility
1.11 Conclusion
2: Figure and Ground Scenes
2.1 Introduction
2.2 A Commercial Event
2.2.1 Selling and Possessive TO
2.2.2 Buying and Source
2.3 A New Definition of Semantic Compatibility
2.3.1 Commercial Events Revisited
2.4 Lexical Representations and Knowledge Representations
2.4.1 The New Proposal
2.4.2 Some Technical Revisions and Stipulations
2.5 The Preposition for and More Commercial Events
2.5.1 Buy, sell and Pay
2.5.2 A Note on Polysemy
2.5.3 Ex-spending the Ontology
2.6 Other Motivations for Figure and Ground
2.7 Conclusion
3: Where are the Case Theories of Yesteryear?
3.1 Why Case Grammar?
3.2 Some Facts about the Semantics of Grammatical Relations
3.2.1 Active and Causal Subjects
3.2.2 Change and Effective Instruments
3.2.3 Experiencers and Causers
3.2.4 Summary
3.3 The Hierarchy and its Function
3.4 Some Issues in the Semantics of Nuclear Terms
3.4.1 Cross-Categorial Claims
3.4.2 Optionality and Instruments
3.4.3 Kajita's Quandary and Subcategorization
3.4.4 A note on Figure, Ground, and POS_TRANS
3.5 The Semantics of Valence
3.5.1 A Hierarchy for Verbs
3.5.2 Subcategorization
4: Valence Alternations
4.1 Optional Advancement
4.1.1 Instrument Promotion
4.1.2 Dative Movement
4.2 Advancement Involving Change
4.2.1 Goal Promotion
4.2.1.1 The Patient Valence
4.2.1.2 The Goal Valence
4.2.2 Other Rules
4.2.2.1 Container Promotion
4.2.2.2 Symmetric Predicates
4.2.2.3 Origin Promotion
4.4 Conclusion: Semantic Options versus Lexical Rule
5: Predication, Control and Lexical Rules Revisited
5.1 Infinitival Complements
5.2 Predicative PP's
5.3 Frame Representations of Predicative PP's
References
Volume7
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Introduction
1: A Background to Semiotics: Saussure and Peirce
2: Semiotics as a Behavioural Theory: Charles Morris
3: Semiotics as a Theory of 'l' Acte Sémique': Luis Prieto
4: Semiotics as a Theory of 'Speech Acts': Austin and Searle
5: Semiology as an Ideology of Socio-Cultural Signification: Roland Barthes
6: Semiology as a Theory of Semiological Systems and of Indices: Functionalism
7: An Integrated Theory of Semiotics: Axiomatic Functionalism
8: Semiotics as a Stylistic Theory: Bureau and Riffaterre
9: Semiotics of the Cinema: Christian Metz
10: Zoo-Semiotics
References
Index
Volume8
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Preface
Table of Contents
Preface
Chapter 1: Adverbial Modification and the Interpretation of Distributive Predicates
0. A Few Preliminaries
1. The Distributive-Collective Dichotomy
2. "Intrinsic" Distributives and Collectives
3. Distributive Predication as Predication on Groups
4. Group-Sensitive Adverbials
5. Conjoining Collective and Distributive Verb Phrases
Chapter II: Group Action and Spatio-Temporal Proximity
0. Introduction
1. Vagueness and Ambiguity
2. The "Proximity"uses of Group-Sensitive Adverbials
3. Semantic Effects of Adverbial Position
4. Quantifier Scope
5. Event Structure and Group Action
6. Fragment 2
7. Examples
8. Locative and Temporal Readings
9. Adjectival and Object-Oriented Readings
10. Distributive Predicates Revisited
11. Conclusion
Chapter III: The Algebra of Groups and the Algebra of Events
0. Introduction: Tightening up the Model
1. The Algebra of Groups
2. A Note on Respectively Constructions
3. Committees, etc.
4. The Algebra of Events as a Semilattice
5. Restricting the Structure of Events
6. Unaugmented and Augmented Frames
Chapter IV: Quantifiers, Group-Level Properties and Agreement
0. Are Quantifiers Sensitive to Event Structure?
1. Different Types of Group-Level Events
2. Event-Sensitive Quantification
3. Additivity, Inclusion and Exclusion
4. An Argument-Restriction Account of Agreement
5. Some Problems
6. Agreement and The Inclusion/exclusion Distinction
7. Distributive Determiners and Plural Agreement
8. Fewer than Two and More than One
9. Remaining Problems
References
Volume9
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Table of Contents
Preface
Acknowledgments
1: The Problem
1.1 Introduction
1.2 Facts to be Accounted For
1.2.1 The Definiteness Effect
1.2.2 The Predicate Restriction
1.3 Previous Analyses
1.3.1 There-Insertion and Its Descendants
1.3.2 Previous Characterizations of the DE
1.3.3 The Predicate Restriction
1.4 Outline of Remaining Chapters
Notes
2: Existential Syntax
2.1 Introduction
2.2 Identifying VP-Adjuncts
2.2.1 Description
2.2.2 A Proposal for Depictive Adjuncts
2.3 Why a DP-External XP Must Be Posited
2.3.1 Arguments for Independence
2.3.2 Williams' Counterarguments
2.4 Supporting the Adjunct Analysis
2.4.1 Stowell's Small Clause Diagnostics
2.4.2 Extraction
2.5 Chapter Summary
Notes
3: The Existential, Descriptions, and Instantiation
3.1 Introduction
3.2 Data to Be Accounted For
3.2.1 Quantificational DPs
3.2.2 Scope
3.2.3 Contact Clauses
3.2.4 Relative Clauses
3.3 Property Theory
3.3.1 The Syntax of PT
3.3.2 The Interpretation of PT
3.4 Interpreting the Existential
3.4.1 A Dynamic PT Fragment of English
3.4.2 Interpreting Existential Sentences
3.5 Definites and the Other Half of the DE
3.5.1 Augmenting the Fragment
3.5.2 A Felicity Condition on Existentials
3.5.3 Summary
3.6 Advantages of the Analysis
3.6.1 Acceptability of Quantified Kind DPs
3.6.2 Quantification and Scope
3.6.3 Contact Clauses
3.6.4 Relativization
3.6.5 Summary
Notes
4: Adjunct Predicates and the Predicate Restriction
4.1 Introduction
4.2 Interpreting Depictive Adjuncts
4.2.1 The Adjunct Rule
4.2.2 Individual/Stage Sensitivity
4.3 Extending the Analysis
4.3.1 Nominalized Functions as Controllers
4.3.2 The Predicate Restriction Revisited
4.4 Eventive Existentials
4.4.1 The Problem
4.4.2 Eventive Participles Are Not Adjuncts
4.4.3 Why These Existentials Are Not Passives
4.4.4 A Proposal
4.5 Chapter Summary
Notes
5: Some Final Remarks
5.1 Introduction
5.2 Definites in the Existential
5.2.1 List Existentials
5.2.2 Focus and the Existential
5.3 Other Expectations
5.3.1 Other Expletive There Sentences
5.3.2 The Existential and Other Indexicals
Notes
References
Index
Volume10
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Acknowledgements
Table of Contents
Introduction
1: A Reichenbachian Tense Logic
1.1 Introduction
1.2 Triple Dependence
1.3 What is Reference Time
1.4 The Logic
1.5 The Preterite and the Indexical Interpretation of Reichenbach
1.5.1 The Motivation for Indexical Treatments of Tense
1.5.2 Vagueness and Indexicality
1.5.3 Indefinite Reference to Time
1.6 Temporal Reference in Connected Discourse
1.6.1 The Treatment of Aktionsarten
1.6.2 Time in Connected Discourse
1.6.3 Some Special Uses of the Preterite
1.7 The Need for Three Indices
1.7.1 The Need for at Least Three Indices
1.7.2 More than Three Indices
Notes
2: A Tense Logical Sketch of German
2.1 Frame Adverbials
2.2 Baeuerle and Stechow's Analysis
2.3 The German Present Tense
2.3.1 Semantics of the German Present Tense
2.3.2 Sample Derivation of Complex Truth Conditions
2.3.3 Why Atelics are Presumed to Refer to Speech Time
2.3.4 Kratzer's Speech Time Pragmatics
2.3.5 The Nonambiguity of the Present Tense
2.3.6 Conclusion
2.4 Duratives
2.5 Frist Adverbials
2.6 (Temporal) schon
2.6.1 Preliminaries
2.6.2 Other Uses of schon
2.6.3 The Truth Conditions of Temporal schon
2.6.4 schon with Telic Aktionsarten
2.7 Summary of Semantic Rules
Notes
3: A Fragment of German
3.1 GPSG: Formalism and Notation
3.2 German Syntax
3.2.1 Constituents of the Sentence
3.2.2 Fronting (of Several Kinds)
3.2.3 Phantoms and Some Recalcitrant Sorts of Fronting
3.3 Basic Rules
3.3.1 Features for Complements
3.3.2 Separable Prefix Verbs
3.4 Fronting Formalized
3.5 The Analyses of Jean Fourquet
3.6 Two Strategies for the Treatment of Temporalia (in GPSG)
3.6.1 Tense as a Verb Operator
Excursus: On Duratives (and Frequentatives) as Complements to the Verb
3.6.2 Tenses as Phrasal Operators
3.7 Metarules for Temporalia
3.7.1 Duratives and Frist Adverbials
3.7.2 Frame Adverbials (that Modify Reference Time
3.8 Some Derivations
3.9 The Syntax of Temporal Schon
Notes
4: Extending the Fragment
4.1 The Perfect Tenses
4.1.1 The Forms of the Perfect
4.1.2 The Meaning of the Perfect Tenses
4.1.3 The Syntax of the Perfect
4.1.4 A Sample Derivation
4.2 Adverbials which Modify Event Time
4.3 Noch
4.3.1 Nontemporal Noch
4.3.2 Immer Noch
4.3.3 The Syntax of Noch
4.4 Passives
4.4.1 The Subjectlessness of Impersonal Passives
4.4.2 The Lexical Nature of the Passive
4.4.3 A Formulation of the Rule
Notes
References
Volume11
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Preface
1: The Possibility of a Theory of Word Meaning
2: Against Semantic Primitives
3: Naïve Metaphysics
4: Theories of Categorisation
5: Verbs, Prototypes and Family Resemblances
6: Semantic Categories
Bibliography
Index
Volume12
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Table of Contents
Dedication
Preface
1: Structural Relations and Restrictions
1.1 Syntactic Domains
1.2 C-Command
1.3 Domain Restrictions on Interpretative Rules
2: Coreference of Definite NPs
2.1 Theoretical Assumptions
2.2 The Non-relevance of 'Precede-and-Command'
2.3 The C-Command Rule
2.4 A Comparison Between the C-Command and the Precede-and-Command Rules
2.5 Coreference in Sentences with Extraposed Clauses
2.6 PPs and Indirect Objects
2.7 Coordinate Structures
3: Prepositional Phrases and Preposed Constituents
3.1 Sentential and Verb-phrasal Prepositional Phrases
3.2 Preposed PPs
3.3 Topicalisation and Left-dislocation
3.4 Summary
4: A Survey of Functional Approaches to Definite NP Anaphora
4.1 Discourse-oriented Approaches
4.2 Relational-grammar Approaches
4.3 Semantic Approaches
5: Bound Anaphora
5.1 Quantified Antecedents
5.2 Reflexive and Reciprocal Pronouns
6: The Indexing System of Interpretative Semantics
6.1 A summary of the Anaphora Conditions
6.2 The Indexing System
7: The Interpretation of Pronouns: A Restatement of the Anaphora Problems
7.1 The Problems with the Current Anaphora Picture
7.2 Coreference and Bound Anaphora
7.3 The Coindexing Procedure and the Interpretation of Coindexing
7.4 Coreference
7.5 Summary
8: Unsolved Problems of Anaphora
8.1 PP Problems
8.2 Possessive NPs
8.3 'Experiencing' Verbs
9: Other Interpretative Rules
9.1 Function-argument Representations
9.2 Relative Scope of Quantifiers
9.3 Theme-rheme Relations
10: The Psychological Reality of the C-Command Conditions
10.1 Syntactic Rules
10.2 Processing Strategies
References
Index
Volume13
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Table of Contents
Preface
1: Introduction
1.1. Overview
1.1.1. What this dissertation is about
1.1.2. Context change: Pragmatic vs. semantic approaches
1.2. The Stalnakerian Model of Presupposition and Assertion
1.3. Dynamic Semantics
1.3.1. The Fundamentals
1.3.2. File Change Semantics (FCS) and the CCP proposal
1.3.3. Discourse Representation Theory (DRT
1.3.4. Dynamic Montague Grammar (DMG)
Notes to Chapter One
2: Disjunctive Sentences in Discourse
2.1. Introduction
2.2. The Discourse Function and Felicity Conditions of Disjunction
2.2.1. The Basic Observations
2.2.2. Relation and Manner in the Stalnakerian Model
2.3. The Enriched Context Change Framework
2.3.1. Presentation
2.3.2. Disjunction and the Relevant Informativity Condition
2.3.3. Disjunction and Simplicity
2.3.4. Disjunction and Rooth's Alternative Semantics
2.3.5. Summary
2.4. Some Exceptions
2.4.1. Floutings
2.4.2. Reasoning contexts
2.4.3. Metalinguistic or
2.5. The Exclusive Interpretation of or
2.5.1. Critique of the ambiguity account
2.5.2. Gazdar's (1979) account
2.5.3. Exclusivity from exhaustiveness
2.5.4. Exclusivity from alternativeness
2.5.5. Summary
2.6. Conclusion
Notes to Chapter Two
3: Presupposition Projection
3.1. Introduction
3.1.1. The Basic Question
3.1.2. The Theoretical Issues
3.2. The Data
3.3. The Satisfaction Account of Presupposition Projection
3.3.1. Basics of the Satisfaction Account
3.3.2. CCPs for disjunction
3.3.3. Critique
3.4. Towards a New Account
3.4.1. Gazdar's cancellation theory
3.4.2. The accommodation view: Van der Sandt (1992)
3.4.3. Translating DRSs into Stalnakerian contexts
3.5. The Account in Action
3.5.1. Basic cases
3.5.2. Entailing disjunctions again
3.5.3. Beaver's counterexample
3.6. Conclusion
Notes to Chapter Three
4: Internal Anaphora
4.1. Introduction
4.2. Anaphora-based Accounts in Dynamic Semantic Theories
4.2.1. DMG: Groenendijk and Stokhof (1990)
4.2.2. DRT: Kamp and Reyle( 1993)
4.2.3. A second DRT proposal: Krahmer and Muskens (1994)
4.2.4. Van der Sandt (1992) revisited
4.3. A Felicity-based Approach
4.3.1. Introduction to the account
4.3.2. The E-type account of anaphora
4.3.3. A felicity-based solution to the internal anaphora puzzle
4.3.4. Summary
4.4. Further Data
4.4.1. Narrow scope antecedents
4.4.2. Non-E-type unbound anaphora
4.4.3. Pleonastic pronouns
4.5. Conclusion
Notes to Chapter Four
5: External Anaphora
5.1. Introduction
5.2. The Basic Data
5.2.1. Anaphora to a disjunction of NPs
5.2.2. Clausal disjunction
5.3. A First Reformulation of the E-type Account
5.4. A Compositional Structural E-type Account
5.4.1. Presentation
5.4.2. Comparison with Chapter Four account
5.4.3. Narrow scope antecedents
5.5. Application to the External Anaphora Data
5.5.1. Anaphora to a disjunction of NPs
5.5.2. Anaphora to clausal disjunctions
5.5.3. Summary
5.6. The Single-antecedent Reading
5.6.1. Derivation of single-antecedent readings
5.6.2. Maximal quantifier antecedents
5.6.3. Summary
5.7. Other Approaches to External Anaphora
5.8. Further Issues for the E-type Account
5.8.1. Inference-based anaphora
5.8.2. Interpretation of plural pronouns
5.8.3. Anaphora to NP disjuncts and conjuncts
5.9. Conclusion
Notes to Chapter Five
Concluding Remarks Grice, Stalnaker and Dynamic Semantics
Appendix
Bibliography
Index
Volume14
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright Page
Dedication
Acknowledgements
Table of Contents
List of Tables
List of Figures
1: Introduction
2: Previous Studies
2.1. Introduction
2.2. Topic-Based Studies of NP Proposing
2.2.1. Halliday
2.2.2. Gundel
2.2.3. Langacker
2.2.4. Rodman
2.2.5. Creider
2.2.6. Bland
2.2.7. Reinhart
2.2.8. Davison
2.3. Non-Topic-Based Studies of NP Preposing
2.3.1. Chafe
2.3.2. Clark and Clark
2.3.3. Prince
2.4. Studies of VP Preposing
2.5. Summary
3: A Theory of Proposing
3.1. Introduction
3.2. Definitions
3.2.1. SCALES
3.2.2. BACKWARD LOOKING CENTERS
3.2.3. OPEN PROPOSITION and FOCUS
3.3. The Functions of Preposing
3.4. Identifying the Open Proposition and Focus
3.4.1. Stress and Focus
3.4.2. Identifying Possible Foci
3.5. Summary
4: Applying the Theory
4.1. Introduction
4.2. The Data
4.3. The Analysis
4.3.1. Preposing and Informality
4.3.2. Proposing and Indefiniteness
4.3.3. Preposing and Specificity
4.3.4. Proposing arid Root Transformations
4.4. A Taxonomy of Proposing
4.5. FOCUS PREPOSING
4.5.1. GENERAL FOCUS PREPOSING
4.5.2. ECHOING
4.5.3. YIDDISH-MOVEMENT
4.6. Summary
5: General Topicalization
5.1. Introduction
5.2. Categorial Restrictions on TOP?
5.2.1. Types of Scalar Relations
5.2.2. Scalar Values
5.2.3. Salience of the Scale
5.3. [+IDENTITY] TOP
5.3.1. ADVERBIAL PREPOSING
5.3.2. NPs and [+IDENTITY] TOP
5.3.3. Bridging via NPs
5.4. Syntactically Distinguishable Types of TOP
5.4.1. INDIRECT QUESTION PREPOSING
5.4.2. 'IF' PREPOSING
6: Proposition Affirmation
6.1. Introduction
6.2. General PROPOSITION AFFIRMATION
6.3. PROPOSITION AFFIRMATION with Modals
6.4. Exclamative PROPOSITION AFFIRMATION
6.5. 'THAT'-TENSE PREPOSING
6.6. A Comparison of PA-Performing Constructions
6.7. 'IT IS' PREPOSING
6.8. Syntactic Arguments Based on PA
7: Ironic Preposing
7.1. Introduction
7.2. Accommodating IRONIC PREPOSING in the Theory
7.2.1. BACKWARD LOOKING CENTER
7.2.2. OPEN PROPOSITION
7.3. Previous Pragmatic Accounts of Irony
8: Conclusion
References
Index

Citation preview

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 1

ON THE SEMANTICS OF WH-CLAUSES

ON THE SEMANTICS OF WH-CLAUSES

STEPHEN BERMAN

First published in 1994 by Garland Publishing, Inc. This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1994 Stephen Berman All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69076-9 978-1-315-53671-2

(Set) (Set) (ebk) (Volume 1) (hbk) (Volume 1) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

ON THE SEMANT ICS OF WH-CLA USES

STEPHEN BERMAN

GARLAND PUBLISHING, INc. NEW YORK & LONDON I 1994

Copyright © 1994 by Stephen Berman All rights reserved

Library of Congress Cataloging-in-Publication Data Berman, Stephen, 19610n the semantics of Wh-clauses I Stephen Berman. p. em. - (Outstanding dissertations in linguistics) Includes bibliographical references and index. ISBN 0-8153-1742-5 (alk. paper) 1. Grammar, Comparative and general-Syntax. 2. Semantics. 3. Generative grammar. I. Title. II. Series. P295.B47 1994 415-dc20 93-46330 CIP

Printed on acid-free, 250-year-life paper Manufactured in the United States of America

Contents vii

Preface

xvii

Acknowledgments I Introduction 1 Overview 2 The Lewis/Heim/Kamp Theory 3 Summary ~otes

II An Overview of Semantic Analyses of Questions 1 Introduction 2 Categorial Approaches 3 Embedding Approaches 4 Propositional Approaches 5 Summary ~otes

III A ~onquantificational Analysis of Wh-Phrases, I: Parallels Between Wh-Phrases and Indefmites 1 Introduction 2 The Quantificational Variability of Wh-phrases 3 Wh-Clauses as Restrictive Terms 4 Deriving the Restrictive Term: Presupposition 5 The Presuppositional Vagueness of Clause-embedding Predicates 6 Recapitulation

3 3

5

13 14 17 17 18 20 24 32 33

39 39 40 42 50 59 72 ~

~~

IV Exhaustiveness 1 Introduction 2 (Weak) Exhaustiveness 3 The Domain of Quantification 4 Summary ~otes

v

81 81 82 88 98 99

vi

On The Semantics ofWh-Clauses

V A Nonquantificational Analysis of Wh-Phrases, II: Asymmetries Between Wh-Phrases and Indefmites 1 Introduction 2 Data and Generalizations 3 Wh-Movement as a Determinant of Wh-Phrase Quantifiability 4 Nonquantifiable Wh-phrases 5 Summary Notes

103 103 104 108 128

133 135

Appendices A Wh-Phrases as Variables Over Functions Notes B Defmiteness and Global Accommodation Notes C Complement If- and Whether-Clauses

147 149 154 155

Bibliography

159

Index

173

141

Preface This book contains the essentially unaltered text of my doctoral dissertation in linguistics, submitted to the Graduate School of the University of Massachusetts, Amherst, in January 1991. I have shortened the title (from On the Semantics and Logical Form of Wh-Clauses), both to make it less cumbersome and to place greater emphasis on the fact that an investigation and analysis of the interpretation of whclauses constitutes the original work presented here; while Logical Form is made use of in chapter V, I do not offer new arguments for this level of representation, and, as I acknowledge there, my results could be maintained (in somewhat different form) without appealing to LF. I have also augmented the bibliography with a few references to published versions of works I originally cited in unpublished form; but I have left the original references in the text, since it is these I used in my writing. The text itself contains no substantive changes; I have merely corrected several typographical errors and changed the wording in numerous places to conform to the book format. In the nearly three years since I finished writing the dissertation, I have pursued the development of certain aspects of it. In addition, several works have appeared that respond directly to some of the central points of the research I report on here; I refer specifically to Lahiri (1991), Groenendijk and Stokhof (1992), and Ginzburg (1992; 1993) (see the bibliography for detailed listings of these and other works cited in this preface). In the remainder of this preface I will briefly summarize both my own further research as well relevant aspects of these works and indicate what I take to be their significance for my proposals. This is not the place for a comprehensive review, nor for in depth argumentation or an explicit formal exposition; these are contained in the manuscript, which as I write is still in progress, upon which these cursory remarks are based. Concerning the formal framework of my analysis, which I referred to as "the Lewis/Heim/Kamp theory," I have more recently implemented it directly in the formalism of Discourse Representation Theory, as codifed in Kamp and Reyle (1993). While my original "flat" vii

viii

On The Semantics ofWh-Clauses

representations have, as I pointed out (chapter I, note 7), the advantage of compactness, Discourse Representation Structures are rather more perspicuous. Along with this superficial change I have also adopted the current position of DRT, different from that of Lewis (1975), Kamp (1981), and Heim (1982), that quantifiers-even adverbial ones-are not inherently unselective but rather, by various means (e.g. focus), in a given context always quantify over specified variables. The reason for this has to do with the proportion problem, which I alluded to (chapter I, note 8) but did not attempt to deal with. To account for instances of simultaneous quantification, which constituted the impetus for unselective binding, quantifiers are allowed, however, to "select" different variables, including multiple variables, in different contexts. Such quantifiers are known as polymorphous; Chierchia (1992) provides a basic discussion of these. Using DRSs has also facilitated elaboration of the role of presupposition within my analysis by enabling me to incorporate the presupposition theory of van der Sandt (1992), which is formulated in terms of DRT. Vander Sandt has proposed a variety of pragmatic and structural principles for accounting for the projection behavior of presuppositional material in discourse. Part of this includes the notion of presupposition accommodation, which plays a central role in my analysis. Van der Sandt argues that there is no theoretical distinction between global and local accommodation, rather there is a general preference principle to accommodate as high up in the DRS as possible, while adhering to the pragmatic and structural constraints. It is the interaction of these principles that results in the descriptive difference between global, i.e. top-level, accommodation and accommodation at various lower levels, which is more or less "local". One of van der Sandt's structural principles is an injunction against creating free variables, which he uses to account for presupposition projection in quantified utterances. I have attempted to apply this analysis to the interaction between presupposition and adverbs of quantification. Consider the derivation of the logical representation (27.a), discussed on pages 53-55. Since the nuclear scope contains presuppositional material, according to van der Sandt's theory this should be accommodated as high up as possible. But this presuppositional material is an open sentence, so accommodating it to the top level will result in a structure with free variables. Accommodating it to the next level down, the restrictive term, however, results in the variables being

Preface

ix

bound by the quantifier, yielding the desired-and required-representation. This account thus seems to provide a motivated explanation for the stipulation (or rather, the observation) that the presuppositions of the nuclear scope of a quantifier become part of its restrictive term. The account cannot, however, be complete as given. For one thing, DRT does in fact allow free variables (i.e., discourse referents) at the top level, which may be existentially interpreted or assigned a contextual referent (I briefly allude to this in the discussion of definiteness in appendix B). This possibility manifests itself particularly in the contrast between generic (or habitual) and episodic sentences. I did not bring this into my analysis, though I drew attention to it in the discussion of if-clauses on page 45. But I have since attempted to treat it, by developing the sketchy remarks of appendix B. It seems that in episodic sentences there is what appears to be simultaneous global and local accommodation, the former providing the context set, the latter the restrictive term. However, this possibility does not appear to be directly compatible with van der Sandt's presupposition theory. An alternative, one which I am continuing to explore, is that the quantifier, as a (possibly contextually triggered) lexical property, requires, or selects, certain variables, and that the context set supplies these. This idea fits well with the treatment of quantifiers as polymorphous rather than unselective. Working it out in detail is, however, rather complicated, involving the introduction of variables (or discourse referents) over situations and other eventualities, and of the accompanying semantic apparatus (these are treated at length in Kamp and Reyle (1993, chapter 5)); I refrain from further elaboration at this juncture. Closely connected with the preceding considerations, and also broached in appendix B, is the analysis of plurality. I have attempted to bring this into the scope of my analysis (also prompted by the discussion in Lahiri (1991)-see below), drawing largely on the treatment in Kamp and Reyle (1993). The need to distinguish plural and singular NPs arises from the quantificational behavior of definites, including which-phrases, alluded to at several points in the text. Their semantics requires enriching the model theory by imposing an algebraic structure on the domain of individuals (thus giving these a similar structure to that of situations). But doing so allows a considerable increase in the empirical coverage of the analysis: bringing into focus parallels and asymmetries not only between indefinites and wh-phrases, which my original work emphasized, but also between these phrase types and defi-

X

On The Semantics ofWh-Clauses

nites; and concomitantly encompassing a wider range of clause types, specifically that-clauses; and finally, revealing the effect of presupposition in all these constructions. Thus, the scope of the project is now much broader than the more modest aims of my dissertation; but there remains much work to do in order to flesh out the ideas hinted at here. There are several others aspects of my original analysis that I have since either elaborated or altered; I do not have space here to touch on all of them, but some I will address in the course of discussing the works I mentioned above that respond to it, to which I now turn, starting with Lahiri (1991). This is a quite detailed critical examination of my original analysis and contains numerous suggestions, based on data I had not taken into consideration, for amending it, while staying within its basic assumptions. But on a couple of points Lahiri concludes that my approach is fundamentally flawed, and therefore proposes an alternative analysis. The "friendly" amendments concern data involving presupposition, plurality, and default quantification. His alternative is prompted by flaws he see in my account of the distribution of wh-clauses and by the behavior of certain multiple-wh-clauses. Let me briefly elaborate on these points. Lahiri rightly notes that I neglected to analyse the role of presupposition in examples where factivity is not at issue; in fact, any presuppositions in the nuclear scope become part of the restrictive term (modulo the remarks above concerning global accommodation and context sets). Lahiri cites as examples the predicates be certain about and agree on: the first presupposes something like 'consider it likely that p', where pis the proposition (or open sentence) embedded under be certain about; the second presupposes something like 'believe that p'. These presuppositions (i.e., the open sentences corresponding to them) serve to restrict the adverbial quantifier in a sentence like For the most part, the women are certain about/agree on who came. This is as expected on my analysis. The predicate agree on also brings to attention the need to treat plurality, since it requires a plural subject. Lahiri shows that my original analysis fails to yield intuitively correct truth conditions for sentences involving this predicate under the assumption of a default universal quantifier. But he also shows that incorporating an adequate semantics for plurality, which as I indicated above is in any case re-

Preface

xi

quired and does not necessitate essential changes in my basic analysis, does produce the desired results. The criticism concerning default quantificational force is somewhat specious, since I cite Hintikka's examples of default existential force and explicitly acknowledge (e.g., chapter III, note 4) that universal force is a simplifying assumption on my part. Lahiri himself does not offer a comprehensive alternative, suggesting only that it may be a contextually influenced lexical matter. Nevertheless, prompted by Lahiri's critique, I have examined a range of data and believe that the ambiguity in default quantificational force only arises in the case of wh-clauses with simple wh-words (e.g. who), and not with whichphrases (compare John told me who he saw at the party. with John told me which students he saw at the party.). I do not yet have a fully satisfying theoretical account of this, but the observation is significant, I believe, in light of the considerations that lead Lahiri to seek an alternative to my basic analysis. I return to these after discussing his criticism of my account of the distribution of wh-clauses. One of the principal claims of my analysis is that the interpretation of wh-clauses is dichotomous, being either a question-denoting entity or an open sentence that serve as a quantifier restriction, due to presupposition accommodation (the denotation in a given context arises from the presence of a question operator or a quantifier, conditioned in embedded contexts by lexical properties of the embedding predicates). As Lahiri notes, on this account it would be expected that all factive predicates can embed a wh-clauses (in fact, Hintikka (1976) comes close to proposing this, in terms of his notion of 'success presupposition', which I refer to on page 23). This is not the case (e.g. regret), and the way I avoid this result is by appealing to subcategorization (though I do not dwell on this, but cf. chapter V, note 18). Lahiri points out that Pesetsky (1982) has suggested that subcategorization can be eliminated from the grammar, its job being done by the independently motivated modules of semantic selection and Case theory. But Rothstein (1992) has shown that Pesetsky's account does not succeed with a wide range of data, and concludes that some reference to subcategorization is (still) necessary. (However, Pesetsky (1993) has suggested amendments to his account that he claims counter Rothstein's objections; thus the matter is still open.) Lahiri further observes that using both subcategorization and selection information, as I do, leads to the prima facie expectation of

xii

On The Semantics ofWh-Clauses

six types of clause-embedding predicate (with respect to the distribution of wh-clauses), while only three are realized. This prompts his alternative analysis, according to which wh-clauses are uniformly questiondenoting entities. To account for their distribution under embedding, however, he proposes that the embedding predicates not only select the semantic type of their complement (question or proposition) but also specify with what type they may grammatically combine. This bifurcation produces the same six possible clause-embedding predicates as on my analysis. Lahiri introduces various mechanisms and stipulations to yield only the three occurring types, but upon examination I have concluded that these are no better motivated than the stipulations I have to make in terms of subcategorization. In short, I find that my account of the distribution of wh-clauses remains viable. Finally, let me address Lahiri's most compelling, if somewhat subtle, argument against my analysis. This concerns multiple-whclauses embedded under predicates that do not distribute over conjunctions in the embedded clause. An example is the following sentence: (*)

It surprised Peter that Prof. Jones spoke to Maria and Prof. Garcia spoke to Bill.

One can construct convincing models in which this sentence is true while at the same time it is neither true that it surprised Peter that Prof. Jones spoke to Maria nor that it surprised him that Prof. Garcia spoke to Bill. Moreover, the same models verify the corresponding sentence with a wh-clause: (**)

It surprised Peter who spoke to whom.

Lahiri's claim is that my analysis, on which quantification is over values of the wh-phrases, cannot render these truth conditions. He shows that an analysis on which quantification is not over an algebraically structured domain of individuals but rather over such a structured domain of propositions-namely, the propositions that result from assigning values to the wh-phrases, does yield the desired truth conditions. He goes on to analyse quantificational variability also in these terms. IfLahiri's claim were correct, my analysis would indeed seem to be falsified. However, in conversation Ede Zimmermann has shown

Preface

xiii

that it is possible to get Lahiri's results while maintaining quantification over a domain of individuals, specifically in the case of (**) pairs of individuals. I refrain from going into the technical details in the present context, but the conclusion is that it is not necessary to go to the level of propositions; rather, one can maintain the basic fonn of my analysis in tenns of quantification over (an appropriately structured) domain of individuals. In connection with this argument let me note that Lahiri does not base his claim especially on sentences like(**), but maintains that the same truth conditions bold for corresponding sentences with whichphrases, e.g., It surprised Peter which professors spoke to which students. It is not, however, clear to me that this is so. That is, I'm not sure that there isn't a distributive entailment in this case, in contrast to (**), though to elaborate on this would be too involved for this sketch. Let me just note that I suspect that default quantificational force plays a role here, based on my observation, referred to above, that which-phrase seem only to allow default universal force, while simple wh-words allow in addition default existential force. As of this writing, I am still pursuing this line of investigation. This concludes my discussion of Lahiri (1991); I turn now to Groenendijk and Stokhof (1992). This article is a response to my argument against analysing wh-clauses as inherently exhaustive, as in Groenendijk and Stokhof (1982; 1984). First, they accept my empirical claims regarding quantificational variability. They proceed to augment their original analysis with dynamic semantics (see Groenendijk and Stokhof (1990; 1991) and Chierchia (1992) for presentations), which they claim allows them to account for this variability. Second, they nevertheless maintain that strong exhaustiveness, in the sense discussed in chapter IV, is desirable even with quantificational variability. Regarding the first point, Irene Heim bas pointed out to me in conversation that Groenendijk and Stokhofs attempt to incorporate dynamic semantics into their semantics for questions contains a fonnal error, having to do with the representation of quantification that is not definable in tenns of the standard truthfunctional connectives of the first-order predicate calculus (see chapter I, pages 8-9). Thus, an account of quantificational variability in terms of their semantics for questions remains wanting. (Ginzburg (1993) refers to Groenendijk and Stokhofs attempt to treat quantificational variability, citing Groenendijk and Stokhof (1993), which I have not seen. This appears to be a later

xiv

On The Semantics ofWh-Clauses

version of their (1992) paper but I do not know if it corrects the error in the earlier version.) As for the case against strong exhaustiveness, I stand by my contention, detailed in chapter IV, that this is not an inherent property of wh-clauses. There I acknowledged that with predicates like know this sometimes seems to have counterintuitive consequences, which I attempted to account for by appealing to Gricean conversational principles. An alternative explanation may be possible by taking into account the results of Heim (1993). She shows that it is possible to generalize Karttunen's (1977) analysis of question-embedding predicates so. as to deliver strongly exhaustive meanings, without making the question-meaning itself strongly exhaustive. This opens the possibility of treating (strong) exhaustiveness purely as a lexical property of given predicates, a strategy in principle compatible with my analysis. And bearing in mind that Karttunen' s analysis is in essence a special case of Hamblin's semantics for questions, where the meaning is restricted to true answers-see the discussion in chapter II-Heim 's results should be replicable in Hamblin's terms, thus directly importable into my original analysis. I close this preface by referring briefly to Ginzburg's (1992; 1993) analysis. This is concerned primarily with providing a comprehensive account of the question/answer relation, which is orthogonal to the main points of my analysis. Nevertheless, Ginzburg's investigation leads him to make a binary division between question-embedding predicates (i.e., in my terms, wh-clause-embedding predicates), that essentially replicates my (pragmatically conditioned) factivity-based distinction. I refrain here from going into any of the formal differences between our approaches (Ginzburg's is implemented in the framework of situation semantics), but I will comment on Ginzburg's discussion of quantificational variability. His contention is that this is not a matter of quantifying over values of the wh-phrases at all but rather that the adverb modifies the degree to which the values of the wh-clause (i.e., the answer to the question the wh-clause denotes) resolve the question with respect to the embedding predicate. Resolution is defined in terms of an algebra of information, with resolving items of information being factual. Ginzburg motivates his proposal with examples such as the following. He says that a situation where Celia only knows that some unruly linguists showed up last night, though she doesn't know who they are,

Preface

XV

may be truthfully reported by the sentence Celia knows to some extent who showed up last night. He observes that in this situation Celia is not claimed to have any knowledge of specific individuals, which is what my analysis (with to some extent translated as a restricted exis~ tential quantifier) entails. I believe that what Ginzburg draws attention to is a use of many adverbial quantificational expressions that is logically independent of the individual~uantifying use I have analysed. Evidence for this is the fact that the same sentence may without contradiction or even a sense of infelicity contain adverbials with conflicting quantificational forces, as in For the most part, Celia knows only to some extent who showed up last night. I understand this as asserting that for most people who showed up Celia bas only limited knowledge of who they are. But on Ginzburg's account, this sentence should assert that Celia's knowledge simultaneously resolved the question of who showed up both to a majority degree and to a minority degree, which appears to be contra~ dictory. (Since Ginzburg also parametrizes resolvedness with respect to goals, he might contend that the two adverbials modify knowledge rela~ tive to distinct goals. But note that one of these "goals" is simply that of counting individuals and not, like the other goal, modifying the quality of knowledge attributed. These seem to me sufficiently different to warrent distinct treatments.) Another observation is that some adver~ bial expressions only seem to function as quantifiers over individuals, e.g., without exception, which is unexpected on Ginzburg's account. In short, I take it that quantificational variability is a distinct phenomenon from the degree of question resolvedness. This concludes my overview of developments of and reactions to since its writing. I hope to have given some indication that thesis my my principal original claims have so far stood up to scrutiny, also that my analysis is flexible enough to incorporate amendments designed to take a wider range of data into account. I also believe that the fact that my research bas generated further research in the same domain shows that this remains a fruitful area of investigation in natural language semantics. Stuttgart, November 1993

Acknowledgments I thank the members of my doctoral committee, under whose guidance I conducted the research documented here: Edmund L. Gettier, F. Roger Higgins, Barbara H. Partee, David Pesetsky, and especially Angelika Kratzer, my committee chair, without whose interest in and attention to my research the resulting dissertation would have been much poorer. I am also grateful to all the people from whom I received feedback on this work throughout its genesis: fellow students at UMass as well as participants at various conferences and colloquia at which I presented parts of this material. And I would like particularly to thank Ede Zimmermann for many helpful discussions during the final months of writing.

xvii

On the Semantics of Wh-Clauses

I Introduction 1 Overview This study is concerned with certain kinds of wh-clauses, whose interpretations are easily and, I will argue, plausibly rendered by a logicosemantic analysis on which wh-phrases translate as open sentences, that is, as expressions of the semantically interpreted representation which contain free variables. The evidence for my analysis is drawn almost entirely from English, though I briefly cite data suggesting the analysis carries over to Japanese as well, and the same holds for German, as I have gathered from discussions with a number of native speakers (though I do not report the evidence here); I am thus reasonably confident that the basic form of the analysis is of quite genereral relevance to the linguistic issues raised by it. The wh-constructions I treat at length are what are traditionally called indirect and direct questions. I The evidence I will be presenting indicates that the sentences that embed indirect questions constitute a semantically inhomogeneous class, consisting of those interpreted as quantified sentences, and those interpreted in the same way as direct questions. On my analysis, (l.a) may have the same interpretation as (l.b), and the embedded clause in (l.c) may have the same interpretation as (l.d); but (l.a) and (l.c) do not have the same intepretation. (1)

a. b. c. d.

The teacher always finds out which students arrive late. If a student arrives late, the teacher always finds out. The teacher always wonders which students arrive late. Which students arrive late?

3

4

On The Semantics ofWh-Clauses

Following the usual line of research in logical semantics, I will use the intuitive truth conditions (which, for explicitness, will be formulated in terms of modeltheoretic semantics) of sentences such as these to argue for a particular logical analysis, and will try to relate the logical representations thus obtained to the syntactic representations of the sentences by means of general grammatical processes. On the analysis I will present, the wh-phrase which students and the indefrnite noun phrase student have the same logical translation, namely, student'(x).2 This means that meaning differences among the sentences in (1) will have to be accounted for in other ways. Two of the aims of my research are to justify this uniform logical analysis and to offer an account of the observed differences. The idea that NPs such as those in (1) are logically open sentences is not new; in particular, this treatment of indefinites is a comerstone of the closely related Discourse Representation Theory of Kamp (1981) and File Change Semantics of Heim (1982). (Kamp originally applied this idea only to indefinites, Heim to definites as well.3) The extension of this idea to wh-phrases is due to Nishigauchi (1986; 1990); he does not, however, develop an explicit semantic analysis for whphrases based on this idea, nor explore the consequences of it for the semantics of wh-clauses. The main goal of my investigation is to do just this; this turns out to result in a semantics for wh-clauses that differs sharply from the infuentialline extending from Baker (1968) through Karttunen (1977) to Groenendijk and Stokhof (1982; 1984), though it bears conceptual similarities to the treatment of Hintikka (1976), for one class of these sentences, and to that of Hamblin (1973), for the other. Since the analysis of indefrnite (and defrnite) NPs as open sentences has been convincingly carried out by Heim and Kamp, the bulk of the original work in this study is concerned with applying this analysis to wh-phrases. I will thus confine myself, in the remainder of this introduction, to a summary of relevant aspects of the Kamp/Heim theory, sufficient to provide the theoretical basis for my analysis of wh-clauses. I will concentrate, in particular, on the form and modeltheoretic interpretation of the logical translations generated in this theory, putting aside numerous significant, but for present purposes extraneous, details, such as the status of discourse referents and the dynamic aspects of the semantics, as well as the compositional constructions of the logical representations. In order to set the stage for my

Introduction

5

analysis, I devote chapter II to a brief review of some of the more influential recent analyses of the semantics of questions, concentrating on issues related to the truthconditional interpretation of these constructions. Then in chapter III I give the first part of my analysis, concentrating on logicosemantic similarities between wh-pbrases and indefmite NPs, i.e., dealing with sentences such as those in (la-b). In chapter IV I take up some challenges to my analysis and attempt to refute them. In chapter V I extend the analysis to account for asymmetries between wh-pbrases and indefinites, and also deal with sentences such as those in (l.c-d). In appendices I discuss several issues raised by my analysis, that are somewhat peripheral to its central points, and in need of further

research.

2 The Lewis/Heim/Kamp Theory 2.1 Motivation One of the most compelling facts suggestive of analysing indefinites as open sentences, apparently first noted in a theoretical context by Lewis (1975), is their quantificational variability, that is, their ability to display a variety of quantiftcational forces, depending on the sentence they are in. The sentences in (2) illustrate this ((2.a,b) are Lewis's (12) and (9), respectively): (2)

a. b. c. d.

A quadratic equation usually bas two different solutions. Riders on the Thirteenth Avenue line seldom fmd seats. If a fanner owns a donkey, be takes good care of it. If a graduate student submits a paper to L/, it is often rejected.

These sentences may be paraphrased as follows: (3)

a. b.

c.

Most quadratic equations have two different solutions each. For few riders on the Thirteenth Avenue line is there a seat that they each find (when they ride on the Thirteenth Avenue line). For all pairs of a fanner and a donkey such that the farmer owns the donkey, the fanner takes good care of the donkey.

On The Semantics ofWh-Clauses

6 d.

For many pairs of a graduate student and a paper such that the graduate student submits the paper to LI, the paper is rejected.

Classically, indefinites are analyzed as existentially quantified (e.g. in Russell (1919); cf Heim (1982) for discussion); for instance, the logical representation of a sentence containing the NP a donkey would always include at least the formula in (4): (4)

3x[ ... donkey'(x) ... ]

But the quantificational variability exhibited by the indefinites in the sentences in (2) seems to be prima facie evidence against such an analysis. But then where do the quantificational forces of the indefmites in these sentences come from? For the first two and fourth sentences, it seems clear that it is the adverb: (2.a) has to do with most quadratic equations (···•Xn))· (Engdahl shows that the function-variable can take multiple arguments, hence the stipulation of n-arity.)2 There might seem to be no substantive issue in the choice between (148) and (149); in any case, it is not a crucial matter for present purposes. Still, it may be noted that (149) avoids the introduction into the translation of another quantified structure, which moreover is not a restricted quantification structure (and although it could be represented as one, this would seem to be unenlightening, since it would be something like ALL [INDIVIDUAL(x)] [wine'(f{x))J). In addition, (148) forces the representation of selectivity, since otherwise the wider-scope quantifier (ALL, here) would never be able to bind the function-variable; with (149), this contingency does not arise. Furthermore, it must be ensured that the variables which are arguments of the function-variable also occur in the translations of the quantified phrases, although the latter are introduced in the translation separately; whereas, with (149), this is guaranteed by the lexical rule. Finally, it is not unreasonable to hypothesize that, given the necessity of n-place function variables as potential arguments in the first place, general principles of type-shifting, as discussed e.g. in Partee (1987) and Hendriks (1990), will automatically yield the needed family of translations for (149), i.e. they needn't be separately listed in the grammar. For these reasons I would tend to prefer (149) as the representation of the functional reading, but again, the choice is not of empirical consequence. In both translations, to return to the issue at hand, I have as usual assumed a default universal quantifier for the wh-phrase; in addition I have represented the quantified NP by means of a tripartite restricted quantification structure with the main restrictive term. Most importantly, the functionfmust be specified: this is already partly done by havingfbe an argument of the predicate wine; in particular, we have that where f is a variable of type (e,e), for any individual variable x, fix) is a wine. But there must be further specification in order to avoid unwanted side effects of quantifying over functions. The restrictions

144

On The Semantics ofWh-Clauses

added by the rest of the clause do not suffice, because there are a great many ways of functionally relating a patron and a wine that he ordered that do not conform to the close relationship between patron and wine expressed by a phrase such as 'his favorite wine,' nor indeed are expressible by any simple (i.e. nonconjunctive) phrase, for instance, by randomly picking element from the domain and range. The issue here has been noticed by G&S and Engdahl, and they suggest that both contextual factors and some notion of natural, or conventional, functional relationships must be incorporated into the analysis in order to satisfactorily restrict the domain of functions to be quantified over (G&S say (1984, 199) that the functions must be "in some sense computable, or they must be made computable by the context.") This restriction can be represented directly in the logical translation by including a free predicate variable over these Skolem-functions, as G&S do. Alternatively, there might be general felicity conditions on the use and interpretation of function variables, as is implicit in Engdahl's analysis. It bears notice, incidentally, as Engdahl and G&S point out, that the need torestrict the domain is a general problem with quantifying over functions; it arises not only on the functional interpretation of wh-phrases, but also on analogous interpretations of indefinites and pronouns. It is worth pointing out, in this connection, that if the Skolemfunctions are not restricted in this way to certain conventional or contextually "computable" kinds, then their specification amounts to giving a list of pairs (or, generally, n-tuples). And indeed, this interpretation is available for (146), which might, for example, be continued by saying "namely, Mary ordered Zinfandel, Bill Petite Sirah, ...." G&S call this the pair-list reading, and distinguish it semantically from the functional reading, deriving it by quantifying the quantified NP into the wh-clause meaning. Such a move is not compatible with my analysis, since it means in effect giving a quantified NP scope over a same-clause whpbrase, which I have argued cannot be the case. It seems to me, bowever, that G&S's grounds for distinguishing the pair-list and functional interpretations are not conclusive, the strongest one, as I noted above, being the impossibility of a list-of-pairs answer when the quantified NP bas less than universal force; but there may well be independent pragmatic or semantic reasons for this, having to do with way items in a list are determined, as G&S themselves suggest. If this is so, there is no need to resort to quantifying in (or scoping the quantified NP over

Wh-Phrases as Variables Over Functions

145

the wh-phrase), and the structural relation I have proposed to hold at LF between the two phrase types can be maintained in full generality. That wh-phrases must be translatable as free function variable in my analysis is indicated by sentences such as the following: (150) John mostly knows which wines every patron ordered. This sentence would be true on a functional reading in a situation in which, for example, each patron ordered all of the following: his favorite wine, the wine his best friend recommended, and the wine he saw advertised in the paper this morning, but John only knows that every patron ordered his favorite wine and the wine his best friend recommended. This shows that a wh-phrase can be functionally dependent on a quantifier within its structural scope while at the same time being itself quantified from a higher adverb of quantification} The treatment of wh-phrases as variables over functions of individuals can be fully assimilated to my analysis of wh-phrases as containing a free individual variable by admitting zero-place function variables, i.e. of type e: these are just individual variables (this is what Engdahl does). But while the zero-place "functional" interpretationi.e. the normal variable interpretation-{)[ wh-phrases is quite general, and where it fails to obtain the reasons are also quite general, functional interpretations of higher arity, i.e. where the interpretation of the whphrase depends on that of some other expressions, appear to be rather restricted; in particular, they seem to be triggered only by the presence of a determiner-quantified NP; it does not arise under purely adverbial quantification, as the following sentences illustrate: (151) a. b.

John always knows which wines which patrons ordered. John always knows which patrons ordered which wines.

These sentences, in contrast to (147), cannot be continued by saying,

e.g. "namely, his favorite," which indicates that the functional reading

is not available. They can, however, like (147), be continued by giving a list of pairs. This can in fact be taken as support for the generality of the use of function variables, if the pair-list interpretation is rendered by means of them, as I suggested above. In that case it still needs to be explained why the "strict" functional reading is lacking in these sentences. I won't attempt to do this here, but a question perhaps worth

146

On The Semantics ofWh-Clauses

investigating is whether there is a connection between quantificationally closed expressions, such as determiner-quantified NPs, and natural or conventional functional relations, as appear to be required on the strict functional reading.4

Wh-Phrases as Variables Over Functions

147

Notes 1. Engdahl (1985, 253(n.10)) maintains that what she calls relational variable are different from variables over Skolem functions, but I fail to see the distinction. 2. Thanks to Ede Zimmermann for helpful discussion of this alternative. 3. I have been told independently by two people that they do not get this reading for (30), at least not saliently. The reading they got was that for most patrons, for every wine they ordered, John knows that they ordered that wine. This has exactly the opposite quantificational relationships as the reading my analysis predicts (and I myself get), and indeed, I see no obvious way of getting such a reading at all within the overall form of my analysis. I have not investigated sentences such as (30) extensively enough to offer any explanation for this discrepancy; but since, at least in the more straightforward examples that I have presented throughout this dissertation, the quantificational variability of wh-phrases in general seems quite robust, and has been corroborated by many people, I will continue to assume that the general form of my analysis is viable, and leave the question of the status of sentences such as (30) open. 4. In this connection it is worth noting that, in addition to determiner-quantified NPs, the strict functional reading is also induced by definite plurals: (i)

John knows which wines the guests ordered, namely, their favorite.

Definites might also be thought of as quantificationally closed, and indeed have been analysed as a kind of universal quantification (e.g. Barwise and Cooper (1981)). (I discuss definites in appendix B, though I do not address this issue.)

Appendix B Definiteness and Global Accommodation In my discussion in chapter III of presupposition accommodation I alluded to the distinction made by Heim (1983b) between local and global accommodation (cf. note 17 there); I returned to this in chapter V, section 3.3, where I suggested that the effect of local accommodation may be plausibly represented at LF, but that this seems less plausible for global accommodation. In this appendix I want to consider in more detail the effect of global accommodation on the logical translations (I will make no reference here to LF). In particular, I will try to exploit this process to take into account a widely assumed property of which-phrases that I have, for simplicity, purposely ignored throughout this study; namely, that they are definite (this is the position, for example, of Katz and Postal (1964)). Definites are often said to carry a presupposition of existence; this is the position going back to Frege (1892) and developed into a classical analysis by Strawson (1950; 1952) (contra Russell (1905)). Within the LHK theory this idea, and the consequent distinction between definites and indefinites, has been given the following formulation by Heim (1982, 233): "Indefinites, the descriptive content of the NP is presupposed, whereas in indefinites it is (merely) asserted." In terms of the context-change semantics that Heim develops, this distinction amounts to the claim that the descriptive content of a definite must be part of the background context prior to the utterance of the definite, in order for its use to be felicitous; conversely, the descriptive content of an indefinite must not be part of the background context prior to the utterance of the indefmite. By building this distinction into a principle regulating the felicity of discourses (the Novelty/Familiarity Condition), Heim is able to treat definites logically the same indefi-

149

On The Semantics ofWh-Clauses

150

nites, that is, as open sentences. I want to suggest that the behavior of definite wh-phrases (e.g. which-phrases), as analysed in this study, can be viewed as essentially consistent with this distinction. First of all, an additional distinction must be made, between singular and plural definites. In all my examples illustrating the quantifiability of wh-phrases, where they contain which-phrases these are plural. In general, singular which-phrase do not appear able display quantificational variability. Compare, for example, the sentence in (105.a) (chapter V), repeated here as (152.a), with the sentence in (152.b): (152) a. b.

The maitre d' seldom knows which patrons of Maxim's are rich. The maitre d' seldom knows which patron of Maxim's is rich.

In contrast to (152.a), the sentence in (152.b) means that there are few times such that the maitre d' knows of the one patron of Maxim's who is rich that s/he is rich. (The oddness of this sentence comes from the adverb's being forced to quantify over states of the maitre d's knowing.) How can we account for this in terms of my analysis? A standard position (along Strawsonian lines) is that a singular definite carries the presupposition that there is exactly one entity (in the domain of discourse) satisfying its descriptive content. Let us also assume that quantifiers carry certain presuppositions (this also accords with the Strawsonian view; cf Strawson (1952, 176); also Hausser (1976)); in particular, an adverb of quantification such as seldom presupposes 'more than once' (thus, in its nonfrequency usage, 'more than one,' like few). Given this, if the wh-clause in (152) is accommodated into the restrictive term, where it would quantified by seldom, there would be a clash of presuppositions. It is worth noting that the same singular/plural distinction with respect to quantifiability obtains with non-wh definites; compare the following two sentences: (153) a. b.

If the students cheat on the final exam, the principal usually

finds out. If the student cheats on the final exam, the principal usually finds out.

Definiteness and Global Accomnwdation

151

While (153.a) has a quantificational reading just like (29.a) (chapter III), (153.b) has no such reading, but the singular definite picks a one particular student, and the adverb bas a frequency reading. The difference between (29.a), with an indefinite, and (153.a) is that the definite in the latter means that its reference is restricted to some contextually determined set of students I (which, moreover, has more than one member -this can be regarded as a presupposition of plurality). The same thing can be said, too, of the plural which-phrase in (152.a). I would like to suggest the descriptive content of definites is accommodated to the context of utterance, i.e., it is globally accommodated, not locally accommodated. Recall that, in the context of a quantified sentence, local accommodation of material in the nuclear scope is to the the restrictive term. In contrast, globally accommodated material remains independent of the restrictive term. What this amounts to in terms of the logical translation is that global accommodation is beyond the scope of the quantifier, so that any free variable within the accommodated material will not get bound by the quantifier. To illustrate this, let us consider what happens in sentences with an embedded that-clause. Recall from chapter III my remark (note 17) that presupposition accommodation of the that-complement of regret is an instance of global accommodation (as are the cases of telescoping (chapter Ill, note 15)). This is supported by sentences such as the following: (154) a. b.

The principal usually finds out that students cheat on the final exam. The maitre d' seldom knows that patrons of Maxim's are rich.

The indefinites in these sentences, unlike those in the corresponding ifclauses, and unlike the corresponding wh-phrases, do not have the quantificational force of the matrix adverbs, but have generic force (coming from an implicit generic quantifier within the embedded clauses. This is straightforwardly accounted for in terms of my analysis if the result of accommodating these clauses, which are presuppositions of the respective nuclear scopes, is not to the restrictive term, as with if- and whclauses, but beyond it-i.e., global accommodation. The resulting logical translations of the sentences in (154) are the following (where G stands for the generic quantifier)2:

152

On The Semantics ofWh-Clauses

(155) a. b.

[G[student'(x)][cheat-on-the-final-exam'(x)]] & [MOST [IlME(t)][find-ou t'(tp,[G[student '(x)][ cheat -on-thefinal-exam'(x) ]],t)]] [G[patron-of-Maxim's'(x)][rich'(x)]] & [FEW[TIME(t)] [know'(tmd,[G [patron-of-Maxim's'(x)][rich'(x)]],t)]]

Although the indefinites here are in any case bound by the generic quantifier, there is independent evidence that that-clauses are globally, not locally, accommodated. Angelika Kratzer observed (p.c.) that the sentence in (156.a) does not have an interpretation corresponding to the sentence in (156.b), where the that-clause appears in an if-clause, which is analogous to being in the restrictive term of a quantifier; rather the interpretation of (156.a) is the same as that of (156.c), in which the that-clause is independent of the if-clause: (156) a. b. c.

If Galileo claims that the earth is round, he knows that the earth is round. If Galileo claims that the earth is round and the earth is round, he knows that the earth is round. The earth is round and if Galileo claims that the earth is round, he knows that the earth is round.

Let us return now to the case of plural definites. I suggested above that their descriptive content is characterized by a contextually specified set. And I have proposed that this descriptive content is globally accommodated. But in addition, there is still the effect of local accommodation: as usual, the wh-clause should be accommodated into the restrictive term. Since the quantity presupposition of plurals is compatible with that of quantifiers such as usually, mostly, seldom, etc., this will not be excluded, as it is with singular definites. I will borrow here from the treatment of plurals in Kamp and Reyle (1990) (though they do not bring the effect of presupposition accommodation into their formulation), and add a condition specifying that the variable in the translation of a plural definite ( wh- or non-wh-phrase) belongs to the context set specifying the descriptive content of the phrase. This results in the following translations, for (152.a) and (153.a) respectively (where X is a variable over sets of individuals):

Definiteness and Global Accommodation (157) a.

b.

153

[patron-of-Maxim's'(X)] & [FEW[x E X & patron-of-Maxim's'(x) & rich'(x)][know'(tmd,[x E X & patron-of-Maxim's'(x) & rich'(x)])] [student'(X)] & [MOST[x E X & student'(x) & cheat-onthe-final-exam'(x)][find-out'(tp,[x E X & student'(x) & cheat-on-the-final-exam'(x)])]

154

On The Semantics ofWh-Clauses

Notes 1. Cf. Westersta.hl's (1985) use of context sets to distinguish definite detenniners from others; in fact, he treats words such as the not as logical detenniners but as indicators of context sets, which is very much akin to the approach I am taking here. 2. I refrain, for simplicity, from adding the translations of the definites the exam, the principal, and the maitre d'; these would all, of course, be globally accommodated, and have their reference restricted to a single individual.

Appendix C Complement If- and Whether-Clauses In chapter III, section 3.1 I remarked that complement if-clauses, unlike adjunct if-clauses, are apparently unable to serve as restrictive terms of adverbs of quantification. In this respect they are like whetherclauses. In this appendix I will try to bring these constructions into the general form of my analysis. Let us consider variants of the sentences in (28) from chapter III; to restrict attention now to the complement reading, I have omitted the correlative pronouns in (28) and allowed an alternation between if and

whether:

(158) a. b. c. d. e.

The principal usually finds out if/whether students cheat on the final exam. Sue usually remembers if/whether a birthday present arrives special delivery. With few exceptions, Mary knows if/whether students submitted abstracts to conferences. Bill seldom acknowledges if/whether he gets a good idea from a colleague. John always discovers if/whether a book is stolen from the library.

It is evident from considering the intuitive meanings of these sentences that the if/whether-clauses are not functioning as restrictive terms. For example, (158.a) does not mean that for most students who cheat on the final exam the principal finds out of them that (or whether) they cheat on the final exam. Instead, the if/whether-clause serves to offer a

155

156

On The Semantics ofWh-Clauses

set of complementary alternatives concerning the possibility of the students cheating on the final exam. By offering a set of complementary alternatives I mean a choice between a given situation and its negation; for example, in (158.a), the alternatives are, for each student, either that s/he cheated on the final exam or that s/he didn't. This might be expressed with the following logical translation: (159) student'(x) & [cheat-on-the-final-exam'(x)] v -.(cheat-on-thefinal-exam'(x))] Now consider what this means with respect to presupposition. The second conjunct in (159) is of the form P v -.P, and this is a tautology. This means that the truth of (159) (under an assignment) depends only on the first conjunct, student'(x). What this suggests is that, when a whether- or complement if-clause is accommodated as a presupposition of the nuclear scope, it is only the indefinite subject that is semantically significant; in other words, in effect, it is only the indefinite that is presupposed. (There is, however, a problem here; see below.) This would account for the inability of the if/whether-clause as a whole to restrict the adverb of quantification. This result accords with the intuitive meanings of the sentences in (158); thus, (158.a), for example, may be paraphrased as: "For most students, if the student cheats on the final exam, the principal finds out that the student cheats on the final exam, and if the student does not cheat on the final exam, the principal finds out that the student does not cheat on the final exam." In order to represent this meaning within the framework of my analysis, I will take advantage of the following logical equivalence: [(P -7 Q) & (-.P -7 R)] ~ [(P & Q) v (-.P & R)]. The paraphrase just given is of the form on the left-hand side of this equivalence; in terms of the right-hand side, it would be: "For most students, either the student cheats on the final exam and the principal finds out that the student cheats on the final exam, or the student does not cheat on the final exam and the principal fmds out that the student does not cheat on the final exam." This more intuitively expresses the offering of alternatives; in addition, with this version we can avoid introducing the material conditional connective into the translation language, which con-

Complement If- and Whether-Clauses

157

forms to our use of restricted quantification without this connective, as noted in chapter I. Here, then, is the translation of (158.a): (160) MOST[student'(x)] [([cheat-on-the-fmal-exam'(x)] & [find-out'( tp,[cheat-on-the-fmal-exam'(x)])]) v ([---,(cheat-on-the-final-exam'(x))] & [find-out'(tp,[-,(cheat-on-the-final-exam'(x))])])] Here I must acknowledge a difficulty for my approach: I see no way of deriving (160) via presupposition accommodation from the nuclear scope. I should note immediately that the difficulty does not lie in the absence of student'(x) from that term; the following translation is logically equivalent to (160): (161) MOST[student'(x)] [([student'(x) & cheat-on-the-final-exarn'(x)] & [find-out'(tp,[student'(x) & cheat-on-the-fmal-exam'(x)])]) v ([student'(x) & -.(cheat-on-the-fmal-exarn'(x))] & [find-out'(tp,[student'(x) & -.(cheat-on-the-fmal-exam'(x))])])] The nuclear scope in this translation has the form [(P & Q) & R] v [(P & -,Q) & S], and this is not equivalent to P. The correct restrictive term would be derived by presupposition accommodation if (159) were taken as the nuclear scope, as alluded to above, but this would yield a wrong translation, paraphrased by: "For most students, the principal finds out that they are students." If there is indeed no way to get (160) (or (161)) via presupposition accommodation from the nuclear scope, I'm not certain what this entails for my general analysis; perhaps the most optimistic conclusion is that the presuppositional properties of whether- and complement ifclauses are different from those of wh- and adjunct if-clauses in ways that remain to be elucidated. Consider the case of a whether- or complement if-clause embedded under a nonpresuppositional predicate, as in the following example:

158

On The Semantics ofWh-Clauses

(162) The principal usually wonders if/whether students cheat on the final exam. Here the indefinite is not quantified by the adverb, which can only have a frequency reading. This is as expected, given the fact that the complement is not presupposed. (162) might be paraphrased as: "Most of the time, the principal stand in the wondering relation to the question: Do students cheat on the final exam?" The alternatives in this case are either that students cheat on the final exam or students do not cheat on the final exam. This in fact follows on Hamblin's semantics for questions, treating the if/whether-clause as a yes/no-question. Note that both alternatives are independent quantificational structures, where the quantifier has generic force, which the indefinite gets. Thus, the set of propositions which is the denotation of the embedded clause is, (in rather clumsy paraphrase): {generically many students cheat on the final exam, generically many students do not cheat on the final exam.}

Bibliography Note: Entries preceded by an asterisk are either published versions of the preceding entry, which is the unpublished form cited in the text, or newer works referred to in the preface. List of abbreviations used in the references: Proceedings of the annual meeting of the Chicago LinguisCLS: tics Society. Published by Chicago Linguistics Society, University of Chicago. Graduate Linguistic Student Association, University of MasGLSA: sachusetts, Amherst. Linguistics and Philosophy. Kluwer, Dordrecht. L&P: Linguistic Inquiry. MIT Press, Cambridge MA. LI: MIIWPL: MIT Working Papers in Linguistics. Department of Linguistics, Massachusetts Institute of Technology, Cambridge MA. Proceedings of the annual meeting of the North Eastern NELS: Linguistics Society. Published by GLSA. Natural Language and Linguistic Theory. Kluwer, DorNLLT: drecht

UMOP:

University of Massachusetts Occasional Papers in Linguistics. Published by GLSA.

WCCFL: Proceedings of the annual West Coast Conference on For-

mal Linguistics. Published by Stanford Linguistics Association, Stanford University.

Altham, J.E.J. and Tennant, Neil W. (1975) "Sortal Quantification." In Keenan, ed., 46-58. Aoun, Joseph, Hornstein, Norbert and Sportiche, Dominique (1981) "Some Aspects of Wide Scope Quantification." Journal of Linguistic Research l, 69-95. Aqvist, Lennart (1965) A New Approach to the Logical Theory of Interrogatives. Filosofiska Foreningen, Uppsala. Aqvist, Lennart (1975) A New Approach to the Logical Theory of Interrogatives. Gunter Narr, Tiibingen. (Republication of Aqvist (1965), with a new preface and additional references.)

159

160

On The Semantics ofWh-Clauses

Asher, Nicholas (1987) "A Typology for Attitude Verbs and Their Anaphoric Properties." L&P 10, 125-197. Bacon, John (1965) "A Simple Treatment of Complex Terms." Journal of Philosophy 62, 328-331. Baker, CarlL. (1968) Indirect Questions in English. Doctoral Dissertation, University of Illinois, Urbana. Baker, Carl L. (1970) "Notes on the Description of English Questions: The Role of an Abstract Question Morpheme." Foundations of Language 6, 197-217. Barwise, Jon and Cooper, Robin (1981) "Generalized Quantifiers and Natural Language." L&P 4, 159-219. Bauerle, Rainer (1979) "Questions and Answers." In Bauerle, Egli, and Stechow, 61-74. Bauerle, Rainer, Egli, Urs and von Stechow, Arnim, eds. (1979) Semantics from Different Points of View. Springer, Berlin. Bauerle, Rainer and Zimmermann, Thomas E. (1987) "Fragesatze." FSN-Bericht-87-20, Forschungsstelle fiir natiirlich-sprachliche Systeme, University of Tiibingen. *Bauerle, Rainer and Zimmermann, Thomas E. (1991) "Fragesatze." In A. von Stechow and D. Wunderlich, eds. Semantics. An International Handbook of Contemporary Research, 333-348. De Gruyter, Berlin. Belnap, Nuel D., Jr. (1970) "Conditional Assertion and Restricted Quantification." Nous 4, 1-12. Belnap, Nuel D., Jr. (1982) "Questions and Answers in Montague Grammar." In Peters, Stanley and Esa Saarinen, eds., Processes, Beliefs, and Questions, 165-198. Reidel, Dordrecht. Bennett, Michael (1977) "A Response to Karttunen on Questions." L&P 1, 279-300. Bennett, Michael (1979) Questions in Montague Grammar. Published by IULC, Bloomington IN. Berman, Stephen R. (1987) "Situation-based Semantics for Adverbs of Quantification." In UMOP 12, 45-68. Also appears, slightly abbreviated, in WCCFL 6, 17-31 (1987). Berman, Stephen (1989) "An Analysis of Quantificational Variability in Indirect Questions." In the Proceedings of the Student Conference in Linguistics, MJTWPL 11, 1-15.

Bibliography

161

Berman, Stephen (1990) "Towards the Semantics of Open Sentences: Wh-phrases and Indefinites." In Stokhof, Martin and Torenvliet, Leen, eds., Proceedings of the Seventh Amsterdam Colloquium, 53-77. Institute for Language, Logic, and Information, University of Amsterdam. Boer, Steven E. (1978) "'Who' and 'Whether': Towards a Theory of Indirect Questions Clauses." L&P 2, 307-345. Bolinger, Dwight (1978) "Asking More than One Thing at a Time." In Hiz, ed., 107-150. Bonet, Eulalia (1990) "Subjects in Catalan." Unpublished manuscript, MIT, Cambridge MA. Bresnan, Joan W. (1972) Theory of Complementation in English Syntax. Doctoral dissertation, MIT, Cambridge MA. Bresnan, Joan and Grimshaw, Jane (1978) "The Syntax of Free Relatives in English." LI 9, 331-391. Bresnan, Joan W., ed. (1982) The Mental Representation of grammatical Relations. MIT Press, Cambridge MA. Carlson, Gregory N. (1977) Reference to Kinds in English. Doctoral dissertation, University of Massachusetts. Published by GLSA. Chierchia, Gennaro (1988) "Dynamic Generalized Quantifiers and Donkey Anaphora." In Krifka, Manfred, ed., Genericity in Natural Language. Seminar fiir natiirlich-sprachliche Systeme, Universitat Tiibingen. Chierchia, Gennaro (1990) "Anaphora and Dynamic Logic." Unpublished manuscript, University of Amsterdam and Cornell University. *Chierchia, Gennaro (1992) "Anaphora and Dynamic Binding". L&P 15, 111-183. Chierchia, Gennaro, Partee, Barbara H. and Turner, Raymond, eds. (1989) Properties, Types and Meaning. Volume II: Semantic Issues. Kluwer, Dordrecht. Chisolm, Roderick M. (1963) "The Logic of Knowing." Journal of Philosophy 60, 773-795. Reprinted in Roth and Galis, 189-219. Choe, Jae-Woong (1987) "LF Movement and Pied-piping." L/18, 348353. Chomsky, Noam (1970) "Remarks on Nominalization." In Jacobs, Roderick A. and Rosenbaum, Peter S., eds., Readings in English Transformational Grammar, 184-221. Ginn, Waltham MA.

162

On The Semantics ofWh-Clauses

Chomsky, Noam (1973) "Conditions on Transformations." In Anderson, Stephen R. and Kiparsky, Paul, eds., A Festschrift for Morris Halle. Holt, Rinehart and Winston, New York. Chomsky, Noam (1986) Barriers. MIT Press, Cambridge MA. Comorovski, Ileana (1989) Discourse ana the Syntax of Multiple Constituent Questions. Doctoral dissertation, Cornell University, Ithaca NY. Cresswell, Max J. (1973) Logics ana Languages. Methuen, London. Cresswell (1985) Structured Meanings. MIT Press, Cambridge MA. Davidson, Donald and Harman, Gilbert, eds., (1972) Semantics of Natural Language. Reidel, Dordrecht. Declerck, Renaat (1988) "Restrictive when-Clauses." L&P 11, 131168. Diesing, Molly (1990a) "Verb-second in Yiddish and the Nature of the Subject Position." NLLT 8, 41-79. Diesing, Molly (1990b) The Syntactic Roots of Semantic Partition. Doctoral dissertation, University of Massachusetts, Amherst. Published by GLSA. *Diesing, Molly (1992) Indefinites. MIT Press, Cambridge MA. Egli, Urs (1973) "Semantische Reprasentation der Frage." Dialectica 27, 363-370. Egli, Urs (1976) "Zur Semantik des Dialogs." Papiere des Sonderforschungsbereich 99, Universitat Konstanz. Elliott, Dale (1971) The Grammar of Emotive ana Exclamatory Sentences in English. Doctoral dissertation, The Ohio State University. In Working Papers in Linguistics 8, viii-110. Computer and Information Science Research Center, The Ohio State University, Columbus. Engdahl, Elisabet (1980) The Syntax ana Semantics of Questions in Swedish. Doctoral disseration, University of Massachusetts, Amherst. Published by GLSA. Engdahl, Elisabet (1985) Constituent Questions. Reidel, Dordrecht. Farkas, Donca and Sugioka, Yoko (1983) "Restrictive If/When Clauses." L&P 9, 225-258. Fiengo, Robert, Huang, C.-T. James, Lasnik, Howard, and Reinhart, Tanya (1988) "The Syntax of Wh-in-Situ." WCCFL 7, 81-98. Fodor, Janet Dean (1978) "Parsing Strategies and Constraints on Transformations." L/9, 427-473.

Bibliography

163

Frege, Gottlob (1892) "Uber Sinn und Bedeutung." Zeitschrift fUr Philosophie und philosophische Kritik 100, 25-50. Gabbay, Dov M. and Guenthner, Franz, eds., (1989) Handbook of

Philosophical Logic. Volume IV: Topics in the Philosophy of Language. Reidel, Dordrecht. Gatlin, Daniel (1975) Intensional and Higher-Order Modal Logic. North-Holland, Amsterdam. Gazdar, Gerald (1979) Pragmatics: Implicature, Presupposition, and Logical Form. Academic Press, New York. Gazdar, Gerald, Klein, Ewan, Pullum, Geoffrey, and Sag, Ivan (1985) Generalized Phrase Structure Grammar. Harvard University Press, Cambridge MA. Gerstner, Claudia and Krifka, Manfred (1987) "Genericity." Unpublished manuscript, Universitat Ttibingen. Gettler, Edmund L. (1963) "Is Justified True Belief Knowledge?" Analysis 23, 121-123. Reprinted in Roth and Galis, 35-38. *Ginzburg, Jonathan (1992) Questions, Queries, and Facts: A Semantics and Pragmatics for Interrogatives. Doctoral dissertation, Stanford University. *Ginzburg, Jonathan (1993) "Resolving Questions." Unpublished manuscript, Human Communication Research Centre, University of Edinburgh. Grewendorf, Gunther (1983) "What Answers Can Be Given?" In Kiefer, ed., 45-64. Grice, H. Paul (1975) "Logic and Conversation." In Cole, Peter and Morgan, Jerry L., eds., Syntax and Semantics, vol.3, 41-58. Academic Press, New York. Grimshaw, Jane B. (1977) English Wh-Constructions and the Theory of Grammar. Doctoral dissertation, University of Massachusetts. Published by GLSA. Grimshaw, Jane B. (1979) "Complement Selection and the Lexicon." LI 10, 279-326. Groenendijk, Jeroen and Stokhof, Martin (1982) "Semantic Analysis of Wh-Complements." L&P 5 175-233. Reprinted as chapter II of Groenendijk and Stokhof (1984), 77-164. Groenendijk, Jeroen and Stokhof, Martin (1984) Studies on the Semantics of Questions and the Pragmatics of Answers. Doctoral disseration, Universiteit van Amsterdam.

164

On The Semantics ofWh-Clauses

Groenendijk, Jeroen and Stokhof, Martin (1989) "Type-Shifting Rules and the Semantics of Interrogatives." In Chierchia, Partee, and Turner, 21-68. *Groenendijk, Jeroen and Stokhof, Martin (1990) "Dynamic Montague Grammar." In L. Kalman and L. P6los, eds., Papers from the Second Symposium on Logic and Language. 3-48. Akademiai Kiad6, Budapest. *Groenendijk, Jeroen and Stokhof, Martin (1991) "Dynamic Predicate Logic." L&P 14 39-100. *Groenendijk, Jeroen and Stokhof, Martin (1992) "A Note on Interrogatives and Adverbs of Quantification." ILLC Prepublication Series for Logic, Semantics and Philosophy of Language LP-9207, University of Amsterdam. Also appears in C. Barker and D. Dowty, eds., Proceedings of SALT II (Working Papers in Linguistics No. 40), 99-124, Ohio State University, Columbus. *Groenendijk, Jeroen and Stokhof, Martin (1993) "Interrogatives and Adverbs of Quantification." In K. Bimbo and A. Mates, eds.,

Proceedings of the 4th Symposium on Logic and Language,

Budapest. Groos, Anneke and van Riemsdijk, Henk (1981) "Matching Effects in Free Relatives: A Parameter of Core Grammar." In A. Belletti, L. Brandi, and L. Rizzi, eds., Theory of Markedness in Generative Grammar, 171-216. ScuolaNormale Superiore, Pisa. Hailperin, Theodore (1957) "A Theory of Restricted Quantification." Journal of Symbolic Logic 22, 19-35, 113-129. Hamblin, C.L. (1958) "Questions." The Australian Journal of Philosophy 36, 159-168. Hamblin, C.L. (1973) "Questions in Montague English." Foundations of Language 10,41-53. Reprinted in Partee (1976), 247-259. Harrah, David (1984) "The Logic of Questions." In Gabbay, Dov M. and Guenthner, Franz, eds., Handbook of Philosophical Logic. Volume II: Extensions of Classical Logic, 715-764. Reidel, Dordrecht Hausser, Roland R. (1976) "Presuppositions in Montague Grammar." Theoretical Linguistics 3, 245-280. Hausser, Roland R. (1983) "The Syntax and Semantics of English Mood." In Kiefer, ed., 97-158. Hausser, Roland R. (1984) Surface Compositional Grammar. Wilhelm Fink, Munich.

Bibliography

165

Hausser, Roland and Zaefferer, Dietmar (1979) "Questions and Answers in a Context-dependent Montague Granunar." In F. Guenthner and S.J. Schmidt, eds., Formal Semantics and Pragmatics for Natural Languages. Reidel, Dordrecht. Heim, Irene (1982) The Semantics ofDefinite and Indefinite Noun Phrases. Doctoral dissertation, University of Massachusetts. Published by GLSA. Heim, Irene (1983a) "File Change Semantics and the Familiarity Theory of Definiteness." In Bauerle, Rainer, Schwarze, Christoph and von Stechow, Amim, eds., Meaning, Use, and Intepretation of Language, 164-189. Walter de Gruyter, Berlin. Heim, Irene (1983b) "On the Projection Problem for Presuppositions." WCCFL 2, 114-125. *Heim, Irene (1993) "Interrogative Semantics and Karttunen's Semantics for Know." Unpublished manuscript, MIT, Cambridge MA. Hendriks, Herman (1990) "Flexible Montague Granunar." Unpublished manuscript, University of Amsterdam. Higginbotham, James and May, Robert (1981) "Questions, Quantifiers and Crossing." The Linguistic Review 1, 41-80. Hintikka, Jaakko (1962) Knowledge and Belief Cornell University Press, Ithaca NY. Hintikka, Jaakko (1974) "Questions about Questions." In Munitz and Unger, 103-158. Hintikka, Jaakko (1976) The Semantics of Questions and the Questions of Semantics. Acta Philosophica Fennica, Vol.28, No.4. North -Holland. Amsterdam. Hint.ikka, Jaakko (1983) "New Foundations for a Theory of Questions and Answers." In Kiefer, ed., 159-190. Hirschbiihler, Paul (1978) The Semantics and Syntax ofWh- Constructions. Doctoral dissertation, University of Massachusetts. Published by GLSA. Hiz, Henry, ed. (1978) Questions. Reidel, Dordrecht. Hiz, Henry, (1978a) "Difficult Questions." In Hiz, ed., 211-226. Hoepelman (1983) "On Questions." In Kiefer, ed., 191-227. Holker, Klaus (1981) Zur semantischen und pragmatischen Analyse von Interrogativen. Helmut Buske, Hamburg. Huang, C.-T. James (1982a) "Move WH in a Language without WH movement." The Linguistic Review 1, 369-416.

166

On The Semantics ofWh-Clauses

Huang, C.-T. James (1982b) Logical Relations in Chinese and the Theory of Grammar. Doctoral dissertation, MIT, Cambridge MA. Hull, R.D. (1975) "A Semantics for Superficial and Embedded Questions in Natural Language." In Keenan, ed., 35-45. Jacobson, Pauline (1988) "The Syntax and Semantics of Free Relatives (in English)." Unpublished manuscript, Brown University, Providence Rl. Jespersen, Otto (1961[1928]) A Modern English Grammar on Historical Principles, Vol. III. Allen & Unwin, London. Kadmon, Nirit (1987) On Unique and Non-Unique Reference and Asymmetric Quantification. Doctoral dissertation, University of Massachusetts. Published by GLSA. Kamp, Hans (1981) "A Theory of Truth and Semantic Representation." In Groenendijk, Jeroen, Janssen, Theo, and Stokhof, Martin, eds., Formal Methods in the Study of Language, 277-321. Mathematisch Centrum, Universiteit van Amsterdam. Reprinted in Groenendijk, Janssen, and Stokhof, eds. (1984) Truth, Interpretation, and Information, 1-41. Foris, Dordrecht. Karnp, Hans and Reyle, Uwe (1990) From Discourse to Logic. Unpublished book manuscript, Institut fiir maschinelle Sprachverarbeitung, University of Stuttgart. *Kamp, Hans and Reyle, Uwe (1993) From Discourse to Logic. Kluwer, Dordrecht. Karttunen, Lauri (1971) "Implicative Verbs." Language 47, 340-358. Karttunen, Lauri (1973) "Presuppositions of Compound Sentences." LI 4, 169-193. Karttunen, Lauri (1974) "Presupposition and Linguistic Context." Theoretical Linguistics 1, 181-194. Karttunen, Lauri (1977) "Syntax and Semantics of Questions." L&P 1, 3-44. Reprinted in Hiz, 165-210. Karttunen, Lauri and Peters, Stanley (1980) "Interrogative Quantifiers." In Rohrer, Christian, ed. Time, Tense, and Quantifiers, 181-205. Max Niemeyer, Ttibingen. Kasher, Asa (1973) "Logical Forms in Context: Presuppositions and other Preconditions." The Monist 57, 371-395. Katz, Jerrold J. and Postal, Paul M. (1964) An Integrated Theory of Linguistic Descriptions. MIT Press, Cambridge MA.

Bibliography

167

Kawasaki, Noriko (1989) "On the Semantics of Wh-Questions in Classical Japanese." Unpublished manuscript, University of Massachusetts, Amherst. Kawasaki, Noriko (1990) "Minimality Condition on Variable Binding." Unpublished manuscript, University of Massachusetts, Amherst. Keenan, Edward L., ed. (1975) Formal Semantics of Natural Language. Cambridge University Press, Cambridge. Keenan, Edward L. and Hull, Robert D. (1973) "The Logical Presuppositions of Questions and Answers." In PetOfi, Janos S. and Franck, Dorothea, eds., Prttsuppositionen in Philosophie und Linguistik, 441-466. Athenaum, Frankfurt. Kiefer, Ferenc, ed. (1983) Questions and Answers. Reidel, Dordrecht. Kim, Soo Won (1989) "Wh-phrases in Korean and Japanese are QPs." In the Proceedings of the Student Conference in Linguistics, MIIWPL 11, 119-138. Kiparsky, Paul and Kiparsky, Carol (1971) "Fact." In Steinberg, Danny D. and Jakobovits, Leon A., eds., Semantics, 345-369. Cambridge University Press, Cambridge. Reprinted from M. Bierwisch and K. Heidolph, eds. (1970) Progess in Linguistics, 143173. Mouton, The Hague. Kitagawa, Yoshihisa (1986) Subjects in Japanese and English. Doctoral dissertation, University of Massachusetts. Published by GLSA. Kratzer, Angelika (1978) Semantik der Rede. Scriptor, Konigstein. Kratzer, Angelika (1981) "The Notional Category of Modality." In H. Eikrneyer and H. Rieser, eds., Words, Worlds, and ContextsNew Approaches in Word Semantics. Walter de Gruyter, Berlin. Kratzer, Angelika (1986) "Conditionals." In CLS 22/2: Papers from the Parasession on Pragmatics and Grammatical Theory, 1-15. Kratzer, Angelika (1988) "Comments on P. Jacobson's 'The Syntax and Semantics of Free Relatives in English'." Delivered at the LSA Annual Winter Meeting, New Orleans. Kratzer, Angelika (1989) "Stage-Level and Individual-Level Predicates." Unpublished manuscript, University of Massachusetts, Amherst. Kuno, Susumo, and Masunaga, K. (1987) "Questions with WH Phrases in Islands." In UMOP 11, 139-166.

168

On The Semantics ofWh-Clauses

Kuroda, S.-Y. (1988) "Whether We Agree or Not: A Comparative Syntax of English and Japanese." Lingvisticae Investigationes 12, 1-47. *Lahiri, Utpal (1991) Embedded Interrogatives and Predicates That Embed Them. Doctoral dissertation, MIT. Lasnik, Howard and Saito, Mamoru (1984) "On the Nature of Proper Government." L/15, 235-289. Lees, Robert B. (1963(1960]) The Grammar of English Nominalizations. IJAL, Vol.26 No.3 Pt.II. Publication 12 of the Indiana University Research Center in Anthropology, Folklore, and Linguistics, Bloomington IN. Levi, Isaac (1967) Gambling with Truth. Knopf, New York. Levinson, Stephen C. (1983) Pragmatics. Cambridge Univerity Press, Cambridge. Lewis, David (1972) "General Semantics." In Davidson and Harman, 169-218. Reprinted from Synthese 22 (1970), 18-67. Also reprinted in Partee (1976), 1-50. Lewis, David (1975) "Adverbs of Quantification." In Keenan, ed., 3-15. Lewis, David (1979) "Scorekeeping in a Language Game." In Bauerle, Egli, and von Stechow, 172-187. Also in Journal of Philosophical Logic 8, 339-359. Lewis, David (1982) '"Whether' Report." In T. Pauli et al., eds.,

320311: Philosophical Essays Dedicated to Lennart Aqvist on his Fiftieth Birthday, 194-206. Uppsala.

Lindstrom, P. (1966) "First-order Predicate Logic with Generalized Quantifiers." Theoria 32, 186-195. May, Robert (1977) The Grammar of Quantification. Doctoral dissertation, MIT, Cambridge MA. May, Robert (1985) Logical Form. Its Structure and Derivation. MIT Press, Cambridge MA. McCawley, James D. (1979) "Verbs of Bitching." In McCawley, Adverbs, Vowels, and Other Objects ofWonder, 135-150. University of Chicago Press, Chicago. Reprinted from D. Hockney et al., eds., Contemporary Research in Philosophical Logic and Linguistic Semantics, 313-332. Reidel, Dordrecht. Milsark, Gary (1974) Existential Sentences in English. Doctoral dissertation, MIT, Cambridge MA.

Bibliography

169

Montague, Richard (1970) "English as a Fonnal Language." In Visentini, Bruno et al., eds., Linguaggi nella Societli e nella Tecnica, 189-224. Edizioni di Comunit.a, Milan. Reprinted in Thomason, ed., 188-221. Montague, Richard (1973) "The Proper Treatment of Quantification in Ordinary English." In J. Hintikka, J. Moravcsik, and P. Suppes, eds. Approaches to Natural Language, 221-242. Reidel, Dordrecht Reprinted in Thomason, ed., 247-270. Mostowski, Andrzei (1957) "On a Generalization of Quantifiers." Fundamenta Mathematicae 44, 12-36. Munitz, Milton K. and Unger, Peter K., eds. (1974) Semantics and Philosophy. New York University Press, New York. Munsat, Stanley (1986) "Wh-Complementizers" L&P 9, 191- 217. Nishigauchi, Taisuke (1986) Quantification in Syntax. Doctoral dissertation, University of Massachusetts. Published by GLSA. Nishigauchi, Taisuke (1990) Quantification in the Theory of Grammar. Kluwer, Dordrecht Partee, Barbara H., ed. (1976) Montague Grammar. Academic Press, New York. Partee, Barbara H. (1984) "Nominal and Temporal Anaphora." L&P 7, 243-286. Partee, Barbara H. (1987) "Noun Phrase Interpretation and Type-Shifting Principles." In Groenendijk, Jeroen, DeJongh, Dick, and Stokhof, Martin, eds. Studies on Discourse Representation Theory and the Theory of Generalized Quantifiers, 119-143. Foris, Dordrecht. Penrose, Roger (1989) The Emperor's New Mind. Oxford University Press, Oxford. Pesetsky, David (1982) Paths and Categories. Doctoral dissertation, MIT, Cambridge MA. Pesetsky, David (1987) "Wh-in-Situ: Movement and Unselective Binding." In Reuland, Eric J. and ter Meulen, Alice G.B., eds. Representation of (ln)definiteness, 98-129. MIT Press, Cambridge MA. *Pesetsky, David (1993) "Topic...Comment." NLLT 11, 557-558. Platts, Mark de B. (1979) Ways of Meaning. Routledge & Kegan Paul, London. Portner, Paul (1990) "Quantification, Situations, and Gerunds." Unpublished manuscript, University of Massachusetts, Amherst.

170

On The Semantics ofWh-Clauses

Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey, and Svartvik, Jan (1985) A Comprehensive Grammar of the English Language. Longman, London. Radford, Colin (1966) "Knowledge-By Examples." Analysis 27, Ill. Reprinted in Roth and Calis, 171-185. Rescher, Nicholas (1962) "Plurality-quantification." Journal of Symbolic Logic 27, 373-374. Rooth, Mats (1985) Association with Focus. Doctoral dissertation, University of Massachusetts. Published by GLSA. Ross, John Robert (1967) Constraints on Variables in Syntax. Doctoral dissertation, MIT, Cambridge MA. Roth, Michael D. and Galis, Leon, eds. (1970) Knowing. Essays in the Analysis of Knowledge. Random House, New York. *Rothstein, Susan D. (1992) "Case and NP Licensing." NLLT 10, 119-139. Russell, Bertrand (1905) "On Denoting." Mind 14,479-493. Russell, Bertrand (1919) Introduction to Mathematical Philosophy. Allen & Unwin, London. *Sandt, Rob A. van der, (1992) "Presupposition Projection as Anaphora Resolution". Journal of Semantics 9, 333-377. Schubert, Lenhart K. and Pelletier, Francis Jeffrey (1989) "Generically Speaking, or Using Discourse Representation Theory to Interpret Generics." In Chierchia, Partee, and Turner, eds, 193-268. Schwarzschild, RogerS. (1989) "Adverbs of Quantification as Generalized Quantifiers." In NELS 19, 390-404. Soames, Scott (1982) "How Presuppositions Are Inherited: A Solution to the Projection Problem." L/13, 483-545. Soames, Scott (1989) "Presupposition." In Gabbay and Guenthner, eds., 553-616. Sportiche, Dominique (1988) "A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure." L/19, 425-450. Stalnaker, Robert C. (1972) "Pragmatics." In Davidson and Harman, 380-397. Stalnaker, Robert C. (1973) "Presuppositions." Journal of Philosophical Logic 2, 447-457. Stalnaker, Robert C. (1974) "Pragmatic Presuppositions." In Munitz and Unger, eds., 197-213.

Bibliography

171

Stechow, Amim von (1989) "Focusing and Background Operators." Arbeitspapier Nr. 6, Fachgruppe Sprachwissenschaft, Universitat Konstanz. Stechow, Amim von and Zimmermann, Thomas Ede (1984) "Term Answers and Contextual Change." Linguistics 22, 3-40. Strawson, P.F. (1950) "On Referring." Mind 59, 320-344. Strawson, P.F. (1952) Introduction to Logical Theory. Methuen, London. Strawson, P.F. (1974) "Positions for Quantifiers." In Munitz and Unger, eds., 63-79. Stump, Gregory T. (1985) The Semantic Variability of Absolute Constructions. Reidel, Dordrecht. Thomason, Richmond H., ed. (1974) Formal Philosophy. Selected Papers of Richard Montague. Yale University Press, New Haven. Urmson, J.O. (1952) "Parenthetical Verbs." Mind, 480-496. Westerstahl, Dag (1985) "Determiners and Context Sets." In Benthem, Johan van and ter Meulen, Alice, eds. Generalized Quantifiers in Natural Language, 45-71. Foris, Dordrecht. Westerstahl, Dag (1989) "Quantifiers in Formal and Natural Languages." In Gabbay and Guenthner, eds., 1-131. Wilkinson, Karina (1986) "Generic Indefinite NPs." Unpublished manuscript, University of Massachusetts, Amherst. Wilkinson, Karina (1991) Studies in the Semantics of Generic Noun Phrases. Doctoral dissertation, University of Massachusetts. Wunderlich, Dieter (1976) Studien zur Sprechakttheorie. Suhrkamp, Frankfurt. Zimmermann, Thomas Ede (1985) "Remarks on Groenendijk and Stokhof s Theory of Indirect Questions." L&P 8, 431-448. Zimmermann, Thomas Ede (1989) "Intensional Logic and Two-sorted Type Theory." Journal of Symbolic Logic 54, 65-77.

Index adverbs of quantification 6-7, 910, 12, 14, 40-41, 45, 72, 74, 81, 130, 155 as frequency quantifiers 52, 104-105, 106, 107 as quantifiers over situations 86-87 sentential scope 109, 123 unselectivity of 7, 10, 41 Altham, J. 14 Aoun, J. 122, 127 Aqvist, L. 20-23, 31-32, 42, 50 Asher, N. 138 Bacon, J. 14 Baker, C.L. 4, 26, 35, 60, 62, 69, 76, 81-83, 88, 124, 126, 129, 132, 134 Barwise, J. 9, 147 Bauerle, R. 18, 33, 81 Belnap, N. 24, 27, 43, 78 Bennett, M. 24, 27, 36 Berman, S. 76, 86, 136 Boer, S. 23, 27, 34. 35, 67-71, 102 Bolinger, D. 137 Bonet, E. 127 Bresnan, J. 76, 113, 135 Carlson, G. 8, 106, 135 Catalan 127 Chierchia, G. 9 Chisolm, R. 94 Choe, J.-W. 137, 139 Chomsky, N. 58, Ill, 118, 122, 124, 135, 139 Comorovski, I. 74, 125, 139 complete awareness hypothesis 90-93, 99

context set 152 conversational background 56 conversational implicature 82, 9698 Cooper, R. 9, 147 Cresswell, M. 34 D-linking 124-126 declarative sentence 24-25 as denoting sets of propositions 24-25, 29 Declerck, R. 46 default quantifier 7, 131 generic 7-8 universal 7, 41, 75, 143 definiteness 14, 78, 136, 149 definites 4, 119, 149-152, 154 and global accommodation 149 functional reading of 147 singular and plural 150 denotation-set 24-25, 26, 34, 35 Diesing, M. 76, 109-111, 127, 135 Discourse Representation Theory 4 Egli, U. 18-19 Elliott, D. 78 Engdahl, E. 141-145, 147 epistemic conditionals 45 epistemic modal 45-46 epistemic operator 21-22 exclamatives 78 exhaustiveness 33, 38, 81-89 intermediate 93 strong 33, 81-82, 88ff., 99-101 weak 81, 82ff., 88-89, 100, 125

174

On The Semantics ofWh-Clauses

existential closure of nuclear scope 11, 15, 46, 77, 107, 113, 137 extensional predicates 30, 37, 66, 81 factivity and presupposition 53-56 and Q operator 129 nullification of 67-69, 70, 71, 72 of predicates 53-54, 59, 78, 136; see also nonfactive predicates and negation 63 pragmatic variability of 60-66, 92, 95 predicates indifferent with respect to 62-63 Farkas, D. 8, 46 Fiengo, R. 137 File Change Semantics 4 Fodor, J.D., 125 free relatives 43, 74, 76, 119-122, 127, 134, 136-137 LF representation of 119-121 as restrictive term 48-50 free variable 8, 12-13, 22, 41, 46, 72, 110, 112-114, 117, 119, 139, 151 Frege, G. 149

Grice, P. 96 Gricean maxims 97 Grimshaw, J. 76, 129 Groenendijk, J. 4, 18, 38; see G&S Groos, A. 76

G&S (Groenendijk and Stokhof) 18-20, 24, 26-33, 35-37, 6667, 69, 81-83, 88-91, 93, 99102, 141-142, 144 Gallin, D. 36 Gazdar, G. 55, 63, 135 generalized quantifiers 9-10 generic quantification 8 German 3, 74, 79, 111 Gerstner, C. 107 Gettier, E. 94 GPSG 135 Grewendorf, G. 18, 35

Hailperin, T. 14 Hamblin, C.L. 4, 24-26, 32, 3435, 37, 128, 138, 158 Harrah, D. 18 Hausser, R. 14, 18-19, 150 Heim, I. 4, 6-8, 10-11, 14-15, 43, 72, 75, 77-78, 109-110, 118, 135-137, 149 Hendriks, H. 143 Hintikka, J. 4, 20-23, 31-33, 42, 50, 75, 77, 81, 94 Hirschbiihler, P. 18 Hiz, H. 18 Hoepelman, J. 18 Holker, K. 79 Hornstein, N. 122, 127 Huang, C.-T.J. 122, 139 Hull, R.D. 18-19, 33 if-clauses 50, 55,152 adjoined to S 110 as adjunct clauses 44-45, 48, 64, 66, 68, 104 as complement clauses 44-45, 155-158 as involving conditional necessity 7, 75 as restrictive term 42-46, 65, 76, 155 in nuclear scope 65, 76 presupposition of consequent 78 imperative operator 21-22 implicit default universal quantifier 7, 41

Index indefinites 4-7, 11, 13, 44, 73, 78, 81, 103-104, 106-107, 130, 132, 134-135, 144, 149, 151-152 ambiguity of 107 as existentially quantified 6, 14, 35, 74 as inherently nonquantified 6 as open sentences 5, 14, 41, 43, 109, 112, 150 default universal force 75 existential reading of 7, 11, 75, 106 nonquimtificational analysis of 6, 39, 41, 98. 103 quantificational variability of 5-6, 39-43, 46, 72, 74, 82, 87, 89, 129, 147, 150 index-shifting effect of embedding 30 intensional logic (IL) 34, 36 intensional predicates 30, 67 interpretational approaches to quantification 9 interrogative structures 69-72. 129 Jacobson, P. 76 Japanese 3, 74, 103, 109, 123124, 127, 134, 137 Jespersen, 0. 44, 76 Kadmon, N. 76 Kamp, H. 4-6, 10, 14-15, 137, 152 Karttunen, L. 4, 23-24, 26-27, 3236, 55, 58, 60, 67, 74, 81-82, 88, 92 Kasher, A. 14 Katz, J. 74, 129, 149 Kawasaki, N. 123, 137-138 Keenan, E. 18-19 Kiefer, F. 33

175 Kim, S.W. 74 Kiparsky, C. 53, 62 Kiparsky, P. 53, 62 Kitagawa, Y. 111 Korean 74 Kratzer, A. 7, 9, 12, 15, 43, 4546, 74, 76, 84, 101-102, 135, 152 Krifka, M. 107 Kuno, S. 137 Kuroda, S.-Y. 111 Lasnik, H. 122, 139 Latin 74 Lees, R. 76 Levi, I. 37 Levinson, S. 55, 58 Lewis, D. 5-7, 11, 14, 20, 29, 41, 43, 46. 56, 109 LF pied-piping 137 LFG 135 LHK (Lewis/Heim/Kamp) theory 17, 32, 39-41, 46-47, 57, 72, 74, 77-78, 81, 103, 107, 109110, 138, 149 formulation of 8-13 motivation of 5-8 Lindstrom, P. 9 Logical Form (LF) 15, 41. 79, 85, 108, 115-116, 118-128, 133135, 137-139, 141-142, 145, 149 mapping from S-structure to 109-113 mapping hypothesis 111-112, 114 Masunaga, K. 137 May, R. 110, 127, 138, 141 McCawley, J. 77-78 Milsark, G. 106 Montague, R. 24, 26, 36 Mostowski, A. 9

176 Munsat, S. 132 Nested Dependency Condition 125-126 4, 39, 74, Nishigauchi, T. 122-123, 137-139 nonfactive predicates 53-54, 59, 66 Novelty/Familiarity Condition 149 nuclear scope 8, 11-13, 15, 39, 41, 49, 53-60, 65, 72, 75, 77-79, 93, 99, 103, 107-114, 118-119, 128, 130, 137, 151, 156-157 objective sense of knowing 9496, 102 open sentences 5, 10, 12, 14, 39, 41, 43, 109, 112, 150 parenthetical verbs 138 Partee, B. 20, 46, 101, 143 Pelletier, F.J. 56-57 Penrose, R. 79 performative utterance 20 Pesetsky, D. 74, 108, 123-126, 136, 139 Peters, S. 36 Platts, M. 8 Portner, P. 120 Postal, P. 74, 129, 149 presupposition 23, 39, 50, 63, 67, 76-78, 103, 108, 114-120, 122, 124, 127, 134, 137, 149152, 156-157 factivity and 53-56 lexical 57 of definites 119, 151 of quantifiers 14, 138, 152 of nuclear scope 39, 53, 55-56, 58, 103, 118 pragmatic 60, 69-71, 76, 124

On The Semantics ofWh-Clauses semantic 76-77 triggers 58 vagueness with respect to 59ff. presupposition accommodation 56-57, 59, 65, 72, 78, 114-120, 122, 127, 134, 137, 152, 157 and LF 116.ff. effect on quantified sentence 56 global accommodation 78, 119, 149, 151 in free relatives 119-121 local accommodation 78, 118119, 121, 149, 151-152 sentential or discourse process 118 proportion problem for restricted quantification 15, 99 proto-question 34 Q morpheme 126, 129 Q operator 124, 129-130, 133, 138 quantificational variability 5-6, 39-43, 46, 72, 74, 82, 87, 89, 129, 147, 150; see indefinites and wh-phrases Quantifier Raising (QR) 110, 118, 126 questions 3, 5, 7, 13-14, 16-22, 31-37, 48-50, 52, 60, 73-74, 77, 82-83, 99-100, 119-122, 127-129, 132, 136-138 and coordination 18-22, 25, 29, 32-34 as denoting propositions 29-31 as denoting sets of propositions 24-29 as A.-expressions 19 associating with truth conditions 17 categorial approach to 17-19, 22

Index complete true answer of 28, 27-28 coordination with declaratives 20, 29 direct 19-20, 128, 141 embedding approaches to 20, 33 embedding of 19, 23 existential reading of 21-23, 31, 34 imperati ve-epistemic approach to 21-23 indirect 22, 31, 76, 85, 138 mention-some readings of 28, 30-31 partial answers to 27, 82-83 pair-list reading of 144 propositional approaches to 17, 24 propositional content of 21-22 quantifying into 36, 142 'Reduktionstheorien' of 33 semantic type of 18-19, 121 true answers to 26-28, 60, 83 universal reading of 22, 34, 75 universal/existential ambiguity 32 Quine, W.V.O. 78 Quirk, R. 44, 47-48 Radford, C. 94 reconstruction 137 Reinhart, T. 75 relative clauses 14, 74 representational approaches to quantification 9 Rescher, N. 8 restricted quantification 8-12, 14-15, 39, 41, 43, 53, 85, 143, 157 restricted quantifier 8 restrictive term 8-13, 15, 39, 41, 43, 65-66, 72, 77-79, 92-93,

177 103, 107-112, 128, 130-131, 134-137, 143, 150-152, 157 derivation of 50.ff. free relatives as 48-50 if-clauses as 43-46 interaction with presupposition 50ff., 59ff. syntactic representation of 114.ff. when- and where-clauses as 4648 wh-clauses as 42.ff. Reyle, U. 10, 14-15, 152 Riemsdijk, H. van 76 Rooth, M. 34 Ross, J.R. 76 Russell, B. 6, 149 Russian 136 Saito, M. 122, 139 Schubert, L. 56-57 Schwarzschild, R. 9, 85 segments of a node 138 semantic selection 119, 121, 129, 136 semantic type 18-19, 121 singular/plural distinction 14, 75, 150 Skolem function 142, 147 Soames, S. 55-56, 58, 77 specifier 135 Sportiche, D. 111, 122, 123, 127, 135 Stalnaker, R. 56, 77 Stechow, A. von 34, 74, 138 Stokhof, M. 4, 18, 38; see G&S Strawson, P.F. 76, 149-150 strict cyclicity of movement 139 strong domain hypothesis 89-91, 93, 96, 100 Stump, G. 46 subcategorization 120-121, 139 Subjacency 139

178 subject positions in S-structure 110-112 success presupposition 23 Sugioka, Y. 8, 46 Superiority Condition 124-125 Tennant, N. 14 that-clauses 21, 23, 26, 29, 34, 50-52, 69-70, 78, 83-84, 88, 151-152 two-sorted type-theory (Ty2) 36 typeshifting 20, 25, 29, 143 Urmson, J.O. 138 vacuous quantification 12, prohibition against 117-118 variable assignment 10-11 verbs of communication 62, 10ff, 95 Westerstahl, D. 9, 154 wh-clauses passim as restrictions on quantifiers 42-50 as questions 52, 128-133 inventory of kinds of embedding predicates 60-61 subcategorization for 139 wh-in-situ 132 wh-movement at LF 108-109, 111, 118-119, 121-127, 133-135, 137, 139, 141-142, 145, 149 in Japanese 74, 103, 123 in multiple-w h-constructions 123-124 landing site for 127, 139 wh-phrases passim functional reading 143-147 multiple 123-4

On The Semantics ofWh-Clauses quantificational variability of 40-42 wide-scope property of 105-107 nonquantifiable 51-52, 128133 unselectively quantified 41 when as a temporal sortal 47-48 when- and where-clauses 46-48 where as a spatial sortal 47 whether-clauses 29, 155-158 strong exhaustiveness as universal conjunctions of 88 which-phrases 78, 100, 135-136, 149-151 as definites 150f Wilkinson, K. 8, 107, 135 Wunderlich, D. 18 X-bar theory 111 Yiddish 127 Zaefferer, D. 18-19 Zimmermann, T.E. 18, 33-34, 36, 81, 101, 138, 147

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 2

THE ANATOMY OF LANGUAGE

THE ANATOMY OF LANGUAGE Saying What We Mean

MARJORIE BOULTON

First published in 1959 by Routledge & Kegan Paul Limited This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1960 Marjorie Boulton All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69070-7 978-1-315-53675-0

(Set) (Set) (ebk) (Volume 2) (hbk) (Volume 2) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

THE ANATOMY OF LANGUAGE SAYING WHAT WE MEAN

by MARJORIE BOULTON M.A., B.Litt.

ROUTLEDGE & KEGAN PAUL LONDON, HENLEY AND BOSTON

First published in 1959 by Routledge & Kegan Paul Limited 39 Store Street London. WC1E 7DD, Broadway House, Newtown Road Henley-on- Thames Oxon. RG9 1EN and 9 Park Street Boston, Mass. 02108 USA Reprinted in 1968, 1971 and 1978 Printed in Great Brita in by Lowe & Brydone Printers Limited Thetford, Norfolk

©

Marjorie Boulton 1960

No part of this book may be reproduced in any form without permission from the publisher, except for the quotation of brief passages in criticism ISBN 0 7100 6288 5 (c) ISBN 0 7100 6351 2 (p)

To IVO LAPENNA a master of words who has given tne a better understanding of the meaning of such words as courage, loyalty, honour and friendship

CONTENTS X1

FOREWORD

PART I

Understanding and Misunderstanding page 3

(. WHAT ARE WORDS FOR? II. THE KEY TO UNDERSTANDING

m.

SOME CAUSES OF MISUNDERSTANDING

IV. VAGUE WORDS AND EXACT WORDS

V. LET US LOOK AT SOME WORDS VI. WORDS THAT HAVE CHANGED VU. INTERNATIONAL SEMANTICS

13

22

36 47 61

7I

PART II

The Misuse

of Language 83

VIII. ADVERTISING

IX. WHAT IS SENTIMENTALITY? X. WHAT I TELL YOU THREE TIMES IS FALSE

92 102

XI. WHITEWASH LANGUAGE

113

XII. LANGUAGE OF QUARRELS

122

XIII. LYING TO OURSELVBS

131 IX

Contents PART III

Literary Semantics XIV. RHETORIC XV. METAPHOR AND ASSOCIATION INDEX

X

FOREWORD

M

purpose in writing this small book was to stimulate a thoughtful, critical and socially useful interest in the meaning of \vords. I have tried to relate problems of elementary semantics to everyday human problems. As I have written the book mainly for older schoolchildren and students, I have tried to choose, as far as possible, examples that may be relevant to their urgent concerns; thinking of a school audience, I have also made a few concessions to conventionality, for which I beg pardon. I hope the book may also be helpful to older people who are interested in language but have no specialized knowledge in this field of study. My book is a mere popularization of what can be a highly academic subject. This is fully intentional: formal and scholarly studies of semantics can be inexhaustibly fascinating to a minority; but I have written for people who will never read such studies. The quest for truth and integrity is not confined to philosophers and lexicographers; it is a true lifetime cause fit for many a young rebel; and our sense of the meaning of words touches our private lives very nearly. I am deeply grateful to Professor W. E. Collinson for checking the chapter on International Semantics. I am also grateful to all the unintentional contributors of examples. I believe I have made no discoveries in semantics, but have merely made some old knowledge rather more accessible and digestible; yet I dare to hope that this little book may provoke Y

XI

Foreword thought and perhaps, here and there, make some small contribution to human happiness. M. B.

Stoke-on- Trent, October 1958.

xu

PART I

Understanding and Misunderstanding

I. WHAT ARE WORDS FORl

M

is the only creature that uses words intention.. ally and habitually. A parrot or a mynah may talk in a fashion, but it would be impossible to discuss any subject with a talking bird. We use words in order to tell somebody somethlng, that is, to communicate. Speech and writing are not essential to human life; deaf and dumb people and idiots who cannot learn to speak may live if their bodily necessities are provided and may even, it seems, enjoy some satisfactions; but all human beings who are able to do so speak sometimes. Indeed, for most of us prolonged silence is distressing. 'I shall never speak to you again!' is regarded as a serious threat, a mark of the most extreme displeasure; and some ascetics regard silence as one of their mortifications. We all seem to have a natural wish to communicate with other people. We may wish to communicate various kinds of information, for example: Facts. 'The train will leave from platform four at 11.20. There will be a restaurant car in the rear of the train. The two front coaches are reserved for passengers joining the train at Doncaster.' Commands or perhaps wishes. This time we do not merely tell someone somethlng; we try to make the person to whom we speak do somethlng, or perhaps we express our desires that no one can satisfy: 'Bring me a glass of water.' 'Hold that prop there and don't 3 AN

What are Words For? move!' 'I wish I had masses of wavy raven hair, liquid dark eyes and long eyelashes !' Emotions. 'I do so love you. You are wonderful. You are everything in the world to me.' 'You make me sick! I can't bear the sight of you!' 'I've won the marmalade competition, ohjoy,joy! joy! Hooroo! Oh glorious day, peal all the bells, bless all the peel!' Arguments. 'If the world is really a flat disc, as you say, why is there no record that anyone has ever fallen off the edge?' Sociable noises. A good many words we use mean hardly anything except that they are conventionally taken as implying that we are fairly well disposed to the person addressed and wish to remain on good terms with him or her. For example, to reply to the question 'How do you do?' with a detailed account of our last medical examination is not at all what is expected. 'Excuse me!' 'I beg your pardon!' 'Please ... ', 'Don't mention it ... ' are examples of such sociable noises. Often when people talk about topics in which they feel no real interest such a statement as 'It's a fme day!' or 'This train is very slow!' is not so much a comment on the weather or on the speed of the train as a noise meant to suggest that the speaker wishes to be civil and sociable. If someone sees me looking about the table, guesses my need correcdy and passes me the salt, I am likely to say 'Thank you very much.' There is a great deal of difference between such a 'Thank you very much', which is more a piece of normal politeness than the expression of an emotion, and 'I am most grateful. . . . I shall never forget your kindness !' which is the kind of expression we use when someone has done a real favour, important to our welfare and probably requiring some trouble or sacrifice. Magic. In primitive communities the idea that words have magic properties is still very much alive. Today in England it is found mosdy in a few taboos: for example, some people 4

What are Words For? show a degree of horror at certain swear-words which is out of all proportion to their literal meaning; children are sometimes afraid of particular words and will not even pronounce them. Ignorant or naive people sometimes mix up the idea of prayer with the idea of magic. Primitive people often repeat certain words to make the crops grow, to kill their enemies and so on, and believe that to know a person's name is to have power over them. (And why is it common even today, for a hostile person in anger to address the enemy by the full name instead of the first name or surname as is usual?) Are magic words communication? Perhaps in a sense they are commWiication with supernatural powers. Enquiry. We may use words not to communicate acrual information, opinions or emotion, but to communicate our own desire to know something. 'Can you tell me the way to the cathedral?' 'What do you think of Phil Mugg?' 'Are you upset about something?' Pure joy in words. People with a fair command of words very often enjoy using them, and practise what we might call redtu1dant communication for the fun of saying things as elaborately and eloquently as possible. For example, in Shakespeare's Henry IV, part I, the Prince wishes to say that Falstaff is much too fat and Falstaff wishes to say that the Prince is ridiculously thin; but their essentially repetitive, yet delightfully varied, invective is much more amusing than a simple statement of weight would be:

Prince: . . . this sanguine coward, this bed-presser, this

horseback-breaker, this huge hill of flesh; Falstaff: 'Sblood, you starveling, you elf-skin, you dried neat's tongue, you bull' s pizzle, you stock-fish! 0! for breath to utter what is like thee; you tailor's yard, you sheath, you bow-case, you vile standing tuck; (Act II, scene iv.) 5

What are Words For? Very often we cannot classify the reasons for using words exactly; for example, it is not always easy to sort out the precise proportions of fact, emotion, opinion and wish in a remark. What happens when we speak without any audience? We are still, in a sense, communicating, for we are communicating with ourselves, telling ourselves to do something, encouraging ourselves, telling ourselves agreeable stories about ourselves (day-dreams), and formulating our emotions. Most people know that to put our emotions into words, even in complete solitude, usually makes it easier to deal with them. We may even talk to ourselves with the specific purpose of communicating with a part of our mind not fully under our own control. This is auto-suggestion-intentional auto-suggestion. It is a better instrument than an alarm-clock for getting up at a particular hour; for instance, to say to oneself firmly several times, just before going to bed and in bed, 'I shall get up at six o'clock' is a fairly good way of being certain to do so; and similar auto-suggestive tricks can improve our health and help in various fields of activity. So though we appear not to be communicating when we talk to ourselves, we may in fact be· conversing with our 'other selves'-an expression common enough. Though the function of words in our lives is to communicate, we do not communicate by words alone. Young babies express emotions and communicate their wants by screams and other non-verbal noises; and so do most of us under the stress of very strong emotion. If I see a girl weeping, wish to say something helpful and cannot think of any useful expression, to place a hand on her shoulder niay be enough to communicate what she will feel as a good deal of sympathy. If, on the other hand, someone is angry with me and indicates the door with a dramatic gesture I shall not suppose that he wants me to eat nuts and raisins with him. Often when we are in a 6

What are Words For? foreign country and do not know the language we resort to more elaborate gesture as a means of expressing our feelings, making requests or being polite to people. Pictures may be a form of communication, either as the work of artists who wish to share their vision with others, or as actual picture stories (comic strips where words are absent) and primitive picture-writing. Pictures, however, may be misunderstood; Kipling's story How the Alphabet was Made illustrates this possibility brilliantly; and there is the old story of the tourist in Spain who, having sketched an ox and a mushroom to indicate that he would like a steak with mushrooms, was given a sunshade and a ticket for the bullfight. A recent publication for tourists who can neither learn languages nor draw consist entirely of clear pictures of things a tourist may be expected to need; it should be very useful; but it would be difficult to say in pictures 'I will do this if it is legal in your country' or even 'I am allergic to eggs'-which is concrete enough, but not easy to represent pictorially! Music has sometimes been said to be an international language. It does certainly communicate; in particular, some music communicates moods and emotiorJ.s; but it does not, on the whole, communicate the same things as words communicate; it is certainly no substitute for words. One of the differences, so far as we know, between hutnan beings and animals is that while the courtship and mating of animals is entirely material, sensitive human beings often feel a kind of personal communication when they caress each other. A clasp of the hand, a kiss, an embrace, may be much more than a pleasure; it may be an expression of emotion, sometimes at least seeming to be more adequate than verbal expressions. Friends as well as sweethearts often caress one another as an expression of emotion-welcome, sympathy, reconciliation, congratulation, gratitude and so on. Indeed, there have on occasions been tragedies because a communica7

What are Words For? tive caress between friends was misinterpreted as an expression of passion. However, there is no substitute for words. All other forms of human communication are, by their very nature, confmed to certain fields of expression: either, as with pictures, to the simplest concrete ideas, or, as with caresses and inarticulate noises, to the expression of emotion. Only words can communicate nearly the whole range of human experience. (Mystics and those who have experienced very deep love, with a few others, will admit that some of their experiences cannot be described in the language at their disposal.) Thus words are very important if we are not to live in solitude. Unfortunately we do not naturally know how to use words; even the most intelligent and articulate people quite often have to say, 'I don't quite know how to put this ... ' and human beings often say to each other something like, 'Why can't you say what you mean?' We have to train ourselves to use words properly, not only from the point of view of the grammarian, but from the point of view of meaning. It does not really matter much if we say, for instance, jimmy has eated Peter's cake !' instead of eaten; we shall still be understood; but if we say, 'Jimmy is always a wicked boy and a thief!' under the provocation of a single, perhaps not at all typical, act of cake-stealing, we are seriously n1isusing words and may do Jimmy great harm. Words, alas, can be misused. They can be used to lie, to evade, and as an excuse for not acting. No more skill in using the vocal organs, no more intelligence, is required to say 'Jimmy has stolen Peter's cake' if this statement is untrue. The idea that it is better to speak the truth than to lie does not seem to be born in us; moreover, very few people live in situations in which it is possible never to tell an untruth. The question of deliberate truth-telling and deliberate lying is a moral question; we can ourselves choose to do one or the other, 8

What are Words For? and the importance vve give to the matter, together with the views we hold on truth-telling-for instance, whether we may tell an untruth to protect someone else, or to protect ourselves from obviously bad and vindictive people-involve questions outside the scope of this book. Words can be used to evade problems and conceal situations. Much propaganda by governments has something of this character. 'The rioting is now at an end' says one governnlcnt when the tnarket square is strewn with dead \vorkmen. 'Reactionary elements have been eliminated' says another, when its prisons are full of suffering and bewildered people who have done nothing generally recognized as wrong. 'We are looking into your complaint', says the firm, meaning 'We hope you will forget about it if we delay long enough.' Words can also be used as an excuse for inaction. It is much easier to sympathize profusely with someone than to help, to make some beautiful noises about our country or our cause than to make sacrifices, to reproach ourselves for our faults than to make the effort to overcome them. Once in a Midland chapel some men began to talk about another member of the congregation who was absent that evening. He had been seriously hurt in an accident at work, compensation was being delayed by legal arguments, his wife was ill and the whole family was in great difficulties. Everyone began to express regret: 'How sad that such a fine man should be in such trouble!' 'Makes you wonder why these things have to happen!' 'Such a pity, and they are such a happy couple.' 'I am very sorry for them.' This went on for some minutes until a simple but sincere tnan, hearing the twentieth 'I feel so sorry for them ... ' suddenly put some notes on the table and said, 'I am sorry five pounds. How much are you sorry?' The result was a large collection that relieved the family's immediate needs. One of the problems of life is to know when to stop talking and begin acting; if we do it too soon, 9

What are Words For? we may make wrong decisions through our haste, and if we leave it to~ late action may be useless. Most of us have had the experience of hearing someone stir up a great dust of words so that people cannot think. Politicians frequently do this; so do advertisers; so do most of us in family or personal quarrels. These uses of words, however, are a matter of wrong conduct not of intellectual mistakes. This book is concerned with saying what we mean when we intend to say it, and understanding what other people mean. The study of the meaning of words is called semantics or semasiology. The former word is the one more often used for the popular or. elementary study of the subject and is the word I shall use hereafter in this book. It might perhaps be said that the aim of popular semantics is to teach us not to lie except on purpose. If we want to lie on purpose, we need moral, not semantic, guidance, just as there is a difference between the man who falls over the kerb because he did not see it, and the one who falls where there is no kerb because he has had too much beer. A great deal of adv2nced research into semantics is being done at present. Much of this is very specialized and difficult, requires a special vocabulary and is unintelligible except to a reader trained in linguistic studies, or philosophy, or preferably both; but everyone can study elementary semantics with advantage. Many people will say at this point, 'But it is not hard to know what words mean! We know the meaning of all the ordinary words, and when we hear or see hard unusual words we can look them up in a dictionary!' In fact the question of what words mean is not as simple as it looks. Suppose we take a very simple word such as air. We do not know what this word means until we have a context for it. It means something different in each of these sentences: 'She strolled in the garden humming a cheerful air.' 'This IO

What are Words For? room is stuffy; we need some fresh air.' 'I don't like his arrogant air.' 'Air my shirt and brush my coat, please.' The easiest words to define are the words that look difficult, the teclmical words, for a teclmical word generally means only one idea: vertebrate, anacoluthon, longitude, ampere, are much more sharply definable than such a word as door, or beautiful, or even blue. It does not seem likely that even specialists in semantics use words as precision tools all day. When such an expert finds that his car will not start, he is quite capable of saying something like 'This diabolical, infernal tin can is misbehaving again !' though the car is not diabolically possessed, was obtained from a second-hand car dealer and not from hell, is not a tin can, and, having no moral consciousness, cannot misbehave. However, once we become conscious that saying what we mean is quite a difficult matter, we shall at least be tnore careful in the use of words when they are importantand more suspicious of words when suspicion is no bad thing. Once I had a class of charming girls whose written work was rather slovenly. Tired of covering their exercises with red streaks, I one day preached them a fierce little sermon on the subject of saying what we mean. 'Don't use words that mean nothing!' I said sternly. 'Don't waffle and pad! Words are supposed to mean something! And be careful how you use words; say what you mean and not something rather like it.' In this vein I continued for some minutes. At the end of my little sermon I said, 'Now, at least for today try to say what you mean and not something vague or inexact.' One girl, who had a sweet and candid nature, but who had probably never before encountered the idea that saying what we mean was an aspect of sincerity and the quest for truth, must have listened with great attention, for later she was asked a question: 'By the way, who wrote Kubla Khan?' 'It was Shelley.... ' She saw the expression on my face. II

What are Words For? 'I think it was Shelley.... 'I gather that the expression on my face did not improve, for she fmally adopted the version: 'At least, er, I hope it was Shelley!' She had grasped the general idea.

12

II. THE KEY TO UNDERSTANDING

C

we defme understanding, from the point of view of the person who hears or reads the words, and saying what we mean, from the point of view of the speaker or writer? Obviously these are two sides of the same coin. Understanding is itself a rather difficult word to define: for example, a person who has had the experience of love or bereavement will in one sense understand a love poem or an elegy far more than a less experienced person; a historian and a plumber may both perfectly well ooderstand an article about some urns and bones recently dug up in Mesopotamia, but the historian is likely to perceive implications in the facts that the plumber cannot perceive; two people may both understand an order to do a certain piece of work, and carry out the order, but one may do it better than the other, having more knowledge, more skill, or more experience of what will best please the employer. For the moment we are concerned only with the verbal ooderstanding of words: that Mary, told 'Please bring me my hat' shall bring a hat, rather than gloves or a piece of cheese, not that she shall realize that the speaker wants her mauve straw hat, not her brown felt hat. A working defmition that will do for the semantic problems of everyday life is: understanding is adequate when a word or group of words means the same to speaker and hearer, or writer and reader. This may be illustrated by reference to a few material 13 AN

The Key to Understanding objects first. A man from Staffordshire who said to a Scotsman: 'Roll up your sausage in an oatcake' would seem to be giving ridiculous advice. Staffordshire oatcakes are large, limp and pliable; Scots oatcakes are smaller and more like biscuits. To me the word starved means 'in serious need of food'. I am somewhat stout, and was much surprised some years ago when, in another part of England, I called on an old lady on a frosty day and was cordially welcomed with the words: 'Come in, come in, my poor dear, you look quite starved!' In her part of the country starved meant cold. To some people a tabby cat is a cat striped in a characteristic pattern, to others a female cat. Most English people use the term salame to refer to any continental sausage of the rich, hard, solid kind that one can hang up in the kitchen and keep for a long time; but when a Pole offered me something of this kind and I said, 'Ah, thank you; I like salame very much ...' he was quite annoyed, saying, 'This is not salame, this is proper sausage. Salame is made from donkey meat!' As soon as we move into the realm of more abstract ideas the possibility of misunderstanding increases. What are young women? Women under twenty-five? thirty? thirty-five? or even forty? To a child of ten a forty-year-old may seem so old as to be utterly remote, incomprehensible, uncomprehending. Yet a fifty-year-old may make the often disastrous mistake of treating an eighteen-year-old like a child, because to him or her the eighteen-year-old seems so very young. 'The Browns are a wonderfully happy couple.' 'You are quite mistaken; it shows how appearances deceive; I once saw them have a nasty quarrel and shout at each other.' 'I can believe that, but fundamentally they are a very happy couple; they care about one another, their temperaments are suited and they are very loyal to one another.' 14

The Key to Understanding 'Well, I don't call that happiness; it isn't a bit like what you see on the pictures. It sounds very dull.' These two people may both be quite sincere; but their definitions of a happy marriage would be considerably different. Even an apparently simple phrase like enough to eat may need to be carefully examined. Such a question may be very important to someone's happiness: 'Are the pLisoners getting enough to eat?' 'The children in the orphanage must have enough to eat, but we cannot afford to give them more than enough: Does enough to eat mean enough to keep a person alive, in whatever state ofhealth; enough of a variety of foods to keep a person in good health; enough food to cause an increase in body weight; or as much food as the person wishes to eat? Obviously the interpretation put on the phrase may make an important difference to the diet. The sentence I love you may reasonably lay claim to be the most beautiful sentence in the English language. It is also one of the most tragically ambiguous. Parents say they love their children, and cramp the children's lives, expect disproportionate returns of love and never really relinquish their children to the marriage partner or the vocation. Friends say they love, and monopolize. Almost every adult person has probably some memory of suffering caused because a man and a woman, perhaps both well-meaning, used the words I love you in two different senses without realizing it. It is important that we should know what we mean by love in ourselves and what we expect other people to mean; and if there is any doubt whatever, the matter needs to be discussed. For 'I love you' may mean anything from 'I value your personality so much that I would like to do all I could to help it to grow and be happy' to 'I want you all to myself and you must never feel any affection or desire for anyone else', and even to 'I want to kiss you so much that it is cruel of you not to let me.'

IS

The Key to Understanding The fact that two people can tmderstand the same words in different senses gives rise to a great deal of amusement; it is the cause of many puns and double-meaning jokes. A friend of mine from Yugoslavia, a very good needlewoman, once saw me in a bad state of nervous tension, and said, 'My sewing helps me to be calm: Marjorie, what a peety you are not a sewer!' I knew a doctor, a man of excellent physique, personal charm and chivalrous manners, who was married very happily and had two children. One day he was asked to visit a sick girl in a farmhouse. It was harvest-time and every well person was out of doors; so no one let him in and he walked slowly upstairs to look for the girl's bedroon1. .As his steps sounded on the stair, the girl began to scream. 'It's all right, it's all right!' he reassured her, but the screams grew more piercing as he approached. As his head appeared round the bedroom door the sick girl suddenly became calm and exclaimed, 'Oh, Doctor, I'm so glad it's you! I thought it was a man!' I was once at a conference where a number of ladies, all most respectable, sat at a long table for lunch. One of us was asked to serve the food. For each course there was a choice of two dishes. We all tried to help by passing our requests to the server. One lady caused some amusement by saying cheerfully, 'I'm a rissole!' but when the sweet couxse arrived there was quite a scandal. Someone who had been helpfully enquiring what her neighbours preferred said in a fme loud conference voice, 'We're all tarts at this end of the table !' 'I can't stop for pudding,' said a busy woman at another lunch. 'I'm all behind already.' The unfortunate thing was that the lower part of her torso was indeed rather disproportionately large. I was lecturing and noticed that a student fidgeted constantly. I raised my eyebrows at her once or twice; no result. I tried to make my lecture more interesting; she still wriggled 16

The Key to Understanding and shifted. Eventually I could stand this no longer and asked her: 'Are you unwell, or what is it?' 'No, thank you; I'm sorry to keep moving, but I think I must have a screw loose!' She undoubtedly meant, in the tubular chair on which she was sitting, but her classmates did not take it that way. Recently I greatly enjoyed an Esperanto lecture tour in Sweden, where I met with magnificent hospitality and was entertained by a number of very charming, kindly and interesting people. I think I must have talked a little too much about my trip, for on one occasion a colleague at the lunch table passed me the vegetable dish with a very pointed, 'You like swedes very much, don't you?' I fell into the trap, too, with a sincere and innocent, 'Oh, yes!' Such ambiguities have their uses. They amuse people and spice our lives. The really dangerous ambiguities are not those which we all notice and which make us laugh, but those which we do not notice at all. 'Enough to eat' ... 'I love you' ... 'Civilized behaviour' ... 'my Christian duty' ... what do we mean by these expressions? It may be very important. 'The prisoners all have enough to eat' we are told, and are satisfied that they are treated with humanity in the matter of food. In fact no one has died of starvation: the prisoners are merely so weak that they cannot work and ridden with scurvy. 'I love you!' says a boy to a girl, and she thinks that when they are married life will be as easy as eating strawberries, a continual ecstasy. Poor girl ... yet her boy may mean something quite sincere and quite worth having. 'Civilized behaviour' to some people n1eans going to theatres and listening to long-playing records, knowing which fork to use and not wearing curlers in the street; to others it may imply trying not to be jealous, tolerating the opinions of others, and setting differences by argument and compromise rather than a punch on the nose. A 'von1an once declared that she burned her small child with a 17

The Key to Understanding hot poker for stealing a few biscuits because it was her 'Christian duty'! I suppose there has never been a political speech or an election address that did not contain examples of language whose meaning was doubtful-even when the words were used in good faith. Two people who officially belong to the same religious group may be found, if their ideas are carefully investigated, to have very widely different concepts about the things they say they believe. A person who says, 'I believe in discipline!' may be anything from a cruel martinet, always punishing, rebuking and regimentating, to someone who merely has a preference for habits of hard work, punctuality and self-control. What does the word 'democracy' mean? To some it refers mostly to constitutional qualities and is almost a synonym for parliamentary government with some safeguards for civil liberties; to others it is mostly a matter of the absence of class distinctions; and to some people in eastern Europe it appears to refer to economic rather than political organization and to be used in quite a different sense from that understood in Arnerica. Let us take two simple words, each of one syllable, each used by everyone at least once a day. What do we mean when we say 'a good man' and 'a bad man'? I am fairly certain that a discussion group of ten people, especially (but not necessarily) if they were drawn from different religious groups and social classes, and were of assorted ages and both sexes, would produce ten different defmitions. This is partly because we all tend to value most those virtues that come most easily to us, and to condone most readily those faults that we fmd it hardest to overcome; partly because we have been brought up with different assumptions about goodness and badness. For example, I have actually heard it said more than once, 'I have been lucky: I had very good parents, very strict.' To me tlus is a contradiction in terms: I, though I do not believe !8

The Key to Understanding in anarchy in the home, could never be convinced that a strict parent can possibly be a truly good parent, or that someone with strict parents can possibly be thought of as lucky; and I could defend my views in much detail. I was once paid a magnificent backhanded compliment when a young girl, of what most people would call very decent character, came, in great anxiety, to consult me about one of those trivial sexual problems that can reduce an adolescent to obsessive misery and can be set right in five minutes with a little rudimentary biological information. I gave the girl the information she needed, reassured her and made her happy again, then said, 'Now, if ever you are worried about anything like that, remember I am here; don't make yourself miserable, but come and see me and I will try to help you.' I then added, not wishing to usurp the functions of a parent, 'But of course the best person to confide in over things like that is your mother.' 'Oh,' cried my young friend, with intense feeling, 'I couldn't possibly talk to my mother about a thing like that! You see, my mother is a good woman!' 'Thanks very much!' I said, laughing; but the remark threw a lurid light on some people's notion of goodness. Most human beings would, if asked, say that justice or fairness was something good, desirable. But justice is not as easy to defme as we may at first think. Suppose, for example, two men each commit the crime of stealing the same sum of money, each from an infirm old woman who kept the same type of shop, and each at dusk. If these cases were reported with only those details in the newspaper, on the same day, ahd one man had been sentenced to two years' imprisonment while the other received a conditional discharge, most people would be shocked at the 'injustice'. Yet suppose one thiefhad been a strong, healthy man of thirty, capable of work, in an area where there was little unemployment, a man with a comfortable home, who was known to the old woman and 19

The Key to Understanding had received various kindnesses from her; and the other thief had been a man whose mind had been disturbed by horrible war experiences, whose wife was in hospital recovering from a dangerous operation and who desperately wanted more money to buy delicacies for her? He would still be doing wrong, but most of us would agree that his crime was not as deserving of severe punishment. I do not care to commit myself to opinions on every detail of politics: I know just enough about politics, from a reading of numerous books and two responsible papers, to be aware that I do not know very much. (This is unfortunately a rare and very valuable piece of knowledge.) In general, however, I know that I stand somewhere to left of centre, that I an1 strongly opposed to all forms of totalitarianism, that I think force in politics should be avoided as much as possible, and that political behaviour should be governed mostly by what many would call humanitarian considerations; to me the business of government should be to pron1ote the happiness of the people governed and, as far as possible, of the rest of the world. I am also very much aware that considerations of immediate expediency do often have to take precedence over ideals for the time being, and that the most desirable reforms can often not be made in a hurry. With this kind of moderate and tolerant left-wing-liberal-humanitarian-rationalist political backgroWld, I have in the course of time been called: A Fascist sympathizer, a Bolshevik, a traitor to my country, hopelessly naive, far too subtle, a dreamy idealist and a terrible cynic; I have also been bitterly and torrentially scolded for not caring enough about the agony ofWorld War Two, and lectured at considerable length about the stupidity of caring as much as I do about the sufferings of the World War Two victims. I hope I deserved none of these things; but I certainly cannot deserve all of them. Name-calling in politics is very largely the result of slack definitions, and will be discussed 20

The Key to Understanding more fully in the next chapter. If political name-calling remained an affair of words, it would not much matter except as a breach of courtesy; but name-calling not infrequently leads to blood-shedding or at least to victimization; so it does matter very much.

21

III. SOME CAUSES OF MISUNDERSTANDING

M

of words can arise in many ways other than lying on the part of the speaker or downright stupidity on the part of the hearer. Some common misnnderstandings may be classified under headings. ISUNDERSTANDING

I. MALAPROPISM

In either (a) the speaker or (b) the hearer. The word is well known to be derived from the popular character Mrs. Malaprop (Mal apropos-inappropriate) in Sheridan's comedy The Rivals. Mrs. Malaprop wants to appear well-educated and uses long words without knowing their meaning, saying epitaphs for epithets, odorous for odious and so on. I have been informed that the housewife using an electric stove should make full use of the residential heat (residuaQ; and that diet should be suited to the digestibility of the family ! The schoolboy with a passion for aeroplanes thought an aviary was a place in which research of aviation was carried out; and a timid girl said she did not like parties because she was not really socialist (sociable). I heard of a woman who was disproportionately offended when a welfare worker advised special teaching for her illiterate young son, and angrily produced her marriage lines. (Illegitimate being the word she thought she heard.) I was much puzzled when in the school laboratory I first saw a desiccator-a lidded jar containing a 22

Some Causes of Misulzderstanding water-·attracting substance and used for drying things-for I knew only the desiccated coconut in the larder at home and could not see how a jar could possibly cut any substance up into tiny pieces. There was the engaging little girl who, told that twins were two babies born at the same time, said, 'Please, miss, I know what it is when there are three-giblets!' There may also be the malapropism of pronunciation: I was once told 'Don't step on that poodle!' by an Oriental, who was in fact warning me not to step in a large puddle. I was once asked by a Japanese girl if I could possibly send her a 'small rocket'. It took some time to realize that what she wanted was a small locket. If a keeper in a zoo warned me 'There's a moose got out!' it would be useful to know if he were English or Scottish! There was recently an argument in the Press as to the precise degree of poverty suffered in an Eastern country. Some said the country people were reduced to eating mice; but others said that this was a mishearing or mispronunciation over the telephone of maize. Most misunderstandings that arise from malapropism can be cleared up by someone who sees what has happened, or by reference to a dictionary. It is, however, wise to remember that a dictionary can be very misleading. A tetrameter does indeed 'have four feet', but a rabbit is not a tetrameter. Claustrophobia, 'fear of being confined' must not be further interpreted as reluctance to have children. Mism1derstandings caused by malapropisms are generally amusing, at least afterwards, and relatively seldom do harm. More serious are the misunderstandings that may arise out of 2. DIFFERENCES OF DEFINITION

These may arise even in the ordinary details of practical life, as with oatcake or salcme quoted in the previous chapter. What is a house? This question may be important with regard 23

Sonze Causes of Misunderstanding to rates, taxes and other legal obligations. There is a story told of a biologist who took a large nun1ber of specimens on a train many years ago. He asked the booking clerk what tickets he would have to buy for his various animals. The booking clerk was not sure and sought the help of the station-n1aster. Together they thumbed through the regulations and finally came to the following startling conclusion-: 'Well, seen1s as if dogs is dogs, and bear cubs is dogs, and anteaters is dogs, and wild cats is dogs, and kestrels is dogs, but tortoises is insects, and goes free!' There was a considerable difference here between zoological and railway-regulation definition. What is a vehicle? For example, if someone lives in an old bus, is it a house or a vehicle with regard to legal obligations? In wartime, in all communities, conduct likely to lower the public will to victory is normally forbidden and punished. Here again questions of definition are important. Perhaps the only punishable offence is to state publicly that the enemy has right on his side, to insist that one's own conn try cannot possibly win or to crge people to revolt against the government. In some countries, however, where such regulations against 'sedition' or 'hindering the war effort' have been in in force, people have been severely punished for grumbling · about a shortage of potatoes. What is sabotage? In England the term is normally confmed to the deliberate hindering of some state or occasionally private activity by damaging machinery, holding up supplies or giving dangerously false information. In a number of countries accidental hindering of state work is punished as sabotage; and indeed the term-implying heavy penaltiesmay even apply to the supporting of opinions or policies which subsequently did not succeed or which went out of favour with the government. What, again, is a traitor? To most people, if they stop to think, a traitor is someone who intentionally injures his country or his cause, and this very 24

So1ne Causes of Misuuderstanding strong word carries the in1plication both of important harnlful action and of malice. It is, however, not unusual for someone to receive abusive letters calling him a traitor because he is a conscientious objector to military service-a man in a very special and difficult moral position, which usually means he has high principles, however much others may disagree vvith him-because he has publicly criticized something wrong in the life of the country or because he has foreign friends. Many people think they agree about religion until they begin to discuss it. What are the real n1eanings of such words as sin, salvation, God, the Holy Spirit, the pure in heart, the Church, the Will of God, or even Christianity? Most of us have read of violent argument as to the implications of 'Christian marriage' and 'Church unity', for example. A very naive person and an archbishop may both believe that 'When I die, I may go to Heaven' but their interpretations of Heaven may be unrecognizable as the same concept. Some people say that the only real sin is hurting other people on purpose or by negligence; others say that sitz is setting our own wishes before the will of God; and a clergyman once told me: 'The trouble with n1ost of my flock is that they think sin is something-pleasant-you-can-do-with-your-body, and that if they have no fun they are sure to be righteous!' People have, in the past, tortured and burned one another over questions of the definition of the terms of religion. Nowadays religious persecution occurs in only a few countries, and seldom extends to murder; but people who are concerned with religion are still apt to misunderstand one another because they have not thought what their words really mean. Such misunderstandings may be very important for mental health, for personal happiness, for good conduct, and, a good many would say, for our hope of salvation. Surely the woman who thought 'Christian duty' could dictate that she burn a child with hot iron was trapped in an appalling 25

So1ne Causes of Misunderstanding misunderstanding that did matter, that was much more than just a small point about words? Surely the mistake of those who think 'purity' is attainable only by those who, having defective glands, never have any thought of sex, is a mistake that is no dictionary quibble, but can matter very n1uch indeed? The most \Vretched, tortured neurotic I ever met had learned in childhood that 'God' meant a kind of gigantic headmaster, vindictive and alert, who was perpetually watching her to see if she 'sinned' and to punish her, far more unreasonably than any human authority, for the most trivial mistake in life. What people mean when they use the word 'God' may colour their whole mode of life: may make a person radiantly patient, unselfish, merciful and helpful, or may create an obsessed inquisitor and persecuting monster. So defi!litions do matter very much indeed. In the field of general morality, too, very serious misunderstandings may arise from differences of definition. Anna, we are told, is a bad girl. On investigation, this may prove to mean that she wears rather a lot of costume jewellery, sometimes stays out at night after ten and likes the company of boys. A religious worker once reported to me that she had heard a woman say of her own child, in public, 'She is a very naughty little girl; she is afraid of pussy-cats!' What is honesty? Not to steal and not to cheat the income-tax? Most of us would agree that these are a part of honesty; but some would add that it involves not pretending to be something we are not; and some, that it includes doing our paid work as well as we can. Large numbers of people seem genuinely to believe that it is not dishonest to cheat the State or to steal or damage Government property, which comes, for them, in a special category. Yet if we reflect that Government property and income come out of our taxes, the person who robs the Government is indirectly robbing his neighbours, not to mention himsel£ 26

Sotne Causes of Misunderstanding A man had for years accepted numerous favours from others without any attempt at making any return when he could; he had shamefully exploited his houseKeeper; and he had eventually cheated a relative of his inheritance. Near to death, he said, 'I am not afraid; I have never done anyone any harm.' He was probably quite sincere; harm n1eant to him giving someone a black eye or throwing him downstairs! What is patriotisnz? To some it n1eans sacrificing various aspects of pers()nal convenience to the country. To others it means mostly taking a pride in the culture and valuable achievements of the country, perhaps being proud of such things (in England) as that the country generally takes its treaty obligations seriously, or that it accepts refugees and usually treats then1 well. To others patriotism means defending their country when its policies are morally questionable, and noisily expressing hatred for all other nations. An an1using aspect of this is that, as strongly natio11alistic governments are those most likely to cause political upheavals, very 'patriotic' people of the latter variety usually show most disapproval for those political leaders who, in their own countries, are themselves the most 'patriotic'. 'I think we have a few things to be proud of,' I was saying to the foreign visitors. 'We really do in England have more freedom than in most of the world. I would like to show you one of our good new schools-! think if I telephone the headmaster he will make arrangements for us. You may smile at our kindness to animals, but I think it is rather a nice tradition and helps to train us in gentleness. You 'von' t often see rJgged people in England today. Of course, I know our cooking is a joke; it is better, generally, in homes than in restaurants, but I must admit we cannot generally cook as well as you do. Please don't look afraid of that policeman; our policemen are always polite and helpful to strangers.' 27

Sotne Causes of Misunderstanding Someone took me aside and hissed furiously in my ear: 'I think you ought to be ashamed of yourself, running your country down in front of those danmed foreigners!' It took me some time to realize it must be the bit about the cooking that he meant. I was once rather sorry to hear that a girl I liked was promiscuous. This properly means much the same as indiscriminate, and is nowadays applied mostly to people's sexual behaviour. In this sphere real promiscuity (having any number of lovers with no sort of genuine choice or affection) is relatively very rare. I thought the girl could hardly be as promiscuous as that, but wondered if she were in fact somewhat lacking in restraint and discrimination. Eventually I found out that my informant meant merely that the girl was in the habit, when opportunity offered, of kissing her fiance and had sometimes been seen doing this! So far from being promiscuous, she was, by all usual standards, the oppositefaithful! But some people use the word promiscuous to refer to any sexual behaviour of which they personally do not approve. My informant could probably have been sued for slander. Misunderstandings may arise, not from actual differences of definition, but more from what might be called 3· DIFFERENCES OF ASSOCIATION

An aunt and I were walking in a rather poor district in an industrial town. My aunt noticed that a boy of perhaps seven or eight had a wound over one eye. She slightly knew him, and asked sympathetically, 'How did you get your eye hurt?' The child replied, as though it was the most usual thing in the world, 'Daddy done it.' To seven-year-old Paul, his daddy is a wonderful person, who gives him treats, tells him stories, takes him for rides on his back, and is fair to him. Sometimes daddy is cross; but he 28

Some Causes of Misunderstanding does listen to what Paul has to say for himself, and Paul adores him. At school these two little boys are told that 'God is like a father to all of us.' The inferences from their experience of fathers will lead them to very different ideas about God! Associations attached to words may differ for several reasons, some of which-not an exhaustive list-being: (a) Social Position. Among poor people 'He has a car' will be said with some tone of admiration, envy, or possibly disapproval of such excessive wealth. Among people whose general income level is such that a car is taken more or less for granted, the statement has no emotional colouring; it is just a statement of fact, probably a confirmation of what was supposed. I can imagine that there may be people, though I have never known any, among whom 'He has a car' implies that he is a poor fellow who cannot afford a helicopter or a yacht. A word such as gentleman or lady may mean very different things for people in different social groups. For example, I heard a rather poor girl, who had had a hard life, say that in the sanatorium where she was being treated she could not lead a lady's life. This to her meant that she was expected to do some work. To another person a lady is someone who, among other things, has a spirit of service and helpfulness and often works very hard. Among some groups to be talked about implies something regarded as desirable; to others it is something to be avoided at all costs. Some people who work with their hands genuinely do not believe that lawyers, teachers, or Civil Servants work; their concept of work does not include mental effort; it is even commoner to suppose that writers and artists do not work, though in fact their work requires great concentration as well as skill, much effort as well as knowledge. Conversely, many 29

Sotne Causes of Misunderstanding professional people do not regard farming, mining, deep-sea fishing or making pottery as demanding intelligence, and may presume to despise those who do such work, forgetting that these tasks require skill, knowledge, concentration and intelligence as well as physical effort. Ideas of what constitutes a hard or an easy life will vary considerably according to the social group. Occasionally the interpretation will not be what one might pardonably expect; for example, in some rich homes where the sons are sent to public schools the parents accept without distress that the boys may be severely corporally punished, whereas many a mother has come out of a factory to make fierce complaints against the teacher who has been 'knocking my lad about'. In this respect at least to be born with a silver spoon in the mouth would seem to be a disadvantage. To some people a bathroom is more of a necessity than a television set; to quite a number of people today the reverse seems to be true. Rich and poor are other words that will very obviously depend for their meaning on the social group to which speaker or hearer belongs. (b) Generation. Some words do actually change with the passage of time; more will be said about this in Chapter VI. The associations of a word, however, also change with the passage of time, and a person of Ot\e generation may have quite different clusters of association round a word from those of another. For example, the word actress has greatly gained in respectability in several generations. The word Empire, on the other hand, once a word always to be spoken with pride and enthusiasm, is now an uncomfortable word for many, savouring too much of jingoism and power-hunger. The word divorce was once a good deal more startling than it is now; and with the shedding of taboos a good many words (most of them perfectly neutral anatomical and medical terms) are spoken naturally by the younger generation, 30

Some Causes of Misunderstanding whereas some people still surviving can pronounce them only in a whisper or with visible difficulty. School will carry very different associations for one who was beaten, bored and humiliated there and for one who was well treated, had interesting lessons and was able to respect his or her teachers. Housewife and housework arouse very different feelings in the minds of two women, one of whom has had to drudge all her life in an ill-planned home and for an unappreciative family, and one of whom has a modern convenient home, a good knowledge of how to do her work happily and a family that gives her love and thanks. Another interesting difference between the generations results from progress in medical science. For instance, only one or two generations ago the word pneumonia carried with it a frightened hush, because pneumonia was very commonly fatal. With the general use of new drugs, it has ceased to be a death warrant and is spoken of in a more casual tone. As the world grows smaller such words as fore(~ner and coloured man lose associations of hostility and terror that they once had. Where a sixty-year-old may still finish her cautionary tale with the words 'and she married a black man!', the thirty-year-old is quite likely to reply-though preferably in more polite diction-' So what?' Disputes between the generations sometimes include semantic difficulties which would be amusing if only they were not so uncomfortable for all concerned. We have all heard this kind of thing: 'You know I am not an unreasonable mother; all I ask is that you behave sensibly. But when it comes to staying out till a quite unreasonable hour . . . . ' 'But, Mother, half-past ten is not an unreasonable hour! The pictures don't finish till quarter-past. It isn't reasonable to expect me to leave until they are over!' 'And this doing yourself up with lipstick and all manner 31

Sonte Causes of Misunderstanding of paint, looking like a trollop! I've tried to bring you up respectable, and look at you!' 'Oh, do be reasonable! All smart girls wear lipstick nowadays!' 'Not a bit of a kid like you!' 'I'm not a bit of a kid at sixteen, and I'm not going to go about like a dowdy old frump!' And so it goes on, with no one trying to think what reasonable tneans, both sides exaggerating, as useless and exhausting as a conversation between diplomatists both of whom are instructed to make no compromises .... The associations often differ for different generations because the invention or other object or experience has become more widely available. To my own mother in youth a banatza was a 1Jxury fruit, rather what lichees are to me; nowadays a banana normally carries practically no luxury or rarity associations. During World War Two, incidentally, the word egg acquired exciting associations that it had not had for a very long time. Shell eggs were rare; hence the shift of association. In my O\Vll childhood television was something one read about in science fiction. Very possibly to a child born two or three generations after me artificial satellite will carry no associations of wonder and hardly arouse interest. (c) Sex. A 1nan and a won1an will sometin1es hear a word with different reactions because their associations are different. The most obvious example of this is the words man and woman, together with boy and girl, husband and tvife, father and n1other. For the opposite sex is always more or less mysterious: the experience of the other sex, not only in family life but in the whole life-process from beginning to end, is something into which we can, it seems, never enter fully; and words denoting the opposite sex will also tend to be loaded with associations of either resentment or interest, 32 1

So1ne Causes of Misunderstanding depending on various factors of society, upbringing and experience. Hotne generally seems to carry more rich and interesting associations for a woman; war and revolution generally seem to carry a few romantic and colourful associations for a man whereas for a woman they almost always imply mere dread and tragedy; this seems odd, in that men still do most of the actual fighting in most wars; perhaps the natural biological interests of the two sexes account for the difference. The sex association-patterns of words also changes as the exact limits of the functions of men and women change; for instance, now that there have been a few women professors, the word professor will have, for both sexes, slightly different sets of associations; in societies where a married woman is a household drudge and has very few civil rights, it may be expected that the word ntistress will carry much more glamour, and even a kind of dignity, than in a society in which n1arricd women are allowed a good deal of freedon1 and initiative. A woman hearing the word hat, with nothing to qualify it, will probably visualize a woman's hat and a man will visualize a man's hat; similarly with other garments. A man hearing the word knife is much more likely to think of a pocket-knife than a woman, who will probably think first of a table or kitchen knife. The word powder by itself once meant gunpowder to the colonel and hair or face powder to his lady. And so on. (d) Occupation. The likelihood of actual misunderstanding arising out of the fact that the same word carries a different set of associations for a person in another occupation is very limited; but the differences of association are real enough. For example: Watch has different first meanings for a watchmaker and a sailor. 33

Some Causes of Misunderstanding Chestnut-for a greengrocer, a comedian and a horsebreeder. Slug-for a gunsmith and a gardener. Cat-for a veterinary surgeon and a prison governor. Case-for a porter and a printer. Scissors-for a wrestler and a seamstress. Plug-for a plumber, a tobacconist, and an electrician. Fortn-for a school-teacher and a Civil Servant. Seat-for a carpenter and a Member of Parliament. Sole-for the shoemaker and the fishmonger. And when a girl on a rather strenuous country walk exclaimed, 'I can't walk fast; I am losing my sole!' her companions wondered if a clergyman might not be needed too. Ham-for an actor and the owner of a delicatessen shop. Again, a word like harvest or warble fly is far richer in associations to a farmer than to a typist; and it is a standing joke that what to a landsman is a gale is to an experienced sailor a gentle breeze. (e) General Personal Experience. It is not at all uncommon to find that people have a few words or names they do not like, not because they dislike what the words or names are supposed to 'mean', but because for them personally some disagreeable associations are attached. Many jokes are based on this fact; for instance, 'I wouldn't ask for a cheque today; the boss has just had a row with our representative in Prague, and he is feeling sore on the subject.' Not only the experience of an individual, but those of the group in which he or she lives, or has been brought up, will give associative colouring to words. For instance, a child can fairly easily be brought up to regard capitalist as an insulting or admiring word, to think that clever people are people who cheat and avoid work, to associate extreme dislike with a word such as Jew, nigger, German, Russian, or British; on the whole education can probably give taboos rather sooner than 34

Some Causes of Misunderstanding it can remove them and inculcate prejudices rather more easily than tolerance or curiosity. 4·

MISUNDERSTANDING OF THE CONTEXT

Very often, when someone asks the meaning of a word, the person asked has to put it in a context, in order to show the meaning, or must even ask the enquirer to provide the context so as to know which of several meanings is required. 'I must have got my threads crossed.' Is the person handling a screw of some kind, sewing or weaving? Or even in a muddle in a lecture? The prevailing human tendency to call things by borrowed names, especially in everyday speech, may complicate matters further; soup can mean nitro-glycerine, crate can mean an aeroplane, mouse a variety of surgical swab, and to give somebody socks or beans may not mean to clothe or feed them. 'I felt wretched, so I went to bed with a bottle.' Of hot water, or of whisky? Is an internal examination by a doctor or in a college? If a professional psychologist says someone is an idiot in a report, the word is very clearly defined. If he calls someone an idiot in the course of a violent quarrel, idiot may merely be an abbreviation for someone-who-disagrees-with-me. A lawyer who says 'I'm worried about my new case' may be speaking of his profession or of his luggage; until we have a little more context we cannot know. The possibilities of misunderstanding arising out of various differences in the use of language are, then, enormous. Important differences are more likely to occur in some fields than in others.

35

IV. VAGUE WORDS AND EXACT WORDS

I

general, the more language deals with concrete objects the more exact it is; and the most exact language is that of those sciences that deal with the concrete. There is no doubt whatever about the semantics of: one molecule of copper N

sulphate; the positive pole of a dry battery; the human vermiform appendix; a 1-dioptre camera lens. Such expressions have only

one mearung. Moreover, scientists in principle aim at being objective; this discourages the formation of confusing clusters of associations round their words. The only scientific words that inevitably become heavily laden with emotional associations are words that stand for things or phenomena which may affect human life and happiness; there is more emotion attached to, say, uranium or penicillin than to heliun1 or ferrous oxide. The scientist is, however, always trying, consciously, to make his language as unambiguous, as objective, as precise as possible. He quite often uses mathematical symbols and formulae, which are even more foolproof than scientific words -provided that one can understand them. I have no real scientific training; but even the little 'science' one studies at school helps to give some awareness of language as an exact tool; the making and studying of defmitions, the following of arguments in mathematics, physics and other sciences, has its uses for showing that some at least of our thinking can be exact.

Vague Words and Exact Words Ordinary concrete phenomena give rise to language which is not quite as unambiguous as that of science, but which seldom gives rise to serious semantic confusion. A pie may be a meat pie or an apple pie, a veal and. ham pie or an egg and bacon pie; it may also be a mess of type upset in a printing works; the ordinary man here is at a disadvantage as compared with the mathematician, who has only one pi to deal with and multiplies it by the square of the radius to give the area of a circle.... But there is a narrow limit to the number of meanings possible for the word pie. There are many breeds of dogs, but a point will always come at which the animal being examined is not a dog, and everyone has more or less the same idea of the characteristics of a dog. A chair may be made of upholstered wood and springs, of steel tubes and canvas, of solid wood, of wood and canvas, for that matter of gold, in which case it is more likely to be called a throne, but need not be; but, again, we all have some limits in our mind beyond which an object ceases to be a chair. Similarly we have limitations in our minds defining roughly at least such things as pencil, carpet, rose, toe, frog, button, larder, basket, attic, oar, trough, postcard and so on. Those limits will not always be precise enough for human requirements; for instance, in assessing objects for Customs duty or purchase tax such questions as 'is this a postcard or a greeting card?' 'Are these buttons or jewellery?' may be quite important and require much discussion. However, in general it is not the names of concrete objects or phenomena that give rise to the most troublesome semantic problems. Moreover, we know when we do not know the meaning of a word of this kind. If we see a less common name for a concrete object, such as lemur, conch, orion, pinking shears, ocarina, alb, malachite, we know immediately whether we understand the term; and if we do not understand it there is usually very little difficulty in finding our what it means. 37

Vague Words and Exact Words We are hardly in danger of basing a wrong action, a catastrophic decision, our own anxiety or misery, on a misconception about the meaning of some word of this kind. And such words generally mean much the same to all the persons who hear or see them. Not quite the same, perhaps; for the surrounding associations may be rich, sparse or even nonexistent; but no normal person hearing the word pie is in any danger of thinking it means what other people mean when they say boiled egg. I have already given son1e evidence that even these words do not mean quite the same to everyone, as in the statement 'He has a car' or the reactions of two generations to the word banana. The semantic advantage of concrete terms is that the limits are much narrower. When we step into the field of abstract language we find ourselves in territory with frighteningly vague frontiers; and, if I may continue my metaphor, the territory is heavily mined; there is real danger in the misunderstanding of many abstract words. The language of politics is often very ambiguous. Some time ago a cartoon was published sho,ving the representatives of two governments shouting at each other, 'Aggressor!' The aggressor, in diplomacy, is supposed to be the government that starts a war or attacks another country. Today almost all governments pay at least lip-service to the idea that aggression is wrong and that an aggressor nation should be restrained and punished by others, if possible. This is a step forward from the open power-worship and threatenings of Hitler, but trouble arises as soon as any nation is accused of aggression. Why? Because it is not at all easy, normally, to decide who is really responsible for starting a war. Sometimes it is obvious. If Utopian troops march straight into Ruritania, kill unarmed Ruritanians, loot, burn and wreck, when yesterday relations between the two countries seen1ed normally quiet, that is clearly aggression; but if large Ruritanian forces had been 38

Vague Words and Exact Words n1assing on the frontier for the past fortnight, if unconventionally large numbers of Ruritanian secret agents have been found in Utopia and if the Ruritanian press and radio have for months been pouring out abuse of the Utopian government and urging Utopians to rise against it, the question of who really started the trouble may not be so simple. Today much propaganda work is devoted to proving, where there is tension or actual armed conflict, that the other party was the aggressor-or is expected to be at any moment. 'Fair shares for all' is an ideal with \vhich I find myself very much in sympathy; but what exactly does it mean? An assembly could easily agree that they all wanted fair shares for all; then, when the problems of putting this into practice were being discussed, would find that the members could not agree on the details. For example, is it Ja'ir that all human beings who work should be paid alike? or that they should be paid more if they are showing special skill or taking more responsibility or working in a dangerous or notably unattractive job? Should a married man be paid more than a single man, a man with children than a childless man? Should those who are unlucky enough to be infirm or too mentally deficient to work receive the same share of the community income, or less? Should there be rewards for special efforts? These are by no means easy questions. Even when the principle of fair shares is applied to the relatively simple problem of fpod rationing in wartime, not all the answers are easy. If everyone has exactly the same ration allowance in a time of shortage, children and adolescents will not have enough and their health may be permanently damaged; and if heavy manual workers such as coal miners have to live on the same rations as desk workers, they are not likely to receive enough calories to provide ·the energy they need. A responsible government has to work out the fairest solutions to such problems that it can devise. 39

Vague Words and Exact Words The word freedom is another word that causes semantic difficulties. I passionately value freedom; so do most English people; and so, I think, do at least all human beings who have tasted it. But freedom is not ever, and cannot be, absolute. I have no wish to be free to hit my annoying neighbour on the nose; and still less do I wish my neighbour to be free to hit me on the nose. Censorship in some form is probably excessive in nearly every country in the world, by the criteria I accept; but though I believe in 'freedom of speech', I would not like to see freedom even of speech made total, permitting people, without penalty, to say false and ugly things about one another. The love ofJreedont as an emotive phrase is most successful; but as a guide to political action and organization it needs a great deal of careful defmition. All peoples and groups who feel oppressed wish for liberation. I remember having as a guest in my home a woman who had spent her most impressionable years in one of the most unfortunate countries in Europe, a country ravaged again and again by war, dictatorship and civil war. One evening I came home from work and set about preparing some supper. 'There's a cake in the tin . . . ' I muttered, opened the tin and found no cake. I looked suspiciously at my guest. With a charming smile she explained, 'I've liberate,l it!' A grim joke, though she did not treat it grimly. As slogans play a great part in the political thought of many people, it is desirable that we should learn to look at them hard and ask very insistently what they mean. Many everyday abstract words are equally vague in definition: good, bad, sensible, strict, cheeky, greedy. Here are two reports on the same schoolgirl, from the same report form: 'Joan has a mind of her own and an enquiring attitude. She shows some originality and thinks for hersel£ She may develop valuable qualities of leadership.' 'Joan has been very cheeky this term and questions the

40

Vague Words and Exact Words decisions of those in authority; she is argumentative and apt to be perverse. She is beginning to have a bad influence on her classmates.' Now some of the difference between these two reports will reflect a real difference in the temperaments and beliefs of two teachers and quite possibly in the objective facts about Joan's behaviour with them; but some of the difference is that their definitions of the key words are different in detail. The statement that Harold is a good husband, that Jill is a good wife, that Ethel is a good mother or James a good father, does not have the san1e meaning for different people. Even the statement that Helen eats very little is not as precise as it sounds; with whose appetite is hers being compared? That Philip has a hot tenzper will probably be taken to mean that he frequently displays violent anger; but the statement does not indicate how often he loses his temper or how quickly he recovers it. One may have a very hot ten1pcr-I have myself-and keep it very much under control after years of struggle to do so-l now really lose n1ine about once a year, but it is still there. Marian is a flirt: this n1ay, for different people, mean anything ranging from a total indifference to sex conventions and the feelings of others, to a mere liking for light badinage with men and rather a bright, wandering eye. A housewife boasted that she was a good cook. Investigation discovered that she could cook a very limited range of recipes and that, though adn1ittedly she hardly ever spoiled a meal-which for her made her a good cook-none of the meals she cooked were original or exciting and nearly all came out of the frying-pan or a tin. We cannot get along without these vague abstract words. Moreover, we need not be endlessly examining them all day; life does not afford the time for so n1uch reflection. When Molly tells her mother, 'I have a very good husband in John ... ', mother knows all that she really needs or wants to know, 41

Vague Words and Exact Words which is that her daughter is content with John and will be comparatively happy. The details of Molly's expectations and definitions do not much matter. If I say 'I have enough money for my needs' the only thing that matters is that I find it enough for me; my own subjective definition will do and my friends or relatives need not be concerned about me. If a doctor says his patient is getting better now, the family may want to know whether this still means another month in bed or two days n1ore, but the statement is not in itself without very important meaning. Too much examining of the meaning of words may be downright rude on occasion; some generalizations, some vagueness, are the small change of conversation. Sometimes, however, the meaning of these words is very important and open to much discussion. In arguments where much emotion is involved, for example, the meaning of a word like love or reasonable or selfish or fair is very likely to be used in different senses by different people. More will be said about this in Chapter XII. Very often the general, abstract words of everyday speech need to be helped out by concrete words that give examples. Concrete advice is generally very plain and not open to misinterpretation: 'You should take two 50 n1g. tablets of vitamin C daily and walk in the open air for at least half an hour.' 'To peel tomatoes easily, put them in a basin first and pour some boiling water over them; then the peel will cotne off loosely.' 'Put an ink-pot at each candidate's place, with one piece of clean blotting-paper and a sharpened pencil.' This is the kind of language that any intelligent person may be expected to understand. Yet even the most concrete language is sometimes misinterpreted where associations differ; I remember a tale of an elderly countryn1an with a sore throat. 'Ah,' said the doctor, 'I will give you a throat wash 42

Vague vVords and Exact Words for that.' So the old man was supplied with a large bottle of a mixture labelled: 'Throat wash; to be used when required' -and, instead of gargling with it, ordered his daughter to warm some of the mixture, put it in a basin and careiully wash the outside of his throat with it. If we are giving advice on general behaviour we shall often fmd it necessary to illustrate the abstract by means of the concrete. 'Yoll are making a bad impression; you really must try to be more polite to people.' This is not in itself much use; a person whose manners are bad often just does not think about manners and has no criteria. 'Well, what 011ght I to do? I didn't think I was being rude.' 'You could hold doors open for people when they are behind you, instead of letting the door bang in their faces; and you could pass the salt and the bread at lunch, instead of leaving all the passing to other people. Oh, and when two people are talking you should not butt in suddenly without saying something like, Sorry to interrupt, but.... And when you bump into someone it is more polite to say, Excuse me, than to glare at them ... .' This is the kind of advice that is of practical use, however unpalatable it may still be. Vague language is sometimes the result of limitations of vocabulary. A woman drove into a garage and said: 'Can you help me? There's a sort of a noise thing that sotmds wrong to me. I've looked, and I think the little whatsit that is usually on the top of the round thing has come off, and I am afraid it must have got into the thingmajig.' This docs not provide a very good basis for investigation. Language is in fact rather inadequate for the expression of shades of emotion; we need at least as many words for the 43

Vague Words and Exact Words different kinds of love and friendship as there are for the different kinds of tree in England; and we do not have them. It is often very difficult to discuss emotions, however much we are trying to be honest. This is apt to lead to misunderstandings and unhappiness. I should think at least 90 per cent of English people-and probably people all over the world-would be n1uch healthier, happier and pleasanter if they were told more often how much they were loved and how lovable they were. But, quite apart from any matter of shyness, of reticence, of psychological difficulties in expressing love and friendship, we very often do not have the language. 'You really are rather a pet,' says the young husband. 'You're not a bad sort yourself,' his wife replies. Hate too is a word that has many shades of meaning; and sometimes the shades are important. There is, I suppose, no word treating of an emotion that is not capable of being ambiguous. There are occasions when it is beneficial to keep this in mind. Vague language is also sometimes used to confuse the reader or hearer, intentionally. 'Why must I not do that?' says one. 'Because it is not right,' says the other. 'But I see no harm in it; why is it not right?' Now, at this point a person who is trying to use words honestly at the moment has three possible replies: I. 'Well, I don't know; perhaps it is right after all, now you ask me; why do you think it is right ?'-that is, readiness to discuss the matter further. 2. 'It is not right because . . .' that, is, readiness to give a valid reason, e.g. 'because it would hurt Mary's and Martha's feelings,' 'because it will be breaking a promise ... ' and so on. 3. 'I am afraid you will have to take my word for it at the moment that it is not right; but just now I can't explain' -an unsatisfactory state of affairs, but one that is sometimes genuine. Whether or not it satisfies the recipient will depend very much on previous experience; if the speaker is usually

44

Vague Words and Exact Words fair, honest and not too authoritative the recipient may be prepared to take the declaration on trust for a time. When we speak not so much from clear notions of what we think is right and wrong, but from our prejudices, something like this is apt to be the reply: 4· 'How dare you set up your judgment above mine? How can a bit of a girl like you possibly know right from wrong? It is very wrong to question your elders' judgment. And if you want to do things like that, it just shows that you do not know right from wrong.' Blah, blah, blah. As soon as we examine words of this kind from the point of view of elementary semantics, we see that they are hardly real cotnmunication at all. The purpose of this kind of language (not always the conscious purpose; more will be said about self-deception in Chapter XIII) is not to communicate; it is to prevent communication.

Politicians are generally very good at using vague language

to confuse. One of the most valuable semantic exercises for those who wish to acquire some grasp of the morality of politics is to translate some vague abstract language used by a politician into concrete, detailed information. This is the kind of thing I mean. The Leader says, and how nice it sounds: 'Our gallant troops have the situation completely in hand. All of them have done their sacred duty with courage and enthusiasm. The seditious and traitorous elements who were disturbing the peace in our beautiful capital have been wiped out or rendered harmless. Our beloved citizens may once again walk in the streets, shop and go about their business in peace. And the peoples of other countries can see that only a handful of sedition-mongers are discontented; the real Utopian people has complete confidence in its government.' As soon as we begin to translate this into concrete detail it sounds rather different: 45

Vague Words and Exact Words 'There is blood in the gutters and there are bullet holes in the walls of the houses round the main square. The bodies of John, Peter, Ralph, Sam, Bill, Alex and Don are buried in a common grave. Agnes has been raving in hospital ever since the stray shot killed her baby Jimmy. Three wounded demonstrators are missing. We hope they are in hospital too, but screams were heard from somewhere underneath the town jail last night. Harry, Frank, Louis, Dorothy and Kathleen were arrested and their families cannot find out what has become of them . . .' and so on. A concrete translation, as it might be called, often makes a big difference to our whole thought about a subject.

V. LET US LOOK AT SOME WORDS

T

first four chapters of this book have already shown that saying what we mean is not as easy as it sounds and that the meaning of most words is open to doubt, depending on such things as knowledge, context, association and background. On the whole, the tnore abstract a word is the more likely are semantic difficulties, unless the abstract word is also a highly technical one like son1e tern1s of philoHE

sophy and the sciences.

The word rich has various meanings as applied to cakes, soil, foliage, colour or silks; but in its most usual sense of having plettty of money its meaning may vary considerably according to the position of the speaker and hearer. Is a man rich if he has a net income of £r,ooo a year? £s,ooo? £Io,ooo? To some people a net incon1e of £soo a year is riches beyond their wildest hopes; to others it is actual poverty. A pathetic little tale is told of an African negro, who, when a missionary was trying to give him some idea of the enorn1ous wealth and power of Queen Victoria, said, 'Could she, then, eat tinned meat every day?' There is another tale of a millionaire who was invited to dine at the house of an aristocratic family. The family was very proud of its massive silver napkin rings, which were beautiful pieces of antique craftsmanship. When the millionaire saw these he did not know what they were for, and had to be told. He was shocked and distressed by the poverty and squalor of a mode of life in which

47

Let us Look at Sonle Words people saved a soiled napkin to use it again at another n1eal! Very few people think of themselves as rich; most people think of someone else as rich, generally with some feelings of envy or disapproval. Being rich is like the previously discussed having enough to eat; it is always defined, not by some absolute criterion, but by the standard of the group in which, at the n1oment, the word is used. We will now examine some words that are open to much discussion, and whose semantic nebulousness frequently gives rise to mistmderstanding, even real distress. I. NATURAL

There are in our everyday speech 'clean words' and 'dirty words'-words which we normally use with a tone of approval and praise and words which carry the implication of disapproval. For instance, practical carries approval; Corttinental for many people in England carries a hint of disapproval, although in fact both words are morally neutral as far as real meaning goes. Now natural is almost always a clean word; it implies something good, desirable, or at any rate permissible. People who are affected or self-conscious are told to be natural. Here the speaker is cotnpletely ignoring the question of what is their nature! It is unfortunately natural, in the sense of being psychologically predictable according to observed laws of nature, that many people should be self-conscious or even affected and insincere. We are quite often told that a person who is deficient in the ordinary human passions is not natural; and even that a person who does not sn1oke, drink alcoholic drinks or play outdoor games is not natural. Again, the question arises of what is his nature. For example, suppose a person has no natural passions in the sense that he does not seem capable of experiencing sexual love or 1ust; if he is found to be a person with defective sex glands, his 'unnatural' attitude is 48

Let us Look at Some Words perfectly 'natural'-i.e. in accordance with the laws of nature. We are often told that some forms of medical treatment are bad because they are not natural. It is true that modern drugs and surgery are not found in, as it were, a wild state; but the same criticism would apply to almost all present-day human activities. In the sense in which modern biochemistry, surgery or bacteriology are unnatural, it is also unnatural to wear clothes, wash regularly with soap and use perfumes, eat cooked food, eat food we have not ourselves killed or cultivated, use spectacles, artificial teeth and limbs, warm ourselves with gas and electricity, proceed from place to place on or in anything but our own two legs or an animal, have water sanitation and polite habits about it, read, listen to the radio, watch the television or consult a lawyer instead of fighting over the limits of our estate! A fully natural life would be intolerable to almost every British human being today; and most of us not only would not enjoy it, but would not survive it for more than a few weeks, even days. The word natural crops up over and over again in discussions of the ethics of sexual and married life; and it is nearly always used very carelessly. If we are to condemn as wrong everything that is not natural, we shall fmd ourselves in the very uncomfortable position that to begin with we shall have to scrap marriage itself, since all the evidence seems to be that monogamy is not natural to human beings and requires considerable effort from them; yet most people would agree that marriage is indeed an excellent institution in its general concept, whatever details of reform or adjustment they would wish to see. We might even fmd we had to scrap all ideas of restraining our sex urges at all, and that brute force and brute desire became our only criteria. The laws of nature on the whole seem to suit the animals very well; but I doubt if a creature as complicated and sensitive as the average human 49

Let us Look at Some Words being, and with the numerous wants of the average human being, would enjoy a genuinely natural sexual life. Incidentally, if the laws of nature were left entirely to themselves, the relatively weak men and relatively plain v.romen, who in a monogamous society have a reasonable chance of finding some satisfaction for their natural needs in marriage, would probably have no chance at all. It will be seen, though I am deliberately being rather melodramatic, that as soon as we begin to use the word natural as a criterion in discussing sex and marriage we are going to land in a very alarming quicksand. We are often told that it is natural for men to rule women, for subjects to obey their sovereigns, for children to obey (and to love) their parents, for human beings to believe in God, and so on. When we begin to consider in detail what the word natural means, we begin to see that these statements are more questionable than we at first supposed. There are in the world matriarchal societies; societies in which c4ildren are not particularly attached to their parents or parents to. children, very democratic societies in which there is no single superior authority, atheistically based societies in which religious hunger is by no means a usual phenomenon, and so on. To use natural as a synonym for approved-ofby-me is a: very misleading habit ! 2. JUSTICE

In a well-known and very powerful French short story, a father shoots his own young son because the child, in a situation that was too hard for him, has betrayed someone. When the horrified mother asks what he has done, the father replies, 'Justice'. Most of us would disagree. Justice is another of those words it is wiser not to talk about too much. Shakespeare had noticed the difficulty in the semantics of the word when he wrote The Merchant of Venice.

so

Let us Look at Some Words 'For, as thou urgest justice, be assur' d Thou shalt have justice, more than thou desir' st' {Act IV, scene i) says Portia, after she has convinced herself that Shylock is not interested in 'mercy' and wants only justice'. Defmitions ofjustice in the mind of a speaker may range over the following-and the definitions are important: 'what the laws of the land say is correct in the set of circumstances'-a nd we can all think of cases where we do not think the laws of the land are fair; 'what makes the wrongdoer suffer exactly the same distress as he has caused, or rewards the well-doer with equivalent benefits'-which is generally impossible to achieve, whether or not it may be desirable-'what satisfies all those immediately concerned as just'-.which is what we often have to think of in settling disputes !-'what takes into account all the various circumstances of the dispute and treats every case or person on the merits of the special case'-ideal justice, I suppose, but hardly achievable all the time; and, of course, on the lips of many, 'what I have always been allowed to have before' or even just 'what I want'. A lawyer's concept of justice will be very different from that of a schoolboy, and the idea of an unbalanced, very selfish or very immature person will differ considerably from that of a mature, thoughtful and co-operative person; but whatever the knowledge, whatever the ethical evolution of the person using the word, there will remain some doubt as to the exact significance of the word. 3·

REALITY

'Reality' (and its relations realism and reaQ are 'clean words', rather in the same way as 'practical'. They carry the implication that the person who claims to be in contact with reality is practical, sensible, honest, not dreamy, muddle-headed or 51

Let us Look at Some T-Vords sentimental. One of the results of this complex of associations has been very odd; the concept of reality as always 'somethin~ unpleasant that must be admitted and faced'. The reason for this is probably that most people are more ready to pretend that things which seem to them ugly or distressing, or that do not suit their theories about life, are not there, than to refuse to believe in things that please them, suit them or fit in with their ideas. Yet even this is not entirely true; we can fmd people, not indeed very rare people, who refuse to accept what is pleasing and cheerful, who seem to be determined to look on the most gloomy and distressing aspect of life, who appear to be as eager to be displeased as some are eager to be pleased. There are some ostrich-minded people who do need a dose of 'stern reality' in being made to see that disease, aggressiveness, slums, sadism, hypocrisy and the desire to rise to success at the expense of others are actual factors in human experience. But there are also people who, in order to achieve a balanced view of life, need rather to go and look at rhododendrons, young sweethearts, goldfish, nurses, sunsets, Swan Lake and the life of Albert Schweitzer; these are also objective phenomena and a part of reality. As a matter of fact the nature of reality is a question of which philosophers differ very much; and once we begin to think about the meaning of the word, instead of using it just as a 'clean word' to tell someone else to be 'practical', we fmd that there are more difficulties than we had supposed. For example, which has more claim to be reality: something that exists in my own mind, or something that, so far as my imperfect senses and information tell me, exists in the outside world? Most people who are not philosophers would probably say that the sparrow is more real than my consciousness of the sparrow; but is it? {I do not myself make the slightest claim to know the answer.) Is there any reason to think that things we find wrong or unpleasant are in some way more 52

Let us Look at Some Words real than things we fmd pleasant or desirable? Are emotions real? That to anyone who has felt emotions at first sight seems a silly question; but people not infrequently use real or reality in a way that suggests that emotions are not real. Emotions are very real factors in life: for example, most industrial and political disputes are hard to solve chiefly because emotions are also involved (e.g. nationalism, memories of the past) ; much human conduct that at first sight appears wildly self-harming may be accounted for by emotions; and also many of our best actions in an objective sense are prompted by emotions. Suppose we act under the influence of emotions that are based on a mistake, e.g. Othello kills his innocent wife believing her to have been false to him; a man is lynched on a false charge; a war is started owing to a misunderstanding of someone' s intentions; what is real in these cases and what is unreal? The emotions, for instance, are real as phenomena, but in another sense they are utzreal, in that they are aroused by mistaken beliefs. And so on and so on; it is easier to tie ourselves up in the problem of defining reality than to solve the problem; so it may be as well not to go about talking of reality and realism as though everyone knew what they were. 4·

HONOUR

There are all kinds of side meanings to the word honour, as in Birthday honours, Honours degree, do the honours (offer hospitality); my word of honour and so on. Let us take just the sense of my personal honour. This is a kind of mixture of my good reputation, especially for being (a) honest in my dealings, (b) brave, {c) loyal, (d) truthful, and, in a woman only, chaste, but this we will ignore for the tnoment; and my good conduct in these matters. Now, honour is certainly a 'clean word'; but for different people the criteria of what is honourable differ enormously. For instance, to some people it is never honour53

Let us Look at Some Words able to tell a lie, except perhaps to the enemy in wartime or to save someone else from grave harm; to others it is very dishonourable to tell a lie within one's particular group, but perfectly honourable to lie to people outside the group. In some societies and groups honour has much more to do with possessions, gifts and hospitality than in others; for example, there are people in parts of Yugoslavia whose honour is seriously offended by a refusal of an invitation, and whose honour obliges them never to receive a gift without giving one. To many people my honour is a thing that I alone am responsible for; but to many other people my honour can be injured, not only by my own bad behaviour but by the slanders of others. I have myself heard a very fme man say that as an honourable man he did not propose to make any attempt to punish some persons who had slandered and insulted him; he clearly felt that to associate with such people even by entering into conflict with them would be to stain his own honour, and, further, that the honourable man is magnanimous and does his best to ignore the misdeeds of pettier people. In past centuries-and it is even true of many people today-honour was thought of as violated if a man allowed another man to slander or insult him and did not immediately attack his insulter, either nowadays by legal means or in the past by challenging him to a duel. Those who saw the fme James Dean fum Rebel without a Cause will remember the boys to whom honour (in American idiom, not being chicken-in British idiom, not being sissy or soppy !) involved a readiness to perform acts which were criminal, dangerous and also useless. This concept is not unusual, especially among school-children and adolescents. Moreover, we all know by experience that one person will say 'I give you my word of honour!' as a formula to obtain some concession and with no sense that it means anything at all, whereas to another person my word of honour has almost the 54

Let us Look at Some Words force of an oath in a court of law-a promise that he regards as binding and indeed in a sense sacred. 'My honour is at stake!' has spurred many people to noble and magnanimous actions, to curing their faults and to dignified self-defence; it has also driven many people to stupid angers, to needless recklessnesses and to futile, even fatal gestures. So what the word really means is quite important. Then, too, to some people honour ('clean word') is oldfashioned ('dirty word') and is associated only with outmoded manners and with priggishness or superstition. I leave it to the reader to think out the semantics of that.

5·

MAKE A MAN OF

In so far as this phrase is not totally meaningless, since every

human being who is not an unlucky medical freak is born either male or female and the male babies will, if they live long enough, become men, it has really two separate sets of associations that give it two different meanings. The idea of man carries in this phrase the idea of those attractive moral qualities that (rightly or wrongly I do not regard myself as qualified to say) are held to be most conspicuous or most important in the male sex: namely, courage, endurance, initiative, responsibility, leadership and dignity. Thus when a probation officer, a headmaster or a good Army officer wants to make a man of Claude, he is thinking that he wants to induce Claude to be more courageous, to be more stoical in enduring discomfort or danger, to show more initiative, to take more responsibility, to prove his fitness for leadership and to have dignity in his general behaviour. He will set about this perhaps partly by rebuke and encouragement, exhortation, psychological assistance; he will also presumably try to provide Claude with suitable company, bring him under suitable influences, cajole or force him to take part in sports and other activities requiring these good qualities,

ss

Let us Look at Some Words present him with difficult situations to test him and give him opportunities and so on. Claude will probably not like this; but the eventual result may make him happier as well as better. The details of the process and the question of how far such discipline should go are both very much open to argument; but the general wish to make a man of Claude in this sense seems to be a proper enough wish. Yet it is not at all unusual for people to use making a man of in a sense which they would probably regard as the same thing-making Claude into what it is proper a man should be; and many people would not only seriously question, but absolutely deny, the desirability of the process. It is not uncommonly practised on young men who go into the armed forces in various countries, or into factories. This making a man of someone implies: getting him into the habit of swearing; teaching him to drink alcoholic drinks and, tnoreover, to get drunk; to smoke and perhaps to spit; to prove his 'manhood' in casual, loveless sex relations; to ignore various questions of ethics; to avoid work, to 'get away with it'; to perform various small dishonesties that are supposed to be clever; to be rude to his parents, or the women of his household; and to do dangerous things that serve no purpose. A well-known example of this odd concept of what makes a man is that police and doctors are generally agreed that all motor-cyclists should wear crash-helmets as a precaution in case of accident, the human head being a part of our body that we cannot manage without and that is particularly vulnerable in motor-cycling accidents. Over and over again I have heard of young men not liking to be seen in a crashhelmet because their friends would think them 'soft'! Alas, it is the human skull that is soft in comparison with asphalt or telegraph poles, and because of this vanity many young and enjoyable lives have been cut short. Making a man of is rather like sporting or a good sport; deS6

Let us Look at Some Words pending on who uses it, it may range from something very valuable and sensible to something which many people would regard as downright evil and nobody, after reflection, could call desirable. 6.

IMMORAL

Some of the words discussed above are weakened by the multiplicity of possible meanings; the word immoral has been weakened and even made a little ridiculous by extreme specialization. It originally means the opposite of moral: that is, not in accordance with the principles of morality. Now, what exactly are the principles of morality is not defmable except within a particular church, country or social group; but in England most of us would agree, at least in general and with some reservations on points of detail, that it is right to work hard at our job, to help people, to speak the truth and keep our word, to be faithful to our mate, to be willing to compromise in disputes, to consider the rights of others; and that it is wrong to kill or hurt another person on purpose, to steal, to break promises avoidably, to do obvious injustices, to go about being rude and intolerant, to take more than what appears to be our fair share of things or to be lazy in our daily work. Morality covers all aspects of behaviour and, unless we belong to some absolutely authoritarian church or political party that lays down defmite and rigid rules for our behaviour in all possible situations, we often have to think what precisely is indeed 'in accordance with the principles of morality'. For what are probably extremely complicated psychological and sociological reasons, connected with a psychological tendency which I should call 'puritanism' were it not that puritanism is another word whose range of possible meanings is so great as to make it rather unhelpful, in England immoral has come to mean, for a very large number of people, only contrary to the principles of morality in the sexual sphere. The

S7

Let us Look at Some Words situation is made more awkward by the fact that there is probably no field of human life in which there is more controversy among well-intentioned and well-informed people as to what exactly the principles of morality are. Even what we take for granted as usual in this field, let alone what we think is right, may vary enormously with social group, standard of education, church, psychological and medical knowledge and-perhaps n1ost of all-personal observation and experience. The result tends to be that for probably a majority of English people immoral means no more really than behaving sexually in a manner of which I personally do not approve. Indeed, the frequent use of the word is nowadays as likely to be evidence of a censorious and intolerant nature, or great ignorance of problems, as of a serious and thoughtful concern with morality-probably more likely. If I said 'I am afraid he is a most immoral schoolmaster; you can tell from his facial expression that he takes pleasure in beating his pupils', most people would be startled at the use of the word; and if I went so far as to say of·a 'respectable' spinster: 'She is one of the most immoral people I have ever known; she is always assuming the worst of others and talking about their sins, and she makes a regular habit of discouraging young people who are trying to achieve something', many people would raise their eyebrows. One of the odd results of this limitation of the meaning of the word is that we may sometimes hear a person referred to as immoral because her cosmetics are too lavishly applied, her skirt is shorter than usual or her bathing costume scantier, or she is hung about rather too glitteringly with jingly costume jewellery. Another and even odder result is that, although all mature human beings at least wish to do right when they think about the matter, the word immoral-being associated with sexual attractiveness, which normal people wish to have, though if they are responsible they mind how they use it58

Let us Look at Some Words has acquired a touch of glamour and desirability, much as certain excellent perfumes have names which by dictionary meaning are repugnant and which by association are attractive: My Sin, Scandal, Primitif, Tabu, Indiscret and so on! 7• OLD-FASHIONED AND MODERN

Both these words are semantically very interesting. They are among those not very numerous words which can according to context be either strongly 'clean words' or strongly 'dirty words'. In fact we very seldom use them with any careful thought about the meaning. Old-fashioned might be expected to be a neutral word meaning 'belonging to past times and no longer customary'; and similarly 'modem' might be expected to be a neutral word meaning 'as is usual in the present time'-the scope in time of the words varying according to the context, for example Modern English is regarded as beginning in about 1450, whereas a modern evening dress is not likely to be more than a few years old. Yet here a psychological fact intervenes. In all of us there seem to be two opposing tendencies, though in most of us one generally predominates: a desire to be 'up-to-date', to move forward, to progress, to leave the past behind; and also a desire to look back or to go back to some Golden Age in the past. Young people are usually more 'progressive' and elderly people more 'retrogressive', but this is to over-simplify. Perhaps the two tendencies reflect the desire to grow up and be more independent, and the wish to revert to childhood or even pre-natal life in which we were more secure; in which case the predominance of one or the other wish in us will be conditioned partly by our earlier experien~es and by whether our own lives have grown happier or less happy. Old-fashioned mint humbugs and a quaint little old-fashioned town are Golden Age associations: we are meant to think 59

Let us Look at Some Words that the sweets are made from a proved traditional recipe and from good, genuine materials; the town is thought of as pretty, quiet and idyllic in atmosphere. Our new· modern railways in a travel brochure carries progressive associations: streamlining, speed, cleanliness; a modern outlook usually, though not always, implies tolerance and up-to-date knowledge; a modern foundation garment is invariably a 'clean' use of the word, suggesting that the garment has been designed after recent research and with the best available materials. On the other hand, an old-Jashioned bathtub suggests an elaborate design and a lack of convenience that strike the present-day speaker as comic; 'You have a lot of old-fashioned ideas about what a girl ought to be!' is hardly complimentary in tone; and an old-fashioned look is a phrase sometimes used to signify a very hostile and disapproving look. Meanwhile one of these modern girls or nasty modern ideas carries a tone of strong disapproval, though the precise fault in the girls or the ideas could be defined by context only. Modern art, modern music and modern poetry are often treated as mere terms of ignorant abuse carrying the assumption that these are bad or meaningless, and are almost always terms used very loosely at best. And so on until the careful listener or reader is tempted to wonder whether these words mean anything at all. The trouble about these words that have acquired special meanings or very loose, vague meanings that are little more than emotional noises, is not merely that they tend to confuse our thoughts. The worst danger is that they can so easily become substitutes for thought; and when we have a handy substitute for thought we generally stop thinking.

6o

VI. WORDS THAT HAVE CHANGED

G

like anything else that depends upon human ings, changes considerably with the passage of time. One result of this is that when we read the literature of several hundred years ago we are likely to be misled by words that have changed their meaning since the book was written; and sometimes much more recent writing contains a few words to which this has happened. Indeed, changes in the meaning, especially the associations, of words are sometimes so rapid that they contribute to difficulties of understanding between two generations. This is one reason why the classics of English literature, especially the works of Shakespeare, often have to be printed with explanatory notes; unfortunately the explanatory notes often look very uninteresting and arouse suspicions that the book is going to be dull. A person who is able to read Old English is not likely to be misled by these historical changes, since the language is so unlike Modern English that he knows he has to look in the dictionary very frequently; a person reading, say, Chaucer's Middle English or Henryson's Middle Scots is aware that he will need the glossary and will expect words to have, often, unfamiliar meanings; but when we are reading Shakespeare or later authors we often misunderstand because we are not looking for changes in meaning. The misleading word is the one that looks the same as the one we are used to and that, in 61 CUAGE,

Words that have Changed the context, could reasonably mean what we at first sight suppose. For example, in Macbeth Ross says, speaking of the misery of Scotland under what would nowadays be called Macbeth's terror regime, '. . . where violent sorrow seems A modem ecstasy;' (Act IV, scene iii.) If we take the words modern and ecstasy as having their presentday meaning, this makes perfectly good sense, though it is figurative: 'Where violent sorrow is the joy we have today'. This, as an indictment of a tyrannous government, is quite good. However, when the play was written modern meant 'ordinary', 'everyday'; and ecstasy had not been fixed in its present-day meaning of 'very great joy verging on an abnormal state of mind' but could mean any kind of 'frenzy', ' rnadness,' ' menta1 conrus1on, r_ • ' • • ' even, somettmes, a cr.ramt. Thus the 'modem ecstasy' implies: 'today violent sorrow is just an ordinary bewilderment' or something like that. Words may have changed owing to changes in knowledge; for example, an element today can be strictly defined scientifically, and the elements are numerous; in Shakespeare's day the elements meant earth, water, air and fire. The request in the Book of Common Prayer, 'Prevent us, 0 Lord, in all our doings . . .' sounds rather odd to presentday ears, implying a request to God to hinder us from doing whatever we wish to do. Not 'in all our wrongdoings', which would make sense, but apparently in all our activities; the explanation is that when the Book of Common Prayer was written prevent was often used in the sense of 'go before' (from praevenire) and so 'to go before to protect or provide'. John Knox's 'Monstrous Regiment ofWomen' did not mean a horrifying large number of females; it referred to govern62

Words that have Changed ment by women, which John Knox regarded as unnatural. There is a book of narrative poems called A Mi"or for Magistrates, which was published in 15 59. Magistrates here does not mean justices of the Peace or Police Magistrates or Stipendiary Magistrates; it means rulers such as kings and emperors. Sad in the sixteenth century often meant merely serious: 'Do you

speak this sadly?' meant no more than what would be meant nowadays by 'Are you pulling my leg?' In much more recent times a favourite hymn had a verse which nowadays it is difficult to sing without an inappropriate smile: 'They climbed the steep ascent of heaven Through peril, toil and pain: 0 God, to us may grace be given To follow in their train.'

Train of course means their escort, their followers; but nowa-

days, on a mountain such as Snowdon, some choose to climb the steep ascent with great efforts, while the weaker or lazier members of the party do indeed prefer to-follow in the train drawn by a locomotive. The hymn ('The Son of God goes forth to war') is a fme one and the change of meaning is nobody' s fault; but it is unfortunate. Matron, from the Latin matrona, once carried ideas of great dignity and nobility of character, because of the high standards of the Roman matrona such as Lucretia, Brutus' Portia, or Octavia; it was a word for a responsible, dignified, influential and virtuous married woman. Of recent times the word has become confined almost entirely to the matron of a hospital or school, who is certainly responsible and in a dignified position, but of a different kind; so that the word in old contexts often sounds rather odd to the reader who is more accustomed to the new context; and with the introduction of matron as a classification for a size in dresses not 63

Words that have Changed associated with beauty or grace, the word has lost some more of its old majesty. It may be interesting to examine a few of these words that have changed their meaniing, in rather more detail. I. NICE

Today this is one of the most meaningless and cor.rupted words in the English language, an adjective of praise that saves us the trouble of thinking of a more apt word. We have a nice girl, a nice book, a nice pudding, a nice house, a nice garden, a nice cat, a nice little business, a nice sum of money ... until the word is no more than an approving noise. It survives in one more accurate and meaningful sense in such phrases as a nice point or a nice distinction-' a point, or a distinction, requiring discrimination to decide . . . .' I have known a comic effect achieved in a mock trial by the judge's stroking his nose meditatively and saying, 'That's a nice point, that is, a very nice point ... ' where we were meant to take the word in the loose and the exact sense simultaneously. Nice, however, was not always so vague; and its meaning has undergone drastic changes. It is actually connected with something not at all 'nice' in the present-day sense, the Latin nescius, 'ignorant'! In I 560 it meant foolish, stupid; in I6o6 it was used to mean wanton (in the sense of lascivious); also in the sixteenth century it was sometimes used to signify rare or strange, difficult to please, fastidious and thus precise or particular as in the same scene from Macbeth previously quoted:

0! relation Too nice, and yet too true! Shakespeare also uses the word more than once in the sense of trivial, unimportant:

Words that have Changed In such a time as this it is not meet That every nice offence should bear his comment. Julius Caesar, Act IV, scene iii. The 1neaning now found in nice distinction-requiring precision, accuracy or minuteness-is recorded as early as Is I 3; and in IS86 it is recorded as meaning able to make such distinctions -finely discriminative. The use of nice as meaning 'pleasant' seems to occur first in I7I3 with reference to food: dainty, appetizing. By I769 the vague agreeable sense is recorded, and in the nineteenth century the word has come to be applied also to character: kind, considerate. So the development proceeds until a word which once had several, but definite, meanings has almost lost its meaning. An interesting further development of nice is that it is now very often used in an ironical sense: 'This is a nice state of affairs!' never means what it says, but quite the reverse: 'This tablecloth is nice and dirty, isn't it?' There is also the polite use of 'not nice' to denote something thought of as vulgar, coarse or 1mproper. 2. SILLY

This word now means foolish and is so debased that it is thought of as a little slangy or vulgar. To tell someone not to be silly almost implies that they are not only foolish but childish. Yet the word is related to the 0 ld English saelig, 'holy', and did not originally carry any idea of foolish or despicable, but merely of poor, helpless, simple, ignorant, unsophisticated. Silly shepherds are found so often in sixteenthcentury literature that the uninformed might be pardoned for supposing that sixteenth-century shepherds were proverbially brainless; in fact they were merely simple and innocent. {Incidentally the word simple has sometimes been used to mean

6s

Words that have Changed stupid and ignorant, or even mentally deficient; nowadays this is a dialect or vulgar use.) Milton's Perhaps their loves, or else their sheep, Was all that did their silly thoughts so busy keep. (Ode on the Morning of Christ's Nativity) does not mean that the shepherds' thoughts were foolish, but that they were not profound or learned. From the poor, simple, helpless, humble senses the idea offoolish developed and the word was used in this sense, but not exclusively in this sense, by the middle of the sixteenth century. Gradually the word lost its non-derogatory senses and became a word meaning only foolish and a very insulting word to use in seriousness. 3·

TABBY

I have stressed that abstract words are more likely to be

ambiguous than concrete words, but some concrete words have had great changes of meaning in the course of history. The word tabby, nowadays either a word for a kind of cat or a technical term used in weaving, has an extraordinary history: it is thought to come from the Arabic through Latin and French and to be connected thus with 'attabiy, a part of Baghdad in which a particular kind of silk was made. First it was this material, with a pattern of stripes, though later the term was extended to cover other silks that were waved or watered. Then the word took other senses by analogy, to describe various other things that were striped in something the same way as the famous silk: a brindled cat in 1774; moths with this type of marking in 1819; and a kind of concrete that had various colours in 1802. As early as 1748 the word is also recorded as ·referring to a gossippy, disagreeable old lady! 66

Words that have Changed 4· DEMOCRACY After this interlude to enjoy something trivial, let us turn to a word of immense importance today, democracy. It is assumed today and has been assumed for some time that democracy is a 'clean word', that it is desirable, worth defending, preferable to autocracy, oligarchy, or other distributions of power. I may add that I heartily agree; but the word has changed its real meaning considerably in the course of history. The word originated in Greece with a combination of the words demos, 'people' and -kratia from kratos, 'rule', and in some form is now very widespread in European languages. It would naturally be rendered as 'government by the people'. But the Greek demos is not nearly as sympathetic to the common people as the English people; and the idea that government by the people was desirable or rational came relatively very late in history. This, for instance, illustrates the seventeenth-century semantics of the word: ... still depending on the crowd, That kingly power, thus ebbing out, might be Drawn to the dregs of a democracy. JOHN DRYDEN,

Absalom and Achitophel

Clearly to Dryden democracy was the rule of the rabble. The word gradually became less derogatory, very possibly as the 'people' became more educated and elections more dignified, until in 1917 President Wilson of America could say 'The world must be made safe for democracy.' This very successful and much-quoted slogan would have sounded to Plato perverse and mad. At present it is customary to refer in non-Communist Europe and the United States to 'the democracies' meaning these countries; but the Communist countries refer to them67

Words that have Changed selves as 'the people's democracies' and do not regard the Western countries as 'democracies'. I doubt if really the 'people', except possibly in Iceland and Switzerland, which are small enough, really exercise enough control over affairs in any country of Europe for it to be truly described at present as a 'democracy'. As a last shift in meaning, the word has been so much misused in propaganda activities that from a very 'clean word' it is sliding into being a 'dirty word' that can be used only ironically or with a rather apologetic air. 5· CLERK

Today this word, except in a few special administrative titles, means 'a person doing written work, accounts and so on (nowadays typing more often than writing) as an employee in an office'. The word has changed its meaning almost completely. Closely related to cleric, the word when first used in England meant a clergyman, a churchman. Since at one time the Church was the chief source of learning, the word came, easily enough, to mean someone who could read and write; and it will be fairly obvious how this could become 'office worker'. Words also frequently change their associations for reasons of social change; some examples have already been given in Chapter III. A few more examples may be considered here. Until fairly recently in history there was nothing embarrassing or disgraceful in being bitten by a flea; the word was neutral until standards of hygiene and housing made fleas unusual. John Donne in the seventeenth century was able to write a witty poem about a flea that had bitten both his lady and himself; it is difficult to read it aloud today without provoking, at least for a moment, a giggle Donne would not have understood. The name of something now obsolete, such as stage-coach or hansom cab, doublet or crinoline, must now have quite a different set of associations than it had for someone to whom one of these things was visible daily and was taken for

68

Words that have Changed granted. Progress in astronomy has given earth, sun and galaxy different associations; and the music of the spheres, once regarded as a fact, is now a beautiful metaphor that we can appreciate at all only with the aid of a footnote. Heaven and Hell have really changed their meanings enormously as theology has evolved. Quaker and Methodist were both once disrespectful nicknames and are now respectable, often indeed honoured terms. Farmer is acquiring a different set of associations as farming becomes more scientific. Operation is losing many of its associations of fear and peril. The word enthusiasm was a 'dirty word' in the eighteenth century, with associations of religious fanaticism and eccentricity; yet it is derived from a Greek word meaning 'possessed by a god' and is again today a 'clean word' such as does us good when it appears in our testimonials or is otherwise regarded as one of our characteristics. LenJur was once known only in its mythological sense of the 'spirits of the dead', as in Milton's 'The Lars and Lemures tnoan with tnidnight plaint'; now it is known almost wholly in the sense of the charming animal of that name. Potato has become rather dull in its associations now that it is the commonest vegetable; in the sixteenth century its associations were luxurious, exotic and slightly disreputable. King and Queen have changed their real meanings with the decline of absolute monarchy and the development of constitutional tnonarchy. Distances have appeared to shorten as transport has become faster. The real meanings of the names of political parties have fluctuated a good deal more than their leaders and propagandists would often care to admit. As weapons have become more in1personal and more drastic war has been stripped of all its romantic, chivalrous associations; Othello's 'glorious war' now sounds a mad combination of words. Mill was once a mill for grinding corn, usually a windmill or a watermill; for the last hundred years or so it has been more 69

Words that have Changed likely to be a cotton mill. Until the eighteenth century nervous meant strong, sinewy; it is never used in this sense today tmless metaphorically with regard to literary style. Words, then, change in the course of history, which is why great works often need uninviting little footnotes. Geography as well as history plays its part in limiting the meaning of words; as soon as we try to work with two or more languages semantic problems are multiplied dramatically.

70

VII. INTERNATIONAL

A

SEMANTICS

years ago the fact that different languages were spoken in different parts of the world inconvenienced only a minority. Most of those who wished or were obliged to go abroad, except in war, were people who had the money, the leisure and the cultural background to learn several major languages. Moreover, the education of the rich was then so much based upon Greek and Latin that the learning of other languages must at least have seemed easier. Today the language barrier is felt far more widely. Most English people have at least some hope of a holiday abroad at some time; trade and cultural exchanges are of increasing importance, and British firms have sometimes lost business, for example, by not giving instructions for the use of some machine in the language of the importing country. Still more important, the problem of somehow preserving peace and promoting active co-operation in the world has become the greatest problem of our time, since all other progress, survival itself, depends on this. English people who have not been abroad, or who have travelled only to large towns and conversed only in the best hotels and the tourist shops, may say that 'someone always speaks English when you are abroad'. This is very far from being my experience. If we are to travel abroad without embarrassment and to see more of a country than can be seen by staying in a hotel that is very like an English one, we still 71 UNDRED

International Semantics have to solve the language problem. The Frenchn1at1 who visits England is very lucky if he can fmd someone who can speak French really usefully to advise him, although French is taught in nearly all our grammar schools. The Scandinavian has even more difficulty, although the Scandinavian countries have long been friendly neighbours. The Yugoslav, the Bulgarian, the Rumanian, the Pole have virtually no prospect of finding someone who speaks their languages without making special arrangements-very likely with an exiled fellowcountryman or a professional linguist. I myself am an enthusiastic traveller and have found Esperanto, the international language, perfectly adequate for all purposes, whether to ask if I may hang my stockings up to dry or to discuss psychological research; but one has generally to arrange in advance to meet fellow-Esperantists; the fact that relatively few people learn the international language is unfortunate, but must at present be accepted. So language barriers still exist. Anyone who has ever tried to learn a language will agree, more or less ruefully, that this is a difficult task. The mere bother of learning to call what we have been brought up to call a man an homme, dog, chien, house, maison and so on through all the words we expect to require, is an enormous labour. The irregularities and exceptions of the grammars of nearly all national languages greatly complicate the task, as anyone who has struggled with French irregular verbs or German strong verbs will at once admit. Swedish has one of the easiest grammars in Europe; Italian perhaps the easiest rules of pronunciation; both have maddeningly awkward difficulties in the way of saying a simple you, from the point of view of the English learner. Esperanto is the easiest of all languages to learn, since it was deliberately designed to be easy; but even in Esperanto the Englishman has to learn to call man viro, dog hundo, house domo; so considerable mental effort is still required.

72

International Semantics The worst problems with foreign languages, however, are those of which we are scarcely aware at first. As soon as we enter the field of international semantics we are confronted with the difficulty that equivalents are often not exact. Even a large dictionary may conceal traps for the unwary. The inexperienced student of a language begins by making obvious howlers, guessing the meaning of words from a resemblance to English words. It is very pardonable to translate chair in French as chair in English, but the French word means flesh. A Swede learning Hungarian and seeing the word apa would naturally suppose that it meant monkey, but in Hungarian the word means father; Rock in German does not mean rock in English, but English coat; morbido in Italian does not mean morbid in the English sense, but soft. Words that do not look the same, but sound much the same, also cause confusion in the spoken language; when in 1956 some young Hungarian refugees were entertained in a college and the students tried to teach them the names of common objects in the room, the word chalk caused a burst of laughter, since it resembles, in sound, the Hungarian word csok, a kiss! Later the student realizes that guessing is dangerous, and learns to check meanings in the dictionary befort deciding what they must mean. Yet the dictionary is not wholly reliable. There is the well-known story of the schoolboy who, working his way painfully through a French text, found the sentence: 'Rose emue repondit . . .' and turned it, by sincere but not very intelligent use of the dictionary, into 'The pink emu hen has laid another egg!' A quite good French speaker wanted to explain that he had just seen laundresses washing clothes with a beetle (the wooden bat used for thumping the dirt out of clothes), realized he did not know the word, tried his pocket dictionary and came out with the word escarbot, which does mean beetle, but the kind that runs across the path and has six legs. Travelling in Sweden with a camera and a 73

International Semantics very small dictionary, I asked in a photographer's shop for lokar, which does mean bulbs, certainly, but usually onions and in any case bulbs that one plants in the garden; glodlampar were what I wanted for taking indoor photographs. The lady crossing the German frontier who told the porter to carry her case into the Gebrauche puzzled him greatly; she wanted the Zoll, but both words in English have to be rendered as

customs.

Howlers are a great source of joy and very seldom a cause of real disaster. Though they do not contribute to international understanding in the intellectual sense, they are quite likely to contribute something to international goodwill. More dangerous traps are those which are less obvious and into which even someone with a fairly good knowledge of the language may fall. The two words do in one sense mean the same; but the associations are very different, or there is a shade of meaning between the two. French correct applies to conduct-proper, conventional-much more often than it does in English. A Swedish-English dictionary will translate smorgas as sandwich; but the Swedish item consists of a single slice of buttered bt'ead heaped with fish, meat, salad, etc.; it is much more palatable and much less convenient to carry than the English sandwich. The Croat kobajica is indeed a sausage; but it is nothing like an English sausage in texture or B.avour. Inversely, I have known a well-bred French girl say she wanted some bloody meat, the French saignant being a perfectly polite word. One does not ask a German in a stuffy room Sind Sie heiss? which does mean 'Are you hot?' but has an unfortunate flavour of what we mean when we say a girl is hot stuff! One must be careful with the French feminine adjective grosse, which does rnean fat but often means pregnant. A word resembling the English professor runs through most European languages, but the precise meaning of the word varies greatly in the various languages; so does the value 74

International Semantics of the academic title denoted by the word resembling the English Doctor. Even when we have mastered these shades of meaning, we may be trapped by association; and this is perhaps the most interesting branch of international semantics. An example which has caused much misunderstanding is the word socialism. It does not appear to present many difficulties: like many semi-technical words, it is international in form: French socialisme, German Sozialismus, Swedish socialism, Italian socialismo, Hungarian szocializmus, Serbo-Croat socializam, Spanish socialismo, Portuguese socialismo; but there is often an extremely important difference depending on historical factors: in England, France, Germany, Scandinavia, socialism refers to the gradual, legal and constitutional introduction of changes in the economic system, the socialists being those who belong to the group known as Labour Party in England and Social Democrats in many other countries. In Russia and all countries under Communist regimes, socialism, instead of being sharply differentiated from communism, is used as a synonym, and we hear of socialist construction, socialist realism and so on. To a Western European Social Democrat is quite a usual and respectable thing to be; to a Communist it is a term of abuse. The meaning and associations of the word bourgeois (another fairly international word) vary for Communists and non-Communists; to a Communist a bourgeois is a rich employer of labour, which to him always implies also an unethical exploiter of labour; a French bourgeois is a comfortable middle-class person (unless the speaker is a French Communist!) ; the word is rarely used in English, but when it is used it generally carries more the sense of cultural inadequacy than of class or economic function. To a Russian Communist it is a term of violent 75

International Sematztics hostility; to an English art student it is a mild 'dirty word' with about the same flavour as old-fashioned. Associations vary for less complicated reasons. Vin rouge in French has as its English dictionary equivalent red wine; but vin rouge is in France one of the cheapest and most usual drinks, whereas in England red wine carries associations of some degree of luxury or celebration. Meat is an emotionally colourless word to an English speaker who is not a vegetarian; the word for meat in some African dialect n1ay be charged with the powerful associations-of great luxury, of festivity, even of magic-that meat will have for someone who thinks it delicious but very rarely has an opportunity of tasting it. Grapes or potnegranates, peaches, bananas, are luxury fruits in some countries, con1monplace in others. My house has been knocked down does not mean the same for a man who can in one day rebuild his house from sticks, mud and leaves, and to a man who has saved up for twenty years to buy himself a brick house with glass windows, a staircase, chimneys and. all the other complications found in the average English house. Marriage does not mean the same in countries where tnarriage laws and customs differ; wife does not mean the satne in polygamous and monogamous comtnunities; widow still does not mean the satne in France and in India. The sun and rain will be friends or enemies depending on the general climate. A tiger does not really mean the same to someone who has seen the creature only behind stout bars in a zoo cage and to someone who has seen a n1an-eating tiger drag a child into the jungle. Expressions of etnotion n1ay differ in real meaning fron1 country to country. A Hungarian writing to a Swedish lady caused a good deal of trouble by finishing his letter 'With true love I kiss your hands.' This was for him a usual, conventional way of finishing a letter written to a respected woman friend. To the woman's Swedish husband it implied 76

International Semantics a decidedly erotic relationship ! I myself once caused some distress unintentionally by fmishing a letter to a foreigner 'With love', a phrase interpreted at the receiving end much more significantly than had been intended. The French Mon Dieu! is not nearly as strong in its implications as the English My God! and is much more often heard among refmed people. To call someone a cochon in French is a great deal more insulting than English pig. (It is interesting to notice the tremendous difference between English you pig, which may be quite affectionate, and you swine, which is almost regarded as foul language and is always extremely offensive. The gradations of this kind of language are often even very local; a word nearly unprintable and not used by people who wish to be thought decent in Northern England survives in my home town, Stoke-on-Trent, as a most harmless term, sometimes heard from the lips of quite respectable ladies !) In many countries, for example, France and Italy, all letters, even the most formal letters of business, are concluded with elaborate and verbally very courteous sentences. English letters have very simple conclusions-'Yours faithfully', 'Yours sincerely' and so on, unless we are writing very intimate letters, when we may make up some more elaborate -but not stereotyped-formula. In some countries titles of respect or profession are much more used than in others. Thanks are more profuse and elaborate in some countries; and taboos are more or less strict. Such social habits, if misunderstood, may cause much embarrassment. Minor misunderstandings that make us seem to be vulgar, insulting or critical when we do not mean to be, that give a wrong impression of our feelings or cause scandal, may greatly hinder happy relationships between two people of different nationality; and as the peoples of the world see more of one another in travel and conferences this becomes more important. Internatiortal semantic problems, however, have the

77

International Semantics larger importance that verbal misunderstandings may add to international tension, creating fears, resentments, confusion where such things are dangerous. English control and French controler do not mean the same; the English word carries a meaning of 'rule over' whereas the French word merely means 'check' or 'inspect'. In 1905 the drafts for the Treaty of Portsmouth (U.S.A.) between Russia and Japan were written in both French and English, with controler and control used as equivalents: which was really intended, 'control' or merely 'inspection' was a very important point that gave much trouble. A not very diplomatic diplomatist more recently gave great offence by saying to some guests from another conntry, 'We shall bury you!' This was generally taken as a rude, brutal threat and was much resented. Later, however, an expert linguist was able to explain that in the speaker's language the statement did not imply, 'We shall kill you!' but merely 'We shall survive you ... 'with the relatively mild implication, 'Our way of life is more successful than yours.' During World War One a mistake of this kind gave rise to a revoltingly gruesome atrocity story. The German High Command issued instructions that the dwindling supply of fats might be augmented by using Kadaver. Now, in English a cadaver is a corpse for dissection, and many people were led to believe that the Germans were using human bodies as industrial raw material. While it might have been argued that the slaughter of human beings was more atrocious than the misuse of their bodies, the latter violates a strong taboo among all civilized peoples. The atrocity story had a macabre success and is occasionally heard as true even today. But the German Kadaver does not denote exclusively a human corpse; it may be the body of an animal; and the instructions were that the bodies of the cavalry horses that had been killed were to be used for fat-making. To most people the idea is somewhat unsavoury, but it does not violate any taboo. 78

International Semantics Careful attention to the meaning of words is very important in politics if such mishaps are to be avoided-even if treaties are to be kept, for people of different nationalities may understand the meaning of what they have agreed to do rather differently-and this is perhaps more so, because dishonest or self-deceiving politicians frequently use language in such a way as to cause confusion and prevent communication: calling names, over-simplification, whitewashing, questionbegging, vaguely edifying language that in fact means nothing definable, meaningless promises, meaningless threats, are all, unhappily, part of the language of politics both national and -much more-international. The mere attempt to discuss the meaning of words internationally is valuable; for when we begin to do this, we are beginning to look for truth; and as soon as we begin to look for truth, we also begin to calm down, we control our extravagant passions and we talk more reasonably. An interest in words and the study of languages both have a moral and social value, as well as being intellectually interesting. Even the most superficial attempt to replace ideas taken from headlines by sincere analysis of problems, even the most amateurish, but friendly, enquiry as to what the other person really means, is a small piece of real progress.

Note: I am sometimes asked what happens in Esperanto about international semantic problems. The answer is, I think, (a) that they continue to exist, for example the word socialismo retains its ambiguity; (b) but competent Esperantists, being able to discuss such things with a neutral language, and being usually rather notably conscious of language problems, generally come rather nearer to understanding and to the solution of semantic problems than those who are struggling to explain shades of meaning without an intermediary neutral language; and (c) that Esperanto words themselves will 79

International Semantics gradually become more perfectly defmed, not only by use . but by dictionary compilation; the first all-Esperanto dictionary is fortunately very good in defmition, and an official Academy exists to solve linguistic dilemmas. I myself think that language will never be an entirely satisfactory means of communication, because human experience is very n1uch more complex than vocabulary is ever likely to be able to provide for; but that is a pathetic feature of the general human situation.

So

PART II

The Misuse of Language

VIII. ADVERTISING

T

limitations of language and of human intelligence are such that we all have some difficulty in saying what we mean and all have some difficulty in being certain that we understand what the ether person means. Sincerity in what we say cannot really be defmed as perfect truthfulness, since we ourselves often do not know what truth is, or lack the words to express it, or know that the person to whom we are speaking can understand only something less than truth; sincerity in human beings as they at present are can be little more than a genuine desire to communicate, in contrast with a desire to deceive. This desire to communicate is what people often call good faith or good will; it is often not satisfied. However, life is very much complicated by the undoubted fact that language is often used, not to commnnicate but to deceive. This is often true of political and religious propaganda, is far from unknown in the field of personal life, and is exemplified in the world of present-day large-scale commercial advertising. Advertising has two functions. One is entirely socially useful, completely ethical and relatively free from semantic problems. This function is to inform people where something they may wish to buy can be bought, how much it costs, what it is like and so on. We all need to know these things; the total disappearance of advertising would n1ake it much more difficult to find what we wanted for all except those who live in cities; and even they would have to wander in the 83 HE

Advertising streets for some time before fmding a shop which sold what they were needing. Assuming that the statements made in it are true, there is nothing whatever wrong with this advertisement: SCRATCHEMS 21

Inky Way, Slaggyford, have a large

stock of Fountain Pens, Ball Point Pens, Propelling Pencils, Writing Cases, Presentation Sets. Agents for all the wellknown firtns. Fountain Pens Repaired. If we do not have what you require, we will try to get it for you. Weekdays 9 a.m.5 p.m. Saturdays 9 a.m.-I p.m. Nor is there anything the matter with this: FARUNA

for Flour, Cornflour, Sentolina, Macaroni, Vermicelli, Custard Powder. We pride ourselves on pure goods of good quality at reasonable prices. We have just brought out a new line: QUICKIE CREAMS

a titne-saving cornflour sweet in STRAWBERRY, LEMON and CHOCOLATE flavours A IS. 3d. packet serves four. Both these advertisentents give genuine inforn1ation that may be of use to someone, and do not make extravagant claims. An advertisement of this kind can of course lie; a Is. 3d. packet of Quickie Cream may serve only three; but provided that there is no actual lie, there is nothing objectionable in the 84

Advertising words. If I want to know where to buy a new pen, or what firm makes flour and similar products, I shall fmd what I want from such advertisements. The second function of modern advertising is very different and frequently leads advertisers into behavour which, to anyone who cares about words or truth, is unethical. This function is to create a want, not merely by stating that something may be bought-a deaf man may not want a hearing aid unless he knows that it is possible to obtain one, but he may be much happier for learning to want one, acquiring it and using it-but by playing on people's emotions, prejudices and ignorance to stimulate an artificial want. As soon as advertising moves from the informative function to the tvant-creating function, it begins to use language dishonestly. Let us suppose that the firm making Kakao's Drinking Chocolate wishes to increase its sales. Now, the drinking chocolate may genuinely be good drinking chocolate: wholesome, palatable, pure, well packed and of reasonable price. If Kakao are going all-out to increase sales, it is just as legitimate to state these things as for me to state, when applying for a lecturing post, the nature of my degrees and my experience of lecturing. HAVE

you

TRIED

KAKA 0' S DRINKING CHOCOLATE? We believe that it is a very good drink. Pure, wholesome chocolate, inspected weekly by our analyst. Packed in air-tight tins in hygienic conditions. Visitors to our factory are welcome without an appoint85

Advertising ment. Note our prices: Is. 8d. half-pound tin, JS. 2d. pound tin, Family Size twopound tin ss. 1od. If you like chocolate and cocoa, try KAKAo' s. We believe that you will be pleased. Again, so long as the statements are true there is nothing objectionable here. The invitation to visitors may be bluff, on the assumption that it suggests that there is nothing to hide and thus makes people believe without investigating; but it may also be quite genuine. This is the kind of thing that is apt to happen; and it happens with extravagant headlines, pictures and repetition in many places:

takes away that tired feeling. Feel young and beautiful again ! You will have the energy to play with your children and go out with your husband if your drink KAKAo' s DRINKING CHOCOLATE every day. KAKAO' S DRINKING

CHOCOLATE

How can a firm possibly be sure that it can promise this? Tiredness may be caused by overwork, psychological distress, anaemia and other physical illnesses, or an inconveniently planned home .... People in the know Smart people Up-to-date people are changing to KAKAO' s DRINKING CHOCOLATE

Have You tried it yet?

This kind of advertisement appeals to our desire to be fashion86

Advertising able, knowledgeable and 'progressive'. It also makes us wonder if we are up-to-date in such a way that it touches not only pride but fear. Most of us feel insecure if we feel that we are 'behind the times' or 'left out of things'. And we all know this kind of advertisement: 'I know Jim can do it!' said Mother. But his teachers were afraid that Jim was not going to pass the I I plus examination. 'The boy is bright, Mrs. Hart,' they said, 'but he cannot concentrate on his work.'

These are the captions to interesting pictures of a crying mother, a worried-looking father, an embarrassed small boy and an interview with a sympathetic-looking but anxiouslooking school-teacher with perhaps a blackboard in the background. Then another teacher, of course in a gown and carrying an impressive-looking chart with scientific formulae on it, appears in the next picture. 'Sorry to hear of your tr.ouble, Mrs. Hart,' said the Science Master. 'But are you sure the lad is getting enough calories? Growing boys need a great deal of energy. Take my tip and try Kakao's Drinking Chocolate. I drink it every night myself.'

In the next picture we see Jim enthusiastically drinking a cup of Kakao's Drinking Chocolate and saying something about its delicious flavour. In the background stands a Family Size tin, and Mother is perhaps saying, 'Drink as much as you like, sonny, it is very economical.' (The punctuation of advertisements often leaves much to be desired.) The last picture shows a delighted family receiving the news that Jim is to go to the 87

Advertising Grammar School and, even as Father is congratulating hin1 and perhaps promising him a bicycle, Mother thinks (people in advertisements always think in balloons above the head) 'But he wouldn't have done it if it hadn't been for Kakao's Drinking Chocolate.' This advertisement now implies tnat a food product will help a boy to pass an examination; indeed the implication is that it was the chief factor. This easily deceives for two reasons; it is quite true that we work better on a good nourishing diet and most have experienced the useful energygiving properties of chocolate and glucose; and we would all like to pass examinations as easily as this chocolate advertisement implies, without mental effort or giving up n1any

hours to hard work. We are told by implication that people with scientific knowledge approve of the product. An appeal is made to the natural and indeed admirable emotion of parental concern and responsibility; and the I I plus examination is a real cause of strain and concern to many parents and children, so that the advertisement attracts attention immediately and appeals to fear. This advertisement, given attraction by interesting pictures and something of a story, is far removed from a mere presentation of relevant information for the possible buyer. It is now the product of much psychological shrewdness, cynical imagination, and a cheerful indifference to truth or probability. The reader will easily think of real advertisements that play upon the reader's desire for something quick and easy ('Slim without Diet or. Exercises'-'Wash without Rubbing!'); to cupidity (£Io,ooo must be won in our new Oval Oranges Slogan Competition-buy your Oval Oranges today !) ; to the sex instinct-'Nobody wanted to dance with Jane until a friend told her ...'-and incidentally the concept of sexual love conveyed by advertisements is generally shallow, cheap and unreal in the extreme, with no reference to friendship, 88

Advertising responsibility or concern for the happiness of another person; -to fear, in a civilization already too fear-ridden ('Hair in your Comb-v.That will you be like in two years?'-'Don't ignore the symptoms of Squodge-Deficiency !') ; to the desire to feel oneself superior ('Our sort of people smoke Aristo Cigarettes'-'Why did her Daughter's Husband not introduce her?'); to gregariousness ('Make friends with Flip's Sweets!' -'Everybody's talking about Anderson's new Seamless Stockings!'); to ambition ('Men who want to get ahead read Biblio Books.'-' Most top executives sign their letters with a Porpoise Pen') and so on. Another advertising technique is to produce a single short and memorable slogan, often one that has alliteration or rhyme to help memory or that is in itself something of a quotable joke; these slogans often show more skill than truth. A common advertising device which is perhaps outside the realm of true semantics is the simple repetition of a name on hoardings, in newspapers and so on. Pictures that associate the product with beautiful women, distinguished-looking men, luxurious homes, exciting holidays, loving caresses, happy family life and the admiration of others are commonly used; these devices might be called a form of non-verbal lie or exaggeration. Advertisements often misuse language in other ways, but this is not the immediate concern of the present book. The incorrect use of words; bad grammar; pseudo-scientific language that sounds very impressive to those who have no scientific background; ugly distortions of words and 'clever' trade names made from combinations or misspellings of words, often apt to set the teeth of anyone with any feeling for words very much on edge; irritatingly childish gush and childish repetitions; these are some of the offences against language committed by advertisements. Here is a fine specimen of semantic dishonesty: 89

Advertising 'SwiTCH OVER TO STULTUS POOLS AND WIN £75,000!'

This at first sight appears to carry a truly wonderful promise; but the word and is used dishonestly. It ought to imply and you will; in fact it means and you may, the winning being not only not certain, but mathematically most improbable. And here is another: 'WISE MOTHERS GIVE THEIR CHILDREN DAILY WOTTO !'

Never mind for the moment whether W otto is a laxative, a vitamin preparation, a breakfast cereal or a milk drink; what is a wise mother? The people who are most likely to know, doctors and specialists in child psychology, are in doubt as to many of the details. Every mother who has any sense of responsibility wishes to be a wise mother and to do her best for her children; an ignorant woman, and especially one who feels unsure of herself, is quite likely to take the advertiser's statement at its face value: other mothers, wiser than she, are already giving their children daily W otto. In fact the advertisement says no more than that a wise mother is a mother who gives her children W otto; it is thus no more than an unsupported assertion. However, as we all know, unsupported assertions are more often believed than disbelieved. Indeed, they have to be; normally if a man says he is unmarried, or a Cambridge graduate, or able to mow the lawn, or to swim, or that he has a bank account in the town, we do not ask for documentary or practical proo£ Ordinary people have seldom any means of checking the truth of assertions made in an advertisement until it is too late. There are different kinds of assertion. 'Wise Mothers give their Children Daily Wotto !' cannot be justified, because it begs the question of what is wisdom. 'Blan~' s Biscuits are Untouched by Hand!' means something and may be either a true statement or a lie. The statements that cannot be 90

Advertising justified usually appeal more strongly to our emotions, and thus are probably more successful in attracting customers, than those that may be genuine declarations of fact. To read advertisements with a critical eye can be very entertaining; to read them with an uncritical eye is often to waste money, to undergo disappointment, and, by accumulation of impressions, to acquire some very unrealistic, psychologically unhealthy ideas about life. The habit of critical reading is not only good exercise for our nlinds; it can make a real contribution to our health and happiness.

91

IX. WHAT IS SENTIMENTA LITY:

S

is one of the greatest obstacles in the way of the honest, careful use of words: and the words sentimental and sentimentality are among other important obstacles. For sentimental is a favourite 'dirty word' and is supposed to put an end to an argument. Unfortunately it is fatally easy to stigmatize all emotions of which we do not approve, or which we cannot understand from want of experience, as sentimental. The tone of voice in which the word is generally spoken makes it clear that this is a term of extreme contempt, possibly touched with amusement, more likely touched with malice. Something tnust now be said clearly that particularly needs to be said in a book on semantics, which is a highly intellectual, rational kind of study. Emotion is not in itself undesirable, contemptible, a sign of weakness, a hindrance to success, or an immature quality. Emotion is a necessary feature of normal human experience; it is probably the biggest driving force in our lives; it is the source of our greatest joys except, perhaps, for a few very rare people who are dedicated to the purely intellectual life; to be emotional is not necessarily to be in the least sentimental. The most emotional woman I know, a beautiful and very artistic Slav with a dazzling personality,. has also staggering capacities for self-control, patience and stoicism. The most emotional man I know is also one of the 92 ENTIMENTALITY

What is Sentimentality? two bravest I ever expect to meet, and the most magnanimous. Occasionally we encounter someone whose emotional capacities have been almost completely destroyed by mental disorder or physical disease, or by exceptional circumstances -displaced persons sometimes seem totally emotionless, for exan1ple. Our usual reaction is to be horrified; these are the faces of pitiful freaks, or incomplete human beings. Emotion sometin1es has a special language of its own, which obeys semantic laws different fronl those of factual non-emotional communication. This will be discussed in n1ore detail in the third part of this book; for the moment I will n1erely say that emotional language is by no means necessarily sen timen tal. I am myself what might be called a professional intellectual; tny intellect is trained to earn my bread and butter for me and I have tnuch satisfaction in using it. The experiences of my intellect are not and have never been one-twentieth as important to n1e, as relevant to happiness or misery, as contributory to my total development and education, as the experiences of my emotions. I believe this to be true of most human beings, including n1ost professional intellectuals. Here are some of the things that sotne people are capable of stign1atizing as sentintentality: The impulse of magnanimity towards a defeated enemy or an act of generosity to an enetny in trouble; The readiness to make some sacrifice for a person we love or admire or for principles we regard as in1portant; The readiness to renounce some of our own rights for the sake of other people; Forgiveness of those who have done us some harm, and an attetnpt to find excuses for then1; The capacity to feel admiration and reverence for a person obviously more intelligent, more creative, more magnanimous or in son1e other way of greater moral achievement than our93

What is Sentimentality? selves, and the desire to imitate the good qualities of that person; The feeling of pity for persons in distress, the impulse to comfort and caress them and the impulse to perform practical actions in order to help them or console them; The practice of commemorating important events in our lives and raising morale by celebrating birthdays, wedding anniversaries and so on; Acts of commemoration to pay respect to those who have died in the service of mankind; Gentleness and patience in our dealings with children, animals, the elderly, and in general those weaker than ourselves; Respect and kindness to our parents when they are growing old and we are in the prime oflife; Attempts to promote friendly feelings and understanding between the peoples of different nations; Attempts to suggest methods of dealing with criminals that shall aim at cure rather than mere punishment; Attempts to protect the weaker members of society by means of various welfare schemes and social services; The love of a parent for a child; Friendship that includes the exchange of confidences and help and is not merely the barren acquaintanceship of bridge partners or talking about the weather; The experience of being genuinely in love, which for millions of people is the greatest illumination that life has ever given them. To obliterate such emotions, such actions, such experiences from our lives would not be to become rational and ·superior beings. It would be tragedy. It would not be the beginning of real human progress; it would be the end of all hope for human progress; it would be the beginning of the human ant-heap, the true reign of the robots. Perhaps fortunately, 94

What is Sentimentality? the end of emotion would also be the end of mankind in one generation; for no more children would be born; it is even imaginable that the very will to live would cease to exist. I was once asked what my idea of Heaven was. I could give many answers, some flippant, some abstract, some allusive; but my answer then was: 'I think that Heaven would be a state of mind in which we were all able to treat everyone else with the generosity, the charity, the understanding and the happiness in their existence that we show when we are genuinely in love with a particular person.' I am not ashamed today of this definition. Experience has so far confirmed in me the belief that what little wisdom we have is acquired mostly through our affections. It has also, on the whole, made me believe that we do well to be demonstrative and affectionate in a love-starved world, and to express freely at least those of our emotions which are not objectionable to others. An investigation of the meaning of sentimentality helps us to combat this wrong use of it as a 'dirty word' to denigrate our emotions and frighten us from our tender, humane, loving or loyal impulses. In fact sentinrental people are often not very strongly emotional. Strong emotions are very often found in conjunction with strong character, considerable selfdiscipline and high principles; those strong emotions themselves tend to provide the dynamic for character development. Sentimentality is a falsifying of emotion; properly used, the word implies something essentially unreal. Let us take a few concrete examples. When we begin to feel sorry for ourselves, we soon become sentimental; we see our case as more pitiful than it is; we imagine how sorry for us other people would be if they knew of our woes; we also blame the harsh world for not understanding us; we begin eventually to feel the kind of glow of self-approval that would be justified if we were really suffering for the sake of some friend or some 95

What is Sentinzentality? great cause. We come to feel a sort of satisfaction in our own wretchedness. Sentimentality! The realistic reaction to suffering, except in very rare circumstances (there are hopeless cases, but not many of them) is to see what we can do to remedy our distresses, not to sit down and wallow in them. Two sisters lived together. Once a week they visited the grave of their dead brother, of whom they were very fond. One day the younger sister was seriously ill in bed on the usual day for the visit to the grave. She begged her elder sister not to leave her, being afraid that she might become worse. The elder sister argued with her, then burst into tears and said, 'You can't have loved poor dear Theodore to be so disrespectful to his memory, and you can't. love me either to ask me to do such a wicked thing !'-Sentimentality. What has happened in the emotional life of the sisters is that, as it were, their scale of priorities is all wrong. A journalist, some years ago, was astonished by the reactions of the audience at a film depicting scenes of the Spanish Civil War. Fearful scenes of death, wounds and destruction flickered across the screen. The audience sat watching with remarkable stolidity as human beings suffered, human beings, incidentally, who were often passionately devoted to principles. Suddenly the film showed an overloaded baggage donkey walking through the gunfire. A woman who had before seemed quite unmoved exclaimed then, in a voice choked with tears, 'Oh, the poor little donkey! Look !-isn't it a shame? How can people be so wicked?' It is not sentimental to love or to be 'in love'. Much sentimentality often attaches itself to love, precisely because the experience is so real and so important that many of us have difficulty in coping with it. A very common form of sentimentality is the line of behaviour that may be summed up as 'If you won't do this for me, you don't love me,' or

96

What is Sentimentality? even, 'If you don't think just as I do, you don't love me.~ Now, it is true that anything I am prepared to call love includes a serious concern for the welfare and happiness of the other person; but someone who claims that love should mean a total surrender of the 'loved' person's will has a very immature and very unrealistic idea of love. Love is a partnership, not a swallowing. Jealousy is an emotion that I suppose we cannot always help, though I think it is one to be restrained as far as possible; but the assertion that jealousy is a proof of love has always seemed to me to be a very sentimenta1 assertion, based on the sentimental premise that true love is possessive, monopolistic. In particular, we find almost unrelieved sentimentality in the not uncommon assertion that unprovoked jealousy in a proof of love. The man who fmds his wife on the sofa kissing another man has a problem to face; he is jealous reasonably and must deal with his jealousy and his wife's behaviour as his own set of values and morals dictate; but the man who makes a scene because his wife has an ordinary letter from another man or is polite to another man at a party is not showing his love; he is showing childish possessiveness and insulting mistrust. Many of us, behaving like this (and most of us do fall into unreasonable and childish jealousy on occasion, if not in love then in some other sphere of life), then make beautiful excuses to ourselves: 'I shouldn't make scenes if it weren't that I am so much in love with you.... ' {Liar!) or 'I am not jealous, but I don't like her getting so much attention.... ' {Liar!) This is sentimentality. It is often difficult to see at what point sentimentality begins when we set a value higher than their market value upon objects because of the circumstances in which we acquired them. There is always a danger of growing sentimental over such things; probably the answer is that we should watch ourselves and aim at keeping such feelings 97

What is Sentimentality? under the control of our common sense. That a gift is more precious because a loved person gave it is perfectly normal and rational; that we should grieve for days because it has accidentally been broken or lost seems to be rather losing a sense of proportion. Before we have any right to call emotional 1nanifestations sentimental, except in such a phrase as 'of sentimental value', which is not a 'dirty word', one of the following three factors must, I think, be present; and usually, where there is real sentimentality, more than one will appear: I. Some kind of unreality in the emotion itsel£ This can often be felt rather than defined. It may be found in an unreasonableness of expectation-'If you lose your temper with me, you can't possibly love me !'-'You ought never to forget that I lost my husband three years ago !'; in a want of proportion as between the stimulus and the emotion, as when someone sobs broken-heartedly for hours and makes the maid's life a misery because the maid has broken a vase the mistress liked; or when someone goes into a passion of anger and cannot be conciliated because someone has said in a good-humoured tone, 'You are a fathead!' or when someone goes off into 'You hate me!' denunciations over a mildly expressed refusal to do something; in an element of deception or self-deception which gives a feeling of play-acting (though here we have to be very careful and hesitant in our judgments, as, depending on temperament, nationality, articulateness and state of health, some people do in fact express their authentic emotions much more dramatically than others); in a very persistent refusal to listen to courteous and reasonable argument after a moderate interval in which to compose ourselves; in a prolonged repetition of some assertion or emotional gesture without developing any further or doing anything constructive; and so on. We always have to be cautious and charitable, and should 98

What is Sentimentality? be more ready to suspect sentimentality in ourselves than to denounce it in others; people of sincerity and good sense do differ considerably in their emotional experiences and reactions and generalization can be very misleading. There is a demonstrativeness that is gushing and insincere, and a demonstrativeness that is a sign of genuine emotion. There is an emotional excitability that springs from a lack of ordinary self-contrcl and common sense, and an emotional excitability that may be combined with great self-control, the outbursts of open em.otion being infrequent or well motivated. We must not go about assuming people are sentimental because their emotional behaviour does not coincide with ours. 2. An exaggeration in the expression of the emotion. Such violent expressions as 'I'd rather die than . . .', 'You have always been cruel to me ... ', 'I shall go mad if you don't do as I ask ... ', 'I will do anything for you! ... ' may be true, but more often they are not. When we begin to think about the meaning of words, we tend to dislike the sound of such expressions on our own lips, and to feel a desire to check their genuineness before saying them. 3. Some degree of deliberately stitnulating our own emotions further. We all know what it is to nurse anger, to wallow in our own misery until, in a sense, we are enjoying it, to make the most of a fit of bad temper, to sulk longer than we can help in order to enjoy the embarrassment of whoever has annoyed us, to search out associations that will tend to make us cry at a stage when we might be able to restrain our grief, to make little pilgrimages and repeat actions from a desire to feel the relevant emotions again (though this may, in a few contexts, have its genuine value), to brood over the photograph of someone we love who is unattainable, and so on. A great deal of political propaganda plays upon the human capacity for enjoying our own emotions; unfortunately, one of the favourite emotions of the demagogue is 99

What is Sentimentality? hatred, which nearly always does harm; if he aroused an artificial tenderness or an artificial desire to co-operate he might accidentally cause a certain amount of objective good. I have tried at some length to differentiate sentimentality from genuine emotion, partly because sentimentality as a dirty word is politically, socially and psychologically a very dangerau~ one, and partly because the people for whom this book is primarily written may fmd this distinction a particularly urgent one. What is the effect of sentimentality on our use of words? Exactly the same: unreality, exaggeration and a use of words to whip up more emotion rather than to achieve communication. Words such as love, hate, angry, broken'-hearted, grieved,

hurt, loyalty, devoted to, adore, crazy, cruel, wicked, wonderful,

lose their proper meaning and become little more than emotional noises. Other words such as home, flag, family, mother, son, may be used to heighten the false or immature emotion by association. The language of sentimentality is also very often either an imitation of the language of strong emotion that has been heard in the past, or an imitati0n of the language in which strong en1otion has been indicated in films that the speaker has recently seen. The imitativeness is part of the unreality. Another common sentimental use of language is to compare ourselves to someone to whom we very clearly do not deserve to be compared, or to compare our opponent or wronger to someone very much worse. In the speech of sentimentality we often use our words as much to keep our own emotion alive as to communicate it to anyone else, or to make another person believe we are feeling the emotion, in order that they will have the reaction we want: will pity us, admire us, be sorry for the dreadful thing they have done to us, love us in return, be suitably grateful to us, help us and so on. The opposite of sentimentality is, then, not lack of emotion, 100

What is Sentimentality? not even self-control-for not all tmcontrolled emotion is unreal or even undesirable-but sincerity. In using our words, as in living our lives, we all need to consider, every now and then, whether we are really being genuine; and we must bear in mind that being fully sincere almost certainly implies, on many occasions, beingfoir and being moderate.

101

X. WHAT I TELL YOU THREE TIMES IS FALSE

E

who has ever learned the multiplication table, a poem, or some irregular verbs by heart knows that repetition is very important indeed in fixing information in our minds. All teachers make great use of this fact. Only very rare people {Lord Macaulay is said to have been one of them) absorb large quantities of information after simply receiving it. Most of us learn things largely by being told, or telling ourselves, over and over again. The methods of hypnosis and of those useful psychological treatments based on suggestion and auto-suggestion make great use of repetition. Thus repetition is extremely important as an educational tool. Unfortunately this valuable tool can be 1nisused. I can, by frequently saying to a child, 'Look both ways before you cross the road!' ftx in the child's mind a habit which may save its life or limbs. I can with equal success say so often to a child, 'You are a clun1sy little boy!' that he becomes incurably clumsy, a failure whenever he has to use his hands. A preacher may tell his congregation repeatedly that they should try to think charitably of their fellow human beings; and he will do at least some good; but a demagogue will be equally successful in convincing his admirers that the inhabitants of the Aleutian Islands are about to attack Canada, if he says so often enough, in spite of the extreme improbability of such an event. VERYONE

102

What I Tell You Three Times is False Because when we have a habit of hearing things we tend also to acquire a kind of habit of at least readiness to believe them, we can usually examine the meaning of some quite unfamiliar statement more critically than that of a statement we have often heard before. Most of us have probably had the disconcerting experience of thinking we believed something until confronted with an unbeliever who argued, and then suddenly discovering that we did not believe, that we had taken the belief on trust from so early an age that we had never examined it. If I say to a group of students or children, 'It is very wrong to squotch wogglegooks,' they will demand a defmition of this mysterious misdeed and not wish to take my assertion for granted. If, however, I say, 'It is very wrong to make other people unhappy!' almost any British child or student who has been brought up in an average well-intentioned home will at once agree with the statement, and will probably not realize that making other people unhappy may be defmed very strictly or very loosely, depending on our definition of unhappiness. It follows that, since we are more likely to be misled by repeated assertions, precisely the assertions we most often heat are those we should examine most carefully with regard to their meaning, in order to consider whether or not they are true. Our most dangerous errors are probably the ones we take for granted. This does not mean that something we often hear is necessarily not true, or even atnbiguous. We have often been told that twice two are four; the statement that twice two are three is more original, but the commonplace is true and the daring innovation is of no value. The fact that a very large number of sane people agree on something, for example that peace is desirable, or that money does not guarantee happiness, or that a hungry man is an angry man, does not prove the truth of the statement; at one time an enormous majority 103

What I Tell You Three Times is False of no doubt sensible people believed that the sun moved round the earth; but it is some evidence in favour of the correctness of the assertion; if a great many sane people all come to the same conclusion about something, they have at least some reason for doing so. Two kinds of repetition that tend to make us uncritical are slogans and cliches. Slogans are, in the present-day sense of the word, which once had a rather more dignified meaning, short, memorable and impressive sentences used to summarize a political aim or to advertise some product or perhaps idea. They include such statements as 'An apple a day keeps the doctor away', which contains some measure of truth, in that apples are a healthy food, but which is demonstrably not fully true; 'Beer is best!', which does not mean anything until we know best of what or best for what; 'A fair day's wage for a fair day's work!' which is most reasonable in tone, and to which, as a principle, most people would agree, but which contributes nothing to policy, since both the fair day's work and the fair day's wage require definition; 'Vote for Putten and Prosperity !'-which implies that Putten and his party can guarantee something no single party in a single country can possibly guarantee, and leaves prosperity undefmedwhose? when? what kind of? what level? at what cost in other things? and so forth. The purpose of slogans is: (I) to impress by repetition and make us do what someone wants, whether something sensible -'Wash your hands before eating' or something less sensible -'Eat Blotto's biscuits and be Beautiful!' (2) to simplify what is usually a complicated matter so that it has an immediate appeal to uncritical minds. Political slogans, for example, generally make large and attractive offers, but do not give any indication of how the goal is to be reached or even how the party can be sure of reaching it; though the most magnificently unintelligent political election slogan I ever saw, 104

What I Tell You Three Times is False which must have been making its appeal entirely to the inarticulate disgruntled, and which was in fact, I understand, quite successful, was: 'It's Time We had a Change !'. This is really clever as propaganda, though utterly stupid as argument; it leaves the disgruntled elector to imagine that a new government will be able to redress all his grievances and that all kinds of wonderful things will happen. Advertising slogans make the names of products conspicuous, so that people will select a named product in a shop when buying; but they also make encouraging generalizations. Advertisements frequently use 'clean words' which immediately sound good-' the scientific soap'-'the modertz way of washing'-' the friendly firm to deal with'-'the drink for the home'-'the latest thing in figure control'-'the paper for top people' and so on. When we stop to analyse the meaning of words,advertising and political slogans can no longer have any power over us, for they are nearly always meaningless in the true sense of meaning; that is, they do not really tell us anything. Then we shall, before buying a product, want to look at it, learn what we can about it, listen to what friends who use it say, and so on; before voting for a political party, we shall want at least to know something of its real programme and the qualifications, moral and intellectual, of its leaders; we shall, in short, make some attempt at making our decisions like fairly rational beings, instead of acquiring what could almost be called a set of mental conditioned reflexes to certain sounds. Cliches are, as it were, the slogans of our everyday conversation. They are those combinations of words which we all use often without stopping to think. They do not improve the style of writers and too many of them make even ordinary conversation dull. Worse, they tend to be a handy synthetic substitute for both expression and thought. I do not think it is possible to converse or write without 105

What I Tell You Three Times is False some cliches. We cannot, at least in ordinary talk, be thinking every second of exactly what we are saying; a good deal of daily talk is not aimed so much at exact communication as at making friendly, obliging noises. When two people talk for a few seconds about the weather they are not usually interested in it; it is merely a topic that will not be controversial and that we discuss to indicate a willingness to make ourselves pleasant. Possibly in some primitive society you might give me a nut and I give you an edible berry with exactly the same real significance, 'l-am-willing-to-be-friendly-with-you-andwish-you-no-harm!' as is evident today when you say, 'Lovely weather we've had for a couple of days, haven't we?' and I reply, 'Yes, much nicer than last week.' That we often say, 'I am sorry to have been tactless!' when we are really thinking, 'You are a nuisance with your morbid touchiness!' or that we say, 'How well you look!' when we are thinking, 'I bet most of that complexion comes out of a box!' is just as well if people are to live together in something like peace. We cannot possibly, in the world as we at present know it, be on terms of close friendship with everyone we 'know'. Similarly, most of us often have to write a letter whose purpose is not really to communicate, but to show another person that we have had a kind thought about them. The traditional postcard from a holiday resort-'Having a wonderful time; wish you were here!' does not express a real longing for the company of the addressee; it is a way of showing friendliness, an expression of a desire, sincere or not, that another person should also have a wonderful time. I was once able to amuse my mother, with whom I am very intinlate, with a holiday postcard, 'Having a wonderful time; just as well you are not here!' When in our speech or writing we really are aiming not at friendly or protective gestures but at communication, cliches are often a great hindrance. Unfortunately it is very easy to 106

What I Tell You Three Times is False use them as if they were fmal, devastating answers, either to our own arguments or to those of other people. They are expressions that we have accepted so often and for so long that we find it very difficult to question them. 'A little of what you fancy does you good.' For mature people in fairly good health this is often true, because a small treat gives psychological satisfaction and that tends to promote health and balance. The human body is normally fairly adaptable. In many circumstances the old lady who feels she wants a strawberry ice or the workman who sees and smells some fresh sausage rolls and thinks he will have a couple for tea instead of the usual sandwiches, will do well to indulge their fancies; and in the experience of many people a 'fancy' is Nature's way of indicating that we need something in our diet. (I hardly ever touch an orange except when I have a cold; then I seem to acquire an almost insatiable craving for them; and vitamin C is said to be very valuable in the treattnent of colds.) But to a person just convalescing from typhoid fever 'a little of what he fancies' may be even fatal, this being an occasion on which a patient is apt to fancy very unsuitable articles of diet; and a young child not infrequently wants to eat something that will do harm. 'Mother knows best.' A good mother has, in the process of bringing up children, acquired a great deal of valuable experience, of balance, of knowledge and of inner wisdom that makes it likely that what she has to say is worth listening to. Not all mothers are wise, kind or conscientious; and not all hun1an beings seem capable of learning n1uch from experience. I think myself that any mother who is making a fair success of her hard and responsible job deserves a great deal of respect, especially frotn her children; but the mother of an atonuc physicist is most unlikely to know better than he does about atomic physics, the mother of a ballet dancer probably knows comparatively nothing about her daughter's art and 107

What I Tell You Three Times is False career; and many wise and good mothers cannot claim to know best about some matter affecting the lives of their grown-up children, merely because the whole set of circumstances is very different from that of Mother's time. 'Love is blind.' This is a useful cliche for sneering at people and often a useful excuse for ourselves. Is it at all true? I do not think so. There is a certain kind of physical infatuation that, for a short time, is 'blind' in the sense of being uncritical, tmcomprehending, without any sense of proportion, without any patience or any forethought. It is a painful and sometimes a tragic experience. But what mature people would call love, either sexual love or some other form oflove -that of mother, father, brother, sister, child, friend, for example-is often extremely realistic, shrewd, perceptive and understanding. In the happiest marriages I have been privileged to see, husbands and wives, so far from being 'blind' about one another, had an extremely detailed and subtle perceptiveness about one another. This included a capacity to perceive some faults and weaknesses; it also included a very great capacity to forgive those faults and excuse those weaknesses, as we must all forgive one another many things if we are to live together peacefully. Real, mature, responsible love is, in my experience and observation, a very important source of knowledge about people and even of wisdom. 'It'll all be the same in a hWidred years!' is a wonderful cliche for excusing ourselves, for refusing to give sympathy to those in trouble and for avoiding responsibility for doing anything about abuses. It is, perhaps, one of the stupidest of all popular cliches. For life is so extremely complicated that almost any circumstance, any word, any action, may have an effect on what happens in a hundred years' time. For example, a girl today may be upset because her boy friend has grown tired of her. It is very easy to say, 'It will be all the same in a hundred years !'-very easy, that is, for someone else, who 108

What I Tell You Three Times is False does not want to 'waste time' sympathizing. Now, suppose in her grief, anger and humiliation the girl too hastily accepts the courtship of someone else who is not really suitable, as has often happened; makes a rash marriage which turns out to be an unhappy one; and then tells her young daughter, eighteen years later, that men are horrible and selfish. The daughter, at twenty-five, marries, for, being a normally constructed human being, she is capable of 'falling in love'; but, because of psychological resistances created by the talk of her disappointed mother, she in tun1 becomes a disappointed and disappointing wife; and the son ofher unhappy marriage is given the impression that n1arriage is not a good idea at all. He is determined to remain 'free'. Thinking himself a 'clever fellow' who is not going to be 'hooked', he, some fifty years after the original episode that was going to 'be all the same after a hundred years', becomes the father of an illegitimate child. This boy has a very unhappy childhood and as a result grows up with a feeling of resentment and grudge against society and a distrustful attitude towards everyone. He also puts a very exaggerated value on the usual family life that he has been denied, thinking that he would be perfectly happy if only he had had a home. Craving to be important, he goes into politics and becomes a clever speaker who knows how to play upon the unhappiness of others, the secret feeling we nearly all have that someone or other is preventing our total happiness. In middle life he becomes a dictator of the Fascist variety, chases women back into the kitchen, undoes all the good that has been done in many years by social workers and psychologists, persecutes minorities in the country, and perhaps plays an important part in starting a war. A girl's mortification a hundred years ago was a link in a chain of consequences that perhaps caused several million premature deaths and uncountable misery. I have, of course, imagined a dramatic and extreme case; but anyone

109

What I Tell You Three Times is False who looks back on life can sec occasions when something that to others seemed trivial played an important part. in, the history of a life; and all lives affect the lives of some other people; so any cliche that renounces responsibility may be a very harmful one. Thus whenever we are presented with some cliche as though it were a universal truth we must immediately accept, we shall be wise to examine it. Two other ways of winning assent to something that is not necessarily true, without having to go to the trouble of defending it, are 'taking for granted' and 'begging the question'. These dishonesties -not necessarily intentional dishonesties-ar e usually not so much a carelessness about the meaning of single words as a general carelessness in thinking. We tend, for instance, to take for granted in argument that: all people wish to be healthy, when many people fmd certain advantages in some ill-health; all women wish to marry, which is not a 100 per cent true; hot sunny weather is better liked than rainy cold weather, though a minority of people feel unwell in hot weather; work and effort are disagreeable, which is very largely a falsehood taught by the too numerous people whose work is excessive or seriously uncongenial; only children are spoilt, when in fact the parents of only children are quite often so anxious not to spoil them that they are over-strict and expect impossible standards; what does another country good does our own country harm, which in many fields of activity is the opposite of the truth; similarly, that in our own country there is only a certain amount of material wealth and that if one group receives more another group must receive less, when other economic changes may modify the amount available and so complicate the arithmetic; home life is desirable for everyone, when a few people have neither the desire for it nor the need for it; all parents do their best for their children and want their children to be happy110

What I Tell You Three Times is False which any social worker can tell us is not always true; and so on. Much of what we are ready to take for granted is 'true' enough for our everyday purposes, in that it applies to most people in the circumstances with which we are likely to be dealing. 'Begging the question' is more truly dishonest. We cannot question everything all the time, or nothing will ever get done; but we can try to argue fairly. When someone says to her teen-age daughter, 'You ought not to wear lipstick because it is wrong,' she is saying no more than 'You ought not to do it because you ought not to do it,' which does not contribute anything to the argument. When a politician says, 'All who truly love their country will accept the necessity of these sacrifices' he is making it very difficult for anyone to argue that the sacrifices are unnecessary, for fear of being thought nnpatriotic. If someone says to me, 'I think you are being too lenient with your pupils; I would like to show you the careless work some of them have been doing lately, and to tell you what Jane, Susan and Pauline have been up to in the vegetable garden! They are getting into the idea that anything will do!' it is quite possible that she may have sotnething to say which will modify my theories and my behaviour; if she says to me, 'All sensible people know that girls need strict discipline, and I wonder that an educated woman like you can be so stupidly soft!' I can safely take no notice of her, for her 'All sensible people .... ' is obviously only a paraphrase for 'All people who agree with me .... ' 'What do you think of the outrageous, wicked views of Freud?' assumes that the other person has already made up his mind to agree with the speaker. If the reply is, 'I think them neither outrageous nor wicked, but am not sure that I understand their full implications!' the reply may never be prononnced for fear of being thought 'wicked', or may be III

What I Tell You Three Times is False greeted with embarrassing astonishment. When people use begging-the-question arguments against us, we can generally, if we are alert, detect the dishonesty, the way in which we have been given a phrase, not a contribution to argument; when we do it ourselves or it is done in defence of something in which we believe, the dishonesty is less obvious! I knew a charming and clever Pole whose knowledge of English was imperfect and who had a special and rather engaging technique for begging the question. When cornered in argument and unable to give a reply, but still obviously feeling that only a very perverse individual would question the truth of his first assertion, he would say, waving his hands appealingly, 'Oh, well, because because.' This sounded rather sweet and was well calculated to appeal to the indulgent maternal instinct. But it was not argument. To do him justice, it was not meant to be, and was usually a signal to put the kettle on.

112

XI. WHITEWASH LANGUAGE

S

neurotic people are often apt to dislike and despise themselves and to think much worse of themselves than the objective facts would justify. The majority of human beings, however, tend to judge themselves and their families, organizations, countries and so on more leniently than they judge their neighbours, other organizations or other countries. Thls comes out in our choice of words. So long as we realize what we are doing it does not matter; but when we are not conscious of this natural human inclination we can be very untruthful, unfair and even dangerous by our misuse of words. There is a joke conjugation: 'I am firm, thou art obstinate, he is pig-headed.' I would continue it in the plural as 'We are united in perfect solidarity, you people all stick together, they are an obstructive clique.' This flippancy contains a great deal of psychological and moral truth. 'I have quarrelled with Green,' says Brown, 'because I can't get him to see reason at all. What an obstinate fellow he is ! And he is an egotist too; he cannot see any point of view at all except his own. Now, I claim to be a man with a will of my own. When I have made up my mind, I know what I want and I set out to get it. I don't allow anything to stand in my way; I don't shilly-shally and waste time with a lot of words; I go ahead and get it. But you can't talk to Green; he just wants his own way and nothing I say can make him see reason.' Translation: Brown has quarrelled with Green because ERIOUSLY

113

Whitewash Language Green, just like Brown, can see no point of view except his own and wants his own way! Now, in such conjugations as: 'I pride myself on my civilized tolerance; you are rather too broad-minded if anything; he has no moral sense at all,' or 'I do know how to relax; you might put more effort into things; he is bone lazy,' or 'I am glad to say I enjoy tny food; I believe you are putting on weight; he gorges like a pig!' or 'I am tactful; no one knows where they are with you; he is a slimy hypocrite!' It is not always easy to say which form of words comes nearest to the truth about the person. To add to the complications of trying to pass judgment, circumstances frequently alter cases; for instance, what may be a wise awareness of how to relax in someone just recovering fron1 a breakdown may well be bone laziness in someone else; an active growing adolescent may have what in a man of fifty would be a hoggish appetite, and n,eed every crumb consumed. We cannot assume that either the praise of the firstperson form or the brutal criticism of the third-person form, or for that n1atter the moderate criticism of the secondperson form, is the way we ought to think. What we ought to bear in mind is that we are extremely apt to whitewash ourselves and also to blackwash other people. 'Well, what do you want?' is not always the right reaction to a compliment; many people deliberately seek opportunities of paying honest compliments in order to give pleasure to other people and to encourage them. We are wise to remember that those who are always attributing the lowest n1otives to others arc giving, unconsciously, a disconcerting picture of how their own motivations work. Before examining whitewash language, the language vve tend to use in speaking of ourselves, I will touch upon a point that may be helpful to some. We are all taught, if we receive any kind of moral instruction, that we should try not to be II4

Whitewash Language selfish, not to think too much about 'I'. This seems to me to be very sound; the more we all try to make others happy and the less aggressively we push our own interests, the more likely we are to make the world a pleasanter place. But the advice not to be selfish is often misunderstood in two ways. One is that legitimate statements of objective fact about ourselves are apt to be taken in England as 'blowing our own trumpets'-a champion tennis player is rather expected to talk to the effect that he pats a ball about a bit for fun; a very brave man, much decorated for valour, is expected, if someone like me who is not brave asks him to give some advice on cultivating courage, to disclaim any title to courage, instead of giving advice which could be useful. A rich man is supposed to pretend that he is poor, a clever man that he is stupid, a staggeringly beautiful woman that she has nothing special about her; and of course a child or adolescent is not supposed to mention anything meritorious he or she may have done. I suspect that this convention may lead to more misunderstanding than it prevents and may deny us all much interesting and useful information; there is all the difference in the world between noisy boasting and a matter-of-fact admission that one has done something rather unusual. A second convention of unselfishness is that it is egotistical and vulgar to use the word 'I' avoidably. Is this true? Nobody likes a person who is endlessly talking about himself or herself and does not seem to show any friendly interest in other people or respect for their achievements; but 'I' is a perfectly respectable and necessary word to describe the person who is speaking. 'Don't know how anyone can eat cod's roe. Disgusting, revolting stuff!' 'I don't like cod's roe.' The first is a silly over-generalization; many people are fond of cod's roe and do not fmd it unappetizing; the speaker 115

Whitewash Latzguage implies that there is something wrong with other people's taste if they can eat it. The second speaker, using 'I', merely makes a reasonable statement of fact that may be useful to a hostess or nurse in planning meals. The same technique of avoiding the terrible taboo 'I', when carried into more important fields of activity, leads to intolerance and rudeness. 'All sensible men are agreed that we should guard our oil interests in the Middle East, by force if necessary-drop a few bombs on the blightcrs if they give trouble! And you will fmd that everyone who cares for his country thinks the same way!' The man who says this implies that if I happen to disagree with him on grounds of morality or expediency I am a silly person, and goes on to imply that if I disagree I cannot love my country. This makes decent argutnent embarrat;sing, perhaps impossible to a shy opponent. 'I think, myself, that our oil interests in the Middle East are so vital that we ought to defend them by force if necessary. To me this is the only patriotic attitude.' The use of the 'I' immediately leaves room for the possibility that the 'you' may think differently and has a right to do so; that the opinion expressed is held only by the 'I', one fallible human being. Surely this is in fact a more modest way of putting things than to imply that what the 'I' happens to think may be attributed to 'all sensible men'? This careful avoidance of the 'egotistical' I leads some people to use we when they are neither royal nor editorial; to use one in a curiously affected fashion; to criticize and censure others as a mode of praising themselves; to make these 'all sensible people' generalizations so as not to admit that 'I belie.ve' ; and to give reasons for doing something that are not in the least genuine, e.g. 'It would do you good to have some exercise !' as a polite expression for 'I don't feel 116

Whitewash Language like cutting the lawn.' A little more admission that we are all selfish and all fallible and that each 'I' generally thinks first of 'me' might tend in the long run to make us less rather than more selfish, and certainly more sincere. We may now turn to the things that 'I' tends to say about 'me'; the whitewash language whose function is to justify 'me'. The most dangerous feature of this habit, which we all have more or less, is not that it deceives other people; very often it does not; but that it helps us to deceive ourselves. Bill gets drunk. This can happen to anyone. If Bill's subsequent attitude is: 'I got drunk; that was foolish and dangerous; I must remember how much I can take in future and not take more,' it is to be hoped that no great harm is done. But this is often not what happens at all. Bill comes home, falls over the milk bottles on the step and breaks them, kicks the cat, smacks the baby who has done nothing amiss, swears horribly at his wife, throws little Jolm' s infant-school drawing on the fire, strews some of his clothes on the landing and falls into bed with his boots on. In the morning his wife is upset. So Bill says pathetically, 'I can't call my soul my own. Can't a fellow have a couple of drinks now after doing a hard day's work?' 'A couple of drinks' is a self-justifying expression for perhaps ten drinks. Later, at work, Bill says something like, 'You know, my missus made no end of a fuss because I was a little merry last night.' 'Merry' sounds quite pleasant, gay, friendly; 'drunk', which is what Bill was, would be too self-condenmatory. There was the experienced magistrate who said that if he was to believe the accused and the witnesses in all the cases concerning road accidents that he had heard, most collisions took place between two stationary cars each on the correct side of the road. The 'I' who is driving too fast is 'letting her rip'; the 'he' who is driving too fast is a 'road-hog'. There is a gruesome story, said to be true, of a man who was en1

117

Whitewash Language dangered by an accident on the river in a small boat. Another man gripped the edge of the boat to save himself from drowning; the speaker actually said later, 'So I had the presence of mind to strike him a hard blow over the knuckles, and he sank.' Some people would have called this something rather different. This kir1d of whitewash language often amounts to downright lying; but it is not usually deliberate lying; it is born from our extreme unwillingness to attribute anything deplorable to ourselves. The most accurate and even pedantic professor of languages may fmd that his carefulness in the use of words disappears when he is explaining why he ought to have some special leave, or could not possibly have done the work he was supposed to do, or did not return that important book to the library! Whitewash language is not confmed to our personal lives; it is also much used by politicians and other persons in high authority. Then the actions of governments and parties, of police and armies, are whitewashed with language that sounds very pleasant. Sometimes this whitewash language is probably quite sincere; our capacity for believing what we wish to believe is such that self-deception is very easy; sometimes it must be cynical dishonesty; at least, it is difficult to see how some politicians can possibly believe the things they say, unless they hypnotize themselves or their senses are defective. This is the kind of thing we are all familiar with: 'We make no claims that are not supported by the principles of truth and justice.' We think so-perhaps. 'We shall defend the dignity and honour of our country to the last drop of our blood !' We refuse to compromise, even to save the lives of many of our young men-and many of yours, too. 'Subversive elements have been suppressed and now we 118

Whitewash Language can all go forward in unity to our glorious goal.' The opposition has been killed, or imprisoned, or is missing, so the government can now do just as it pleases. 'Unproductive members of the community will no longer be allowed to share in the products of the community.' At first sight this sounds fair, but it quite likely means, in politicians' language, that old and sick people have lost their ration cards. 'Un-American Activities.' Anything that I, being an American with a particular set of rather inflexible opinions, disapprove o£ 'Reactionary intrigues.' Any relationships that I, being a Communist leader, disapprove of. 'Un-British.' Anything that I, being British, think ought not to be done in Britain. An amusing story is told of a magistrate, who must have been rather pompous. A Dutch sailor was brought before him, charged with being drunk and disorderly. In the course of the trial evidence was given that the Dutch sailor had bitten someone's ear. The magistrate rebuked him with the words, 'It is very un-British to bite people.' Perhaps he thought that in Holland it was regarded as quite usual behaviour, but the captain of the ship gently intervened: 'It is very nn-Dutch, too, my lord.' 'We hold a firm and serene conviction of the rightness of , our cause. We have never stopped to think if we might possibly have made a mistake or if some concession or compromise might be wise and ethical. 'The Ruritanian Way of Life.' The general structure of conventions, social order, pattern of activities, that I, being a Ruritanian, have become accustomed to in my particular class and group and generation. 119

Whitewash Language 'Our gallant soldiers have wiped out a few nests of insur gents. ' The Army has killed, with overwhelming odds on its side, some people who disagree with the present Government and who were resorting to force to attack it. 'Our troops have retired to consolidate suitable positions further behind the lines.' Our troops have had to retreat but hope they will have to retreat no further. 'Our troops have made a planned evacuation of Katzenjammerburg and are moving as fast as possible southwards.' The retreat has turned into a rout and the Army is running away as fast as it can. 'Katzenjammerburg has been liberated and has welcomed the democratic forces with enthusiasm.' Katzenjammerburg is full of scattered loot, broken windows, outraged women and terror-stricken children. A few flags are hanging from a few windows. 'The enemy will not escape the just wrath of our people.' We think we are succeeding in working up mob hysteria. 'The Minister is giving the matter his consideration.' Nothing is being done. 'The Minister is giving the matter his active consideration.' He might do something if you wrote another half-dozen nagging letters. 'We all passionately desire peace, but--' We hope the other side will be blamed if war breaks out. Yes; we all know the kind of language. At first it often sounds very reasonable; later it gets a laugh and is parodied; and a time comes when it makes people cynical and defeatist. There is in political life a great deal of sincere service and selfdedication, a great deal of idealism, and even a fair amount of practical sense. It is, however, a field of activity in which hypocrisy is particularly prevalent and words are often mis120

Whitewash Language used abominably. The habit of seriously examining the meaning of words is thus a protection to the public-from being swindled, from suicidal mob excitement, from allowing injustice to be done in its name, from those obsessive hatreds which in the end are as self-destructive as they are otherdestructive. The habit is also a protection for the politicians themselves. There is, after all, no intrinsic reason why trying to administer the large-scale affairs of a country-a job that somebody has to do-should involve the alternatives of headily total success and an undignified end in a prison yard or dangling from a lamp-post. . . .

121

XII. LANGUAGE OF QUARRELS

A

of communication may cause a quarrel. For example, two schoolgirls were in the cloakroom putting their outdoor clothes on. Joan's hat fell to the floor, and for some reason bounced several times. 'Goodness!' Evelyn exclaimed, 'the thing must be alive!' This remark, meant merely as a comment on the unusual liveliness of an inanimate object, was taken by Joan as implying that her hat was infested with lice, and she made a scene. We can even mishear a remark by failing to catch the tone of voice, and suppose ourselves insulted when we are merely being teased, or believe a word of praise to be sarcastic when it is sincere. Often one person hears something said about an absent friend, reports it in a slightly garbled form from genuine misunderstanding, and makes a great deal of mischie£ A sincere and sensible woman said to another, in the course of conversation, 'I suppose I am rather strongly sexed.' This woman had great self-control and the observation was merely on the temperament she knew herself to have. Her hearer, not accustomed to this particular kind of frankness, went to a third party, and with the best intentions, in genuine anxiety for a younger person's happiness, told the third person she feared for the safety of the youngster, who was obviously about to do something very wrong and reckless ! She was confusing a desire with the inability to handle it. People who are trying to avoid quarrels are, therefore, wise AILURE

122

Language of Quarrels to be careful with words and to try to ensure that they say what they mean and understand clearly what others mean. What happens to language when a real quarrel has already begun? Truth, it has been said, is the first casualty in war; and it is the first casualty in any quarrel in which speech plays a part. For language in a quarrel rapidly degenerates, and that not merely in the sense that he who was at frrst tiresome may become pig-headed, then a swine and finally a - - swine. Our choice of invective and abuse will always depend mostly on our upbringing and habits; among some people a string of swear-words is no more than a recognized ornament to speech, and in other groups any use of such words is an unpardonable insult. A worse degeneration of language is that in a quarrel it gradually ceases to communicate. Language in a quarrel may be as useless and even as harmful as cancer cells are to the body; instead of fulfilling its proper function of communication, it erects new barriers. The analysis of a typical quarrel-conversation may be unpleasant, but it may also be illuminating. The quarrel arises because A has carelessly dropped some cigarette-ash on B's carpet while calling.on B to discuss the date for a committee meeting. B. Do you mind using the ash-tray, please? (A reasonable request, couched in moderate terms, and not offensive provided that the tone of voice is friendly.) A. Oh, sorry; didn't see it. (A reasonable reply which is the end of the matter.) But suppose A is in a bad temper; then the conversation n1ay go sotnething like this. A. Oh, all right, all right, don't fuss. B. Here it is. Sorry, but it is my carpet, you know. A. Never known such aJusspot as you! {A, perhaps slightly conscious of being in the wrong in this 123

Language of Quarrels small matter, and not wanting to feel in the wrong, tries to defend herself by an accusation against B. It might still not be the beginning of a quarrel, since Jusspot is a good-humoured word, provided that the tone is not markedly hostile. But by now B too may be feeling a little annoyed.) B. ~Vell, I'd rather be a fusspot than slovenly. A. Oh, so I'm slovenly, am I? (B never said so; but A can argue that it was implied.)

B. No, no. Don't be silly. Of course not.

(The irritable tone is justifiable, but not wise, for A is now in a mood to take words in their most unfriendly sense, and no one likes to be called silly.) A. Oh, so now I am silly as well, am I? B. No, but I shall begin to think you are if you take offence at

every little thing.

(At this point words still mean something and either A or B can prevent a serious quarrel; B by an affectionate or consoling remark instead of this rather self-righteous tone, A by a simple, brief apology, after which the subject should be changed. As A is the person most in the wrong, it is likely to be B who will have to take the initiative in making peace; such, rather unfairly, is life. This point, at which both persons are annoyed, is the danger point; and if the situation is not saved now the language will rapidly degenerate.) A. You think I'm bad-tempered! And do you know why I am

bad-tempered?

{One of those questions to which a tactful answer is almost impossible.)

Because I've had hardly any sleep for three nights, trying to work things out about this wretched committee! I'm working myself to death for the club and nobody appreciates me in the least! You've no idea what a burden I am carrying!

(Self-pitying language with obvious exaggerations.) 124

Language of Quarrels B. Well,

if the

work is too much for you, why haven't you

said so? {A fair comment, but A does not want to be taken literally. B's question, on the face of it reasonable, sounds like further criticism.) A. And who says the work is too much for me? No one has ever complained that it wasn't properly done! {Unnecessary self-defence.) And I couldn't ever say anything to you, however tired I was! No one can talk to you! You are too much tvrapped up in yourself! (Accusation again as a means of defence. By now B too is losing patience at the repeated unfairness; and as soon as she loses patience she will become, not only rather self-righteous, but herself downright unfair.) B. Look here, there are plenty of people in the club tvho can talk to me; look at Ethel and Flora and how they tell me all their troubles! (Adding something inflammatory to what may be a true defence): I'd like to see Flora telling you anything! A. I don't care to make a cheap play for popularity with a bunch of ttJoonstruck adolescents. (Denigratory language.) B. You wouldn't make yourselfpopular with adolescents, either; if there's one thing they won't stand, it is whining insincerity. (Oh dear! Now B too has reached a point of no return. That cannot be withdrawn except in total apology.) A. So now I know what you really think of me! {No, she does not. What she now knows is that B, under some provocation, is also liable to say unkind, foolish and largely meaningless things.) B. Well, if the cap fits, wear it. {A cliche that is no help; and for some reason such cliches generally sound smug.) A. You cat! 125

Language of Quarrels B. You silly woman! (Language has now degenerated into mere impolite expressions; communication has come to an end after decreasing for some time.) A. Oh, shut up! I'm sick of you! B. Don't shout! I've come to the end of my patience with you! A. That's the end of that! I' 11 send you my resignation in the

morning! (And at this point either A goes out slamming the door, or A and B, having come to an end of meaningful speech, begin to scream, swear or weep, depending on their background and habits.) I suppose no one old enough to read this book has escaped being, at some time, in one of these painful, ridiculous and almost always unnecessary situations. Not all quarrels can be avoided by ordinary, fallible and ordinarily selfish human beings; some people are extremely provocative, some differences of opinion extremely difficult to settle amicably; but well-intentioned people wish to avoid quarrels as far as possible. One check we can use is a certain perceptiveness about our own words; when we fmd that we are beginning to use words because they somehow give satisfaction in a mood of anger, and no longer because we genuinely wish to communicate something, we can try to pull ourselves up; we are about to plunge into a quarrel. Often the mere question, spoken in a mild and kindly tone, 'What is it we are really arguing about?' may be enough to tum language back from primitive noises into the channel of real communication. Soothing words are often as semantically empty as angry words; but they are tnore useful socially. Another check we can use is a perceptiveness about our neighbour's words; when we hear words slipping into this kind of meaninglessness, it is time to say something soothing and to try to restore communication. The last stage of a quarrel, 126

Language of Quarrels when words have degenerated progressively away from communication, is a series of ugly animal howls; and these may easily turn into the opposite of human communication: blows. A poignant and impressive example of the restoration of communication was seen some years ago at a conference in which Esperanto was the language used. Esperanto is available for use to anyone who finds it useful; but it is historically associated, for obvious reasons, with a general ideal of human understanding, brotherhood and the peaceful settlement of disputes, so that most active users of Esperanto are in some sense internationalists. At this conference some delegates from Western Europe and some delegates from Eastern Europe had a serious difference of opinion. As the argument developed, with strong feelings on both sides and a good deal of misinformation on both sides, words began to degenerate from instruments of communication to instruments of selfrighteousness, then to the cliches of the propagandists, then to vulgar angry abuse and shouts. It seemed that in a few seconds the delegates would actually come to blows; the noise and facial expressions were already ugly.... A plucky delegate who fortunately had a loud voice began to sing the international anthem of the Esperanto Movement. The shouting stopped just long enough for people to notice what they were doing ... they realized that the whole point of learning Esperanto had been to understand one another, and that they were paradoxically quarrelling in it . . . words which had once seemed rather hackneyed were heard afresh with their full meaning . . . and the delegates, ashamed and calmed, began to join in the song. A few minutes later the discussion was resumed in a different spirit, using words to try to seek out the truth. It might be said, then, that a serious, violent quarrel is temporarily the death of language; words lose their meaning, 127

Language of Quarrels largely because the attempt at communication has ceased; in the last verbal stages of a quarrel we are generally not talking to the other person, but saying things that give satisfaction to ourselves. There is a less brutal kind of quarrel in which no settlement is possible until the meaning of words has been considered. Two people mean to communicate; the failure is not from temper but from real misunderstanding. This is very common in dispute between two people of different generations. Defmition is one of the keys to achievement. Here are some time-saving and nrisery-saving questions, for example: 'What is it exactly that you want?' 'Could you give me an idea of how much you expect 0

f

•••.?'

'Can we make an exchange, you do this and I do that?' 'Could we discuss a concrete example?' 'Can you tell me how?' 'If that isn't possible, what else can be done?' 'Would you consider settling for . . . ?' 'I am quite willing to do this .... Will that do for a start?' 'Shall we see how it works?' 'Can we agree first on . . . ?' For instance, a mother tells her daughter she. wants her to behave properly. The daughter is likely to turn round at this and say her mother is fussy and old-fashioned. This is the way to start a quarrel. However, while an unreasonable mother who wants to be a dictator will be angry if her daughter asks, 'What do you mean by behaving properly?' a mother who wants to educate her daughter will be willing to give some sort of example or defmition. The daughter is then able to take part in a meaningful discussion and the two may reach some practical conclusions; but they will almost certainly have to include compromises on the way. For instance, Mother may then say that proper behaviour includes not wear128

Language of Quarrels ing slacks, Daughter suggest that slacks are surely all right in the house and garden, Mother concede this but remain firm on wearing them for visiting other homes; Mother may say that proper behaviour forbids the use of lipstick, Daughter protest that all girls today wear it and that therefore it is embarrassingly unconventional not to do so, and Mother see her point, but suggest that too much lipstick looks jammy and that the colour should be suitable to the clothes and con1plexion of the wearer; Mother may say that it is not proper to walk along the street arm-in-arm with others so as to occupy the whole pavement and send other people into the gutter, and Daughter, on reflecting how inconvenient this behaviour is for others, may agree. And so two wellintentioned people, sitting by the fire and talking in a quiet manner, tnay learn much that is helpful, become better friends and practise the precious art of compromise instead of having a painful and useless quarrel. The attempt at defining proper behaviour instead of taking it for granted that Daughter knows what it means and wilfully disobeys, that Mother is nothing but a prim and proper old fuss-box to use such silly words, immediately leads to practical, meaningful discussion. One of the tragedies of political life today is that so much discussion goes on without defmition. Sometimes I find myself aching to hear of some statesman saying 'Gentlemen, would it not be a good idea to decide what we are talking about?' Nearly all political speeches of a controversial nature have in them some element of quarrel-language, of meaningless language functioning not as communication, but as selfapplause or as abuse of the other side. Real wants are disguised in clouds of language suggestive of shining virtue and disinterestedness; sometimes the man in the street (who is also likely to become the man in the hole in the street when bombs begin to fall) feels that it might be better if the politi129

Language of Quarrels cians said what they wanted-though of course this would not in itself solve the very complex problems of present-day politics. Co-operation between groups of people whose immediate interests differ is never going to be easy; words will not in themselves abolish selfishness or produce forethought; but co-operation and reasonableness are aided whenever human beings show determination to use words as instruments of communication, and firmly reject the degenerate use of words.

130

XIII. LYING TO OURSELVES

T

are very few people who do not now and then tell a deliberate lie. What might be called a full, deliberate lie is not merely the careless use of words, but a statement known to be objectively untrue. 'It was Peter who stole the buns!' says Jim, who stole them himsel£ 'I have not seen Anne for six weeks!' says Philip, who has been meeting her regularly behind the bam. 'This is my last territorial claim in Europe!' said Hider, shortly before making a further claim. If, however, we consider the nun1ber of things said by the average person ·every day, the number of full, deliberate lies is a small percentage of our whole communication; much social life is based on the assumption that most people tell the truth most of the time, at least in the sense of not making deliberate total mis-statements. Modem commerce, for instance, has to be based on a large measure of mutual trust. So have personal relations; so has the appointment of people to jobs; so, indeed, has ordinary conversation. We recognize the difference between the person who occasionally, under some kind of pressure, tells a lie, and the person whose words are never to be trusted, the 'pathological liar'. Most of us regard direct lies as 'wrong' and would differ only in assessing those situations in which some would say a lie is excusable {someone whose business it clearly is not asks me an impertinent question about my private life) or even a positive duty (a homicidal maniac runs into the room with a knife in his HERE

131

Lying to Ourselves hand and asks me where my mother is) or accepted as a social convention ('Mrs. Warwick is not at home'). To me one of the great evils of a totalitarian or even an over-conventional society is that it tends to force wellmeaning but unheroic people to tell more lies than they otherwise would, from fear. Where there is constant fear of, and expectation of, unjust and unreasonable treatment, most people will get into the habit of lying; it is a needful part of self-protection; and the habit may soon extend beyond what is necessary. Even in so unimportant a tyranny as a highly authoritarian and punishment-ridden school or family this mechanism may be seen at work. However, most people would agree that telling deliberate lies is wrong, except perhaps in certain special situations where more harm will be done by telling the truth. Even the most truthful people probably tell a good many more lies that might be regarded as semantic lies; their use of words contains some measure of falsehood, more or less deliberate. 'Do you think it is right to rag Mr. Dickenson? Inkpellets flying, a white mouse in his desk, silly sentences in your exercises-do you realize how badly you are behaving?' 'Oh,' the schoolboy replies, 'Mr. Dickenson doesn't mind; he is always a good sport.' To this boy 'a good sport' is someone who will meekly put up with stupid insults and time-wasting; this praise is an excuse for continuing to treat the master in a very unsporting fashion. 'I can't see you now; I am too busy.' I say to one of my pupils. This may be true; but it is more likely that I do not want to take the trouble to replan my replannable work so that I can see her. 'I don't think that bathing costume is decent, showing all 132

Lying to Ourselves that much of herself!' says Mrs. Lumpe indignantly; what she really means is, 'I wish she would not show her lovely figure, because I know I should look like a piece of pink pork in a costume like that.' And so on. The average human being tells a direct lie relatively rarely and is conscious of this as wrong. He or she tells semantic lies more often, probably does not regard them as wrong and is often barely conscious of their insincerity. I should say that all of us lie to ourselves, in part, almost as a continuous process. Not all of this is verbal; we do not think continuously in words; but only when we put our thoughts into words can we make any examination of their truthfulness. The process of becoming conscious is a difficult one; many people do not even wish to be conscious; and modern psychology tells us that a large part of what goes on in our minds is normally inaccessible. The reactions of many people to modern psychological theory, especially the theories of Sigmund Freud, provide interesting and pathetic examples of our taste for lying to ourselves. Freud laid great stress on the importance of our sexual desires and experiences and curiosities in shaping our lives. To someone who is trying to be honest, such theories are, like any other theories, something to be examined, investigated, accepted or rejected, or-as is usually the best reaction to most theories-partly accepted as containing some useful truths or working hypotheses, but not accepted unconditionally as a new revelation, in that all scientific 'truth' is provisional and tnay be outdated by further research. Many people, however, found it distasteful to suppose that their sexual feelings played as great a part in their livt~ as Freud said. Not many such people told themselves, 'I do not want to think about such theories; they cn1barrass n1e and make me feel I am just an animal, not dignified. I feel they are not pleasant ideas, and it willtnake me unhappy to think about 133 K

Lying to Ourselves them, so I will not think about them; the theories I have about life seem to work all right for me and I do not need new ideas.' While this is perhaps rather an ostrich-like attitude, it would be fundamentally honest; and there may indeed be good sense, for a nervous or timid person, in refusing to think about subjects that cause anxiety. But the people who refused to pay any attention to Freud's startling theories often commended themselves for their 'virtue' in fmding them distasteful; this was really an excuse made to themselves for not having the courage to face what they found unpalatable. 'He must have been a dirty old man to have had such filthy ideas. I am a good woman and I am not going to soil my mind with such nastiness. The world is going to the dogs; when I was a girl no one would even have dared to say such things; but I shall keep my mind pure, even if everyone else swallows all this modern wickedness.' The woman who says this is almost certainly not deliberately lying. She is to some extent lying semanticallyscientific research described scientifically may err even wildly, but it cannot be 'dirty'; and she is obviously using a 'good woman' and 'pure' in an extremely limited and inadequate sense; but she is not doing this consciously. What she is doing is lying to herself; and this we all do every day. I tell myself I am honest; and I sincerely believe, when I say this, that I am: I never steal, I do not cheat the Customs or the Income Tax; I pay my debts promptly; I have never sold anything under false pretences . . . and then I stop to think a little harder. I am lying to myself; I am not honest. Yesterday I was five minutes late for my work, for which I am being paid on the assumption that it begins at a given time. This morning I took rather more than my fair share of butter at breakfast. And I said last night that I had no time 134

Lying to Ourselves to deal with a problem and would see to it tomorrow, when the truth was that I badly wanted to write some private letters. . . . Well then, I may be a human being trying hard, but I cannot call myself honest. John does not care for cultural activities-music, painting, reading, serious discussions. He is entirely at liberty not to; there are plenty of other ways of enjoying life and of being useful. But does he tell hin1self: 'I haven't much brains and I am mentally rather lazy, so I am not intending to bother with the arts'? No! He is far more apt to say, 'I am a real man, I am not a sissy; I an1 not one of those degenerate eggheads, those drawing-room intellectuals.' The people I know with the most insatiable appetite for culture happen to include an ex-guerilla fighter and freedomradio broadcaster, a professional explorer and a former bomber pilot; but the legend that cultured people are soft and unmanly is a very useful excuse for the mentally lazy and the insensitive. Katherine is lacking in mental adventurousness, is afraid of new ideas and new activities, and is afflicted with the kind of intolerance that arises from timidity. On her lips we constantly hear such words as wholesome, healthy, normal, nice, decent, the done thing, not quite the thing, respectable. She disguises her fears as virtues and learns to be proud of her limitations. On the other hand, Roger, with no sense of responsibility, who goes through life causing unhappiness to others and leaving others to clear up his messes is quite likely to praise himself as adventurous, unconventional, enterprising, daring, uninhibited and so forth; and, like Katherine, he does not stop to think if a few careful definitions might not lead him to sotnc uncomfortable opinions about life and the sel£ We do not usually lie to ourselves wilfully; but we all do it tmintentionally. People who think they are in love but whose 'love' is really shallow and immature sometimes deceive on 135

Lying to Ourselves purpose; but more often they deceive themselves first. How easily we can convince ourselves that our anger is righteous anger ! and how easily we can convince ourselves that we are broken-hearted when we are merely angry and disappointed! How easily we can believe that we never had a chance when what has really prevented our success is that we never took one ! While there are certainly some people who are deliberately dishonest, insincere, hypocritical, the vast majority of us are probably the self-deceived, and most of the time deceive others only as a by-product. What are the reasons for this? One reason is certainly the desire of a healthy person to think fairly well of himself; and the unhealthy mind that is determined to think ill of itself is just as self-deceiving with the disadvantage that it is miserable too. This might be called the moral reason: we have a weakness of pride, a rather pathetic little vanity of wanting to have good motives for all we do; and we may add to this, often, a desire that if we cannot be good we may at least live rather dramatically. I have myself probably deceived myself more in the process of trying to think myself interesting than in trying to think myself good! Then there is what might be called an intellectual reason: the simple and inevitable fact that none of us are sufficiently intelligent to understand even ourselves completely, let alone other people or the general meaning of human life. Very little can be done about this. The great Greek exhortation 'Know Thyself' is excellent advice, is the foundation of psychology and the beginning of charity; but it is impossible ever to obey it adequately. Third, there is what might be called the verbal reason. There are enough words for all the kinds of grass in the world, all the cloud formations, all the possible chemical compounds. There are not, and I do not see how there can ever be, enough words to differentiate human emotions and experi136

Lying to Ourselves ences. A word like love, faith, mercy, justice, pride, anger, pity, wisdom, glory, or fear has to carry a load of assorted meanings that makes it well-nigh impossible for any two people, at a given moment, to understand it in precisely the same sense. And it is in the realm of emotions, human relations and ethics that we are most apt to fall into self-deception. Only a fool or a deliberate liar says, 'There are six eggs in the basin!' when he can see that there are four pears in it; but in the course of history millions of decent kindly men must have said to millions of women, 'I love you!' and brought them immeasurable unhappiness because the words were understood in two different senses by speaker and hearer; and millions of parents must have said, 'It is all for your own good!' and so appeared to be callous hypocrites, when in fact the total situation was far more complex than anyone could provide words to explain. It seems to me, then, that it is very important to try to use words honestly and to think about what words mean; and that a certain reverent care for meaning will tend to improve human relationships on both a small and a large scale; but also that full communication is impossible and that, because of human limitations, we must to some extent resign ourselves to life being a tissue of lies, not deliberate lies, but a mixture of semantic and psychological confusion. What is the inference from that last fact? Charity. The study of meaning may at first give us a sense that we are very clever, with a lawyer-like subtlety; we enjoy our new perceptions and we think for a time that we have the key to truth. Then, as we realize how much all communication is limited, even within our own minds, we come to the painful knowledge that at best we know very little. The inferen'.:e from that need not be any kind of despair; it can be a generous, sympathetic tolerance, arising from the awareness that every human being is trapped 137

Lying to Ourselves in the satne ignorance, the same inadequacy. Truth, it n1ight be said, is the inspirer and the instrument of justice; but the awareness of our own fallibility is the foundation of mercy.

PART III

Literary Semantics

XIV. RHETORIC

M

of this book is devoted to what might be called the semantics of everyday life; and if we are thinking about the meaning of words we can never go far away from human beings, human problems and human relationships, since the purpos-e of words is to communicate and we do not communicate with the empty air. Moreover, semantics would be a mere pedantic exercise if human relationships were not involved, and on the contrary it is a study of great importance to human sincerity and happiness. However, the last two chapters will deal superficially with the meaning of words as used in poetry and rhetoric; in the verbal arts, which are in one sense operating at one remove from 'real life' and in !mother sense, being created by human beings, are still as much a part of real life as eggs and bacon. Today rhetoric is very nearly a dirty word. Rhetorical language is almost a synonym for exaggerated and insincere language. It is not fashionable to like 'purple patches', and today much writing that was once regarded as beautiful is thought of as high-flown, even comical. Have we gone too far now in the direction of distrusting rhetoric? and are we not perhaps falling into a new sort of cliche, the 'dirty-word' use of rhetorical as a term dismissing all emotional, all metaphorical. all picturesque language as insincere? I have tried to point out that there is a great difference between sentiment and sentimentality, that emotion is a necessary and indeed a OST

141

Rhetoric desirable part of human life; a distinction should also be n1adc between legitimate and illegitimate rhetoric. If we remember that the purpose of language is communication, we tnay be able to arrive at some concept of the function of rhetoric. Rhetorical language is not trying to communicate in the san1e manner; but it may still be communicating; and the best rhetorical language may be conlmunicating n1ore intensely, more, so far as we can judge, truthfully, than any literal, objective statements. 'The toast is on fire!' I exclaim. Everyone who understands the language knows what I mean. The bread I was roasting is blackening and flaming. There is no doubt. This may possibly be a lie, but it cannot be an an1biguous statement. Another day I say something rhetorical instead of literal: 'I am on fire with love!' As a joke, someone tnay reply, 'Fetch a bucket of water!' However, no one thinks for one second that I mean that I am blackening and flaming. This hackneyed metaphor is an attempt at conveying a particular kind of rather obsessive emotion which is in part a physical experience. Rhetoric is often needed, because language is not adequate for expressing emotion. Metaphor, sinille, association, personification, analogy can indeed mislead when used dishonestly, but they can also be used honestly in an atten1pt to communicate. Indeed, one of the criteria of good rhetoric is that it should be sincere; inflated rhetoric is essentially insincere. The soggy pseudo-rhetoric found in some political speeches and some sermons, in the worst kind of family quarrels {'I work my fmgers to the bone for you ... ') and in some of the love scenes of very poor plays and films is an imitation of the rhetoric generated by real emotions. Bad over-flowery writing, too, is an imitation of great rhetorical writing; and some rhetorical writers tend to parody themselves in their less creative moments; for example, D. H.

142

Rhetoric Lawrence can at times convey impressions of intense and complicated emotion with electric vitality; but he can also sometimes imitate himself and become inflated. It is not 'unnatural' to use figurative language. I have heard a barely literate but fundamentally sensitive man say, gazing with awe at a tiger in a cage, 'Eee, but its fur looks like hot coals behind bars!' and a quite unliterary woman suddenly refer with delight to the 'meringues on the hawthorn trees' (the masses of blosson1). Some time ago I teased a very sincere and unaffected woman whose husband, though a most kind and humane man, had a rather formidable appearance: 'I don't wonder you collect stones; I suppose you were looking for a lump of granite when you found your husband?' 'No,' she replied at once, 'you forgot that I collect pieces of pure crystal too.' Strong emotions often bring surprisingly rhetorical language out of people with no literary pretensions; the struggle to say something during courtship that is adequate to express feelings for which no human language can have words often produces metaphors and hyperboles; very angry people also often come out with rhetorical expressions; and so do people in great distress. Proverbs, which are not literary but folk art, bear witness to the naturalness of such language: 'There's never smoke without fire', 'All that glitters is not gold', 'The cat would eat fish and not wet her feet' are not useful contributions to knowledge if we take them literally. Affected and tiresome rhetoric occurs in conjunction with sentimentality or in the imitation of genuine rhetorical speech. Wordsworth explains this admirably in an appendix to the Preface to the Lyrical Ballads: 'The earliest poets of all nations generally wrote from passion excited by real events; they wrote naturally, and as men : feeling powerfully as they did, their language was daring, and figurative. In succeeding times, Poets, and Men ambitious of the fame of Poets, perceiving the influence of 143

Rhetoric such language, and desirous of producing the same effect without being animated by the same passion, set themselves to a mechanical adoption of these figures of speech, and made use of them, sometimes with propriety, but much more frequently applied them to feelings and thoughts with which they had no natural connection whatsoever. A language was thus insensibly produced, differing materially from the real language of men in any situation. The Reader or Hearer of this distorted language found himself in a perturbed and unusual state of mind: when affected by the genuine language of passion he had been in a perturbed and unusual state of mind also: in both cases he was willing that his common judgment and understanding should be laid asleep, and he had no instinctive and infallible perception of the true to make him reject the false; the one served as a passport for the other. The emotion was in both cases delightful, and no wonder if he confounded the one with the other, and believed them both to be produced by the same, or similar causes. ' There are occasions when it is perfectly permissible for a politician to say something like 'a crime which has shocked the conscience of every Christian community in Europe'. What kind of crime? Perhaps Hitler's persecutions of the Christian Churches; or the action of Christian priests who swear oaths of allegiance to dictator regimes; or the fact that some priests had been using the threat of excommunication to influence voting in an election; . something large, clearly wrong or at least violently controversial; comprehensible to everyone and of general importance. When F. E. Smith said that the Welsh Disestablishn1ent Bill-sotnething essentially local, highly technical and not a black-and-white n1oral question, was 'a Bill which has shocked the conscience of every Christian community in Europe', G. K. Chesterton wrote his famous 'Chuck it, Smith!' satirical poen1, pointing 144

Rhetoric out that the language was ridiculously immoderate in relation to the circumstances:

'In the mountain hamlets clothing

Peaks beyond Caucasian pales, Where Establishment means nothing And they never heard of Wales, Do they read it all in Hansard With a crib to read it with"Welsh Tithes: Dr. Clifford Answered." Really, Smith?'

Chesterton himself was one of the most rhetorical writers imaginable; his objection was not to F. E. Smith's rhetoric but to its disproportion to the subjec~, in other words, its sentimentality. There is inevitably something comic in an egg-bald man's plea that his ungrateful children will bring down his grey hairs in sorrow to the grave, or a fat woman's declaration that she is wasting away for love, though both expressions nught be meaningful and touching in another context. When someone says 'I wish I was dead!' he very seldom means it, though on rare occasions someone does say this with full meaning and it is then a very terrible thing to hear. Similarly 'I've got no one but you in all the world!' is very rarely true ... but just occasionally it may be. The functions of rhetoric seem to me to be as follows: 1. To communicate emotion and other experiences that cannot be communicated by means of factual, literal language. 2. To induce emotion, perhaps emotion leading to action, in other people. 3· To adorn speech and writing. We must accept the communication of emotion as a reasonable human desire; and if we also accept, as I do, that our greatest satisfactions are found in our emotional lives, it is 145

Rhetoric very important for our happiness and evolution that we should talk about emotions. Legitimate rhetoric in this field, then, seeks really to communicate emotion. It is illegitimate when it falls into sentimentality, which has already been defmed. Emotion may be communicated by hyperbole: Were you the earth, dear love, and I the skies, My love should shine on you like to the sun, And look upon you with ten thousand eyes, Till heav'n waxed blind and till the world were dun. JOSHUA SYLVESTER, 1563-16I8 (from a sonnet) and it may be said that the hyperbole is meant to convince us of the violence of the emotion; by metaphor: A single violet transplant, The strength, the colour, and the size, (All which before was poor and scant) Redoubles still, and multiplies. When love with one another so lnterinanimates two souls, That abler soul, which thence doth flow, Defects of loneliness controls. JOHN DONNE, 1573-163 I: The Ecstasy

where an analogy is used to try to convey to us a psychological experience for which we have no adequate vocabulary: by drastic, adventurous and unexpected language: I am soft sift In an hourglass-at the wall Fast, but mined with a motion, a drift, And it crowds and combs to the fall; GERARD MANLEY HOPKINS, 1844-1889: The Wreck of the Deutschland 146

Rhetoric where the function of language is not only to convey emotion by analogy, but also, as it were, to wake us up, to starde us into noticing by some unusual stimulus. People with little understanding of d1e use of poetry often complain that a poem is 'difficult'. I have heard it brought more than once as a serious complaint against a poem: 'You have to read it several times before you understand it' ! There is an obscurity which is caused by careless writing and insufficient concern for communication; but most 'difficult' poetry is 'difficult' precisely because the poet is trying to do something very difficult indeed, to convey a complicated and delicate experience to other people. If an adequate vocabulary for emotion and the inner life existed, there might be no 'difficulty' poetry. As it is, the remarkable thing is that some poetry and rhetorical prose do seem to come fairly near to conveying a satisfactory impression of emotional phenomena. It is very difficult to assess the sincerity of an emotional statement; most of us do so by methods which we could not ourselves fully explain, which are often more intuitive than intellectual; they depend very much on that indefmable quality we often call the tone ·of a speech, poem, personal conversation and so on.. One of the biggest difficulties here is that what is a mark of insincerity in one person may be perfecdy authentic in another. For example, in general it is n1ore likely that rhetorical language which is an imitation of someone else's language will be insincere, than that something original will be; the passage fron1 Wordsworth quoted above explains this admirably; but, on the other hand, people of relatively little verbal ingenuity are naturally apt to tum to expressions they have previously heard to express what they themselves cannot fmd words for; it is natural to quote works with which we are familiar, such as the Bible; and the expressions of strong and usual human emotions may tend to coincide without imitation being involved. For example, 147

Rhetoric when some years ago my father died, and my mother said to me (an only child), 'Now we have only got each other!' this might well have sounded to an outsider like something sentimental, imitative and out of a book; but since such a thing has often been in some sense true it will also be often said; and my mother is an exceptionally sincere person. It was the same person who, wishing to appeal to a Government department a little later for some concession, looked up in the process of wording a letter to say, 'Can I say, My husband has recently died in tragic circumstances? No, I don't suppose I can; if in the middle of a world war he died a natural and painless death, I have no business to use the word tragic, have I?' This struck me at the time as something like heroic semantic honesty. Moreover, some people genuinely feel much more strongly than others about various things. There are people who do not ever 'fall in love' and who, if they marry, marry for companionship or convenience; there are people who are surprisingly philosophical about misfortune; there are people who hardly react at all to slanders and insults; and so on. To people of naturally unemotional temperaments many expressions of strong etnotion may well sound like a great fuss over nothing; the most genuine, passionate and urgent real emotions may sound like sentimentality. In England, a country in which the violent expression of emotion tends to be disapproved of, with an excellent effect on our formal manners and political stability and probably often a bad effect on the psychological health of individuals (and possibly one factor in the high proportion of gastric ulcers and chronic catarrh!), we are very apt to suspect insincerity in the more emotional expressive and dramatic manner of a Slav or Southern European. I have certainly not had this suspicion confirmed by my intimate knowledge of many such people; and we should remember that many foreigners suspect us of 148

Rhetoric being much more hypocritical than the average English person really is, because of our tradition of restrained expression. My own view is that we should be slow to suspect real insincerity, and should judge sincerity by the relation of people's deeds to their words, rather than by the nature of the words themselves. If a woman says, 'My religion is everything to me; I could not breathe without it!' she may or may not be sincere; but if she then puts a button in the collection and spreads sto:.-ies about the minister that are known to be untrue, she must be insincere. If a man says to a woman, 'I love you; I adore you; I want to protect you against all the rough ways of the world and look after you as my treasure!' he may be absolutely sincere; but if he then humiliates the woman in front of others and stands wondering what to do when another man insults her, he must be insincere. Even then we must remember that insincerity is by no means always the same as deliberate deceit, that most of us lie to ourselves before we lie to others. Being honest is difficult, even strenuous, not only morally but intellectually. What about the usc of rhetorical methods to induce emotion in other people? Here words are used with the hope that their forcefulness or unusual quality will startle: 'Are you dogs, or worms, or little bits of slime, or grown men?' (A rhetorical question assuming that the hearers will wish to reply, 'Grown men!') 'I do not have a soul!' said the priest from the pulpit, and shocked his orthodox hearers. 'I have a body'-and shocked them more. 'I am a soul.' (Rhetorical use of an ambiguity in the word have to remind his congregation that to a Christian the soul is more important than the body and the body merely a vehicle or vessel.) Or words may be used in the hope that the rich structure 149

Rhetoric of association around them will stimulate the emotions of the hearers. 'I implore you to think again before you pass this law. Do not think only of your own prestige, of the prestige of this parliament. Think also of bewildered women and of the children who may be made homeless by an adn1inistrative formula. I have explained how injustice might be done; think of your own wives; how would you feel if they were to suffer the humiliation this law might bring them? Think of your own children; think of the horror that would sicken you if you knew your own children were to be made homeless ! Those innocent children you love, with their bright trusting eyes and the light on their hair. . . .' It may be supposed that earlier in the speech the speaker's appeal was objective, explanatory and factual; but now he is trying to arouse emotion, by appealing to personal experience and so !nake things more real. Is this legitimate? I think it depends entirely on the context and motive. There is honest rhetoric, which aims at skilfully applying psychologicalverbal techniques in order to arouse emotions in others, without deliberate deceit and for a good motive; and there is dishonest rhetoric, which uses the same techniques deceitfully, for a discreditable reason . . . and cold-bloodedly. A great and good orator is sincere; it is from his own emotions that he produces the words with which he can arouse emotion in others. A demagogue uses words skilfully, but this is the use of a formula, a deliberate, cunning working upon other people's feeling. If \Ve read the wartime speeches of Sir Winston Churchill, or the most famous speeches of Edn1und Burke or of John Bright, we may fmd that we disagree with many of the opinions expressed; but we do not, I think, fmd outselves muttering 'Cant!' The opinions are such as we feel a rational and well-meaning person might hold; the rhetorical dc.vices 150

Rhetoric make the expression of them more forceful. The brief and very impressive Gettysburg Oration of Abraham Lincoln is a superb piece of rhetoric; but it rings true; Lincoln had a reputation for sincerity that few politicians have enjoyed before or since. The emotion itself seems to be generating the rhetoric that moves us. Was Hitler sincere? Perhaps he deceived himself before he deceived his people and wrecked a world; but some of his lies were so obvious, and he himself admitted so openly to the deliberate technique of 'the big enough lie', that we may safely assume at least some insincerity; was his rhetoric good in the aesthetic sense? So far as I have been able to find out, it seems not. His power over an audience seems to have depended relatively little on eloquent words, but to have been a matter of endless reiteration and a rather hysterical manner -hysteria is notoriously very catching. It is, of course, not possible to say that every orator of whose principles or aims we do not approve is insincere; this is merely making insincere another 'dirty word'. For example, today some political orators, especially in the Middle East, try to propagate an exaggerated nationalisn1 that is quite Wlsuitable to their epoch and will not, in the long run, be the most beneficial attitude for their peoples to take; but most of these fanatical nationalists are probably sincere. Insincere rhetoric occurs only when orators try to induce in other people emotions that they do not themselves feel, largely by pretending to feel them. Fortunately, emotion is so much the driving power of art that the best orators artistically may generally be expected to be sincere, however wrong-headed they may seem. Related to this is the fact that much very bad public speaking includes not only poor verbal expression, but actual muddle-headedness in thought. It is very difficult to arouse emotions in other people unless we both feel them ourselves, and are able to justify them to ourselves. 151

Rhetoric The strictly moral criterion that we n1ay at times apply to rhetoric is a good deal more limiting than the artistic criterioiJ of sincerity; when we are considering whether it is right to make a given speech, we must also consider whether the emotions the orator wishes to arouse are desirable and whether the action to which these emotions may lead is a good action. Few people would object to a passionate, rhetorical speech appealing to an assembly to contribute money to aid refugee children; indeed, even if the speaker was deliberately working his emotion up a little artificially, the cause is so good that most of us would not be unduly anxious about the details of the method. But a man might be absolutely sincere in his passionate, fanatical hatred of, say, the country's president; if he used his passionate oratory to induce a mob to burn down the president's house and trample the president to death, very few people could feel that his sincerity was a justification for his behaviour. These, however, are questions of ethics, not of semantics. Lastly, rhetoric may be used simply to adorn speech and writing. Figures of speech may not be necessary to clarify, or to stimulate emution, but may still be used for the sheer pleasure they give. I sometimes give a humorous lecture in English or Esperanto on some completely unimportant subject such as 'My Well-Planned Kitchen' (the title itself is rhetorical, the figure of speech used being irony!); 'His Mewing Excellency-From the Life of a Siamese Cat'; or some journPy I have made, with emphasis on the funny side. These talks, if they were simply straightforward accounts of the everyday and valueless subjects, would be of no interest at all; the whole point of a humorous lecture is the way the stories are told. Irony, comical exaggeration, climaxes, sudden anti-climaxes, surprises of every kind, puns, odd comparisons and so on, ludicrous understatements, unexpected epithets, are what make a talk of this kind; the I

52

Rhetoric pleasure of the listener is not in the valueless subject-matter, but in the comical language. I have at times been restrained only by sympathy and courtesy from yawning at what might be expected to be a very affecting story of some dramatic, tragic personal experience; I remember once laughing until tears ran from my eyes when a fan1ous orator was doing nothing but describe a nervous old lady eating her meal in a hotel. I once heard someone describe the death of a sow so skilfully that it was impossible not to feel grie£ I have heard a professional lecturer give an account of a minor car accident in such a way that it was a remarkable comedy-drama. Delight in language, as it were for its own sake, is, then, an aspect of rhetoric. This is a direct unadorned statement: 'Percy had a very much larger appetite than the average , man. Here is what is essentially the same statement, made into a piece of frivolous rhetoric: 'To say that Percy was fond of eating would be rather like saying that a lighted match is fond of petrol. Nothing edible could be placed within his reach without being instantly consumed. To him four pork chops with fried potatoes were a snack to stay his stomach between meals; and a full-grown boiling fowl was not enough for two persons. He wished that oranges grew as large as footballs, and found that a melon merely stimulated his appetite without satisfying it. A fullsize family fruit cake was a fortress to be sacked and razed to the ground; and when gooseberry or apple pie was on the table, his piety was exemplary. When the housing problem was at its worst, there was always room inside Percy; when the buses were crowded, Percy always had room for a few more inside. To offer him a sweet was like dropping a stone into a well; the stone disappears but the well still gapes. Percy could not see a pretty girl without thinking that the meat off her would be tasty with apple sauce; to him the 153

Rhetoric clouds were meringues and the copse in the valley was a salad bowl; the sand made him think of ginger cake and the rocks of parkin; the snow seemed to him like the icing on a global cake and we suspected that he regarded the sun itself as a cosmic cooking stove. The only abstraction he could appreciate was Time, since that is said to devour all things, and this at once waked some fellow-feeling in Percy, the respect of a man for a competent senior colleague.' This is not the language of statement; its exaggerations are so obvious as to be free from danger, its details are really repetitions not contributions to our knowledge. It is the language of fun. Any study of the meaning of words which never takes into acconnt that language is son1etimes just fun, or just emotional outlet, is inadequate. An obvious example of this is that when people swear they are very, very rarely conscious of any relationship between the swear-word and its apparent n1eaning. Without going into the details of the more grossly offensive language, we can consider such expressions as this-which an uncle of nline claims really to have heard, 'So I told 'im to sit down for five minutes, for 'e was as white as a bloody sheet!' Most people who allow themselves to swear at all have almost certainly, at some time, consigned something or someone to 'hell' when in fact their feelings towards that thing or person were usually friendly. A foolish and vulgar habit; but perhaps not as bad a habit as sentimentality, if we are to measure our behaviour by its consequences. As a Hungarian friend of mine once put it, a good deal of what we say is giving our mouth an airing, not communication. This is, indeed, probably a necessary function of language; we all know how suffering is relieved by talking the trouble out, and some psychologists today positively encourage their patients to use very violent, emotional language in order to get rid of the harmful emotional tensions that have 154

Rhetoric caused the illness. Trouble arises from the loosely emotional use of language, or from language used in fun, only when the ornamental aspect of language is mistaken for communicative language. If I say to sonteone, 'Really, you are an idiot; look, you have punctuated that sentence so that it n1akes no sense!' and she takes 'idiot' as my giving my mouth an airing, no great harm is done; but if she receives the impression that I genuinely regard her as mentally defective, she may become quite unable to work for me. So usually I prefer, when merely being playfully abusive, to use some term such as 'silly sausage', which, since my pupil is never in any doubt that she is not a sausage, cannot be harmfully misunderstood. It would be a pity to renounce the fun, the gaiety, the ornamental quality, of figures of speech, the delight of rhetoric and its comfort, from a puritanical, Platonic feeling that such language was telling lies. What matters is that we should try to say what we mean in all those contexts in which it matters to human happiness that we should fully mean what we say.

ISS

XV. METAPHOR AND ASSOCIATION

M

is probably the most important and frequent figure of speech in poetry and in rhetorical prose.

ETAPHOR

But true love is a durable fire In the mind ever burning.... ANONYMOUS (sixteenth century) We seek to know the moving of each sphere And the strange cause of th' ebbs and floods of Nile But of that clock within our breasts we bear The subtle motion we forget the while. SIR JOHN DAVIES (1569-1626): Nosce Teipsutn My hasting days fly on with full career, But my late spring no bud or blossom shew' th. JOHN MILTON

(1608-1674)

He gave us this eternal spring Which here enamels everything; ANDREW MARVELL

{1621-1678):

Song of the Emigrants in BernJuda

How the Chimney-sweeper's cry Every black'ning Church appalls; 156

Metaphor and Association And the hapless Soldier's sigh Runs in blood down Palace walls. WILLIAM BLAKE

(1757-1827)

This Sea that bares her bosom to the moon .... WILLIAM WORDSWORTH

(177D-1850)

I fall upon the thorns of life! I bleed! P. B. SHELLEY

(1792-1822): Ode to the West T¥ind

and in n1y breast Spring wakens too, and my regret Becomes an April violet, And buds and blossoms like the rest. ALFRED TENNYSON (1809-1892): ln Memoriatu In paradise, the fruits were ripe, the first n1inute,

and in heaven it is alwaies Autumne, his mercies are ever in their maturitv. JOHN DONNE (1573-16~1): from a sermon ~

Who are you, that you should fret and rage, and bite the chains of Nature? Nothing worse happens to you than does to all Nations who have extensive Empire; and it happens in all the forms into which Empire can be thrown. In large bodies, the circulation of power n1ust be less vigorous at the extremities. EDMUND BURKE (1729--1797): from his speech on conciliation with America Such figures of speech as simile and personification have a very similar purpose and effect. For the purposes of this chapter they can generally be considered as a part of meta157

Metaphor and Association phorical language, which differs so much from literal language that its meaning cannot be regarded as meaning in the same sense. Metaphor is an atten1pt at communicating son1ething, usually but not always something emotional, which literal language cannot communicate, or which metaphorical language will communicate more vividly and intensely. It may seem to be merely ornamental; but in good writing these 'ornaments' generally add something to intensity, if not to meanmg. One thing metaphor is not; it is not untlatural. We are often tempted to talk as if it were; but metaphor is a part of everyday speech, so much so that we often use metaphors without even rc~lizing that we are doing so. It is to be found in the very history of language: zest, for example, originally meant lemon peel used for flavouring; zest is the quality of mind and manner that gives life a piquant, pleasant 'flavour'; stupid comes from a word meaning stunned; we speak of moral obliquity just as we speak coloquially of someone who is not straight; a flagrant offence is philologically a blazing one. The most scrupulous avoider of evocative or rhetorical language cannot carry on a conversation without sotnetimes using a word in which a dead metaphor lies hidden; and everyday slang is full of metaphors. We give somebody beans, we are in the soup, we tell someone to go and boil his head, we are like a dog with two tails, our tiresome children have sent us up the wall, we lose our heads, we show a leg, we wash our hands of the matter, we get something off our chest, or we listen with half an ear. We are not intentionally using figurative language when we use such expressions, highly figurative though they are; we are using expressions that we have long taken for granted. Metaphor thus seems to be an innate and entirely natural aspect of the language, not only of artists but of non-artistic, IS8

Metaphor and Association even banal, human beings. It is closely related to analogy; a metaphor when it is original is an attempt to explain something by means of something else. A savage who had never seen an aeroplane might be given some notion of the meaning of the word by saying, 'It is a very large bird, as large as a hut, but it is not alive; it is the image of a bird made from metal, and it can fly for many hours. Men and women can sit in the ston1ach of the bird and so make journeys.' Scientifically speaking, this is most inaccurate, but it is an analogy which would mean something to the savage and would enable him to recognize an aeroplane when he saw one. Anyone who knows what cancer is understands such an expression as: 'Don't let yourself become envious, for envy is a sort of cancer of the soul.' If I say that in a crisis of my life some friend has been a rock, or that some other friend has turned out to be a jellyfish, tny hearer will understand me just as well as when I say that there is a hairbrush on the dressing-table. A good piece of advice is 'Don't keep your wishbone where your backbone should he~' This is a little more complex than an ordinary metaphor, for there is also a kind of play upon words in it. It means, of course, 'Do not tvish for things that you could achieve by n1aking a disciplined effort; daydreaming weakens the will to success.' Backbone is used metaphorically to imply cot-1rage and persistence (I suppose because our backbone enables us to stand upright and we look tnore enterprising and bold when we stand upright). Wishbone, a part of a chicken, is not usually thought of as a human bone, but here it is used to suggest 'a part of our anatomy with which we daydream'. If we stop to analyse the advice, it is thus somewhat lopsided and illogical; but everyone can understand what is meant at once. A false analogy may be very misleading. 'Do not criticize the Government; you know how tiresome back-seat drivers are' may sometimes contain some measure of sense, but it is 159

Metaphor and Association possible to reply, 'If the driver is about to take the car over a precipice, it is reasonable for the passengers to object!' 'You would pluck weeds out of a beautiful garden, so why not remove useless citizens from the country?' sounds very reasonable until we realize that this is going to be an excuse for acts of horrifying cruelty to some human beings. But all metaphor is bound to be imperfect; since no two things of different species can resemble one another in every detail, it is possible to point out the limitations of any metaphor, to be flippant by taking it too literally: 'Her goodly eyes like sapphires shining bright' (SPENSER: Epithalamion) refers to the blueness and the clearness of the lady's eyes; we are not tneant to suppose that the eyes are hard stones, or inanimate, or expensive. When Donne says to his God: 'In what tom ship soever I embark, That ship shall be my emblem of Thy Ark;' (On the Author's Last Going into Germany) we are meant to think of the protection of God, of the mercy shown to Noah, of a divine order and plan governing human experiences, perhaps too of an unbroken tradition in the Church-Donne was in later life a priest in the Church of England-but we are certainly not meant to have a picture of the 'animals going in two by two'. Indeed, a large part of the real understanding of metaphor in poetry is our selection of those aspects of the picture that are relevant, and our firm dismissal of possibilities that are not intended to be noticed. Humorists often point out the irrelevant content of metaphors in order to obtain a comic effect; superficially this 160

Metaphor and Association appears to be clever, but if done too often it can create habits of wanton misreading and spoil poetry for people: 'Like a pale flo\\?er by some sad ntaiden cherished, And fed with true-love tears, instead of dew. . . . ' (sHELLEY,

Adonais)

- ' Ah !' says our silly-clever humorist, 'but plants will not grow in a salty soil!' 'Those green-robed senators of mighty woods, Tall oaks. . . . ' {KEATS:

Hyperion)

and our humorist says, 'I hope there is no Senator McCarthy in the forest!' Such irrelevant images are to be excluded from the mind when we read poetry; the poet is entitled to that courtesy from the reader. He is even, I think, entitled to say when a reader claims that, for instance, 'The pansy at my feet Doth the same tale repeat. (woRDSWORTH: Intimations of Immortality) suggests to him an effeminate man lying prostrate, 'Well, that's your fault; I can't help that!' For the poet cannot hope to produce a metaphor that will carry exactly the same associations for everyone and that will not in any way bring in irrelevant possibilities. It is the reader's business to see that portion of the idea, that resemblance between two unlike things, that makes the metaphor illuminating; metaphors, honestly used, as by good poets and orators, are not meant to save us the trouble of thinking; they are meant to help us to think more deeply and adequately. Let us now look at some very well-known lines and analyse the 'meaning' of the metaphors-a repulsive proI6I

Metaphor and Association cedure, as all dissection of living bodies must be, but, perhaps, illuminating. That time of year thou may'st in me behold When yellow leaves, or none, or few, do hang Upon those boughs which shake against the cold, Bare ruined choirs, where late the sweet birds sang.' (sHAKESPEARE:

Sonnet LXXIII)

This means that the writer feels he is old, and implies also that he feels somewhat lonely and broken. Now, the statement, 'I am a poor, lonely old man!' would have a tone of rather unpleasant self-pity and emotional blackmail in a lovepoem, which the sonnet is. The metaphor of the tree in autumn first removes the idea one degree from the realm of the embarrassingly personal. There is, of course, one huge imperfection in the metaphor: autumn is merely temporary and every autumn is followed by another spring; old age is a permanent physical deterioration and, at least for the human body on earth, there is no possibility of a new spring. But we must dismiss this association as quite irrelevant to what is now being said; there are huge imperfections in all metaphorical language, just as there are in all other forms of communication. Additional associations are brought in to stress the aspect of autumn that Shakespeare here intends to emphasize: the cold, the bareness, the silence (absence of bird-song). The yellow leaves by itself might suggest too much of colour and beauty, the 'riches' of autumn, so the leaves are described as 'yellow ... or none ... or few ... '. If gold or golden had been used instead of yellow, the colour adjective would have been equally truthful, but would have brought in agreeable associations such as Shakespeare needed to avoid. The boughs that shake may be meant to suggest the tremulousness that is a common weakness of old age. Cold carries ideas of lone162

Metaphor and Association liness, poverty, deprivation, sexual inability and emotional starvation, as well as winter. The fact that the birds no longer sing on the tree in autumn in itself suggests a deprivation of past joys, a sense of loss, loneliness and melancholy silence; but this is further stressed by the bare ruined choirs; and the use of a word that suggests a church helps to suggest also the sanctity of the human being, to suggest that something worthy of sad reverence still clings about the ruin. Lastly, the music of the four lines helps to take away the possibility of unpleasant self-pity and to replace this ugly emotion by a gentle wistfulness that is attractive. Good metaphorical language may in a sense be far more truthful than prosaic non-figurative language; that is the point of writing poetry, perhaps. Responsible users of metaphor are trying to communicate what literal language cannot convey so well. A fairly common human experience is to wish to d(\ something that in itself seems to us innocent enough, but that would not be appropriate to our situation-that might, for example, discredit our cause or arouse the disapproval of others who cannot understand our point of view. We wish to do the deed, and half decide to do it; why are we being so silly, why restrain ourselves when there is no harm in it?but then something, call it conscience, or conditioning, or inhibition-at all events it is not a rational reaction-intervenes and we cannot do what we wish. Obviously this may at times be desirable and useful; at other times this inner hindrance may merely be a nuisance and cripple us in desirable activities. At all events, most readers of this book are likely to know the experience. I have tried to describe the experience in very general and abstract language, as objectively as I can. Here is a politician describing it and applying the experience, not to himself but to others.

Metaphor and Association 'Discipline is what n1atters; self-discipline' (I). Anyone with any moral sense will know what I mean (2). Anyone who loves his country lives continually under certain restraints (3). We all know this (4). You rnust all have experienced many times (5) those sacred hesitations, those restraints, those noble refusals, that make a man out of an animal (6). Consider: the temptation comes to you. You are tempted by it (7). But you are a man (8). You cannot even explain why this act is wrong; we cannot always explain the most important and sacred things in life (9). You hesitate; you would very much like to do it; you would so much like just this once, to give up all you stand for, to enjoy yourself (1o). Yet something holds you back. Possibly you curse it, possibly you even ask yourself, why am I so silly? But that strange inner force holds you back. You know that if you do not listen to it you will be unhappy {II}. You feel it as something sacred, as something more important than satisfaction or happiness {I2). And in the end you do not yield to temptation. Yours is the victory' (I 3). I do not deny that a man who talks like this may conceivably be sincere. But to the average listener, I think such a portion of a speech has a certain flavour of unctuousness, hypocrisy, unreality. We may briefly look at some of the things that jar: (I) One of those words that can mean almost anything we want it to mean at the moment. (2) The implication is the common one in these generalizations: 'no one who disagrees with me has any moral sense'. (3) Moral blackmail: we dare not question the assertions lest we be supposed not to love our country. (4) Do we? Perhaps so; but this is flattery of the audience and a further attempt at making dissent or query difficult. (s) Another trick to render dissent or query difficult. (No one likes to indicate that he may not be 'normal'.) 164

Metaphor and Association (6) We have already discussed 'make a man of'! (7) Of course one is tempted by a temptation; that is what

a temptation is for; but for a second this sounds most dramatic and as though something is happening further. (8) 'Man' as a 'clean word'-what does it really mean? (9) This appeal to the common inarticulateness of most of us is a kind of flattery of the audience, and also tends to suggest that we have no business to question certain things. (10) Enjoyment is surely a good thing if it does not do harm to others or to ourselves in the long run. However, most of us have a secret and rather morbid distrust of pleasure (psychologists explain this by reference to what is known as the super-ego); and the suggestion that we are capable of something nobler than enjoyment is subtly flattering as well as appealing to our secret fear of pleasure. (1I) Yet now in fact the speaker implies that happiness is to be pursued. . . . (12) ... and now, that it is something to be sacrificed.... (13) And 'victory' is a very good dramatic-sounding 'clean word' with which to end the paragraph. George Herbert, I59J-I6JJ, was a priest in the Church of England and lived a somewhat austere life, thinking this his duty. Sometimes he h«d impulses of rebellion against his strict mode of life; but some inner voice always called him back to his difficult duty. He writes of this experience in a very figurative poem; the drastically figurative language helps to suggest the violent, emotional thoughts, at times almost chaotic. THE COLLAR

I struck the board, and cried, No n1ore! I will abroad. What? Shall I ever sigh and pine?

My lines and life as free, free as the road, 165

Metaphor and Association Loose as the wind, as large as store. Shall I be still in suit? Have I no harvest but a thorn To let me blood, and not restore What I have lost with cordial fruit? Sure there was wine Before my sighs did dry it. There was com Before my tears did drown it. Is the year only lost to me? Have I no bays to crown it? No flowers, no garlands gay? All blasted? All wasted? Not so, my heart! But there is fruit, And thou hast hands. Recover all thy sigh-blown age On double pleasures. Leave thy cold dispute Of what is fit and not. Forsake thy cage, Thy rope of sands, Which petty thoughts have made, and made to thee Good cable, to enforce and draw, And be thy law, While thou didst wink and wouldst not see. Away! Take heed! I will abroad. Call in thy death's head there. Tie up thy fears. He that forbears To suit and serve his need Deserves his load. But as I raved and grew more fierce and wild At every word, Methought I heard one calling, Child/ And I replied, My Lord. Sharing neither George Herbert's religious convictions nor

166

Metaphor and Association his belief that asceticism is meritorious, I immediately feel a s=ncerity and a profundity of experience in this poem. The very violence of metaphor helps to suggest the violence and the knotty complexity of the emotion. This poem is in anything but literal language; but it is certainly about something

real.

The examination of language in this book has hardly scratched the surface. It is possible to read many books on semantics. It is not possible to learn how to communicate perfectly by means of words, for it is not in the nature of language to communicate perfectly. The reason for studying the meaning of words is that by so doing we may perhaps become a little more honest and communicate a little better. There are many occasions in human life when that 'little' is of great importance to human happiness and morality.

A FEW QUESTIONS To speak of a person as obedient or docile is usually meant as a term of praise. Can these words ever be legitimately used in dispraise? 2. The expression just five minutes has one objective meaning and a number of other personal meanings. Discuss. 3. What do we mean when we speak of a person as a normal or average person? When are these terms useful and when are they meaningless or misleading? 4· Give ten definitions of a nice girl such as might be given by ten different people.. s. What are the various (and numerous) meanings of the word self? 6. Explain how the word protection has come to have, at times, a sinister meaning. I.

Metaphor and Association 7· Differentiate the following pairs of words: lazy, idle;

enthusiastic, fanatical; loyal, uncritical; vanity, pride; normal, usual; passionate, emotional.

8. What motive lies behind the use of such expressions as

the fuller figure, matronly appearance, mature lines? 9· What is a great man?

What is the difference between 'I found a large parcel in my room' and 'I found a great parcel in my room'? II. Make a list of words that you have heard people use as synonyms for not-agreeing-with-me. 12. What is the correct meaning of the following words: psychology, complex (in psychology), inhibition, allergic, Bol10.

shevik, romantic, bureaucracy, imperialist? How have these words come to be popularly used with

inexact and even incorrect significance? 13. Why are highbrow and intelleciual often used as terms of abuse instead, as might be expected, of respect? 14. 'He works hard!' is praise. Why do some people use the word working-class in a tone of contempt? IS. What is a gentleman? And a lady? 16. What is the difference between weather and climate? I7. Can you account for the totally wrong use of the word native, in England, to denote a coloured man? And why are black or brown men coloured and English people presumably colourless? 18. Study the numerous n1etaphorical uses of the word

black.

19. Examine in as much detail as possible the very numerous

and often widely different meanings of the word common.

What different meanings may be attached to the words my and mine? 20.

168

INDEX Absalom and Achitophel, 67 Adonais, I6I Advertisements, IO, 83 et seq., I 3 I Analogy, 142, I59 Arabic, 66 Association, 28, 32, et seq., 36, 38, 59, 61, 68 et seq., IOO, I42, I 50, I 56 et seq. Bible, I47 Blake, William, I 57 Book of Common Prayer, 62 Bright, John, 150 Burke, Edmnnd, 150, IS7

Dean, James, 54 Defmition, 23 et seq., 36, I04, I28

et seq. Dictionary,

IO,

23, 26, 59, 61, 73,

74, 76

I>onne;John, 68, I46, I57, I6o

Dryden, John, 67 Ecstasy, The, I46

Falstaff, 5 French, 66, 72, 73, 74, 75, 76, 77, 78 Freud, Sigmund, I I I, I 33, I 34 German, 72, 73, 74, 75, 78 Gesture, 6, 7 Gettysburg Oration, I 51 Greek, 67, 69, 7 I

Henry W, S Henryson, Robert, 61 Herbert, George, I65 et seq. Hitler, Adolf, I 3 I, I44, I 5I Hopkins, G. M., 146 How the Alphabet was Made, 7 Hungarian, 73, 75 Hyperbole, 143, I46 Hyperion, I 6 I

Censorship, 40 Chaucer, Geoffrey, 6I Chesterton, G. K., I44 Churchill, Sir Winston, I so Cliche, I04 et seq., I2S, I4I Collar, The, 165 et seq. Croat, 74 Davies, Sir John, I 56

Epithalamion (Spenser), 16o Esperanto, 17, 72, 79, I27, 152

In Memoriam, I 57 Irrelevant Images, Italian, 72, 73, 75

161

julius Caesar, 65 Keats, John, I 6 I Kipling, Rudyard, 7 Knox, John, 62, 63

I6g

Index Kubla Khan, II

Proverbs, 143

Latin, 64, 66, 7I Lawrence, D. H., 143 Lincoln, Abrahan1, I 5I Lyrical Ballads, I43

Rebel without a Cause, 54 Rhetoric, I4I et seq. Rhetorical Questions, 149 Rivals, The, 22

Macbeth, 62, 64 Macaulay, Lord, I02 Magic, 4 Malapropism, 22 Marvell, Andrew, 156 Mathematical Symbols, 36 Merchant of Venice, The, so Metaphor, 142, I43, I 58 et seq. Middle English, 61 Middle Scots, 61 Milton,john,66,69, 156 lvfirror for Magistrates, A, 63 Monstrous Regiment of Women, The, 62 Music, 7

Schweitzer, Albert, 52 Semantics, IO, II, 45, 55, 92, 141, 167 Semasiology, 1 o Sentimentality, 92 et seq., 141, 145, I46, 148, I 54 Serbo-croat, 7 S Shakespeare, William, 5, 50, 6 I, 62, 64, 162 Shelley, P. B., II, I2, 157, I6I Sheridan, R. B., 22 Shylock, 51 Simile, 142 Slang, 158 Slogans, 40, I 04 et seq. Sociable noises, 4, 106 Song of the Emigrants in Bermuda, 156 Sonnet (Shakespeare}, 162 et seq. Spanish, 75 Spenser, Edn1mtd, 160 Swan Lake, 52 Swear-words, 5, I2J, I 54 Swedish, 72, 74, 75 Sylvester, J., 146

Nosce Teipsum, IS6 Ode on Intimations of Immortality, I6I

Ode on the Morning of Christ's Nativity, 66 Ode to the West Wind, I 57 Old English, 61, 65 On the Author's Last Going into Germany, I 6o Othello, 53, 68 Personification, 142 Portia, 51 Portuguese, 75 Propaganda, 9, 39, 68, 99, 105, 127

Taboo,4, 30, 34, 77,78 Tennyson, Alfred, 157 Wilson, President (U.S.A.), 67 Wordsworth, William, 143, 147, I57, I6I Wreck o.f the Deutschland, 146

170

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 3

PRINCIPLES OF SEMIOTIC

This page intentionally left blank

PRINCIPLES OF SEMIOTIC

D.S. CLARKE

First published in 1987 by Routledge & Kegan Paul Inc. This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1987 D.S. Clarke Jr All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69180-3 978-1-138-69183-4 978-1-315-53377-3

(Set) (Set) (ebk) (Volume 3) (hbk) (Volume 3) (pbk) (Volume 3) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

PRINCIPLES _OF_

SEMIOTIC nSCLARKE

First published in 1987 by Routledge & Kegan Paul Inc. in association with Methuen Inc. 29 West 35th Street, New York, NY 10001 Published in the UK by Routledge & Kegan Paul Ltd JI New Fetter Lane, London EC4P 4EE Set in Times New Roman by Columns, Reading, Great Britain and printed in Great Britain by Richard Clay Ltd, Bungay, Suffolk Copyright

© D.S. Clarke Jr 1987

No part of this book may be reproduced in any form without permission from the publisher except for the quotation of brief passages in criticism Library of Congress Cataloging in Publication Data Clarke, D.S. (David S.), 1936Principles of semiotic. Bibliography: p. Includes index. 1. Semiotics. I. Title. P99. C54 1987 401.41 86-21945 British Library CIP Data also available ISBN 0-7102-0981-9 (cased) ISBN 0-7102-JI36-8 (pb)

TO MY TEACHER

CHARLES HARTSHORNE

This page intentionally left blank

C

ONTENTS

Preface

1 Introduction 1.1 Logical analysis 1.2 Ordinary language description 1.3 The role of semiotic

ix 1 2 5 7

2 History of semiotic 2.1 The Classical tradition 2.2 Augustine and his successors 2.3 Peirce and Saussure 2.4 Behavioral semiotic 2.5 Semiotic's critics

12 12 19 25 30 36

3 Natural signs 3.1 Signs and evidence 3.2 Images 3.3 Natsigns: some basic features 3.4 Dynamic interpretation

43 43 51 58 65

4 Communication 4.1 Communicative intent 4.2 Conventional signs 4.3 Signals 4.4 Features of communicative systems

73 73 81 90 96

5 Language 5.1 The role of subjects 5.2 Denotation and reference 5.3 Meaning, truth and illocutionary force 5.4 Addresses 5.5 Discourse

104 104 111 117 124 130

Postscript

137

Notes

139 vii

Vlll

Contents Name index

159

Subject index

161

P

REFACE

Semiotic has a long, continuous history that dates back to the debates between the Stoics and Epicureans over the status of the sign as an object of interpretation. Despite this history the present prospects for the subject in British and American philosophy seem very dim indeed. Interest in the subject in the English-speaking world is confined primarily to those either with historical interests or in fields such as literature, theatre, or film influenced by the tradition of European semiology (or 'cultural semiotics'), an area of study with a much different focus from that of past periods. This work is an attempt to reestablish semiotic on the basis of principles more consistent with its past history and especially with the guiding ideas found in the writings of Peirce and Morris. This task has been made easier by the work of such recent philosophers as Grice, Strawson, Austin, Bennett, and Lewis, who have introduced into current philosophy of language - often without explicitly acknowledging it - discussion of issues directly relevant to those of the Peirce-Morris tradition. It is because of this recent work and also because of advances in our understanding of sub-human systems of communication that the nature of semiotic as a discipline has changed and advanced beyond that defined forty years ago in the writings of Morris. The book is divided into two parts, with the first two chapters providing the background for the more systematic discussions of signs at different levels taken up in the last three. Much of the interest in semiotic stems from the comparative insights it contributes to our understanding of language use. In the final chapter issues that have become the focus of recent philosophy of language regarding the reference, meaning, and truth of sentences are discussed in the light of the analogies to more primitive signs developed in the preceding two chapters. I have tried to make both background and systematic discussions accessible to a wide variety of readers, an introduction to what I lX

x Preface conceive as the key issues of semiotic as a viable part of presentday philosophy. The notes to be found at the end are designed to inform those interested in the extensive literature on the subject and more specialized problems. The origins of this work date back to studies of Peirce's semiotic and participation in seminars conducted by Charles Hartshorne in which he developed his philosophy of panpsychism - both of which took place while I was a graduate student at Emory University in the early 1960s. Hartshorne's vision of human experience being continuous with that found in all forms of life seemed to me then, and still does now, an exalted one, an essential correction to the anthropocentric bias of postCartesian philosophy. I was then swept up in his project of constructing a metaphysical system in which basic features of human experience, as basic as the substance, form, and matter distinctions of Aristotle, were to be extended to all sentient organisms. But the basis for this extension used by Hartshorne and the Whiteheadean tradition he represents seemed to me defective, one no longer viable in the face of the successful critiques of a priori psychologizing by the linguistic philosophers. Instead, the basis, I thought, must be our use and interpretation of language, with semiotic the means for extending their principal features. It is this revision which this work attempts to carry out. I am very grateful for the help of students and faculty colleagues at Southern Illinois University with whom I have discussed these issues over the past few years. In particular, I owe much to the careful reading of an earlier manuscript version by Thomas Mitchell and Mark Johnson; both have provided valuable criticisms leading to revisions. lowe a similar debt to an anonymous reader for Routledge & Kegan Paul. To Stratford Caldecott I am grateful for encouragement and guidance in the early planning of the work. Part of its preparation was made possibly by a research grant from Southern Illinois University at Carbondale for the summer of 1985. Carbondale, Illinois December, 1985

1

INTRODUCTION

In contemporary philosophy the focus of study is the use and interpretation of language. To study scientific inquiry is to study the language of science and how it is used; to study legal decision-making is to study the language used in applying laws; to study aesthetic expression is to study the symbolic forms for this expression; and so it is for every human activity. For philosophy seeks to understand the nature of thought, and to think is to use and interpret language, sometimes as it is publicly expressed by ourselves and others, sometimes in the form of what Plato calls in the Phaedo the 'inner dialogue of the soul with itself' where public expression is suppressed or delayed. If we had special introspective powers by which we could directly intuit the nature of our thought processes, then language might playa less central role in philosophical investigations. But we have no such powers, and previous attempts to describe directly these processes and their contents through such terms as 'concept', 'idea', 'sense data', etc. can be shown to be in fact importing distinctions derived from language. Hence the shift instituted at the end of the nineteenth century by Frege and Peirce from a conception of philosophy as the study of psychological processes to one concentrating on the description and analysis of linguistic expressions. On these points almost all can probably agree, and the justification for making the shift is generally accepted. Differences arise mainly over the method to be used in studying language and the objectives to be pursued. It is helpful at this preliminary stage to review briefly the two methods that have dominated recent philosophy in the English-speaking world and to raise some difficulties that confront them. This will help us to determine the special role semiotic can hope to play in relation to contemporary studies of language. 1

2 Introduction 1.1 Logical analysis Bertrand Russell can be credited with instituting the widely adopted method for applying the symbolism and paraphrase techniques of modern predicate logic to sentences in natural languages in order to clarify their meaning, and thereby correct philosophic conclusions based on a misunderstanding of their logical form. Russell's paradigm sentence, the one whose analysis can be regarded as starting the logical program, was (1) The present king of France is bald which he paraphrased by (2) Exactly one (material) thing is such that it is a king of France and bald Its representation in symbolic notation becomes finally (3) 3x[Kx & Vy(Ky => y=x) & Bx]

with 'K' for 'is a king of France' and 'B' for 'is bald'. In (3) the existential quantifier '3x' and the first two conjuncts within its scope have the effect of stating that there exists at least one king of France and at most one (any other thing which is a king of France will be identical with the one which has been stated to exist). The third conjunct as quantified states this king is bald. The effect of the paraphrase is to change the subject from the original grammatical subject 'the present king of France' in (1) to the general subject 'thing' or 'material thing' of (2). This latter is the logical subject, and is represented by the variable 'x' within the existential quantifier. The 'true' meaning of (1) is revealed, Russell claims, only by the paraphrase, and this meaning is represented by the sentence's logical form as (3).1 The project of analysis begun by Russell was to be carried out by similarly paraphrasing and representing all those other sentences of ordinary language whose grammatical form disguised logical form in a way giving rise to philosophical confusions. The symbolism of formal logic was originally designed to represent the form of sentences within the context of inferences in order to evaluate these inferences as valid or invalid. In order to make such evaluations the logician must make certain assumptions. The most fundamental are those which insure a

Introduction

3

connection between the premisses and conclusion of a given inference. Consider an example of the simplest of all inferences, (4) John is sitting Therefore, John is sitting in which the conclusion simply repeats the premiss. It is, of course, valid. But in order to establish its validity we must assume that the sentence 'John is sitting' as it occurs in the premiss expresses the same content in the conclusion. This common content necessary for validity is labelled the 'proposition' expressed by the two occurrences or tokens of the sentence. Sameness of proposition in turn requires first that what logicians call the 'denotation' of its subject term 'John' remains constant, i.e., that there be a unique individual which both occurrences of the name denote. And second, granted that we have the same proposition, we must assume its truth value remains constant, that if it is true in the premiss it remains true in the conclusion. John could stand up between the time a speaker utters the premiss and then utters the conclusion, and hence the premiss would be true and conclusion false. Logic must assume that this does not occur. The effect of these assumptions is to abstract the sentence 'John is sitting' as it occurs in the content of inference (4) from its use or interpretation by a person on a given occasion. All we need to know in evaluating (4) as valid is that if the premiss were true then the conclusion would be true also. For this a person's assertion of or assent to the sentence at a given time and place is irrelevant. Outside inferential contexts, of course, sentences function very differently. On different occasions of use the name 'John' may be used by a speaker to refer to different individuals. The triadic relation between an utterance of a name within a sentence, a person using or interpreting the sentence, and a referent, is thus different from the dyadic relation between the name and what it uniquely denotes as considered by the logician for his special purposes of evaluating inferences (cf. below Section 5.2). And similarly for change of truth value. 'John is sitting' as used on one occasion can be asserted as true, while on another denied as false, contrary to what must be assumed when the sentence is used as a premiss and our task is inference evaluation.

4 Introduction It is small wonder, then, that Russell's project of employing techniques and symbolism designed for inferences where truth and denotation constancy are assumed, would encounter difficulties when applied to sentences isolated from inferential contexts. As Strawson has shown, it is because the assumptions do not generally hold of these sentences that Russell's paraphrase and representation of sentence (1) is defective. 2 To assert the proposition it expresses as true, or to assent to it, presupposes the existence of its referent on that occasion. Only if we can identify the referent of the singular term 'the present king of France' can we judge the proposition true or false. Since this condition does not hold (in fact, there is no such referent), (1) cannot be asserted or assented to. Yet (2) must be asserted as false, since its first conjunct as quantified falsely states there is at least one king of France. 3 Hence, (2) cannot be regarded as having the same logical meaning expressed in more perspicuous form as (1). Implicit in Russell's project of analysis is the assumption that every sentence with a specific meaning has a unique logical form which paraphrase and logical representation can reveal. The meaning of a sentence is regarded as equivalent to the totality of consequences that it (or the proposition it expresses) logically entails. To specify a sentence's meaning then becomes the task of representing its form in such a way that the inferences to all these consequences can be shown to be valid. But, in fact, no one logical representation will justify this indefinite number of inferences. Consider, for example, Donald Davidson's analysis of action sentences such as (5) John walks in the street which Davidson proposes to paraphrase by 'There is an event which is such that John walks in it and it (the event) is in the street.' This is finally represented by (6) 3x(Wjx & Ixs) This is claimed to represent the form of (5), since it justifies the inference to its consequence 'John walks'.4 Indeed, this inference is justified, but the cost in intuitive plausibility is a high one, since we are forced into introducing a variable ranging over events, names of individual things such as persons and things, and strange relations between them, e.g. the relation of walking in

Introduction

5

between John and an event. Moreover, there are other consequences which (6) will not justify, consequences such as 'John sometimes walks', 'John walks somewhere', and 'It is possible that John walks'. To represent them we will have to introduce additional variables ranging over other 'objects' such as times, places, and actual and possible worlds. Such objects require even more bizarre relations between them. All of this seems to show that there is not some meaning of (5) which some unique logical form can specify. Instead, there is an indefinite variety of inferential contexts in which (5) can occur, and as these contexts vary we vary our logical representation in order to justify those inferences which on the basis of prelogical intuitions we accept as valid. Logical representation is not a means of specifying the meaning of an isolated sentence, but instead a device for evaluating a specific inference in which the sentence may occur. Logical analysis has thus its principal application to a special use of language, that of deductively inferring conclusions from premisses. It can also be extended to assist in the solution of specific philosophical problems, and has proved successful as a means of criticizing mistaken inferences from grammatical form to ontological conclusions. The symbolism of logic is also useful as an abbreviating device for stating general conclusions not restricted to this or that particular example. But as a general theory of language and meaning it has not had and cannot have the success its advocates have hoped for it. 1.2 Ordinary language description The so-called 'ordinary language' school of philosophy instituted by the later Wittgenstein, Austin, Ryle, and Strawson used a very different method. Instead of paraphrasing and representing sentences in the manner required for inference evaluation, they undertook to describe the use of a variety of categories of sentences, including imperatives, expressions of feelings and emotions, and reports of sensations, in addition to the fact-stating indicative sentences that occur in standard deductive inferences. There are rules governing the use of sentences in these various categories, and it becomes the role of philosophy to make them explicit. Of special interest are what Austin terms 'performa-

6 Introduction

tives,' sentences containing prefixes such as 'I promise that ... ' or 'I state that ... ' which are used by speakers to perform the speech acts described by their main verbs, verbs such as 'promise' or 'state'. But the explication of rules of use is also extended to sentences containing words such as 'believe', 'certain', 'voluntary', 'ought', etc., often with the intent of criticizing a traditional philosophic theory which is claimed to be based on a misunderstanding of their use. Doubts must be raised about the long-term viability of this method. First of all, granted its success as a means of criticizing assumptions made in modern philosophy since Descartes, once the criticisms have been completed there seems little else to accomplish. There is a limited number of errors that can be presumed to have been made in the historical tradition because of a misunderstanding of key words of ordinary language. Having exposed them, at least this reason for studying the rules governing the use of ordinary language will come to an end. Indeed, this prospect is envisaged by Wittgenstein when he predicts the demise of philosophy once the 'therapy' of disclosing its violation of rules of ordinary language has been completed. As for the constructive task of simply describing linguistic rules, this also threatens to come to an end as a distinctively philosophical project. The science of linguistics has gradually incorporated most of philosophy's main findings as it has extended beyond the study of the syntax of language to the fields of semantics and pragmatics. The rules governing the use of performatives and the presuppositions for this use, for example, have become the subject matter of pragmatics as a branch of linguistics, and the statement of these rules is virtually indistinguishable from that given by the ordinary language philosophers. Just as the a priori introspective psychology of modern philosophy was, in Austin's words, 'kicked upstairs' to become part of empirical psychology, so philosophic conclusions reached about language have supplied the foundations for new areas of linguistics. An attempt to distinguish the philosophy of language from linguistics can be made by claiming that philosophy attempts to state necessary or essential features of language, features which every natural language must have, while linguistics states contingent features which languages happen to have as the result

Introduction

7

of the special needs and arbitrary choices of the communities in which they have evolved. Thus, it can be claimed, and with justification as we shall see in Chapter 5, that every sentence must have a subject-predicate structure, while the specific grammatical rules which specify agreement between nouns and verbs will vary from one language to another. These rules will then be the subject matter of linguistics, while the subjectpredicate distinction is a philosophical one. 5 But linguistics is also concerned with specifying universal features of languages, features which are in fact shared by all the world's surveyable languages. Included among these are the subject-predicate structure of sentences, as well as universal phonological laws which place constraints on the permissible sequences of speech sounds, for all languages. These universal features are inferred on the basis of an inductive inference from empirical evidence, and are in this sense contingent features. How do we distinguish them from the necessary features sought by philosophy? Within the framework of language itself it is difficult to make this allimportant distinction, and vague intuitions seem to become the final court of appeal. 6 In fact, few philosophers in the ordinary language tradition (Strawson is a notable exception) have even attempted to confront this problem. Its critics have noted the almost indefinite variety of linguistic forms studied and the haphazard selection from among them that has often been made. 7 Often selected for attention, as we have noted, has been a word or phrase which is the source of what is claimed to be a confusion made by the historical tradition. But relevant though they may be for historical corrections, such selections seem to be totally irrelevant to a general theory of language. Again, certain features of language are presumably central for any such philosophic theory, but ordinary language descriptions do not themselves provide a basis for selecting some instead of others. 1.3 The role of semiotic Semiotic attempts to remedy this defect by singling out for attention features of language which are both invariant for signs of different levels of complexity and which distinguish sentences as linguistic signs from more primitive signs. Instead of what can

8 Introduction be termed horizontal classifications of linguistic forms within natural languages or within special language frameworks such as those of science, law, or art, semiotic attempts to develop vertical classifications for signs at different levels of complexity, including environmental natural events as objects of interpretation, signals without grammatical structure, and finally sentences as the basic elements of human communication. Necessary features of language then become those which are invariant at all levels of signs and which distinguish linguistic signs as a category fulfilling certain functions from those at lower levels. The subjectpredicate structure of a sentence is more than a feature which happens to be shared by known natural languages; it is a feature that must appear at the linguistic level if the sign is to function there in a way different from what is characteristic at the levels of signals and natural events. The starting point for semiotic is the same as for contemporary philosophy of language: the use of ordinary language. As language users this is what we are most directly aware of as constituting in Peirce's words the 'warp and woof' of human thought. Unique to semiotic is its attempt to extend analogically features initially arrived at by examining language use to more primitive signs, with logical features of language becoming the archetype on which analysis of these latter signs is developed. 8 By characterizing a sign Xl as 'more primitive' than another X 2 I mean that the interpretation of Xl lacks one or more features present in the interpretation of X 2 , while features of Xl'S interpretation are preserved in X 2's. Thus, a signal such as a warning cry is more primitive than a linguistic sentence, in so far as it lacks an internal subject-predicate structure which enables reference to distant objects. On the other hand, devices used to secure reference for signals are present in the use of sentences. A natural sign such as an odor interpreted by a deer as the sign of a predator is, in turn, more primitive than a signal in not being produced with a communicative intent that must be recognized in order to interpret the sign. But a signal can direct its interpreter's attention to a referent occasion in a way similar to that for natural signs. After describing the interpretation of more primitive signs on the basis of analogically extended features we can then determine more exactly which features of language use are in fact necessary. There is nothing at all circular in this

Introduction

9

procedure. We proceed from a rough characterization of language use to a more exact understanding of it later based on comparisons made to signs at lower levels. In this way terms such as 'sign', 'sign type', 'interpretation', 'significance', and 'reference' take on a generality of application beyond the starting point from which they are initially derived. The procedure may be compared to that used by neurophysiologists in unlocking the mysteries of the human brain. Because of its overwhelming complexity neurophysiologists first study the far simpler structures of lower organisms, starting with the flat worm and proceeding on through the phylogenetic scale of reptiles, birds, and mammals. With this as a background the functioning of the parts of the human brain becomes more clearly understood. So too for language. As the product of a long .evolutionary history it also displays bewildering complexity. The key to understanding its central features is to trace their development from more primitive signs, signs whose use and interpretation we can recover with some difficulty from our own experience, but which we can speculate as occupying the central part of the experiencing of the lower forms of life from which we have evolved. 9 Philosophers of language have in fact selected out for their chief attention those features of language which allow extended application to primitive signs. Out of the great variety of possible inferences philosopher-logicians, at least in the early development of modern logic, selected out those whose representation exhibited the same fundamental features of sentence form that comparisons made by semiotic reveal. Similarly, not all features of ordinary language have been the objects of philosophical descriptions, and in this respect they differ from those found in the writings of linguists. Major consideration has in fact often been given to those features uniquely characteristic of linguistic signs or shared in common by signs at more primitive levels. But rarely have the reasons for the choice of these features been made explicit. Paralleling the relation of the philosophy of language to linguistics are relations of semiotic to a host of empirical sciences, including ethological studies of animal communication systems, comparative psychology, psycholinguistics, kinesics, and paralinguistics. Philosophical semiotic is not a competitor with these

10 Introduction empirical disciplines, since it is concerned with invariant and contrasting logical features of the special types of signs which each singles out for special study. But as we shall see, these sciences do provide the materials for the comparisons that semiotic seeks to make, and key discoveries within them inevitably have an important effect. Advances within these sciences force present-day semiotic to be a very different discipline from what it was for Peirce in the nineteenth century. We can speculate that semiotic may also have a positive influence on the development of new sciences. Given the basic methodology of the biological sciences to seek descriptions of the more complex in terms of the relatively simple, it is surprising to find the contemporary discipline labelled 'cognitive psychology' or 'cognitive science' attempting to construct models describing the computational processes of the human brain which ignore comparisons to information processing in lower organisms. The causes of behavior for cognitive scientists are 'mental representations' as states of the brain, with these representations (or signs) identified with encoded sentence tokens. 1O Thus, the task of explaining a person bringing his umbrella in terms of his belief that it will rain and his desire to stay dry becomes reformulated as the task of explaining how the behavioral response of umbrella carrying is caused by the brain state encoding a token of the sentence 'It will rain' and an encoded token of 'I shall stay dry'. But obviously, it will be difficult to explain the operations of a system as complex as the human brain. Far more tractable would seem to be the problem of explaining the causal relation between the encoded internal signs of lower organisms and behavior as their effects. The restriction of cognitive science to models postulating sentence tokens as the processed elements has led Paul Churchland to reject this entire approach as a means of explaining behavior in terms of psychological states. Infants prior to language acquisition and other infra-linguistic organisms, he argues, certainly are not processors of encoded sentence tokens. Yet how their wants and expectations affect their behavior and how they learn from experience seem continuous with how behavior is affected and learning accomplished for adult human language users. Hence, Churchland concludes that any model requiring processing of encoded sentence tokens cannot be

Introduction

11

adequate to the essential features of human cognition. l l At least the beginning of a remedy may be provided by semiotic with its investigations of logical functions of signs interpreted by infralinguistic organisms and their similarities and differences with the functioning of sentences as used by adult humans. An understanding of such functions may suggest some day ways of constructing empirically testable models which can successfully explain the continuities that Churchland is correct in emphasizing. Semiotic as a sub-discipline in philosophy has been in existence since at least the early Middle Ages. Why has it been virtually ignored by the two main schools of the philosophy of language in this century? Why has language been analyzed and described in isolation from comparisons to non-linguistic signs? To answer these questions we must first review in the next chapter the history of the subject, paying special attention to difficulties which were later to block its progress and prevent it from assuming the role in the study of language which it should now be occupying. After this we can proceed to the constructive task of outlining the main similarities and differences between sentences and more primitive signs.

2

HISTORY OF SEMIOTIC

This chapter is devoted to providing the background essential for the systematic account of different levels of signs that begins with the next. I am not concerned here with providing a sequel to the excellent historical studies available in the recent literature. It is sufficient for our purposes to show the origins of some issues that will concern us later and to outline the main sides in the controversies arising from earlier solutions to basic problems. The most important of these involves the status of natural signs, signs not produced for the purposes of communication, and the relationship of these signs to linguistic signs. In the process of conducting this brief survey we will see how the conception of a sign has evolved through a long history of philosophical discussions. This presents us with the problem of selecting among alternative conceptions in defining the scope of semiotic in the final section of this chapter. 2.1 The Classical tradition The Greek word for sign was 'to semeion' (plural, 'ta semeia'). The modern 'semiotic' is derived by way of a Medieval intermediary from the Greek semeiotikos, meaning an observant of signs, one who interprets or divines their meaning. From the earliest writings available to us, beginning with Hippocrates and Parmenides in the fifth century Be through to the discussions of such Roman writers as Cicero and Quintillian, writers in the classical period used the term 'semeion' as a synonym for 'tekmerion' to mean evidence, proof, or symptom of what was at least temporarily absent or hidden from view. 1 Examples commonly given were smoke as a sign of fire, clouds as a sign of an impending storm for the sailor at sea, and a flushed complexion as a sign for the physician of a fever. In all such examples we have one natural object or event which can be

12

History of semiotic

13

directly observed in the present standing for another which cannot be, though sometimes (rarely in philosophic discussions) 'semeion' was also used in the sense of a sign or omen of the supernatural. It was also used to stand for facial expressions or behavior as signs of mental states themselves inaccessible to the observer, as blushing is a sign of shame or a grimace a sign of felt pain. But the paradigm sign in the most extensive discussions available to us, those found in the writings of Sextus Empiricus, 2 was the medical symptom as a means of diagnosing the condition of a patient. The principal semeiotikos for classical philosophers was thus the physician seeking to determine a hidden disease in order to apply a cure. For Hippocrates, diagnosis involved discovering signs signifying the past and future as well as the present: I hold that it is an excellent thing for a physician to practice forecasting. For if he discover and declare unaided by the side of his patients the present, the past, and the future, and fill in the gaps in the account given by the sick, he will be more believed to understand the cases, so that they will confidently entrust themselves to him for treatment. 3 Quintillian was later to explicitly point out the three temporal directions of signs. He notes that a woman giving birth is a sign of past sexual intercourse, with the inference from sign to what is signified going from effect to prior cause, while a serious wound is the sign of death, with the inference now from a presently observed cause to a future effect. 4 In contrast, waves are a sign of contemporaneous wind, just as smoke is contemporaneous with fire, flushed complexion with a fever, etc. Classical authors also distinguished, following Aristotle, infallible or necessary signs from 'refutable' or probabilistic signs which do not guarantee the presence of what they signify. As examples of necessary signs Aristotle gives a woman's giving milk as a sign that she lately bore a child and a fever as a sign of illness. Probabilistic signs are illustrated by fast breathing as a sign of fever, 'since a man may breathe hard without having a fever.,5 The context of this distinction by Aristotle is a discussion of the rhetorical syllogism in his Rhetoric, a context which had, as

History of semiotic

14

we shall see, an important influence on the conception of signs later developed by the Stoics. The most important controversy developed in these early discussions was that between the Stoics and the Epicureans, and centered on the status of the sign itself. As reported by Sextus Empiricus, the Stoics held that the term 'semeion' stands for a proposition or 'intelligible' (lekton) describing an observable fact. As such it is a constituent of an inference in which we infer what the sign signifies or stands for. The form of this inference is modus ponens, with the sign as one premiss and the antecedent of the conditional or hypothetical which constitutes the other premiss. What the sign signifies is then inferred by detaching the consequent. In Sextus's words, for the Stoics, 'the Sign is an antecedent proposition in a valid hypothetical major premiss, which serves to reveal the consequent.'6 The example of a hypothetical premiss given is 'If this man has had a viscid bronchial discharge, he has a wound in his lungs'. Having a bronchial discharge is a sign of the lung wound. For this premiss is valid, as it begins with the truth 'This man has had a viscid bronchial discharge', and ends in the truth 'he has a wound in his lungs'; and besides, the first serves to reveal the second; for by observing the former we come to an apprehension of the latter. 7 More generally, if X is the sign and Y what it signifies, we can infer from X to Y by way of the inference.

X If X then Y

Y In opposItIon to this inferential theory of the Stoics, the Epicureans are reported by Sextus to have held that the sign is a sensible particular, an object of direct observation rather than a proposition expressed within an inference. It is the observed smoke that is the sign of fire, not the proposition expressed by the sentence 'There is smoke over there'. First, they contended that the sign must be what signifies, and while utterances or written inscriptions signify propositions, propositions as intellectual conceptions do not themselves signify. Hence, 'as proposi-

History of semiotic

15

tions are signified, but not signifying, the sign will not be a proposition.,8 Second, they argued that illiterates and lower animals seem to be incapable of reasoning by modus ponens, but are capable of interpreting signs: ... if the sign is a judgment [or proposition] and an antecedent in a valid major premiss those who have no conception at all of a judgment, and have made no study of logical technicalities, ought to have been wholly incapable of interpreting signs. But this is not the case; for often illiterate pilots, and farmers unskilled in logical theorems interpret by signs excellently .... Yet why do we talk of men, when some of the Stoics have endowed even irrational animals with understanding of the sign? For, in fact, the dog, when he tracks a beast by footprints, is interpreting by signs; but he does not therefore derive an impression of the judgment 'if this is a footprint, a beast is here'. The horse, too, at the prod of a goad or the crack of a whip leaps forward and starts to run, but he does not frame a judgment logically in a premiss such as this - 'If a whip has cracked, I must run'. Therefore the sign is not a judgment, which is the antecedent in a valid major premiss. 9 The Epicurean view, as we shall see in the next chapter, is the superior for framing a general theory of signs inclusive of language, and we must recognize the soundness of both arguments. But it is important to notice how the choice of examples of signs selected by classical writers and the use of the medical symptom as the paradigm sign virtually dictated the Stoic view. In fact, it is by way of an inference capable of linguistic formulation that we infer from evidence such as smoke, a fever, a bronchial discharge, or a scar, to their past or present causes or future effects. Moreover, this is an inference that those unschooled in logic can make as easily as the educated. It does not, after all, require an ability to formulate the rule of modus ponens in order to use the rule for evidential reasoning, and in this sense the illiterate pilot or farmer is the equal of us all. If the examples of signs interpreted by animals had been used by the Stoics generally, then indeed they would have been guilty of the inconsistency being charged. But it is only 'some of the Stoics' who used these examples. More typical were examples given of

16 History of semiotic signs which could only be interpreted by humans with their capacity to use language and draw inferences, and for these the inference from sign to what it signifies does conform to the Stoic pattern. For such objects or events - and they include all the socalled 'natural signs' of the Medieval tradition to follow - no significant analogy exists to the linguistic signs used in human communication. Instead, they are a constituent part of the use and interpretation of language at the sophisticated level of inferential reasoning. This relationship between signs conceived by classical philosophers and inferential reasoning is illustrated by Sextus's criticisms of the Stoics' conception of what they called 'indicative' signs, signs which stand for what is unobservable, as sweating can be interpreted as a sign of invisible pores in the skin or behavior a sign of motions in the soul. These were contrasted with what were called 'associative' (or 'commemorative') signs. These latter are signs which stand for what is observable and for which there is a correlation in past experience between the sign and what it signifies, as illustrated by the observed correlation between smoke and fire, a scar and the wound that caused it, and a puncture of the heart and impending death. It is only the associative in Sextus's view that should be included as signs. He argues that a sign should be capable of being interpreted in a uniform way by all that observe it, but the indicative signs of the Stoics are interpreted in many different ways, with many different hidden causes assigned by physicians to symptoms where no observable correlation exists: 'The symptoms of fever, the flush, the moisture of the skin, the high temperature, the rapid pulse, when observed by doctors of the like constitution, are not interpreted in the same way.,l0 This controversy between the Stoics and Skeptics can be seen to involve contrasting views of scientific causal explanation, with the sign playing the role of the explanandum (thing to be explained) and what it signifies stated as a premiss in the explanation. Sextus admits only generalizations established by induction from observations as premisses of an explanation. The presence of smoke can be explained by fire because we have observed a past correlation between the two. The Stoics, on the other hand, were willing to admit also theoretical explanations of events in which unobservables were posited. To explain sweating

History of semiotic

17

is to posit invisible pores in the skin, and in this sense the sweating is a sign of the pores. Again, the nature of examples appealed to, and especially the use of the medical symptom as the paradigm, requires sign interpretation to be understood as the use of an inference in which the sign is now reported in the conclusion. The question whether Y is what is signified by sign X becomes the question whether Y can be used to explain X as the premiss of an inference. In none of these early discussions by classical philosophers do we find attempts to include within the extension of the term 'semeion' words or sentences as linguistic signs. In the fragmentary writings of Parmenides there even seems to be a contrast drawn between signs (semeia) as reliable indicators of what they stand for and words (onoma) which are arbitrarily posited names introducing distinctions where none exist in the objective world. Thus, we have signs 'that Being is ungenerated and imperishable, whole, unique, immovable, and complete,' while the mistaken belief that 'things were born and now are, and from now on they will grow and will afterwards perish', arises from the act of introducing names for these things.l1 Sextus reports that the Stoics did include within the scope of their discussion certain conventional signs such as a torch signifying an approaching enemy or a bell signifying the selling of meat. These signs are 'determined, as they say, by the lawgivers and are in our power, whether we wish them to make known one thing or to be significative of many.'12 But while they are used for purposes of communication, they are not signs formed within a rule-governed language, and do not present exceptions to the general exclusion of the linguistic. It is a short step to move from signs such as the torch and bell to linguistic signs, but it is one taken only later by St Augustine and the Medieval logical tradition that followed him. Another apparent exception to this exclusion of the linguistic is presented in the opening passage of Aristotle's De lnterpretatione, a passage destined to have an important influence on later discussions: Spoken expressions are symbols [symbola] of mental impressions, and written expressions [are symbols] of spoken expressions. And just as not all men have the same writing, so

18

History of semiotic not all make the same vocal sounds, but the things of which [all] these are primarily the signs [semeia] are the mental impressions for all men. . . .13

This seems to state that verbal utterances are to be included within the class of signs, along with natural events such as clouds, smoke, and scars. Recall that behavior and facial expressions were regarded as signs of mental states, as laughter is a sign of joy. It seems to be a natural extension also to include a scream as a sign of experienced fear, and finally a verbal utterance of 'I am afraid' as a sign of this fear, as evidence from which we can infer an emotion which itself we do not directly observe. Finally, utterances of indicative sentences such as 'It is raining' can be regarded as a basis for inferring the speaker's belief and utterances of 'Close the door' for inferring his desire for the door to be closed. By using semeia interchangeably with symbola for spoken and written words as signs of the mental, Aristotle himself strongly suggests this interpretation, for this is the same term used in the Rhetoric to stand for necessary and probable signs as evidence. As Eco notes, however, Aristotle customarily refers to spoken and written words, not as signs or semeia, as he does at the end of the passage just quoted, but as symbols (symbola), arbitrarily instituted marks with significance. Because they are arbitrary they vary from one culture to another, and in this respect are unlike spontaneous cries and most facial and behavioral expressions of emotions. 14 Moreover, the analogy Aristotle is making in this De Interpretatione passage is that between the relationship of written to spoken words on the one hand and spoken words to mental experiences on the other. Now it is incorrect to say that written words are evidence of the spoken. Rather, we should say they replace spoken utterances, serving as more permanent substitutes which enable communication at greater distances, both spatially and temporally. If the analogy is to hold, then spoken words are indicated in this passage also to be a kind of replacement or substitute for the mental experiences, and not evidence for them. And just as the same spoken words can be replaced by a variety of written marks, depending on the culture's system of writing, so the same mental experiences as shared in common by all can be replaced by a

History of semiotic

19

variety of speech sounds. As we shall see in the next section, this is precisely the interpretation of this passage given by Ockham at the end of the Middle Ages. For the classical period, then, the paradigm sign was a medical symptom, a natural occurrence serving as evidence for what it stands for. The sign could signify what is in the past, present, or future, and as a necessary sign could guarantee this other object or event or make it more or less probable as a 'refutable' sign. Spoken and written words were termed 'symbols' and contrasted with signs, though they were like signs if, as one possible interpretation of Aristotle suggests, they could function as evidence of the mental states of those producing them. 2.2

Augustine and his successors

During the Middle Ages a very different conception of the sign emerged. As represented by the Latin 'signum' (plural, 'signa') as the translation of 'semeion', it now came to take on an extended generic sense that included both natural occurrences and linguistic expressions, distinguished now as 'natural' and 'conventional' signs. Increasing attention was focused on the latter, transforming the linguistic expression into the paradigm sign, and even by the time of Locke's Essay, leading to its supplanting the classical evidential sign as the paradigm sign. The writer generally acknowledged as responsible for bringing about this fundamental change was St Augustine. 15 He accepts the Epicurean view of the sign as a sensed particular which signifies that which is not presently sensed: 'A sign is something which is itself sensed and which indicates to the mind some thing beyond the sign itself. [Signum est quod et se ipsum sensui et praeter se aliquid animo ostendit. ]'16 Under this very general heading are distinguished two kinds of signs. Natural signs [signa naturalia] are those which, 'without any intention or desire of signifying [sine volantate atque ullo appetitu significandi], make us aware of something beyond themselves, as smoke signifies fire. >17 By using the Stoic example of smoke as a sign of fire Augustine demonstrates the link between his natural signs and the evidential signs of the Greeks. He seems to show no awareness, however, of the presence of linguistic mediation in the interpretation of such signs and the use of an inference noted by the Stoics.

20

History of semiotic

Contrasted to natural signs were signa data, signs given or produced. Such signs are: those which living creatures show to one another for the purpose of conveying, in so far as they are able, the motions of their spirits or something which they have sensed or understood. Nor is there any reason for signifying, that is for giving signs, except for bringing forth and transferring to another mind what is conceived in the mind of the person who gives the sign. I8 There is nothing in this characterization of signa data to warrant for them the customary label 'conventional signs.' The contrast to natural signs being drawn is between signs which are not intentionally produced for the purposes of communication and those that are. Signs produced with communicative intent, whether the intent is to express 'motions of spirits' or information about objects 'sensed or understood,' may not be conventional in the sense of conforming to a rule established within a linguistic community. A gesture made to communicate in a foreign country with someone not speaking one's language would seem to be a signum datum of the kind being defined here. But it may not be a conventional sign. Elsewhere Augustine characterizes the meaning of signa data as not given by nature, but arbitrarily stipulated and reached by consensus ('non natura, sed placito et consensione significandi,).19 This does characterize so-called 'conventional signs,' but the addition is not necessary to distinguish a sub-class of signs distinct from natural signs. We shall see later in Chapter 4 the importance of distinguishing these two aspects of communicated signs. Augustine is also responsible for a second innovation of fundamental importance. Recall Aristotle's analogy between the relation of written to spoken words and of spoken words to the mental states of the speaker. Augustine draws out this analogy to its logical conclusion. We must postulate, he concludes, 'mental words' as the mental correlates of spoken words, just as spoken words are the correlates of the written: The word heard sounding outside is the sign of the word which is luminous within, which is more appropriately called a 'word'. For what is brought forth by the mouth of the body is the

History of semiotic

21

utterance of the word [vox verbi]; and though this, too, is called a 'word', it is so only on account of that which it is being used to manifest externally. 20 The mental word or conception will be one which is common to all men, but for the purpose of 'transferring to another mind' the speaker must employ an arbitrarily selected conventional word unique to his linguistic community. If we identify these mental words with the mental states of the speaker, then it seems clear that spoken words will not signify or be evidence of these mental states. Rather, they will signify what the mental words do, that is, signify as equivalent public translations of private signs. William of Ockham clearly draws this implication towards the end of the Middle Ages. Ockham pays lip service to the evidential sign of the Classical period, giving the traditional examples of smoke as a sign of fire, a groan the sign of pain, and laughter of a 'certain interior joy.' These are distinguished from words as conventional signs whose meaning is established by 'voluntary institution. 121 But, like Augustine, Ockham's chief concern is with linguistic expressions and the relation between them and the thoughts or conceptions they express. This leads him to suggest a drastic revision of the natural-conventional sign distinction. It is now the mental conception itself which signifies naturally: The concept or the passion of the soul naturally signifies whatever it signifies [naturaliter significat quidquid significat]. The pronounced or written term, on the other hand, does not signify anything save by voluntary institution. 22 Ockham explicitly rejects the view that the spoken word or term stands for or signifies the mental concept or 'passion of the soul.' Instead, it signifies those same independent objects which the concept or 'mental word' signifies. The conceived term is an intention, or passion of the soul, which by its nature signifies or cosignifies something. . . . These [mental] words remain in the mind only and cannot be brought to light externally, although words, as signs subordinated to them, are pronounced externally .... Words are used to signify those same things that are signified by the concept of the mind, in such a manner that first the concept signifies

22 History of semiotic something naturally and secondly the word signifies the same thing.23 For Ockham, then, there is a second type of natural sign, Augustine's mental word, and with attention focused increasingly on linguistic signs, the primary natural-conventional contrast was between mental words and the spoken and written words by which they are made public. 24 The shift in focus towards the linguistic culminates in the restriction of the term 'sign' to linguistic expressions and their mental correlates in the writings of Hobbes and Locke. For Hobbes there are both private and public signs. Private signs are 'marks' (notae), that is, Augustine's mental words, which enable us to 'remember our own thoughts.' Public signs are 'signs by which we make our thoughts known to others.' 'The difference ... betwixt marks and signs is this, that we make those for our own use, but these for the use of others.'25 Both types of signs, public and private, are included in the scope of the branch of philosophy Locke later terms 'semiotic,' 'the business whereof is to consider the nature of signs, the mind makes use of for the understanding of things, or conveying its knowledge to others. ,26 Here the term 'sign' has virtually no recognizable affinity with the 'semeion' of the classical period, which had excluded from the domain of signs exactly what it is now being restricted to. There are uses of the term 'natural sign' in the period after Descartes in which the link to the early classical evidential signs is acknowledged, though in these the term is significantly modified and extended. Arnauld, for example, makes the traditional distinction between natural and conventional signs, but takes the former to stand in an iconic or picturing relation to what they signify, and distinguishes them by the presence or absence of a causal relation to their objects. Signs may be divided into natural signs and conventional signs. Natural signs do not depend on the whim of man fjantaisie des hommes] - an image in a mirror is a natural sign of the person mirrored. A conventional sign is a sign established by convention and may, but need not, have any connection with the thing signified. 27 The mirror image is the causal effect of the object mirrored or a

History of semiotic 23

process emanating from this object, just as smoke is the effect of fire and symptoms of some underlying disease. But unlike the classical evidential signs, we seem to infer from the mirror image to the object only in special circumstances, and Arnauld has thus introduced a novel category of natural sign. Maps and pictures are also listed by him as signs, but since they fail to stand in a causal relation to what they signify and depend on the 'whim of man,' they would seem to fall under the heading of his conventional signs, here widened also beyond Augustine's paradigm of the linguistic expression and the Stoics' torch and bell. Still a further extension of the natural sign was made later when sensations themselves were regarded as natural signs of the objects producing them, thus combining Augustine and Ockham's mental aspect with Arnauld's iconic. This conception is perhaps most clearly stated by Reid in the eighteenth century. There are for Reid three categories of natural signs. The first includes the Stoic evidential signs, 'those whose connection with the things signified is established by nature, but discovered only by experience. ,28 But while the classical evidential sign was typically the effect of a cause, Reid regards interpretation as predictive: 'What we commonly call natural causes, might, with more propriety, be called natural signs, and what we call effects, the things signified.' Reid's second category includes signs of human thoughts, purposes, and desires which are part of the 'natural language of mankind' prior to the institution of 'artificial language.' For all artificial languages presuppose some compact or agreement to affix a certain meaning to certain signs; ... but there can be no compact or agreement without signs, nor without language; and therefore there must be a natural language before any artificial language can be invented. 29 Signs within this 'natural language' are said to be 'modulations of the voice, gestures, and features.' Finally, Reid adds as a third category of signs those whose signification is determined, not by prior experience of correlations, but by a kind of innate mechanism which suggests what is signified. These are signs which,

24

History of semiotic

though we never before had any notion or conception of the things signified, do suggest it, or conjure itup, as it were, by a natural kind of magic, and at once give us a conception, and create a belief in it. Included in this third category are sensations as signs of external objects, even though, like the Stoic indicative signs, the connection between sensations and the conception and belief of external existences, cannot be produced by habit, experience, education, or any principle of human nature that hath been admitted by philosophers. They are thus to be distinguished from signs of the first category. The later controversy between realists and idealists then became one regarding the status of natural signs. Realists admitted signs of Reid's third category, though differed about what they signified, with some (including Reid) holding that they are signs only of the bare existence of objects, others that they picture certain specific aspects of these objects. Idealists such as Berkeley and perhaps Hume, on the other hand, admitted signs only of the first category. They held that we do not perceive objects, but only sensations, and a sensation such as that of the sound of a coach is a sign of the coach only in the ~ense that in prior experience the sound and the sight of the coach have been correlated. Thus, in 'truth and strictness' for Berkeley, 'nothing can be heard but sound; and the coach is not then properly perceived by sense, but suggested from experience.'3o Rather than being signs of objects, sensations are signs of other sensations with which they have been previously correlated. By the beginning of the nineteenth century Augustine and his modern successors had thus bequeathed a motley variety' of natural signs with differing relations to the so-called 'conventional' signs with which they were contrasted. Still retained were the classical evidential signs, illustrated by the same examples as those of the Greeks. But in addition there were also mental conceptions or 'mental words,' mirror images as iconic signs, and finally sensations as the 'internal' mental signs of the 'external' objects that cause them. With Augustine spoken and written words as the basic elements from which sentences are formed became the paradigm signs. But as the variety of natural signs

History of semiotic 25

increased, it became increasingly difficult to find common features that warranted applying the same term 'sign' to both these and the signs used in human communication. 2.3

Peirce and Saussure

The influences of these earlier views and many of their resulting confusions are evident in the writings of the philosopher generally credited to be the founder of modern semiotic. For Charles Peirce the term 'semiotic' is applied to what he calls the 'quasi-necessary, or formal, doctrine of signs.' By describing the doctrine as 'quasi-necessary,' or formal, I mean that we observe the characters of such signs as we know, and from such an observation, by a process which I will not object to naming Abstraction, we are led to statements, eminently fallible, and therefore in one sense by no means necessary, as to what must be the characters of all signs used by a 'scientific intelligence', that is to say, by an intelligence capable of learning by experience. 31 The empirical studies by linguists of human language and ethologists of animal communication systems have as their aim the description of signs as actually used and interpreted. In Peirce's view semiotic as a branch of logic and philosophy has the same observational basis as do these related empirical sciences, but it aims at singling out necessary, as opposed to contingent, features of signs interpreted by creatures capable of learning. Just as a logical analysis of a sentence differs from a grammatical description that will vary from language to language, so too the semiotic description of a sign will differ from any empirical description. It is obvious from this definition of semiotic that Peirce intended its scope to extend beyond the linguistic signs used in human communication. This is also shown by the generality of his definition of a sign or 'representamen' as 'something which stands to somebody for something in some respect or capacity.m Under the heading of a sign as so defined Peirce includes single-word sentences such as 'Red' accompanied by a gesture, sentences with a subject-predicate structure, and even inferences consisting of several sentences, one of which, the conclusion, is inferred from

26 History of semiotic

the remaining as premisses. But also included are several kinds of non-linguistic signs. Among them under the heading of what he terms an 'index' are the examples of classical evidential signs where the sign is the causal effect of the object it is said to represent, e.g. the bullet hole as a sign of the bullet, the weathervane as a sign of wind direction, the falling barometer as a sign of coming rain. Also included under the heading of 'sign' are Arnauld's iconic signs which are related to their objects by similarity of structure, e.g. a painting, map, or diagram. But though this very inclusive scope is intended by Peirce, his characterizations of signs seem often to show that the paradigm sign is in fact for him the sentence with subject-predicate structure. The sign, he tells us, can only represent the Object and tell about it. It cannot furnish acquaintance with or recognition of that Object; for that is what is meant ... by the Object of a Sign; namely, that with which it presupposes an acquaintance in order to convey some further information concerning it. 33 Neither a bullet hole nor a map would seem to represent in the way that is being described here, for they need not convey information about a particular object (a bullet or a town) with which the interpreter has some prior acquaintance. The bullet hole stands for some bullet or other, the dot on the map (without a name as label) for some indefinite town. In contrast, the predicate of a sentence such as 'John is tall' does convey information about an object with which the interpreter is assumed to have prior acquaintance; only with this can the referent of 'John' be identified. Thus, the passage seems directed towards the sentential paradigm, not towards signs in general. This divergence between a general characterization extending to the traditional natural signs and one that is language-specific is reflected also in Peirce's treatment of two of his three major categories of signs, indices and symbols. 34 An index, he says, 'is a sign which refers to the Object that it denotes by virtue of being really affected by that Object. ,35 Besides bullet holes, weathervanes, barometers, etc. Peirce includes under this heading photographs which, though they resemble their objects, and in this respect are like icons, have been 'produced under such circumstances that they were forced to correspond point by point

History of semiotic 27

to nature,' and are thus indices. An icon such as a diagram or map, in contrast, is produced in Arnauld's words by the 'whim of man.' Under this definition of an index its relation to its object is a dyadic causal one which would exist whether or not there were someone present to interpret it. 36 But Peirce also characterizes the index in a way that does necessarily involve an interpreter and that seems formulated with the sentence again in mind. 'Anything which focusses attention,' he says, is an index, and he gives as examples the demonstratives 'this' and 'that' as they occur in the subject position of a sentence. Such indices are not themselves the causal effect of objects, but devices which he says enable their interpreter to 'place himself in direct experiential or other connection with the thing meant. >37 A similar divergence of application arises in Peirce's treatment of a symbol. This is characterized as a sign which signifies by virtue of being an instance or token (or 'replica,' as he sometimes calls it) of a type (or 'legisign'). Thus, a verbal utterance or written inscription of the word 'red' signifies as a token of a type of expression. In this respect symbols differ from indices as particular effects of the objects they represent and icons as particular sensible images similar to their objects. This aspect of the symbol is captured in the following definition: A Symbol is a sign which refers to the Object that it denotes by virtue of a law, usually an association of general ideas, which operates to cause the Symbol to be interpreted as referring to that Object. It is thus itself a general type or law, that is, is a Legisign. As such it acts through a Replica. 38 In this general sense of a sign which signifies by virtue of 'an association of general ideas,' the symbol would seem to have a scope far beyond linguistic signs, including, for example, even the flash of lightning which stands for thunder by virtue of associations in past experience. It would seem also to apply to the crack of a whip as a sign of running for the horse, the example used by the Epicureans in their rejection of the Stoic conception of the sign. 39 Both signify as tokens or replicas of general types of events as the result of prior learning, and are neither indices nor iconic representations. But Peirce also defines the symbol in a way more consistent with the traditional sense bestowed by Aristotle of a conventional sign whose meaning is determined by

28

History of semiotic

a linguistic rule: 'A Symbol is a Representamen whose Representative character consists precisely in its being a rule that will determine its Interpretant. All words, sentences, books, and other conventional signs are Symbols. ,40 As for the index and the sign in general, Peirce fails to adopt consistently a general definition which is then applied to the linguistic as a special case. Instead, he vacillates between an account applying only to linguistic signs and one with considerably wider scope. Peirce has relatively little to say about the use of a sign by someone to communicate with another. It is perhaps for this reason that he fails to repeat Ockham's view that a spoken utterance or written inscription is a public manifestation of a private mental sign. But he does restate this doctrine in terms of the effect of a sign on one who interprets it. The 'logical interpretant' of a sign, he says, is the effect the sign produces in its interpreter, or its 'significate outcome.' This effect can be a thought, which he equates in common with Augustine and Ockham with a 'mental sign,' which represents the same object as does the original sign it translates. In this case it can also be interpreted and produce another mental sign, which in turn can be interpreted. But there is another type of effect that a sign can produce, and one that allows this process of sign translation to eventually terminate. Shall we say that this effect [produced by a sign] may be a thought, that is to say, a mental sign? No doubt, it may be so; only ... it must itself have a logical interpretant; so that it cannot be the ultimate logical interpretant of the concept. . . . It can be proved that the only mental effect that can be so produced and that is not a sign but is of general application is a habit-change. 41 This view that a habit is induced as the 'mental effect' of a sign, was destined to have, as we shall see in the next section, a decisive influence on behavioral formulations of semiotic that developed later. After Peirce the study of signs took two very different directions, with the divisions marked largely along geographic lines. In Continental Europe the science of signs became known as 'semiology' and came to have an anthropocentric orientation. In the United States the key terms of semiotic were defined

History of semiotic 29

relative to the concepts of the emerging science of comparative animal psychology. The initial formulation of the program of semiology was provided by Ferdinand de Saussure. Semiology is to have as its subject matter, Saussure says, all the devices used in human society for the purposes of communication, including both linguistic expressions and non-linguistic devices such as gestures and signals within non-linguistic codes. Semiology in this conception is an empirical science, with linguistics a sub-division dealing with language as a special means of human communication: Language is a system of signs that express ideas, and is therefore comparable to a system of writing, the alphabet of deaf-mutes, symbolic rites, polite formulas, military signals, etc. But it is the most important of all these systems. A science that studies the life of signs within society is conceivable; it would be a part of social psychology and consequently of general psychology; I shall call it semiology (from the Greek semeion 'sign'). Semiology would show what constitutes signs, what laws govern them .... Linguistics is only a part of the general science of semiology; the laws discovered by semiology will be applicable to linguistics, and the latter will circumscribe a well-defined area within the mass of anthropological facts. 42 Excluded by Saussure from semiology are the traditional natural signs and Peirce's indices, as well as signs used for communication by lower organisms. In another departure from Peirce's semiotic as a 'quasi-necessary' science studying features that must obtain of signs used by interpreters 'capable of learning by experience,' semiology restricts itself to features that in fact obtain of devices used in human communication. Logical analysis is thus abandoned in favor of empirical description characteristic of such sciences as sociology and anthropology. Later European writers have accepted the main outlines of Saussure's program, with differences arising chiefly over its scope. All seem to agree in excluding natural signs. Guiraud, for example, excludes 'natural indications' such as clouds as signs of rain on the grounds that 'the cloud-laden sky has no intention of communication. ,43 There seems also to be general agreement that signs used for communication among lower animals should also

30

History of semiotic

be excluded, even though there may be reasons for believing they are used in some species of primates with communicative intent. Semiology is conceived as having an exclusively human orientation. Finally, there is agreement with Saussure that the methods of linguistics are to provide the model for studying non-linguistic modes of communication. In particular, the attempt is made to identify meaningful units analogous to the morphemes of human speech and rules analogous to linguistic syntactic and semantic rules by which such units are combined to form complex wholes. 44 Not aU writers have followed, however, the project of Roland Barthes and others to extend semiology to include complex cultural forms of communication, including film, theatre, music, dance, literature, architecture, advertising, and even food, clothing, and perfume by which persons indicate social status, sexual availability, etc. Such forms of expression have in Barthes's view a linguistic background constituted by areas which define the elements of the artistic or personal code being studied. It is 'the great signifying unities of discourse lIes grandes unites significantes du discours], providing this background that Barthes takes as the subject matter of semiology, and he claims it on these grounds to constitute a sub-division of linguistics. 45 The effect of making this extension to such a complex subject matter is to make it difficult to apply the methods of such sciences as linguistics, sociology, and anthropology, and this appears to threaten the goal of constructing a discipline in which consensus can be gained by co-operating investigators. Semiology becomes a vehicle for conveying personal interpretive insights rather than either Peirce's logical, philosophical study or the empirical science envisaged by Saussure. 2.4 Behavioral semiotic A very different direction was taken by those writers who took as their project the extension of the methods of behavioral learning theory to the study of signs. In early formulations the paradigm sign became the conditioned stimulus of the conditioned reflex learning experiments, e.g. the bell paired with the unconditioned stimulus of food particles on a dog's tongue evoking the reflex response of salivation. The bell as an acoustic stimulus becomes a sign of the food as a tactual stimulus so far as it comes to evoke

History of semiotic 31 by itself at least a part of the response previously evoked only by the food. One of the earliest formulations of this approach was given by Ogden and Richards in their The Meaning of Meaning. It is a feature of the interpretation of a sign, they say, . . . that when a context has affected us in the past the recurrence of merely a part of the context will cause us to react in the way in which we reacted before. A sign is always a stimulus similar to some part of an original stimulus and sufficient to call up the engram formed by that stimulus. An engram is the residual trace of an adaptation made by the organism to a stimulus. 46 In the classical reflex learning experiments the paired bell and food constitute the whole 'context' and the bell on a given occasion is similar to some 'part' of this whole. As a sign the bell 'calls up' the response or 'residual trace of an adaptation' to the whole stimulus. The attempt is then made by these writers to extend this analysis of primitive signs to the interpretation of linguistic expressions. Behavioral semiotic was an attempt to transform semiotic into an empirical science by proposing operational definitions of such traditional terms as 'meaning', 'denotation, and 'truth' which enable the process of sign interpretation to be investigated in terms of correlations between stimuli and responses as publicly observable events. Of these the term 'meaning' or 'significance' was of central interest, and became the focal point for the project of correcting what was regarded as the 'mentalistic' bias of traditional semiotic. In Ogden and Richard's version, what we shall refer to as the 'direct response theory,' the significance of a sign was explicated in terms of a given type of response to a type of stimulus. If we use XY for the entire context of which the sign X is one part and Y is the unconditioned stimulus evoking in isolation the response R the other, then Y can be said to be the significance of X for an interpreter I, according to this theory, if and only if X by itself evokes a response R' in I similar to or part of the R evoked by Y. Thus, the tactile stimulus of food particles on the dog's tongue is the significance of the bell for the dog in the classical experiments, since the bell in isolation evokes a salivation response similar to that evoked by the food. The direct response theory then attempts an extension of this

32 History of semiotic

analysis to the significance of sentences for language users. Suppose that the visual stimulus correlated to an apple evokes a salivation response in a person. Then if the single-word sentence 'Apple' evokes a similar response, this visual stimulus is now regarded as the significance or meaning of the sentence. 47 In this form the theory seems to be implausible, since for few of the sentences a person hears and understands are there the overt responses R' it requires. It is highly unlikely that for a given person hearing 'Apple' there will be a detectable salivation response. Later versions of the direct response theory, the most influential of which is that formulated by C.E. Osgood, attempted to remedy this difficulty. For Osgood the responses evoked by signs are internal 'representational mediational responses,' neurophysiological in nature and detectable therefore only by special forms of observation. If a sign X causes an internal process R' in a subject I which is similar to a process R caused by a type of stimulus Y with which it has been paired, then Y can be said to be the significance of X for I. 48 But as critics of this version have noted,49 there are formidable difficulties it must also face. First of all, there seems to be no experimental evidence that sentences as linguistic signs cause internal processes similar to those caused by the stimuli with which they are correlated in language learning, and the 'mediational' processes seem thus to be simply postulated as devices for saving the theory. Second, and more important, there seems to be no criterion for deciding whether one internal process is similar to another in the way required by the theory. Everything is similar in some respect to every other. Without a specification of that respect in which process R' is to be claimed as similar to R it would seem to be impossible to establish empirically the significance of a sign X. Yet it is precisely this sort of empirical determination that is the avowed central goal of behavioral semiotic. As an alternative to defining significance in terms of a direct response to the sign itself, several philosophers have proposed an alternative that we shall refer to as the 'dispositional theory.' According to it a sign's significance is to be determined by a disposition of its interpreter to respond in certain ways to some subsequent stimulus. In its original formulation by Charles Morris the sign is defined as a 'preparatory stimulus' whose

History of semiotic

33

significance is a type of stimulus Y if it causes a disposition in its interpreter to respond in a way similar to that with which it responds to Y in isolation. 50 The disposition that is induced can be regarded as Peirce's habit change as the sign's 'logical interpretant.' Thus, a buzzer is a sign of food for a dog if it causes the dog to search for food and if on finding it at a given location the dog eats the food. The properties of the food evoking this subsequent eating response are regarded as the buzzer's significance. 51 This account is fraught with difficulties, as the responses that Morris requires include those to the sign itself (searching for food) and a subsequent stimulus (the visual stimulus of the food). The complexity of this response, the differing ways it can be interpreted, and again the problem of establishing criteria for similarity of responses - all of these factors ensure that this formulation can never be fruitful as a basis for experimental studies of the process of sign interpretation. Later formulations of the dispositional theory by Carnap 52 and Quine53 directed towards language interpretation, however, offer more promise. In them the significance of a sentence for a person is defined in terms of the disposition of that person to respond when queried by either 'Yes' or 'No' when presented with a given type of object. The subsequent response is in this manner simplified as one of two readily identifiable alternatives. Quine states his version in terms of the problem of determining the meaning of expressions in an alien language. Suppose a linguist were confronted with the problem of determining the meaning of a single-word sentence such as 'Gavagai' and had access to a native speaker. Then, by successively pointing to different kinds of objects and asking 'Gavagai?' the linguist should be able to solve his or her problem by noting those objects which evoked an affirmative response from the native (e.g. a nod of the head) and those which elicited a negative response. The 'affirmative stimulus meaning' of the single-word sentence 'Gavagai', Quine says, is simply that type of visual stimulus that evokes an affirmative response, while the 'negative stimulus meaning' is that which evokes a negative response. If on varying the objects and their properties and noticing that the informant consistently responds affirmatively to rabbits and negatively to other objects, the linguist can hypostatize that 'Gavagai' in the native language is synonymous with 'Rabbit' in English.

34

History of semiotic

Difficulties confront this theory as a successor to either the direct response theory or Morris's alternative. One is that it will be difficult to identify a given gesture as an affirmative or negative response; such gestures are conventional ones which can vary markedly from one culture to another. Another difficulty is that the affirmative and negative responses by informants are voluntary responses that can be withheld. Informants certainly have the capacity to distinguish between objects to which an expression applies and those to which it does not, but whether this can be termed a disposition to respond to given types of stimuli is questionable. Certainly the term 'disposition' is being applied in a way different from that in which it is applied to the elasticity of rubber, the brittleness of glass, etc., and this difference is left unexplained. Some may want to interpret 'disposition' as synonymous with 'tendency', and claim that it only implies a probability that a response will be elicited by a given stimulus. But this then leaves it totally mysterious as to why the response is forthcoming on some occasions, while absent on others. Quine's theory differs from his predecessors' in being restricted to language interpretation. After the dog hears the buzzer there may be a detectable affirmative response when it later sees the food (e.g. muscular relaxation or an identifiable neurological process) and a negative resonse (e.g. muscular tension) when the food is absent at a given location. If so, the significance of the buzzer for the dog could conceivably be defined in terms of the types of visual stimuli that evoke these responses. But the task of identifying both the stimuli and the affirmative and negative responses is a formidable one, and at present there is only a hope that it could be solved. 54 More fundamental objections to the dispositional theory have been raised by Chisholm, and are directed towards what appear to be irreducibly intentional aspects of sign interpretation. 55 Consider a native informant of whom the linguist asks the question 'Gavagai?' and points to a brownish, furry object that is not a rabbit but instead a squirrel. The informant, however, mistakenly believes it is a rabbit and hence responds in the affirmative. From the native's response the linguist cannot infer that the visual stimulus from the squirrel is the affirmative meaning of 'Gavagai' or that the word is synonymous with the English 'Squirrel'. Hence, the possibility of error seems to rule

History of semiotic 35

out the dispositional theory. Of course, normally the informant is correct, and after several repetitions the linguist should be able to translate the word correctly. But then we must revise our statement of the theory and say that the language disposition being characterized is only a tendency for assent or denial relative to types of stimuli. As before, there is the problem of explaining divergences of response. Negation poses a parallel problem. The negative stimulus meaning of a sign is for Quine constituted by those stimuli which evoke a negative response. But it seems possible for the linguist to point to a blank space with only background 'noise' impinging on the native's sense receptors. Here it is the absence of any stimuli which would evoke the negative response, not some stimulus other than that correlated with rabbits. The native has the capacity to recognize the absence of a rabbit at the location pointed to, but this capacity cannot be explicated in terms of a disposition to respond to a specific type of stimulus. So far we have been discussing signs conveying information. For signs prescribing actions, e.g. the Epicurean example of the crack of the whip as a sign of running for the horse or the verbal commands 'Run' or 'Go', a version of the direct response theory would seem applicable, and the significance of the sign defined as the response it elicits as its effect. 56 But the obvious objection to this theory is that a person may disobey a given command, and even the horse may not run at the sound of the whip. The sign may have a certain action as its significance, though no response is observed. Again, for such a sign to have significance for an interpreter I there must be a capacity on I's part to recognize what action constitutes obedience to it. The behavioral theories considered above have all been attempts to formulate empirical tests for determining the significance of a sign for an interpreter. B.F. Skinner has developed a version of behaviorism directed towards the production of signs by a communicator. The model for him is provided by instrumental learning, learning in which an organism's response to a given stimulus are determined by a schedule of positive or negative reinforcements. One standard experimental situation is that in which an animal learns to press a bar on seeing a flash of light if the pressing is positively reinforced by the reward of a food pellet. The light is the 'controlling stimulus,'

36 History of semiotic while the learned response to this stimulus is the 'operant.' The animal learns to discriminate that controlling stimulus (e.g. the color of the light), which, when followed by a given operant, is reinforced from other stimuli for which reinforcement is absent. Skinner then extends the analysis of this primitive model to language learning. For descriptive words (e.g. 'red' or 'book'), or what he terms 'tacts,' learning occurs when verbal responses as operants, that is, utterances of the word, are selectively reinforced in the presence of controlling stimuli. 'A child is taught the names of objects, colors, and so on when some generalized reinforcement... is made contingent upon a response which bears an appropriate relation to a current situation.'s7 For prescriptions or 'mands' (e.g. the commands 'Go!' and 'Run!'), learning takes places in the absence of a definite controlling stimulus, with reinforcements in the form of obeying the command shaping the child's verbal behaviour. Skinner's variant of behaviorism has had little direct influence on philosophic formulations of semiotic. It did inspire, however, an attack on the entire behavioral program for analyzing language learning that was to have considerable influence. Behavioral theories, especially those of Morris and Skinner, have had an important role in establishing the conceptual framework for behavioral studies of animal communication by ethologists such as Tinbergen, Lorenz, and Thorpe. Sign use and interpretation in this domain is studied in what Sebeok terms 'zoosemiotics,' as contrasted to 'anthroposemiotics,' the study of signs used in human communication, an area including linguistics, paralinguistics, the study of non-verbal behavior accompanying language use (e.g. pointing gestures, voice pitch), and kinesics, the study of non-verbal behavior used independently to communicate (e.g. bodily movements, facial expressions).s8 2.5

Semiotic's critics

From this brief historical survey we can distinguish three main versions of semiotic, with variants for two of them. The first is the Greek conception of it as the theory of evidential signs, with the paradigm being the medical symptom as evidence of a disease as its cause. This should be regarded only as a stage on the way to the conceptions that followed, and not having a major

History of semiotic 37

influence on recent discussions. 59 The second conception is that of Locke for whom, as we have seen, semiotic is the theory of language as the primary means of communication, a conception brought about by Augustine's shift to the word as the paradigm sign. As a variant of Locke's conception we have the extension of the term 'sign' by Saussure and his followers in the semiological tradition to non-verbal means of human communication and the forms of discourse they presuppose. And finally, we have Peirce's conception of semiotic as the study of necessary features of signs used by creatures capable of learning from experience, including signs used for the purposes of communication and natural events that are not. The precursor of this conception was the Medieval application of 'sign' to both 'natural' and 'conventional' signs. A variant of Peirce's conception is found in attempts to analyze sign interpretation and use in terms of the controlled learning situations with animal subjects studied by behavioral psychologists. Signs to which this behavioral variant can be extended include those used in animal communication, the subject of Sebeok's zoosemiotics. We shall understand in this work the term 'semiotic' to stand for the discipline delimited by Peirce, with its subject including linguistic and non-linguistic signs, and including natural events not produced for the purposes of communication. We shall also follow Peirce in understanding the term to apply to the logical study of necessary features of signs as interpreted and used, thus distinguishing it from those related empirical sciences which describe contingent features which observed signs happen to have. By adopting Peirce's conception we distinguish semiotic both from what is known as the 'philosophy of language,' the modern term for Locke's semiotic, and from semiology. Any such selection from among alternatives is arbitrary, but at least ours can be justified by the continuous history of Peirce's inclusive conception from St Augustine until the present. Why has this discipline received so little attention from English-speaking philosophers in recent years? Its neglect seems to have been based on several important methodological concerns which serve to throw into question the possibility of semiotic as a general theory of signs. The first of these is directed towards any generic definition of a sign that is constructed with sufficient generality to include linguistic and non-linguistic signs.

38

History of semiotic

Recall Peirce's general definition of a sign as 'something which stands to somebody for something in some respect or capacity.' For a sign to stand for something or for it to mean something for an interpreter would seem for it to stand in a relationship that can vary markedly, depending on the type of sign being considered. Gilbert Harman states this objection as follows: Smoke means fire and the word 'combustion' means fire, but not in the same sense of 'means.' The word 'means' is ambiguous. To say that smoke means fire is to say that smoke is a symptom, sign, indication, or evidence of fire. To say that the word 'combustion' means fire is to say that people use the word to mean fire. Furthermore, there is no ordinary sense of the word 'mean' in which a picture of a man means a man or means that man. This suggests that Peirce's theory of signs would comprise at least three different subjects: a theory of the intended meaning, a theory of evidence, and a theory of pictorial description. There is no reason to think that these theories must contain common principles. 60 If there are no common principles relating the smoke, the word

'combustion', and the picture of a man as signs, then semiotic as the theory of such principles would seem to be impossible. This same criticism can be directed towards Karl Buhler's definition of a sign as a 'representative' ('Stellvertrendes') which 'stands in for something or someone,' and is applied to linguistic expressions, an actor portraying a character in a play, a lawyer representing his client, and a canvas picturing a landscape. 61 It is again difficult to see how such diverse senses of 'standing in for something' can exhibit common features which constitute a domain for study. This reasoning can be extended to attempts to frame a sign theory that includes language and animal communication systems. Chomsky has criticized such attempts for arriving at a level of generality which must include much that is noncommunicative and bears no obvious analogy to language. Suppose we define a communicated sign as that which is purposive, exhibits organization of elements or a syntax, and is informative. Then, Chomksy argues, this definition can be applied to a behavioral activity such as walking, for it is purposive or goal-directed, and there is an ordering among the separate movements, the flexing of muscles, positioning of the

History of semiotic 39

heel, etc. It is also informative, for the rate at which a person walks can indicate to another how strong is his want for the goal towards which he is directed. But there is obviously no significant analogy between walking and a sentence used to communicate. 62 The similarities between semiotic and that pseudo-science known as 'general systems theory' that enjoyed a brief vogue in the 1960s and 1970s, and the type of enthusiasm both have inspired, should serve as an ominous warning. The latter theory proposes a definition of what is called a 'system' that is general and indefinite enough to be applied to humans, lower animals, a lake (an 'ecosystem'), the solar system, atoms ('systems' with a nucleus and orbiting electrons), and even a galaxy. This variety and scope of applications may serve to evoke in its adherents attitudes of wonder and awe. But no common features seem to hold of the types of objects called 'systems' that have justified the establishment of a science with identifiable problems, nor has there been the development of a method for reaching that consensus on conclusions essential for progress in any discipline. In the same way nearly everything within a suitable context can be classified as a 'sign' if the term is defined with sufficient generality. But it is unlikely that this chaotic variety will be one for which significant problems are formulated and solved by means of an agreed upon method. Another major reason for the rejection of semiotic is the belief that linguistic signs possess unique features that resist attempts at assimilation to signs of more primitive types. The most wellknown statement of this view is provided by Chomsky's criticism of the analysis by Skinner of language acquisition in terms of instrument learning outlined at the end of the previous section. The extension of such terms derived from animal learning as 'controlling stimulus', 'operant response', and 'reinforcement' to language acquisition, Chomsky contends, is a 'metaphorical extension' or a series of 'analogic guesses' which only serve to obscure the important differences between the two types of learning. What is created is only the 'illusion of a rigorous scientific theory with very broad scope. ,63 Also criticized have been the attempts of the semioticians to extend terms from their application to sentences to non-linguistic signs. Austin criticizes the extension of 'truth' to pictures and the classical natural sign, e.g. smoke as a sign of fire:

40

History of semiotic

A picture, a copy, a replica, a photograph - these are never true in so far as they are reproductions, produced by natural or mechanical means: a reproduction can be accurate or lifelike ... , but not true (of) as a record of proceedings can be. In the same way a (natural) sign of something can be infallible or unreliable but only an (artificial) sign for something can be right or wrong. 64

It is only an utterance or inscription of a sentence as a conventional sign which can be said to be 'true.' To extend the term beyond this application is again to ignore the unique features of language. This criticism can be extended to Morris's attempt to apply the term 'true' to a sign such as the buzzer interpreted by the dog. 'From the point of view of behavior,' says Morris, 'signs are "true" in so far as they correctly determine the expectations of their users. ,65 But for such signs, as for the picture or photograph, there are no rules governing their use, and hence conditions for applying such normative terms as 'true' or 'correct' seem to be absent. Just as suspect is Morris's application of the logical term 'denotation' to primitive natural signs. 66 The subject of a sentence such as 'John' in 'John is tall' can be used to refer to an object on a given occasion, but a natural sign with no internal structure such as the buzzer interpreted by the dog has, as we shall we see in the next chapter, very different features. These are important criticisms which cannot be simply dismissed, and indeed are indicative of serious difficulties in early expositions. One of these arises from the controversy over the status of the sign between the Stoics and Epicureans noted in Section 2.1, with the Stoics holding that signs as evidence are premisses of an inference. Another is suggested by the vacillation by Peirce between definitions for signs with potentially broad scope and those restricted to the sentential paradigm. Finally, through the influence of Morris semiotic has been identified by many with the behavioral variant of Peirce's conception of the subject. The difficulties of extending a terminology derived from experimental situations involving animal learning to human language use are thus regarded as fundamental to semiotic. To the extent that contemporary semiotic has perpetuated such difficulties its critics are on strong ground.

History of semiotic 41

In the chapters that follow I attempt to provide two remedies to these difficulties. The first of these is to specify the differences as well as similarities between the various levels in a more careful way than has been done before. The alternative to a vacuous generic conception of a sign is to restrict ourselves to simply describing the features which hold of signs as interpreted natural events, primitive signs used in communication, and sentences as the complex signs used in communication, with the goal of isolating features which signs at these different levels must have. We impose no requirement from the beginning of our inquiry that such comparisons will lead to some feature or set of features shared by signs at all levels. The second remedy is to restrict the term 'sign' more narrowly than has been the practice, with only those objects of interpretation which bear important analogies to linguistic expressions included within its scope. As we shall see in the next chapter, this requires excluding the classical evidential signs, the 'natural signs' of the Medieval tradition, and also certain physical and mental images, e.g. photos, mirror images, and Reid's sensations as signs. It will also require, as we shall see in Chapter 4, that only a very selective sub-class of animal signals be listed as communicated signs. Finally, I concede to semiotic's critics the inadequacy of behavioral formulations that extend analysis of animal learning to language use. There are important analogies between the capacity of a lower animal to associate an event in its environment with either another event or an action and the human capacity to interpret linguistic expressions. But given the complexity and variability of human behavior, behavioral reductionism has proven an unsuccessful method for specifying these analogies, and I make no attempt to provide still another version. A more fruitful method, as I hope to be able to show, starts with the logical analysis of necessary features of language and extends the results of this analysis to features of more primitive signs encountered in human experience by abstracting from language's special features. This analogical extension of logical features is designed to replace the vague metaphors derived from animal learning experiments of the kind criticized by Chomsky. All can agree that there is a neurophysiological basis to the capacity of all organisms, human and non-human, to interpret signs. But reductionist accounts of logical terms such as 'significance',

42

History of semiotic

'denotation', and 'truth' will not advance our scientific understanding of the physical structures and processes that provide this basis. These terms have their own appropriate application to language use as described by those who are engaged in it. Their application to more primitive signs can only be derived on the basis of our understanding of this primary use by extending the standard logical analyses of language, not by inventing for the terms entirely new applications.

3

NATURAL SIGNS

We turn now to our central task, that of specifying the essential features of signs of differing levels of complexity. We begin with the simplest level, that of natural signs as objects or events in the environment of their interpreter which are not produced by some agent with the intent to communicate. Our historical sketch has shown a great variety of signs to have been classified under this heading, including the classical evidential signs, material images, and sensations. We now look more critically at these classifications with an eye towards developing a conception of a natural sign for which important analogies to linguistic signs hold. We thus consider evidential signs in the first section of this chapter and images, both material and sensory, in the second. In the final two sections are outlined the essential features of a severely restricted range of signs that we label 'natsigns' and distinguish from the natural signs of the tradition. 3.1

Signs and evidence

We have already discussed in Section 2.1 the basic features of the classical evidential signs. What is signified by such signs is thought to occur either in their past, be contemporaneous with them, or occur in their future. When it lies in the past the sign seems invariably to be regarded as an effect and what it signifies the sign's cause, as the scar is said to be a sign of the past wound that caused it. If we include such signs of what lies in the past under the heading of what Peirce terms an 'index', e.g. a fossil as a sign of past life, a bullet hole as a sign of a fired bullet, and boulders as a sign of past glacial activity, 1 then the cause can lie in the remote past of the sign which is its effect. Indeed, the temporal interval between them can span millions of years. Where the sign is contemporaneous with what it signifies the relation again is normally understood to be causal, though 43

44 Natural signs

examples have been given where this is not the case. Certainly medical symptoms of a disease (e.g. fever and vomiting as symptoms of influenza, spots as a symptom of measles) and smoke signifying fire are causal effects of that which they signify. But Peirce also gives the example of man's rolling gait being a sign he is a sailor (see note 5 of 2.3) and classical writers spoke of flushing as a sign of shame or laughter as a sign of joy. For such examples the term 'sign' seems to be used in the sense of 'criterion for': the rolling gait constitutes at least part of the criteria we use in identifying an individual as a sailor; the blushing is part of the criteria used in describing a person as ashamed. Being a sailor does not cause the rolling gait, nor does shame as a mental state cause the blushing. 2 Where what is signified is to occur in the future, again the relation is normally, but not necessarily, causal, with the sign now the cause and what it signifies or means its effect. As for signs of past events, the temporal interval between the two may be of a long duration. H.H. Price distinguishes between "shortrange" natural signs where the interval is relatively short and there is direct association in prior experience and "long-range" signs where what is signified lies in the relatively remote future. 3 Lightning as signifying thunder is an example of the former, since thunder follows the flash of lightning in a matter of seconds. Another example would be Hume's sight of the flame as a sign of intense heat for the learning child. In contrast, clouds are a longrange sign of the rain to follow, as would be a crack in a building's foundation as a sign of its eventual collapse and the rate of a star's combustion a sign of its later explosion. Here the interval may be hours, years, or even millions of years, and no direct association in prior experience between the sign and what it stands for is required for interpreting the sign. For these longrange signs of what occurs in the future the relation between sign and signified is not necessarily causal. The metabolic rate of the potato usually rises about forty-eight hours before a fall in barometric pressure, and thus this rising rate can be interpreted as a sign of the falling pressure. But the former clearly cannot be said to be the cause of the latter. 4 Besides distinguishing natural signs with respect to temporal direction the tradition has followed Aristotle in also distinguishing necessary from probable signs. Thus smoke could be said to

Natural signs

45

be a necessary sign of fire, a striated rock a sign of past life, a red shift in a spectral line a necessary sign of a receding galaxy, etc. These signs are 'necessary' in the sense that their presence invariably assures the presence of what they signify. In contrast, spots are only a probable sign of measles and clouds a probable sign of rain. As examples of indices Peirce sometimes cites signs which only signify with a high degree of probability, as a low barometer with moist air is an index of rain, for 'we suppose that the forces of nature establish a probable connection between the low barometer with moist air and corning rain. ,5 These distinctions and virtually all of the examples used to illustrate them indicate that what are called 'natural signs' in the tradition that begins with Augustine are not primitive signs to be contrasted or compared with linguistic signs. Instead, they are examples of evidence whose interpretation requires the use of a linguistic empirical generalization. This generalization is usually a causal generalization and the evidence either an instance of either the cause of effect described within it. The interpreter infers from the evidence to what it stands for by means of the generalization. Thus, to interpret on a given occasion clouds as evidence of rain requires first formulating or being aware of the causal generalization 'Clouds cause rain', or better, 'Clouds are a sufficient causal condition of rain.' From the clouds that are observed we infer by way of the generalization to the predicted rain. More generally, if evidence x is an instance of an event, state of affairs, or object6 of type X and its effect is of type Y, then given that X is the sufficient cause of Y a person can infer from the presence of x to a later instance of Y. The form of inference used in making the prediction, is thus X is the sufficient cause of Y Evidence x occurs (is present)

There will be (is) an occurence of Y Upon observing the evidence x its interpreter infers to the conclusion by means of the first premiss. Where the relation is not causal, as for the metabolic rate of the potato and falling atmospheric pressure, the first premiss becomes a generalization of the form 'Whenever X occurs Y will also.' Where we can anticipate an occurrence of Y with only a certain degree of

46

Natural signs

probability, the first premiss is of the form 'X is the probable sufficient cause of Y' or 'The probability of Y given X is r', where r is some rational number less than 1 but greater than zero (symbolically p(Y/X) = O A is just as plainly analytic. Given the meanings of 'and' and 'if ... then' it could not possibly be denied. To deny that the analytic-synthetic distinction applies to these sentences is to confuse the status of sentences within theoretical contexts with those outside such a context. 34 Still another feature unique to discourse contexts is to be found in fictional narrations. Here information transference from predicates to subjects occurs of the kind just discussed above for non-fictional narrative discourse. Thus, a novelist may write 'The old man walked with a limp. He wore a blue suit with a red necktie. He smiled frequently'. If then the novelist continues 'The old man knocked at the door', this last occurrence of 'the old man' is synonymous with the definite description 'the old man who walked with a limp, wore a blue suit and red necktie, and smiled frequently' in which occurs the predicates of the sentences that preceded. For the novelist to 'develop a character' is for the meaning of a name or definite description to change from earlier occurrences in the discourse to later. A term within a fictional narration can be termed a structural subject term when it occupies the position of a sentence's subject. Such a subject exploits the mechanisms of information transference which operate wherever there are repeated occurrences of a subject followed by a variety of predicates. It is not, however, a functional subject, one which is used to refer to some object to be identified. There is, of course, no spatial or temporal occasion at which such identifi-

Language

135

cation could or could not have taken place, and without this possibility there is no reference. We must reject, then, any theory which attempts to assign referents to fictional subjects. This obviously applies to Meinong's theory of 'subsistents' as referents of a term such as 'Pegasus' in 'Pegasus is white', in contrast to 'existents' as referents of ordinary proper names such as 'Aristotle' and 'John F. Kennedy'. It also applies to Russell's theory outlined in Section 1.1 in which 'Pegasus' is first replaced by a definite description, e.g. 'the winged horse captured by Bellerophon' to form the sentence 'The winged horse captured by Bellerophon is white'. As outlined in 1.1, a new subject 'material thing', the 'true' logical subject, is then introduced, and the sentence is paraphrased by an existential sentence in which 'is a winged horse captured by Bellerophon' and 'is white' are predicates. Still more recent theories postulate a possible world in which exists an individual named 'Pegasus' and satisfying the unique attributes expressed by the associated definite description. The task set for all such theories is to specify the referent of the fictional sentence 'Pegasus is white'. But this is to assume that sentences within discourse contexts function in all respects as do sentences isolated from them, and we have seen this not to be the case. Just because subjects of isolated descriptive sentences are used to refer, it does not at all follow that subjects of sentences within fictional discourse must also have this function. Some may object to this attempt to distinguish features of discourse-relative sentences from isolated sentences by arguing that all sentences occur in the context of some discourse or other. Even a sentence such as 'It is raining' normally occurs in the context of a conversation, and is an answer to at least an implied question ('What is the weather like?). An 'isolated sentence' is therefore a fiction, and cannot sustain a contrast to a sentence within a discourse context. The contention being made is, of course, correct, but does not constitute an objection to the contrasts we have been drawing. Developed forms of discourse - whether inferences, scientific theories, novels, directives, or laws, etc. - are employed for specific social purposes, and their special features derive from these purposes. The sentences that occur in everyday conversation, in contrast, are used for a variety of purposes, and their

136

Language

principal features enable this flexibility. By an 'isolated sentence' has been meant a sentence with this potential flexibility of use. The error committed by the theories so far discussed has been either that of projecting features of a specialized use of language on to sentences with a potential for a variety of uses or of ascribing to a specialized use (as in fiction) what holds of sentences for which flexibility is essential. There are thus dangers in restricting analysis - as has been the practice of much recent philosophy - to the relatively complex language forms to be found in developed areas of discourse, dangers which often outweigh the advantages of the explicitness and amplification these complex forms provide. We have seen in this chapter how generalizing from features of discourse-relative sentences can lead to distorted views of reference, meaning, and truth. The remedy is to adopt the strategy followed in the last three chapters and focus attention on features of language shared in common between relatively simple sentences and more primitive types of signs.

P

OSTCRIPT

Our discussion of language forms has been cursory and partial. There has only been brief mention of general sentences and only an indefinite sketch of a few forms of discourse. I have ignored the special features of first person avowals, whether expressive of feelings and emotions such as 'I feel tired' and 'I am sad' or reports of sensations such as 'I feel a pain'. As forms of language parasitic on ordinary descriptive language their use poses special problems. There has been no mention of the iconic symbolic forms characteristic of artistic expression, nor of the symbolic formulas employed in mathematics. There is a good practical reason, I think, for excluding these more complex signs from the scope of semiotic as conceived of in this work. It is simply that the study of signs at these higher levels of complexity becomes indistinguishable from established branches of philosophy such as the philosophy of language, deductive and inductive logic, philosophy of science, philosophy of mathematics, philosophy of law, and aesthetics. There is nothing unique that semiotic has to offer regarding the structures and use and interpretation of these forms of discourse studied by other branches. Instead, its focus is properly on features of language revealed by contrasts and comparisons between simple forms of sentences and more primitive signs. There is, however, what might be called a 'semiotic perspective' that can be brought to bear on the study of more complex lansigns. It consists of extending the focus on comparative features of the kind developed in this work for singular sentences and primitive signs to those holding between singular sentences and sentences embedded in the various frameworks of discourse. The previous section may provide some indication as to how this extension is to be made. How does the reference of terms such as 'quark' and 'neutrino' standing for theoretical entities postulated by a physical theory differ from that of singular terms such as 137

138

Postscript

'this book' as occurring in 'This book is red'? How does information transference to such terms as a result of confirmation of a theory in which they occur compare with that discussed in the previous chapter for singular terms? How do the iconic representations of paintings or other forms of artistic expression compare to the gestures or sounds of primitive iconic comsigns? But though such questions may be generated by extending the study outlined in this work, answers to them we should expect to find in the branches of philosophy linked to the special forms of discourse.

N

OTES

1 Introduction

1 See Russell's Introduction to Mathematical Philosophy, London, Allen & Unwin, 1919, Ch. XVII. Russell's interests in logical analysis at this early stage were ontological. Specifically, by replacing a sentence's grammatical subject by its logical subject he sought to show the kinds of objects denoted by the sentence. For later developments of this same project see Gilbert Ryle, 'Systematically Misleading Expressions' in R. Rorty (ed.), The Linguistic Turn, Chicago, Chicago University Press, 1967 and W.V.O. Quine, 'On What There Is' in A Logical Point of View, New York, Harper & Row, 1953. Later Russell, in common with the early Wittgenstein and Carnap, used logical representation techniques as a devise for pursuing the epistemic goal of analyzing complex propositions into simple atomic propositions reporting the immediate data of experience. For this project see his 'The Philosophy of Logical Atomism' in Logic and Knowledge, New York, Capricorn, 1956. 2 See 'On Referring,' Mind, 59 (1950): 30-44. Straws on introduces the term 'reference' to stand for the triadic relation between a subject term and an object as used by a person on a given occasion in contrast to the logician's term 'denotation' as standing for a dyadic relation between subject and object in abstraction from such use. 3 More exactly. 3x(Kx & Vy(Ky:::ly = x) & Bx) logically entails 3xKx, 3xVy(Ky:::ly = x), and 3xBx. Since 3xKx is false. it follows that 3x(Kx & Vy(Ky:::ly = x) & Bx) must be false. 4 See Donald Davidson, 'The Logical Form of Action Sentences' in N. Rescher (ed.), The Logic of Decision and Action, Pittsburgh, University of Pittsburgh Press, 1967. The example used here is from Gilbert Harman's exposition of Davidson's theory in 'Logical Form' in Davidson and Harman (eds), The Logic of Grammar, Eicino, Dickenson, 1975. Davidson'S, project is again briefly discussed below in Section 5.3. 5 Exactly such a distinction is argued for by Strawson in 'Grammar and Philosophy' in his Logico-Linguistic Papers, London, Methuen, 1971. Here the contrast is made between the 'essential grammar' of 'nonempirical linguistics,' i.e. philosophy, and the transformational grammar of a given language as constructed by the procedures of empirical linguistics.

n9

140 Notes to pages 7-12 6 In his Philosophy of Language, New York, Harper & Row, 1966, Ch. 3, and The Underlying Reality of Language and Its Philosophic Import, New York, Harper & Row, 1971, Chs 1-3, Jerrold Katz argues that there is no distinction, that the features of language sought by both the philosophical traditions of logical analysis and ordinary language descriptions are deep structures of language discovered by the empirical procedures of transformational grammar. For the equation of logical form with the linguists' deep structure see also Gilbert Harman, 'Deep Structure as Logical Form' and George Lakoff, 'Linguistics and Natural Logic,' both in Davidson and Harman (eds) , Semantics of Natural Language, Dordrecht, D. Reidel, 1972. 7 For this criticism see Michael Dummett, 'Can Analytical Philosophy be Systematic and Ought it to Be?' in his Truth and Other Enigmas, London, Duckworth, 1978. The ordinary language philosophers, claims Dummett, espouse a 'doctrine of total particularism' in their descriptions of the various uses of sentences. 8 Cf. Max Black, Models and Metaphors, Ithaca, Cornell University Press, 1962, p. 241: 'By an archetype I mean a systematic repertoire of ideas by means of which a given thinker describes, by analogical extension, some domain to which those ideas do not immediately and literally apply.' 9 For a discussion of the link between sign complexity and evolution see T.P. Waldron, Principles of Language and Mind, London, Routledge & Kegan Paul, 1985. 10 See Zenon Pylyshyn, Computation and Cognition, Cambridge, Mass., MIT Press, 1984 and Steven Stich, From Folk Psychology to Cognitive Science, Cambridge, Mass., MIT Press, 1983. I present here a version of what Stich calls the 'sentential theory' of psychological attitudes. According to this theory, for one to have an intentional attitude (a belief or desire) 'is to have a sentence token suitably instantiated or encoded in one's brain' (p. 29). 11 Paul Churchland, Scientific Realism and the Plasticity of Mind, Cambridge, Cambridge University Press, 1979, pp. 128ff. Extending investigations in the way Churchland proposes is to adopt the program of what Jerry Fodor calls 'naturalistic psychology' and accept the challenge (which Fodor claims cannot in practice be met in the foreseeable future) of correlating internal signs with environmental stimuli and behavioral effects. See Fodor's 'Methodological Solipsism Considered as a Research Strategy in Cognitive Psychology,' The Behavioral and Brain Sciences, 3 (1980): 63-73. 2 History of semiotic 1 For good general summaries of this early history see Umberto Eco, Semiotics and the Philosophy of Language, Bloomington, Indiana University Press, 1984, Sections 1.6 and 1.7 and Thomas Sebeok,

Notes to pages 12-18 141

2

3 4 5 6 7 8 9

10

11 12 13

'Semiotics: A Survey of the State of the Art' in his Contributions to the Doctrine of Signs, Bloomington, Indiana University Press, 1976. More specialized studies can be found in A. Eschbach and J. Trabant (eds), History of Semiotic, Amsterdam, Verlag, 1983. Sebeok claims that the modern term 'semiotic,' introduced by Locke, corresponds to the Greek semeiotike, which is derived from semeiotikos. It has been pointed out to me by Robert Hahn, however, that the term 'semeiotike' cannot be found in extant classical writings. Luigi Romeo in 'The Derivation of "Semiotics" through the History of the Discipline,' Semeoisis 6 (1977): 37-49, accounts for this by speculating that Locke's term is instead derived from the Medieval semeiotice, a term then mistakenly assigned by Renaissance lexicographers as corresponding to a Greek noun. The principal ones are Against the Logicians (Adversos Mathematicos) and Outlines of Pyrrhonism in Sextus Empiricus, by R.G. Bury, Loeb Classical Library, Cambridge, Mass., Harvard University Press, 4 vols., 1949-57. Prognostic, 1, 1-8 in Hippocrates, trans. by W.H.S. Jones, Loeb Classical Library, Cambridge, Mass., Harvard University Press, 19231, Vol. II. Institutio aratoria, Bk. V, Ch. IX. See The Institute aratoria of Marcus Fabius Quintiliamus, trans. by C. Little, Nashville, George Peabody College for Teachers, 1951, Vol. I. Rhetorica, 1357bI8. Translation by W. Rhys Roberts in R. McKeon (ed.), The Basic Works of Aristotle, New York, Random House, 1941. Against the Logicians, II, 245. ibid., 253. See also VII, 245, 358, 364-7. ibid., 264. In this and the quotation which follows Sextus is not simply presenting but defending the Epicurean view. ibid., 269-71. Note that the first example of a sign interpreted by a lower animal, the footprint, stands for an object, while the second, the prod or the crack of the whip, stands for an action. For the latter Sextus reports that there is a 'must' or 'ought' conclusion. ibid., II, 188 and VIII, 215ff. The argument is presented by Charlotte Stough in Greek Skepticism, Berkeley, University of California Press, 1969, pp.97-137. Stough points out that Sextus's argument is anticipated by the Greek skeptic philosopher Aenesidemus. See also C.K. Ogden and LA Richards, The Meaning of Meaning, New York, Harcourt & Brace, 1923, Appendix C. Fragment VIII, 1-5 and Fragment XIX as translated by Leonardo Taran in Parmenides, Princeton, Princeton University Press, 1965. Against the Logicians, VIII, 200. De Interpretatione, 16al-7. Translation by H.G. Apostle in Aristotle's Categories and Propositions, Grinnell, Peripatetic Press, 1980. The bracketed Greek words are mine; the bracketed expressions in English are Apostle's.

142 Notes to pages 18-22 14 See Semiotics and the Philosophy of Language, pp. 27-9. Eco concludes that Aristotle's use of semeia to stand for words is only 'incidental. ' 15 But the extent of Augustine's originality has been a subject of controversy. R.A. Markus emphasizes this originality, noting that the early Christian writers applied the term 'signum' to Scriptural events and biblical miracles. In the period before Augustine (with the possible exceptions of Plotinus), concludes Markus, 'for no writer is the theory of signs primarily a theory of language, nor is reflection on language carried on in terms of "signs".' See 'St. Augustine on Signs' in Markus (ed.), Augustine: A Collection of Critical Essays, New York, Doubleday, 1972. B.D. Jackson claims to find more similarities between Augustine's theory and the Stoic theory of evidence, and claims the Stoics as anticipating Augustine's view. See 'The Theory of Signs in St. Augustine's De Doctrina Christiana' in this same collection. 16 De Dialecta, V, 9-10. Translated by B.D. Jackson in Augustine: De Dialecta, Dordrecht, Reidel, 1975. 17 De Doctrina Christiana, II, 2, 12-15. Translation by Jackson in 'The Theory of Signs in St. Augustine's De Doctrina Christiana.' 18 ibid., 3, 1-6. Translation again by Jackson. 19 ibid., 24, 37. The contrast between the intentional and conventional aspects in Augustine's characterization is noted by Jackson. Markus, in 'St. Augustine on Signs,' regards the conventional aspects as essential for signa data, for he excludes from this category facial expressions and 'sounds whereby animals communicate to one another their desires, their perceptions of food or dangers and so forth.' 20 De Trinitate, XV, 10, 18. Translated by Markus in 'St. Augustine on Signs.' Thomas Mitchell has pointed out to me that a similar view can be found in the writings of Aquinas. In his De Differentia Verbi Divini et Humani, Sec. 287, we find: 'It is clear, nevertheless, that that which is signified as existing internally in the soul [illud quod significatur interius in anima existens] is prior to the word itself that is proffered by the exterior voice [verbum exteriori voce prolatum], which is its existing cause.' Later in Section 288 Aquinas says the external word is the 'sign' of the internal word's existence. See Aquinas's Opuscula Philosophica, R.M. Spiazzi (ed.), Taurini, Mariett, 1954, p. 101. lowe the translation of this passage to Charles Speck and G.K. Plochmann. 21 William of Ockham, Super Quatuor Libros Sententarium Subtilissimae Earumdenque Decisiones, II, quo 25. Translation by Stephen Tornay in Ockham: Studies and Sections, La Salle: Open Court, 1938, p. 19. 22 Summa Totius Logicae, Pt. II, Ch. 1. Translation in Tornay, pp. 92, 93. 23 ibid. For an exposition of Ockham's views see Gordon Leff, William

Notes to pages 22-7 143 of Ockham, Manchester, Manchester University Press, 1975, pp. 42, 43. 24 How thorough-going is the shift in focus towards language is shown by the curious inversion of the classical evidential sign noted by Ian Hacking in The Emergence of Probability, London, Cambridge University Press, 1975, pp. 42, 43. The primary evidence became in the Middle Ages the testimony of some source, with the authority of this source determining the reliability of the evidence. Hacking points out how natural events then came to be regarded as God's testimony to man, and the interpretation of these events as natural signs was regarded as the task of giving a correct reading of nature as God's 'Book.' It has been pointed out to me by G.K. Plochmann that St Bonaventure regards creatures as 'shadows' and 'traces' from which God's nature can be inferred. See his Commentary on the Four Books of Sentences of Peter Lombard, Pt. I, Qu. II, trans. by R. McKeon in Selections from Medieval Philosophers, Vol. II, New York, Scribner's, 1930. 25 Elements of Philosophy, Pt. I, Ch. 2, 2. 26 Essay Concerning the Human Understanding, Bk. IV, Ch. XXI, Sec.

4.

27 Antoine Arnauld, The Art of Thinking [L'Art de Penser] , First Discourse, Pt. I, Ch. 4. Translation by J. Dickhoff and P. James, Indianapolis, Bobbs Merrill, 1964. 28 Thomas Reid, An Inquiry into the Human Mind, Pt. II, Ch. 5, Sec. III. 29 ibid., Pt. II, Ch. 4, Sec. II. The passages that follow are also from this section. 30 Three Dialogues between Hylas and Philonous, 1. 31 C. Hartshorne and P. Weiss (eds), The Collected Papers of Charles Sanders Peirce, 6 vols., Cambridge, Mass., 1934-36, 2.227. 32 ibid., 2.228. 33 ibid., 2.231. Shortly after this passage Peirce offers as an example the sentence 'That vessel there carries no freight at all, but only passengers' as uttered by someone to a hearer who is looking at the vessel being referred to. 34 The other category is that of icons. Peirce also has a more complex classification of signs in which they are sub-divided into 64 classes by a formal scheme that follows his metaphysical categories of firstness, secondness, and thirdness. See 2.234-42. 35 ibid., 2.248. He also gives as an example the rolling gait of a man as a 'probable indication that he is a sailor' (2.285), where the index is a classical evidential sign but not a causal effect. 36 The bullet hole, Peirce says, would be an index of the bullet 'whether anybody has the sense to attribute it to a shot or not' (2.304). 37 ibid., 2.288. 38 ibid., 2.249. 39 The crack of the whip as a sign of an action would seem to be the

144 Notes to pages 27-33

40 41

42

43 44 45

46 47

48

49 50

object of what Peirce refers to as 'dynamic' or 'energetic' interpretation. See Collected Papers, 4.536 and 5.475. ibid., 2.292. ibid., 5.476. As George Gentry notes, this view of the sign's interpretant as a habit is a later view of Peirce's; in earlier versions the interpretant of a sign is always another sign. See 'Habit and the Logical Interpretant' in P. Weiner and F. Young (eds), Studies in the Philosophy of Charles Sanders Peirce, Cambridge, Mass., Harvard University Press, 1952. Course in General Linguistics, trans. by W. Baskin, New York, Philosophical Library, rev. ed., 1974, Intro., Ch. III. Pierre Guiraud, Semiology, trans. by G. Gross, London, Routledge & Kegan Paul, 1975, p. 22. For a review of other contributions to semiology in the Saussurean tradition, including those of Buyssens, Hjelmslev, and Uldall, see Sebeok, Contributions to the Doctrine of Signs, pp. 17-21. Roland Barthes, Elements of Semiology, trans. by A. Lavers and C. Smith, New York, Hill and Wang, 1967, pp. 11-15. British and American writers sometimes use the term 'semiotics' of Barthes's extended study of cultural forms. For a good review of the contributions of Barthes and other Continental writers see Kaja Silverman, The Subject of Semiotics, New York, Oxford University Press, 1983. C.K. Ogden and I.A. Richards, The Meaning of Meaning, New York, Harcourt & Brace, 1923, pp. 139, 140. This extension is made by J.B. Watson for whom 'words function in the matter of calling out responses exactly as did the objects for which the words serve as substitutes.' See his Behaviorism, New York, Norton, 1925, p. 233. A similar theory is developed by Bertrand Russell in An Inquiry into Meaning and Truth, London, Allen & Unwin, 1940. 'When you hear an object-word which you understand,' Russell says, 'your behavior is, up to a point, that which the object itself would have caused. This may occur. . . by the ordinary rules of conditioned reflexes, since the word has become associated with the object' (p. 68). See C.E. Osgood, 'On Understanding and Creating Sentences,' American Psychology, 18, 1963: 735-51 and Osgood, G. Succi, and P. Tannenbaum, 'The Logic of Semantic Differentiation' in S. Saporta (ed.), Psycholinguistics, New York, Holt Rinehart & Winston 1965. See especially Roger Brown, Words and Things, Glencoe, Free Press, 1958, pp. 93ff. The precise definition given in Signs, Language and Behavior, New York, Braziller, 1946, p. 10, is as follows: 'If anything A, is a preparatory-stimulus which in the absence of stimulus-objects initiating response-sequences of a certain behavior-family causes a disposition in some organism to respond under certain conditions by response-sequences of this behavior family, then A is a sign.'

Notes to pages 33-8 145 51 ibid., pp. 13, 18. Morris's early theory can be found in Foundations of the Theory of Signs, Vol. I of O. Neurath (ed.), International Encyclopedia of Unified Science, Chicago, University of Chicago Press, 1938, and seems to be a version of the direct response theory. For criticisms of this version see C.l. Ducasse, 'Some Comments on C.W.Morris's 'Foundations of the Theory of Signs',' Journal of Philosophy and Phenomenological Research, 3 (1942): 43-52. 52 See Rudolph Carnap, 'Meaning and Synonymy in Natural Languages,' Appendix D of Meaning and Necessity, Chicago, University of Chicago Press, 1947. 53 W.V.O. Quine, Word and Object, Cambridge, Mass., MIT Press, 1960, pp. 26-40. 54 Even for Quine's linguistic version there are difficulties in identifying the types of stimuli required for the definition of meaning. For a discussion of the difficulties in making the interpersonal comparisons required for translation see D. Follesdal, 'Meaning and Experience' in S. Guttenplana (ed.), Mind and Language, Oxford, Clarendon Press, 1974. 55 Roderick Chisholm, 'Intentionality and the Theory of Signs,' Philosophical Studies, 3, 1952: 56-63. Chisholm's criticisms are directed towards Carnap's version of the dispositional theory, and are here reformulated to apply to Quine's. Translations of the alien language into the linguist's language Quine regards as fallible empirical hypotheses, and hence he can admit (and has admitted) the difficulties Chisholm raises. Nevertheless, these difficulties serve to throw into question the basic definitions of meaning that are central to his theory. For criticisms directed specifically against Quine's theory see Noam Chomsky, 'Quine's Empirical Assumptions' in D. Davidson and l. Hintikka (eds), Words and Objections, Boston, Reidel, 1969. 56 This application is made by Friedrich Waismann in Principles of Linguistic Philosophy, London, Macmillan, 1965, p. 13: 'By means of words grouped and combined in different ways we can move the limbs of other people as with invisible strings or wires, and in this way direct their actions. In the last analysis there is nothing there except the signs and their effect; so the effect must be its meaning.' 57 Verbal Behavior, New York, Appleton-Century-Crofts, 1957, p. 84. 58 Thomas Sebeok, Contributions to the Doctrine of Signs, pp. 65, 66 and Approaches to Semiotic, The Hague, Mouton, 1964, pp. 277-87. Note the plural's' in Sebeok's 'semiotics'. It seems appropriate because Sebeok seems to apply both 'zoosemiotics' and 'anthropo semiotics , to the collection of those empirical sciences having a common subject matter. 59 Though Sebeok notes in Contributions to the Doctrine of Signs, p. 49 that the term 'Semiotik' until quite recently was also applied in Germany to the diagnosis of medical symptoms. 60 G. Harman, 'Semiotics and the Cinema' in G. Mast and M. Cohen

146 Notes to pages 38-46

61 62 63 64

65 66

(eds), Film Theory and Criticism, London, Oxford University Press, 1979. I have replaced Harman's italics with quotes to indicate mentioned expressions. In Semiotics and the Philosophy of Language, p. 19, Eco replies to Harman by appealing to the 2,000 years of tradition of semiotics. This is, of course, in itself unconvincing, since it is entirely possible that this tradition was based on a misconception. See Buhler's 'The Axiomatization of the Language Sciences' in R. Innis, Karl Buhler: Semiotic Foundations of Language Theory, New York, Plenum Press, 1982. Chomsky, 'The General Properties of Language' in F.L. Daley (ed.), Brain Mechanisms Underlying Speech and Language, New York, Grune and Stratton, 1967. See his 'Review of Skinner's Verbal Behavior,' Language, 35, 1959: 25-58. J.L. Austin, 'Truth' in J.O. Urmson and G.J. Warnock (eds), Philosophical Papers, Oxford, Clarendon Press, 1961. Signs, Language and Behavior, p. 33. ibid., p. 17. 'Anything which would permit the completion of the response-sequences to which the interpreter is disposed because of a sign,' Morris says, 'will be called the denotatum of the sign. A sign will be said to denote a denotatum.'

3 Natural signs

1 The boulder example is from William Alston's Philosophy of Language, Eaglewood, Prentice-Hall, 1964, p. 50. 2 But if we identify emotional states with brain states these items of behavior can be claimed to be symptoms and causal effects of these states, not criteria. This position is argued for by Hilary Putnam in 'Brains and Behavior' in R.l. Butler (ed.), Analytical Philosophy, Vol. 2, Oxford, Blackwell, 1965. Establishing the identities, however, would seem to require first independently identifying the emotional states, and for this behavior would certainly seem criterial and not a contingent effect. 3 H.H. Price Thinking and Experience, Cambridge, Mass., Harvard University Press, 1953, p. 106. 4 The example is given by Kenneth Friedman in 'Analysis of Causality in Terms of Determinism,' Mind, 89 (1980): 544-64. The relation is non-causal, Friedman argues, because the metabolic rate is not also a necessary condition for the rising pressure: suppressing the former does not eliminate the latter. 5 Collected Papers, 2.286. 6 I follow our ordinary ways of speaking of objects and states of affairs as well as events as being the relata of the cause-effect relation. Thus, the presence of boulders, fossils, and spots is said to be the effect of past glaciers, life forms, and present measles. 7 'Semiologie et Medecine' in R. Bastide (ed.), Les Sciences de la

Notes to pages 46-54

147

Folie, Paris, Mouton, 1972. 8 See Sebeok, Contributions to the Doctrine of Signs, pp. 126, 127. 9 Eco, Semiotic and the Philosophy of Language, pp. 36-43. A similar error is made by Andrew Jones in Communication and Meaning, Dordrecht, Reidel, 1983, where he compares a person's regular walking habits as a sign of the time of day to a conventional signal such as the hanging of a lantern in Boston as a sign the British were coming. Jones says he has 'no hesitation in assigning truth conditions to actions' such as the regular walking (p.14). 10 It would not be accurate to claim that these signs constituted the extension of 'semeion' for the Epicureans, who it will be recalled from 2.1 cited sign interpretation by lower animals in their attempt to refute the Stoic inferential theory. By retaining the examples of the classical evidential signs and including medical symptoms they clearly did not intend our restriction. Nevertheless, the Epicureans provide an anticipation of the view defended here. 11 Signs of the kind characterized here seem to be what Ogden and Richards in The Meaning of Meaning refer to as 'initial signs:' 'Directly apprehended retinal modifications such as colors are therefore initial signs of "objects" and "events'" (p. 82). They cite with approval Helmholtz's view that 'sensations which lie at the basis of all perceptions are subjective signs of external objects' (p. 79). 12 In Sense and Sensibilia, G.L. Warnock (ed.), Oxford, Clarendon Press, 1962 and Concept of Mind, New York, Barnes & Noble, 1949, Ch. VI. See also Richard Rorty, Philosophy and the Mirror of Nature, Princeton, Princeton University Press, 1979 for an exposition and criticism of this theory that dominated post-Cartesian philosophy. 13 For the opposing view that mirror images are representations see Virgil Aldrich, 'Mirrors, Pictures, Words, Perceptions,' Philosophy, 55 (1980): 39-56. Aldrich does concede, however, 'that any x in the role of representing y tends to function simply as a point of view on y, or to disappear in favor of the y it intends.' This disappearance of the image, he says, 'makes the question of their resemblance slippery.' 14 See Kendall Walton, 'Transparent Pictures: The Nature of Photographic Realism' and Fred Dretske, 'Seeing Through Pictures,' both in Nous, 18 (1984): 67-74. The issue defined by Walton is whether photos are 'transparent' (we see objects by means of them) or 'opaque' (we see photos as representations of objects). Dretske argues that photos are natural signs, and includes TV images also within this category: 'In watching a football game on television ... we get information about the game by getting information about what is on the television screen' (p. 74). The screen image, he concludes, is 'informationally' but not 'perceptually' transparent. A version of Dretske's view is presented (and eventually rejected) by Max Black in 'How Do Pictures Represent?' in E.H. Gombrich, J. Hochberg,

148 Notes to pages 54-62

15

16 17

18

19

20

and M. Black, Art, Perception, and Reality, Baltimore, Johns Hopkins University Press, 1972. On this view, says Black, a photograph P is regarded as a 'trace' of the object it depicts, and 'the interpretation of P is a matter of inference to an earlier time in a causal sequence.' E.M. Gombrich calls a photograph a 'natural trace, a series of tracks left ... on the emulsion of the film by the variously distributed light waves which produced chemical changes made visible and permanent.' See his 'The Evidence of Images' in S. Singleton (ed.), Interpretation, Theory and Practice, Baltimore, Johns Hopkins Press, 1969. It is, however, not normally interpreted as a trace (though it could be) by one who views it, unlike Peirce's indices. This Wittgensteinian notion of 'showing' as contrasted with 'saying' is applied to photos by Black in 'How Do Pictures Represent?'. Certainly the showing relation between image and object cannot be explained in terms of resemblance between the two. For criticisms of such an explanation for pictures that can be extended to material images see Nelson Goodman, The Languages of Art, Indianapolis, Hacket, 1956, pp. 3-12. For singular sentences complete interpretation thus coincides with what is called 'verification' and 'falsification.' For generalizations, whether uniform or statistical, conducting at least one experimental test constitutes complete interpretation in the sense intended here, even though acceptance or rejection of the generalization may not result. Morris, for example, writes of the food found by the dog as what is 'denoted' by the buzzer. The denotatum of the sign, he says, is 'anything which would permit the completion of the responsesequences to which the interpreter is disposed because of a sign,' in contrast to its significance (or 'significatum'), which is 'those conditions which are such that whatever fulfills them is a denotatum.' See Signs, Language and Behavior, pp. 17, 18. Peirce's triadic relation of sign, interpretant, and object seems to assume a referential relation between every sign to some object for its interpreter. Over and over again we find this error perpetuated in the semiotic tradition following Peirce in the form of diagrams with arrows pointing from sign to an object representing what is taken as a referential relation. The contrast to Morris's views should again be noted. In Signs, Language and Behavior signs are said to be 'true' 'in so far as they correctly determine the expectations of their users' (p. 33). But in the absence of a conventional rule there is no sense to a 'correct' determination. Morris describes a sign as 'reliable' 'to the degree that it denotes in the various instances of its appearance. If an animal always found food at a certain place when a buzzer sounded the buzzer would be completely reliable.' (p. 106). Here again terminology is misapplied. For us a sign can be reliable in the sense that

Notes to pages 62-71

21 22

23

24 25 26

27

149

it is supported by a generalization with high probability and speakers can be more or less reliable depending on our confirmation of what they have to say. But at the level of natsigns 'reliable' has no application. P.F. Strawson, The Bounds of Sense, London, Methuen, 1966, p.273. This recognition as well as sign discrimination seems to be what Jonathan Bennett includes under the heading of what he calls 'registration' in Linguistic Behavior, Cambridge, Cambridge University Press, 1976, p. 46. Registration, he says, is a 'perceptual or epistemic matter' which depends on whether goal-seeking behavior is modified. Our observation of this modification may be our only means of identifying registration in lower animals, but it seems possible for there to be registration without behavior change. Peirce also distinguishes what he terms the 'emotional interpretant' of a sign. Emotional interpretation can be regarded as a third mode in which the sign's significance evokes an emotional attitude on the part of its interpreter, whether that of fear or hope, aversion or wanting. Thus, the sight of the candle signifies intense heat to which a person's attitude is that of fear or aversion. In the discussion that follows we assimilate this mode to dynamic interpretation. For higher-level signs it is easier to distinguish it, but since signs as objects of this interpretation seem to display no unique logical functions we ignore emotional interpretation in later chapters. Our cognitive-dynamic contrast corresponds also to H.H. Price's distinction between the 'cognitive aspect' of significance and the 'practical' aspect. For the former, Price says, X is the sign of Y; for the latter X is the sign for an action taken in order to secure or avoid Y. See Thinking and Experience, pp. 91ff. For a more complete account of both forms of inference see my Practical Inferences, Routledge & Kegan Paul, 1985, pp. 15-22,51-5. See Saul Kripke, 'A Puzzle About Belief' in A. Margalit (ed.), Meaning and Use, Dordrecht, D. Reidel, 1976. The inapplicability of belief ascriptions to non-language users is noted by Davidson in 'Thought and Talk' in S. Guttenplan (ed.), Mind and Language, Oxford, Clarendon Press, 1975. But he draws the more general conclusion that 'thoughts' cannot be ascribed to those without linguistic capacity. This more general denial of any psychological ascriptions other than feelings or sensations seems mistaken. See Ralph Barton Perry's distinction between signs as 'representative stimuli' whose interpretation is based on prior learning and stimuli triggering instinctive behavior in General Theory of Value, New York, Longman, 1926, pp. 177-9.

150 Notes to pages 73-7 4 Communication

1 The term 'comsign' is borrowed from Charles Morris who assigns it a somewhat different meaning. What Morris, following G.H. Mead, calls a 'gesture' is a non-conventional sign used to communicate, e.g. the snarl of a dog. Only when a gesture 'has the same effect on the individual making it that it has on the individual to whom it is addressed' does it become for Mead a 'significant symbol.' Significant symbols are said to 'have the same meaning for all individual members of [a] given society or social group.' See Mind, Self and Society, C.W. Morris (ed.), Chicago, University of Chicago Press, 1934, pp. 46, 47. Mead's significant symbol is what Morris terms a comsign in 'Foundations of the Theory of Signs,' p. 36. It proves convenient, however, to take 'comsign' in a generic sense that contrasts its extension with that of natsigns and then distinguish within the class of comsigns conventional comsigns, signs which coincide with Mead's significant symbols, and non-conventional comsigns which include Mead's gestures. The bristling-snarling example that follows is used by Mead to illustrate the contrast between a movement by an animal and a vocal gesture. 2 See 'Meaning,' Philosophical Review, 66 (1957): 377-88. 'Comsign' is preferable to 'utterance' because of its more natural application to gestures and other means of non-verbal communication. For the same reasons I replace Grice's 'utterer' by 'communicator' and 'audience' by 'interpreter'. The generality achieved allows the extension of Grice's analysis to the examples of lower animal communication discussed below. 3 The policeman standing in the way of the motorist is, of course, not evidence from which to infer some other fact, but instead a circumstance from which to infer by way of a practical inference. Cf.3.4. 4 The original counterexamples are by Strawson in 'Intention and Convention in Speech Acts,' Philosophical Review, 73 (1964): 439-60 and John Searle in Speech Acts, Cambridge, Cambridge University Press, 1969, pp. 43-50. Perhaps the most sophisticated later elaboration is by Stephen Schiffer in Meaning, Oxford, Oxford University Press, 1974, p. 63. Simon Blackburn proposes that to avoid deception we simply add the additional condition that the communicator intends that all his intentions be recognized, thus ensuring what he calls 'openness.' See his Spreading the Word, Oxford, Clarendon Press, 1984, p. 116. A similar solution is proposed by Jonathan Bennett in Linguistic Behavior, Cambridge, Cambridge University Press, 1976, p. 127. 5 Strawson, 'Convention and Intention in Speech Acts,' p. 460. In Austin's terminology no specific perlocutionary effect may be intended, though the speaker must intend his utterance to have some illocutionary force.

Notes to pages 77-84

151

6 That belief as an effect is not necessary is noted by H.L.A. Hart in 'Signs and Words,' Philosophical Quarterly, 2 (1952): 59-62. 7 In Biological Bases of Human Social Behavior, New York, McGrawHill, 1974, p. 114. 8 E.W. Menzel, Jr, 'Communication About the Environment in a Group of Young Chimpanzees,' Folia Primatologica, 15 (1971): 22032. 9 Hinde, Biological Bases, p. 114. 10 ibid., p. 115. The low gutteral bark of a chimpanzee reinforced by turning and playing is offered as an example. Some philosophers have ruled out ascribing intentions to lower animals on logical grounds. Thus Stuart Hampshire in Thought and Action, London, Chatto & Windus, 1959: 'The difference here between a human being and an animal lies in the possibility of the human being expressing his intention and putting into words his intention to do so-and-so .... It is a . . . difference . . . correctly expressed as the senselessness of attributing intentions to an animal which has not the means to reflect upon, and to announce to itself or to others, its own future behavior' (p. 98). This seems a totally arbitrary stipulation which is at variance with ways in which we actually describe animal behavior. G .E.M. Anscombe seems closer to this usage when she ascribes intentionality to animals without language in Intention, Ithaca, Cornell University Press, 1957, p. 85. 11 W. John Smith, The Behavior of Communicating, Cambridge, Mass., Harvard University Press, 1977, p. 264. N. Tinbergen regards almost all signaling behavior as innate and reflex: 'Except perhaps in the highest mammals, all signaling behavior is immediate reaction to internal and external stimuli. In this respect there is a great difference between animals and Man.' See Social Behavior in Animals, London, Methuen, 1953, p. 74. A stronger conclusion, and one that does not seem warranted by available evidence, is reached by Bennett. He concludes in Linguistic Behavior that animal displays are 'due to the existence in [a] species of certain inflexible, wired-in behavior patterns whose instances do not exhibit intentionalness of any sort' (p.204). 12 Biological Bases, p. 90. 13 See Grice, 'Utterer's Meaning, Sentence-Meaning, and WordMeaning,' Foundations of Language, 4 (1968): 255-42 for a detailed attempt to state the relationships between occasion and timeless meaning. 14 This objection is formulated by Paul Ziff in 'On H.P. Grice's Account of Meaning,' Analysis, 28 (1967): 1-8. 15 For a similar view of the role of iconic comsigns and transfers from natsigns such as involuntary cries see Bennett's Linguistic Behavior, pp. 140, 141, 206. The historical transition of Chinese written characters from iconic to non-iconic conventionalized signs is another obvious example of what is being discussed here. The issue of

152 Notes to pages 84-90

16 17 18 19

20

21

22

23

whether and to what extent human languages initially evolved from a 'natural' iconic link to what is signified is, of course, one much debated by philosophers and linguists. Roger Brown in Words and Things, Glencoe, Free Press, 1958, pp. 11Off. cites evidence for a 'phonetic symbolism' for which there is iconic representation between speech sounds and magnitudes. In Convention, Cambridge, Mass., Harvard University Press, 1969, p.42. Exactly such a combination of the two sets of conditions is proposed by Bennett in Linguistic Behavior, pp. 177-9. See Paul Ziff, Semantic Analysis, Ithaca, Cornell University Press, 1960, pp. 42ff. Such correlations are termed 'conditional regularities' by Schiffer in Meaning, p. 155, where he summarizes Lewis's theory. That linguistic semantic conventions are not semantic regularities is argued for by Saul Kripke in stating Wittgenstein's 'skeptical paradox' that arises out of the possible use of two expressions with different conventional meanings, e.g. 'plus' and 'quus' or 'green' and 'grue', where there is no observable difference in linguistic behavior, whether in computing numbers or in applications to colors. It is the practice of mutual correction within the linguistic community that determines the meaning of expressions for Kripke (following Wittgenstein), not simply coincidence of behavior. 'Truth' in 1.0. Urmson and G.l. Warnock (eds), Philosophical Papers, Oxford, Clarendon Press, 1961. I avoid in these discussions the term 'statement' as standing for an object of which 'truth' and 'falsity' are predicated, preferring instead to speak of utterances as being judged true or false, or being assented to or denied. In 'Sceptical Confusions About Rule-Following,' Mind, 93 (1984): 423-9. Contributions to the Doctrine of Signs, p. 120. Georg Klaus distinguishes a sign as an object of consciousness (bewustein) from a signal producing a determinate behavioral effect. See Semiotik und Erkenntsnistheorie, Berlin, Deutscher Verlag der Wissenschaften, 1969, pp. 55ff. Suzanne Langer contrasts signals as 'substitute stimuli' evoking responses normally evoked by objects for which they are being substituted from symbols which 'are not proxy for their objects, but are vehicles for the conception of objects' in Philosophy in a New Key, Cambridge, Mass., Harvard University Press, 1951, p. 61. Our use of 'signal' shares many features with Peirce's 'rheme' (Collected Papers, 2.272), and it is tempting to use Peirce's term as a way of avoiding associations of 'signal'. But in a discipline where artificial terminology is overly abundant, it seems more advisable to choose the more familiar term and run the risk of its mechanistic associations. Also, Peirce's rheme is a sign of 'qualitative possibility' (cf. 2.250), and this would seem to rule out applying it, as we do here, to proper names. Biological Bases of Human Social Behavior, p. 83.

Notes to pages 91-9 153 24 For the contrast between the information of a 'signal sequence' and 'semantic information' see Yehoshua Bar-Hillel, Language and Information, Cambridge, Mass., MIT Press, 1964, pp. 275-97. 25 See Signs, Language and Behavior, p. 58. 26 Cf. Quine, Word and Object, pp. 94, 95 for an account of the transition from recognizing Mama-like features to recognizing Mama as an individual. 27 'Teleology and the Great Shift,' Journal of Philosophy, 81 (1984): 647-53. De Sousa formulates his distinction in terms of goal-directed behavior; I reformulate it here in terms of sign interpretation. 28 Cf. W.J. Smith, The Behavior of Communicating, pp. 59, 60. Smith cites studies of animal calls among species as diverse as chickadees, elephant seals, and reindeers in support of conclusions about signatures for animal communication. Long-distance communication with signatures can be accomplished also by chemical transmission, e.g. female sex pheromones which inhibit the approach of members of other species. Badges found in certain animal species, e.g. bird plumage indicating species, sex, and even mating availability, though they are of course not intentionally produced, can be regarded as primitive precursors of addresses. 29 All but the third and sixth features outlined are included in C.F. Hockett's list of 'design features' for communication systems. See his 'Logical Considerations in the Study of Animal Communication' in W.E. Lanyon and W.N. Tavolga (eds), Animal Sounds and Communication, Washington, D.C., American Institute of Biological Sciences, 1960 and C.F. Hockett and S.A. Altmann, 'A Note on Design Features' in T. Sebeok (ed.), Animal Communication, Bloomington, Indiana University Press, 1968. In the latter work altogether sixteen features are listed. In 'The Problem of Universals in Language' in J.H. Greenberg (ed.), Universals of Language, 2nd ed., Cambridge, Mass., MIT Press, 1966 Hockett lists duality of patterning, displacement, and openness, the capacity to generate infinitely many signs, as the three features whose joint presence is most important for distinguishing human languages from animal signal systems. 30 The possibility of complex signals for animal communication is raised in A.M. Wenner, 'The Study of Animal Communication' in Sebeok and Ramsay (eds) , Approaches to Animal Communication, The Hague, Mouton, 1957. 31 Hockett in 'Logical Considerations in the Study of Animal Communication' speculates that bird songs may exhibit the second level of patterning, while Wenner raises the possibility of a signaling system exhibiting the first level in 'The Study of Animal Communication.' Diebold speculates that of all the sixteen design features listed by Hockett and Altmann only duality of patterning is unique to human languages in 'Anthropological Perspectives' in Sebeok (ed.), Animal Communication.

154 Notes to pages 100-7 32 With syntactic rules for forming sentences comes also the possibility of sentence embedding and the formation of relative clauses and prepositional phrases. The formulation of these rules for given languages is the topic of transformational grammar as presented originally in Chomsky's Syntactic Structures, The Hague, Mouton, 1957. 33 In 'Logical Considerations in the Study of Animal Communication' Hockett cites this dance as the unique case of displacement occurring without language. But this is because he uses 'displacement' in the behavioral sense of orienting towards distant objects, not in the interpretive sense of enabling extension of a referent occasion. 5 Language 1 See Signs, Language and Behavior, p. 36. 2 The combination of nucleus with qualifiers is what Otto Jespersen calls a 'junction,' in contrast to the combination of 'nexus' connecting subjects with predicates. A junction, he says, is a 'unit or single idea, expressed more or less accidentally by means of two elements', while a nexus 'always contains two ideas which must remain separate.' See The Philosophy of Grammar, London, Allen & Unwin, 1924, p. 116. 3 Roman Jakobson notes how 'decisively important' is the capacity to use subject-predicate sentences in the child's development: 'It liberates speech from the here and now and enables the child to treat events distant in time and space or even fictitious.' See his 'Verbal Communication,' Scientific American, 227 (1972): 73-80. R.M. Yerkes cites an experiment that seems to illustrate the different psychological capacities of interpreters of signals and sentences. A chimpanzee can be shown food put in one of four differently colored boxes, but if their positions are interchanged it will not be able to discriminate between the boxes and find the food after a delay. It will find the food, however, if the positions of the boxes remain unchanged. Thus, lower animals seem capable of delayed reactions relating a place to a kind of object (the food), but only humans as interpreters of sentences seem to relate one kind of object (a colored box) to another kind. See R.M. Yerkes, Chimpanzees, A Laboratory Colony, New Haven, Yale University Press, 1943. 4 We are restricting ourselves here to what Quine calls occasion sentences, sentences whose truth or falsity varies with occasion of utterance. These are contrasted to standing sentences, general sentences such as 'All crows are black' whose truth values are at least relatively independent of this occasion. See Word and Object, pp. 35, 36. 5 The view that temporal determination belongs logically in the subject term is argued for by Balzano. It is a given object A at time to which we ascribe an attribute b, says Bolzano, not an attribute b at t which we ascribed to A: 'It does not happen at time t that the attribute b is

Notes to pages 108-13

6

7

8

9 10

11

12

13 14

155

claimed for the object A; but the object A, inasmuch as it is thought to exist at time t (hence to have this determination) is claimed to have attribute b.' See his Theory of Science, trans. by R. George, Oxford, Blackwell, 1972. See O.H. Mowrer, 'The Psychologist Looks at Language' in L. Jakobovits and M. Miron (eds), Readings in the Psychology of Language, Englewood Cliffs, Prentice-Hall, 1967, for an attempt to provide a behavioral account of this meaning transference from predicates to subjects. For 'Tom is a thief', says Mowrer, the meaning of 'thief' gets transferred to that of 'Tom' in such a way that our response to Tom is modified. It is also sufficient to refute Kant's formulation of the analyticsynthetic distinction. For Kant a sentence of the form 'All Sis P' is analytic if P is used as a criterion for identifying individuals as an S, or in Kant's terms if P is 'part of the meaning' of S. But if P expresses a criterion for an S based on prior acceptance of what was false, then 'All S is P' is itself false, and hence not analytic. Jespersen notes how verbs like 'rain' and 'snow' were originally used without subjects, and then later pronoun subjects were added in the evolution of language to conform to the grammatical requirement for a subject. See The Philosophy of Grammar, p. 25. See footnote 27 of Chapter 4. Cf. Strawson, Individuals, London, Methuen, 1959, pp. 208ff. When criteria for identification are associated, nouns become capable of functioning as general subjects having what Strawson calls 'divided reference' to a plurality of objects. Peirce includes portraits with names within his category of signs having subject-predicate structure, contrasting them with portraits without names as 'icons.' See Collected Papers, 2.320. See also E.M. Gombrich, Art and Illusion, London, Phaedon Press, 4th ed., 1972, p. 59 for the view that a picture itself is never true or false, but can become so when it is accompanied by a label or caption. For this denotation-reference distinction see L.J. Cohen, 'Geach's Problem About Intentional Identity,' Journal of Philosophy, 65 (1968): 329-35. For the same distinction see also Keith Donnellan, 'Speaker Reference, Descriptions, and Anaphora' in P. French and H. Wettstein (eds), Contemporary Perspectives in the Philosophy of Language, Minneapolis, University of Minnesota Press, 1979. Searle, 'Proper Names,' Mind, 67 (1958): 166-73. Saul Kripke, 'Naming and Necessity' in Davidson and Harman (eds), Semantics of Natural Language. A proper name is for Kripke a 'rigid designator' designating the same individual in all possible worlds in contrast to definite descriptions which designate an individual in some but not all worlds. I ignore here Kripke's application of this distinction to modal logic except to note that possible worlds seem invoked to preserve the dyadic denotation relation for definite descriptions.

156

Notes to pages 114--25

15 This theory is outlined in Kripke's 'Naming and Necessity.' 16 I have in mind here only the referential use of a subject of an isolated sentence in contrast to what Donnellan calls an 'attributive use' of a subject in 'Reference and Definition Descriptions,' Philosophical Review, 75 (1966): 281-304. An attributive use approximates to the conditions for logical denotation, since it presupposes a narrative context in which predicates PI> Pz, ... , P n are ascribed. The subject term then becomes the definite description 'whoever (whatever) satisfies 'P1 +P2 + ... +Pn'. 17 P.T. Geach, 'Assertion,' Philosophical Review, 74 (1965): 449-65. 18 That subject negation exists was first pointed out to me by Geoffrey Nathan. I have argued these points at greater length in my 'Negating the Subject,' Philosophical Studies, 43 (1980): 349-53. 19 Philosophical Investigations, trans. by G.E.M. Anscombe, New York, Macmillan, 1953, Section 304. 20 Perhaps the clearest statement of this relation between meaning and truth conditions is in Davidson's 'Truth and Meaning' and 'The Method of Truth in Metaphysics,' both in Inquiries into Truth and Interpretation, Oxford, Clarendon Press, 1984. In the second of these papers Davidson says that a 'theory of truth' must 'show us how we can view each of a potential infinity of sentences as composed from a finite stock of semantically significant atoms (roughly, words) by means of a finite number of applications of a finite number of rules of composition. It must then give the truth conditions of each sentence ... on the basis of its composition.' That stating truth conditions has the effect of providing a synonymous paraphrase is argued by Dummett in 'What is a Theory of Meaning?' in S. Guttenplan (ed.), Mind and Language, Oxford, Clarendon Press, 1975. 21 See his 'Davies on Recent Theories of Metaphor,' Mind, 93 (1984): 433-9. Cooper raises the doubt whether any sentence meanings 'are computable on the basis of the meanings of their parts and modes of combination alone.' 22 See A. Tarski, 'The Concept of Truth in Formalized Languages' in Logic, Semantics, Metamathematics, Oxford, Clarendon Press, 1956. This formula becomes the basis for the theory of meaning developed by Davidson in 'Truth and Meaning.' 23 See J.L. Austin, How to Do Things with Words, Cambridge, Mass., Harvard University Press, 1962. 24 Louis Althusser calls this a kind of 'hailing,' in effect saying to another 'Hey, you there!'. See his Lenin and Philosophy, trans. by B. Brewster, London, Monthly Review Press, 1971, p. 174. 25 See The Language of Morals, Oxford, Oxford University Press, 1952, pp. 17, 18. The imperative shares with the indicative for Hare the common 'phrastic' 'Your closing the door' with subject-predicate structure. 26 See Kant's Anthropology from a Pragmatic Point of View, trans. by

Notes to pages 125-34 157

v.

27 28 29

30

31

32

33 34

Dowdell, ed. by H. Rudnick, Carbondale, Southern Illinois University Press, 1978, Bk. I, Sec. 39. Here Kant includes as 'voluntary signs' class signs of hereditary superiority (e.g. coats of arms), uniforms as signs of service, badges of honor (e.g. ribbons), and signs of dishonor such as brandings. This contribution by Kant to semiotic was pointed out to me by Thomas Wilson. Though frequently they function relative to a linguistic background and thus as what we have termed 'post-linguistic' signs. G.E.M. Anscombe, 'The First Person' in S. Guttenplan (ed.), Mind and Language. For my amendment to Anscombe's view see my 'The Addressing Function of "I",' Analysis, 38 (1978): 91-3. For a summary and criticism of what he calls Wittgenstein's 'noownership theory' see Strawson, Individuals, London, Methuen, 1959, pp. 95-101. I defend Wittgenstein against these criticisms in 'A Defence of the No-Ownership Theory,' Mind, 81 (1972): 97-101. This defense extends only to Wittgenstein's contention that the 'I' in avowals is not used by the speaker as a referring expression. See Strawson, Individuals, Chapter 3. See also Sydney Shoemaker, Self-Knowledge and Self-Identity, Ithaca, Cornell University Press, 1963 for a defense of Strawson's view. We must postulate the existence of persons as referents of'!', Shoemaker contends, because of the fact 'that corresponding to any first person statement there are third-person statements that are in a certain sense equivalent to it and certainly 'about something.' So first-person statements must also be about persons' (p. 13). See David Harrah, 'A Logic of Message and Reply,' Synthese, 63 (1985): 275-83 for an analysis of a formal memo that distinguishes its 'vector-specifier' from the 'body' of a message conveying information or instructions. The relationship between bivalence and the realist correspondence view is noted by Dummett in 'Truth' and 'Realism', both in Truth and Other Enigmas, London, Duckworth, 1978. Instead of the bivalence principle, however, Dummett uses acceptance of the law of excluded middle 'A or not-A' as dividing the realist from the 'antirealist.' The law follows as a direct consequence of the principle. For a discussion of the bivalence principle and the three-valued logics that result from denying it see my Deductive Logic, Carbondale, Southern Illinois University Press, 1974, Section 25. See Quine's 'Two Dogmas of Empiricism' in From a Logical Point of View, New York, Harper & Row, 1953. For a defense of the analytic-synthetic distinction see Strawson and Grice, 'In Defense of a Dogma,' Philosophical Review, 65 (1956): 141-58. For sentences of the form 'All S is P' where S is a natural kind term (e.g. 'All gold is yellow' or 'All tigers are felines') application of the analytic-synthetic distinction is more difficult, as was noted in footnote 7 of this chapter. For an account of the reference of natural kind terms that has the effect of denying

158 Notes to page 134 analyticity for predicates such as 'is yellow' and 'is a feline' see Kripke, 'Naming and Necessity' and Putnam, 'Is Semantics Possible?' in Mind, Language and Reality, Cambridge, Cambridge University Press, 1975.

N

AME INDEX

Aldrich, Virgil, 54n Alston, William, 43n Althusser, Louis, 124n Anscombe, G.E.M., 78n, 128 Aquinas, 20n Aristotle, 13, 18,20,27 Arnauld, Antoine, 22, 23, 27, 50 Augustine, 17, 19-23,37,84 Austin, J.L., 5, 40, 51, 77n, 122, 127 Bar-Hillel, Yehoshua, 91n Barthes, Roland, 30, 46 Bennett, Jonathan, 64n, 77n, 80n, 84n,86n Berkeley, George, 24, 51, 64 Black, Max, 8n, 54n Blackburn, Simon, 77n Bolzano, Bernard, 107n Bonaventure, 22n Brown, Roger, 32n, 84n Buhler, Karl, 38 Carnap, Rudolph, 2n, 33 Chisholm, Roderick, 34 Chomsky, Noam, 34n, 38, 39, 41, lOOn Churchland, Paul, 10 Clarke, D.S., Jr, 67n, 128n Cohen, L.J., 112n Cooper, David, 120 Davidson, Donald, 4, 70n, 119n, 121n De Saussure, Ferdinand, 29, 30, 37 De Sousa, Ronald, 94 Descartes, 6, 22, 64, 103

Diebold, A.R., 99n Donnellan, Keith, 112n, 115n Dretske, Fred, 54n Ducasse, c.J., 33n Dummett, Michael. 7n, 119n, 133n Eco, Umberto, 12n, 18, 38n, 48, 49 Empiricus, Sextus, 13, 14, 16 Fodor, Jerry, lOn Follesdal, D., 34n Frege, Gottlieb, 112 Friedman, Kenneth, 44n Geach, P.T., 116 Gentry, George, 28n Gombrich, E.M., 57n, 111n Goodman, Nelson, 57n Grice, H.P., 74-7, 81, 82,115 Guiraud, Pierre, 29 Hacking, Ian, 21n Hahn, Robert, 12n Hampshire, Stuart, 79n Hare, R.M., 125 Harman, Gilbert, 4n, 7n, 38, 48 Harrah, David, BIn Hart, H.L.A., 77n Hinde, R.A., 78-80, 90 Hippocrates, 13 Hobbes, 22 Hockett, C.F., 96n, 99n, lOIn Hume, David, 24, 44 Jackson, B.D., 19n, 20n Jakobsen, Roman, 106n Jespersen, Otto, 105n, 110n

159

160

Name Index

Jones, Andrew, 49n Kant, Immanuel, 63, lIOn, 125 Katz, Jerrold, 7n Klaus, Georg, 90n Kripke, Saul, 70, 86n, 113, 134n Langer, Suzanne, 90n Leff, Gordon, 22n Lewis, David, 85 Locke,John,22,37,51,120 Lorenz, Konrad, 36 Markus, R.A., 19n, 20n Mead, G.H., 73n, 78 Meinong, Alexius, 135 Menzel, E. W., 78 Mitchell, Thomas, 21n Morris, Charles, 33, 36, 40, 61n, 62n, 73n, 91, 105 Mowrer, O.H., 108n Nathan, Geoffrey, 116n Ockham, William of, 19,21,22, 28 Ogden, C.K. and Richards, I.A., 16n,31,51n Osgood, C.E., 32

Reid, Thomas, 23, 51, 52, 57 Romeo, Luigi, 12n Rorty, Richard, SIn Russell, Bertrand, 2, 3, 32n, 116, 119,135 Ryle, Gilbert, 2n, 5, 51 Schiffer, Stephen, 77n Searle, John, 77n, 113 Sebeok, Thomas, 12n, 30n, 36, 37, 47,90 Shanker, S.G., 89 Shoemaker, Sydney, 129n Silverman, Kaja, 29n Skinner, B.F., 36, 39, 69 Smith, W.J., 79 Speck, Charles, 26n Stich, Steven, IOn Stough, Charlotte, 16n Strawson, P.F., 4, 5, 7, 63, 77, 110, lIOn, 128n, 129; and Grice, H.P.,134n Succi, G., 32n Tannenbaum, P., 32n Tarski, Alfred, 121 Thorpe, L.P., 36 Tinbergen, N., 36, 80n Von Frisch, K., 79

Peirce, C.S., 8, 10,25,26,28,37, 38,40, 43,44,58,61n, 65,90n, ll1n Perry, Ralph Barton, 71n Plochmann, G.K., 22n Poincare, J.H., 133 Price, H.H., 44, 50 Putnam, Hilary, 44n, 134n Pylyshyn, Zenon, IOn

Waismann, Friedrich, 35n Warnock, G., 127 Watson, J.B., 32n Wenner, A.M., 99n Wilson, Thomas, 126n Wittgenstein, L., 2n, 6, 57n, 87n, 117,128

Quine, W.V.O., 33-5, 93n, 106n, 110, 133 Quintillian, 13

Ziff, Paul, 82n, 86

Yerkes, R.M., 106n

S

,UBJECT INDEX

acceptance, 107, 112 address, 130; as distinguished from subject, 125; non-linguistic, 125, 126; of sentence, 124; of signal, 95; source, 124; target, 95 addressee, 125 addressor, 125 affirmative stimulus meaning, 33 analogical extension, 8, 41 analytic-synthetic distinction, 133, 134 animals: learning by, 47; communication by, 89; experiences of, 62-3; sign interpretation by, 15 assent, 106, 107 associative sign, 16, 49

basic element, 98, 99, 105 basic particular, 129 behavioral semiotic, 31, 64 belief ascription, 70 causal explanation, 16 causal generalization, 45, 48, 49, 57 code, 91 cognitive science, 10 communicative intent, 8, 73, 80, 82,88,92, 115; referential, 117; of illocutionary force, 123 communicative system, 96 comsign, 73-6, 81, 82, 111; conditions for, 74-9; conventional, 84-8; degenerate, 78, 84; non-conventional, 82, 92; proper, 78, 84 conditioned reflex learning, 31

controlling stimulus, 35, 36, 69 convention, 84, 97; conditions for, 85-7; cultural, 83; as norm, 87; regularity theory of, 85 conventional sign, 17, 20-2, 24, 27,83,84 coordination problem, 85 definite description, 108, 112, 134; within imperative, 125 denotation, 3, 40, 111, 116, see also reference; causal theory of, 113, 114; as time-invariant, 116 denotatum, 111, 114 designation, see denotation direct response theory, 31, 32 directedness, 79 discourse, 130 disobedience, 88 displacement, 100-2 dispositional theory, 32-4 disquotational principle, 70 duality of patterning, 98, 99 Epicurean theory of signs, 14-16, 35,50n ethology, 9, 25, 78-80, 89,95,96, 102 evidence, 15; as interpreted, 75; as made available, 75, 77 falsity, 60, 88,106,119; of subject, 115,116 fiction, 134, 135 general systems theory, 39 grammaticality, 99, 100; strong, 100, 102; weak, 99

161

162 Subject Index habit, 28; change of, 33 horizontal classification, 8

logical form, 4, 5 logical interpretant, 28, 33

iconic sign, 22, 24, 26, 57, 121; as comsign, 83, 111; gesture, 101; map, 23, 111; as natsign, 56, 57; as natural sign, 51 identification, 106, 114, 115; criteria for, 110 illocutionary force, 94, 95, 122, 123, 131; indicator of, 94, 122, 123 illocutionary speech act, 122 imperative sentence, 65, 66, 71, 107, 118, 122, 124, 125, 131, 132; single-word, 94 index, 26-8, 43, 45, 50, 56 indicative sign, 16 inferential context, 3-5, 14,45, 132,133 information theory, 91 information: storage of, 102, 107, 108; transfer of, 107-9, 131, 132 instrumental learning, 35, 36 intentional inexistence, 94 intentionality, 34, 35, 79 intentions: as ascribed to animals, 73,74,80 interpretation: 59, 60, 118; cognitive, 65, 91; complete, 60; dynamic, 65, 67, 68, 71, 82, 91; by lower animals, 62; as meaning specification, 118, 119; of natsigns, 60-2; partial, 60, 62, 107; as recognition, 50; of utterances, 59, 60

mand,36 mass noun, 110 material image, 52-7 meaning, 31, 49, 81, see also significance; change of, 120; of iconic sign, 121; natural, 75; non-natural, 74, 75; occasion, 81,82; of sentence, 117, 118; specification of, 118, 119; timeless, 81; transference of, 112,113 mental schema, 48 mental word, 20-3, 24 mirror image, 22, 23, 52, 53 morpheme, 98, 99,104

judgment, 106 language: contingent features of, 6; definition of, 101, 102; necessary features of, 6, 7; universal features of, 7 lansign, 105 lexicon, 104 linguistics, 6, 25, 29 logical analysis, 2-5, 116, 119, 120, 136

natsign, 60-71, 75, 118; definition of,50 natural sign, 8, 19,20,22-4,26, 29,43-50,66, 75, 76 necessary sign, 13,44 nucleus of term, 105, 108 object: as referent, 116, 117; concept of, 93 onoma, 17 operant, 35, 36 ordinary language, 5; philosophy of,5-7 performative, 5, 6, 127, 129 performative prefix, 122, 123 philosophy of language, 37 photograph,40, 54,55,57, 111 picture, 23, 40, 111 post-linguistic sign, 91 practical inference, 66, 67 predicate: function of, 105; of imperative, 66 primitiveness, 8 principle of bivalence, 132 probable sign, 13, 45 pronoun: demonstrative, 27, 58; first person, 126-9; second person, 125, 129, 130

Subject Index proper name, 108; description theory of, 112, 113 proposition, 3, 14, 15 psychological prefix, 123 qualifiers, 105 quantifier, 106 quantitative sameness, 92, 93 quasi-intentionality, 94 quasi-subject, 127, 128, 130 reference, 112; its particularity, 93; of signal, 91; of subject term, 112 referent occasion, 8,59, 61, 63, 67,81,91,98,101,106,107, 118; extention of, 91, 106, 107; of natsign, 60; of utterance, 59 reidentification, 92, 93, 110 representational mediational response, 32 representational theory of perception, 51, 52 semantic field, 97, 98 semantic regularity, 86 semanticity, 96 semeiotikos, 12, 13 semiology, 29, 30 semion, 12-14,17,18,22, 50n semiotic, 7-11, 22, 25, 31, 36-41, 96,120 semiotic inference, 49 sensation: as referent, 117; as sign, 24,51,52 sensory image, 51; see also sensation sentence, 100, 105; analytic, 133; complete, 90; feature-placing, 110; general, 105; single-word, 25,90,91; singular, 58,105; synthetic, 133 shadow, 52, 56 sign disposition, 32, 33 sign: as conditioned stimulus, 30, 31; as criterion, 44; as evidence, 12, 19,21,23,36,43-50; longrange, 44; as medical symptom,

163

13,44,46; necessary features of, 25, 37, 96; as preparatory stimulus, 33; short-range, 44, 50 signa naturalia, 19 signal, 8, 90-2; complex, 98, 99; non-conventional, 92; simple, 98 signaling display, 78-80 signaling system, 96 signature, 95 significance, 31, 32, 82, see also meaning; change of, 61, 62, 68, 108, 109; dynamic, 67; of natsign, 60, 117, 118; of sentence, 58, 68; of signal, 92 significate occurrence, 61, 64, 65, 106,118 signum, 19 signum datum, 20 social releaser, 95 speaker reliability, 109 statement, 105 stellvertrendes, 38 stimulus generalization, 69 stimulus meaning, 33 Stoic theory of signs, 14-16,20,84 sub-element, 99 subject: function of, 105, 115, 122; functional, 134; grammatical, 116; of imperative, 66; logical, 2, 116, 135; structural, 134 subject-predicate asymmetry, 116 subject-predicate radical, 122 subject-predicate structure, 8, 26, 90,100,101 symbol, 17,27,28 symbola, 17 tact, 36 term, 105 theoretical context, 133, 134 token, 27, 28; of natsign, 61, 62; of sentence, 3; of signal, 92 truth, 40, 60, 88, 118; conditions of, 119, 121; constancy of, 3, 132; correspondence theory of, 88,89, 121; of subject, 115, 116; redundancy theory of, 89, 121 TV screen image, 52-7, 111

164

Subject Index

type, 27; of natsign, 61, 62; of signal,92 vertical classification, 8

vocal gesture, 78 zoosemiotics, 36, 37

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 4

DESCRIPTIONS IN CONTEXT

DESCRIPTIONS IN CONTEXT

CLEO A. CONDORAVDI

First published in 1997 by Garland Publishing, Inc. This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1997 Cleo A. Condoravdi All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69699-0 978-1-315-52189-3

(Set) (Set) (ebk) (Volume 4) (hbk) (Volume 4) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

DESCRIPTIONS IN

CONTEXT

CLEO A. CONDORAVDI

GARLAND PUBLISHING, INc. NEW YORK & LONDON /1997

Copyright© 1997 Cleo A. Condoravdi All rights reserved

Library of Congress Cataloging-in-Publication Data Condoravdi, Cleo A., 1962Descriptions in context I Cleo A. Condoravdi. p. em.- (Outstanding dissertations in linguistics) A revision of the author's thesis (Ph. D.)-Yale University, 1994. Includes bibliographical references and index. ISBN 0-8153-2871-0 (alk. paper) 1. Definiteness (Linguistics). 2. Semantics. 3. Grammar, Comparative and general-Number. I. Title. II. Series. P299.D43C66 1997 401'.43-dc21 97-2202

Printed on acid-free, 250-year-life paper Manufactured in the United States of America

Contents

PREFACE

IX

1

INTRODUCTION

2

BARE PLURALS AND GENERICITY 1 Introduction 2 Genericity and the Readings of Bare Plurals 2.1 Genericity 2.2 Bare Plurals 3 The Kind Analysis 3.1 Basic Outline 3.2 Arguments for a Unified Treatment

5 5 6 6 8 9 9 17

3.3 Limits and Limitations of Uniformity 3.4 Summary 4 The Indefiniteness Analysis 4.1 Basic Outline 4.2 Individual-Level and Stage-Level Predicates 4.3 The Generic Operator

31 33 33 33 37 39

4.4 Conclusion Appendix Notes

45 46 47

FUNCTIONAL READING OF BARE PLURALS 1 Introduction 2 The Functional Reading of Bare Plurals 2.1 Initial Observations 2.2 Functional Reading with Individual-Level Predicates

55 55 57 57

3

3.2.1 Bare Plurals Are Unambiguous 3.2.2 Bare Plurals and Indefinite NP's 3.2.3 A Generic Operator and Bare Plurals

4.3.1 Implicit Domain Restrictions 4.3.2 Modal Dimension

3

v

18 19 26

39 41

58

vi

Contents 2.2.1 Genericity and the Functional Reading 2.2.2 Contextual Restrictions 2.3 Functional Reading with Stage-Level Predicates 2.4 Other Indefinites 2.5 Excluding a Purely Pragmatic Account 2.5.1 The Implicature Approach 2.5.2 The Referentiality Approach 2.6 Functional Reading in Quantified Contexts 2.6.1 Dependent Functional Reading 2.6.2 Quantificational and Modal Subordination 2.7 The Presupposition of Existence 2.7.1 Simple Cases 2.7.2 Frojection of the Existential Presupposition 2.7.3 Functional Reading with Adverbs of Quantity 2.8 Overview 3 An Operator Analysis 3.1 Degenerate Genericity 3.2 Degenerate Genericity and the Functional Reading 3.2.1 Implicit Contextual Restrictions 3.2.2 Implication of Existence 3.2.3 Scopal Interaction 3.3 Stage-Level Predicates 3.4 Positive Contextual Sensitivity 3.5 Summary 4 Conclusion Notes

4

STRONG AND WEAK NOVELTY

1 Introduction 2 The Dynamic View on Meaning 2.1 Assertions and Contextual Update 2.2 Presuppositions and Contextual Admittance The Novelty-Familiarity Theory of Definiteness and 3 Indefiniteness 3.1 Files as Information States 3.2 Informativeness of Files 3.3 The Felicity Conditions of Definites and Indefinites 3.4 The Truth Conditions for Definites and Indefinites 4 The Functional Reading 4.1 Weakly and Strongly Novel NP's

59 67 70 73

74 75 79 81 82 87 89 90 92 95 100 102 103 106 107 114

115 121 123 129 130 130 141 141 143 143 152 157 157 159 162 165 168 168

Contents

vzz

4.2 4.3 4.4 4.5

Contextually Salient Functions Negative Contextual Sensitivity Strong and Weak Novelty and NP Strength Existential Force and Strong vs. Weak Novelty 4.6 Maximality 4. 7 Consequences of the Existential Presupposition 4.7.1 Positive Contextual Sensitivity 4. 7.2 Dependent Functional Reading

5 Conclusion Notes

172 174 175 177 178 179 179 180

181 182

BIBLIOGRAPHY

187

INDEX

199

Preface

This is a slightly revised version of my dissertation, submitted to Yale University in 1994. The references have been updated, some stylistic changes were made throughout the text, and some clarificatory comments were added in chapter 2. Since the dissertation was finished, my thinking about some of the issues involved has changed but not in ways that affect the central claims of this work. Acknowledgments. I would like to thank the members of my committee, Donka Farkas, Bill Ladusaw, Annie Zaenen, and Larry Horn, for their advice and encouragement through the course of this work. My advisor, Donka Farkas, has followed its development from vague hunches, to formed ideas, to a spelled out proposal with insight and enthusiasm. She has been very generous with ideas, suggestions, comments, support, and, not least, hospitality during my trips to Santa Cruz. Both the general argument and the specific workings of details owe a lot to our discussions. I hope that, after all the versions she has seen, this one is closer to the simplicity and explicitness she was urging all along. Many thanks to Bill Ladusaw and Annie Zaenen for readily agreeing to serve on my committee. Bill has a surprising way of turning a marginal point into a decisive one, of redirecting the focus of an argument and of bringing out what is essential in a proposal, often with one well-chosen phrase. I am very glad for having had the opportunity to discuss with him much of the material in this thesis. Annie's faith in me has made a big difference in more ways than I can say and, more tangibly, this thesis is much the better for her involvement with it. She never failed to detect a weakness or a hole in an argument, even in an area that is not her own, but also to acknowledge that some point had finally become convincing. Thanks to Larry Horn for many useful comments on a near-final draft and for taking care of the logistics of submitting from afar.

X

Preface

Tony Davis and Chris Pifi6n have been wonderful friends throughout this time and have contributed in no small ways. They have ungrudgingly read countless versions of various parts and their critical comments helped clarify my claims and improve the style in which they are presented. Tony was always ready to listen to an idea, to help sort out the central from the peripheral or sharpen an example, and brought sanity and rationality in moments of panic. Chris's skeptical scrutiny and meticulous attention to detail were invaluable checks for hasty conclusions or uncalled-for optimism. I have greatly enjoyed and benefited from our discussions, in the confines of my office or the expanse of the Stanford hills. I am lucky that Sandro Zucchi, Henriette de Swart and Kees van Deemter were at Stanford during various stages of this work. I would like to thank them for the interest they took in my work and the challenging questions they asked. Discussions with David Israel, a maven on just about everything, have always proved revealing both in the short and in the long run. I am also grateful to Mark Gawron, Eric Jackson, Sam Bayer, and Louise McNally for discussions and comments that shaped the direction of this work early on. Thanks to Henriette de Swart and Paul Kiparsky for thorough and detailed comments on draft versions of chapters 2 and 3 that led to considerable improvements and to Andras Kornai for a close reading of a pre-final version of chapter 2. Finally, many thanks to Annie Zaenen, Ivan Sag, Geoff Nunberg, John Perry, and Dikran Karagueuzian, who have made it possible for me to be at CSLI before and after the dissertation was completed, and to Tom Burke for his expert technical help in preparing the book manuscript.

Descriptions in Context

1 Introduction

This work is a contribution to the semantics of definite and indefinite descriptions. I take as a starting point the presuppositional theory of definiteness and indefiniteness proposed by Heim. Heim's theory shares with more traditional approaches a conception of definiteness as a simple binary opposition. I show that there exists a special type of indefinites that have an interpretation commonly associated with definites. More specifically, in the novelty /familiarity theory of indefinite and definite NP's that Heim (1982) develops, there is a one-to-one correspondence between the definiteness of an NP and the felicity conditions it is associated with. Indefinite NP's are associated with a novelty condition for their index, definite NP's are associated with familiarity conditions for their index and their descriptive content. I argue that the felicity conditions associated with indefinite NP's can vary. In addition to indefinite NP's that always assert existence, there are indefinites that may presuppose existence. Bare plurals in English exemplify this special type. I develop a more fine-grained theory of novelty within the framework of File Change Semantics. The theory accounts for the full range of readings of the special type of indefinites, explains why their additional interpretations appear in a precisely circumscribed range of contexts and gives a unified characterization of indefiniteness within which the parameters of difference between regular and special indefinites can be naturally formulated. Formally, the analysis I propose allows the felicity conditions of a given noun phrase to interact with its file change potential to the following effect. The admissible contexts can be partitioned into those that entail in part the information added by the processing of the noun phrase and those that either fully entail that information (for 3

4

Introduction

definites) or do not entail any part of that information (for indefinites). Definites and indefinites of the traditionally recognized kind are then those that, given their felicity conditions, would have as admissible contexts only those of the latter type. Indefinites of the newly discovered type are those whose admissible contexts, given their felicity conditions, would be contexts of the former kind. More generally, this work can be seen as providing an empirical argument in favor of a dynamic theory of meaning and against the more traditional truth-conditional theory of meaning with its strict separation between semantics and pragmatics and its restricted conception of context-dependency. The empirical evidence is of a new kind-most of the evidence in the literature concerns anaphoric binding-and helps sort out two ideas. One is the idea that NP's should be distinguished according to whether they are inherently quantificational or not, allowing the semantic scope of non-quantificational NP's to extend beyond their syntactic scope. It was meant to account for donkey and intersentential anaphora and the variable quantificational force of indefinites. The other is to view meaning as a dynamic notion, an idea that can be characterized with the motto: the meaning of a sentence is the way in which it changes a given body of information. These two ideas, which came together in the work of Kamp and Heim, have been very influential and have guided a large part of recent work in semantics. Since, however, one way of implementing the first idea is in terms of the general setup that dynamic semantics provides, they are also often conflated. That they are independent of each other was already demonstrated by Heim {1982), with the difference in the semantic systems of chapters II and III of that work. The same point has also manifested itself implicitly in the direction of subsequent work that has adopted the former but has rejected the latter, such as Kratzer {1995} and Diesing {1990, 1992a). The facts and the analysis I present argue in favor of both ideas.

2 Bare Plurals and Genericity

1. INTRODUCTION This chapter provides the background on bare plurals and genericity necessary for the discussion of the following chapters. It contains an overview of the familiar problems posed by bare plurals and generics and of two influential analyses that have so far been proposed. Its main focus are the empirical and theoretical motivations behind each analysis and the crucial assumptions underlying it. The semantics of bare plurals is inextricably linked to the semantics of genericity and therefore they must be treated as interdependent in any analysis. Bare plural NP's pose a particular challenge because of their apparently divergent behavior. On the surface, at least, they can be interpreted as indefinite descriptions, as quantificational NP's with varying force, or as expressions referring to kinds, individuals of a special sort. The basic question then is whether they are ambiguous and, if so, in how many ways. The central issue concerning the linguistic analysis of genericity is whether the regularity expressed by generic sentences should be built into their semantics as a generalization/quantification of some sort. 1 I will distinguish two kinds of approaches to generics: direct generalization approaches, which specify the truth conditions of generic sentences in terms of quantification, and inferred generalization approaches, which specify the truth conditions of generic sentences in some other way. I will concentrate on Carlson's ( 1977b) classic kind analysis and on the indefiniteness analysis along the lines of Heim ( 1982), Gerstner & Krifka (1987), Krifka (1987), Krifka et al. (1995) and Wilkinson (1988a, 1991). Carlson's analysis is an inferred generalization approach to generics that takes bare plurals to be expressions de5

Descriptions in Context

6

noting kinds. The indefiniteness analysis is a direct generalization approach that takes bare plurals to be indefinite descriptions with ordinary individuals in their denotation. Other approaches combining features of these two will be mentioned along the way. Although Carlson's theory and the indefiniteness theory differ in terms of the semantics they give to genericity and to bare plurals, they are similar in making a clear separation between the meaning of a generic statement, which the semantics they propose is supposed to capture, and the kind of reasoning or action that belief of a generic statement may give rise to, which they consider as outside the purview of a semantic analysis. 2 In section 2 I give a brief overview of the problems raised by generics and bare plurals. In section 3 I present and discuss the kind analysis and in section 4 I present and discuss the indefiniteness analysis. I will ultimately side with the indefiniteness analysis of bare plurals and the direct generalization approach to generics and these two positions are crucial for the material to follow in this book. However, given the uniformity, intuitive appeal and great empirical coverage of Carlson's analysis, I will present it in some detail and will show that some of the arguments he provided in favor of a kind analysis can be turned around as motivating the indefiniteness analysis of bare plurals and the direct generalization approach to generics.

2. GENERICITY AND THE READINGS OF BARE PLURALS 2.1

Genericity

Generic sentences express regularities and non-accidental generalizations, which among other things play an important role in reasoning. On the basis of a generic statement we can draw inferences, albeit defeasible, about particular occurrences or particular objects. The generalizations can be either descriptive or normative. The linguistic means of expressing genericity are varied: there is nominal and verbal genericity, and often no special expression is designated as exclusively generic. Thus definite and indefinite descriptions may have a generic use in addition to their usual uses; similarly, often the same tense can be used in an episodic and in a generic context. The sentences in [1) are all intuitively identified as generic sentences, yet the source of genericity, the means used to express it,

Bare Plurals and Genericity

7

and the inferences it supports about particular instantiations are different in each case. (1]

a. b. c. d. e.

Punica granatum is an Old World tree. The 3-R4* generation robot will soon be antiquated. A pirate ship flies a black flag. Pomegranates have a crowned end. In this place one whistles a lullaby to show approval.

(1a) contains the name of a biological species and asserts something about the origin of that species. Members of the species planted in America are not, strictly speaking, Old World trees, rather they are descendants of trees that originated in the Old World. In (1b) the definite description also identifies a kind whose instantiations satisfy the descriptive content of the NP. [1b) is similar to [1a) in that the conditions placed on a kind for it to be considered antiquated are different from those placed on its particular instantiations. For example, the kind could be antiquated well before any one of its instantiations. Thus, in [1a) and [1b) the NP denotes a kind and the predicates can be meaningfully predicated of both a kind and an individual of the usual sort but the conditions under which these properties can be said to hold of a kind and of an individual are quite different. [lc]-[le] are more complicated. Do apparently indefinite descriptions identify a kind as well, and is that the source of genericity in [lc) and [1d)? The question becomes particularly vexing with respect to (1e), where it is unclear what the relevant kind should be. Moreover, it seems that, in contrast to [1a] and [1b], the truth or falsity of [1c]-(le] has something to do with whether particular instances of pomegranates, pirate ships or people who are in this place have the relevant property. (1c]-[le] seem to express a generalization about actual and potential entities satisfying the descriptive content of the singular indefinite, that of the bare plural, or the property of being in this place, respectively. For descriptive purposes, I will follow Link (1988, 1995) and use the term proper kind predication for cases like (1a) and (1b) and the term derived kind predication for cases like [1c)-[1e). Generic generalizations have two hallmarks that distinguish them from actual, universal generalizations: they are not sensitive to the way things actually are but they are sensitive to alternative ways things might be or ought to be. As a consequence, generic generalizations are at the same time weaker and stronger than the corresponding actual accidental generalizations. They are weaker

Descriptions in Context

8

in that they tolerate exceptions, that is they can be true even on the face of exceptional cases. For example, [ld) is true even though, as it happens, there exist pomegranates whose crowned end has been chopped off. In fact, [ld) could still be true even if the crowned end of all pomegranates in existence were to be chopped off. They are stronger in that their truth does not depend simply on actual instantiations. 3 For example, it does not suffice for [lc) to be true that all the pirate ships in existence fly a black flag since this could happen as matter of accident, without any such convention in place. Therefore, actual generalizations are not necessarily generic.

2.2 Bare Plurals In some cases a bare plural uncontroversially denotes a kind. In [2a], for example, the bare plural must denote a kind since the predicate be extinct can apply only to kinds and not to ordinary individuals. In addition, bare plurals exhibit both a generic or (quasi-)universal reading and an existential reading, each one of which is related to the kind of reading the whole sentence receives. The universal reading is tied to a generic or habitual generalization reading for the whole sentence. The existential reading is tied to an episodic reading for the whole sentence. [2b] and [2c], in which the bare plural appears to have universal and quasi-universal force, 4 respectively, express a characteristic property of dogs in general, while [2d), in which the bare plural appears to have existential force, is about an instance of barking by some dogs. [2)

a. b. c. d.

Dinosaurs are extinct. Dogs are mammals. Dogs bark. Dogs barked last night outside my window.

Moreover, while the bare plural can get the universal/generic reading with all predicates, the existential reading is confined to a subclass of predicates (all episodic and a few stative predicates). Observing these correlations, Carlson (1977a, 1977b) argued that the bare plural itself is not ambiguous, its different readings being the result of the semantics of the predicates it combines with. Subsequent work has in part accepted Carlson's arguments for a unified treatment of some of the readings of the bare plural but has sought it elsewhere, guided by different assumptions on the origin of genericity. Specifically, it has sought a unified treatment for the

Bare Plurals and Genericity

9

existential and generic readings of [2d], [2b] and [2c] but not of the one involving reference to kinds, as in [2a].

3. THE KIND ANALYSIS 3.1

Basic Outline

Carlson, in a series of works (1977a, 1977b, 1979, 1982), provides a uniform analysis of bare plurals, in which they rigidly denote kinds. 5 The multitude of their readings is derived from the different types of predications they appear in. Central to the analysis is the ontological assumption that the domain of individuals must be enriched to include kinds and the semantic assumption that the source of genericity resides in the intensionality of individuals. 6 Carlson proposes that kinds are individuals in their own right which have different realizations (or manifestations) relative to different points of evaluation (i.e., worlds and times). In that sense they are intensional individuals. Following a proposal of Gabbay & Moravcsik (1973), Carlson goes a step further in assuming that ordinary individuals are also intensional in that they too have different realizations relative to different points of evaluation. 7 The semantic proposal consists of two main claims. The first claim is that the regularity or generalization expressed by a generic statement is essentially indirect and, crucially, not part of its truth conditions: a generic statement attributes a property to an individual which can then be inferred to be "inherited" by a certain number of its extensional manifestations. A property of a kind is inherited by some of its realizations, and a property of an object is inherited by some of its spatio-temporal manifestations, but these conditions do not constitute part of the truth conditions of a generic statement. As Carlson (1977b:109) puts it: "Instead of asking the wrong question ('What is the generic quantifier?'), we are now asking the right question ('How can we infer quantification from generic sentences?')." 8 The second claim, crucially connected to the first, is that a generic sentence comprises simply a predication, albeit one of the right type. As Carlson (1979:65) states: "generic attribution is considered to be nothing more than attribution of a predicate to an individual, rather than to a stage of that individual." Within the model structure proposed by Carlson kinds, objects and stages are all basic entities of the model. Kinds and objects are the intensional individuals, stages their spatio-temporal man-

Descriptions in Context

10

ifestations. NP's may denote kinds or objects but no NP denotes stages. The three types of entities are related by two realization relations: R relates stages to individuals (objects and kinds}, R' relates objects to kinds. Meaning Postulates ensure that kinds, objects, stages and the realization relations satisfy the conditions in (3] (see Carlson (1977b:414-416} for a complete list of Meaning Postulates and their exact formulation). 9 [3]

a. every stage is the stage of some individual (object or kind} b. different objects cannot have the same stages c. if a stage realizes some object and that object realizes some kind, then the stage realizes that kind d. a stage realizing a kind is also a stage of an object realizing that kind e. individuals with a stage realization in some world exist in that world f. any two kinds with exactly the same realizations in all worlds are identical. A stipulated equivalence between kinds and their object realizations allows for a correspondence between the kind denoted by a kind-denoting NP and the intension of the common noun {CN) predicate of the NP. Namely, if the NP denotes the kind k, the intension of the CN and the set of sets defined by AxR'(x, k) at each point of evaluation are identical. In other words, at every world and every time, the set of objects satisfying the descriptive content of the NP will be exactly the same as the set of objects which (at that world and that time) realize the kind denoted by the NP. Thus, the translation of the bare plural NP dogs is as in (4] .10

[4]

AP-P(~xk

(0 'Vz 0 (dog'(z 0 }

H

R'(zo,xk))))

The denotation of the bare plural then is the property set of the unique kind whose realizations relative to every point of evaluation are exactly those belonging to the extension of the CN dog relative to that point of evaluation. Given the meaning postulate described in [3f], uniqueness is guaranteed for all models. Thus, the equivalence between kinds and their object realizations that is built into the translation of bare plurals and the meaning postulate guaranteeing uniqueness make bare plurals rigid designators. In accord with the tripartite split of entities into stages, objects and kinds, predicates are sorted according to whether they take stages, objects or kinds as arguments. Individual-level predicates

Bare Plurals and Genericity

11

take individuals (i.e., kinds or objects) as arguments, stage-level predicates take stages of individuals as arguments. Individual-level predicates and stage-level predicates denote properties of different types of entities; giving a predicate an argument of the wrong type results in sortal incorrectness. Empirically, it appears that the same predicate (corresponding to a full VP) can have both an episodic and a characteristic property reading, or, in terms of the analysis, it can apply both to objects and to kinds. Based on that evidence the finer typology in [5] emerges. 11 [5]

a. predicates that can apply only to kinds (strictly kind-level predicates) b. predicates that can apply both to objects and to kinds but not to stages (strictly individual-level predicates) c. predicates that can apply to stages, objects and kinds and have a characteristic property reading d. predicates that can apply to stages, objects and kinds but do not have a characteristic property reading

Strictly kind-level predicates can be either stative (e.g., be extinct, constitute a species) or episodic (e.g., become extinct, populate) _12 Bare plurals construed with such predicates are intuitively understood as referring to a kind. Strictly individual-level predicates are always stative and have a characteristic property reading. They are comprised of be plus an adjectival predicate or a predicative NP (e.g., be intelligent, be a bore)P Bare plurals construed with such predicates have a generic/ (quasi-)universal reading. Predicates of the category [5c] are episodic when applied to stages, while they express a characteristic property when applied to objects or kinds; they comprise all verbal predicates except for be (e.g., bark). Bare plurals construed with such predicates have an existential reading when the predicate is episodic (as in [2d]) and a generic reading when the predicate has a characteristic property interpretation (as in [2c]). Finally, predicates of the category in [5d] are comprised of be plus an adjectival predicate or a predicative PP (e.g., be available, be in the corner). Bare plurals construed with such predicates have an existential reading. 14 Schematically, the generalizations are summarized in [6], where individuals comprise both objects and kinds.

12

[6]

Descriptions in Context

Predicate

Argument

Type of

Interpretation of Predicate

Interpretation of Bare Plural

(5a]:

kinds

stative or episodic

kind-referring

(5b]:

individuals

characteristic property (stative)

universal

[5c]:

stages individuals

episodic characteristic property

existential universal

[5d]:

stages or individuals

stative (but not characteristic property)

existential

Type of

extinct intelligent

bark

available

Carlson's analysis captures these generalizations by distinguishing between basic and derived predicates within the system of Intensional Logic {IL), into which expressions of English are mapped by the translation mapping. Derived predicates are derived from basic predicates in two ways: (a) a stage-level predicate is mapped to an individual-level predicate containing existential quantification over stages, (b) appropriate generalization operators raise the level of a given predicate creating an intensional context. There are no derived stage-level predicates. Derived predicates of the first kind are needed so as to make a basically stagelevel predicate capable of composing with an individual-denoting NP. If a' is a stage-level predicate of (the sorted) type , then ..\xi3y 8 [R(yS, xi)&a' (y 8 )] is an individual-level predicate of (the sorted) type . Derived predicates of the second kind characterize the characteristic property reading associated with the majority of generic statements. Gn is the generalization operator mapping stage-level predicates to individual-level predicates and Gn' the generalization operator mapping object-level predicates to kind-level. Both Gn and Gn' are intensional VP operators. More precisely, if a' is a predicate of type or , then application of the generalization operators Gn or Gn' results in the predicates GnCa') and Gn'Ca'), which are of type and , respectively. Basic predicates and derived predicates of the first type give rise to predications unmediated by a generalization operator; derived predicates of the second type give rise to mediated predica-

Bare Plurals and Genericity

13

tions, which create an intensional context for the VP. 15 Unmediated predications arise when basically object-level predicates combine with object-denoting NP's, 16 when basically kind-level predicates combine with kind-denoting NP's, or when individual-level predicates with existential quantification over stages combine with individual-denoting NP'sP Mediated predications arise when an individual-level predicate derived from a basically stage-level predicate combines with an individual-denoting NP, 18 or when a kindlevel predicate derived from a basically object-level predicate combines with a kind-denoting NP. All individual-level predicates with stage-level counterparts are derived from the latter either via the mapping described in (a) above or via the generalization operator Gn. Most kind-level predicates with object-level counterparts are derived from the latter via the generalization operator Gn'. Exceptions are predicates like be popular, be well-known, etc., which do not give rise to a generic reading with singular indefinites. These are assumed by Carlson to be both basically object-level and kindlevel. The typology of predicates according to their surface distribution outlined in [5] and [6] corresponds to the classification within the system of IL given in [7]. [7]

a. basically kind-level: be-extinct', populate', be-popular' b. i. basically object-level: be-intelligent', be-popular' ii. derived kind-level: Gn' (~be-intelligent') c. i. basically stage-level: smoke' ii. derived individual-level with existential quantification over stages: ..\xi3y 8 [R(y 8 , xi) & smoke'(y 8 )] iii. individual-level derived via a generalization operator: Gn(~ smoke') d. i. basically stage-level: be-available' ii. derived individual-level with existential quantification over stages: ..\xi3y 8 [R(yS,xi) & be-available'(y 8 )]

Predicates of the categories [7ci] and [7di] do not correspond directly to the translation of any natural language predicate. Natu-

14

Descriptions in Context

ral language predicates that are stage-level need to compose with individual-denoting NP's, hence their translation is always a derived predicate. Carlson can analyze bare plurals and other generic NP's as uniformly kind-denoting by having (i) a sufficiently rich ontology, (ii) constraints on models (meaning postulates) which regulate the elements of the ontology, (iii) predicates selecting for the sorts of entities they can apply to, (iv) mappings from basic predicates to predicates which can apply to entities of a different sort. A kind-denoting NP, such as a bare plural, always combines with a kind-level predicate; different readings for the NP arise depending on whether the predicate is basic or derived, and, if derived, on whether it is mediated by a generalization operator or not. If the predicate is basic or mediated by a generalization operator (categories [7a], [7bii], [7ciii]), then there is no quantification either over objects or over stages in the semantics. If the predicate is derived but unmediated (categories [7cii] and [7dii]), then there is quantification over stages in the semantics. Thus the apparent universal force of the bare plural, arising when the kind-level predicate it combines with is derived from an object-level predicate or has an object-level counterpart but is not derived from it (e.g., be popular, be well-known,) is not captured in the semantics. The intuition that there is quantification over {unexceptional) objects satisfying the descriptive content of the bare plural is attributed to pragmatic inferencing. Distinguishing predicates such as be popular, be well-known from the rest is an artifact of Carlson's analysis, which tries to capture the fact that singular generic indefinites do not have a generic reading with such predicates. Carlson analyzes singular generic indefinites as kind-denoting but assumes that they combine only with derived kind-level predicates. While this also correctly excludes basically kind-level predicates like extinct from applying to them, it is a rather ad hoc way of accounting for the difference between bare plurals and the singular generic indefinite {see Farkas (1985)). The existential reading of the bare plural, arising when the kind-level predicate is derived from a stage-level predicate, is attributed to the existential quantification over stages of individuals built into the translation of such predicates. The apparent existential quantification over objects satisfying the descriptive content of the NP is the result, on the one hand, of the semantics, which gives us existential quantification over stages, and, on the other, of

Bare Plurals and Genericity

15

the meaning postulates spelled out in [3d] and [3e] above, which guarantee that the relevant stage also realizes an object existing in the actual world and realizing the same kind. The analysis, moreover, predicts that bare plurals will not have an existential reading with basically individual-level predicates, thus correctly capturing a very important and otherwise mysterious generalization. Specifically, [2a], [2b], [2c], [2d] receive the translations in [8a], [8b], [8c] and [8d], respectively (ignoring tense and adverbial modifiers). A bare plural, analyzed as denoting a unique kind, receives the translation AP~P(ak), where ak is a constant picking out a kind-level entity. 19 The predicate extinct is basically kindlevel, hence [8a] involves an unmediated predication. The predicate mammal is basically object-level and can be raised to kind-level; hence [8b] involves a mediated predication. The predicate bark is basically stage-level and can either be raised to object-level or kind-level via a generalization operator or simply by mapping to an individual-level predicate and introducing existential quantification over stages; hence [8c] involves a mediated predication and a generic reading for the bare plural, while [8d] involves an unmediated predication and an existential reading for the bare plural.

(8]

a.

b. c.

AP~P(din)Cextinct')

exti net' (din)

=>

AP~P(dog)CGn'Cmammal')

Gn'(Amammal')(dog))

AP~P(dog)CGnCbark')

=>

=>

GnC bark') (dog)) d. AP~P(dog)(A Ayi 3x 8 (R(x 8 ,d) & bark'(x8 ))) => 3x 8 (R(x 8 ,dog) & bark'(x 8 )) Relative to a given world and time, [8a] is true iff the individual kind dinosaurs is in the set of kind-level entities that are extinct in that world and that time, [8b] is true iff the individual kind dogs is in the set of kind-level entities that are mammals in that world and that time, [8c] is true iff the individual kind dogs is in the set· of kind-level entities that bark in that world and that time, and [8d] is true in a given world and time iff the kind dogs has stage realizations in that world and that time which are in the set of stage-level entities that bark. 20 The truth conditions make no reference to how many object or stage realizations of a kind possess the property denoted by the predicate in order for the predicate to be truthfully predicated of a kind. In a given model and relative

Descriptions in Context

16

to a given point of evaluation a kind is either in the denotation of a kind-level predicate or not, and that is independent of how many stage or object realizations of that kind, if any at all, are in the denotation of the equivalent stage or object-level predicate, either in that point of evaluation or in any other point of evaluation. This is what makes this analysis fall within the category of inferred generalization approaches. 21 Since the truth of derived kind-level predications does not depend on the truth of object-level predications in any way, the analysis sidesteps the problem of exceptions and captures the distinction between generic and actual generalizations. Generic generalizations, in contrast to actual generalizations, are only apparent, involving in reality a kind-level predication. But, in doing so, the analysis has a serious shortcoming in that it does not capture the distinction between different kinds of generic generalizations, such as descriptive, dispositional, and normative generalizations. While the operators Gn and Gn' are supposed to capture all the different kinds of generic generalizations, they cannot discriminate among them. In the same way that the truth of a kind-level predication relative to a given point of evaluation does not depend on the truth of the corresponding object-level predication relative to any point of evaluation, similarly, it does not depend on the truth of the kindlevel predication relative to any other point of evaluation. But this is precisely what we need if we are to capture normative generalizations.22 This shortcoming is serious, as different types of generic generalizations must be distinguished since they are not always truth-conditionally equivalent. For example, a descriptive generalization might be true without the corresponding normative generalization being true as well, as in [9a), and vice versa, as in [9b]. [9)

a. People around here shoot at each other indiscriminately. b. A country with rich natural resources shares them with its neighbors.

Carlson runs into further problems with respect to dispositional and normative generalizations because of two meaning postulates he posits requiring that if GnC a') can be truthfully predicated of some individual and that individual has at least some (past) stages then a' can be truthfully predicated of at least one stage. This might be true of descriptive generalizations but it is certainly not true of normative generalizations. For example, 'In chess bishops move diagonally' may be true in the actual world even

Bare Plurals and Genericity

17

if no bishop has ever moved diagonally because, say, no game of chess has ever actually been played (although the rules of the game are in place) or every game of chess has been played incorrectly. It is important to note that an inferred generalization approach to generics is independent of a uniform analysis for the bare plural. For example, one could have a direct generalization approach to generics (at least those that do not involve proper kind predication) and a uniform analysis for the bare plural as kind-denoting. This is the position taken by Farkas (1985}, who assumes that there is a generic operator quantifying over object realizations of the kind denoted by the bare plural, and by Schubert & Pelletier (1988, 1989}, who assume that bare plurals denote kinds and have a meaning postulate requiring the existence of an object realizing the kind in every world. 23 Similarly, one could have a uniform analysis of the generic readings of the bare plural and a separate analysis for its existential reading.

3.2 Arguments for a Unified Treatment Carlson designed his theory to provide a uniform analysis of generics and a unified analysis of the different readings of bare plurals. All generic sentences involve the attribution of an individual-level property to an individual (object or kind) and a bare plural always denotes a kind. This approach was justified by the following three claims, which he defended in great detail: (1) The ambiguity resides in the predicate and not in the NP. (2) On their existential reading bare plurals are not indefinite NP's. (3) The generic reading of the bare plural is not due to a generic operator. If they are indeed correct, (1} and (2) jointly support a unified analysis of the existential and the generic reading of bare plurals, and (3) an inferred generalization approach to generics. In sections 3.2.1, 3.2.2, and 3.2.3 I discuss the evidence which led Carlson to these claims. I have tried to bring an updated perspective with respect to the arguments supporting (2) and (3}, so I present them in a way that does not always follow Carlson's organization or characterization. I will argue that a critical reexamination of the evidence leads to different conclusions from the ones Carlson drew. In some cases, the evidence is compatible either with the assumption that the bare plural is kind-denoting or with the assumption that it is

18

Descriptions in Context

an indefinite description. In other cases, the evidence in fact favors the assumption of the indefiniteness of the bare plural and of the presence of an operator in generics. 3.2.1

Bare Plurals are Unambiguous

Let us take for granted that some predicates can denote either episodic or characteristic properties. If, in addition, bare plurals were ambiguous between an existential and a generic reading, then there would, in principle, be four possible readings resulting from the combination of a bare plural with such a predicate, as schematized in [10]. Correspondingly, for those predicates that can denote only characteristic properties {basically individual-level predicates) there would in principle be two possibilities, those in [lOa] and [toe]. [10] a. generic NP- characteristic property b. existential NP - episodic property c. existential NP - characteristic property d. generic NP- episodic property Of these, Carlson points out, only [lOa) and [lOb) actually occur. Consider the sentences in [11]. [11] a. Dinosaurs ate kelp. b. Dinosaurs were intelligent. [lla] has a reading corresponding to (lOa], according to which all (at least all unexceptional) dinosaurs had the characteristic property of eating kelp, and a reading corresponding to (lOb], according to which some dinosaurs ate kelp on one particular occasion. The readings corresponding to [lOc] and [lOd] are lacking: there is no reading according to which some dinosaurs had the characteristic property of eating kelp, nor is there a reading in which there was an instance of kelp eating by all (unexceptional) dinosaurs. 24 Similarly, [llb] has only a reading corresponding to [lOa], according to which all (unexceptional} dinosaurs had the characteristic property of being intelligent. There is no reading corresponding to [toe], according to which there were some dinosaurs that were intelligent. As we have seen, the reading of [lla) corresponding to [lOa] is the result of a derived mediated predication, of the type specified in (7ciii], and the reading corresponding to [lOb] is the result of a derived unmediated predication, of the type specified in [7cii]. Given the typology in [7] and taking eat kelp to be a basically stage-level predicate, there are no other possibilities. The reading of [llb] corresponding to [lOa] is the result of a derived mediated

Bare Plurals and Genericity

19

predication, of the type specified in [7bii). Given (7) and the fact that be intelligent is a basically individual-level predicate, no other possibilities exist. While Carlson is correct in claiming that the readings of type [toe) and [10d) are missing, another generalization that he makes, namely that a bare plural subject cannot have an existential reading in a generic sentence, is too restrictive. Cases exemplified by [12) (originally discussed in Milsark (1974)) constitute a serious problem for his analysis, which predicts that the only reading available for [12a) is one equivalent to [12c). In fact [12b) is also a possible reading, and the only plausibly true one in this case. 25 [12)

a. Typhoons arise in this part of the Pacific. b. In this part of the Pacific there arise typhoons. c. In general, typhoons arise in this part of the Pacific.

Notice, moreover, that [12a) (with the existential reading for the bare plural) and [12b) are interpreted as generic while they contain no NP in them that is kind-denoting, which can then give rise to a kind-level predication. 26 Given the overall architecture of the theory there is no way of accommodating such cases since the predicate can either be derived via a generalization operator, thus giving a generic reading to the bare plural, or have existential quantification over stages, giving an existential reading to the bare plural and an episodic reading for the whole sentence. 3.2.2

Bare Plurals and Indefinite NP's

In support of the claim that bare plurals are not indefinites, even on their existential reading, Carlson brings three types of arguments: (a) scopal restrictions, (b) anaphoric properties, (c) similarity with definite, kind-denoting NP's. a. Scopal Restrictions

When interpreted existentially, bare plurals always have narrow scope with respect to sentential operators (e.g., negation), count nominal quantifiers, attitudinal predicates and partitive quantifiers. For example, while the singular indefinite can have both narrow and wide scope with respect to negation, thus allowing for a non-contradictory reading of [13a], the bare plural has only narrow scope with respect to negation, allowing only for a contradictory reading of [13b). [13]

a. A cat attacked John and a cat did not attack John.

20

Descriptions in Context

b. Cats attacked John and cats did not attack John. Since the existential reading for bare plurals arises as a result of existential quantification over stages introduced within the translation of the VP predicate, the analysis allows only for translations in which the existential quantification has narrow scope with respect to every other operator. If the bare plural were analyzed as an indefinite, Carlson's argument goes, then the existential quantifier associated with it would scope freely with respect to other operators resulting in more readings than are actually attested. The facts about scope have since been widely discussed and sometimes disputed so I will not devote too much attention to them here (see Kratzer (1980), Link (1984), Wilkinson (1988a), Rooth (1995)). The interaction between bare plurals and partitive quantifiers (labelled differentiated scope phenomenon by Carlson) is well-known from studies on aspect. The facts pointing to a difference between bare plurals and indefinites NP's in terms of their aspectual properties, while robust, cannot be taken as a definitive argument against the bare plural's being indefinite since they can be accounted for in a different manner than assuming existential quantification introduced by the predicate over stage realizations of a kind. See, e.g., Krifka (1989), Verkuyl (1989), Moltmann (1991). Similarly, a more in-depth analysis of the other scopal restrictions might reveal that they are a by-product of some other property of bare plurals and are therefore consistent with the position that bare plurals are indefinite NP's. b. Anaphora

The argument from anaphora can be summarized as follows: bare plurals are not indefinite NP's because pronouns anaphoric on them have a wider range of readings than pronouns anaphoric on indefinites. In what follows I look more closely at Carlson's analysis of anaphora with the aim of showing that not assuming that bare plurals are indefinite at least on their existential reading misses certain important generalizations. Pronouns outside the syntactic (c-command) scope of their indefinite antecedents are interpreted as E-type, that is as synonymous with a definite description: consider the synonymy of [14a] and [14b]. Carlson notes that pronouns with bare plural antecedents have an additional interpretation. They can be interpreted as E-type, as in [14c], but they can also be interpreted as pronouns of laziness standing for a description identical to that of

Bare Plurals and Genericity

21

their bare plural antece~ent. 27 This is illustrated by [14d], which has a reading corresponding to that of [14e] (in other words, Mary need not have sold the strawberries I grew). In contrast, [14f], in which the pronoun has an indefinite antecedent, does not have such a reading. [14) a. I bought some strawberries yesterday. Mary washed them. b. I bought some strawberries yesterday. Mary washed the strawberries I bought yesterday. c. I bought strawberries yesterday. Mary washed them. d. I grew strawberries. Mary sold them. e. I grew strawberries. Mary sold strawberries. f. I grew some strawberries. Mary sold them. g. # Strawberries are fragrant. I bought them yesterday. Carlson assumes that pronouns in general receive two translations.28 The interpretation of the antecedent ultimately influences the range of readings for the pronoun. He analyzes E-type pronouns along the lines of Cooper (1979) as implicit Russellian definite descriptions. Their translation, AP (3xi ('Vyi CQ(z 8 ))(yi) +-+ xi=yi)) & ~P(xi)), involves two free variables, which must be given a value of the appropriate type by the context of use. The relation variable Q can pick as its value the realization relation R that relates stages to individuals when that is contextually available; the stage-level variable Z 8 can then be assigned as its value the contextually available stage-level entity that is an argument of R. For example, the pronoun in [14c) picks out those individuals that are strawberries and have a stage-level realization that was bought by me yesterday. The existence of such individuals is guaranteed by the first sentence of [14c] since for that sentence to be true there must be stages of the kind strawberries that were bought by me yesterday. The meaning postulate described in [3d) guarantees that these are also stages of an object realizing the same kind. The second sentence of [14c) then asserts of those individuals that they also have a stage-level realization that was washed by Mary. A pronoun may also translate as the set of properties associated with some individual entity: AP~P(xi). It is this translation that results in a pronoun of laziness interpretation if the antecedent is a bare plural. The context must make available a unique individuallevel entity which can be given as the value of the free variable xi. In [14d), where the first sentence contains a kind-denoting NP, the pronoun appears in a context which makes available a unique

22

Descriptions in Context

individual-level entity. The free variable in the translation of the pronoun is assigned as value the kind-level entity denoted by the bare plural strawberries. Although a pronoun always has the option of translating as the set of properties associated with some individual entity, whether a true pronoun of laziness reading will arise depends on whether its antecedent is kind-denoting or object-denoting. If the antecedent is object-denoting, either translation for the pronoun would amount to the same reading. For example, in the context of the first sentence of (14f], where the NP introduces existential quantification over objects, the free variable in the translation of the pronoun can be given as value the object-level entity which makes the first sentence true. 29 But any such entity must satisfy both the condition of being a strawberry and of having stages that were grown by me. So, in effect, the pronoun is interpreted as if it were Etype (modulo the origin of existence and uniqueness). 30 In other words, having two different translations for the pronoun does not lead to two distinct interpretations except when the antecedent is kind-denoting. However, the two interpretations are not always available for a pronoun anaphoric on a bare plural. Carlson observes that such a pronoun can be E-type only if its antecedent is in an episodic context. The pronoun in [14g], whose antecedent occurs with an individual-level predicate, lacks an E-type reading (i.e., the reading according to which I bought the strawberries that are fragrant). This reading is excluded by Carlson on the grounds that the first sentence does not introduce existential quantification over stages and hence there is no contextually available value for the free variables in the translation of the pronoun. 31 However, why should the translation of the pronoun be specified in such a way as to contain a two-place relation variable or a free variable sorted for stages? There is no independent evidence for specifying the adicity of the contextually recoverable property or sortally restricting its arguments. Is there a way of excluding the relevant reading without this stipulation? Suppose we follow Cooper and take E-type pronouns to translate as .XP (3xi (Vyi (-Q)(yi) +-+ xi=yi) & -p(xi)), with Q being a property denoting expression that may contain only free variables (of any sort) and parentheses. Then for the pronoun in (14g] we can reconstruct the free property variable as the property of being fragrant and assign to Q the value Gn'(..:. fragrant'). Since the first sentence of [14g] guarantees the existence of an individual-level

Bare Plurals and Genericity

23

entity that is fragrant, namely the kind strawberries, the second sentence would assert that I bought stages of the unique kind that is the kind strawberries and which is fragrant. Nothing will require that the objects whose stages I bought had to be fragrant. 32 In this case, therefore, the E-type interpretation and the pronoun of laziness interpretation would amount to the same thing. However, when object-level entities are implicated in the truthconditions of the sentence containing the antecedent, this approach predicts that the pronoun will have a distinct E-type interpretation. Sentences with adverbs of quantification introduce quantification over object realizations of a kind, as we will see in section 3.3, and are therefore predicted to give rise to such a reading for subsequent pronouns. [15a] exemplifies such a case: the first sentence of [15a] would translate as in [15b] and the second as in [15c], assuming we assign to Q as value the property of being fragrant. [15]

a. Strawberries are sometimes fragrant. I bought them yesterday. b. 3z 0 (R'(z 0 ,strawb) & fragrant'(z 0 )) c. 3x 0 ('Vy 0 (fragrant'(y 0 ) t-+ x 0 = y 0 )) & 3z 8 (R(z 8 ,X0 ) & bought(l,z 8 ))

But then the pronoun will be taken to be synonymous with the definite description 'the fragrant objects (that are strawberries)', contrary to fact. Therefore, unless a sortal restriction is put on the free entity variable, the analysis predicts more readings than are actually available. However, by putting sortal restrictions on the free variables in the translation of E-type pronouns, the analysis seems to stipulate what it should explain, namely, why the interpretation of the predicate that the bare plural antecedent is construed with affects the range of interpretations of the pronoun. In other words, the analysis fails to capture the generalization that a bare plural supports E-type anaphora only if it receives an existential reading, i.e., only when it is in an episodic context. Moreover, the reverse generalization appears to hold as well: if both the bare plural antecedent and the pronoun are in an episodic context, then the pronoun must be interpreted as E-type. 33 Gerstner and Krifka (1987) have noted the deviance of examples like [16], which can be explained on pragmatic grounds assuming the E-type interpretation is the only possible one for the pronoun. 34 [16]

# John ate applesi and Mary ate themi too.

Now, Carlson's analysis can exclude the E-type reading on ,pragmatic grounds as well (since it would imply that John and Mary

24

Descriptions in Context

ended up eating the same apples) but has no way of ruling out the pronoun of laziness interpetation. An assumption underlying Carlson's argument from anaphora is that a pronoun may have a kind or generic reading only because its antecedent is kind-denoting. But as various authors (e.g., van Eijck 1983, Wilkinson 1991) have demonstrated, a wide variety of NP's give rise to this kind of anaphora; the best way to analyze it is to treat the pronoun as a pronoun of laziness whose descriptive content is determined on the basis of the CN of its antecedent. The factors determining when such a reading for the pronoun could arise (since it is not always available) is predicted to be uniform for the various types of NP's. 35 Such an approach is of course entirely compatible with treating the bare plural as an indefinite NP, at least on its existential reading. A treatment of the bare plural as an indefinite would then account for the possibility of Etype anaphora with bare plurals. On the other hand, a bare plural involved in a proper kind predication would directly support kind anaphora so we would expect such anaphora to be freer than the one involving a pronoun of laziness interpretation. 36 c. Definite Kind-Denoting NP's

Definite NP's referring to kinds, like this kind of wolf, exhibit a parallelism with bare plurals in episodic contexts. Although they have an existential reading (see [17a]), we would not want to analyze them as indefinite NP's. In addition, they have the same properties with respect to scope as bare plurals (see [17b], [17c]). [17]

a. I saw this kind of wolf at the zoo. b. # I saw this kind of wolf at the zoo and I did not see this kind of wolf at the zoo. c. John believes that this kind of wolf will appear in his garden.

Carlson concludes that what such NP's share with bare plurals is precisely reference to kinds and takes the parallelism between these two types of superficially dissimilar NP's as further evidence that bare plurals are, despite appearances, disguised definite descriptions of a special sort. However, one can cast doubt on the claim that definite kindreferring NP's really have an existential reading with episodic predicates. Although [17a] implies that I saw a wolf in the zoo that was a wolf of this kind, it is not beyond doubt that this is in fact part of its truth-conditional meaning. I will not offer an analysis of

Bare Plurals and Genericity

25

definite kind-referring NP's here but I will try to establish that, unlike bare plurals, they do not have an existential reading and that, therefore, an episodic predicate does not introduce existential quantification over stages. More generally, I will argue that definite kind-referring NP's do not behave (even superficially) in ways that exactly parallel bare plurals. Because the implication of existence of an individual that is a realization of a kind is so strong, it is easy to be convinced that the bare plural and the kind-referring NP receive the same interpretation in episodic contexts (except, of course, for the fact that an NP like this kind of wolf refers to a subkind of the kind wolf), which according to Carlson's analysis comes about because the predicate introduces quantification over stages of the kind. Nevertheless, there is some subtle evidence that can distinguish between the existential reading of the bare plural and that of a definite kindreferring NP. The former is part of the truth-conditional content of the sentence, the latter is only an implication. 37 Consider [18). [18]

a. I got rid of weeds from the garden yesterday. b. I got rid of this kind of weed from the garden yesterday.

[18a) and [18b) are true under somewhat different circumstances: if I got rid of some weeds but left some still standing, [18a] would be true while [18b] would be false. [18b] requires that I get rid of every single specimen of this kind of weed in the garden, a requirement linked to the fact that get rid of is interpreted as a basically kindlevel predicate with respect to its second argument in [18b]. So the implication associated with the NP this kind of N as to how many realizations of the kind have the relevant property can vary, depending on the lexical semantics of the predicate. This can also be taken as indication that the existential reading of the bare plural in [18a] is not due to the predicate but to the NP itself. A pronoun anaphoric on an NP like this kind of N behaves like a pronoun anaphoric on a definite NP and unlike a pronoun anaphoric on a bare plural. In contrast to [16], [19a] is perfectly acceptable. Moreover, if the episodic predicate which the NP this kind of N is an argument of introduced existential quantification over stages, an E-type interpretation for a pronoun anaphoric on that NP would be possible. For example, the pronoun in [19b] would be synonymous with the definite description the raccoon that I saw yesterday that is a realization of this kind of raccoon and the continuation with the although-clause would not be contradictory. This, however, is not the case: the pronoun is synonymous with the

Descriptions in Context

26

description this kind of raccoon, hence the contradiction of the continuation. In (19c], on the other hand, the pronoun is interpreted as E-type and no contradiction arises. 38 (19]

a. John ate this kind of apple and Mary ate it too. b. I saw this kind of raccoon in the forest. It was magnificent (# although this kind of raccoon is (generally) repulsive). c. I saw raccoons in the forest. They were magnificent (although raccoons are (generally) repulsive).

This perspective on the apparent existential reading of definite kind-referring NP's in episodic contexts is compatible with the scopal properties of such NP's. The contradictory reading of [17b] is the same as that of any definite in the place of the kind-referring NP and we need not appeal to narrow scope existential quantification to account for it. As for [17c], if there is no existential quantification over realizations of this kind of wolf in the semantics, then no existential quantifier needs to scope with respect to the attitude predicate. John may have a de re belief about the kind itself without having a belief about any particular realization of the kind. 3.2.3

A Generic Operator and Bare Plurals

Carlson offers a reductio argument that bare plurals should not be treated like other quantificational NP's. He explores the consequences of assuming that there is a generic nominal quantifier and then rejects such an assumption for the reasons discussed in (a)-(g) below. A theme common to most of them is that such a quantifier would not resemble any known nominal quantifier in certain significant respects. From this he infers that a direct generalization approach to generics is untenable. In presenting his arguments, my aim is to establish that although he is right on the first point, the second conclusion does not necessarily follow. As we will see in section 4, it is possible to analyze generics as involving a generic quantifier which is not nominal. Moreover, we can construe some of Carlson's evidence as showing that an operator is present. a. Exceptions

Although it would appear at first sight that the generic quantifier has the force of a universal quantifier, unlike all other universal quantifiers, it tolerates exceptions, which at times might even

Bare Plurals and Genericity

21

outnumber the non-exceptional cases. The intuition about exceptional cases is of course that they are not normal in some relevant respect. 39 Carlson considers appending the restriction 'normal' to a universal quantifier but finds such a move inadequate. For instance, he argues that normality with respect to having a mane would require all normal lions to be male while normality with respect to feeding their young milk would require all normal lions to be female, hence we end up with a contradiction. Or, if all normal dogs are mammals, could an abnormal dog not be a mammal? His criticism against 'all normal,' however, does not take into account the context-dependency of 'normal'; the argument would have force only if the criteria of normality were identical for all properties. There are no universal criteria for normality but this does not mean that there are no criteria for normality relative to a given property. b. No Fixed Quantificational Force

Not only does the generic quantifier not have universal force, it does not even seem to have a stable quantificational force. Its varied force depends, among other things, on the predicate with which the bare plural is construed since that seems to be responsible to a large extent for determining the criteria of normality, and on external circumstances that seem to vary from case to case. So although few shoplifters actually get prosecuted and few alligators survive long enough to attain their full length, [20a] and [20b] are true, while [20c] is false even though most books are actually paperbacks. [20] a. Shoplifters are prosecuted in criminal court. b. Alligators grow twenty feet long. c. Books are paperbacks. Furthermore, even if we claimed ambiguity and allowed its force to vary, the generic quantifier could not be identified with any of the known nominal quantifiers since the truth conditions of generic sentences and that with a nominal quantifier are different. For example, while [21b] and [21d] are true, [21a] and [21c] are not. [21]

a. b. c. d.

Seeds do not germinate. Most seeds do not (in fact) germinate. Crocodiles die before they get two weeks old. Most crocodiles (actually) die before they get two weeks old.

Descriptions in Context

28

This kind of argument, however, is valid only if we try to link the truth of a generic statement to the way things actually are, that is if the generic quantifier is taken to quantify over actual entities. In other words, Carlson has successfully argued against using an extensional operator to account for genericity but this falls short of showing that no operator can be used to account for the semantics of genericity. Carlson takes (a) and (b) to be strong and decisive arguments against a quantificational analysis of generics in general. On the basis of them, he concludes that a quantificational analysis would be "profoundly misdirected" (Carlson 1982:167). However, the real import of the argument is that we should not build quantification over instantiations into the semantics because inferences we get about actual instantiations of the generalization vary widely, depending both on the content of the generalization expressed and other background assumptions we bring to bear. In any case, the kind-denoting analysis that Carlson advocates solves this problem only by allowing itself to take no stance on what inferences we can draw from the characteristic property of a kind about its individual instantiations. 40 The analysis not only misses something significant in not connecting in any way the truth of a derived kind predication to the truth of predications about individual instantiations but it does not even give a general answer to what Carlson took to be the right question to ask about generics, namely 'How can we infer quantification from generic sentences?' (Carlson 1977b:109). c. The Port-Royal Puzzle

The generic quantifier, unlike other nominal quantifiers, is not right monotone. This is known as the Port-Royal Puzzle and is exemplified by [22]: [22a] does not entail [22b], whereas [22c] entails [22d]. [22]

a. b. c. d.

Dutchmen are good sailors. Dutchmen are sailors. Every Dutchman is a good sailor. Every Dutchman is a sailor.

Interestingly, Carlson's ultimate analysis of the contrast in [22] does not rely on the monotonicity properties of universally quantifies NP's vs. those of bare plurals. (In a sense, it removes the preconditions for testing monotonicity for bare plurals by relying on the presence of the Gn' operator prefixed to the predi-

Bare Plurals and Genericity

29

cate Agood'Csailor').) See Carlson (1977b:295ff) and Wilkinson (1991:17ff) for discussion. d. Interaction with Adverbs of Quantification If a generic quantifier were associated with the bare plural, then

like other nominal quantifiers, it would force a temporal reading for an adverb of quantification. This, however, is not the case; unlike [23a], [23b] has a felicitous reading, according to which many Texans are tall. 41 [23]

a. # Every Texan is often tall. b. Texans are often tall.

Points (c) and (d) are the decisive arguments against a nominal generic operator but they are perfectly consistent with, if not providing evidence for, an adverbial generic operator. Research on natural language quantification has shown that quantifiers of natural language belong to two main categories: nominal quantifiers, associated with an NP, and so-called adverbial quantifiers, which are not associated with an NP and whose paradigmatic cases are adverbs of quantification. An implicit quantifier is considered adverbial if it exhibits the semantic properties of overt adverbial quantifiers. e. Kind-Level Predicates

Bare plurals can co-occur with kind-level predicates, like extinct, which cannot take as arguments expressions denoting ordinary individuals. At most, however, this shows that not all forms of genericity can be analyzed as involving quantification, not that none can. f. Scopal Restrictions

Generically interpreted bare plurals exhibit scope restrictions, a fact which can be accounted for if bare plurals are analyzed as rigidly denoting kinds rather than as quantified NP's. The semantics of rigid designation is such that rigid designators always exhibit wide scope with respect to any other operator. Specifically, bare plurals have wide scope with respect to negation and narrow scope in attitudinal contexts. 42 For example, [24a] expresses the (false) generalization that all (normal) pomegranates have the characteristic property of not having a crowned end, rather than the weaker generalization that pomegranates do not have the characteristic property of having a crowned end. 43

30

[24]

Descriptions in Context

a. Pomegranates do not have a crowned end. b. Bishops should not move diagonally. c. Pomegranates do not normally /typically have a crowned end.

However, as seen in [24b] and [24c], both necessity modals and overt generic operators like normally or typically seem to favor wide scope with respect to negation too, thus favoring the stronger generalization reading even when this leads to falsity. 44 Therefore, the evidence from examples like [24a] could lead either to the conclusion that bare plurals behave like names, or, alternatively, that the implicit generic operator behaves like modals and other overt generic operators in taking wide scope with respect to negation. 45 With respect to attitudinal contexts the issue is more complicated since the argument based on the absence of a wide scope reading for the bare plural is tied to an extensional version of the hypothesized generic operator. For an example like [25a], Carlson takes wide scope to give rise to the following reading, which he claims is missing: for each entity that is actually a pomegranate John believes of it that it is poisonous. This interpretation is consistent with John's lacking the belief that all pomegranates are poisonous. [25] John believes that pomegranates are poisonous. The kind analysis predicts that John has a belief about the kind itself without having a belief that all pomegranates are poisonous or a belief about actual instantiations of the kind. The generalization operator must indeed have narrow scope with respect to the attitude predicate. In general, adverbial operators do not take scope beyond their clause. Therefore, the property of exhibiting narrow scope with respect to an attitude predicate is shared by modals ([26a]), overt generic operators ([26b]), and adverbs of quantification both on their generic ([26c]) and their temporal reading ([26d]). [26]

a. John believes that bishops (should) move non-diagonally. b. John believes that pomegranates normally /typically have a crowned end. c. John believes that pomegranates are always poisonous. d. John believes that during last winter Mary always went for a walk at night.

Bare Plurals and Genericity

91

Since only nominal operators within the complement of an attitude predicate can take scope over the attitude predicate, these facts simply establish that the generic operator is not nominal but are perfectly consistent with its being an adverbial operator. g. Definite Kind-Denoting NP's

Definite NP's referring to kinds, like this kind of animal, exhibit a generic reading and yet we would not want to analyze them as involving a generic quantifier. This, however, is not inconsistent with having a generic operator and a kind-denoting NP without the former being contributed by the latter.

9. 3 Limits and Limitations of Uniformity One of the most compelling arguments of Carlson's in favor of the kind analysis is that kinds provide the intensionality necessary for the semantics of genericity without the need to posit any quantification. However, an across-the-board unified analysis of generics as involving no quantification is untenable. Carlson himself has to allow for a "mixed" analysis in which some quantification is built into the semantics. This is necessary for generic sentences with adverbs of quantification. In Carlson's {1979) analysis, (atemporal) adverbs of quantification are VP operators applying to object-level predicates and yielding kind-level predicates. In that respect they are like the generic operator Gn1 • They differ, however, from Gn 1 in that they quantify over object-level realizations of a kind. I will now show that the same kinds of arguments Carlson advances against a generic quantifier, summarized in (a) and (b) in section 3.2.3, can be made with respect to his analysis that postulates a quantifier associated with adverbs of quantification. 46 Let us take the adverb of quantification always, which, treated as a VP operator mapping object-level predicates to kind-level predicates, translates as in [27). [27) ).P).xk(Vyo(RI(yo,xk)----+ -p(yo))) Although it is incontestable that the adverb has universal force, one could object to positing a universal quantifier in ways that parallel Carlson's arguments against a quantificational analysis of generics by showing that it leads to inadequate predictions about what the domain of quantification and the force of the quantifier are. [28a), for example, translating as [29a), comes out as too weak under this analysis since in order for it to be true in the actual world it suffices

32

Descriptions in Context

that all actual dogs be intelligent. But, intuitively, [28a] expresses something stronger. [28b), on the other hand, translating as [29b], comes out as too strong since it would require of male ducks that are well-bred to lay eggs, whereas, intuitively, [28b] is judged to be true without that strong requirement holding. [28]

a. Dogs are always intelligent. b. Ducks that are well-bred always lay eggs.

[29]

a. 'v'y 0 (R'(y 0 ,d)-+ intelligent'(y 0 )) b. 'v'y 0 (R'(y 0 , wbd)-+ Gn(Alay-eggs'(y 0 )))

In other words, once some other operator than the implicit one is introduced, we can avoid neither the modal aspect of genericity nor the requirement for normality. Carlson, on the other hand, is committed to extensionality once an adverb of quantification is present. Neither can we avoid readjustments in the domain of quantification, resulting in narrowing of the domain, so as to satisfy certain conditions associated with the predicate. Again, Carlson's analysis makes wrong predictions in this respect. For example, [30] comes out as false in Carlson's analysis, given that most shoplifters are never caught, let alone prosecuted. [30] Shoplifters are always prosecuted in criminal court. To sum up, Carlson's analysis of adverbs of quantification suffers precisely from the problems he identified with positing an operator for generic sentences. The reason for that is not that the quantificational force of the operator cannot be specified with any precision-adverbs of quantification wear their force on their sleeve-but rather because of the assumed extensionality of the operator. Here the intensionality of individuals is not helpful precisely because the truth conditions make no reference to alternative points of evaluation. If one were to revise this analysis so as to make the adverb of quantification intensional, the need for making the bare plural kind-denoting in these cases would be diminished. Moreover, once we recognize the need for intensionality in the evaluation of sentences with adverbs of quantification, it takes only a small step to make a similar assumption about generic sentences without an overt operator.

Bare Plurals and Genericity

3.4

33

Summary

Carlson concludes that all genericity should be given a uniform analysis, that bare plurals are neither indefinite nor ambiguous, and that generics do not involve a generic quantifier. As we saw, the last point confounds three distinct claims: (i) there is no nominal generic operator associated with bare plurals, (ii) no extensional semantics would do for generics, (iii) there is no quantification over objects at different points of evaluation in the specification of the truth-conditions for generics. (i) and (ii) are undoubtedly true but (iii) is not. In fact, none of the evidence Carlson provided against a direct generalization approach to generics was really evidence in favor of (iii). Some of the crucial facts leading Carlson to these conclusions about genericity and bare plurals can be dealt with within a quantificational approach to generics and a more sophisticated framework for the treatment of indefinites, as we will see in the next section. In contrast to Carlson's claims summarized by (1)-(3) in section 3.2, this approach is committed to (1)'-(3)': (1)' Ambiguity resides neither in the predicate nor in the NP; the apparent ambiguity results from the presence or absence of a sentential operator and from the interpretation of indefinites within and outside quantificational structures. (2)' Bare plurals are indefinite both on their existential and generic reading. (3)' An operator is present in the generic reading of bare plurals but it is not associated with the NP.

4. THE INDEFINITENESS ANALYSIS

4.1

Basic Outline

The indefiniteness analysis brings together the conception of genericity as involving direct generalization and a semantics for bare plurals in which they are treated as ordinary indefinites. The analysis of bare plurals as indefinite NP's proposed by Krifka (1987), Gerstner & Krifka (1987), and Wilkinson(1988a, 1991) relies on' the treatment of adverbial operators; and indefinites in Lewis (1975), Kamp (1981) and Heim (1982) and claims that one type of genericity is to be attributed to the presence of an appropriate operator. 47

34

Descriptions in Context

The analysis assumes t'hat there are two sources of genericity, one arising from the presence of a sentential dyadic intensional operator, the other from the presence of a kind-denoting term and a kind-level predicate. In that sense it partitions genericity in such a way that one type is amenable to a direct generalization approach and the other type to an inferred generalization approach. Krifka's distinction between I-genericity and D-genericity is precisely along this dimension; moreover, it connects 1-genericity to indefiniteness and D-genericity to definiteness. 48 Bare plurals are claimed to be ambiguous; on one interpretation they denote kinds, and hence can appear within D-generics, on another interpretation they are ordinary indefinites, and hence can appear within !-generics. The singular indefinite is only an ordinary indefinite; hence it can appear only within !-generics. The singular definite generic, on the other hand, is unambiguously kinddenoting; hence it can appear only within D-generics. Thus examples [1a], [1b] and [2a], repeated in [31], involve D-genericity, while [lc], [ld], [le], [2b] and [2c], repeated in [32], involve 1-genericity. [31]

a. Punica granatum is an Old World tree. b. The 3-R4* generation robot will soon be antiquated. c. Dinosaurs are extinct.

[32]

a. b. c. d. e.

A pirate ship flies a black flag. Pomegranates have a crowned end. In this place one whistles a lullaby to show approval. Dogs are mammals. Dogs bark.

From here on I will concentrate on 1-genericity. This analysis of bare plurals and 1-genericity makes three basic claims: (i) 1-genericity involves a sentential dyadic intensional operator; with an interpretation similar to that of adverbs of quantification, (ii) bare plurals are indefinites, (iii) indefinites are nonquantificational, variable contributing elements. If the bare plural is an indefinite, then its apparent ambiguity between an existential and a generic reading can be seen as a special case of the apparent variable quantificational force of indefinites. Its generic reading is a quantificational reading which comes about when the bare plural is in the restriction of an appropriate operator, such as an adverb of quantification, a modal or an implicit generic operator. When not in the scope of an operator, the bare plural is caught by existential closure; hence its existential reading in episodic sentences. The bifurcation in the readings of bare plurals is, therefore, not

Bare Plurals and Genericity

35

a property unique to them but one shared by all indefinite NP's. In this respect, the analysis is in accord with Carlson's conclusion that there is no genuine ambiguity in the semantics of the NP: if we exclude the true kind reading of D-genericity, the variability in the readings of bare plurals is not due to an ambiguity per se but is traced to a more basic property of the bare plural. In this case the relevant property is indefiniteness, which interacts with the interpretation of operators and free variables to yield the desired range of interpretations. The logical form of a generic sentence has the schematic representation in [33]. (33] G (ljl, 1/J) If an indefinite is within the restriction ljJ of the generic operator, it will be bound by the operator and will have a generic reading; 49 if it is in the nuclear scope 1/J, it will get an existential reading. A bare plural that receives an existential reading in a generic sentence, as in (12], is an instance of the latter case. Whether an indefinite is mapped in the restriction or nuclear scope of the generic operator is not fully determined on the basis of surface syntactic structure, as is clearly demonstrated by [12]. This problem has come to be known as the issue of semantic partition (Diesing 1990, 1992a) and is an active issue of current research. In general, the indefiniteness analysis predicts that a given sentence may have several generic readings as a result of different partitionings of its material between the restriction and the nuclear scope. Several factors seem to play a role in determining semantic partition, most notably syntactic structure (Diesing 1990, 1992a, 1992b, Kratzer 1995) and focus (Rooth 1985, 1995, Krifka 1992). The generic reading of singular and plural indefinites arises when the variable of the indefinites is bound by a generic operator. The plural indefinite and the singular indefinite are thus analyzed in the same way, as contributing an open formula. The open formula corresponding to the indefinites in [34a] and [34b] appears in the restriction of the implicit generic operator G and its free variable is bound by G, as in [35a]. In (34c] and (34d], on the other hand, there is no operator and the variable in the open formula corresponding to the indefinites is subject to existential closure, as in (35b]. 50 (34]

a. Whales are mammals. b. A whale is a mammal. c. Whales are roaming the coast.

36

Descriptions in Context

d. A whale is roaming the coast. (35]

a. Gx (whale(x), mammal(x)) b. 3x (whale(x) & roam-the-coast(x))

[35a] is interpreted according to [36], where the generic operator is assumed to have universal force. How this can be reconciled with the tolerance to exceptions will be discussed in section 4.3, where the intensionality of G will be taken into account. 5 1 [36] [Gx(whale(x),mammal(x))]~ c w ' '

=1

iff for every g' ~ g

such that [whale(x)]1£ c w = 1, [mammal(x)]~ c w = 1. A quantificational analysis of generics is consistent with incorporating only a subset of (i)-(iii) above, or with modified formulations of any one of (i)-(iii). This is for two different reasons. One is that an analysis of bare plurals as uniformly kind-denoting can be combined with a direct generalization approach to generics. The other is that the apparent variable quantificational force of indefinites can be captured within a different set of assumptions, which associate an existential quantifier with an indefinite but have an appropriately revised semantics for the existential quantifier and adverbial operators. Farkas (1985) and Farkas & Sugioka (1983), for instance, operated only with a modified version of (i); they assumed an implicit sentential operator binding variables corresponding to stage or object realizations of the kind denoted by a bare plural. An analysis incorporating (i) and (iii) but treating bare plurals as kind denoting terms is proposed in Rooth (1995), where it is the object realizations of kinds that behave like indefinites. 52 The advantage of preserving a uniform kind denotation for bare plurals is that it allows the same bare plural to support both kind-level and object-level anaphora. The narrow scope properties of bare plurals, however, do not follow anymore once the existential quantification is not part of the semantics of stage-level predicates. Finally, recent analyses analyzing indefinites as dynamic existential quantifiers and adverbs of quantification as generalized quantifiers, such as Chierchia (1992) and de Swart (1996), applied to generics, can be seen as modern incarnations of the (i)-(ii) analysis. 5 3 [37] gives an overview of the similarities and differences between the different analyses with respect to three significant features. FV-Indef. stands for the analyses taking indefinites to be quantifier-free, DynIndef. for the analyses taking indefinites to be analyzed in terms of a dynamic existential quantifier. I

',

I

,

t

37

Bare Plurals and Genericity

[37]

Carlson FV-Indef. Farkas Rooth Dyn-Indef.

Uniform denotation Uniform analysis Direct Generalization of bare plurals for of bare plurals in episodic/generic kind/generic contexts readings

-

+ + + +

-

+ +

-

+ +

+

+ +

-

4.2 Individual-Level and Stage-Level Predicates Once existential quantification need not be built into the predicate and once the bare plural can be taken to denote ordinary individuals, we get a simplification in the domain of predicates, for we no longer need derived kind-level or derived object-level predicates. 54 Nevertheless, although stages as ontological entities can be dispensed with, we still need to maintain the distinction between stage-level and individual-level predicates since a number of linguistic phenomena are sensitive to it. The two types of predicates have different distributional patterns and influence the interpretation of a variety of linguistic elements. Specifically, stage-level predicates, in contrast to individual-level predicates, permit an existential reading for bare plurals, are acceptable as post-nominal predicates in there-insertion sentences (Milsark 1974, Carlson 1977b), allow for a conditional reading as adjuncts and absolutes (Stump 1985) and are acceptable in conditional clauses without necessitating the presence of an indefinite NP (Kratzer 1995). How the distinction is to be theoretically construed, however, remains an open question. There have been several recent proposals but their degree of overall success remains to be ascertained. For example, Kratzer (1995) proposes that the difference between individual-level and stage-level predicates is in their lexical argument structure: stage-level predicates have an extra Davidsonian argument in their argument structure while individual-level predicates do not. Diesing (1990, 1992a) proposes that the difference between individual-level and stage-level predicates is in the syntactic structure they project to: individual-level predicates project to control structures, stage-level predicates to raising structures. Some discussion of these proposals, to the extent that they are relevant to the material presented there, is to be found in the next chapter.

Descriptions in Context

38

Assuming we adopt a basic bipartite division of predicates, the correspondence between Carlson's classification and that of the indefiniteness analysis is as follows: basically stage-level and individual-level predicates correspond to stage-level and individuallevel predicates, respectively. Derived kind-level predicates correspond to (non-kind-level) individual-level predicates. Derived object-level predicates correspond to stage-level predicates within a quantificational context: a generic sentence containing a stagelevel episodic predicate is taken to express a generalization over episodes. The question is whether this is always the case. An interesting issue arises with what Krifka (1988) calls 'unconditional generics,' exemplified by [38b), as opposed to conditional generics, exemplified by [38a). [38)

a. When John hikes in the woods, he always smokes. b. John smokes.

Is a generic operator involved in [38b)? 55 Krifka (1988) assumes that a generic operator is indeed involved in [38b] whose restriction is contextually recoverable. Kratzer (1988, 1995), on the other hand, distinguishes between derived genericity, brought about by an operator, and intrinsic genericity, residing in the predicate. In derived genericity there is quantification at the sentential level, in intrinsic genericity there is no quantification. The intuitive difference between derived genericity and intrinsic genericity is exemplified in [38], where we have a generalization over episodes in [38a], hence derived genericity, but intrinsic genericity in [38b]. According to Kratzer then, the predicate smoke is lexically ambiguous between a stage-level interpretation, as in [38a], and an individuallevel interpretation, as in [38b). All three approaches agree on the truth conditions of a sentence containing a basically individual-level predicate but interesting differences arise with respect to sentences with a basically stage-level predicate that express a generic or habitual generalization. Let us take [39] as our test case. [39]

a. John is a smoker. b. John always/normally smokes. c. John smokes.

[39a] contains a basically individual-level predicate and, under all three analyses, it would be true in a given world iff John is in the set of things that are smokers in that world. [39b] contains a derived stage-level predicate and all three analyses specify its truth conditions in terms of quantification: Carlson over stages of John, 56

39

Bare Plurals and Genericity

Krifka and Kratzer over situations. Carlson and Kratzer would give (39c] truth conditions equivalent to those of [39a], 57 while Krifka would give it truth conditions equivalent to those of [39b]. If the adverb of quantification is taken to have a modal dimension, considering non-actual situations will be crucial in specifying the meaning of [39b] and [39c]. [40] provides a schematic summary. [40] Carlson Kratzer Krifka

Quantification yes no b a,c b a, c b,c a

Modality yes no a, b,c b a, c b,c a

The discussion above shows that the issue of the (non-)uniformity of genericity manifests itself quite independently of bare plurals, in the domain of individual-level predicates.

4.3

The Generic Operator

Carlson argued that no known quantifier could be substituted for the generic quantifier and took that to be a strong argument against a quantificational analysis of generics. How can the quantificational analysis meet this challenge? Farkas {1985) and Farkas & Sugioka (1983) have claimed that the generic operator is adverbial and Dahl (1975), Heim (1982), Krifka (1988), and Krifka et al. {1995) that it, moreover, has a modal dimension. Coupling this with the analysis of indefinites as involving free variables, we can assume that the generic operator is a universal operator quantifying over pairs of worlds and assignment functions. 58 The restriction of a generic operator may be supplemented by extra conditions in addition to those contributed by the indefinite NP's mapped in the restrictor (Wilkinson 1988a, 1991, Krifka 1988, Krifka et al. (1995)). The tolerance to exceptions that generics exhibit has at least two sources: implicit restrictions in the domain of quantification and the modal dimension of the operator. Both assumptions are necessary and neither is sufficient without the other. 4.3.1

Implicit Domain Restrictions

If the generic operator is adverbial, then not only material from the subject NP but also material from the VP can enter the restriction of the operator. For example, the articulation of a sentence into

40

Descriptions in Context

a focus-background structure affects the mapping, with focussed material being mapped into the nuclear scope and material in the focus background being mapped into the restriction (Rooth 1985, 1995, Krifka 1992). Moreover, sortal restrictions of the predicate and presuppositions of material in the nuclear scope would be accommodated into the restrictor. Once we have this freedom, a lot of the "criteria of normality" can be incorporated as extra conditions in the restriction of the operator. For example, as a descriptive generalization, [41] is a generalization about caught shoplifters, not shoplifters at large, since being prosecuted has as an enabling condition having been caught. [41] Shoplifters are prosecuted in criminal court. That presuppositions of the predicate implicitly restrict the domain of quantification can be observed with adverbs of quantification, as in [42] from Schubert & Pelletier {1989). [42] A cat always lands on its feet. If the implicit conditions for the satisfaction of the predicate are not

taken into account, [42] would just be false. But these conditions clearly affect the evaluation of [42] and this is also why [42] implies that sometimes some cats fall, or get dropped, or find themselves in the air in some way. If the generic operator is adverbial, it would be right monotone only under certain conditions, i.e., if we keep their first argument constant (see de Swart (1991)). The Port-Royal Puzzle can be accounted for keeping this in mind and assuming that part of the VP predicate belongs in the restrictor. In [43a] quantification is over Dutchmen who are sailors, not over Dutchmen, hence [43a] does not entail [43b].59 [43]

a. b. c. d.

Dutchmen Dutchmen Dutchmen Dutchmen

are are are are

good sailors. sailors. always good sailors. always sailors.

Notice that a similar case obtains with overt adverbs of quantification; [43c] does not entail [43d]. In the move from [43a] to [43b] we have not kept the first argument constant, since in [43a] we quantify over Dutchmen who are sailors while in [43b] we quantify over Dutchmen. [43c] and [43d] are analogous in this regard. Implicit domain restrictions arise to satisfy presuppositional requirements of the restrictor. Another source for implicit domain restrictions, namely contextual restrictions, is not available

Bare Plurals and Genericity

41

for generic sentences. I will discuss this property in more detail in the next chapter. 4.3.2

Modal Dimension

Positing a modal dimension for the generic operator captures the fact that the existence of actual exceptions does not suffice to make a generic generalization false. The determination of the modal dimension is, as with overt modals, heavily context-dependent. As Kratzer (1977, 1981) argues, modals are unambiguous but contextdependent; they require the context of use to fix two parameters of their interpretation. 60 These two parameters are conversational backgrounds, assigning to each world a set of propositions, and determining in turn a modal base and an ordering source. The apparent multiple ambiguity that we can detect with generic adverbial quantifiers can thus be explained away if we take into account the context-dependency of the modal base and the ordering source. The question about generics that Carlson took to be misguided, 'What is the generic quantifier?,' is reformulated in this analysis as 'What is the modal base and the ordering source associated with the generic quantifier?' The generic operator can be interpreted analogously to Kratzer's (1981) human necessity operator. This and alternative interepretations for the generic operator are discussed by Krifka et al. (1995). For an alternative modal treatment see Asher & Morreau (1991, 1995). [44] gives the interpretation of a dyadic human necessity operator, concentrating on the modal aspect of the interpretation of the generic operator (i.e., and '1/J are assumed to contain no free variables to be bound by the operator or by existential closure over the nuclear scope). In the informal exposition below I will simplify and talk of the maximally normal worlds.

[44] [G(,'1/J)]L,c,w = 1, where c determines a modal base Rw and an ordering source $w iff for every w 1 E Rw such that (JL,c,w1 = 1, there is w2 E Rw such that W2 $w w1 and for every W3 E Rw such that W3 $w w2, ['I/JJL,c,w3 = 1. In view of these assumptions about the modal dimension of the generic operator, let us consider, informally, the generalizations expressed by [1c]-[1e], repeated in [45a]-[45c]. [45]

a. A pirate ship flies a black flag. b. Pomegranates have a crowned end. c. In this place one whistles a lullaby to show approval.

42

Descriptions in Context

[45b) expresses a descriptive generalization. The modal base is circumstantial and the ordering source stereotypical. In constructing the modal base we consider facts about the inherent properties of pomegranates, so the worlds in the modal base are worlds where the biological facts about pomegranates (e.g., their evolutionary history) are identical to those of the actual world. In constructing the ordering source we consider facts which determine an ideal where no mutations have come about and no other external intervention has altered the appearance of pomegranates. The generalization expressed by [45b) is that everything that is a pomegranate in the worlds in the modal base has a crowned end in all those most normal worlds determined by the ordering source. If there exist pomegranates whose crowned end has been chopped off, then the actual world is not among those closest to the ideal and therefore the actual generalization 'every actual pomegranate in existence {right now) has a crowned end' would not be entailed by [45b). [45a) expresses a descriptive generalization too. The modal base is constructed on the basis of facts having to do with the conventions about the operation of pirate ships and the ordering source on the basis of an ideal where these conventions are upheld and adhered to. [45c) expresses both a descriptive and a normative generalization. The modal base is constructed on the basis of facts having to do with the habits, social rules, or conventions of the people in this place. The ordering source may be stereotypical, whereby it is constructed on the basis of an ideal where people follow their habits, or deontic, whereby it is constructed on the basis of an ideal where people adhere to the social rules. Conventions and social rules induce certain habits and vice versa so that's why both readings appear to be equally prominent. However, this is not always the case so a descriptive generalization might be true without the corresponding normative generalization being true, as well, and vice versa, as discussed with respect to [9), repeated in [46). [46)

a. People around here shoot at each other indiscriminately. b. A country with rich natural resources shares them with its neighbors.

Let's assume that the area around here is such that it induces violent habits in people but that violent behavior is neither morally nor legally sanctioned. Then [46a) would be true if interpreted with respect to a circumstantial modal base taking into account

43

Bare Plurals and Genericity

facts about the habits and dispositions of the local population and a stereotypical ordering source. But it would be false if interpreted with respect to a circumstantial modal base taking into account facts about what morality or the law prescribes and a deontic ordering source where the ideal is such that what morality or the law prescribes is adhered to. Let us assume in turn that in fact countries carefully protect their natural resources. Then [46b] would be false with respect to a circumstantial modal base taking into account facts about the practice of countries with rich natural resources and a stereotypical ordering source where these practices are followed. [46b] may also have a normative generalization reading: in that case it is interpreted with respect to a modal base determined on the basis of facts having to do with what is considered morally proper behavior and an ordering source which ranks worlds according to their closeness to an ideal where that behavior is realized. The generalization expressed by [46b] is that a country with rich natural resources which it does not share with its neighbors is less close to the ideal of morally desirable behavior than a country with rich natural resources which it shares with its neighbors. In these cases, both the descriptive and the normative generalization readings correspond to identical logical structures. The difference between the two types of readings has to do with what the modal base and ordering source are taken to be in each case. For example, the logical form of (46b] is as in [47a] and is interpreted according to [4 7b]. [47]

a. G:r:,y,z(country(x) & resaurces(y) & have(x, y) & neighbar(x, z), shares(x, y, z)) b. ([47aJJ~,c,w = 1, where c determines a modal base Rw

4

and an ordering source ~w iff for every g' g where A= {x,y,z} 61 and for every w 1 E Rw such that (cauntry(x) & resaurces(y) & have(x,y) & neighbor(x,z))~,c,w 1 = 1 there is w2 E Rw such that w2 ~w w1 and for every w3 E Rw such that W3 ~w w2 (shares(x, y, z))~,c,wa = 1. In the logical form of [46b] given in [47a] the dependent definite its neighbors is accommodated within the restriction and there is symmetric quantification over countries, their neighbors and natural resources. A reading where there is asymmetric quantification over countries and their neighbors is also available but I will not I

44

Descriptions in Context

consider it any further here except to note that the familiar problems which arise with respect to conditionals and quantification arise with respect to generics as well. This analysis of I-genericity also accounts for Dahl's (1975) observation that indefinite NP's are associated with a non-accidental generalization reading, under the assumption that an implicit generic operator is present in [48a] and [48b] but not in (48c]. 62 [48]

a. A member of this club does not drink whisky. b. Members of this club do not drink whisky. c. The members of this club do not drink whisky.

The interpretation of the generic operator and its contextdependency with respect to a modal base and an ordering source also help predict when a generic generalization would entail the corresponding actual generalization. This would be the case when the modal base and ordering source are determined by the context to be realistic, or when the modal base is realistic and the ordering source trivial. A modal base is realistic if the worlds in the modal base are selected on the basis of their similarity with the actual world with respect to certain facts, that is if the actual world is among those in the modal base. In all the examples we have considered the modal base is realistic. An ordering source is realistic if worlds are ordered according to how closely they approximate a set of certain actual facts. [49a] is an example where a realistic ordering source is involved. An ordering source is trivial if any two worlds are related by the partial ordering relation. This situation arises when the relevant conversational background is empty. (49b] is an example where the ordering source is trivial, assuming that in all worlds the same physical entity cannot be in two different locations at the same time. [49]

a. These days if a building is designed by an avant-garde architect, it has a spiral-shaped dome. b. A monument on Naxos is not in Athens.

As with modals, certain generic operators may be conventionally associated with and therefore select particular modal bases and ordering sources. For example, overt generic operators like typically and normally select only stereotypical ordering sources. This is why generic sentences with these operators, such as those in [50], express descriptive generalizations, rather than normative ones.

Bare Plurals and Genericity

[50]

45

a. A country with rich natural resources typically shares them with its neighbors. b. #A monument on Naxos is normally not in Athens.

[50a] and [50b] have only the descriptive generalization reading. [50b] in fact is odd since it implies that there might be monuments on Naxos which are (also) in Athens. Conventional association with particular modal bases and ordering sources may also ·account for the seemingly differential force or even opposite truth values of otherwise identical generic sentences containing operators with the same quantificational force. For example, [50b] expresses a weaker generalization than [49b]. The interpretation of the generic operator as a human necessity operator is weak enough and flexible enough, thanks to its relativization to the two context-dependent parameters, but it is still not without some problems. There are cases where it is plainly too strong. This problem is discussed by Krifka et al. (1995) and Asher & Morreau (1991, 1995). Consider [51a]. [51]

a. A turtle lives to be very old. b. A turtle normally lives to be very old.

[51a] is intuitively true but the semantics of the generic operator would make it false if we assume that maximally normal worlds with respect to the characteristics of turtlehood have to be worlds where the same biological facts hold as in the actual world. The semantics of the generic operator in this case is too strong as it requires of all individuals that are turtles to live until very old in all maximally normal worlds. But since such maximally normal worlds are also worlds with the same ecological pressures as the actual world it cannot be true that all turtles survive, even in those maximally normal worlds. The fact that this problem shows up with the singular indefinite indicates that one cannot resort to a kind reading, which involves no quantification, in order to avoid the problem. Although there is no apparent way to fix this problem within this approach by using only one operator, it is not clear that this type of modal analysis is entirely on the wrong track since the problem is contained. Note that [51b] is intuitively false, which is what we would expect if we start with a circumstantial modal base and a stereotypical ordering source. 63 See the appendix for an alternative interpretation for [51a].

Descriptions in Context

46

4.4 Conclusion The indefiniteness analysis assimilates the bifurcation in the readings of bare plurals to that of other indefinite NP's. It predicts that, all else being equal, the range of interpretations for a bare plural are the same as those of any other indefinite NP. Furthermore, in all phenomena that depend on indefiniteness bare plurals should behave like indefinites. For instance, bare plurals should participate in all phenomena involving the interaction of indefiniteness, quantification and anaphora, such as donkey-anaphora (see Diesing (1988), Wilkinson (1991)). Unlike Carlson's analysis, which crucially distinguishes between generic sentences with adverbs of quantification (the former involving quantification, the latter not), the indefiniteness analysis unifies the two cases. The indefiniteness analysis is also more explicit on how context affects the type of generic statement involved (e.g., normative vs. descriptive regularities): the contextual effects are manifested in the choice of modal base and ordering source. The indefiniteness analysis and the kind analysis differ on the predictions they make about the clustering of readings for a given NP in a given language. Since it does not distinguish between 1genericity and D-genericity, the kind analysis predicts that if an NP has a kind reading, then it can have an existential reading as well since existential quantification is introduced by the predicate. Definite generic and kind-denoting NP's such as this kind of N present problems for this prediction. Given the way it captures the variable quantificational force of indefinites, the indefiniteness analysis predicts that an NP can have an existential reading iff it can have an !-generic reading. 64 A prediction that both analyses share is that a bare plural will have universal force only in generic sentences. The functional reading of bare plurals, which is the topic of the next chapter, presents a problem for this prediction.

APPENDIX An alternative strategy for deriving the interpretation of [52] is outlined below. [52) A turtle lives to be very old. Let us assume that the interpretation of [52) involves two operators: in addition to the standard generic operator there is a monadic

41

Bare Plurals and Genericity

operator 0 taking scope over the nuclear scope. The logical form of (52] then is as in (53]. [53] Gx(turtle(x),0live-long(x)) The operator G has a stereotypical modal base determined on the basis of facts having to do with biological laws and inherent properties of turtles and a trivial ordering source. The operator 0 has the same stereotypical modal base as G and an ordering source which depends on the restriction of G: for each individual verifying the restriction of G there is an ordering source which orders worlds according to how close they come to being optimal for the survival of that individual. (54]

a. ([53])~ c w ''

= 1, where c determines a modal base Rw

and an ordering source ~w iff for every g' ~ g and for every w1 E Rw such that [turtle(x))~,c,wx = 1, I

(0live-long(x)]~,c,w 1 = 1.

b. [0live-long(x))~,c,w 1 = 1 iff there is a c' that determines a modal base Rw 1 and an ordering source ~w 1 , 9 1 such that for every w2 E Rw 1 there is W3 E Rw 1 such that W3 ~w 1 , 9 1 w2 and for every W4 such that W4 ~w 1 ,g 1 w3, [live-long(x)]~,c~,w., = 1. I

The interpretation of 0 is non-standard in one respect. In Kaplan's (1989) terms, the operator 0 is a monster, since the formula in its scope is interpreted relative to a different context than the formula headed by 0.

NOTES 1. This already presupposes that generic sentences can be true or false like any other declarative sentence. Of course, the evidence for their truth might be of a different type than that of non-generic sentences. 2. Other theorists, however, may place a stronger condition on the semantics of generics, requiring that it interact with an appropriate notion of defeasible consequence so as to account for the kinds of reasoning supported by generics. Such a view is clearly articulated and developed by Asher & Morreau {1991, 1995). 3. See Dahl {1975), who makes the point very persuasively. 4. As is common practice, I will use the term 'quasi-universal' in order to indicate the tolerance for exceptions.

48

Descriptions in Context

5. See also Carlson (1987, 1988, 1989). For a more comprehensive and critical review of Carlson's analysis and analyses in its steps see Schubert & Pelletier (1987). 6. See also Carlson (1989) for discussion on this point. 7. Gabbay & Moravcsik and Carlson differ, however, on how they execute the idea formally. Gabbay & Moravcsik explicitly construe individuals as functions from points of evaluation to basic elements in their ontology, corresponding to the temporal stages of individuals. Carlson, as we will see, assumes that both the intensional individuals and their extensional manifestations are basic elements of the ontology. 8. For a similar position see also Nunberg & Pan (1975). 9. The model-theoretic construal of kinds and stages need not be as in Carlson. For instance, Chierchia (1982) proposes that kinds be construed as nominalized properties, Ojeda (1991) takes kinds to be the mereological sum element of subsets of the universe of discourse, while Hinrichs (1985) construes stages as spatiatemporal locations. To a large extent, these reconstructions are compatible with Carlson's overall analysis of genericity although they differ from it in certain significant ways. For example, Ojeda's proposal has the consequence that the use of a definite generic carries with it the presupposition of existence of actual instantiations, something that is not true of Carlson's analysis. 10. Superscripts on variables indicate the restriction on the sort of entity the variables can take as their value: k stands for kinds, o for objects and s for stages. z is the definite descriptor. 11. In what follows I limit attention to bare plurals in subject position. 12. Although Carlson does not consider episodic kind-level predicates, they can be easily accommodated in his analysis as long as kind-level episodic predicates do not introduce quantification over stages. This accommodation is possible because the analysis does not dispense with times as parameters of evaluation in favor of stages, a choice that can otherwise be criticized as introducing a lot of redundancy (for such criticism see Schubert & Pelletier (1987)). 13. For purposes of this discussion I will ignore the contribution of be and will treat the whole predicate as stage-level or individuallevel. 14. Whether such predicates disallow the universal reading for the bare plural and lack a characteristic property reading can be contested (see Diesing 1988, 1990). If we admit the characteristic

Bare Plurals and Genericity

49

property reading, they would not constitute a separate category and would be treated on a par with those in category [5c]. 15. Thus, 'Unskilled thieves left a visible trace,' involving an unmediated derived predication, entails the existence of a visible trace in the actual world, while 'Unskilled thieves leave a visible trace,' involving a mediated derived predication, does not. 16. Technically, Carlson follows Montague's (1974) PTQ system in having NP denotations apply to VP denotations. 17. Predications involving stage-level predicates applying to stage-denoting NP's do not exist since there are no stage-denoting NP's. 18. When applied to an object-denoting NP, we get a habitual reading, as in 'John walks in the woods.' 19. This is simplifying in that the translation of the bare plural involves a constant; the complete translation is that given earlier in [4]. For the purposes at hand this is an innocuous simplification. 20. Notice that truth of a sentence involving a kind-level predication does not imply necessity. That dinosaurs are extinct is true in the actual world but nothing in the analysis forces it to be necessarily true. Diesing (1988), in effect, criticizes Carlson's analysis for treating all generic sentences as non-accidental generalizations when in fact it treats all generic sentences as accidental generalizations, in the sense that in order to determine whether a sentence is true in a given world we need not look at any other world. See the discussion below and in section 3.3. 21. An inferred generalization approach to generics, while not having any quantification in the semantics, might encode the connection between predicating something of a kind and inferring something of its instantiations by putting further constraints on models in the form of meaning postulates. Carlson posits no such further constraints (the quasi-meaning postulates in Carlson (1979) have no formal status). Heyer (1985), aiming to characterize different types of kind predications in terms of the inferences about particular instantiations they give rise to, does precisely that. He has direct predications over kinds but places constraints on models such that if a certain predicate is true of a kind then it is true of all its realizations (or all its typical realizations) in each world. 22. I take it that at least normative generalizations, if not dispositional ones as well, are irreducibly modal, that is, their truth depends on the denotation of the relevant predicate across a subdomain of worlds.

50

Descriptions in Context

23. Also, one could have a uniform analysis of genericity (at least the kind of genericity which does not involve proper kind predication) without assuming that all generic NP's are kinddenoting. Farkas (1985) and Schubert & Pelletier (1988, 1989) argue that generic singular indefinites are object-denoting indefinite NP's while bare plurals are kind-denoting NP's. 24. It must be noted that a kind-denoting NP with a basically kind-level episodic predicate is possible, as in 'Dinosaurs ate kelp but at some point they resorted to grass.' However, this sentence means that the kind changed its eating habits and not that all dinosaurs resorted to eating grass on a particular occasion. 25. See Carlson (1989) for the significance of such cases. 26. Even if one were to claim that in this part of the Pacific is a locative subject in (12b] and hence provides a kind-level entity, this would not help us account for one of the readings of (12a] since there the subject is uncontroversially the bare plural typhoons. 27. In the characterization of the two readings, I am following the terminology of Evans (1977). Some authors use the two terms interchangeably. 28. Since I am concentrating here on cases where the pronoun is outside the scope of its antecedent I will ignore bound variable readings. 29. I am disregarding plurality. 30. For an E-type pronoun existence and uniqueness are part of the meaning of the pronoun itself-they are part either of its truth-conditional content, if we follow the Russellian line, or of its presuppositional content, if we follow the line of the presuppositional analyses. For a pronoun of laziness, on the other hand, existence and uniqueness are a by-product of the interpretation of free variables and of the general requirement that the context should provide a unique value for each free variable. 31. This presupposes that the stages of an individual-level entity denoted by an individual-denoting NP do not become salient as a matter of course. They only become salient when they are implicated in the truth-conditions of the sentence containing the individual-denoting NP. 32. If we require that the context should entail the existence of some unique individual-level entity that is fragrant (thereby making the existence and uniqueness part of the presuppositional content of the pronoun}, we must distinguish between existence entailed by the previous discourse on the basis of the truth conditions of the sentences comprising it and existence pragmatically inferred.

Bare Plurals and Genericity

51

The former is the case in [14c], where the meaning of the first sentence involves existential quantification over stages, the latter in [14g), where no existential quantification over stages or objects is involved. Nevertheless, if the first sentence of [14g) is true, then it can be inferred (but not guaranteed by the semantics} that some object realizations of the kind strawberries are fragrant. In order to exclude the reading for the pronoun in [14g) that is equivalent to the definite description the strawberries that are fragrant, we must, therefore, require that existence should be entailed by the previous context in the strict sense. 33. This holds if the bare plural and the pronoun are not in the background of a focus structure. For instance, in contrast to [16), I find (i}, and therefore the pronoun of laziness interpretation, acceptable: (i) John ate prickly pearsi [yesterday)p, while Mary ate themi [today)p. I will not address this problem. 34. [16) is non-deviant if the predicate is given a habitual interpretation; in that case the pronoun would be a pronoun of laziness. 35. This appears to be true for the factors discussed here. Uniform episodic interpretation of the two predicates, as in [16), excludes it: (i) # John ate no applesi. Mary, on the other hand, ate themi a lot. Uniform habitual interpretation of the two predicates or a particular focus structure make it available: (ii} John used to eat no applesi. Mary, on the other hand, used to eat themi. (iii) John ate most apples [yesterday)p, while Mary ate themi [today]p. 36. A further argument for distinguishing proper kind anaphora and anaphora involving a pronoun of laziness interpretation comes from cross-linguistic considerations. In Modern Greek, only pro can have a a pronoun of laziness interpretation; overt pronouns, clitic and non-clitic alike, exclude this kind of interpretation but they can, nevertheless, be anaphoric on kind-denoting NP's. 37. This reading of a kind-denoting NP could be subsumed under what Krifka (1987) and Krifka et al. (1990} have termed representative object interpetation.

52

Descriptions in Context

38. We can also use this kind of evidence to argue against analyzing this kind of N as synonymous with an N of this kind, as proposed in Wilkinson (1988b, 1991). An N of this kind is a true indefinite and, therefore, gets an existential interpretation and gives rise to E-type anaphora. McNally (1992) uses evidence from anaphora and scopal interaction with adverbs of quantification to contrast that kind of N and an N of that kind, claiming that the former does not involve existential quantification over instantiations of the kind. The strongest piece of evidence in favor of analyzing this kind of N as a disguised indefinite comes from its acceptability in there-insertion sentences. However, McNally (1992) has proposed an account of this fact that is compatible with taking the NP to be definite and kind-referring. 39. Carlson (1987) draws the distinction between non-verifying instances and exceptions. 40. One could try to account for this by introducing meaning postulate, as does Heyer (1985}, or by adding a non-monotonic logical apparatus to the kind analysis in order to determine the inferences we get. 41. Wilkinson (1991) points out that Carlson's analysis itself has problems with [23b] since the truth conditions it gives it amount to there being many objects that are Texans and tall (see Carlson (1977:207) ). ·The problem is that Carlson gives [23b] a first-order translation involving unrestricted quantification and conjunction of the two arguments of the quantifier. 42. Carlson's examples are: 'John doesn't like wombats' and 'Jill believes professors are insane'. 43. As L. Horn (p.c.) points out, the weaker reading, corresponding to wide scope negation, is most natural as a direct denial. Obligation-type modals show similar behavior: 'he shouldn't go' in isolation has only the narrow scope negation reading, as opposed to 'he shouldn't go' in response to 'he should go,' in which case it can have either the wide or the narrow scope negation reading. Krifka (1987) and Krifka (1988) also contain discussion of the relative scope of negation and the generic operator. Krifka (1988) claims that the readings corresponding to both scopes are available. 44. Possibility modals favor narrow scope with respect to negation, which again results in a stronger statement. See Horn (1972, 1989). 45. The absence of the narrow scope reading might not be arestriction on scope per se but a consequence of the interpretation of negated generics. (I have in mind Stalnaker's (1968) proposed in-

Bare Plurals and Genericity

53

terpretation for negated conditionals which results in the following equivalence: -.(-+ 1/J) iff-+-.¢.) 46. See Farkas & Sugioka (1983), Stump (1985), Diesing (1988), Wilkinson (1991) for arguments against treating adverbs of quantification as VP operators. My criticism is independent of whether adverbs of quantification are treated as monadic VP operators or as dyadic sentential operators. 47. Heim (1982) applied this idea to the singular indefinite generic. 48. We might need to further subdivide D-genericity; see, for instance, Heyer's (1985) distinction between absolute generic reference and personal generic reference. 49. This is a bit of a simplification since it excludes cases of existential quantification in the restriction. For the time being I am concentrating on the generic reading of indefinites and hence on the cases where they are directly bound by the generic operator. 50. The plurality of the NP might, of course, impose further restrictions on its predicate, such as that it include plural entities in its denotation. See Hinrichs {1985), Schubert & Pelletier (1987) and Wilkinson {1991) on this issue. 51. The notation g' ~ g means that g' is exactly like g except possibly with respect to the value it assigns to x. 52. Rooth sets aside the distinction between stages and objects. 53. Chierchia (1992) takes the generic reading of indefinites, which he uniformly analyzes as dynamic existential quantifiers, to arise when the indefinite is selected as topic and is subject to existential disclosure, an operation that results in the creation of a free variable and the binding of the indefinite by the adverb of quantification. 54. These correspond to Carlson's [7bii], [7cii], [7ciii], [7dii], as discussed in section 3. 55. This issue is also related to Lawler's (1972), Dahl's (1975) and Kleiber's (1985) discussion of existential generics. See also de Swart (1987). 56. Carlson does not actually consider stage-level predicates with adverbs of quantification or other generic operators but it is reasonable to assume that this is how he would analyze such cases. 57. Carlson would, of course, take the predicate of [39c] to be derived from a stage-level predicate via the operator Gn; Kratzer would take it to be an underived individual-level predicate. 58. Some proposals have opted for claiming vagueness of the operator with respect to the number of assignments it takes for

54

Descriptions in Context

the sentence to be true rather than intensionality (Farkas (1985), Wilkinson (1991)). Vagueness alone without intensionality, however, does not suffice; Carlson's arguments against an extensional account carry over. Schubert & Pelletier (1989), although they take the generic quantifier to be intensional, do not construe it as universal since it allows for exceptions. 59. There seems also to be an implication for actual instantiations with [43a]. 60. See also Wertheimer {1972).

4

61. The notation g' g means that g' is exactly like g except possibly with respect to the values it assigns to the elements of A. 62. Of course, [48c] may contain a dependent definite, as in 'The members of this club never drink whisky.' 63. An analysis like Asher & Morreau's {1991, 1995) accounts for the truth of [51a] but not for the falsity of [51b]. In other words, the argument from the turtle problem cuts both ways. 64. Potential problems arise with des-NP's in French, which can have an existential but no generic reading (de Swart 1992, 1996), and with definites in languages such as Romanian (Farkas 1985), Greek (Condoravdi 1992), and French (de Swart 1996}, which can have an 1-generic reading but do not have an existential reading. The latter problem with regard to Greek is addressed in Condoravdi (1992) along the lines of the proposals made in chapter 4.

3 Functional Reading of Bare Plurals

1. INTRODUCTION Indefinite descriptions have played a central role in natural language semantics. Their properties relating to their intuitive quantificational force inside and outside the scope of various operators and their potential for serving as antecedents to pronouns have had pervasive consequences on the overall design of semantic theories. Starting with Russell (1919), indefinite descriptions have traditionally been analyzed as inherently existentially quantified. More complex facts, having to do with their variable quantificational force when in the restriction of operators, originally noted by Lewis (1975), and their ability to serve as antecedents to pronouns outside their syntactic scope, as in intersentential and donkey anaphora, have led to an analysis of indefinites as non-quantificational expressions. In such theories, the intuitive quantificational force of indefinites arises as a result of either construal rules, such as operator indexing and existential closure (Heim 1982), or of principles of interpretation, such as the definition of truth and the satisfaction conditions of dyadic operators (Kamp 1981, Heim 1982). Standard theories of indefinite NP's, including both those that analyze them as inherently existentially quantified and those that analyze them as non-quantificational, variable-contributing elements, predict that indefinites always have existential force if outside the scope of any operator, that they assert rather than presuppose existence, and that they are never anaphoric. Those theories, moreover, in which indefinites can inherit the force of the operator whose scope they are under (Kamp 1981, Heim 1982, Chierchia 1992) predict that an indefinite will have non-existential force only in the restriction of an operator. Finally, indefinites are not con55

56

Descriptions in Context

sidered as essentially context-sensitive, although it is acknowledged that their interpretation may be supplemented by implicit contextual restrictions. In this chapter, I show that there are indefinites which, in one of their readings, do not have existential force even in the absence of any operator, presuppose rather than assert existence and are crucially context-sensitive. Bare plural indefinite descriptions in English are of this type. In the previous chapter I discussed two influential analyses of the bare plural that have sought a unified treatment for its existential and generic readings. One treats bare plurals as kinddenoting terms (Carlson 1977b). The other treats bare plurals as indefinite NP's, relying on the analysis of indefinites developed by Kamp (1981) and Heim (1982) as inherently non-quantificational NPs contributing a free variable and conditions on that variable (Gerstner & Krifka 1987, Krifka 1987, Krifka et al. 1995, Wilkinson 1988a, 1991). It distinguishes between two types of genericity and analyzes one type as involving a sentential dyadic modal operator binding free variables in its restriction. Both analyses are designed to account for the following two generalizations: (i) if the bare plural has universal force, then the reading for the whole sentence is generic; (ii) if the reading for the whole sentence is episodic, then the bare plural has existential force. 1 In this chapter I show that English bare plurals exhibit a universal reading which arises both with individual-level predicates in non-generic sentences, violating (i), and with stage-level predicates in episodic sentences, violating (ii). I refer to this reading as the functional reading since, as will be shown in the next chapter, a contextually salient function is implicated in its analysis. I first present the reading and its associated properties (section 2) and then explore the ways it can be analyzed within a framework of standard assumptions about genericity and indefiniteness (section 3). For purposes of this discussion I will assume the DRT-Heimian quantifier-free treatment of indefinites. However, the argument does not depend crucially on these assumptions and can be reconstructed within the framework of more recent proposals, whereby indefinites are analyzed in terms of first or higher order dynamic existential quantifiers (e.g., Groenendijk & Stokhof 1991b, Chierchia 1992).

Functional Reading of Bare Plurals

57

2. THE FUNCTIONAL READING OF BARE PLURALS 2.1

Initial Observations

A prediction of the standard indefiniteness analysis of bare plurals is that the universal reading of a bare plural, as of any indefinite, will arise only in quantificational contexts. As outlined in the previous chapter, such contexts require the presence of an overt adverb of quantification, a modal, or implicit genericity. A universal reading, however, arises in a wider range of contexts which cannot be straightforwardly assumed to be quantificational. Consider [55a] and three possible continuations, [55b]-[55d]. In 1985 there was a ghost haunting the campus. Students were aware of this fact/the danger. The students were aware ofthis fact/the danger. There were students who were aware of this fact/the danger. Intuitively, [55b] appears synonymous with [55c]. Unlike [55d), [55b) does not make an existential assertion but, like [55c), it is an assertion about the totality of the contextually relevant students, whose existence in the actual world seems to be presupposed by both [55c] and [55 b). Although the bare plural receives a universal reading, [55b] is not generic in any obvious way; it does not express a non-accidental generalization about students in general, nor a regularity about the occurrence of awareness in other situations in which a ghost was haunting the campus. The individual-level predicate is not understood as expressing a characteristic property. 2 That the bare plural in [55b] lacks an existential reading is not surprising given that the predicate be aware is individual-level. The question is whether genericity is involved in [55b). If implicit genericity is involved, how does this square with our intuitions about the meaning of [55 b)? If not, what does the universal reading amount to in the absence of genericity? Although this use 3 of the bare plural in English is quite pervasive, as even a casual look at actual texts makes clear, it has gone virtually unnoticed in the literature. The only works I am aware of that acknowledge it are Lahiri (1991) and Prince (1992). Lahiri observes parenthetically that certain bare plurals take a universal reading with stage-level predicates and assumes that bare plurals may have a definite reading. Prince notes that there appear indefinites in episodic contexts with an unexpected universal read[55]

a. b. c. d.

58

Descriptions in Context

ing in the text under scrutiny in her article. 4 Interestingly, these indefinites are categorized as 'inferrable' within the typology of information-status she proposes. 5 However, inferrable indefinites with a universal reading are inconsistent with the generalizations she advances about the formal marking of an NP, its informationstatus and its interpretation. According to these generalizations, only inferrable definites can have a universal reading. Prince ultimately assumes that such uses of indefinites are deviant and that the text would have been more natural with definites in their place. I will show that the appearance of such indefinites, far from being deviant or a marginal phenomenon, stems from a hitherto unrecognized semantic property of some indefinites, including bare plurals in English. This property gives rise to a universal reading with both individual-level and stage-level predicates in the absence of genericity or any operator. In the remainder of section 2, I discuss the properties associated with such indefinites and show that the synonymy with the definite, the presence of contextual restrictions, the lack of genericity and the presupposition of existence constitute a real and pervasive phenomenon and are not incidental to example [55]. Throughout section 2 my interest is in ascertaining the properties of the functional reading, and therefore the discussion is cast in such a way as to be neutral with respect to the question of whether this reading is due to the presence of an implicit operator. Section 3 spells out a possible analysis in terms of an implicit operator. I will eventually reject it on the grounds that it fails to provide a unified account of the phenomena described. I will present an alternative analysis in the next chapter.

2.2

Functional Reading with Individual-Level Predicates

A crucial step in the argument that the functional reading constitutes a distinct interpretation is to establish that it is not a special case of the generic interpretation. If the functional reading were a special case of the generic interpretation, then it would arise in one of the following three ways: (a) as an entailment of the generic reading, (b) as an implicature of the generic reading, or (c) because of the presence of an extensionalized generic operator. I consider and reject the first possibility in section 2.2.1, where I provide evidence that can tease apart the functional from the truly generic reading. I consider and reject the second possibility in section 2.5. I con-

Functional Reading of Bare Plurals

59

sider and reject the third possibility in section 3, where I develop an account in terms of an extensionalized generic operator, taking into consideration the full range of facts presented in section 2. A generalization that will emerge is that contextual restrictions are consistently associated with the functional reading and that they constitute an integral part of that interpretation (sections 2.2.1, 2.2.2). However, not any contextual restriction is possible, as I show in section 2.2.2, where I discuss the limitations on what can constitute a contextual restriction. 2. 2.1

Genericity and the Functional Reading

As we saw in the previous chapter, both the kind analysis and the indefiniteness analysis of bare plurals take for granted that the universal reading for the bare plural depends on a generic reading for the whole sentence. In [55b], however, the two do not coincide. [55b] expresses a generalization restricted to the actual students on campus on a particular occasion. As discussed in the previous chapter, actual generalizations are not equivalent to generic generalizations since a generic generalization can be true even when the actual generalization is not, and vice versa. The issue is whether the functional reading, which can be described as a contextually restricted actual generalization reading, is in fact the interpretation of [55b], or whether it is a consequence of the generic interpretation under certain circumstances. Now, in order for [55b] to be true, the actual students on campus in 1985 have to have been aware of the danger or of the fact that a ghost appeared. In other words, if [55b] is true, then so is [55c]. But this by itself is not necessarily inconsistent with [55b] 's having a generic interpretation. For instance, the functional reading could be an entailment of a descriptive generic generalization when the generic operator is construed with a realistic modal base and trivial ordering source. If we are to definitively distinguish the functional reading from the truly generic reading, we must show not only that the generalization expressed by [55b] is true in the actual world but also that whether it is true or not depends on nothing but the actual world. If the latter is true, then the bare plural is not in the scope of a modal operator. I present three arguments to distinguish between the generic interpretation and the functional interpretation. The first argument is based on the contextually restricted nature of the generalization expressed by the functional reading. The second argument is based

60

Descriptions in Context

on the implication of existence in the actual world associated with the functional reading. The third argument is based on the scopal interaction between the bare plural and quantificational adverbs and modals. The presence of contextual restrictions and the implication of existence show some fairly straightforward dissimilarities between the generic and the functional reading. The implication of existence and the scopal facts show that the bare plural has wide scope with respect to overt generic operators and modals, hence its interpretation is not dependent on a modal operator. a. Implicit Contextual Restrictions

Contextual restrictions, supplied by the previous discourse, are present with the functional reading of the bare plural in [55b], as they are with the definite in [55c]: [55b] is no more general than [56b), where the restrictions are part of the descriptive content of the NP, and similarly for [55c] and [56c]. [56]

a. In 1985 there was a ghost haunting the campus. b. Students on the campus at that time were aware of this fact. c. The students on the campus at that time were aware of this fact. Moreover, if no students other than the ones on campus at that time were aware of the ghost's appearance, [55b] would still be true. In other words, individuals satisfying the descriptive content of the bare plural NP but not the contextual restrictions are irrelevant in ascertaining the truth of [55 b). While the context of utterance affects the descriptive content of a bare plural NP with the functional reading, this does not happen with generic indefinite NP's. Generic statements do not accept implicit contextual restrictions. This is a general way of stating the observation made by Dahl {1975), Croft (1986) and Krifka {1987), on the basis of examples like [57] and [58], that nominal quantifiers are easily amenable to contextual restrictions while adverbial quantifiers and the implicit generic operator are not, at least with respect to individuals. Specifically, [57b] and [57d] have a reading equivalent to that of [57c] but neither [58a] nor [58c] or [58e] have a reading equivalent to that of [58b] or (58d]. [57]

a. (Out of the blue:) Every lion has a mane. (non-restricted) b. There are lions and tigers in the cage. Every lion has a mane.

Functional Reading of Bare Plurals

61

(restricted or non-restricted) c. There are lions and tigers in the cage. Every lion in this cage has a mane. d. (Context: we are near a cage with lions and tigers) Every lion has a mane. (restricted or non-restricted) [58]

a. There are lions and tigers in the cage. A lion always has a mane. (non-restricted only) b. There are lions and tigers in the cage. A lion in this cage always has a mane. c. There are lions and tigers in the cage. A lion/Lions has/have a mane. (non-restricted only) d. There are lions and tigers in the cage. A lion/Lions in this cage has/have a mane. e. (Context: we are near a cage with lions and tigers) A lion/Lions has/have a mane. (non-restricted only)

The crucial point here is the difference between overt restrictions, which are part of the linguistic descriptive content of an NP, and implicit restrictions, which are provided by the context of utterance. The context of utterance in [57] and [58] provides information on the basis of previous linguistic discourse ([57b], [58a], [58c]), or some salient facts established by the extralinguistic context ([57d], [58e]). The assumption that the functional reading is an entailment of the generic reading which is present when the generic operator is construed with a realistic modal base and ordering source does not by itself explain why the bare plural appears to be contextually restricted. If contextual restrictions are not part of the generic interpretation, then they cannot come for free in the functional reading. We might be able to supplement the entailment assumption with some pragmatic story to the effect that the contextual restrictions are present to guarantee coherence with the previous discourse but I will not pursue this direction any further since there is overwhelming evidence against it. b. Implication of Existence

Unlike generic statements, [55b], repeated here as [59b], implies the existence of students on campus in 1985 in the actual world. 6 That

62

Descriptions in Context

this is a non-trivial implication can be perhaps better appreciated if we consider bare plurals with additional descriptive content, as in (59c]. (59]

a. In 1985 there was a ghost haunting the campus. b. Students were aware of the danger. c. Students with police connections were aware of the danger. d. (But) there were no students (with police connections) on the campus in 1985.

Although the existence of individuals satisfying the descriptive content of the bare plural in (59c] cannot be taken for granted, (59c] certainly implies that there were actually students with police connections on the campus in 1985. Continuing the discourse comprised of (59a] and [59b] or [59c] with [59d] leads to a contradiction. If the generic interpretation were the only interpretation for (59b] and (59c], no implication of existence would be guaranteed since (59b] and [59c] could be true, and even entail the equivalent actual generalization, even if no students with police connections actually existed on campus in 1985. Intersentential anaphora provides an additional piece of evidence for this implication. In order for intersentential anaphora with an indefinite antecedent in a modal/ generic environment to be possible, the indefinite must be asserting existence in the actual world. This generalization was already made by Karttunen {1976) and most theories capture it by assuming that the indefinite takes wide scope with respect to the modal/generic operator. 7 For example, within the framework of Kamp (1981), in order for this kind of intersentential anaphora to be possible, the discourse referent introduced by the bare plural must be at the top-level DRS and therefore accessible to the discourse referent of a pronoun outside the scope of the modal or generic operator. Within the framework of Heim (1982, Ch. II), the indefinite must have wider scope than any other operator, in which case it will be captured by text-level existential closure and will bind the pronoun. 8 Of course, the problem with respect to both frameworks is how the indefinite can have widest scope, on the one hand, and not be interpreted existentially, on the other. This is something I will address in detail later, as it is the focus of the revised theory of indefiniteness that I will propose. At this point, I want to establish that the evidence from anaphora places precisely the constraint for entailment of existence and therefore the requirement for wide scope.

Functional Reading of Bare Plurals

63

The bare plural of [55b] and [59b] can be the antecedent of a pronoun in intersentential anaphora: each of these sentences can be felicitously continued by either [60a], which contains an individuallevel predicate, or [60b], which contains a stage-level predicate. The contextual restrictions are present in the anaphora as well: they picks out the students (with police connections) on campus at the time of the ghosts's appearance, not just the students (with police connections) in the actual world at large. Therefore, [55b] and [59b] must entail existence in the actual world in order for anaphora to be possible. [60]

a. They were well-informed. b. They had been informed by the police.

Since modal subordination and an interpretation for the pronoun as a pronoun of laziness are alternative options which are consistent with the bare plural antecedent's being within the scope of a generic operator, we must make sure that the right type of anaphora is involved. We must, therefore, exclude the possibility of modal subordination and the pronoun of laziness interpretation for [60a] and [60b]. If we assumed that there was modal subordination in [60a] with the pronoun being a pronoun of laziness, this assumption would be open to the same problems as assuming genericity for [55b] or [59b]. Of course, no modal subordination is involved in [60b] since it is an episodic sentence. The pronoun they is not a pronoun of laziness: being aware of the danger and being well-informed are understood to be attributed to the same individuals. 9 If they were a pronoun of laziness, then in the episodic context of [60b] it would have an existential reading, equivalent to that of some students {with police connections) had been informed by the police, a reading that is absent. Moreover, in chapter 2 we saw that if the bare plural antecedent is construed with an individual-level predicate while the pronoun anaphoric on it is in an episodic context, the result is rather deviant (examples [14g] and [15a]). [60b], on the other hand, is perfectly acceptable. There is one remaining option that we must consider. If we take the functional reading to be an entailment of the generic reading relative to a realistic modal base and ordering source, we might still be able to account for the possibility of anaphora by treating the pronoun as an implicit definite description and by assuming accommodation of the information that the actual world did indeed have the relevant individuals at that time. This kind of

64

Descriptions in Context

anaphora is presumably involved in [61], where they is construed as the pomegranates in actual existence right nowin [61b] and as the pirate ships in actual existence right now in [61d]. [61]

a. Pomegranates do not have a crowned end. b. ?? They had it/them chopped off. c. Pirate ships fly a black flag. d. ?? They (all) bought it/them from the same shop. We can assume, along the lines of Heim (1990} and Chierchia (1992), that these pronouns correspond to free functor variables which in these cases take as value a function from worlds and times to the individuals satisfying the descriptive content of the NP at the relevant world and time. The additional contextual restrictions involved in the anaphora in [60] can be accounted for by taking the functor variable to have as value a function from worlds, times and individuals to the individuals satisfying the descriptive content of the NP, i.e., a function from worlds, times and campuses to the students associated with each campus at the relevant world and time . However, there is a clear difference in acceptability between the anaphora in [61] and the anaphora in [60]; in fact, some speakers find the anaphora in [61] totally unacceptable. The difference in acceptability can be linked to the readiness with which the information about the existence of the relevant individuals can be accommodated. Of course, if (55b] and [59b] themselves entail the existence of the relevant individuals, then no accommodation is necessary and the anaphora should be perfectly acceptable, as it in fact is. c. Wide Scope

That [55b] is not generic can also be seen by providing the sentence with an adverb of quantification or an overt generic operator. If the truth of the actual, contextually restricted generalization, which is associated with the functional reading in [55b], were an entailment of the generic reading, we would expect that in the presence of an overt operator the bare plural would be in its scope and we would get the familiar descriptive generic generalization reading, with the entailment about the actual world depending on the force of the operator. But this is not what we find. The universal reading of the bare plural persists even when an overt adverb of quantification with non-universal force is present, which shows that no direct binding by the adverb of quantification is involved. Consider [62]: in (62b], as in [62c], there is an assertion

Functional Reading of Bare Plurals

65

about the totality of the contextually relevant students, with the adverb of quantification taking on a temporal reading. In fact, because the adverb of quantification has a temporal reading, the sense of be aware shifts to be consciously aware in [62b] and [62c), so that the predicate can be temporally relativizable. [62]

a. b. c. d.

In 1985 there was a ghost haunting the campus. Students were usually aware of this fact. The students were usually aware of this fact. Most students were aware of this fact.

Since their interpretation does not depend on the adverb of quantification, the bare plural of [62b] and the definite of [62c] must outscope it. If the bare plural were within the scope of the adverb of quantification, it would be bound by it, and if we took the adverb of quantification to have a modal dimension with a realistic modal base and ordering source, [62b] would at most entail [62d], and not something about the totality of the contextually relevant students. 10 In other words, if the functional reading were simply an entailment of the generic reading, the intuitive reading of [62b] would appear to be [62d] rather than [62c]. In the previous chapter, I claimed that adverbs like normally or typically are generic operators with particular requirements on the nature of their modal base and ordering source. At this point, they can be used as diagnostics for a true generic reading because they cannot be stripped of their modal force. When these adverbs are added to examples like [55], they lead to infelicity, as in [63]. [63] a. In 1985 there was a ghost haunting the campus. b. (#)Normally /Typically students were aware of this fact. c. (#)Normally /Typically the students were aware of this fact. d. (#)Normally /Typically there were students who were aware of this fact. e. The normal/typical students on campus were aware of this fact. The bare plural still patterns with the definite in picking out the totality of the contextually relevant students in the actual world. [63b]-[63d) are infelicitous to the extent that quantification is vacuous, which is a consequence of the fact that the bare plural is not within the scope of the operator in this case. In order to preserve coherence with [63a], a contextually restricted reading is sought for the bare plural which, however, results in the bare plural's outscop-

Descriptions in Context

66

ing the operator. Crucially, the reading that is missing is one where the operator has been stripped of its modal force and the domain of quantification consists of the contextually relevant students with the adverbs contributing an extra restriction on that domain. In other words, [63b] does not share a reading with [63e]. 11 [63b] and [63d] may, of course, have an interpretation in which the adverbs are interpreted as modal operators and the bare plural is within the scope of the adverb but on that interpretation the bare plural is not contextually restricted, no claim is being made about the actual students associated with campus during that time and the sentences are not natural continuations for [63a]. The operators in [63b], [63c] and [63d] can also be interpreted as quantifying over temporally individuated situations. In that case, the bare plural in [63b] outscopes the operator and exhibits the functional reading, a situation parallel to that of [62b]. 12 Thus, we have seen that the bare plural can outscope both an adverb of quantification and an overt generic operator. The wide scope of the bare plural with respect to a quantificational adverbial affects the interpretation of the adverbial in terms of what provides its domain of quantification and results in the functional reading for the bare plural. The functional reading surfaces with modals as well, with the bare plural taking wide scope with respect to the modal. [64b], like [64c], has a reading in which the students are taken to be the actual students. That the modal base of the deontic operator is not realistic is illustrated by [64d], which is a perfectly felicitous continuation for [64b] and [64c]. (64]

a. b. c. d.

A ghost is 4aunting the campus. Students should be aware of the danger. The students should be aware of the danger. Unfortunately, they are not.

Note that anaphora in [64d] is possible (without the need for modal subordination) precisely because (64b] and [64c] entail the existence of students in the actual world. If the bare plural indeed takes wide scope with respect to generic and modal operators, then the truth of sentences like [55b], [59b], [62b] and (64b] depends on nothing but the actual world, and therefore the functional reading is not an entailment of the generic reading.

Functional Reading of Bare Plurals 2.2.2

67

Contextual Restrictions

In the previous section it was shown that bare plurals with the functional reading can be contextually restricted, in contrast to bare plurals within the scope of adverbs of quantification and generic operators. In this section I will show that bare plurals must be contextually restricted on their functional reading. I will call this effect the positive contextual sensitivity of the functional reading. However, the functional reading is systematically blocked if the context providing the contextual restrictions has certain properties. I will call this effect the negative contextual sensitivity of the functional reading. NP's are often dependent on the context of utterance to supply information that determines their interpretation either by restricting their domain of quantification or by providing additional conditions for their descriptive content. The information supplied by the context may be based either on information provided directly by the previous linguistic discourse, or on the shared beliefs between speaker and hearer and the speaker intentions recoverable by the hearer. The role that the context plays in restricting the interpretation of NP's may be viewed either as a purely pragmatic phenomenon, or as the result of an interplay between semantics and pragmatics: the semantics makes available a certain parameter whose value is determined by pragmatic factors, such as saliency, discourse coherence, etc. In any case, whether an NP in a given context of utterance is interpreted as contextually restricted or not, or which contextual restrictions are chosen, depends solely on pragmatic factors. Bare plurals with the functional reading are different in this respect; I will argue that the positive contextual sensitivity and the negative contextual sensitivity of the functional reading depend on semantic properties of the bare plural. a. Positive Contextual Sensitivity

The bare plural in [55b], [59b], [62b], or [64b] must be contextually restricted. The implicit contextual restrictions on the bare plural seem to be part of its meaning as they not cancellable. The discourse in [65] is contradictory. [65]

a. In 1985 a ghost was haunting the campus. b. Students were aware of the danger. c. However, none of the students associated with the campus was aware of the danger.

Descriptions in Context

68

If the bare plural is construed as having a generic reading, no con-

tradiction arises, as in [66]. [66]

a. A ghost is haunting the campus. b. In general, students are aware of this kind of danger. c. However, none of the students associated with the campus is aware of this kind of danger.

For indefinites, in general, there is a preference for a contextually restricted reading, which, however, can always be overridden if, for instance, it leads to contradiction. Therefore, the discourse in [67] is perfectly coherent. [67]

a. In 1985 a ghost was haunting the campus. b. Some students were aware of the danger. c. None of the students associated with the campus was aware of the danger.

Although the indefinite in [67b] can, in principle, be interpreted as contextually restricted with the contextual restrictions contributed by [67a], the presence of [67c] rules out this kind of interpretation. How we take the context of utterance to affect the interpretation of the indefinite in [67b] is irrelevant to the argument. We can attribute the presence of contextual restrictions purely to pragmatics, in which case [67b] would be true depending on the existence of any students having the relevant property. Or we can allow the context of utterance to provide an additional condition to the descriptive content of the indefinite. In the former case, the hearer would draw the inference that some students on campus have the relevant property, so as to make [67b] both more informative and more relevant to the current discourse. In the latter case, the previous discourse could provide the relevant restriction while the continuation with [67c] would indicate to the hearer that he was wrong in assuming the context provided a restriction and should reconsider. The crucial point is that the meaning of the indefinite in [67b] is such that it is compatible with any number of contextual restrictions, or none at all, while that of the bare plural requires a specific kind of contextual restriction. Not only should the bare plural be contextually restricted but the contextual restrictions must be provided by the discourse prior to the utterance of the bare plural. Consider the difference between the discourses in [68] and [69] .13 [68]

a. A burglar was roaming Santa Clara county. b. Deputy sheriffs were aware of the danger.

Functional Reading of Bare Plurals

69

c. #They had been sent from LA county to investigate. [69]

a. A burglar was roaming Santa Clara county. b. Several deputy sheriffs were aware of the danger. c. They had been sent from LA county to investigate.

The same point is also made by cases in which the sentence containing a bare plural is first in a discourse, with the NP that can provide the contextual restrictions in the following sentence, as in [70]. [70] a. Students were roaming the streets. b. A school nearby had ended classes early. In that case, the bare plural has only an existential reading although the next sentence can, and in fact does, provide additional restrictions for it. b. Negative Contextual Sensitivity

Necessary as the contextual restrictions may be for the functional reading to arise, not any contextual restriction is possible. There are contexts which provide extra information that can in principle constitute a further restriction for the bare plural but which in fact does not. [71c] still expresses the same generalization as [55b] in the context of [71a] and [71b], not the more contextually restricted one corresponding to every student in this dormitory. The contextually restricted reading is easily available for the definite in [71d] and the quantificational NP in [71e]. [71]

a. b. c. d. e.

There is a ghost haunting the campus. There are 500 students in this dormitory. Students are aware of the danger. The students are aware of the danger. Every student is aware of the danger.

Admittedly, a discourse comprised of [71a], [71 b] and [71c] sounds incohesive. Rather than being at odds with the point being made here, this can be taken as further support for it, since the more contextually restricted reading is absent even when discourse cohesiveness would require it. Such a discourse would in any case improve if it were followed by something that would justify the shift from one group of students to the other, like e.g. 'The students of the dormitory have, moreover, taken strict precautions.' Similarly, in a deictic context, like that of [72], the bare plural receives the same reading as [55b], not the more contextually restricted one picking out the perceptually salient students.

70

Descriptions in Context

[72] (Context: We know that there is a ghost haunting the campus. We are standing in front of the library and we can both see several students.) Students are afraid to enter the library. While the perceptually salient students can be taken as providing the evidence for the generalization, the generalization is not restricted to them as far as the meaning of [72] is concerned. As descriptions and nominal quantifiers in general readily accept contextual restrictions constrained only by pragmatic considerations, the selectivity of the bare plural is highly surprising and unprecedented. The following question arises at this point: is what matters the contextually supplied information, or the kind of context that supplies it? In [71], as in [55] and in all the examples we have looked at so far except for (72], the context supplying the information and hence the relevant restrictions is the previous linguistic discourse. In (72) the context is extralinguistic. However, the fact that the bare plural in [72] can have a contextually restricted reading that encompasses all the students on campus shows that the extralinguistic context can supply the relevant information. We are driven, therefore, to viewing the kind of information that is contextually supplied as being responsible for what is an acceptable and what is an unacceptable contextual restriction. Later I will argue that this is not a limitation on contextual restrictions per se but a consequence of the fact that the bare plural cannot be anaphoric in some technical sense. A theory that would allow us to account for this limitation in these terms is preferable to one that would identify the phenomenon as direct limitation on the contextual restrictions. Contextual factors affect the common ground, that is the information state in which speaker and addressee presume themselves to be and against which new linguistic utterances are evaluated.

2. 3

Functional Reading with Stage- Level Predicates

If the functional reading were an entailment of the generic reading,

we would expect it to be limited to bare plurals construed with individual-level predicates or with stage-level predicates within a quantificational context. In this section, I demonstrate that the universal reading of bare plurals shows up with stage-level predicates in purely episodic contexts. Consider (73], where the bare plurals opponents and proponents appear in an episodic context.

Functional Reading of Bare Plurals

71

[73) Although the odds still seem to favor Senate approval of Thomas, opponents redoubled their effort and tried to delay a floor vote on confirmation ... Proponents, in contrast, demanded a vote next week. (San Francisco Chronicle, Sept. 28, 1991) [73) can be understood either as an existential statement about some opponents and proponents of the approval, or as a statement involving the totality of the opponents and proponents of the approval. The former reading is the expected reading for a bare plural in an episodic context but the latter is rather surprising. Moreover, the totality effect associated with the second reading is independent of the kind of predication involved. Both readings allow for a distributive or a collective predication, that is both readings are compatible with there having been either individual attempts for delay and demands for a vote, or a single collective attempt and a single collective demand. Similarly, the examples in [74) are ambiguous: on one reading, they are synonymous with the corresponding ones in [75], and on the other, they are understood as involving the totality of the entities specified by the NP. On the latter reading, for example, [74a] is a statement about all (relevant) linguistic theories, [74b] implies that the total number of victims rescued by rescue teams is 28,950, 14 [74c) is a promise about all (relevant) details and [74d] an announcement about all {relevant) prices. In none of these cases do we have a generalization over episodes. [74)

a. Linguistic theories have posited abstract representations. b. Rescue teams have rescued 28,950 victims. c. Details will be presented tomorrow. d. Prices went up today.

[75]

a. There are linguistic theories that have posited abstract representations. b. There are rescue teams that have rescued 28,950 victims. c. There are details that will be presented tomorrow. d. There are prices that went up today.

Let us look a bit more closely at the kinds of contexts that give rise to the functional reading with stage-level predicates and the kinds of contextual restrictions present. In [73] what is explicitly uttered provides the contextual restrictions: the approval of a nom-

Descriptions in Context

72

ination is at issue and the bare plurals opponents and proponents are understood as elliptical descriptions whose full form would be opponents of the approval and proponents of the approval, respectively. I have not provided any explicit context for [74]; however, for [74b] and [74c] a situation is easily conjured up that could provide the contextual restrictions: in [74b] a natural disaster, in [74c] a presentation. [74a] and [74d] are somewhat different in that they can be uttered without any explicit context. Rather, the existence of the relevant entities relies on some general background facts that seem to constitute general background knowledge and taken for granted by a given linguistic community, for example, an economy such that goods have prices, or one with intellectual disciplines that have theories. The generalization emerging from these cases is that the existence of the entities satisfying the description of the bare plural is stereotypically tied to some other entity or to some world where some typical facts hold and which we take the actual world to be. These observations relate to Prince's observations about 'inferrable indefinites' with a universal reading. The fact that a bare plural can have a universal reading with a stage-level predicate in an episodic context contradicts the generalization stated in the previous chapter that a bare plural with a universal reading does not combine with an episodic predicate. It also provides rather striking evidence that the functional reading is not generic. Adding a generic operator results in complete illformedness, as in [76], where the possibility of a regular generic reading is absent because the verb corresponds to an episodic predicate. 15 [76)

a. #Normally /Typically rescue teams rescued 28,950 victims yesterday. b. #Normally /Typically details will be presented tomorrow. c. #Normally /Typically prices went up today.

A correlate of the ambiguity exhibited by the bare plural with stage-level predicates in episodic contexts is that a pronoun outside the c-command domain of the bare plural has a totality reading in addition to an E-type interpretation. TheE-type interpretation correlates with the existential reading for the bare plural, the totality reading with the universal reading for the bare plural. Consider [77] and [78]. [77]

a. Prices went up today. b. They were expected to go up.

Functional Reading of Bare Plurals

[78]

73

a. There are prices that went up today. b. They were expected to go up.

If [77a] is interpreted as synonymous with [78a], the pronoun in [77b], as in [78b], is interpreted as equivalent to the prices that went up today. In that case [77b] and [78b] would be true in a situation in which only some prices were expected to go up and those were the prices that actually did go up. If, on the other hand, the bare plural in [77a] has the functional reading, then the pronoun in [77b] is interpreted as equivalent to all prices. The functional reading also arises in quantified contexts, where, as with individual-level predicates, the indefinite outscopes a modal or an adverb of quantification, the latter having a temporal reading. [79]

a. Prices must come down (for the good of the people). b. Unfortunately, they will not.

[80]

a. There were several alerts today. b. Rescue teams usually reacted promptly.

One could claim that the wide scope reading, and, more generally, the functional reading of bare plurals with stage-level predicates, is really a specific existential reading. I argue against this possibility in section 2.5.

2.4

Other Indefinites

The functional reading does not arise with other indefinites. [81 b] shows that the singular indefinite, in exactly the same context as that of [55b], has only an existential reading. Similarly for the plural indefinite with the determiner some in [81c]. [81] a. In 1985 there was a ghost haunting the campus. b. A student was aware of the danger. c. Some students were aware of the danger. Given that the bare plural lacks an existential reading and the singular and plural indefinites the universal reading, [8th] and [81c] have no reading in common with [55b]. 16 No such discrepancy between the singular indefinite and the bare plural occurs in standard cases of generic quantification. For instance, in the context of [82a] both [82b] and [82c] have a generic interpretation. [82]

a. A single bad grade ruins a student's reputation forever. b. Students are (always) aware of this danger. c. A student is (always) aware of this danger.

74

Descriptions in Context

In parallel fashion, indefinites other than the bare plural construed with a stage-level predicate show no ambiguity between an existential and a universal reading. (83]

a. A linguistic theory has posited abstract representations. b. A detail will be presented tomorrow. c. An opponent redoubled his effort. d. Some details will be presented tomorrow. e. Some food prices went up today.

This is another piece of evidence that the usual notion of genericity is not involved in the contextually restricted universal reading.

2. 5

Excluding a Purely Pragmatic Account

Before continuing, I want to establish that the functional reading of a bare plural cannot be given a purely pragmatic account, thereby supporting the claim that it constitutes an interpretation distinct from either the generic or the existential. A pragmatic account might seem prima facie desirable, at least on methodological grounds, since it would avoid multiplying ambiguities. In this section, I will outline the general character of a pragmatic account and identify the inadequacies of failing to recognize the functional reading as the actual interpretation of the bare plural in the cases considered so far. Further facts associated with the functional reading that will be presented in later sections exclude a purely pragmatic account altogether. However, a pragmatic analysis is untenable even for the facts discussed so far and I believe it is interesting to see why. A purely pragmatic account would claim that the facts of the functional reading are not facts about the meaning or interpretation of bare plurals but solely facts about their use. An account in which pragmatics plays a role by fixing some aspect of the interpretation, such as the analysis based on degenerate genericity presented in section 3 of this chapter, is not a purely pragmatic account by this criterion. In a purely pragmatic account, the role of context consists in generating implicatures and not in satisfying preconditions for interpretation. Such an account would capitalize on the Gricean distinction between sentence meaning and speaker meaning and would attribute the effects of the functional reading to speaker meaning. In Grice's seminal distinction, sentence meaning is that given by the semantic interpretation, and, relative

Functional Reading of Bare Plurals

75

to the specifications a given context of utterance provides for the context-sensitive expressions in the sentence, it determines what is said. Speaker meaning, or what is meant, is what the speaker intends to communicate relative to a given occasion of utterance and is determined in part by general principles underlying rational communication. According to a pragmatic account, therefore, bare plurals with stage-level predicates in episodic sentences have just the usual existential reading and bare plurals with individual-level predicates have just the usual generic reading, and that is all the semantics should be responsible for. The functional reading arises when, due to certain contextual factors, what is meant is distinct from what is said. There are two ways I can see of working this out. One is to analyze the functional reading as an implicature that can be associated with both the generic reading and the existential reading. I will call it the implicature approach. The other is to assimilate the functional reading to the referential or specific uses of indefinite descriptions. I will call it the referentiality approach. In what follows I will spell out these two approaches in more detail and examine them in turn. For the sake of concreteness, let us assume that what is said and what is meant are (possibly distinct) propositions and let us follow common practice (e.g., Ludlow & Neale (1991)) in referring to the former proposition as the proposition expressed and to the latter as the proposition meant. 2.5.1

The Implicature Approach

The implicature approach would aim to derive the actual, contextually restricted generalization reading as an implicature from the generic reading, in the case of individual-level predicates, and from the existential reading, in the case of stage-level predicates. Thus, for an example like [55b], where the predicate is individual-level, the proposition expressed corresponds to a generic generalization, while the proposition meant corresponds to an actual, contextually restricted generalization. For an example like [73], where the predicate is stage-level, the proposition expressed is existential, while the proposition meant is the (stronger) actual, contextually restricted generalization. The question then is what contextual factors give rise to the difference between the proposition expressed and the proposition meantP

76

Descriptions in Context

The functional reading with individual-level predicates could arise if the generic generalization is patently false and this fact is obvious to both speaker and hearer. Under the assumption that the speaker would not try to communicate something for which he has insufficient evidence or something that is mutually believed to be false, the hearer would draw the inference that the speaker actually intends to communicate something weaker but true, namely an actual and contextually restricted generalization. The functional reading with stage-level predicates could arise if the speaker has every reason to believe that the individuals of whom the existential statement is true are exactly all the individuals of whom the statement is true and intends to communicate that to the hearer, while the hearer is aware of the speaker's relevant beliefs and intentions. If the functional reading is attributed to speaker meaning, the presence of contextual restrictions and the implication of existence could be accounted for .18 The speaker would intend to communi-· cate the proposition corresponding to the functional reading when his or her beliefs are formed on the basis of some actual individuals in a particular context. Such a take on the issue, while having these advantages over the entailment alternative--which in any case could not account for the existence of the functional reading with stage-level prediactes-faces some serious problems as well. First, the implicature approach gives two disparate explanations for the functional reading in episodic sentences and in generic sentences. The explanations are, in general, distinct because the mechanisms responsible for getting from what is said to what is meant are not the same. In the reading with stage-level predicates what is meant is stronger than what is said (universal vs. existential), while in the reading with individual-level predicates what is meant is weaker than what is said (accidental vs. generic generalization). To illustrate let us consider a variation of [73], in which one of the NP's is construed with an individual-level predicate and the other with a stage-level predicate. [84] The odds still seem to favor approval of the nominee but proponents believe it will be a close vote. Opponents, for their part, tried to delay the vote yesterday, hoping time is on their side. According to the implicature analysis, the first sentence is false and the second sentence is true while the speaker, by uttering them,

Functional Reading of Bare Plurals

77

intends to communicate with both something about the entire set of proponents or opponents of the nomination. More seriously, an indefinite NP does not aquire a universal reading just because the equivalent universal statement also happens to be true. In general, there is no intuition of a universal reading, in the same way that there is one in the case of the functional reading, in contexts in which both speaker and addressee have evidence supporting the truth of the universal statement. For instance, suppose I utter [85] while looking outside the window, having every reason to believe that the dogs I can see are all the dogs of the neighborhood and all of them are tearing up my backyard, and intending to communicate that belief, while the hearer is aware of that fact as well as of my intention. [85] Dogs are tearing up my back yard. [85] does not acquire the functional reading in such a context. Although what I intend to communicate could well be something about all the dogs in my neighborhood, this does not affect the reading of the indefinite NP, whose force remains existential, even when the hearer draws the appropriate inference. If generic sentences could be used in order to indirectly communicate information about actual entities in the way outlined above, then, on the one hand, all else being equal the type of the NP should make no difference, and on the other, a wider range of false generic statements should be subject to this charitable reinterpretation. Neither of these implications is true. Utterances of generic sentences are not appropriate for such a task and fail to communicate a contextually restricted actual generalization. As shown in section 2.4, the functional reading is restricted to bare plurals. The fact that an actual generalization is manifestly true does not affect the interpretation of the corresponding generic generalization so as to make it true, nor does it give rise to the functional reading. False generic statements remain false and do not give rise to the functional reading even when the context of utterance is such that it would support the truth of the contextually restricted actual generalization. False generic statements, as in [86] (from Carlson (1977b:37)), where the equivalent non-generic, contextually restricted, universal statements are true, are a case in point. [86]

a. Books have between 100 and 150 pages. b. Sleds are black. c. Dogs are collies.

78

Descriptions in Context

Carlson uses these examples to show that the existential reading is absent with individual-level predicates since they are judged as false; if the existential interpretation were available, all the sentences in [86] would be true on that reading since there exist books that have 100 to 150 pages, sleds that are black, and dogs that are collies. We can also use these cases to show that the functional reading does not arise even when the context of utterance supports its truth. On the implicature account, the proposition expressed by each one of [86] is false but the proposition meant, which is one embodying an actual and restrictive generalization, is true. 19 Not only do the sentences in [86] remain false, no functional reading for the bare plural is detected, even if they are uttered in a context where all the contextually relevant books do indeed have between 100 and 150 pages, or all the contextually relevant sleds are indeed black, or all the contextually relevant dogs are collies. On the other hand, the functional reading arises regardless of what we believe about the truth of the corresponding generic statement, i.e., a sentence could be true both on the generic and the functional reading of the bare plural, as illustrated by [87]. [87]

a. A ghost is haunting the campus. b. Students are aware of dangers of this kind. c. Unfortunately, the students on this campus are not. Finally, the implicature account both overgenerates and undergenerates functional readings. Negative contextual sensitivity and false generic sentences lacking the functional reading exemplify overgeneration and infelicity or contradiction due to the functional reading of the bare plural exemplify undergeneration. Negative contextual sensitivity cannot be accounted for since there is nothing that would exclude the relevant speaker's meaning in the presence of a previous NP with identical descriptive content. Previous discourse in general helps provide contextual restrictions for NP's and there is no way it can be excluded here on any principled grounds. As for undergeneration, if the actual generalization reading is simply an implicature, why is the implicature not cancelled, or to put it somewhat differently, why does the implicature arise in the first place in cases where it is strongly disfavored? For example, the presence of normally or typically in [63] leads to infelicity rather than cancellation of the implicature, and in the discourses of [59] and [65], [59c] and [65c]lead to contradiction rather than can cella-

Functional Reading of Bare Plurals

79

tion of the implicature. The functional reading of the bare plural has a direct influence on the interpretation of the quantificational elements and consequently on sentence meaning. 2.5.2

The Referentiality Approach

The referentiality approach would claim that the functional reading is the manifestation of specificity or referentiality with bare plural indefinite descriptions. For the sake of concreteness, let us frame the issue in terms of the conception of referentiality and specificity of Ludlow & Neale (1991). Ludlow & Neale's primary aim is to argue against an ambiguity for indefinites, and in the course of their argument they develop a typology of uses of indefinite NP's that is useful for our purposes. Ludlow & Neale's aim is to defend the Russellian account of indefinites as uniformly quantificational elements against the evidence arguing for an ambiguity between a quantificational and a referential interpretation. According to Russell's account, a sentence containing an indefinite expresses a general proposition: it makes an existential statement and therefore its truth or falsity depends on there being an object, any object, that satisfies the relevant conditions. A sentence containing a referring expression, on the other hand, expresses a singular proposition: 20 it is a statement about the object referred to by the referring expression so its truth depends on whether that object satisfies the relevant conditions. 21 As part of their argument, Ludlow & Neale develop a taxonomy of the distinct uses of indefinites and characterize the utterance contexts that give rise to them. They make a three-way distinction between the proposition expressed by a given utterance, the proposition meant, and speaker's grounds, the proposition which is the object of the speaker's belief providing the grounds of an utterance. For an utterance containing an indefinite, the proposition expressed is always a general proposition with existential import. But on different contexts of use the speaker's grounds and the proposition meant could well be singular. An indefinite has a referential use if the speaker grounds and the proposition meant constitute a singular proposition. An indefinite has a specific use if the speaker grounds constitute a singular proposition and the proposition meant a general proposition. As a matter of course, bare plurals have both referential and specific uses, at least when interpreted existentially. An example of referential use is [85]. An example of specific use is an utterance

80

Descriptions in Context

of [88] where the speaker knows directly the individuals providing the information about the president and intends to communicate that she knows. [88] Sources close to the president told us that he is ready to resign. However, referential or specific use is neither necessary nor sufficient for the functional reading to arise. Deictic contexts are prime candidates for supporting the referential use of descriptions. Let that be the context for (89]: suppose I utter it while my addressee and I are looking outside the window, intending to communicate the belief that the particular dogs that are perceptually salient will tear up my back yard, while my addressee is aware of my intention to communicate something about those dogs. [89] Dogs will tear up my back yard. (85], however, does not acquire the functional reading in such a context. The context is optimal for the referential use of the bare plural dogs (in fact, the bare plural is used referentially) and yet the functional reading does not arise. A context in which I have some particular individuals in mind and assert something of them while expecting my addressees to recognize that I am intending to communicate something about those particular individuals (although they may not know who the individuals are) is a good candidate for the specific use. Let that be the context for (90]: suppose I know exactly which raccoons ate my flowers-say because they have made multiple appearances in my garden, so I have a way of identifying them uniquely, and I saw them at work the night before-and that my addressees also know that I have some particular raccoons in mind. [90] Raccoons with a ferocious appetite ate my flowers last night. Again, however, (90] does not acquire the functional reading in such a context. The context is optimal for the specific use of the bare plural raccoons with a ferocious appetite (in fact, the bare plural is used specifically) but the bare plural has only existential force. So specific and referential uses of a bare plural are not sufficient conditions for the functional reading to manifest itself. Conversely, the presence of the functional reading does not imply that the bare plural indefinite is used referentially or specifically. I may utter (84] without having the slightest idea who the

Functional Reading of Bare Plurals

81

opponents and proponents of the nomination are, nor do I indicate by such an utterance that I do in fact know. The arguments against the referentiality approach carry over to the analysis of the functional reading as a semantic phenomenon involving specificity jreferentiality. The position that indefinites, such as the singular indefinite, can be ambiguous between a referential and a non-referential interpretation is defended, among others, by Chastain (1975), Fodor & Sag (1982) and most recently by de Hoop (1992) with respect to indefinites and the individual/stagelevel contrast. De Hoop claims that individual-level predicates require a strong reading whereas stage-level predicates simply allow it, where the strong reading comprises a family of interpretations including the referential interpretation. Under such a view, the existence of the functional reading would be seen as evidence for semantic ambiguity. 22 The proposition meant would be taken to be identical with the proposition expressed and they would both be singular propositions because the bare plural would be given a directly referential interpretation. This account would have the advantage of giving a unified explanation for the functional reading both with individual-level and stage-level predicates (since in both cases the reading would come about because of the semantic ambiguity of the indefinite), and of explaining the wide scope of the bare plural with respect to adverbs of quantification and generic operators. But if referential or specific use is not sufficient for the referential or specific interpretation to arise, and if the referential or specific interpretation does not require referential or specific use, how can the bare plural be directly referential but not context-sensitive? As I understand it, the proponents of a referential interpretation for indefinites must be committed to both theses: direct reference and context-sensitivity Although direct reference by itself does not presuppose contextsensitivity, the case for the referential interpretation of descriptions is made precisely by instances when the context of utterance provides a referent satisfying the descriptive content of the description. Unless the context of utterance provides a referent, the interpretation of such descriptions cannot be fixed. More importantly, the functional reading does not always correlate with widest scope possible for the indefinite, as we will see in the next section. This, of course, more than anything else undermines any pragmatic account, as well as the semantically referential approach.

82

Descriptions in Context

2. 6 Functional Reading in Quantified Contexts In section 2.2.1 I established that the interpretation of the bare plural on its functional reading does not depend on an adverb of quantification, a generic operator, or a modal. A crucial piece of evidence is that in the presence of an overt operator of this kind the bare plural can take wide scope. But is wide scope a necessary characteristic of the functional reading? After all, if the functional reading precluded anything but widest scope possible for the bare plural, then this would give credence to the directly referential interpretation analysis outlined in the previous section. If, however, the functional reading is a distinct interpretation and not reducible to direct reference, it should in principle be independent of the scopal interactions of the bare plural with respect to various operators. In this section I show that this is indeed the case. 2. 6.1

Dependent Functional Reading

The functional reading can arise when the bare plural is within the scope of an adverb of quantification and in that case it affects the way the elements in the domain of quantification are individuated. The bare plural is dependent on some other variable which is directly bound by the operator (a situation variable) and the contextual restrictions are sensitive to that variable as well. Consider [91], and take [91b]-[91d] to be possible continuations for [91a]. The adverb of quantification usually is given a non-modal interpretation and its domain of quantification is determined, at least in part, by the occasions of a ghost's appearance. [91}

a. b. c. d.

Ghosts have occasionally haunted this campus. Students were usually aware of the danger. The students were usually aware of the danger. A student was usually aware of the danger.

To begin with, there is a difference between [91b) and [91c) on the one hand, and [91d] on the other. When the bare plural and the definite are within the scope of the adverb of quantification, they affect the restriction of the operator. The singular indefinite, on the other hand, is not part of the restriction; an indefinite cannot in general constitute the restriction by itself when the adverb of quantification is non-modal. 23 [91d] has only the reading in which the indefinite has wide scope relative to the adverb of quantification or a nuclear scope existential reading. 24

Functional Reading of Bare Plurals

83

The bare plural and the definite can scope freely with respect to the adverb of quantification. They take wide scope in [91] if it is assumed that the students on campus have remained the same throughout the time during which occasional appearances of ghosts took place. They take narrow scope if it is assumed that the choice of students depends on the choice of occasion in which a ghost appeared on campus. In either case the bare plural exhibits the functional reading. For [91b] as well as for [91c] to be true, it must be the case that for most ghost appearances the totality of the students on campus at the time were aware of the danger. In other words, the force of the bare plural does not directly depend on the adverb of quantification. Let us consider more closely how that interpretation of [91 b] could come about. (92] contains two possible logical form representations for [91b]. In [92a] quantification is symmetric over situations of a ghost's appearance and students. In [92b] quantification is asymmetric over student-containing situations of a ghost's appearance. [92]

a. Usuallya,x(s: 3y(ghost(y) & be-on-campus(y,s)) &

student(x) & be-on-campus(x,s), be-aware-af -the-danger( x)) b. Usually 8 (s: 3y,x(ghost(y) & be-on-campus(y,s) & student(x) & be-on-campus(x,s}}, be-aware-af -the-danger( x))

The bare plural has universal force regardless of whether we take quantification to be symmetric or asymmetric. Both representations in [92], if unsupplemented by any further conditions on the individuation of the elements in the domain of quantification, give us the wrong reading for [91b]. For [92a] to be true it must be the case that for most ghost appearances most students on campus at the time were aware of the danger. For [92b] to be true it must be the case that for most ghost appearances in which there were students on campus those students were aware of the danger associated with that appearance. Now whether this gives us the right reading or not depends on how we individuate the situations in the domain of quantification. 25 For each situation we must select the maximal collection of students on campus during the occasion of a ghost's appearance. This is necessary, or otherwise we would run into the proportion problem. For example, assume there have been four appearances of a ghost, and during one of them the number of students happened to exceed the number of students of the other

84

Descriptions in Context

three occasions taken together. If in the former case the students were aware of the danger while in all the others they were not, then, unless we allow for maximality, [91b] should be true, whereas intuitively [91b] is judged to be false. Imposing maximality on the student variable in [92a] will also give us the right reading for [91b]. Thus, if we analyze the bare plural in [91b] as bound by the adverb of quantification, then the corresponding variable must pick out the maximal collection of students in each case. Alternatively, if we quantify asymmetrically over situations, the situations must be individuated in such a way as to contain the maximal collection of students in each case. Interaction with nominal quantifiers, which always quantify asymmetrically, shows that the problem is not simply in needing asymmetric quantification (and therefore not quantifying directly over students) but that the whole restriction is affected by the maximality effect of the functional reading, as in [93]. Of course, this is also shown by the fact that [92b] is not an adequate logical form representation for [91 b] even though it involves asymmetric quantification. [93] Most of the ghosts that students liked were good-spirited. In order for [93] to be true most of the ghosts that were liked by the totality of the students must have been good-spirited. The cases of the proportion problem that are familiar from the literature arise when a quantificational sentence is analyzed as involving symmetric quantification over (at least) two variables when in fact it should be analyzed as involving asymmetric quantification over one variable with the second variable being dependent on it. In the case described here, the maximality forced by the functional reading of the bare plural constrains the individuation of the entities in the domain of quantification; if it is not taken into account the proportion problem arises. What is common to both cases is that the individuation of the entities in the domain of quantification is the wrong one; what determines the right individuation is, however, different in each case. A complex array of factors is responsible in choosing between symmetric and asymmetric quantification, as various authors 26 have argued. For the case involving the functional reading, as I will argue in the next section, the reason is simple: individuation is determined by a presupposition of existence associated with the functional reading. In that respect, the behavior of bare plurals with the functional reading under quantification parallels that of definites under quantification.

Functional Reading of Bare Plurals

85

In particular, the maximality effect of the bare plural on the domain of quantification parallels cases in which the presence of a donkey pronoun analyzed as a definite description affects the individuation of situations/eventualities in the domain of quantification of a given adverb of quantification, as argued by Kadmon (1987, 1990) and Heim (1990). In those analyses, definite descriptions are associated with existence and uniquenessfmaximality presuppositions. Similar effects to those in [91] can be seen in cases where the restriction of the adverb of quantification is provided directly by a conditional clause rather than by previous discourse and the indefinite or definite NP is in the main clause, as in [94], or in cases in which the restriction is provided partly by previous discourse and partly by an overt conditional clause which contains the indefinite or definite NP, as in [95]. [94]

a. Usually, when a ghost haunted this campus, students were aware of the danger. b. Usually, when a ghost haunted this campus, the students were aware of the danger. c. Usually, when a ghost haunted this campus, several students were aware of the danger.

(95]

a. A ghost has occasionally haunted this campus. b. Usually, when students were aware of the danger, the ghost was satisfied. c. Usually, when the students were aware of the danger, the ghost was satisfied. d. Usually, when a student was aware of the danger, the ghost was satisfied.

The indefinite several students in [94c] and the bare plural students in [94a] differ in force, and even in the effect they have on the domain of quantification. Several students has a nuclear

scope existential reading and it does not affect in any way the individuation of the elements in the domain of quantification, which consists of the temporally maximal occasions of a ghost's haunting the campus. The bare plural has the dependent functional reading and it does affect the domain of quantification, which consists of those occasions of a ghost's haunting the campus in which there are students present on campus and such that they are distinct if the student body changes. The definite has a dependent reading and affects the domain of quantification in the same way. For example, assume a situation in which there have been 7 distinct occasions of

86

Descriptions in Context

a ghost's haunting the campus; during 4 of them no student was in attendance while during the other 3 students were in attendance and during 2 of those latter all students were aware of the danger. Now in such a situation [94c) can be judged false whereas [94a) and [94b) are judged true, the reason being that the irrelevant occasions in which no student was in attendance are part of the occasions in the domain of quantification for [94c) but not for [94a) or [94b). 27 This is not to deny that [94c) may be interpreted with respect to a more contextually restricted domain of quantification, one which excludes the irrelevant occasions. The point is that the occasions in which there are no students in attendance simply do not count for [94a) and [94b) but they may count for [94c). 28 In [95b) and [95c) quantification is taken to be over situations and maximal collections of students. In [95d) quantification is either symmetric over situations and students or asymmetric over situations in which at least one student was aware of the danger. With stage-level predicates, where the bare plural is ambiguous between the existential and the functional reading, the maximality effect arises only when the bare plural is not interpreted existentially, as seen in (96] and (97]. Therefore, the maximality effect is not due to the plurality of the bare plural. [96]

a. Ghosts have occasionally haunted this campus. b. Students usually protested in the main square.

[97]

a. This proposal has been made many times in the last 100 years. b. Proponents usually tried to defend it on the basis of... while opponents usually tried to attack it from within.

The functional reading within a quantified context preserves the crucial properties of the functional reading in an unembedded context. The presence of contextual restrictions is necessary: whether students other than those associated with the relevant campus are aware of the danger or not is irrelevant to the truth of [91b], [93], [94a], [95b) and [96b). Moreover, all of these sentences have an implication of existence of individuals satisfying the descriptive content of the bare plural relative to the bound situation variable. Again, this point can be made more convincingly with a bare plural with additional descriptive content, as in [98b). (98]

a. Ghosts have occasionally haunted this campus. b. Students with police connections were usually aware of the danger.

Functional Reading of Bare Plurals

87

[98b] implies that there were students with police connections during all the occasions of a ghost's appearance. It must be said that there is a certain variability of judgment with respect to this implication: rather than a universal implication, the sentence may have a (weaker) existential implication, namely that in some such situations there were students with police connections. 29 I will come back to this issue in the next section. 2.6.2

Quantificational and Modal Subordination

In the examples we considered in section 2.6.1, the choice of students varied with respect to the choice of situation but the element associated with the NP providing the contextual restrictions did not. Now, if the NP providing the contextual restrictions for the bare plural is itself within a quantified environment, the bare plural requires quantificational subordination. Consider [99]. [99]

a. Every campus I visited had a ghost. b. Students were aware of the danger. Students loved the ghost. c. The students were aware of the danger. The students loved the ghost. d. Some students were aware of the danger. The bare plural in [99b), like the definite in [99c), has to be relativized to the choice of campus. 30 The indefinite in [99d), on the other hand, does not need to be subordinated even when it is contextually restricted. Therefore, it is not the contextual restrictions alone that trigger the subordination. The need for quantificational subordination must be related to the requirement of the bare plural for contextual restrictions. If the NP providing the contextual restrictions for the bare plural is in a modal environment, modal subordination is necessary for the functional reading to arise. Consider first [100] and take the indefinite a campus to have narrow scope with respect to the epistemic possibility modal may. [100]

a. A ghost may be haunting a campus. b. Students must be careful. c. Students are careful.

The modal base of the modal in [lOOb] is relativized to those worlds in which a ghost is haunting a campus and the students are taken to be those associated with that campus. [lOOc], where there is

88

Descriptions in Context

no modal subordination, is taken to be a descriptive generalization about students in general. Consider next [101] and [102], where the NP providing the contextual restrictions for the bare plural is within the scope of (and bound by) a deontic necessity operator. [101] a. A campus should have its own ghost. b. Students must be careful (in that case). c. The students must be careful (in that case). d. Students are careful (in that case). e. The students are careful (in that case). f. Some students are careful/Some students are careful in that case. [102]

a. It is required that the discussion about a proposal of this kind finish within a day. b. Proponents usually /must make a case for it quickly.

[lOlb], [lOlc], [101d] and [lOle] are interpreted as if they had an implicit antecedent of the form if a campus has a ghost. [101 b] and [!Ole] contain on overt modal which is interpreted deontically with a modal base determined by the [lOla]. [lOld] and [lOle] must be interpreted as descriptive generic generalizations, 31 where the worlds in the modal base are those in which what is commanded in [lOla] holds and the choice of students is relativized to the choice of campus with a ghost. Similarly [102b] is interpreted as if it had an implicit antecedent of the form if a proposal of this kind is under discussion. Moreover, the choice of students is relativized to the choice of campus ([lOlb], [lOld]), the choice of proponents to the choice of proposal ([102b]). We can say that in those cases we have both quantificational and modal subordination. Interestingly, [lOlf] can be making a claim about actual students. Apart from underscoring the necessity for contextual restrictions, the modal subordination facts are also significant in that they show that any attempt to explain away the functional reading as a by-product of a generic generalization involving a realistic modal base and trivial ordering source is bound to fail. The generalization from the quantificational and modal subordination facts is that if the licensing NP is under the scope of an operator, the bare plural too must be under the scope of that operator or an operator subordinated to the former. In other words, the bare plural does not scope independently of its licensing NP. This requirement is similar to that of anaphoric definites with indefinite antecedents in the scope of an operator, as originally ob-

Functional Reading of Bare Plurals

89

served by Lakoff (1972) and Karttunen (1976). Quantificational and modal subordination have become prominent in recent discussions of anaphora. Their more general characteristic, of which anaphora is a special case, is that quantificational and modal structures provide a context which serves as the context for subsequent sentences in the discourse. The problem of discourse subordination, in its most general form, is characterized as follows by Roberts (1989:717): "in each case, the second sentence in a discourse is interpreted as involving an operator (explicit or implicit) whose force is relativized so that it ranges only over the type of situation given in part by the first sentence." There are at least three different sources for discourse subordination: 32 (a) domain selection, (b) presupposition triggering elements, (c) elements interpreted as bound variables. 33 At first sight, we might be inclined to add necessary contextual restrictions to the list but as I will argue later this can be subsumed under (b). Therefore, (b) is involved in inducing quantificational subordination in [99); (a) and (b) in inducing modal subordination in [101), [102). 34 To sum up section 2.6, the bare plural on its functional reading exhibits scopal interactions which have an effect on the individuation of the domain of quantification parallelling that of definite descriptions. The need for contextual restrictions induces quantificational and modal subordination. Subordination should be taken here to be a descriptive term and need not imply subordination at some level of representation. The real issue is that the interpretation of some expression is relativized to elements in the previous discourse not in the usual way; for instance, it may be relativized to elements whose scope it is ostensibly not under.

2. 7 The Presupposition of Existence When not in the scope of a modal operator, a bare plural on its functional reading entails existence in the actual world, without, however, having existential force or asserting existence. 35 When the bare plural is in the scope of a modal operator, what is entailed depends on the previous discourse. For example, [lOOb) on its own does not entail that there are students but in the context of [100a) it implies a conditional statement of existence: if there is a campus that is haunted by a ghost, then that campus has students. Where does the information about existence come from? And why is it sensitive to the discourse preceding the sentence containing the bare plural?

Descriptions in Context

90

In this section I will argue that the functional reading presupposes existence. To say that the bare plural presupposes existence means that entities satisfying the descriptive content of the NP plus the additional contextual restrictions are presupposed to exist (in the actual world, when in the outermost context). In section 2.7.1 I argue that simple sentences of the sort we have been looking at possess this presupposition. In sections 2.7.2 and 2.7.3 I concentrate on more indirect effects that the presence of the presupposition has. One effect is that the presupposition exhibits the usual presupposition projection effects (section 2. 7.2). If an adverb of quantification is present, its domain of quantification is restricted as a result of the projection of this presupposition. If the bare plural is in the antecedent of a conditional then the previous context must entail the existence of the entities satisfying its descriptive content. A different and more striking effect is that bare plurals can co-occur with a special class of quantificational adverbials, which I call contextually restricted proportional adverbs of quantification. I will argue that these adverbials presuppose the existence of the group forming the basis of the proportion and quantify over the atomic parts of this group (section 2.7.3). In the discussion below I take presupposition failure to result in a truth-value gap and infelicity. This implies both a semantic and a pragmatic notion of presupposition. How these two notions can be unified will be discussed in chapter 4. 2. 7.1

Simple Cases

That students with police connections exist is a precondition for a felicitous utterance of [103b] and of the negated [103c] and must be taken for granted by the discourse participants. 36 [103]

a. In 1985 there was a ghost haunting the campus. b. Students with police connections were aware of the danger. c. Students with police connections were not aware of the danger. Moreover, the contextual restrictions are part of what is presupposed. What must be taken for granted for a felicitous utterance of [103b] or [103c] is not simply the existence of students with police connections at large, nor the existence of such students on some campus or other, but the existence of students on the campus that makes [103a] true.

Functional Reading of Bare Plurals

91

The fact that the implicit contextual restrictions are part of the presuppositional content of the bare plural explains why the information contributed by them is not cancellable, as shown by [65] and [68], and why they are only provided by prior discourse, as shown by [68] and [70]. That a presupposition is present can be established most convincingly by cases in which it is known by the discourse participants that the relevant entities do not exist, as in [104]. Both [104b] and [104c] are infelicitous and cannot be assigned a truth-value since it is the case that Yale has no fraternities and therefore there are no fraternity members associated with it. a. A ghost was haunting Yale last year. b. #Fraternity members were aware of the danger. c. #Fraternity members were not aware of the danger. If [104b] were false, then [104c] would be true, contrary to intuitions.37 An interesting contrast arises when the bare plural is construed with a stage-level predicate, which allows for either the existential reading or the functional reading. If the bare plural in [105] is interpreted existentially, the existence of opponents of the proposal is asserted and certainly it need not be taken for granted for a felicitous utterance of [105]. If, on the other hand, the bare plural receives the functional reading, the existence of opponents of the proposal must be taken for granted for a felicitous utterance of [105]. [105] The proposal will be voted on tomorrow. Opponents demanded that the vote be shifted to next week. [104]

A more indirect effect of the presence of the existential presupposition arises with if any elliptical clauses. Bare plurals pattern with definites and contrast with other indefinites in supporting the hedge of an if any elliptical clause, as seen in [106]. [106]

a. There is a ghost haunting the campus. b. The students with police connections, if there are any, must be aware of the danger. c. Students with police connections, if there are any, must be aware of the danger. d. #Some students with police connections, if there are any, must be aware of the danger.

If any elliptical clauses, in my view, do not simply cancel a conversational implicature; if they did, the existential implication asso-

92

Descriptions in Context

ciated with definite descriptions would simply be a conversational implicature. Rather, they make the assertion of the sentence they are associated with conditional. In fact, they might signal that the speaker takes himself to be in two conversational backgrounds, one in which the existence of the relevant individuals is taken for granted and one in which it is not. If any elliptical clauses then can be seen as relativizing the assertion to the former conversational background; elements presupposing existence would be felicitous with respect to such a conversational background, whereas elements asserting existence would not. 2. 7.2

Projection of the Existential Presupposition

If an existential presupposition is present, then, given the way presuppositions are inherited by complex constructions containing the presupposition-triggering element, it should surface, in a modified form, when the bare plural is within the scope of an operator. Indeed, the existential presupposition associated with the functional reading has the following consequences. (a) It affects the domain of quantification of a given operator, (b) it shows the usual presupposition projection effects in one-case (epistemic) conditionals, (c) it shows presupposition projection effects and affects the domain of quantification in multi-case conditionals. In section 2.6 I discussed how the functional reading of the bare plural restricts the domain of quantification and affects the individuation of the elements in it. We also saw there that a sentence like [98b], repeated here as [107b], implies that for each occasion there were students with police connections, or that for some occasions there were students with police connections.

[107]

a. Ghosts have occasionally haunted this campus. b. Students with police connections were usually aware of the danger.

In fact, it is the existential presupposition associated with the bare plural inherited by the whole sentence that is responsible for these effects. The instability in the judgements of what the actual presupposition is is manifested in all cases where a presuppositiontriggering element is within the scope of an operator. Let me give some background on the standard views on the effects of presuppositions on domain selection and on presupposition projection in conditionals. It is an interesting and unsettled issue what the presuppositions of sentences with various operators are, or in other words, what the filtering effects of various operators

Functional Reading of Bare Plurals

93

are. 38 It is an unsettled issue because intuitions oscillate between a strong (universal) version and a weak (existential) version and because existing theories are equipped to handle one but need special provisions to handle the other one. 39 Let us consider the problem with respect to nominal quantifiers and the existential presupposition associated with a definite description. What the filtered presupposition is depends, among other things, on whether the presupposition-triggering element is in the restriction or the nuclear scope of the quantificational structure and on what the force of the quantifier is. In [108a] and [llOa] the presupposition-triggering element is in the nuclear scope; in [109a] it is in the restriction. The universal presupposition of [108a], [109a] and [110a] is the proposition expressed by [108b], [109b] and [110b], respectively; the existential presupposition is the proposition expressed by [108c], [109c] and (110c], respectively. (108]

a. Every nation cherishes its king. b. Every nation has a king. c. Some nations have a king.

[109]

a. Every nation that cherishes its king will be rewarded. b. Every nation has a king. c. Some nations have a king.

[110]

a. No nation cherishes its king. b. Every nation has a king. c. Some nations have a king.

The theory of Heim {1983) without provisions for local accommodation predicts universal presuppositions for [108a], [109a] and [110a]. Universal presuppositions can be gotten by global accommodation, existential presuppositions by local accommodation. Cooper (1983) predicts universal presuppositions too. Beaver (1992) opts for the weakest presuppositions possible with the caveat that anything stronger could be built on top. For universal quantifiers he basically builds the presupposition as domain restriction (see the inference table on p. 12).40 Similarly, in van der Sandt's (1992} analysis presuppositions end up as domain restrictions, as a consequence of intermediate accommodation. Intuitions are indeed unstable. However, the degree of variability and the factors affecting the choice in particular instances do not vary with respect to the existential presupposition of definite descriptions and the existential presupposition of bare plurals with the functional reading. For example, some speakers find that

94

Descriptions in Context

in cases of restricted quantification where the proposition corresponding to the universal presupposition is definitely not within the common ground (because, for instance, it is presupposed to be false) a definite description in the nuclear scope is deviant, as in [llla]. Those speakers also find that a bare plural in a similar environment cannot have the functional reading, as in [lllb]. [111]

a. Every European Community country adores its royal family. b. Always, if a ghost is present on an Ivy league campus, fraternity members are aware of the danger.

The inheritance properties of presuppositions in conditionals work out in the following way: the presuppositions of the antecedent are inherited by the whole conditional, the presuppositions of the consequent are filtered through in the form of a conditional whose antecedent consists of the content of the antecedent of the main conditional and whose consequent consists of the presuppositions of the consequent of the main conditional. Another way of formulating the inherited presuppositions of the consequent is the following: those presuppositions that are not entailed by the antecedent are inherited by the whole conditional, those that are entailed are filtered through in the form of the conditional mentioned above. The presupposition (in their terms, conventional implicature) that Karttunen & Peters (1979) associate with 'if then 1/J' is 4>i & (e---+ 1/Ji). 41 Subsequent works have accepted these as the presuppositions that ought to be associated with conditionals although they have tried to derive the projection properties on the basis of the interpretation of conditionals. Let us then see that the existential presupposition associated with the bare plural on its special reading is projected in the usual fashion in conditional sentences. Indeed, [112a] and [112b] as a whole presuppose that there are students with connections in the police department. [112]

a. If students with connections in the police department are aware of the danger, they will inform the rest. b. If a ghost is present on the campus, students with police connections are aware of the danger.

The conditionals in [112] are meant to be one-case conditionals, so the implicit necessity operator (assumed to be present in all conditionals) is to be taken as having an epistemic modal base. Also in [112b] the presupposition is inherited by the whole conditional under the assumption that the presence of a ghost makes no differ-

Functional Reading of Bare Plurals

95

ence one way or another with respect to the existence of students with police connections on the campus. More accurately, we might say, following Karttunen & Peters {1979}, that the actual presupposition is 'If a ghost is present on the campus, there are students with police connections' and that we get the stronger 'there are students with police connections' by pragmatic strengthening. In the consequent of a conditional, the singular indefinite may have an existential reading, as in (113b]. 42 In the same position, the bare plural in [113a] has only the universal reading. Moreover, [113a], but not [113b], seems to presuppose that there are students with connections in the police department in all campuses in the domain of quantification. This must be because of the existential presupposition associated with the bare plural. [113]

a. Usually, if a ghost is present on a campus, students with police connections are aware of the danger. b. Usually, if a ghost is present on a campus, a student with police connections is aware of the danger.

Assuming the account of presupposition projection proposed by Karttunen {1974} and Heim (1983), the existential presupposition associated with the bare plural in the consequent of (113a] affects the domain of quantification as follows. Given that the presuppositions of the consequent must be entailed by the previous context augmented by the local context provided by the antecedent, the existence of students with connections in the police department must be entailed for each choice of campus and occasion of a ghost's presence on a campus. Therefore, we assume either (i) that there aren't any campuses without students with police connections (universal presupposition and global accommodation), or (ii) if such campuses exist, that they are not in the domain of quantification (existential presupposition and local accommodation resulting in domain restriction). 2. 7.3

Functional Reading with Adverbs of Quantity

Bare plurals on their functional reading can co-occur with a class of adverbs which are sensitive to certain properties of the NP's they co-occur with. The bare plural in [114b], like the plural definite in [114c] and in contrast to the singular indefinite in [114d] or the plural indefinite in (114e], is compatible with such adverbs. The only possible reading for [114d] and [114e] is one in which mostly/for the most part are predicate modifiers, arguably within the AdjP headed by aware, specifying the degree of awareness. 43 Such a

96

Descriptions in Context

reading is excluded in [114f] because of the syntactic position of the adverb, hence [114f] is unconditionally unacceptable. [114g], on the other hand, is acceptable, indicating that the quantificational reading for the adverb arises with the bare plural as well as with the plural definite. [114]

a. There is a ghost haunting the campus. b. Students are mostly /for the most part aware of the danger. c. The students are mostly /for the most part aware of the danger. d. (#) A student is mostly /for the most part aware of the danger. e. (#) Some students are mostly /for the most part aware of the danger. f. # A student/Some students for the most part is/are aware of the danger. g. The students/Students for the most part are aware of the danger. The pattern in (114] shows that the distribution of mostly/for the most part is not just a matter of the plurality of the accompanying NP, nor a matter of definiteness alone. The ambiguity of the bare plural with a stage-level predicate disappears once such an adverb is added to the sentence. The bare plural in the variation of [73] given in [115] has only the functional reading. [115] Although the odds still seem to favor Senate approval of Thomas, opponents, for the most part, redoubled their effort and tried to delay a floor vote on confirmation. This is telling evidence that what the adverbs are sensitive to is the interpretation of the NP they co-occur with rather than some superficial co-occurrence restrictions. Let us call this the selective affinity of bare plurals for such adverbs. Similarly, in contexts mirroring those of [71], [72] and [58] where contextual restrictions for the bare plural are impossible, the functional reading does not arise and the adverbs of quantity are not interepreted as contextually restricted. Consider [116], where only the non-contextually restricted reading is possible for the bare plural-giving rise to falsity, in fact-while the contextually restricted reading is present for the definite NP. [116]

a. There are lions and tigers in this cage.

Functional Reading of Bare Plurals

97

b. The lions are mostly /for the most part old. c. Lions are mostly /for the most part old. (no contextually restricted reading) Once again, the constraints on the selection of appropriate contextual restrictions are not determined entirely by the quantifier but depend crucially on the indefinite NP. That bare plurals can co-occur with such adverbs is an observation that was independently made by Lahiri (1991), who named the adverbs 'adverbs of quantity.' In fact, it was in this connection that he noted that bare plurals may have a universal reading with stage-level predicates. He also noted that the existential reading for the bare plural excludes an adverb of quantity. The examples he offered are given in [117]. 44 [117] a. Experts, for the most part, are blaming government policies for the recession. b. ?Men, for the most part, are playing in the garden right now. Lahiri attributed the co-occurrence of bare plurals with adverbs of quantity to their denoting kinds, relying on Carlson's analysis of bare plurals as kind-denoting terms, and as such having part structures. 45 He attributed the selective affinity of bare plurals for such adverbs to the stage-level predicate's introducing existential quantification which binds the variable corresponding to the bare plural and thus leaving no variable for the adverb to bind. As will become clear below, I will adopt one of the central insights of Lahiri's analysis of adverbs of quantity but will provide an alternative account of what determines the class of NP's these adverbials can co-occur with, an account which does not rely on an analysis of bare plurals as kind-denoting. First let us see that these adverbs are special in certain respects. Within the Lewis-Kamp-Heim tradition, the paradigm case of quantificational adverbials are adverbs of quantification like always, usually, never, which are analyzed as quantifying over the parameters of evaluations, i.e., worlds and assignment functions. Adverbs of quantification bind variables corresponding to indefinites and when they have a modal dimension they give rise to the generic reading of indefinites. However, there is evidence coming form different sources that the class of quantificational adverbials is not homogeneous. Evidence from indirect questions, free relatives and correlatives indicates that the class is split between the standard adverbs of quantification and another class into which adverbs

98

Descriptions in Context

of quantity fall (see Berman (1990), Srivastav (1991), Lahiri (1991), Ginzburg (1992)). On the basis of different considerations, Lohner (1985, 1987) drew the distinction between referential and generic quantification. Standard adverbs of quantification participate in generic quantification. Adverbs of quantity participate in referential quantification. According to Lohner, referential quantification, as in [118a], requires a specified, contextually restricted domain of quantification, which is provided by a definite NP. Generic quantification, as in [118b], operates with an open set as the domain of quantification. [118]

a. The apples were for the most part sour. b. Apples are usually sour.

Kroch (1979) observed that adverbs of quantity can quantify over times (or occasions), that they can always co-occur with a definite NP on the relevant reading, and that they cannot co-occur with a quantificational NP on the relevant reading. For instance, he observed that [119a] is ambiguous between the reading of [119c] and the reading of [119d], while [119b] only has the reading of [119c]. (119] a. My friends are generally /mostly /by and large honest. b. All of my friends are generally /mostly /by and large honest. c. All my friends are honest on most occasions. d. Most of my friends are honest. The various discussions in the literature in which the contextually restricted proportional adverbs of quantification have figured share the concern with the following two issues. One of the issues is whether they are monadic VP operators or dyadic sentential adverbs of quantification. The other issue is the proper characterization of the NP's they co-occur with. Kroch (1979), whose interest was to distinguish plural definites from universally quantified plurals, assumed that adverbs of quantity quantify over elements of the set denoted by the plural subject NP and claimed that they cannot co-occur with universally quantified NP's because there can be no double quantification over the same variable. Dowty & Brody (1984) focused on floated quantifiers like all, each, both but one can extend their proposal to the adverbs under discussion. They analyzed floated quantifiers as monadic VP operators which, given their semantics, must be hosted by NP's that denote principal filters. That is, the host NP's must denote fam-

Functional Reading of Bare Plurals

99

ilies of sets having a non-empty intersection. This includes both definites and quantificational NP's with a universal determiner. 46 De Swart (1991) takes the quantifying adverbials of referential quantification to be floating quantifiers, and following Dowty & Brodie (1984), analyzes them as relations between sets of individuals: the set that is the generator of the definite NP and the set that is the set of individuals denoted by the VP. If we follow the Dowty & Brodyjde Swart line of analysis, we would have to analyze the bare plural exactly like a definite in order to account for its co-occurrence with these adverbials. In other words, we would have to assume that a bare plural is ambiguous between an indefinite and a definite interpretation. I take the negative contextual sensitivity facts to argue strongly against this ambiguity analysis. A bare plural is never anaphoric even on its functional reading. Somehow the analysis of the functional reading should not give up the assumption that the bare plural is indefinite. Lahiri (1991) proposed that adverbs of quantity involve amount quantification and analyzed them as denoting a function from ordered pairs of individuals to truth values. The individuals that are the arguments of such a quantifier are the maximal elements of which a given predicate is true and the quantifier compares their extent, that is it compares the number of their atomic subparts. Lahiri suggested that bare plurals are compatible with adverbs of quantity because they are kind-denoting and kinds have a part structure. Now if this is the reason why bare plurals can co-occur with these adverbs this cannot explain their co-occurrence with bare plurals on the functional reading. As I have been assuming so far, the kind interpretation is not responsible for the functional reading. But even assuming that bare plurals are kind-denoting (and disregarding the functional reading for the moment), it is not obvious why stage-level predicates disallow the co-occurrence of bare plurals with adverbs of quantity. For example, why aren't the arguments of the quantifier for the most part in [120a] construed as in [120b]? [120]

a. #Boys, for the most part, are playing. b. Most(ux[boys(x)], ux[boys(x) & 3y(R(y, boys) )(play(y ))]

According to the interpretation that [120b] receives, the number of boys that have stage-level realizations that are playing is greater than the number of boys that do not. Moreover, whatever the

Descriptions in Context

100

means by which that interpretation is excluded they should be such that they do not exclude it for pronouns anaphoric on a bare plural. As can be seen in [121], such a pronoun can co-occur with an adverb of quantity. [121 b] is interpreted according to [121c]. [121]

a. Boys are hiding in the garden. b. For the most part, they are playing. c. Most(ux[boys(x) & 3y(R(y,boys)) (hide-in-the-garden(y)], ux[boys(x) & 3y(R(y, boys))(play(y))]

Therefore, as far as I can see, Lahiri's proposal cannot account for the co-occurrence of bare plurals with adverbs of quantity only if they do not have an existential reading {which of course includes the functional reading). I will follow Lahiri (1991) and Ginzburg {1992) in assuming that these adverbials quantify over atomic elements. I will, moreover, assume that they presuppose the existence of the element whose atomic parts they quantify over. It is exactly the presupposition of existence that makes both definites and bare plurals on their functional reading acceptable with these adverbials. Adverbs of quantity seem to be the adverbial equivalent of partitives.

2. 8

Overview

Contrary to what any current theory of indefinites would lead us to expect, bare plurals exhibit a reading which is closer to the interpretation of definite NP's. This is most immediately obvious with stage-level predicates in episodic sentences, where the force of the bare plural is universal rather than the expected existential. But it is also the case with individual-level predicates, although in that case one has to provide subtler arguments and more extensive evidence to distinguish the functional from the generic reading. A successful analysis of the functional reading would have to account for the following facts: • the universal force of the bare plural • the non-generic interpretation of the bare plural • its occurrence with both individual-level and stage-level predicates • the scope dependence of the bare plural on its licensing NP • its apparent synonymy with the definite in certain contexts • its positive and negative contextual sensitivity • its implication of existence

Functional Reading of Bare Plurals

101

• its sole association with bare plurals rather than all indefinites • its effect on the domain of quantification of adverbs of quantification • the existential presupposition • its acceptability with adverbs of quantity An obvious possibility is to say that the bare plural is ambiguous between an indefinite and a definite interpretation. After all, it appears to be in free variation with the corresponding definite in certain contexts, such as in [55], [62] and [63], and the quantificational contexts of [91], [94], [95]; it co-occurs with an adverb of quantity on a non-generic reading, as in [114]; and it has an existential presupposition, the hallmark of definiteness. Moreover, unlike an indefinite, it is context-sensitive in that it must be contextually restricted in a particular way. But ultimately, we do not want the bare plural to be a disguised definite description of the usual sort. [71d], [72], [58c], [58d], [58e], [86] and [116c] constitute evidence against such an approach. On the disguised definite description analysis the question is what kinds of antecedents are acceptable for the bare plural on its definite interpretation. The generalization emerging from the facts considered is that the bare plural never has an anaphoric reading, that is a definite reading where an explicit antecedent is available. In other words, putting together the evidence from the presupposition facts and the negative contextual sensitivity facts, we arrive at the following generalization: a bare plural on its functional reading is associated with an existential presupposition but it is never anaphoric. In fact, it is in free variation with the corresponding definite only when the definite is neither anaphoric nor deictic. At this point, I use 'anaphoric' and 'deictic' as descriptive terms. The DRT and Heimian analysis of definiteness makes all definites anaphoric semantically, even those that one would not characterize this way on descriptive grounds. The descriptive term for those definites that appear to have something in common with bare plurals on their functional reading is 'associative' (see Hawkins (1978), to whom the term is due, and Heim (1982), who discusses how associative definites can be assimilated to anaphoric definites). 47 I will come back to this issue in the next chapter. The real challenge, therefore, is to account for the properties of the functional reading while maintaining that the bare plural is an indefinite NP. The question then is in what way the theory of indefiniteness should be revised in order to accommodate the

102

Descriptions in Context

functional reading. Let us take the facts listed below to be the central facts of the functional reading; already in my presentation of the full array of facts connected to the functional reading I have indicated how they all revolve around these three basic facts and how once we have an account of these three the rest would follow. • the universal force • the positive and negative contextual sensitivity • the existential presupposition In what follows I will explore two alternatives: one, in keeping with standard assumptions about indefiniteness, attributes the functional reading to an operator; the other, in a more radical move, attributes the functional reading to the fact that a certain felicity condition of familiarity is associated with bare plurals. In the next section I propose an operator analysis, which starts from the universal force of the bare plural and accounts for the other properties by certain additional assumptions. In the next chapter I will propose an account that rests on a more fine-grained conception of novelty and gives a formulation of it that will allow for the functional reading and for a characterization of the difference among indefinites.

3. AN OPERATOR ANALYSIS Is the functional reading of the bare plural due to the presence of an implicit operator? Are there cases in which apparent nongeneric quantification is a trivial case of generic quantification? In this section I will consider how these two questions can be answered affirmatively. I will develop an operator analysis, which starts with the assumption that nothing special needs to be said about the bare plural and no revision in the analysis of indefiniteness is required. The bare plural is not ambiguous and its range of readings depends on whether it is under the scope of an operator or not, and, if it is, on the type of operator it is under. Rather, it is our analysis of genericity and of adverbial operators that needs some revision. An operator analysis of the functional reading will assimilate it to the universal reading of indefinites arising in quantificational contexts. This is of course the most obvious way of accounting for the functional reading as it would stay in line with all other assumptions about indefinites. According to this view, what appears to be an unbound indefinite is in fact bound by an operator with the caveat that this operator bas somewhat different prop-

Functional Reading of Bare Plurals

103

erties from those standardly associated with adverbial operators. However, as I show, the operator analysis fails to provide a unified account of the functional reading, and it leaves unanswered the central question how the positive and negative context sensitivity and the presence of a presupposition of existence relate to each other and to the reading. In the process of developing the operator analysis, I also relate the functional reading to other proposals about indefinites that have been made recently within the overall program of analyzing indefinites as non-quantificational, and show that none of them is sufficiently equipped to account for the functional reading either. In what follows, I will adopt the theory of indefinites presented in Heim (1982, ch. II}, with which I will assume basic familiarity. The analysis of indefinites will have to be supplemented with additional assumptions, as will become clear, but not in ways that require a major departure from the central assumptions of that theory.

3.1

Degenerate Genericity

We have established that the functional reading cannot be explained away simply as an entailment or as an implicature of a realistic generic generalization reading. Instead it must be recognized as the actual interpretation of the bare plural. We can reconcile this fact with the standard view that connects the universal force of a bare plural with genericity if we assume that the functional reading is a special case of the generic reading when we have degenerate genericity. Degenerate genericity is the case when a generic generalization reduces to an actual generalization. It will arise from an extensionalized generic operator, that is a generic operator with a trivial modal dimension. In that case, quantification is vacuous with respect to worlds and the generic operator ends up quantifying only over individuals. Once we have a degenerate generic generalization, the corresponding actual generalization is not just entailed by the generic generalization but it concides with it. Degenerate genericity should thus be an option allowed by the semantics, with the pragmatics determining when it arises. Degenerate genericity parallels the case in which conditionals, which are essentially modal, 48 reduce to material implication. This happens when the implicit necessity operator is associated with a totally realistic modal base and trivial ordering source (see Kratzer

104

DP.scriptions in Context

(1981) and Heim (1982)). A totally realistic modal base is one which contains only the actual world, i.e., for all WI, WI E Rw iff WI = w. A trivial ordering source is such that WI ~ w for any WI and w. As the modal dimension of a generic operator is contextdependent, the apparent ambiguity of the bare plural is due to the choice of modal dimension for the implicit generic operator: the generic reading results when the operator has a non-trivial modal dimension, the functional reading results when the operator has a trivial modal dimension. On that analysis then, our initial example (55b) involves the familiar quantificational structure in (122a) and ends up expressing an actual generalization because the modal base is determined by the context to be totally realistic and the ordering source is determined to be trivial. In addition to a trivial modal dimension, we must assume that restrictions supplied by the context of utterance further limit the domain of quantification to the relevant entities. For example, for (55b) the context restricts the domain of quantification to the set of students on the campus in 1985 during the ghost's appearance. (122] a. G~(student(x)&CR(x), be-aware(x,)) b. In order for the text of which (122a] is part to be felicitous relative to sothe context c, c must provide a modal base and an ordering source for the operator G'. c. Relative to a totally realistic modal base and a trivial ordering source, (g, w) satisfies (122a] iff for every (g', w), where g' ~ g, if (g', w) satisfies student( x) & CR(x), then it also satisfies be-aware(x, ). In the representation (122a), G' stands for the extensionalized generic operator. CR is a cover designation for the additional contextual restrictions, such as being on campus in 1985. We can follow Cooper (1979) and assume that it is a property-denoting metavariable containing only variables and parentheses. In this case, it can be instantiated as P(y)(x) (ignoring the temporal restriction), where P is given the value be-on by the context and y is the variable corresponding to the NP the campus in (55a]. 49 For the sake of concreteness, I have given the predicate be aware a formula as its second argument. I am assuming that be aware of is a propositional attitude predicate even when subcategorizing for an NP but I will not be concerned here with how the content of 4> is construed. 50 The interpretation of (122a] as given in (122c]

Functional Reading of Bare Plurals

105

coincides with that of a sentnce with an extensional universal quantifier. On this approach then, nothing special needs to be said about the operator itself or the bare plural. Any special properties the operator might have ought to follow from the modal dimension it is associated with and therefore, ultimately, from certain facts about the context of utterance which determines this kind of modal dimension. In what follows, I will consider to what extent this can be maintained and at what cost. Since, as argued in section 2.4, in exactly the same contexts that give rise to the functional reading of a bare plural, a singular indefinite has only an existential reading, the generic operator with a trivial modal dimension must somehow be prohibited from binding a variable contributed by a singular indefinite. The kind of explanation forced upon us is to say that the singular indefinite does not have the functional reading because it is incompatible with an extensionalized generic operator. We can assume that the singular indefinite selects for the kinds of modal bases and ordering sources that the operator can be construed with in such a way that it excludes the combination of a totally realistic modal base and a trivial ordering source. An analysis capitalizing on the context dependency of the modal dimension of the implicit generic operator raises the question of whether we can in general use the implicit generic operator with a totally realistic modal base and trivial ordering source to make non-generic universal statements. For instance, why isn't the modal dimension determined to be trivial in cases like [58], [61], or [86]? We have to spell out the conditions under which a trivial modal dimension is selected and their connection to contextual restrictions. Does the trivial modal dimension correlate with the interpretation of the predicate? One could claim that the modal dimension of the generic operator is trivial because of the particularity of the propositional argument of the predicate, which has the result that the predicate is not interpreted like a characteristic property. How likely would it be that students in general, actual and potential, would be aware of this particular fact or the actual danger? However, the particularity in the nature of the predicate is, in general, neither a sufficient nor a necessary condition for the functional reading to arise. That it is not sufficient is shown by examples with an overt generic operator, such as the one associated with used to. In [123b]

106

Descriptions in Context

and (124b], as opposed to (123a], the argument of the predicate is a particular fact. However, the operator is not restricted so as to pick out only the actual students on campus or a contextually restricted set of financial wizards, hence the oddness of [123b] and [124b]. In [124c], on the other hand, where the bare plural exhibits the functional reading, and, according to the proposal under discussion, there must be an implicit operator involved, the operator is construed with a trivial modal dimension and quantifies over a contextually restricted set of financial wizards. [123]

a. Students used to be aware of that kind of danger. b. ??Students used to be aware of the danger.

[124]

a. Yesterday's fundraising event solved the university's fiscal problems. b. ??Financial wizards used to be aware that yesterday's fundraising event would bring in millions of dollars. c. Financial wizards were aware that yesterday's fundraising event would bring in millions of dollars.

The conclusion from this contrast is that not all generic operators can be construed with a trivial modal dimension, even when the interpretation of the predicate requires it. That the particularity of the predicate is not necessary is shown by examples where the predicate is given an argument which does not constitute a particular fact. In that case the functional reading of the bare plural is still present, although the modal dimension need not be trivial. For example, (125a] is ambiguous between a truly generic and a functional reading (in the right context}, and the bare plural in [125b] has the same reading as in [55b]. [125]

a. Students are aware of dangers of this kind. b. Students were aware of dangers of this kind.

This shows that the functional reading is compatible with an interpretation of the predicate as a characteristic property, 51 and that therefore a generic operator is not determined to be degenerate simply on the basis of the interpretation of the predicate, when no other construal of genericity would do. 52 Therefore, it remains an open question exactly under what conditions the modal dimension is determined to be trivial.

Functional Reading of Bare Plurals

3. 2

107

Degenerate Genericity and the Functional Reading

Since I identify the functional reading with degenerate genericity, I must reexamine the properties distinguishing between the functional and generic readings discussed in section 2.2.1, namely: the presence of contextual restrictions, the implication of existence, and the scopal interaction between the bare plural and overt operators. The presence of contextual restrictions must be connected to the nature of the modal dimension. The implication of existence is easily seen to be compatible with the view of the functional reading as a case of degenerate genericity. Scopal interaction poses some more challenging problems for the operator analysis, which lie not with the fact that genericity is degenerate but rather with the presence of an operator. These three points will be taken up in order. 3.2.1

Implicit Contextual Restrictions

Why are implicit contextual restrictions associated with the functional reading if generic operators do not accept implicit contextual restrictions? Can we relate the presence of contextual restrictions to the triviality of the modal dimension? In order to understand why degenerate generic generalizations can be contextually restricted, we must first understand why regular generic generalizations cannot be contextually restricted. I will suggest an account of the impossibility of contextual restrictions with generic operators in this section. This account relates the possibility of contextual restrictions in the case of the functional reading with the triviality of the modal dimension of the operator. However, it is independent of the analysis of the functional reading as involving an operator, so it can stand even if the degenerate genericity account of the functional reading is ultimately to be dismissed. There are two alternative ways of construing contextual restrictions that I will concentrate on here. Up to now, I have construed them as additional conditions on the descriptive content of the NP, as exemplified in [122a]. But we can also relativize the interpretation of the NP to a contextually given set of entities, following proposals by Westerstahl (1984), van Deemter (1991, 1992) and En·cion

as

and

sometimes

is

commonly

it

understood in Artificial Intelligence and model theoretic semantics as formulated most explicitly for natural language by Richard Montague. The most common methodological assumption for those working in model theoretic,

or denotational semantics,

is that semantic representations

may be dispensed with entirely in favor of rules of interpretation in a model.

Quite naturally, then, those working in this tradition have had thus need

little to say about the internal semantics of lexical items.

to make a case that these two traditions can be blended in a coherent way. I will begin by arguing for lexical

representations on grounds

which are far removed from the central concerns of this dissertation. The

heart

of

the

argument

is

that

any

theory

of

discourse

understanding will need to be anchored in a fairly rich system of lexical representations.

This

is

a somewhat peculiar place from which to

launch the enterprise, because almost nothing I have to say in the rest of

this

dissertation

understanding.

will

have

any

direct

The need for the appeal,

bearing

on

discourse

I think, boils down to the

9

matter of balancing profit versus effort. Many

of the

following

pages

will

be

taken

up

with

discussions of matters relating to very small lexical families. the statements made will apply only to one verb. three or four verbs and one preposition sense of that

detailed Some of

Some will apply to

one preposition in a special

In Chapters 1 and 2, I will set up a broad

prepo~.ition.

framework for lexical descriptions, taking as my focus an account of the selection of prepositions by verbs.

In chapters 3, 4, and 5 -

which

form the linguistic heart of the dissertation -

I will discuss some of

the grammatical payoffs of such representations.

It should be conceded

at

the

outset

independently

that of

much

lexical

grammatical

semantic

inquiry

questions.

can Those

proceed

quite

interested

in

pursuing the cross linguistic behaviors of long distance dependencies, bound anaphora,

coreference,

cliticization,

and

agreement will

most

likely find questions of which particular verb has been used peripheral to their concerns.

I would even go so far as to venture that the

grammatical payoffs alone are too small to justify the enormous research effort of constructing detailed lexical point of this dissertation

representations.

The starting

is to assume on independent grounds that

lexical representations are necessary.

Once that is given, there ar·e

numerous questions of interest both to the grammarian and the artificial intelligence researcher. with

the

grammar?

How do such lexical representations connect up How

do they

contribute

to the

semantics

of

10

sentences?

What

consequence

grammar as a whole?

Can

does

their

existence

have

for

the

some things that used to be stated as

"grammar" simply fall out of the lexical structure?

Chapters 1 and 2

will deal with the first two questions, chapters 3,4, and 5 with the last two. wish plausibility grammar,

to

do

two

argument and

for

second,

things lexical argue

in

this

section:

representations that

such

first,

pr·esent a

independently

representations

can

of be

coherently treated in a denotation a I (model theoretic) semantics. In dealing with word meanings we are dealing with that part of language which is flush with the physics and the metaphysics of the world.

If knowing the word horse means

knowing

horses,

then the

same can be said of run and running situations, and in and situations of containment.

We seem to commit ourselves ourselves to a potentially

endless enterprise.

Clearly we need some formal definition of the object

of description; and just as clearly that formal definition cannot come from within grammar.

What we are interested in here is some approach

to viewing the relationship of language with the world; presumably the domain of grammar is purely language internal relationships.

We thus

need to begin by defining some larger enterprise outside of grammar within which lexical description can proceed. This larger enterprise is already a going concern,

and it has

11

been very clearly articulated by Charles Fillmore:

1

(1) I believe that the linguist, in his consideration of a number of issues in semantic theory, can profit from the exercise of examining these issues within a larger view of language production and language comprehension .... [T] he task I am taking on is that of locating a concern with meaning within a larger theory of language processing; the aspect that I will concentrate on here is text comprehension." Suppose, Surely

then, such

that a theory of text comprehension were the goal. a

extra-grammatical

theory

would

structures,

with

need text

to

concern

conventions,

itself

with

story frames,

character r·oles, models of reader and writer, with knowledge of the world and structured events in the world; but it will also need at some point to return to the chunks of text, to the language tokens and the language types they instantiate, and at some point it will need to have access to "meanings," and those meanings will finally bottom out at words.

It will be crucial, in any theory of text understanding, to have

some understanding of how words structure the world, and just what pieces of it they evoke.

A theory of lexical semantics that has some

promise of being of more interest than a dictionary is a theory that can contribute to an understanding of text comprehension.

Indeed, the

view of lexical meaning set forth in the article just cited, in which word meanings are represented with structured templates called scenes, one that can be summarized: words are very short texts.

1.

Fillmore 1977b, p. 77

is

12

It is a very short step to recast things in a totally language independent fashion.

Suppose we call the object of study discourse

understanding,

the

using

term

discourse

to

mean

any

interaction,

however brief or extended, governed by social conventions.

The best

term

probably

for

Saussure"s

the

sorts

term

of

sign.

soc1al

convention

A discourse

in

of

the

interest broadest

is

sense

1s

an

interaction involving those social conventions w:1i.:.l• have meaning. We see a man walk up to a door and take out a key.

That we can in some

sense be said to "understand" this event is shown by the fact that we have

expectations

about

it.

The

same

knowledge

we

use

in

understanding the event will play a 1·ole in our understanding of a text about it. The question the linguist asks sometimes characterized: what do we know when we know a language? of

a

student

of

discourse

might

The somewhat different question

be:

understand the discourse of a culture? of sounding impossibly large, impossible to exclude anything.

what

do

we

know

when

we

This second question has a way

a domain

of

study

from

which

it

is

At the very least it is clear that any

representation of such knowledge will be far more complex than current representations of grammar. The "knowledge" at issue here is sometimes referred to as the encyclopedia, sometimes, by linguists, as real world knowledge, usually to

distinguish

it

from

the

linguist's

business,

which

is

linguistic

13

knowledge.

At this point I would like to introduce a term current in

Artificial Intelligence circles, which wi II be in force for the remainder of this dissertation.

The term is ''knowledge representation;"

representation

AI

in

knowledge

is a field quite separate from natural language

processing (though of course often linked to it in actual systems); it is generally thought of as just one side of a coin whose other half is reasoning.

systems work l> Pl: Pl> V P2: (P2 V)> The format of the rules

here is taken from Gazdar 198lb, with the

material before the colon intended as a syntactic statement, material

after

as

a

semantic

statement.

In

GPSG,

every

and the phrase

structure rule must be accompanied by translation rule expressing how the

translations

of

the

daughters are to be combined to yield the

translations of the mother.

In the example rules given,

the actual

51

lexical predicate argument structure is concealed in the entry for for and is that of choice (b).

If we insist on heads as functors,

syntactic rules, we need only complicate the translation of wish (LAMBDA PP (PP WISH)).

in our to be

A decade of lambda conversion has taught

us that the notions of function and argument are fixed only when there is some

ind~!pendent

theory of the denotations of the objects being

combined. The moral to be drawn from lambda shuffling is u nderdetermi nation of our logical

representations.

one about the

We need

whatever

help we can get when confronting the wealth of choices they offer. Thus, a theory that constrains the logical type of non-lexical syntactic categories,

like that of Klein and Sag 1982,

is welcome because it

immediately rules out choice (b) above, which forces P2 to assume a disallowed type.

1

The specific choices of that theory, however, are

problematic for me, because the type it DOES choose for P2"s is that of N2. And the assumption I have made is that prepositions contribute to meaning. I would like to investigate an alternative kind of constraint on our

choices:

suppose

instead

of

constraining

nonlexical

syntactic

1. The idea of constraining syntactic categories was not first suggested in Klein and Sag 1982. It is fully explicit in Montague 1970, though his notion of a syntactic category was somewhat different than the one linguists are used to.

52 categor·ies we tried to work things bottom up. lexical items are where semantics starts,

Acquisitionally speaking,

so there is some potential

payoff in stipulating that their logical forms, at least, be simple (as we shall see later in the chapter, the conceptual structures behind those logical forms can still be ar·bitrarily complex).

I have in mind letting

prepositions be of type eai We also need a new truth definition for such three-place preposition predicates,

differing

(transitive verbs,

from

the one for

for example)

when the predicate is defined.

three-place

head

predicates

in that it lacks a clause stipulating

149

1. 9

Conclusion In this chapter I have proposed a basic framework in which to

undertake lexical description.

I have argued that the ultimate goal of

constructing

lexical

a

full

set

of

representations

is

that

lexical

representations are an integral part of a full-blown representation of cultural

categories,

representation"

pose that scenes in general may be said to take a point of view on events.

That is, they are based on certain regularities in the

flow of things,

and

sometimes they will be underinformed about the

event-types they describe.

The BUYING scene will not directly encode

the information that the scene it describes always co-occurs with a SELLING scene.

This will be the job of a special kind of scene called a

ground scene, which coordinates the views of one or more scenes into a single new scene that is neutral as to point of view.

Looking ahead a

bit, we will say the BUYING scene views the buyer as actor, and the SELLING scene views the seller as actor; the the ground scene relating these two scenes will be the COMMERCIAL_EVENT scene proper.

In

that scene, both buyer and seller will be actors. 2.4.1

Are there COMMERCIAL_EVENTS? In the last section I was unhappy with my

representations for

buy, sell, and pay because the complexity of the COMMERCIALEVENT scene

licensed

some

implausible

semantic

compatibilites.

want

to

consider two stronger examples of this before proposing an alternative

188

account that seems to offer a solution. verbs cost and spend. these

verbs

takes

The cases in point are the

The first thing to notice is that neither of

either

to

or

from,

so

that

considerations

of

compatibility would not bring us to represent either with a POS_ TRANS scene.

More striking, as Fillmore 1977b points out, neither allows the

seller to be realized in ANY form. in the

~C:tnre

That is, the seller may not appear

clause; to bring this participant into the picture at all,some

intervening predication is required: (25) That old grammophone cost Johann 2K, installments to his uncle in Jersey.

payed

in twenty

On the one hand, one might look at this as par for the course.

\Ve

have all along talked about facts of compatibility that are not rules; as such we are allowed to claim that lexical idiosyncrasy has once again intervened, and that it is just an arbitrary fact about a pair of highly arbitrary verbs, that they do not provide any means for realizing an expected participant. But to say that is to put this fact about spend and cost on a par· with the facts about idiosyncrasies of preposition selection that we saw before. these

Beside decide on we have opt for verb-preposition

pairs

express

roughly

despite the fact that the

same

meanings;

whatever difference there may be in meaning is not likely to be one that we can ascribe precisely to a difference in the relevant meanings of on and for.

Thus, the selection of on with decide is in some sense

189

arbitrary given the meaning of decide. with cost is a very different

But the impossibility of seller

kind of arbitraryness.

Somehow the

choice of this lexical item reflects a choice to leave the seller out of the accounting.

Indeed,

it may even describe a situation

in which the

seller, in any conventional sense, is absent. (26) Parking on High Street cost me twenty dollars. This fact about cost is indeed an arbitrary fact, but it is an arbitr·ary fact about its meaning; it is a fact which ought to be reflected in our choice of among

lexical

the

complementation seems

to

me,

semantic

"syntactic

representation, afterthoughts··

possibilities. which

rules

This out

a

instead of being that

mark

the

included actual

is a conclusive consideration, syntactic

recourse,

even

it

without

entering into the technical difficulties of devising a form of syntactic marking that would rule out the realization of some core participant, in any ·of its possible forms. I want to tackle this problem by first considering in some detail what I consider to be the wrong solution, which is to attempt to extend the

same

techniques

we

have

used

till

now.

Pointing

out

the

inadequacies of this course will go a long way towards motivating the special status lexical representations will have in the "right'' account. Suppose we try the following solution: cost and spend distinguished from the other commercial event verbs

will be

by having them

190

reference a situation different from the

COM~IERCIAL_EVENT

situation.

The outlines of a fairly plausible account are clear.

We could call the

new

have

scene

participants,

a

MONEY-TRANSACTION;

buyer,

intuitions of lexical

money,

relatedness

scene we have

already

tntroduces

Pd•"'ticipant

the

and

would not reference the

goods.

by

presented

seller.

it

would We

saying the is

AKO

this

could

only

three

preserve

COM~IERCIAL_EVENTS

scene,

and

that

The MONEY-TRANSACTION

POS_ TRANS

scene,

our

and we will

it

scene

not expect

verbs like pay and spend to be compatible with to and from. So far nothing we have said addresses the problem that troubled us in the last section, namely that semantic compatibilities. done

will

solve

that

buy and sell exhibited too many

A simple extension of what we have already problem,too.

Suppose

that

we

split

out·

COMMERCIAL_EVENT scene into its two component POS_ TRANS scenes, and made each scene.

of these

separately

AKO

The AKO links would look like this:

the

MONEY-TRANSACTION

191

MONEY- TRANSACTION POS_ TRANS

("Cost"

"Spend")

POS_ TRANS

MONEY-POS_ TRANS

GOODS- POS_ TRANS ACTION

("Pay")

ACTION

BUYER-ACTION ("Buy")

SELLER-ACTION ("Sell")

A drawback to this scheme is that it predicts that buy

I

sell

I

and pay

are semantically compatible with all the same complements as cost and

spend.

The most obvious problem with that claim is the double object

construction. Although cost takes a double object construction it does so with the buyer and money following the verb.

Neither sell nor buy

allows that configuration, so that their double object construction must be semantically founded

cost. with

on a different frame structure than that of

As the framework stands, a lexical frame is not just compatible a particular subcategorization;

it is compatible with a particular

assignment of roles to the complements in that subcategorization. non-occurence

of

buy

and

construction could, of course,

sell

with

the

cost-type

double-object

be called a syntactic accident,

more plausible representation would be:

The

but a

192

MONEY-TRANSACTION

?

?

COST ("Cost")

SPEND ("Spend")

POS_ TRANS

POS_ TRANS

GOO:)S-POS_ TRANS

MONEY-POS_TRANS

ACTION

("Pay")

ACTION

BUYER-ACTION ("Buy")

SELLER-ACTION ("Sell")

In the alternative sketched here, a MONEY-TRANSACTION is still the root of the COMMERCIAL_EVENT "tree'', scenes

are

not

inherited

by

the

other

but the verbs.

COST

The

and

SPEND

question

marks

represent other scenes that distinguish COST and SPEND from a simple MONEY-TRANSACTION, thus allowing cost and spend to be semantically compatible with complements that the other verbs do not take. face

of

it,

compatibility

then,

we

simply

COMMERCIAL_EVENTs.

have

solved

all

by

com pi icati ng

our problems our

"sort

about

On the semantic

tree"

for

The COMMERCIAL_EVENT scene as such with

its four participants has ceased to exist; but the same information has been distributed out over separate scenes. I now want to argue that this solution is inadequate, chiefly on the basis of entailments among some of the commercial event verbs.

193

This

point

can

be

made

most

clear·ly

MONEY-TRANSACTION scene and ask just what encode.

if

we

look

information

at it

our must

It must satisfy the two needs of all our representations; it

must adequately describe a

certain

type of

situation,

and

it

must

adequately represent the semantics shared by a group of verbs. A

general

begins to emerge.

pattern

for

making

statements

about

compatibility

A verb is related to a class of verbs by being AKO

the abstract object that unites the the class; its differences from the other members of the class, if any, can be represented either by the idiosyncrasies of its particular lexical frame or by making it AKO some new object or objects, No constraint is placed on the nature of this new object.

So,

in principle a MONEY- TRANSACTION verb could be AKO

INTENTIONAL ATTITUDE, and we could have a lexical item specifically about intentional attitudes connected with sales. We now come to the crucial question. group of verbs

We have been examining a

Fillmore has dubbed Commercial Event

verbs.

The

question is: should there BE any constraint on what sort of object a commercial event verb can be?

The intuitive answer is yes: at least if

we return to our naive pre-theoretic notion of what a commercial event is. Why should there be a verb buy that presents a situation in which the buyer is actor, and another verb sell which presents the seller as actor.

Because, intuitively, in a real life commerical event, both the

buyer and seller act.

What the set of representations above lacks is a

194

single description of what a real-life commercial event is like. Instead the description has been scattered over a number of different frames. If

we

believe

that

the

commercial

structure that ought to be a then,

event

is

an

important

kind

of

unit in our knowledge representation,

in our eagerness to capture the lexical facts, we have failed to

capture a representdtional fact. The scheme offered above makes lexical scenes highly specified, with

the

MONEY-TRANSACTION

scene

an

underspecifed

compatible with with all the COMMERCIAL_EVENT verbs.

object

It is thus not

particularly clear what it actually represents, apart from an abstraction of what a group of verbs

has

in

common.

COMMERCIAL_EVENT scene as such is really backwards.

The unifying

If we believe that the

an important structure, this is

scene of a

lexical class should be

highly specified, and particular verbs should pick aspects of that scene to highlight.

They should be underspecified versions of that scene. So

SELLING is a POS_ TRANS scene with the seller as actor because it is a COMMERCIAL_EVENT verb,

and a COMI\1ERCIAL_EVENT

that kind of POS_ TRANS scene. not

represent

an

INTENTIONAL

involves just

But a COMMERCIAL_EVENT verb can ATTITUDE

because

there

is

no

intentional attitude involved in a COMMERCIAL_EVENT. The important claim here is that a COMMERCIAL_EVENT should be a scene in which both buyer and seller are actor.

That is, we should

modify our COMMERCIAL_EVENT scene to look like this:

195

(COMMERCIAL_EVENT: SIMULTANEOUS SCENES ( Fl RST _SCENE: COt\IPONENT (ACTION_POS_ TRANS (ACTOR_DONOR ) (ACTOR_RECI PI ENT ) (PATIENT))) (SECOND_SCENE: COMPONENT (POS_ TRANS lDONOR ) (RECIPIENT) (PATIENT )))) Here

ACTION_POS_ TRANS

is

a scene del'lveJ by the following AKO

chain: POS_ TRANS

ACTION

RECIPIENT _ACTION

ACTION

DONOR_ACTION

ACTION_POS_ TRANS RECIPIENT_ACTION is a POS_TRANS scene in which the recipient is an

actor,

symbolized

with

a

new

slot

name,

ACTOR_RECIPIENT;

DONOR_ACTION a POS_TRANS scene in which the donor is an actor, given

the

slot-name

ACTOR_DONOR.

An

ACTION_POS_ TRANS

will

have two ACTORS. The nature of the ACTION-POS-TRANS frame deserves a bit of comment, since it is a frame which must inherit an actor slot from two different parents, one slot identified with donor,

one with

recipient.

Let us say that when a frame inherits the same slot from two different sources,

there are two possible things that can

happen.

First,

the

196

constraints

on the slot are merged and the resulting slot is one for

which all defined values must meet both

sets of constraints.

followed this course for ACTION_POS_ TRANS

we would get one actor

slot whose value would have to be both donor and recipient. words,

we

would

have described

something to himself.

Obviously,

If we

In other

a scene in which some actor gives this

is not quite what

we

had

in

mind. Let us define another· possibility as follows: when a frame inherits the same slot from two different sour·ces, we may chose to interpret the resulting slot as taking an ordered-pair as its value; each member of the ordered-pair satisfies a separate inherited set of constraints on the slot.

This is what we want in the case of ACTION_POS_ TRANS. (DONOR_ACTION (AKO (POS_ TRANS)(ACTION)) (PATIENT) (RECIPIENT) (ACTOR_DONOR )) (RECIPIENT _ACTION (AKO (POS_ TRANS)(ACTION)) (DONOR) (PATIENT) (ACTOR_RECIPIENT )) (ACTION_POS_ TRANS (AKO (DONOR_ACTION) (RECIPIENT _ACTION)) (PATIENT) (ACTOR1 ) (ACTOR2 ) (ACTOR_DONOR ) (ACTOR_RECIPIENT ))

The ACTION_POS_ TRANS frame illustrates both of the options possible

197

when a slot has been inherited from two sources.

constraints have been simply been

donor, recipient, patient slots, the merged

the

Since

requirements on any value.

interpreted as simultaneous

and

from

constraints

In the case of the

both

are

sources

the

same

(neither

DONOR_ACTION nor RECIPIENT _ACTION adds anything to those slots except the specifications donor _actc•r and recipient_actor, respectively), Let us call this option the dd,

, SLOTl>.

The elaboration sequence SLOTl

in

FRAt-vlEA

1s

an

The elaboration sequence is «S LOT2

SLOT4 is a direct elaboration of SLOTl in FRAI\1EA,

and is therefor·e an indirect elaboration of SLOT3.

The sequence, not

suprisingly, is «SLOT3 SLOT2>, , SLOTl, SLOT4>. We are now

ready to define the way in which a lexical frame

references the "structure" in a reference frame. (1)

Rule of Referenced Structure

204

( 1) Suppose we have a lexical template scene T and a lexical scene L that references a scene R. Reference is always defined with respect to a path P in R; the value of P is constrained to be some frame That there will always K such that K is AKO T. exist such a path in R follows from the fact that R must be an extension of T, provided that we handle the trivial cases as follows: We may think of a frame as "constrained to be" AKO itself, and we may think of the nullpath in any frame as giving that frame. Thus, if L references R with the null path in R (wr·itten ""), R must be AKO T. K that ar·e be slots in (2) There will elaborations of slots in T (since K is AKO T). Furthermore, there may be other slots in R that are elaborations of those K slots, and thus elaborations of slots in T. We now construct all pairs , such that Sl is a slot in T and S2 is an elaboration of that slot in R, and such that the first member of the elaboration sequence linking Sl and S2 begins with the path

(all the elaborations of T slots in R whose elaboration sequences are rooted in K). Then, for each such the Rule of Referenced the Structure makes the following stipulation: lexical scene L will have S2 as a slot and Sl will be constrained to equal S2: (Sl )

All of this preparation is so that we can talk about a lexical scene accessing only some of the structure in a reference scene.

When a

lexical scene acquires a slot because of the rule of reference, we will speak of it as

"referencing" that slot.

The introduction of the notion elaboration sequence into the Rule of

Referenced Structure is to allow for the possibility that a slot in R

is an elaboration of the same slot in L in more than one way.

As we

205

see

shall

COM~IERC

is

this

shortly,

I AL_EVENTS,

All of this is best illustrated by example. trivial

of

case

the

in

precisely

crucial

example

a

as

preparation

for

of

verbs.

event

commercial

the verbs

Suppose that the template scene for

Let us start with a

longing

like wish

is

PURSUIT-LONGING, and that the lexical scene for wish in particular is WISHING, with a reference scene, WISHING-R. (PURSUIT-LONGING (AKO (PURSUIT) (LONGING) (PURSUER (=longer)) (PURSUIT (:frame)) (LONGER) (KARr-.IAN)) (\VIS HI ~G (AKO (PURSUIT-LONGING)) (REFERENCE (PATH ) (WISHING-R))) Then

WISHING-R

be

must

AKO

PURSUIT-LONGING

as

by

well,

our

trivial case in the Rule of Referenced-Structure. And we have: (WISHING-R (AKO (PURSUIT-LONGING))) The path is "''

This is a trivial case.

Since a slot is an elaboration of itself, all the slots

the template scene. in

WISHING-R

the reference scene equals

(that

PURSUIT-LONGING), Referenced-Structure.

are

is,

all

added But

the to

slots

WISHING

WISHING,

which

it

inherits

by

the is

Rule also

from of AKO

PURSUIT-LONGING, already has those slots, and we already know for

206

any

slot that its value is equal to itself.

acquires no new meaningful structure. even

in a trivial case like this,

lexical scene from reference

However, this is not to say that

ther-e

scene.

So here the lexical frame

is

no

point

in

distinguishing

WISHING and WISHING-R are still

two different frames, and they can carry different information.

If we

need it, WISHING-R stands as a record of which structural facts about situation wish.

have not been built

into the

lexical

semantics

I do not insist that there be any such facts.

the ground scene may just be a formal device with uniformity. Now

let

us

turn

to the We

vastly

will

Suppose the template scene for is,

the

verb

In some cases,

no advantage but

1

COMr-.1ERC1AL_EVENTS.

That

of

suppose

the

more

use the BUYING

BUYING

scene

complicated

verb

buy

is AKO is

as

business our

of

example.

RECIPIENT_ACTION.

AKO

RECIPIENT_ACTION;

recall that RECIPIENT _ACTION is the POS_TRANS scene in which the recipient

is

BUYING.

actor.

Intuitively,

being AKO

the donor with from,

the

RECIPIENT _ACTION

but not to;

as manipulating any instrument. hardly

is

template

we

want

for

We want the lexical scene to determine semantic compatibility

requirements;

is

this

soecific

enough

will

license

marking

of

and the recipient will be interpreted Yet the notion of RECIPIENT _ACTION

to

characterize

BUYING.

Through

1. Another possibility is that in such trivial cases the ground scene equals the template scene.

207

referencing the COMI\1ERCIAL _EVENT frame, we will build into BUYING all

the

necessary

specifications.

To

begin

with,

here

is

the

COMMERCIAL_ EVENT scene once again: (COMI\1ERCIA L_EVENT: SIMULTANEOU S SCENES (FIRST _SCENE: COMPONENT (ACTION_POS _ TRANS (ACTOR_DONO R ) (ACTOR_RECI PIENT ) (PATIENT))) (SECOND_SCE NE:COMPONEN T (POS_ TRANS (DONOR ) (RECIPIENT) (PATIENT)))) Here is BUYING: (BUYING (AKO (ACTOR_REC I PI ENT)) (REFERENCE (PATH ..:FIRST_SCEN E>) (COMMERC IAL_EVENT)) (ACTOR_REC I PI ENT) (DONOR) (PATIENT) (ACTOR )) It is in this case that the notion of reference relative to a path will do some work.

Here are the major frames

and

slots

involved,

together with the variables they instantiate in the Rule of ReferencedStructure: BUYING = L COMMERCIAL _EVENT = R RECIPIENT_A CTION = T = P ACTION_POS_ TRANS = K

208

Note that K is AKO T, as required in the rule.

Here, then, are the

pairs , and the elaboration sequences that justify them.

Note:

I have left out any vacuous sequences where BUYING would inherit a slot it already had: Pair

Sequence



«FIRST _SCENE ACTOR_RECI PI lENT>, BUYER>

«FIRST_SCENE PATIENT>, GOODS>·

Note that all the elaboration sequences begin with a path of the form , where "slot" is the slot whose elaboration is being traced.

This is the

in the definition.

these sequences in a particular location in crucial.

Note also that "rooting'' COMMERCIAL_EVENT was

Although the template scene has only one patient slot,

reference scene has two,

and they have different elaborations.

the

If we

had not specified which elaborations of each slot in the template scene were

allowed,

both

a

money slot and

goods slot would have been

declared in BUYING, and both would have been constrained to be equal to "patient''. Similarly, there would be two donors and two recipients, and it would have been quite unclear who was giving what to whom. Here,

then,

is BUYING "before" and "after" application

Rule of Referenced-Structure. (BUYING (AKO (ACTOR_RECI PI ENT)) (REFERENCE (COMMERCIAL_EVENT))

of the

209

(ACTOR_REC I PI ENT ) (DONOR) (PATIENT) (ACTOR )) Reference In herita nee

====================>

(BUYING (AKO (ACTOR_RECI PI ENT)) (REFERENCE (COM~IERC IAL_EVENT)) (ACTOR_RECIPIENT ) lDOi'iCJF. ) (PATIENT )) Note that BUYING has lost its money slot under the new analysis. the status of money and the provenance of the preposition for will be treated in the next section. Also

dealt

with

next

the

in

section

some facts

be

will

about

argument structure, noteably the inability of the verb spend to realize a

buyer.

The

entailment

COMMERCIAL_EVENT

(at

facts

least when

suggest used

spend

that with

the

references

preposition

for);

what we want to do is arrange for· it to do so in such a way that it referenceially inherits only three slots, and the seller slot is not among

cost,

them.

reference the

because

it

does

not

COMMERCIAL_EVENT,

entail but

a true one

that

purchase, is

more

will

not

general,

which will not include a seller slot. Since our primary purpose in certain

kind

of

situation

takes

referencing is to establish that a place,

for

purposes

entailments, we will want at least the following property:

of

correct

210

(1)

Rule Of Entailment

(2)

If a scene A references a scene B then for every instance of A, there is an instance of B such that the filler of every slot in A that is an elaboration of a slot in B is the filler of the corresponding slot in B.

That is to say every for every of

the

BUYING

scene)

CO~H.1ERCIAL_EVENT

there

BUYING situation (every instantiation is

also

an

instantiation

of

the

scene such t,at the buyer, seller, money,

and

goods are the same.

To guarantee that we will need to put a rider on our Rule of Referenced-Structure.

To each

slot countenanced

by

the

Rule of

Reference we add "(metonymy 'reference)." (52 (metonymy 'reference)) So for example buyer in BUYING (BUYER

will be:

( = actor _recipient) (metonymy 'reference))

The METONYMY procedure on buyer says that value of the buyer slot for any BUYING frame must be the same as the value of buyer slot in the frame that is the value of the COMMERCIAL_EVENT frame).

reference slot (that is,

in some

As always, we interpret the appearance

of a scene A in the slot of another scene as a constraint on possible values for that

slot.

All such values must be AKO A.

all the other slots referenced by BUYING.

Similarly, for

Intuitively, we can think of

the METONYMY procedures on all the slots

as

mapping

a

BUYING

211

"subsituation" onto the total

COt\1MERCIALEVENT situation it is part

of. To

get

entailments

COMf\IERCIAL_EVENT to they point to it.

between

lexical

items.

we

will

need

point to its lexicalized subsituations,

a

just as

Thus far, we have guaranteed that for every BUYING

situaticn there is a COt\1MERCIAL_EVE.NT situation with a superset of the BUYING situation's participants.

To get an entailment from buy to

sell, we also need a pointer from a COMMERCIAL_EVENT to SELLING

situations,

guranteeing that for ever·y C0Mt\1ERCIAL_EVENT, there is a

SELLING

situtaion

participants. that

with

In fact,

reference

a

subset

we will

of

the

COMf\1ERCIAL_EVENT's

need pointers to all the lexical scenes

COMt\1ERCIAL_EVENT,

so

that

this

considered part of the rule of refer·enced structure. declaring

the

reference

slots

in

the

lexical

may

as

well

be

Simultaneous with

scene,

we

declare

a

subsituation slot in the reference scene, of the form: ( subsituation: COMPONENT (subsituation)) Thus,

for

example,

we

declare

in

COMMERCIAL_EVENT

a component

slot, called buying, constrained to take

BUYING scenes as its value.

The

means

fact

accessed slots.

that by

it

is

a

metonymy,

component which

slot

guarantees

the

that

all

correct

its lining

Let us add a convention like the convention that

scene does not have its AKO slot accessed: reference

slot accessed.

slots

are

up of

a component

neither does it have its

With all this apparatus granted,

then,

for

212

every BUYING scene, there is a COMr-.1ERCIAL_EVENT scene with the same buyer, seller, scene,

and goods.

And for that COMMERCIAL_EVENT with the same buyer, seller, and

there rs a SELLING scene

goods. Thus a BUYING scene entails a SELLING scene,

just as we

wanted. The

La•.v of Entailment

Referenced Structure.

can thus b£ made part of the Law of

Most importantly,

stipulation on the frame representation,

it is

not left as

an extra

but is built into the defining

procedur·es of the frames associated with a reference slot. The somewhat.

role

of

compatibility

in

this

revised

and prepositions will both have figure scenes. scene

compatible

will

change

Our definition of compatibility will remain as it was, but will

only apply to the templates of figure scenes.

figure

system

is

with

an

extension

everything

a

of

its

template

This is because verbs While we know that a

template, scene

and

was

will

thus

compatible

be

with,

nothing in our definition of compatibility guarantees tnat figure scenes that are extensions of compatible template scenes will be compatible with one another.

Thus, in the next section we will develop a figure scene

for for in one meaning, and a new figure scene for buy; those two figure scenes will not in

fact be compatible, but their templates will.

The sortal constraints on linguistic predicates will thus be template scenes.

It is the template scenes

that

actually

defined by impose

a

semantic structure on the lexicon, since they define the properties of

213

linguistic predicates that determine other predicates. they

possibilities of combination with

If two lexical items share the same template scene,

are semantically compatible with all the same complements.

This

amounts to saying that the template scenes taken together constitute a significant level of linguistic representation, which we can call lexical structure. We can close this section with a terminological proposal.

We nave

of frames: one kind has a

essentially divided the world into two kinds

slot called reference; it takes a particular ''view" of the scene in its reference slot defined by the structure it "acquires" from that scene. The other kind of

Call scenes with reference slots figure scenes.

scene appears in the reference slot of a figure scene; let us ca II it a ground scene. scenes.

A gr·ound scene represents a generalization over figure

It collects the facts about a certain

some sense underlies all the figure scenes.

situation-type which in

It is tempting to say that

the ground scene represents the real world scene on which the figure scenes take different views, but to put it this way is to give the two kinds of scenes a different ontological status.

I would rather say that

the ground scene represents a recognition that the figure scenes are situationally-related, category.

1.

and

as

such

defines

a

new

more

1

The terms figure and ground were suggested by Fillmore.

complex

215

2.4.2

Some technical revisions and stipulations Our current definition of extension plays two independent roles in

the

framework.

It

directs

compatibility

traffic,

and

crucially

it

constrains the Rule of Referenced Structure used to define figure and ground scenes. In this last capacity,

it is a little too

restrictive.

As

things

stand now, a fr·ame A is an extension of another frame B if it fits one of three clauses in the following recursive definition: (1) A is AKO B.

(2) B is a component of A. (3) A is an extension of C and C is an extension of B. This leads us into an undesirable situation with respect to figure and ground scenes, which can be illustrated with the following mock frames: (FRAMEA (SLOTl)) (FRAI\1EB: FRAMEA (SLOT:!) (SLOT3)) (FRAMEC: FRAMEA (SLOTl) (SLOT:!)) (GF: FRAMEC (SLOTl) (SLOT:!) (SLOT3)) As things stand,

GF is an extension of FRAMEA and FRAMEC, but not

216

of FRAMES.

Nevertheless, we may wish to think of SLOT3 in FRAr-..1EB

as being the same slot as SLOT3 in GF.

This would make a difference

if FRAMES were the template of some verb scene V, not be a

legal ground

scene

for

V,

even

because GF could

though,

structure contains any structure defined in FRAMES.

intuitively,

its

One way to make

GF an appropriate ground scene is to simply stipulate that it is also Af\0

FRM1EB.

GF then

inherits SLOTl

from two differ·ent places.

Suppose that

no new information is added to SLOTl in either FRAMES

or FRAMEC.

The generalization about SLOTl is captured in FRAr-..1EA;

but we ar·e losing it again in GF by having it inherited twice. We can avoid this sort of redundant AKO stipulation by loosening our·

definition

of

constituent hood.

extension.

Let

us

define

a

new

notion

called

A Frame A is a CONSTITUENT of a FRAME B if

either: (1)

B is AKO A.

(2)

We can construct a frame s· which is AKO both B and A, and B' has a structure identical to B.

Intuitively, what we are after here is this: A is a constituent of B if making B AKO A would make no difference in the structure of A, that is, if the structure of A contains the structure of B. special

case

of

one

frame

containing

the

structure

AKOness is a of

another.

Constituenthood generalizes this notion of contained structure to include cases where the structure is

"incidental,'' rather than stipulated by

217

AKO links. In our example with mock frames both FRAI\!ES and FRAMEC are constituents of GF, even though GF is

not AKO FRAMES.

We now

generalize our definition of extension: (1) S is a constituent of A.

(2) S is a component of A. (3) A is an extension of C and C is an extension of S. Now since the Rule of Referenced Structure depends on the definition of extension, FRAMES will be a legal template scene for GF. making

use

of

this

revised

definition

of

extension

encounters with the COMMERCIALEVENT verbs.

there

is

any

use

our fur·ther

1

One might wonder whether with the notion available,

in

We will be

of constituenthood

for AKOness at all.

We could define

''contained structure" in terms of sets of requirements on particular slots, and say that one frame contained the structure of another if it contained

all

of

its

slot

requirements.

Elsewhere,

AKOness could be discarded in favor of contituenthood.

any

appeal

to

For example,

our truth definition for one place predicates requires that denotation of

1. The Rule of Referenced Structure also makes use of the notion AKO at one point, where it requires that K, the frame rooting the reference path, be AKO T. In order for the trivial cases with an empty path to work, we must relax this requirement; T need only be a constituent of K.

218

a term be AKO the denotation of the predicate.

We might just as well

require that the denotation the predicate be a constituent denotation of the term.

of

the

Thus, for "(DOG XJ'' to be tr·ue, X would

have to have all the slots defined in DOG and satisfy all the constraints on them defined in DOG, but it would not need to be AKO DOG. The trouble with :;witching over to this way of looking at things is that it gives str·ucture.

us

no way of stating generalizations

about

sha.red

Without stipulated AKOness, there is no way for one frame

to make use of structure already defined

in

another.

Since

lexical

structure is our chief concer·n, we will thus continue to use the idea of AKOness.

However, the idea of extension, and thus both the Rule of

Referenced Structure and the definition of semantic compatibility, will make use of the more general notion of constituenthood. Two more formal points need to be made before we continue with the

COM~1ERCIAL_EVENT

First,

we will

verbs.

be making the assumption that no lexical scene

assigns the same role to two different participants.

Call this constraint

uniqueness, after the LFG principle of the same name which constrains each grammatical function to have only one realization (see Kaplan and Bresnan 1982).

Uniqueness is not only a standard assumption common

to many varieties of case grammar; make

our

semantics

it is also technically necessary to

work in the general case.

The mapping from

219

pr·epositions

to

participants

must

be

determinate;

if

a

preposition

marked some role R, and the governing head assigned two participants to that role, how would we know which participant was being marked? Uniqueness lexical

scenes,

a·onor and the

will

rule

because in

out an

scenes

like

ACTION_POS_ TRANS

ACTION_POS_ TRANS

recipient have the actor role.

scene,

This

both

as the

is of course a

special stipulation that distinguishes lexical representations from others, but it is a stipulation that can be confined to the grammar, and need not be expressed in the frame representation as such.

We can think of

it instead as a requirement imposed on the mapping from lexical items to their denotations as defined in the lexicon.

Any scene which assigns

the same role to two participants is simply not suitable for inclusion in a lexical denotation.

We will be discussing the mapping from scenes to

denotations in greater detail in chapter 3. The relevance of uniqueness at this juncture is that it will help us decide which material belongs in a figure scene and which mater·ial must be

in

a ground scene.

For example, the COI\1MERCIAL_EVENT

scene will have two actors, but no COMMERCIAL_EVENT verb will. The second formal

point also concerns separating figure scene

material from ground scene material, and will lead to another constraint imposed on lexical scenes by the grammar.

We assume that all of the

participants in a lexical scene will be grammatically realized.

Call this

220

completeness.

assumption

To

define

completeness

requires

some

sharpening of the notion participant. First, certain kinds of slots will not count as participant slots. In particular, ako and reference slots will not; neither will component or subscene slots (see 2.3).

Excluding these slots, a participant may

be thought of as a maximal elaboratior set, where an elaboration set is a set of toplevel slots in some frame F the

set,

one

such that for any two

is an elaboration of the other,

elaboration set in F is an elaboration

slot~

in

and where a maximal

set of F that is not contained by

any other elaboration set in F. For a participant to be grammatically realized means that one of its slots is grammatically realized. We will specify in chapter 3 what it means

for

a

slot

to

be

grammatically

realized.

Intuitively,

governing head it means the slot is either a subject, marked

by

a

subcategorized-for

complement.

for a

an object,

or

If a governing head

grammatically realizes all its participants, then any subcategorized-for preposition must also, since the head is an extension of the preposition scene.

A participant in an

adjunct

preposition

is

realized

if

it

contains a slot that is either realized as object of the preposition, or is an elaboration of some slot contained in the situation slot. In what follows, we will appeal to completeness to decide what material belongs in a figure scene and what belongs in a ground scene.

221

We

have

principle,

already be

in

seen the

an

example

ground

of

scene.

material The

that verb

must, spend

by

this rs

a

COI\1MERCIAL_EVENT ver·b, but does not realize the seller. Therefore, the seller must not be included in the lexical scene for spend.

222

the preposition for and more

2.5 2.5.1

verbs

COM~IERCIAL-EVENT

Buy, sell, and pay Throughout most of our discussion of COMMERCIAL-EVENT the

participant money has played only a supporting role.

We have analyzed

one Vf·rb, pay as having a figure scene concerned with the POS_ TRANS of the money, with buye,· c.s donor and seller as recipient, but we have totally ignored the realization of money with the verbs buy and sell. To deal with this question we shall need to consider with some care various occurrences of the preposition for.

Let us begin with two old

standbys:

(27) John bought the Porsche from Mary for S200. (28) John sold the Porsche to Mary for s200. On the easiest readings, other he gets it. these

occurrences

in one case John gives the money,

in

the

It would be desirable, of course, to ascribe both of

for

to

the

same

meaning,

one

that

in

the

COMMERCIAL-EVENT scene, picks out the money.

Under the proposal

concerning figure and ground scenes introduced

in the last section,

this would mean that the figure scenes for both buy and sell would have to be some variety of the POS_ TRANS scene compatible with the relevant meaning of for.

The following examples, all involving other

POS_ TRANS verbs, make this approach look promising:

(29) John gave Harry the book for a dollar. (30) Trish assigned the task to a prelate for a small bribe. (31) Trish acquired the old book for a very reasonable price. (32) *Trish received the old book for a very reasonable price. (33) The

seller transfers the goods to the buyer for a good price.

Note the unsuitability of for with receive.

A first hypothesis is that

this meaning of for requires POS_TRANS verbs with an ACTOR.

\'/hen

the actor is donor, the object of for is interpreted as coming into the actor's

possession

recipient,

(the actor is also a recipient); when the actor is

the object is interpreted as

(the actor is

also a donor).

leaving the actor's possession

In what appears to be the relevant

sense, for also occurs with non-POS_TRANS verbs. (34) John vacationed in Sienna for under 5300. (35) Mary fixes computer·s for an outrageous salary.

(36) John made a film for lOK. (36)

appears to allow both a reading on which John got the money and

one on which he spent it.

We will pursue an account that grants for a

meaning something like "exchange," and will assume that the apparent "ambiguity" of (36) is due to semantic vagueness.

John may get or

give the lOK; all that for means is that the lOK is one of the items in an exchange involving John's making of a film. A slightly different class of examples that can be subsumed under

224

the same meaning is discussed in O'Connor 1983: (37) John rewarded Mary for her· loyalty. (38) Mary punished John for his disloyalty. (39) Mary sent John to bed without his supper for marking up the walls. (39)

involves a participial clause controlled by an element in the main

clause,

but

that

element

dot!~

not

appear

to

be chosen

either

by

syntactic function or by semantic role: (40) The Italians turned the copy them recover the original.

over to

1\lary

for

helping

(41) The fort commander turned Mathers over to the Indians for behaving so savagely. We thus assume that this meaning of for combines with ACTIONS, but that no role within the modified ACTION is singled out as crucially involved in the EXCHANGE.

This is in distinction to our account of

MANIPULATION,

actor

understood

as

where

the

manipulator

flexibility of exchange for,

of

the

in

a

modified

instrument.

action Given

was the

always greater

something like the following ground scene

will do: (EXCHANGE: CONDITIONED_SCENE (PRECONDITION (ACTION)) (SCENE (POS_ TRANS (DONOR) (PATIENT ) (RECIPIENT))))

Here

we

call

distinction

EXCHANGE

between

the

a notion

CONDITIONED_SCENE. of

cause

We

(appealed

to

draw 1n

CAUSAL-SEQUENCE scene), and the notion of a conditioning event.

a the A

conditioning event constrains the occurrence of some other event. but does not necessarily cause it. event it condition•;.

It may, for example, occur after the

Often we pay for things before we get them, but

we can still speak of the acquis.ii.ion as a condition on the payment. Conversely, if John afterwards,

makes a film for a salary of lOK, he may be paid

but we still speak of his directing the film as a condition

on the payment of his salary. The basic idea of the EXCHANGE scene, then, is that some ACTION conditions a POS_ TRANS scene.

No role in

that ACTION scene is distinguished; there is not even the requirement that the actor must be in the POS_ TRANS scene as either a donor or a a recipient, although all the examples we've seen would satisfy such a requirement.

We could impose some such restriction, but its disjunctive

nature would require some new notation, and it would have little effect on what follows. Now the above scene cannot be the immediate representation of

exchange for, 2.4.2.

because of the completeness principle we

adopted

in

When for is an adjunct, the ACTION will correspond to the

scene of the matrix verb, but the only part of the POS_ TRANS scene that is consistently realized is the exchanged object, which is object of the

preposition.

Therefore,

to

have

a

grammatically

feasible

226

representation of for,

we will need to construct a figure scene which is

missing the unrealized elements.

This is easy enough to do.

We begin

by constructing TRANSITION, a scene of which TRANSFERENCE will be an extension, and proposing that

POSSESSION be constructed out of a

single-participant scene called UNDERGOING: (TRANSITION (FIGURE)) (TRANSFERENCE (FIGURE) (SOURCE)) (UNDERGOING: THING (PATIENT)) (POSSESSION: UNDERGOING (POSSESSOR) (PATIENT)) Note that the

extended

definition

of extension given in 2.4.2 makes

TRANSFERENCE an extension of TRANSITION, even though there is no AKO

relation

between

TRANSFERENCE).

By

them more

extension of UNDERGOING.

(TRANSITION conventional

is

means,

constituent

POSSESS ION

is

of an

Using the TRANSITION and UNDERGOING

scenes, we can construct a bigger scene of which extension.

a

Call this scene CAUSED_ TRANS IT ION:

POS_ TRANS is an

2'27

(CAUSED_ TRANSITION (ANTECEDENT:COMPONENT (TRANSITION (FIGURt: ))) (CONSEQUENT:COMPONENT (UNDERGOING (PATIENT)))) Since

TRANSFERENCE

is

an

extension

of

TRANSITION,

and

CAUSED_ TRANSITION makes rro requirements at all on the consequent slot,

CAUSED_TRANSITION is a constituent oi PCS_TRANS.

\1/e have

in effect an underspecified version of POS_ TRANS with no recipient and no donor.

We now have the necessa r·y material for the template scene

of exchange for. (EXCHANGE_ TEMPLATE: CONDITIONED_SCENE (CONDITION (ACTION)) (SCENE: COr-.1PONENT (CAUSED_ TRANSITION (ANTECEDENT:COMPONENT (TRAr-..!SITION (FIGURE ))) (CONSEQUENT:COMPONENT (UNDERGO! NG (PATIENT)))))) The lexical scene for exchange for will be called EXCHANGE_FOR; will reference EXCHANGE, and it will be AKO EXCHANGE_TEMPLATE. (EXCHANGE_FOR: EXCHANGE_ TEMPLATE (REFERENCE (EXCHANGE))) Rule of Referenced Structure

=================================> (EXCHANGE_FOR:EXCHANGE_TEMPLATE (REFERENCE (EXCHANGE)) (CONDITION (ACTION))

rt

228

(SCENE: COivlPONENT (CAUSED_ TRANSITION (ANTECEDENT:COMPONENT (TRANSITION (FIGURE ))) (CONSEQUENT:COMPONENT (UNDERGOING (PATIENTd EXCHANGED_OBJECT>)))))) As desired, the lexical scene lacks botn donor and recipient slots. Now let us turn specifically to

COM~1ERCIAL_EV£iH

verbs.

The

first question is: do we wish to make the money a core participant with

buy and sell? adjunct,

it

Given that we have an account of exchange for as an

would

be

plausible to

claim

that

this

freely

occurring

meaning of for is sufficient to account for its interpretation as money with verbs like buy and sell.

Selectional r·estrictions argue otherwise:

(42) John gave Mary his Buick for a grand. (43) John sold Mary his Buick for a grand. (44) John gave Mary his stamp collection for a bottle of Scotch. (45) ?John sold Mary his stamp collection for a bottle of Scotch.

Sell seems to require real currency as its medium of exchange.

Under

the current proposal for adjuncts (which take matrix verb scenes as participants), there is no way for a matrix selectional

restriction

subcategorized for.

on

an

adjunct.

verb to impose such a

Therefore,

for

must

be

This is a new criterion for core participanthood:

given an independently motivated preposition meaning,

involving some

scene S, any occurrence of that preposition meaning which

involves

:!29

some new restriction on S must be subcategorized for; the participant marked by the preposition must be a core participant of its governing head. Besides the selectional restriction that money be the medium of exchange, there is another restriction.

We have seen that in general

exchange for can mark either things tF1at the actor gets, or things H:at

the actor gives: (46) John made the film for lOK. When used with buy to mark the medium of exchange, for must mark something the actor gives. occur as an adjunct with buy.

am not claiming here that for can never Indeed, it is easy to construct examples

in which for simply marks some object whose exchange takes a purchase as its condition: (47) For a bottle of Chivas, John bought a Porsche. This sentence has a special offer interpretation: dealership,

every

at

some

Porsche

Porsche purchaser gets a bottle of Chivas.

generous offer inspires John to buy. something John gets,

In this case, exchange for marks

but it does not mark something which can be

interpreted as the medium of exchange in a COMMERCIAL_EVENT. claim

This

I have made is that when for

The

occurs with buy to mark the

medium of exchange in a COMMERCIAL_EVENT,

the medium must be

money, and it must be money the actor surrenders.

To get this result,

230

we shall have to impose some extra structure on an EXCHANGE scene, and we can do this only by making the EXCHANGED_OBJECT a core participant. Given that, the COMI\1ERCIAL_EVENT scene will be quite different from what we proposed last section. general

scene

involving

exchange

First, of

goods

let us assume there is a which

will

be

used

in

building up the ground scenes of verbs like exchange and trade: ( POS_ TRANS_EXCHANGE: EXCHANGE (CONDITION (ACTION_POS_ TRANS)) (SCENE (POS_TRANS (DONOR) (PATIENT ) (RECIPIENT)))) Essentially, this is just an exchange scene in which the ACTION been constrained to be an ACTION_POS_ TRANS scene;

has

recall that an

ACTION_POS_ TRANS scene is just a POS_ TRANS scene in which both the donor and the recipient are actors.

Now let COI\11\1ERCIAL_EVENT

be AKO POS_ TRANS_EXCHANGE. (COMMERCIAL_EVENT: POS_ TRANS_EXCHANGE (CONDITION: COMPONENT (ACTION_POS_TRANS (ACTOR_RECI PIENT ) (ACTOR_DONOR ) (PATIENT))) (SCENE: COMPONENT (POS_ TRANS (DONOR ) (PATIENT ) (RECIPIENT ))))) The

BUYING

template

will

be

the

RECIPIENT _EXCHANGE_ TEMPLATE

231

scene, a scene which has POS_TRANS_EXCH ANGE as an extension, and thus has COMr>1ERCIAL_EVENT as an extension: (RECIPIENT _EXCHANGE_ TEMPLATE: EXCHANGE_ TEr>1PLATE (CONDITION: COMPONENT ( RECIPIENT _ACTION)) (SCENE: SU BSCENE (CAUSED_ TRANSITION (PATIENT )))) Recall that a RECIPIENT _ACTION is a POS_ TRANS scene 1n which .y,Jiy the recipient is an actor. ACTION_POS_ TRANS,

RECIPIENT_ACTIO N is thus a constituent of RECIPIENT _EXCHANGE_ TEr>1PLATE

and thus

POS_ TRANS_EXCHANGE as an extension.

has

The construction of BUYING

is then: (BUYING: RECIPIENT _EXCH.A "JGE_ TEMPLATE (REFERENCE ( COMr>1ERCIAL_EV ENT))) Rule of Referenced Structure

=============================> (BUYING: RECIPIENT _EXCHANGE_ TEMPLATE (BUYER ) (SELLER ) (GOODS ) (MONEY ) (CONDITION: COr>1PONENT (RECIPIENT _ACTION (ACTOR_RECI PI ENT) (DONOR) (PATIENT))) (SCENE: SUBSCENE (CAUSED_ TRANSITION (PATIENT )))) In

fact,

in

this

version

of

buying,

the

only

explicit

element

of

232

COM~1ERCIAL_EVENT

that has been left out is the assignment of the

role actor to the seller.

This is because the way ACTION_POS_ TRANS

was originally defined, donor is an elabor·ation of actor but not vice versa.

The actor is not the elaboration of any slot, and thus can only

become part of the figure scene if it already exists in the template scene. SELLING

will

differ

only

in

that

DONOR_EXCHANGE_ TEI\1PLATE,

a

its

template

scene

scene

will

just

be like

POS_ TRANS_EXCHANGE except that only the donor in the POS_ TRANS scene is an actor. The Rule of Referenced Str·ucture will operate just as it did with BUYING, with donor replacing recipient as the role that is ·an elaboration of actor. Note that the above representation of COMMERCIALEVENT two

POS_ TRANS

scenes.

However,

POS_TRANS scenes symmetrically;

it

does

not

represent

one is a condition,

one a

has those

scene.

Given that representation, no COMMERCIAL_EVENT verb will be able to subcategorize for exchange for to mark the goods.

However, pay is a

COMMERCIAL_EVENT verb, and one of the criteria we used to argue that buy subcategorizes for for also applies to pay.

Namely, the actor

must be understood as recipient of the goods in examples like: (48) John payed a grand for the Buick. In order to allow for to mark the goods, we shall have to complicate our

233

representation of a COI\11\1ERCIAL_EVENT. What

propose

is

that

we

allow

COI\IMERCIAL_EVENT

to

be

composed of two different views of a situation, one of which views the

money as the exchanged_object, one of which allows goods to be. can be done with only minimal new apparatus, it, let us separately compose the two views.

This

but before introducing

Let COMMERCIAL_EVENTl

be roughly scene we already have, which views money as the exchanged

object.

COMMERCIAL_EVENT2 will have to be:

(COMI\1ERCIAL_EVENT2: POS_ TRANS_EXCHANGE (CONDITION: COMPONENT (DONOR_ACTION (ACTOR_DONOR ) (PATIENT ) ( RECIPIENT ecifying a ground scene; an

encode a

They will

arbitrary

amount

of

this is because a ground background

information;

it

includes all possible views of some object which, intuitively, is a single type of situation in the world. 2.5.2

A Note on Polysemy Any theory of polysemy will prefer· one meaning to two; but it is

far from obvious that it will prefer three meanings to six. representations differing wildly in structure might very

Thr·ee

well be less

perspicuous than six clustering together, or even six representations that were ordered by a transitive

resemblance relation so that each

successor was tolerably close to its predecessor, even though the ends of

the

"family semantic

chain

were quite distant

resemblances").

If

representations

the

are

(this

is the case of Wittgenstein's

defineable

that

complex,

relationships some

among

decisions

our about

individuating meanings can only be made after looking at all the uses for a pa rticu Ia r form.

But if deiifer·ent meanings for the same form

share some non-trivial structure, that will count as evidence for them. If matters have gotten that far along, no decision about individuating meanings can be made without looking at all the uses for a particular form.

Thus,

we

might

derive

encouragement for

our account of

238

exchange for by looking for other meanings of for built up out of the

same components. Along these lines, there is a well-known use of for that seems to be representable with a POS_ TRANS skeleton: (BENEFACTION (AKO (POS_ TRANS)) (DONOR) (BENEFICIARY ) (PATIENT (ACTION (ACTOR )))) This is the sense in which the actor is donor of an action, the classical benefactive: (59) John made the sweater for 1'.1ary (60) Mimi sang an aria for the audience It

is

of

interest

that

these

sentences

have

paraphrases

with

double-object VP's, s1nce we have already suggested that one use of the double object frame.

construction

be

related to the bare

POS_ TRANS

Note that any verb semantically compatible with the above frame

would automatically be compatible with the POS_TRANS scene, and thus, with the double object construction related to POS_ TRANS. One more meaning of for deserves mention here, and this is our old friend, karman.

Is it possible that we can decompose the LONGING

scene as it stands into something a little more familiar? move is to reduce karman for

to exchange for.

One obvious

Against that move we

239

have the possibility of intensional readings with karman (or,

but not

exchange for: (61) John wished for a tootsie roll.

(62) John gave me an apple for a tootsie roll. The question is, could we get this intensionality contrast with a single meaning of for, with the intensionality when it occurs always due to the meaning of the governing head? The real answer to this question depends on an explicit account of

intensionality,

which

do

speculative considerations present

framework.

that

First,

not

have.

argue

for

wl-':tever

Yet

can

separate the

offer

meanings

correct

Here I mean ""object" in the broadest sense.

is not a two-place relation between entities,

in

the

treatment

intensionality, it will involve slots that take a very special object.

some

of

kind of

Obviously, seek

but even on 1\-lontague"s

type-jacked possible-world account, it can be thought of as a two-place relation demands

between as

properties.

peculiar objects,

a

relation

which

in

one

position

its "object'" a function from possible worlds to sets of Call the object type r·equired to fill the ''direct object slot"

of seek a "peculiarity."

Suppose that we assume that no elaboration of

a entity slot can be a peculiarity slot, basing this on the convention in our

system

subtracted.

that

information

can

be

added

to

a

slot,

but

never

Then a single meaning of (or that covers all the cases

would have to have a peculiarity slot;

in the cases where (or was

2-lO

extensional in use, the have

to

be

elaboration of karman in the head scene would

extensionalized.

But

then,

this

leaves

no

way

to

extensionalize the object of the preposition in the case of adjunct uses of

for,

because

a

head

can

non-situation slots in an adjunct.

have

no

"selectional"

effect

on

the

A single meaning of for that is never

an adjunct and takes only peculiarities solves this problem. there is enough similarity between inducement and

Nevertheless,

karman so that something like this looks promising: (LONGING: INTENTIONAL-STATE (LONGER ) (KARMAN (PECULIARITY )) (INTENDED (POS_TRANS (RECIPIENT) (DONOR) (PATIENT))) This frame presupposes a general INTENTIONAL-STATE frame that will cover

verbs

like

believe,

imagine,

and

expect,

proposition connected with the state involves POS_ TRANS,

"coming

in the case of LONGING situations.

a ''black box" function I have called

assuming to

that

the

have,"

i.e.

Somehow, through

"peculiarity," the karman slot of

the toplevel frame is associated with the patient slot in POS_ TRANS. Although, the proposal is quite vague, I think it is quite plausible that a

representation

of

karman for

will include some POS_ TRANS scene

within it. If so,

we would have three different meanings

of the multiply

241

ambiguous preposition for which all use the POS_ TRANS scene in their definition. meanings

While this hardly ''motivates" the coincidence of these three single

a

with

occurring

form,

it

at

least

suggests

a

connectedness of meaning which ought to be less costly than three utterly

unr·elated

structures.

That

is,

using

the

metaphor

of

acquisition, given two of these meanings of for, a third meaning with a POS_ TF.AI\:S scene should be easier to learn than one with no shared structure at all. 2.5.3

Ex-spending the ontology When philosophers discuss questions like "Are there qualities?''

what they are trying to do is settle on a technical ontology, a collection of objects they must believe in in order to stay open for business.

At

times they have thought such questions are best settled by inspection of the world, at times by inspection of the language, and at times, it is hard to tell.

It should be clear by now that ontological parsimony is

not in the cards here,

but this state of affairs is not necessarily

offensive to parsimonious philosophers.

Rarely, if ever, has the wol'd

"parsimony" been in used this connection to mean that there should be few objects.

This

disorderly universe.

would

be curious

restriction to impose on our

Rather, the term applies to positions in which the

criteria for objecthood are few and stringent, so that only the things we are most likely to want to call objects can get by.

While

I admit that

I have already allowed

creatures into the bestiary, admission section,

are at least clear, want

to

some fairly curious

want to argue that the standards of if not particularly stringent.

briefly

address

some

points

In this

of

possible

indeter·minacy, and assist at the induction of a new member. When we discussed the 1uestion of why one should believe that any

such

thing

situation-type, verbs.

as

a

COMMERCIAL-EVENT

was

an

individuable

cited as evidence a few entailments among

English

We are thinking of an entailment between two verbs as showing

that their· use involves reference

to the same situation type.

Since

our goal is the construction of a discourse ontology, this is sufficient grounds for admission. admitting ordinary

Curiously enough, it gives us no grounds for

unnamed objects like the

keyboard this is

being

composed at, but this only reprises a point we have noted before in passing; ontology.

a knowledge representational

ontology

is

not

a

physicist"s

But when we get to specifics, a bit more has to be said.

Let us take as given that the very special relation between buy and sell licenses a scene we call COMMERCIAL-EVENT and a verb class we call COMMERCIAL-EVENT verbs.

Then sentences like the following

mutually entailing pair license adding pay to that class: (63) John paid twenty bucks for a funny hat.

243

(64) John bought a funny hat for twenty bucks. But, what now do we say about the following: (65) John paid a thirty dollar fine for parking on Hyde Street. (66) *John bought a thirty dollar fine for parking on Hyde Street. That this is a different verb pay seems implausible.

What appears more

likelv is that a singie Vt!l'b pay covers a wider range of situations than buy.

One recourse, then,

is to posit a scene slightly more general

than COI\Hv1ERCIAL-EVENT (call it TRANSACTION), and say that pay references

TRANSACTION,

rather

COMMERCIAL-EVENT.

than

If

COMMERCIAL-EVENT is AKO TRANSACTION, that will mean that every COMMERCIAL-EVENT is (entails) a TRANSACTION, and that will get us entailments from buy and sell to pay.

The problem is that it will get

us nothing going the other way; no sentence involving pay will entail a sentence involving buy or sell. This presents difficulties for the proposed treatment of pay, but perhaps not insurmountable ones. structured

sets

of

procedures

Suppose that we think of scenes as test objects

that can

whether they are a certain kind of thing.

to determine

COMMERCIAL-EVENT is a

single large procedure that returns yes or no depending on whether a particular object satisfies the COMMERCIAL-EVENT tests. possible that we can tell if some objects, TRANSACTIONS,

are further

specified

Then it is

specified by a verb as

by the

nature

of

its

core

participants to be

In particular, if the goods

COtvlt-.IERCIAL-EVENTS.

and money are suitable -

if one is a merchandiseable item, and the

other a real sum of money -

then a TRANSACTION would invariably be

a COMMERCIAL-EVENT. I have no objection to such an account, exceptn that such test procedures ar·e hard to formulate and somewhat of a departure from what we

have

distinguish

done

between

thus

far.

scenes

But

it

which

is

clear

are

that

one

needs

to

to

be

constrained

COMMERCIAL-EVENTS with certain kinds of slotfillers, and those which are constrained only by a larger context, and not by their slotfillers. Consider cost: (67) The printer breakdown cost me three hours. (68) *I bought a printer breakdown for three hours. Thus far,

cost is parallel to pay in that it

ranges

beyond

simple

COMMERCIAL-EVENTS, but now consider the following pairs: (69a) John bought that sweater for thirty dollars (69b) John paid thirty dollars for that sweater. (70a) John bought that sweater for thirty dollars. (70b) That sweater cost John thirty dollars. My claim is that the sentences in the first pair are In

(70),

mutually entailing.

however, the sentence with buy entails the sentence with

cost, but not vice versa.

"That sweater cost John thirty dollars" is

consistent with a situation in which John has paid for extensive repairs

245

to his favorite cardigan. One way to talk about these facts is the following: buy, sell, and

pay all reference the same situation type, but buy and sell have extra selectional restrictions on their core participants, forcing the money to be real money and

the goods to be what, for want of a better term,

we will call ''mer·chandise."

On this account, :t would be an

accident

for buy and sell to have the same selectional restrictions, and in fact, there seem to be some differences: (71) Tom bought three hours of Willhelmina's time for a C-note. (72)

?Willhelmina sold three hours of her time for a C-note.

(73) Jack sold his kidney to Menninger's. (74) ?Menninger's bought Jack's kidney from him.

cost, COST,

on the other hand,

references a more general scene, call it a

such that every COMMERCIAL-EVENT is a COST.

This

is

consistent with the much broader range of contexts cost is appropriate with: (75) Jack's coronary cost him a pint of blood. (76) Phil's greasy comment cost him a million votes. The account sketched makes a figure scene encode a bit more than we have imagined till now.

Thus a ground scene encodes as much

of the situational information as is consistent with all the verbs that access

it;

a figure scene is allowed to add selectional

restrictions.

246

Furthermore, let us try the following constraints; a figure scene can add restrictions only to its core participant bearing slots (not to slots that

take grammatically unrealized situations); and those restrictions mean

local to the slot.

must be entirely

by this to

rule

out

statements of the foliowing form: if slotl is sor·tA, slot2 must be sortB. Such facts about lexical items must follow from the ground scene. chief motivation for this constraim

i~

The

·i:o prevent figure scenes fr·om

bearing extra ''relational .. information, constraining the roles that that slotfillers play to one another; all that information should be found in the ground scene. Let us now turn to the case of spend, which was one of the original motivations for introducing the figure/ground scene distinction, because it does not allow the seller to be realized. a real COM!\1ERCIAL-EVENT verb then the

Now if spend is not

fact that it does not allow

seller to be realized does not have to be represented as a difference

between figure and ground scenes, and spend offers no motivation for the distinction.

Consider:

(77) John spent thirty dollars for that sweater. (78) John spent thirty dollars on that sweater. Only the first sentence entails that John bought the sweater. second, as with cost, suggests the following:

is consistent with a heavy tailoring bill. in one valence spend is

verb; in another it is not.

a

The This

commercial-event

One entry is compatible with on and not

247

for;

and conversely for the other.

To capture this fact, we will have

to give spend two different figure scenes, and two Certainly

lexical

economy argues we should

preposition meaning if we can.

different templates.

hang the differences on

But pairs like this suggest different

situation-types are involved: (79) In the 1960's we spent billions of dollars on an unwinnable war. ?for This pair seems to show that the on marks a different role from goods, not something that

is owned after a transaction,

consumes a resource during it. liability.

but something that

A better name for this slot might

be

Nor· is spend with on limited to commercial transactions:

(80) I spent three days on that ?for

paper.

Like cost, spend on seems to reference a more general situation type than cm..1r-.1ERCIAL-EVENT. Since nothing more seems to stand rn the way of spend for,

let

us propose a template scene: (SPENDING-TEMPLATE: EXCHANGE_FOR (CONDITION (ACTION_DONATION: CAUSED- TRANSITION (DONOR) (PATIENT))) (SCENE (CAUSED- TRANSITION (PATIENT)))) An

ACTION_DONATION

has an

DONOR_ACTION

as an extension;

it

248

differs in that it lacks any participant corresponding to recipient. template scene

will thus have no slot whose elaboration is seller in the

COMMERCIAL-EVENT participant

of

This

scene,

SPENDING.

COMMERCIAL_EVENT with

and The

PATH2,

thus

seller

SPENDING

will

not

scene

because the goods

exchanged_objf•ct, just as they were with pay.

will

be

a

core

reference

are viewed

as

249

2.6

Other motivations for Figure and Ground I have argued for a division of the information about lexical items the

pr·imarily on

grounds

of

representational

adequacy,

where

the

criterion of adequacy was essentially capturing the right entailments, and semantic adequacy, where the criterion of adequacy was essentially semantic compatibility. In

this,

the concluding section of the chapter,

will try to

pr·esent some further arguments for the approach adopted. To summarize the direction of the argument: first, an examination of

a

fairly

complex

interaction

tentative lexical representations. differences frames

in

among a

situationally

configuration

that

of

compatibility

facts

led

to

some

In order to capture the full range of related lost

predicates,

certain

we

entailment

had

to

facts.

posit The

alternative we then pursued was to have the COMMERCIAL_EVENT scene related to the various lexical scenes in a new way; we introduced a new kind of slot called a reference slot, which related scenes through

a

Rule of Referenced Structure, using mechanisms quite different from the inheritance machinery we have relied on till now. Generalizing this to a method for the entire lexicon, we were left with two important frames for every verb, one the referenced situation, the other the lexical frame proper.

We called the lexical scene the

figure scene and the referenced scene the ground scene.

The ground

250

scene may in principle encode more information than is available in all the figure scenes that access it.

It may encode more when there is

situational information about a ground scene that was not a necessary part of any of the figure scenes.

Fillmore points out,

for example,

that there is always some process of entering into an agreement that precedes the actual comme•·cial transaction This

is

part

background event.

of

its

temporal

assumptions

we

and

share

causdl in

in

any commercial s·l:ructure,

understanding

part

any

event. of

the

commercial

Yet we have not discussed this aspect of a commercial event till

now because we have no compatibility facts that depend on it.

Yet, the

"contract" nature of a COMt\IERCIAL_EVENT is arguably

intrinsic

an

part of cultural significance; for example, discourse involving the legal consequences contract

of

a

aspect.

commerical As

event

will

such the "contract"

nevessar·ily

refer

to

the

implicit in any commercial

event should be part of the fully expanded generic scene. We posit scenes basis

like

BUYING and

of linguistic evidence.

Thus,

SELLING exclusively on the

the material

in a lexical frame

ought to be what is conveyed or asserted by the linguistic form, those aspects

of

the

situation

the

predicate

actually

"presents'"

or

''foregrounds."

Most obviously, we can thus think of the slots in the

template

as

frame

foreground.

The

"foregrounded" the

chosing first

certain

figure

roles

in

representation

the we

ground chose

role of the buyer as an actor and

frame for

to

buy

recipient, in a

251

RECIPIENT_ACTION scene, the goods as

a patient in that scene, and

the seller a donor; the modification proposed in the next section added to that the role of money as payment.

It will be apparent to readers

versed in case grammar and other theories of semantic roles that some of the

roles

''foregrounded"

in

the

figure

frame are more or less

classical case roles like actor and possibly recipient," and some, such as payment, are definitely not. COI\li\!ERCIAL_EVENT are in

Note that all bui. i.Le roles specific to

some way or· another

motivated

by

the

compatibility facts of buy. The

roles

in a template frame can be thought of as

roughly

analogous to the "deep" case roles of classical case grammar, with the following important differences: (1)

There ar·e many more of them.

(2)

As we shall see in the next chapter, only a subset of the of the roles in a figure frame will play a part in selecting nuclear terms from among the core participants.

(3)

The role assignments are motivated by compatibility facts. In fact, one might say they are little more than an encoding of them. But this encoding induces some rather natural lexical semantic classes, such as the POS_ TRANS verbs. As we shall see in the next chapter, it also defines at least a partial mapping onto lexical predicates, that is, a partial definition of the semantics of nuclear terms.

In the theory presented in this dissertation there will be three kinds of facts the figure frame is responsible to, compatibility facts, "-arity" facts (how many core participants a lexical head has),

and

252

facts

about the realizations of particular· core participants as nuclear

terms.

In addition to this, a lexical head may be associated with extra,

derivative

claims

compatibilities; frame.

The

figures, men.

on

the

real

world

not

revealed

by

its

semantic

capturing such claims is the function of the gr·ound

ground can be thought of as a projection made up out its

like the vision of the elephant constructed by the six blind

The relation

mediated

by

the

between a figure frame and a gruunJ

Rules

of

Reference and

frame

Referenced Structure.

rs A

figure frame must chose its structure from the available structure of the ground frame. Entailments facts will follow from the structur·e of the ground scene and its linkage to lexica\ frames.

This is in keeping with the

'"structural" view of lexical entailment adopted at the outset.

Lexical

entailments ar·e not just scatter·ed facts; they are consequences of the knowledge

representational

structure

and

should follow from correct

lexical representations. A structural view of entailment an

axiomatic

approach,

is intended as a replacement for·

where the model theoretic interpretations of

lexical items are constrained by ''meaning postulates." On that sort of account, each lexical item would be associated with certain axioms, and certain groups of lexical items would share axioms, and be related by them.

To give an

example,

one axiom might specifically state that

every COMMERCIAL_EVENT is an ACTION, and thus yield an entailment

253

relating, say, buy and do. (81) John bought a VW. (81) John did something.

Suppose

the

frame

framework adopted follow

for

do

here,

and

simply

the

ACTION

BUYING

has an

RECIPIENT _ACTION

is

In

scene.

the entailment between buy and

from the fact that

component,

is

do

the

would

RECIPIENT_ACTION as a AKO

ACTION.

Our

interpretation of the AKO mechanism is already such that anything that is

AKO

RECIPIENT_ACTION

is

automatically

AKO

ACTION;

our

interpretation of component scenes is such that any BUYING scene must automatically involve a RECIPIENT_ACTION with buyer as actor. On this view, lexical entailments are never specifically stated, as such.

Methodologically, they are among the intuitions that guide our

design of the procedures implemented in a frame representation. are tests of the theory.

They

They are not themselves the objects of our

description. Some comment ought to be made about the contrast between figure and ground representations and other classical semantic contrasts,

in

particular,

an

informal

the contrast

sense,

both

between figure

extension

and

ground

and

intension.

frames

are

In

intensional

descriptions; that is, they are descriptions of sets rather than sets. The "referenced" frame is not really referential in the model theoretic

254

sense, since the COMMERCIAL_EVENT frame doesn't change even if the set

intension/extension

distinction

is

generally

intension gives you more information.

the

Moreover,

changes.

COMMERCIAL_EVENTS

specific

of

interpreted

so that the

That is, given the intension you

have a pr·ocedure for determining the extension, but not conversely. In Montague's treatment,

intensiors are functions from

for example,

possible worlds to extensions; so given a world and intension, you irc1ve in hand an extension. In the case of figure and gr·ound :;cenes, it is the ground which is more fully specified (more informative) and the figure which is less Similar

specified.

comments

apply

distinction, in so far as I undpr·stand it. is primarily

to

Frege's

sense/r·efer·ence

The distinction offered her·e

one between the structure of the real world situation-type

('"real" in the sense of culturally real) and the structure of a particular view of that situation-type. On the other hand,

this

view of

is

description

lexical

compatible with others that have been advanced.

quite

First, of course, it

seems to be a reasonably straightforward interpretation of Fillmore's dictum, "Meanings are relativized to scenes,·· where scenes in our terms are ground scenes, and "meanings" are figure scenes. good

deal

of

"profile" and

conceptual "base"

scenes, respectively.

(see

similarity

between

Langacker 1983)

There is also a

Langacker's and

figure

notion

and

of

ground

Both these proposals differ from mine in that

255

have

But

compatibility.

through

scenes

figure

motivated

some

the

of

same

of

considerations semantic

semantic can

be

in chapter 1,

the

intuitions

captured. Consider a

simple case that was discussed

notion of a "figure" as a preposition role: (83) The ball is over the box (8-0 The box is under the ball. Intuitively, we can think of the prepositional phrase in both cases as predicated of the subject (this notion will be elaborated in chapter 5). We can also think of the subject in

both cases as the object being

located; in Talmy's work, figure, in Gruber's, theme, and the classical term, Patient.

At the risk of confusion with the notion of figure and

ground scenes, let us adopt Talmy's more specific ter·m and call the role here figure

(we

already

have

made this terminogical choice in the

TRANSFERENCE scene associated with from).

Both over and under can

be thought of as referencing the same real world situation, vertically oriented in participant they

space.

encode

as

two objects

But the figure scenes differ in figure.

compared to COMMERCIAL_EVENTS.

This

is

rather

a

which

simple case

The relevant scenes are:

256

(ORIENTATION_IN_SPACE: THING (FIGURE) (GROUND)) (VERTICAL_ORI ENTATION: SIMUL TANEOUS_SCENES (FIRST_SCENE: COMPONENT (ORIENTATION_IN_SPACE (FIGURE q INFERIOR>) (GROUND)) ( S ECON D_SC ENE: COMPONENT (ORIENT ATION_IN_SPACE (FIGURE ) (GROUND))) (OVER: ORIENTATION_IN_SPACE (REFERENT ((PATH ) VERTICAL_ORIENTATION))) ORIENTATION_! N_SPACE

simply

concerning two objects in space, thought

of

as

being

defines

one of which,

oriented

with

unspecified

an

may

the figure,

respect

to

scene

the

be

other.

VERTICAL_ORI ENTATION is a ground scene involving two objects and By referencing

two orientation scenes.

VERTICAL_ORIENTATION with

the path , the figure fr·ame OVER makes the superior the figure. path

UNDER will

,

reference the same

reference VERTICAL_ORIENTATION with the

making the inferior the figure.

situation,

Both

frames

but assign different roles to the same

ground participants. When we define the mapping to nuclear terms for prepositions, figure will outrank ground and will thus be mapped to the predicate

position

corresponding

to

subjects.

Thus,

the

predicates

UNDER and OVER will have the right entailment relation, but still have different nuclear term realizations.

257

Now the above representation of VERTICAL_ORIENTATION poses certain

problems

which

were

implicit

in

the

representation

of

but which can be discussed a bit more clearly

COM~1ERCIAL_EVENT,

with this simpler example.

What I seem to have done is to create a

ground frame which is simply two preposition meanings welded together·. That is, without the notions of figure and ground, one might take what I have in the first-scene

slot as the meaning of

the second-scene slot as the meaning of over.

In

ver~

just

under, and

what

sense

is

VERTICAL_ORIENTATION a description of a new situation-type, instead of just a cross-indexing of two related meanings? To make my defense here, I will need to make a small digression into the domain of. situation semantics (Barwise and Perry 1980),

where

some very similar issues have arisen in a different connection. The starting point of the Barwise and Perry program seems be this:

to

a theory of how language car·ries information about the world

must be founded on a theory of information.

But information, and

thus meaning, is in the world to begin with in the form of regularities across situations. building

blocks

Among these uniformities we find the traditional of

logicians,

Barwise and Perry thus propose

objects,

properties,

and

relations.

a theory of meaning that focuses on

the described world rather than on the mind that describes it. enters the picture primitive

here:

uniformities.

particular organisms These

primitives

individuate define

the

Mind

particular meaningful

258

relationships

for

that

organism.

Naturally,

those

we

are

most

interested in ar·e the unifor·mities important to humans. Two points ar·e of interest her-e. The first will help us answer the charge that we have smuggled linguistic material into our description of situations.

The Bar-wise and Perry line is that the

salient regular·ities

among situation types are part of the information available in situations; they are not imposed by language.

They are exploited by it.

At the ver-y beginning I said that the purpose of a knowledge representation was to explain discour·se understanding, with discourse understood in the widest sense, including nonverbal inter-action. Barwise

and

Perry

The

program might be taken as even more general,

depending on how one wants to inter-pret the particulars.

One of the

objects of study of situation semantics is meaningful relationships among situations.

Examples given by Barwise and Perry are the following:

Smoke means fire. Kissing means touching. The ringing bell means class is over-. "Cookie" means cookie. Barwise

and

Perry

treat

the

last

of

these

examples,

the

most

straightforwardly linguistic, as essentially similar to the rest (which is not to say identical); relationship,

an

"means"

essential

each

uniformity,

individuate among phenomena. interest,

in

case

that

picks humans

The second example is

out are of

a

lawlike able

to

linguistic

as well, because it can be recast in a form we might call

a

259

Of

lexical entailment ("Mary kissed Bill" entails "Mary touched Bill"). that case, they write:

1

Suppose the issue is whether Mary has ever touched Bill. If you learn somehow or other that ~:lary has kissed Bill, then information available to conclude that Mary has you have the This has nothing to do with language per se; touched Bill. situations across uniformities are touching and kissing relational recognized by human beings in this culture activities. And it falls out of our recognition of these activities that kissings are touchings. Kissing is just a more fine-grained uniformity than touching. The situation semantics view of a "meaningful congenial

to

the

relationship"

is quite

here.

Lexical

view of lexical entailments adopted

entailments ought to fall out from the structural relations of situation they

types;

are

not

linguistic

statements.

are part of the

They

uderstanding of the world imposed by our structuring of it (read, the uniformities we are attuned to). sources

chief

uniformities,

of

evidence

in

Of course language will be one of the formulating

hypotheses

about

such

meaningful relationships are not constrained to be

but

linguistic, and we should not expect all the regularities expressed in scenes to be reflected in lexical entailments. Let us return now to figure and ground scenes.

We have briefly

sketched the situation semantics view that linguistic meaning can be viewed as a special better,

1.

a special

kind of uniformity among situations, or perhaps

kind of information available in

Barwise and Perry 1982. p. 16.

situations.

Both

260

are

VERTICAL_ORIENTATION

and

ORIENTATION_IN_SPACE

scenes,

Both express a

situation-types in Barwise and Perry's terminology.

certain uniformity among situations, irrespective of the fact that one is The claim made by

directly associated with a particular lexical item.

the particular r·epresentation above is that ORIENTATION_IN_SPACE is a component of VERTICAL_ORIENTATION, that, in effect, it expresses

a

primitive

more

among

situations.

can

always

be

situations

_OR I ENTATION

one

subsituations,

uniformity

inferior

the

which

in

VERTICAL-

is

object

into

two

''oriented"

with

down

broken

respect to the superior, another in which the superior is oriented with respect

to

the

Now

inferior.

is

it

clear

that

in

this

definition

"oriented" conceals an appeal to a per·ceptual category; but that appeal may be made entirely independently of language.

philosophical

of

risk

the

At

blundering to

want

landscape,

into

some

argue

that

rather

well-traveled

concealed

perceptual

categories are the rule rather than the exception in frame descriptions, as

they

must

interesting

be

in

uniformities

human perceivers.

the

Barwise

across

Perry

and

situations

One of the salient

program,

where

the

are those happened on by

uniformities about squares for

humans is that, viewed from most angles, they are parallelograms.

It

is no accident that much spatial vocabulary, cross linguistically,

can

encode implicit information about the location of a and in front of are simple cases).

perceiver (behind

If our task is to model the structure

261

of the uniformities among situations, then it is quite reasonable to posit "orientation

with

respect

to"

as

a

more

primitive

uniformity

than

"absolute orientation," and to compose certain cases of the latter out of the former. It

is

share

because

the

biases

of

situation

semantics

just

out'ined that I have chosen to make lexical frames proper (the figur·e frames) the same kinds of objects as all other frames. scene

is

ever·y

bit

as

COMr-.IERCIAL_EVENT scene.

much

a

type

of

The BUYING

situation

as

the

It is simply that they bear a very special

relation to each other which is mediated by a special kind of slot.

The

reason the ground frame contains all the raw situational information is for economy of encoding; this is information shared by all the figur·e frames.

All frames, in situation semantics terms, capture regular·ities

among situations.

A ground frame can be thought of as one step of

abstraction higher, a regularity among certain of these

regularities.

In this light, there 1s no particular reason why a lexical scene needs to have a ground scene. particular

situational

Nevertheless, definitions,

to

If no other scenes material,

facilitate

the the

share this

ground

scene

statement

of

lexical

item's

is

superfluous.

our

compatibility

we will continue to assume all lexical items have ground

scenes, allowing a figure scene to reference itself on occasion. now want tum back to the question of motivating two kinds of frames, figure and ground, and move the field of argument back to a

262

linguistic plain.

One of the chief payoffs in this kind of abstraction is

in helping us maintain a useful distinction between ground participant and core participant.

The simplest case of this was the verb spend, a

COMMERCIAL_EVENT verb that simply didn't include the seller in its figure frame.

Because of this,

the question of

that participant"s

realization as a complement simply doesn't arise. A little reflection shows that such non-occurring participants will be far from rare in lexical description.

Consider the word alimony,

cited by Fillmore as one that presupposes quite a bit of background knowledge for felicitous use.

Minimally, the necessary script for this

word involves two once-married participants. be

a

core

participant of

alimony?

Should the alimony payer

Probably

not,

considering

the

unsavoriness of examples like these: (85) The alimony from Chris arrived every third Thursday of the month. (86) Chris's alimony came to two grand a month. Results for the alimony recipient seem to be similar: (87) Tracy's alimony

came to two grand a month.

(88) The alimony to Tracy arrived every third Thursday of the month. All of these examples improve drastically if alimony is replaced with alimony payment.

We can describe this state of affairs as follows.

the ground alimony frame ALIMONY;

Call

it includes two participants with

26211

the

appropriate

itself

is

and

ALIMONY

references

the

Call

history.

figure AKO

frame THING

(giving

it

it no

But it has no participants, and thus by our

compatibility privileges).

definition of core participant, no cor·e participants. ground participants,

ALIMONYP;

Again we represent

important to the understanding of a lexical item,

without calling them core par·ticipants. In general, when we get to nouns the fairly robust intuitions we have about what constitutes a "participant in a situation" will leave us, and we shall have to get by with a much vaguer tag like "associated Consider a noun

thing."

phrase lih;e the father of our discipline.

Should the child be considered a core participant in the FATHER scene? Or perhaps our picture fatherhood as a condition is clear enough so that we can say, with some confidence, what sorts of participants enter into it.

A word

like skill may

be more slippery.

It seems fair·ly

reasonable to claim that both the owner of a skill, and its domain are important participants

fairly

in defining the concept; and perhaps a

noun phrase like "John's skill at plowing" will convince us that both participants belong in

lexical frame as full-fledged core participants.

In some cases, it seems clear, there will be participants important to the definition of a noun that we do not wish to grant a grammatical status. The basic account here follows participants

are

chosen

from

among

Fillmore's in this respect: the

participants

of

a

core scene.

Scenes,

of

course,

are

justified

in terms of the situational

mainly

knowledge we extract from a lexical item,

and

if a participant in a

scene is "left out," we'll know about it only because of our knowledge of the relevant ground scene.

Until now the only motivation I have Another

given for ground scenes has come from lexical entailments. kind of motivation, perhaps linguistically stronger,

is the phenomenon

of morphologicdi der·ivation. Consider the verb and noun blame: (89) Bilbo blamed the mishap on Blaise. (90) Bilbo blamed Blaise for· the mishap. (91) The blame for the mishap. (92) *The blame on Blaise (for the mishap). (93) *The blame of Blaise (for the mishap). It appears to be the case that the noun does not allow the any

realization,

reminiscent

of

spend

and

the

seller.

culprit in A

similar

relationship, where one core participant is mysteriously excluded, seems to hold between the verbal and nominal forms of stems like envy and

credit.

We can even find such ''exclusions" in the morphology of the

semantic domain we have been discussing: (94) A buyer of Chaldean pottery (95) *A buyer of Chaldean pottery from Macy's

26SC

(96) A seller of antique books (97) *A seller of antique books to children All of the nouns in these examples are describeable in terms of figure scenes that import only some of the participants from the ground scene, where the corresponding verbs have inherited at least one participant more.

There

is

a

good

deal

more

going

on,

of

course,

since

presumably we want tv r·.,present some difference of meaning between the noun and the verb, but whatever account we choose will need to capture the difference in argument structure and the similarity in the situations

evoked,

morphological

a

very

derivation.

mechanism for doing that.

common Figure

feature and

of

ground

different scenes

kinds give

us

of a

161D

2. 7

Conclusion

In this chapter

I

have

elaborated the bare bones apparatus for lexical

description outlined in chapter 1.

The elaboration

has resulted in a

basic split-up of the semantic information about a lexical item into what I have called figure and gr·ound scenes. The initial motivatron fur thrs split was to capture the differences in complementation possibilities among lexical items which can in some Linking verbs like buy and sell

sense be linked with the same scenes. with

a

single

entailments

COMI\1ERCIAL_EVENT

among

these

verbs,

scene gives

although

it

was

us the conventional argued

that

such

entailments are facts about the way we individuate the world, not facts about language.

This appr·oach to entailment has been most clearly

articulated by Barwise and Perry, although it is clearly implicit in much artificial

intelligence

research

on

reasoning

and

knowledge

representation. It was also argued that the distinction between figure and ground scene gives us a very handy way of representing the differences in argument structure among semantically related lexical items, in articular among items related by morphological derivation.

3.

Where are the case theories of yesteryear? The question I wish to address in this chapter has been very

succinctly posed by Paul Kay (personal communication): "Why can't you say, 'Three' o'clock ate the hamburger by John?"' Less succinctly, it is the question of how meanings are mapped into grammatical functions like subject and direct object.

In the "case grammar" proposed in Fillmore

1968 the answer involved two major theoretical decisions: ( 1) an analysis of what sort of semantics can be fit into a single lexical item, stated in terms of a (hopefully small) number of semantic roles which the arguments of a head may play, and placing co-occurrence restrictions on these roles (prohibiting, for example, two occurences of the same case role with the same head); (2) a hierarchy determining which participants of a head become subjects. 3.1

of

the

semantic

Why Case Grammar? One of the central

functions

of case

roles

as formulated

in

Fillmore 1968 was to encode semantic information about how clauses are canonically organized, especially in terms of grammatical relations. was the motivation for the subject hierachy, a fragment of which is:

This

164

Ag > I > Pa > Go > So Things work very simply: if there is an Agent in a lexical entry it will become subject; if not the subject will be the next available member of the hierarchy.

The possibility of an absolute hierarchy which decides

the issue for every lexical item seems remote, because of pairs like like and please,

but I will argue below that H·e basic premise of such a

hierarchy, that there is a '"semantics" behind subjecthood, i.; bvt'n out. Let us extend the idea of a case hierarchy as Fillmore does in later works

to

include

direct

grammatical functions, of a clause.

object

selection,

and

refer

to

the

two

subject and direct object, as the nuclear terms

We can state the central hypothesis of this chapter as

follows: nuclear terms are chosen in accordance with a partial ordering on semantic roles. Early formulations of case hierarchies suffered because they failed to

distinguish

the

situation-type

a

predicate

references

particular point of view the predicate takes on a situation. simple

examples,

which

we

have

already

encountered,

from

the

Two very show

the

relevant problem. (la) (lb)

The ball is in the box. The box contains the ball.

(2a) (2b)

The sphere is over the table. The table is under the sphere.

No one would deny, I think, that the sentences in each of these pairs "mean" different things. The problem is that no description of the real

265

world

situations

these

sentences

describe

can

ever

capture

the

''meaning" difference, and therefore that no such account can capture any semantic factor motivating the difference in choice of subjects here. To speak

simply of

situations

involving

spatial

location

will

never

predict for us which of the objects in a location scene makes a good subject.

What we need to talk about is which object is being located

with respect to which.

In (a) the sphere is being located with resf.>t:d

to the table; in (b) just the opposite has happened.

In (a)

the ball is

being located with respect to the box; in (b) the communicative not one of locating one object with respect to another.

act is

What is being

expressed is a property of the box which is, of course, independent of its location.

Modifying some examples in Gruber 1976, one can show

this with question-answer pairs: (3)

Where is the ball? The ball is in the box. ?The box contains the ball. ?The ball is contained by the box.

(4)

Where is the sphere? The sphere is over the table. ?The table is under the sphere.

Only the first answer in each example functions as answer to the locative question. of

the

sentence

with

contain

an

appropriate

The fact that the passivized version is

also

inappropriate

shows

that

appropriate responses cannot simply be characterized by saying "the located object must be subject."

2.66

At the end of Chapter 2 we gave a semantics for over and under which assigned the two prepositions different figure scenes and the same ground scene.

The mechanisms developed there for linking figure

and ground guaranteed that sentences of the form "a is under b" and "b is over a" would entail each other. scenes of the two prepositions differed:

At the same time the figure for over, the superior was the

for under, the inferior was the figure.

figure;

Suppose nuw ·\J,c.t ',he

lexical representations consulted by something like the Fillmorian Case Hierarchy are figure scenes, and that the roles are interpreted simply as

slot

names.

And

suppose

we

made

the

following

addition

to

Fillmore's Case Hierarchy: Ag > I > Pa > Go > So > Figure This would pick out the right participants as subject for over and under.

In this chapter I want to accomplish three things: ( 1)

Review some of the obvious right generalizations made by classical case hierarchies, and add a few new ones.

(2)

Exploit the fact that the figure/ground distinction encodes the sort of "predicate point of view" displayed by under and over. I will also argue that there are still very strong constraints on what roles we assign to particular predicates. Those constraints are basically of two kinds, situation-type and semantic compatibility. Without some such constraints on role assignment, a theory of the semantics of nuclear terms is vacuous.

267

(3)

Present an "implementation" of the theory sketched: an logical to algorithm that maps from figure scenes predicates, where, in the grammar proposed in Chapter 1, predicate position corresponds to grammatical role. The That is, not every algorithm will not be complete. predicate will have its nuclear terms completely determined "Saliency" by the partial ordering on roles that the As Hierarchy (to use the term in Fillmore 1977b) gives. far as the theory goes, these cases are "pot luck." will begin these various tasks in the next sect1on,

briefly discuss some other examples of clear semantic constraints in choices of nuclear arguments.

I will then discuss some more recent

work of Fillmore's in which the case hierarchy as such (and with

it

cases) is discarded in favor of something called the Saliency Hierarchy, which does much the same work but is,

for various

reasons,

more

desireable.

3.2

Some Facts about the Semantics of Grammatical Relations

What

kinds

What kinds

semantic objects become subjects?

of

become direct objects?

What kinds become adjuncts?

These are the

kinds of questions linguists have, in one terminology or another, been asking ever since Panini.

With over two millenia of the semantics of

core participanthood behind us, we can with some confidence draw two conclusions:

(1)

there

are

some

solid

generalizations

connections between semantic role and grammatical function; is no complete algorithm; that is,

about

the

(2) there

in some cases, it is best to say,

268

either that the choice is arbitrary, or that only is involved.

This might, for example, be the

"grammatical meaning

most perspicuous way to

talk about the difference between like and please. In what follows, roughly,

when

say subject or object,

logical or underlying subject or object;

I shall mean,

I shall have nothing

to say about "derived" grammatical functions like that of the beans in

The beans were overcooked. Let's begin by trying to back up conclusion (1), that there are some solid generalizations about the semantics of grammatical relations. It will be useful to begin by selecting the roles which will be used in the informal statements of Fillmor·e's Saliency Hierarchy in the next section.

In this preliminary discussion I am only concerned with

describing the intuitive content of the roles; later we will look at the kinds of evidence that can justify their assignment. The something."

obvious

role

to

start

with

is

actor.

An

actor

··does

We will speak of actors as the primary source of

an

ACTION;, where only creatures capable of volition can be the primary source of an activity.

Ultimately, the notions defined here must apply

to the linguistic presentation of reality in clauses. walk," John is active in the matrix clause, active in the downstairs clause.

In ''John made Bill

Bill inactive;

but Bill is

In "Henr·ietta moved the Book," both

the book and Henrietta are most naturally viewed as moving, but only

269

Henrietta as primary sour·ce of that activity is active. notion

of causer,

(although ultimately we will dispense with this category).

Again, we

Along

with

this

we

use a closely

related

are confined to single clauses; if "Bill hits the ball across the par·k, ·· Bill causes the ball to move across the park.

Bill's role can now be

thought of as causer in addition to actor In "The key opened the door," the key is causer but not actor. Finally, we shall need a notion of changed.

If the vase breaks,

it changes. We can now get down to cases. 3.2.1

Active and Causal subjects As conceived thus far, actors correspond exactly to the agents of

Fillmore 1968 (although this understanding of actors will be revised later).

Like Agents in Fillmore's system, they will be the paradigm

examples of subjecthood in this one.

What evidence is there,

then,

that something is an actor? In this framework, all roles are defined with respect to scenes, and all scenes have their place in the lexical system as a whole.

actors occur

in ACTIONS, which are compatible with instruments, exchanged objects, and beneficiaries, all preposition frames we declared in Chapter 2. first is a meaning of with, the next two meanings of for.

The

Here they

270

are: (EXCHANGE_FOR: EXCHANGE_TEMPLATE (CONDITION (ACTION)) (SCENE:COMPONENT (CAUSED_ TRANSITION (PATIENT)))) (MANIPULATION: THING (INSTRUMENT) (ACTION (ACTOR )) (BENEFACTION (AKO (POS_ TRANS)) (DONOR) (BENEFICIARY [recipient]) (PATIENT (ACTION (ACTOR )))) Saying a verb has an actor entails that it combines with with and for with just the specified interpretations,

For example,

with

instrument

with , the actor is understood as manipulator of the instrument; with benefactive for the actor is understood as donor of the ACTION. so

on.

In

short,

there

distributional

both

are

consequences to calling something an actor. determine

if

something fills

that

role,

and

And

semantic

We have concrete tests to

and therefore the claim that

actors are subjects becomes testable. Let us also identify a role causer, which will semantically be like

theactor role, except that the requirement that the actor be capable of volition is dropped.

Actor is thus a further specification of causer,

and like actors, causers will make excellent subjects.

The subjects of

271

the following sentences will be causers: (Sa)

The key opened the lock.

(Sb)

The magnet attracted the filings.

The status of causer is somewhat problematic, and we will dispense with this role later on; for the time being it is a useful repository for those participants

called

instruments

both

forces

and

(see

Fillmore 1971)

which are realized as nuclear terms. 3.2.2

Change and Effective Instrument Before discussing roles other than actor,

clarify the ultimate goal.

it will be useful to

What we desire is a partial ordering of

semantic roles which constrains the choice of nuclear terms in a clause. We have begun by discussing the role actor; as it happens actor will be the nuclear term role par excellence, outranking all other roles; verb has an actor the actor becomes subject.

if a

Causer is similar but its

status with respect to actor is unclear.

As we move on to other roles,

we

we

will

ordering.

need

to

keep

in

mind

Given

that

some

role

that

plays

are a

constructing a partial

part

in

determining

the

semantics of nuclear termhood, we need to ask how it is ordered with respect to the other roles that also play a part -

if it is.

The fact

that we are constructing a partial-ordering leaves open the possibility that any two roles may be unordered with respect to one another. Our next example involves the notion changed.

Consider as one

272

special

subcase of ''changing" the property of "moving." Suppose we

claim that objects.

moving

objects

make

better

subjects

than

non-moving

This predicts that no verbs will fit into the following sentence

frames, with the indicated meanings: (6)

The store the meteorite. The meteorite approached the store.

(7)

The store the meteorite. The meteorite approached the store.

(8)

Zeus the store the meteorite. Zeus hurled the meteorite at the store.

(9)

Zeus the store the meteorite. Zeus hurled the meteorite at the store.

Of course, the role changed needs to be located in the ordering of roles.

It appears, for example, to rank lower than actor.

In "John

blew the filings across the table,'' John does not move, but John's role is still the subject role.

We can get this effect by including in the

ordering of roles the statement: Actor > Changed Thus any choice between the role actor and changed will be decided in favor of actor.

Causer will also have to rank higher than changed,

because a motionless object outranks a mover if it acts as a causer (for example, "the magnet attracted the filings."). Consider

a

different

kind

of

semantic

generalization

about

grammatical functions, a negative one; effective instruments may not be

273

nuclear terms/ By an effective instrument I mean an instrument that occurs in the same clause with an actor and a changed element.

The

following examples are from Fillmore 1977a: (lOa) (lOb)

( 11 a) ( 11 b)

He broke the vase with the hammer. He broke the hammer against the vase. hit the fence with the stick. hit the stick against the fence.

Let us assume that the verb break has a figure scene necessarily changes.

"breakee" instrument.

Note

that

it

can

The hammer is no

longer

instrument when realized as a direct object.

be

then

1n

an

interpreted

which its effective as

an

The direct object of break

is always interpreted as the ''breakee." This is not the case with hit. The fate of the fence and the stick remain the same in the (a) and (b) versions of (11).

Hit does not take effective instruments; that is, it

does not predicate any necessary change of its direct object in either syntactic valence.

In classical case

terms, we might call the "hittee'' a

Goal, where we would call the "breakee" Patient.

In our terminology

here, the ''breakee" is changed. The claim I made above was that

effective instruments cannot be

How can we talk about this in terms of a ranking of

direct

objects.

roles?

By definition effective instruments imply both an actor and a

changed participant in the same clause. changed

participant always

If both the actor and the

rank higher than the instrument,,

instrument is ineligible for either of the nuclear term positions.

then In

274

particular, objects,

this

prevents

effective instruments from

and rules out the following verb frames,

becoming direct

with the indicated

meanings: (12) (13)

John the needle the pillow. John pierced the pillow with the needle. John the needle the pillow. John pierced the pillow with the needle

The examples here ar·e chosen to evoke one of Gruber's (Gruber 1977), which seems to pose a possible counter-example. sentence, along with ( 14a) ( 14b)

Here is the relevant

one that differs minimally:

John pierced the needle thr-ough the pillow. John poked the needle through the pillow.

find (a) quite grating and (b) fine, so that I disagree with Gr·uber's judgements.

I think that mine may be explained in terms of effective

instruments.

For me, the piercee is necessarily changed with the verb

pierce, and therefor·e

the instrument is

not a

happy candidate for

direct objecthood; with poke, on the other- hand, the state of the pokee may be utterly unchanged.

I propose that for anyone who can

use

pierce in Gruber's fashion, pierce y through x means "force y thr·ough x;" change in x rs of course implied, but not foregrounded. such

speakers,

suspect

this

use

grating to me; (14•1) is absolutely out.

is

rather marginal;

Even for(a)

sounds

275

(15) *The needle was pierced through the pillow. Some further

are

comments

order

in

In

about these claims.

making the case for effective instruments, I implicitly appealed to the distinction

between

a

feature

of

feature of the real world situation.

the

linguistic

presentation,

and a

Suppose that Gruber and I both

have the same truth conditions for sentences with pierce, despite our differences in judgement about the appropriateness of some of them. maintain that our differences in judgement reflect a difference in the semantics we have learned, and that for Gruber,

pierce really means

something differ·ent than it does for me. This puts me rn something of a methodological bind.

Whenever

someone s lexical judgements disagree with mine, I can claim they have learned a different lexicon.

The move is quite parallel to the standard

generative appeal to "different dialects.·· are obvious.

The dangers of both moves

On the one hand, something like this notion is obviously

needed, especially at the lexical level; we all know words at the limit of our competence as speakers, words we can use in limited contexts, and then only with great uncertainty.

Even better,

we have all

known

words which later linguistic authority has shown us to know imperfectly The lexicon is the clearest place where the differences in

or wrongly.

our linguistic competence some

hypothesis

community

of

about

speakers

emerge. a is

common

But even in describing the lexicon body

essential,

of

and

knowledge appeals

shared

by

to differences

a of

276

idiolect need to be backed up with some evidence for these differences in

the

assumed

grammar.

At

present

have

no

independent

justification to offer for my claims about Gruber's sense of pierce, but note that the difference

have ascribed to his lexicon is a difference

involving one lexical item. in

the

lexicons

of all

I assume that in everyone's lexicon, in fact, languages,

the

role changed

still

plays

an

attributed

to

divergence

in

important role in determining nuclear terms. How Gruber's

would lexicon

we and

represent to

the

mine?

meanings The

have

hypothesized

understandings can be described straightforwardly in terms of figure and

ground

scenes.

Since we

have

assumed

both

speakers

have

roughly the same truth conditions for sentences involving pierce, we assume they have the same ground scenes. But now in my version of the

figure scene of pierce, the "peircee" fills a slot called changed; in

Gruber's it does not. The claim I have made about effective instruments is that they cannot be direct objects.

Hit and poke take non-effective instruments

and thus allow the instrument to be become direct object.

Hit also

offers a further option, that of tacking on a Locative complement that gives us a sort of resultative location for the same participant. "John hit the ball across the park." Note that in this valence the instrument cannot become direct object.

We do not have:

"*John hit the bat

against the ball across the park.·· We could explain this by noting that

277

rn this valence there clearly HAS been a change in the "hittee", a change of location.

Then the instrument in

this valence is effective.

We shall return to these resultative locatives in Chapter 4. 3.2.3

Experiencers and Causers Suppose we

admit

another semantic category often

invoked in

Filimoriar• Cc.se Grammar, that of the experiencer, taking this to be a necessary participant role for any verb involving perception, emotion, We have already recognized the features actor, causer,

and cognition.

and changed as playing an important role in the selection of subjects. Suppose we now add experiencer to the list. Here are some examples: ( 16) Mary sees Bilbo. ( 17) Bilbo suprised Mary. ( 18) Mary thought about Bilbo. ( 19) Mary liked Bilbo. (20) Bilbo pleased Mary. (21) Bilbo seems funny to Mary. (22) Bilbo's soup tastes funny to Mary. If the various realizations of experiencers in these examples are to be accounted

for,

we will

participants with verbs

experiencer.

need

to

identify clearly the roles of those

like surprise and please which

outrabk the

One move in the right direction is made in Fillmore 1971.

278 Fillmore distinguishes those verbs where the "intentional object" (the content of the cognitive state,

the perceived,

or the target of the

emotional state) can be thought of as causing the intensional state.

If

the intentional objects of these verbs are called causers and if causer outranks

experiencer

intensional

objects

experiencers as

in

the

which

are

subjects.

ordering

causers

of

roles,

should

then be

Some of the relevant

just

those

selected

over

verbs:

surprise,

astonish, disgust, thrill, amuse, excite, sadden, impress, please and stun. Note that these verbs allow complements with by: (23) John amazed me by arriving on time. (24) Bill pleased Roberta by arriving on time. In contrast we have: (25) *Tina liked Bill by focusing on his strong points. (26) *John believed Slim by performing an act of faith. In

Chapter 1,

PURSUITS.

we called the situations these by adjuncts selected

If we suppose that PURSUITS contain causers, then the

fact that the experiencers aren't subjects with these verbs and the fact that they allow by adjuncts can be explained together. Along with causer noted was actor.

another

Clearly,

high-ranking

Saliency feature we

certain intentional attitude verbs involve

participants we would like simply to call actors; these, too, outrank the

279

ex periencer s: (27) John persuaded Bill classical arguments.

to

use

case

grammar

with

several

(28) John persuaded Bill to use case grammar for his advisor. Thus, experiencer seems to be involved in at least the following orderings: Actor· > Experiencer Cuaser > Experiencer Clearly problematic for this simplistic account are the verbs like

seem and taste. yet

experiencer

Neither of the ordering statements appears to apply, rates

only

an

oblique

realization.

Moreover,

the

behavior of these verbs is something more than a case of random lexical misfire.

They appear to be examples of a phenomenon observable in a

number of

languages,

both

Indo-European

1977), which is sometimes called inversion.

and

others

(see

Nichols

Loosely speaking, inversion

involves verbs in a number of semantic domains, including possession, perception,

and

modality,

with

animate

arguments that are

obliquely, with the subject often a dummy or absent.

realized

Some typical

examples from Polish:

(29)

Mnie Me (Dat.) I'm cold

zimno cold (neuter)

jest is (3rd sg)

(30) Nam jesc kapusty. nie wolno Us (Dat.) not permitted(3rd sg.) eat (inf.) cabbage (gen.)

280

We're not allowed to eat cabbage. The term "inversion" is motivated by the fact that these oblique animate arguments exhibit a number of subject properties, including, as in the second of the Polish examples, control of infinitives. property

is

extensionality:

for

note,

Another subject

seem

that with

example,

the

subject position is not extensional, as is often the case with raising but the experiencer

''),

verbs ("A unicorn seems to be

approac.loi!"o~.

argument (realized with to) is.

Moreover that position strongly favors

definiteness, another property typical of subjects: (31) *?A unicorn seems to a centaur to be approaching. (32) A unicorn seems to me to be approaching. Perlmutter 1975 postulates that these arguments begin as subjects, but that there is a syntactic inversion rule causing them to be realized as surface Indirect Objects.

A general problem for the GPSG framework to

which I have hitched my wagon, and for the Dowty-style representation of grammatical

relations which give

phenomena

that

relations,

pa rticu Ia rly

it adopts, for

evidence where

multiple

obi igatory

Inversion is one such phenomenon.

is how to handle syntactic levels

processes

of

grammatical

have

applied.

I have little to say about it here,

except to note that it shows both syntactic and semantic symptoms, and that a general account of the semantics of nuclear terms must ultimately find an account for it. What I propose to do with verbs like seem and taste, then, is

281

class them among the myster·ies.

The only truly unsettling thing about

the particular mystery of inversion is that it is not very clear,

in

English, just what its boundaries are. Are the following examples cases of inversion? (33) It takes me four hours to dr·ive to Eureka. (34) It is amazing to her that you even came. The it

preceding

take

is

presumably

subject is the infinitive clause.

experiencer?

extraposition-it,

so

that

the

Why does this clause win out over an

A parallel question arises for amazing. The appeal to a

causer element is plausible in sentences like "John amazed Mary with his impudence," but (34)

clear·ly has an adjectival head (only the present

participle form of amazing takes the preposition to).

The problem is

that in 3.3, when we do consider hierarchical rankings for adjectives, we shall want experiencers at the top;

Let

ignoring that point,

never have causers,

seems likely that adjectives stative.

even

it

because they are

Why, then, does the experiencer occur obliquely here? me

now

back

up

intentional attitude verbs. we call an experiencer.

and

summar·ize the

current

state

of

All of them by definition involve something If there is no causer or actor element in the

same clause, the experiencer becomes subject.

Complicating this simple

picture, there is a special class of verbs called inversion verbs which seem to be immune to this

rule.

The class of exceptions (if the

inversion verbs are indeed exceptions) shows some evidence of being

282

semantically characterizable.

Apart from these verbs, there is a large

class of verbs whose subject ar·guments can be correctly characterized with the notions, actor, causer, and experiencer. 3.2.4

Summary In

this

section,

we

have

used the

notions

actor,

changed, and experiencer to make some fairly broad about what

kinds

causer,

geo1eralizc.ti.::.i1s

of

semantic features promote subject- and direct

object-hood in English.

Before that we noted that that the role figure,

already invoked in Chapter 2, also played a role in some significant semantic observations about subjecthood.

In the next section we will

discuss Fillmore's proposal for a Saliency Hierarchy which uses notions like actor, causer, experiencer and figure to state facts about the semantics of nuclear terms. 3.3

The Hierarchy and its Function One can distinguish among a number of different theories of deep

case

two

function

served

by

the

cases.

First,

a

paradigmatic

classification of lexical heads according to the semantic roles played by their nominal arguments; semantics.

this enterprise falls into the domain of lexical

If successful,

such a classification elucidates the lexical

semantic structure by defining a useful kind of lexical relatedness; it can capture natural verb classes, predict entailment relations, and help reduce

the number of lexical entries.

But case theories were never

283

meant to complete the task of lexical semantics.

In a representation

rich enough to show the detailed local relationships of verbs like, say,

buy and sell, a good deal of extra and very specific semantic apparatus is needed. such

One of the goals

specific

semantic

of this dissertation is to show how having

apparatus

around

helps

us

formulate

the

generalizations of case grammar in a testable way. The second role of deep cases was our principal concern in the last

section:

to

represent

semantic

generalizations

realization of grammatical relations in clauses.

about

the

This enterprise takes us

away from the mere cataloguing of situation types and involves us in the problem of the possible linguistic encodings of those situations.

A

theory that concentrated primarily on this second function of case roles might well omit them entirely from specific lexical descriptions, might constrain the

and

use of classical case roles to its metalanguage.

This seems to be one reading of the theory presented in Fillmore 1977a and Fillmore 1977b.

In Fillmore 1977b, a procedure for selecting

nuclear arguments is formulated, and case roles as such play no part in it.

Instead

an

assumption

is

made

that

there

are

semantic

representations rich enough to encode the kinds of features discussed in the last section.

Those features are ranked in a hierarchy which

governs the selection of nuclear terms. Saliency Hierarchy:

Fillmore calls his hierarchy the

284

(1) An active element outranks an inactive element.

(2) A causal element outranks a noncausal element. (3) A elements.

human

(or

animate)

experiencer

outranks

other

(4) A changed element outranks a nonchanged element. (5) A complete or individuated element outranks a part of an element. (6) A "figure' outranks a 'ground.' (7) A 'definite' element outranks an 'indefinite' element. These seven statements 1 are consulted in order and when one that fits is reached, the ranking element becomes nuclear if a nuclear slot is still available.

If we think of this as a formal

procedure, then of course

everything depends on the particular choice of lexical representations, Fillmor·e

and how the necessary information is extracted from them. does

not

make

a

commitment

about

the

exact

nature

of

those

representations, but he does assume that lexical representations proper are

frames

actions,

linked with schemata that charaterize the structure of

institutions,

and

objects

in

the world.

In

sum,

although

Fillmore makes none of the notational choices subscribed to here,

this

venture in lexical representation is in very much in the spirit of the program outlined there. It remains to talk a bit about the hierarchy itself.

1.

Fillmore 1977b, p. 102.

The seven

285

kinds of choices indicated

by the hierarchy can be illustrated with

some examples from the lore of case grammar (where NP's denoting elements relevant to each choice are italicized): (1) The man hit the fence. (2) The duck surprised John. (3) John Ii ked the duck. (4) John loaded the truck with hay. (5) Bill hit the car on the bumper. (6) The ball is on top of the table (i) Roman made the log into a canoe.

The

hierarchy is doubtless incomplete, and some details of its

formulation may be open to question, but it offers, I think, just

the

right sort of scaffolding on which to build. There are a number of advantages to approaching the semantics of nuclear terms with a saliency hierarchy instead of a more classical case theory.

The difference here is not simply in the choice of a more

abstract set of features; analyst. relations,

It is in the kinds of decisions forced on the

Rather than seeking an exhaustive taxonomy of verb argument we are seeking subject-inducing semantics.

When we finish

listing the statements in a Saliency Hierarchy, there is no need to claim that we have defined some set of primitive semantic entities which must play a role in the semantics of every open class lexical item.

It is

286

perfectly consistent with the enterprise to have verbs that lack any semantic features mentioned by the hierarchy. Hierarchy simply has nothing to say.

For these verbs the

Indeed, the fact that Hierarchy

does not completely determine subjects and objects for every verb will play an impor·tant role in our char-acterization of possible valences. The in he rent shift in the natu r·e of the enterprise be best illustrated with an example. us a

might perhaps

The example will also provide with

addition to Hierarchy to use in the revised Hierarchy in the next

section.

Fillmore tells the story of the early days of case grammar

when

skeptical

a

Paul

Postal

conceivably assign to a verb

asked

deep

like outnumber.

framework there was no easy answer. of semantic roles, one is only

what

In

cases

he

could

the then-current

If, instead of a small inventory

interested in the semantics of nuclear

terms, that particular question does not arise.

The relevant question

is: what if anything is there about the semantics of outnumber gives it the particular choice of subject that it has? The question could be rephrased: that means "is less than

in

number,"

"Could there be a verb blik whose subject was the lesser

object and whose object was the greater one? this

If the answer is no, then

is a fact about the semantics of subjecthood, and needs to be

represented in the Saliency Hierarchy. The constraints imposed by

the

Hierarchy

presumably

apply

to

all

287

verbs,

both

among

the

basic and derived. ordinary

verbs

and

prefixation, that apply to them.

Hence we can find our data both the

morphological

processes,

like

The verb outnumber is the output of

a productive prefixation process in English.

Out-

roughly means "to

exceed in," and occurs quite productively, in words like outnumber,

outshine, outplay, and outeat.

Bresnan

1980 argues

that

t,·ar.::.itivizing prefix which attaches only to intransitive verbs.

2

it

is

a

What is

important here is that there is no corresponding prefix meaning '"to do less well in.

In general it seems that verbs whose content implies

ranking on some scale, whether of quality or quantity, will take the superior element as their subject. to

We will capture such facts by adding

the Saliency Hierarchy the clause:

The Superior outranks the

Inferior. The claim that the Saliency Hierarchy must also apply to the output of the mor·phology is quite important, considering the general unpredictability of the semantics of morphological processes.

Despite

the enormous range of semantic creativity discussed for denominal verbs in Clark and Clark 1979, it seems correct that one kind of use that

won't occur for "The Duchess teapotted the dean is one in which the dean

instigates some action with a teapot.

The same kind of evidence

2. outnumber and outweigh may seem like counterexamples, but they can quite plausibly be called intransitive, since their complements are exclusively quantitative, and they do not passivize.

~88

can be adduced for our proposed addition to the hierarchy.

dwarf,

despite its meaning as a noun,

magnitude as its subject.

Compare tower.

Note that

takes the entity of

greater

28811

3.4

Some Issues in the Semantics of Nuclear Terms This section ties together a number of loose ends that need to be

addressed

before

actual

the

present

we

subcategorizations in the next section.

procedure

for

assigning

We will begin by examining

possible extensions of the Hierarchy to other categories besides verbs. We will then present a long-overdue discussion of optionality, taking as our paradigm example the case of the instrumental.

Next, we will turn

to a subcategorization puzzle raised in Kajita 1967 concerning the verb

serve,

and

finally

POS_ TRANS

scene,

we

present

will

for

crucial

a

further

developments

decomposition in

of the

Chapter 4,

which

extends the current interpretation of the roles figure and ground. 3.4.1

Cross-Categorial Claims Throughout this dissertation I have concentrated mainly on verbs,

but have maintained that the same principles and methods would car-ry over to the other lexical heads,

the adjectives and the nouns.

In

dealing with Saliency Hierarchy, however, I need to revise this position somewhat. It has been fashionable, particularly among proponents of X-bar theory, to draw an analogy between subject noun-phrases and genitive noun-phrases in noun-phrases.

in sentences

Jackendoff 1977 formalizes

the analogy by assigning both nouns and verbs the feature [•SUBJ], and

defining

subject

as

a

noun

phrase

immediately

dominated

by

289

leaves

This adjectives. analogy

somewhat

as

problematic

the

status

of

One of the original motivations for drawing the noun/verb place goes all the way back to Chomsky 1970,

in the first

where the problem of representing the relationshsip of such heads as

criticism and criticize is addressed.

Opting against a transformational

account, Chomsky suggests that the lexicon contain entr·ies unspecified for

features

the

distinguishing

nouns

and

ver·bs.

In

effect

this

provides a place to record the shared selectional restrictions of the two Jackendoffs

forms.

definition

of

subject

provides

the

necessary

syntactic account of which argument positions correspond in the fully specified versions. Unfortunately

for

such

a

syntactic

problems ar-ise in connection with adjectives.

account,

the

very

same

The adjective perverse

corresponds to the noun perversity, and this time it becomes impossible within the system of projection rules Jackendoff adopts, to state the parallel:

1

(1)

John is perverse

(2)

John's per·versity

(3)

I consider John perverse.

Jackendoff adopts a very different kind of account here, claiming that for adjectives the correspondence between noun and adjective arguments

1.

Jackendoff 1974, p. 502.

290

must be stated at the level of thematic relations. The same sorts of problems do not arise in the framework adopted here.

have assumed logical predicates

with

argument positions to grammatical function. state correspondences

on

argument

fixed

mappings

from

Therefore I am free to

positions.

Thus,

can

define

subject to be a particular argument position for nouns, verbs, and adjectives, or indeed, point

is

that

the

something more complicated,

intrusion

if needed.

the verb be between an

of

The

attributive

adjective and its "subject" need not be reflected in the logic. Nevertheless, there are some problems in pursuing this analogy down the line. head

noun.

First of all, if a genitive is to be an argument of the

what

are

we

to

do

with

the

head

nouns

which

lack

genitives? Either, "the destruction" and "the city's destruction'' involve different lexical

items

as

their

heads,

or the first

has

a

missing

argument, existentially quantified over or filled with a deictic variable. And do we strive for complete uniformity? Is ''John" an argument of "lamp" in "John's lamp?" One

solution

to these problems is to generate more than one

logical predicate from a single

figure scene.

Thus the lexical semantic

representation for destruction might yield two logical predicates, one for

genitives,

one

not.

Then

"Tyre's

destruction"

and

"the

destruction'' would involve different predicates, but would map to frame

291

representations that were AKO the same lexical frame, DESTRUCTION. We have already had recourse to multiple logical predicates for the same lexical representation: this was the move taken for the argument and adjunct predicates of prepositions like against; we will take the same course with ver·bs when we tur-n to dative movement. I now recommend it wholesale for- nouns.

But what about nouns like lamp? I believe that

the right answer is that the English genitive is ambiguous between a lamp should map to only

grammatical and a non-grammatical function.

one logical pr·edicate because the lexical scene for LAMP countenances And whatever restrictions there are on

no extra argument positions.

what can occur as possessor· of a lamp are determined by non-lexical factors. The

road

thus

seems

open

for

a

complete

cross

categorial

definition of the notion ··nuclear term.··

I will not, however, carry

this enterprise out in all its gory detail,

because of two complicating

factors.

First, there is good reason to believe that differ·ent categories

require different saliency

hierarchies,

that

this

may

essential par·t of the "semantics" of syntactic categories.

even

be

Second,

an an

explicit semantic account of genitives as subject in the noun phrase is a major enterprise,

involving problems in the semantics of control that

would take us too far afield here. The second point is just a plea for clemency. some evidence.

The first requires

292 The relevant evidence is to be found in Amritavalli 1980; using a framework of Jackendoff-style thematic roles,

Amritavalli argues that

the privileges of subject selection are different for nouns and adjectives than for verbs. (4a) (4b) (4c) (4d)

John amused the children with his stories. *John's amusement of the children with his stories. The children's amusement at John's stories. The children were amused at John's stories.

Once again the semantic class at issue is what we analyzed as the causal

intentional attitudes in the last section.

example

(b),

where

the

noun s

The problem is that

subject corresponds

to

subject in (a), is ungrammatical.

Instead we have (c).

If, with Wasow

1977,

m

we call the head amused

formation

rule,

then

pattern together.

(d)

the output of

the

an

verb's

adjective

the noun amusement and the adjective amused

If this were an isolated instance it would have little

consequence for· Amritavalli"s theory or anyone's.

However, she cites a

large body of nouns and adjectives, all in the same semantic class, that behave identically.

A sampling of her data:

(5) annoyance, astonishment, boredom, conviction, delight, disappointment, disgust, dismay, distress, elation, emba rassment (6) annoyed, astonished, bored, convinced, delighted, disappointed, dsigusted, dismayed, distressed, elated, embarassed. In the account we have been pursuing, we could describe this data by having a different Saliency Hierarchy for nouns and adjectives. All we

293

need do is move the statement involving experiencers to the top of the list, interpreting it as the highest ranked of the roles. think

Amritavalli's

category-independent Saliency

data

shows

Hierarchy.

we

can

not

have

a

In the remainder of this

section, I want to explore some possible consequences of this view in the cbmain of nouns, before,

where Amritavalli's case is stronger (as noted

it may be that we do not want our lexical representations for

adjectives to ever include the roles actor or causer, if adjectives are In particular, I want to explore the possibility that for some

stative). categories,

not only is the Saliency Hierarchy different from that of

verbs, but there are more ways of realizing a particular nuclear term. The obvious candidate with nouns is the grammatical function that can be realized in either of the following ways: (7)

John's arrival

(8)

The arrival of John

Suppose that the Saliency Hierarchy for nouns was something like this: ( 1) An Experiencer outranks a non-experiencer. (2) A Changed element outranks a non-changed element. (3) An Actor element outranks an Inactor element.

.294

(4) A Causer element outranks an Non-Causer element. The Hierarchy may or may not continue, but this is all we need for now.

The most str·iking thing about it is that I have moved not only

the statement about Experiencers, but also the statement about Changed elements. Let 'Js assume that the notion transitive may be carried over to from

verbs

to

nouns,

where

a

transitive

noun

realizes one of its

arguments as a genitive and the other as object of the grammatical preposition of.

I propose that nuclear terms in nouns be r·ealized as

follows: the grammatical possibilities of the top-ranked argument differ depending

on

whether

intransitive nouns,

the

noun

is

transitive or intransitive.

the top-ranked noun may

object of the case-marking preposition of.

either

be

genitive

For transitive nouns,

For or the

genitive slot will be taken up by the second-ranking noun, and the top ranking

noun

will

case-marking of.

only In

have

one

possible

realization,

as object of

the examples we have considered, then, of always

functions as a grammatical preposition marking the top-ranked noun (in Chapter 4, we will examine some cases where of has meaning). obvious

inspiration

languages.

for

split

marking

convention

is

ergative

As we shall try to show below, this slight complication of

the grammatical mapping ''valence"

such

The

possibilities for

simplifies nouns.

the

description

of

the

numerous

Taking as our paradigm case,

destruction, the hierarchy allows (a) - (c) as basic realizations of the

295

It blocks (d) on the r·eilding

noun, without any syntactic derivation. that paraphrases the other examples: (9a) (9b) (9c) (9d)

Tyre"s destruction by heaven The destruction of Tyre by heaven Heaven"s destruction of Tyr·e. *The destruction of heaven ( Tyre).

I am assuming that in (a)

and (b), the option to have an intransitive

noun hds been taken, and that by in the noun phrase is a semantically an

marking

argument.

oblique

the

(c)

In

contentful

preposition

transitive

option

possible.

The intransitive option (d) is blocked because the destroyer

would

to

have

is

and

taken,

outrank

the

only

syntactic configuration

one

Note

destroyee.

that

is

saliency

any

hierarchy in which actor outranks the changed in nouns would have to resort to some extra mechanism to block (d); this is the chief pay-off in an ordering of the noun hierarchy which

ranks changed elements

above other·s. The following examples would then seem to be problematic: (lOa) (lOb)

The shooting of the deer The shooting of the hunters

( 11 a) (llb)

The approval of the program The approval of the deans

If program

and

deer

are

assigned

the feature

changed in

these

examples, then both the (b) examples are the sort that did not occur with

destruction.

If

the

role changed

is

not

involved,

then the

examples are not problematic; let us for the sake of discussion assume

296

For approval and shooting, then,

they are.

be omitted while the other element is not.

the changed element can The way to handle these

examples within the framewor·k of the cur-rent proposal is to have the

changed element be semantically optional for both these nouns.

\\'e will

deal with the question of optionality in somewhat greater detail in the next section, but the relevant point here is this:

when an element is

marked as O!Ji.ional rn a lexical representation, the grammar may treat that lexical item as if the element wer·e not present at all. Hierarchy may then safely ignore it. little.

Now

in

order

to

account

The Saliency

This move complicates things a for

the

non-occurrence

of

the

destruction of heaven in the relevant reading we need to say that with destruction the changed element is semantically obligatory.

This may

seem a peculiar claim, given noun phrases like the destruction, but we will establish a general basis for making such claims in the next section when

we

draw

obligatoryness.

a

distinction

between

semantic

and

syntactic

The case of destruction will be taken up then.

In sum, what I am proposing is that the Saliency Hierarchy the mapping into the syntax be arranged so as to give nouns options than verbs.

conventions

"intransitive" position

nouns

more

The result would be that the scene DESTRUCTION

could gives rise to three separate logical predicates. predicate

and

for map

nouns into

is a frame-position;

quite

parallel

two-place

"transitive"

to

Let us assume those for verbs:

predicates,

one

of

whose

nouns map into three-place

297

predicates. may

Nouns also have an option which verbs do not have; they

have all of

their core

non -argument-taking

participants

nouns

DESTRUCTION, for example,

map

into

realized

obliquely.

one-place

yields three predicates,

one three-place, and one two-place.

Such

predicates.

one one-place,

Some examples:

(12a) The destruction wa~ awful. ((Tt1E X li)~SIRUCTIONl X)) (AWFUL X)) (12b) The destruction of Tyre by heaven was awful. ((THE X (AND (DESTRUCTION:! X Tyre) (BY x heaven))) (AWFUL X)) (12c) Tyre"s destruction by heaven was awful. ((THE X (AND (DESTRUCTION:! X Tyre) (BY x heaven))) (AWFUL X)) (12d) Heaven"s destruction of Tyre was awful. ((THE X (AND (DESTRUCTION3 X Tyre Heaven))) (AWFUL X)) The most unconventional choice I have made above is to repr·esent versions of destruction often assumed to be the output of passive as in some sense basic.

That is, I posit no relation-changing rules that link

these different predicates; they exist solely in virtue of

the option in

the grammar to '"create"" both intransitive and transitive nouns from the same semantic material.

Again,

I have taken the choice of creating

several predicates from a single lexical representation, as adjunct and argument prepositions like against.

did with

In doing this

I am

exploiting a possibility generally available to us once we recognize a level of lexical

representation,

the possibility that the grammar will

298

leave us certain options in the mapping to grammatical form, and that sometimes multiple options may be taken. like rules

Thus, alternations that seem

need not in any formal sense be rules at all.

They are

simply areas of flexibility.

This will be the account adopted for Dative

Movement

of

and

promotion

Instruments

to

Objects

as

well.

The

-;trongest argument I know for placing these alternations on a different level from

relatiur.-changing rules like passive is the morphology.

In

particular, there is no mor·phological reflex for the "passive" in noun ph rases, or Dative rv1ovement in verbs. The

same

case

(cross-linguistically)

for

instrument-promotion

and will

be discussed in

is a

more

bit more

difficult detail

in

Chapter 4. One of the chief advantages of this treatment is that, without a syntactic

passive

rule

for

nouns,

we

do

not

have

to

explain

a

semantically systematic class of ··exceptions": (13) *John's love by t\1ary (14) *Kant's perception by John (15) *physics's knowledge by John These facts fall right out of the noun hierarchy as given. A final point to be taken up about nominal passives is by. analysis

sketched

above

assumes

that

by

in

contentful a preposition as, say, against in verbs.

noun-phrases

The is

as

It has been argued

299

(Bresnan 1976) that by can always be analyzed this way, on the basis of examples like: (16) A symphony by Beethoven The problem with

this

claim

is

that

Passive affects

a

bewildering

semantic variety of ver·bs, and the roles of the "demoted" subjects are simply too varied to be characterizab!:! with anything but the vaguest meanings.

To bor·rrow an example from Anderson 1977:

(17) The top of the hill was occupied by an obelisk. thus prefer to call by with

verbal passives a grammatical preposition,

and limit its active semantic life to marking actor or causer with nouns. That by occurs in this sense in nouns in other than passive contexts is shown by both the Beethoven example above and others like: (18) The decision by I Bl\1 to go public Note that if nominal by has no grammatical status, it should compete with other semantically appropriate prepositions to mark complements. This seems to be the case: (19) The order by the general to leave. (20) The order from the general to leave. Note that in

() and

marked with of.

() the participant marked with by can also be

This can be accounted for by allowing the lexical

representations of decision and order to be realized both as intransitive nouns and non-argument-taking nouns.

The intransitive versions take

300 grammatical of; the non-argument versions take by. thus exhibit two of the three possibilities

Decision and order

exhibited

by

destruction

above. If the distribution of by in nouns is semantically governed, then it ought to show a more restricted distribution in nouns than in verbs; the emotion nouns in () constitute an entire cl;-ss whose verbal passives take by and whose nominal "passives" do not.

Less systematically,

there are nouns which exhibit passive-like valence alternations, neither of which has any grammatical claim to being more basic, and neither of which uses by: (21a) John's gift to Deirdre (21b) Deirdre's gift from John (21c)*Deirdre's gift by John (22a) Joan's interview with John (22b) Joan's interview (22c) *John's interview by Joan hope to have shown that the idea of different

By this point,

saliency hierarchies for different syntactic categories at least has some interesting possibilities.

I have presented a case in some details for

nouns, but Amritavalli's evidence argues that a similar case could be made for adjectives.

If that is so, the treatment proposed for nominal

passives could be extended to adjectival passives, including those cases discussed in Wasow 1977 as instances of a second "lexical" passive rule which creates adjectives.

301

There is an interesting complication in the Saliency Hierarchy for adjectives, though.

If we followed a line somewhat like that which we

followed with nouns, we would simply take the Experiencer statement in the Hierarchy,

and move it to the head of the list,

statements involving actors and causers.

ahead of

the

But there is some question

whether adject:ves ever involve actors and causers, given their stative natur·e,

and

thus

whether

a''Y

!.ubstantive

change

to

the

verbal

hierarchy is necessary for adjectives. A final point to note in passing is that some such division among categories is necessary if our proposed addition to Saliency Hierarchy, "The Superior outranks the Inferior," is to stand.

This is because

there are both adjectives and nouns whose subjects can rank lower on some scale than an oblique argument. (23) Two is less than three. (24) The inferiority of object to subject Thus the scope of this Saliency Statement must be limited to verbs. 3.4.2

Optionality and Instruments One last piece of business needs to be transacted

return to the hierarchy itself.

instruments.

This

before we

is to revise our treatment of

Until now we have been treating instruments just like

other adjuncts. The problem with this is that, as we have seen, they can sometimes become nuclear terms.

We thus want to grant them

302

genuine ar·gument status so that we can manipulate them just along with other arguments in the hierarchy.

We need to do this in a way that

does not lose what we had captured in the previous account, the fact that the actor was understood as manipulator. at all,

Also, to get anywhere

we are finally going to have to say something about optional

arguments. That optionality That

there

need

overlooked.

to

exists be

is

two

uncontroversial

kinds

in

of optionality

anyone"s is

point

often

To make this point it will be useful to r·eturn to several of

the verbs of LONGING we discussed in chapter one.

apply,

a

theory.

wish

and

complements.

cannot

appear

discourse

I claim that

initially

without

try, their

This point can be nicely illustrated with a test originally

suggested in Panevova 1974 and elaborated in Peter Sgall 1980. following

The

exchange indicates that a misunderstanding or omission has

occurred: (25)

John applied

(26)

What did john apply for?

Compare: (27)

John read.

(28)

What did John read?

The question

in

()

can

be read as a

request

information the addressee may or may not have.

for

amplification

on

The question in (),

303

however,

had better be answerable, or else the first speaker had no

business saying that John applied in the first place. Another "well-formed" example might serve to make the contrast clearer: (29)

John loaded the truck

(30)

What die he load it with?

Here,

it seems to me, the fir·st speaker can felicitously reply:

don't know, but he had to do it all alone, because Mary wouldn't lift a finger.

That is, the contents of the truck can be utterly irrelevant

with load;

the karman cannot be with apply.

Another sort of test involves negation: (31a)

John hasn't applied yet.

(31 b)

John hasn't loaded the truck yet.

believe (a) can be uttered so that it commits the speaker only to a claim about a specific application.

That is, if the speaker utters (a) in

a context where applications for Florida trucking jobs are at issue, (a) will still be true even if John has applied for a Berkeley post-doc. contrast, even if the last hour's

In

conversation has revolved round a

half ton of bananas stored in a nearby warehouse, (b) is literally false if John has packed the salient truck full of oranges. The right way to talk about apply seems to be to say that it is

304

possible to omit overt syntactic expression of the karman, but that the

karman

must

communication) optionality

semantically

be

always

Fillmore

present.

(personal

has suggested calling such cases of purely syntactic

omissibility,

and

reserves

term

the

optionality

semantically optional elements like the direct object of read

for

and the

locative complement of load. Returning

to our example from 3.4.1,

destruction,

a

similar

distinction can be made. (3.2)

The destruction was wanton and willful.

(33)

The destr·uction of what?

Here, the first speaker must know of which carnage he speaks, or he is not engaged in cooperative communication. of destruction here which is not issue,

There is a related sense

meaning something like the

substance resulting from the destruction, or· perhaps, in some poetic sense, the effect itself:

"The destruction went on for miles."

One can

imagine such a description uttered by a speaker confronted with miles In

of rubble, with no idea what structures once stood in its place.

this use destruction seems to be a mass noun denoting a substance (perhaps an abstract substance), representation

would

and presumably the correct lexical

be one without any

core

participants.

What

concerns us here is destruction in the use where it denotes an event. Having drawn this distinction between omissibility and optionality

305

we return to instruments and ask, which are they?

The Panevova test

tells us they are truly optional. There is no misunderstanding or failed communication in the following exchange: (34)

Bill hit John.

(35)

What did he hit h1m with?

proposed that we represent such truly optional element; as optional slots

in

frames.

Whereas

elements

that

merely

are

syntactically

optional, such as the karman of apply, will be represented at the frame level

as ordinary

slots;

is,

that

every

applying situation will have to fill its karman slot. the

karman

is

omitted

will

be

of

an

Sentences in which

treated

as

cases

of

Thus, the missing syntactic material will have to be

null-anaphora. reconstituted

syntactically

instance

satisfactory

in

appropriate type,

the

semantic

interpretation

with

objects

of

the

just as it must be in resolving any of the cases of

Anaphora Hankamer and Sag 1979 call Deep Anaphora. In the case of instruments I propose to represent the optional material not just as an optional slot, but as an entire optional subscene of the frame.

Consider the following ground scene:

(MANIPULATION: CAUSER-SEQUENCE [ANTECEDENT (DIRECTED-ACTION: ACTION; UNDERGOING (ACTOR ) (PATIENT))] (CONSEQUENT (DIRECTED-ACTION: ACTION; UNDERGOING (ACTOR)

306

(PATIENT)))) When all the slots are filled this scene works as follows: it consists of a first ACTION scene in which an actor does something to an instrument; then there is a second ACTION scene in which the actor does something When the first subscene is absent, matters reduce to a

to a patient.

actor doing something to a patient. The

representation

inspired

is

here

offered

by

comments

in

Fillmore 1977a that suggest that thinking of instrument as a special kind of patient may help explain its occasional elevation to direct objecthood. Here

I have represented instrument as the patient of an embedded

scene, with the "ultimate" patient lodged in a resulting scene.

Note

that the instrument is not a participant in that resulting scene.

This

choice has made a number of technical details of the treatment that follows

go

smoother.

But

it

is

also

supported

by

some

crucial

entailment facts: (36a)John hit the fence with the stick. (36b)John hit the fence. (36c)The stick hit the fence. It seems to me that (a) entails (b) much more naturally than it entails (c).

(c) seems to have two readings, one "personification" reading,

clearly not at issue, and another on which its strikes the fence not under any particular control; it may be flying through the air, agent

may

be

randomly

swinging

it

about.

Thus,

on

the

some true

307

instrument reading of (c), the crucial feature of control is missing, while on the true actor reading of (a) it is required.

John may hit the

fence accidentally, but he must be controlling the stick when he does. We will return to this concept of control below. We now need to articulate the figure scene for the instr-umental meaning of with.

The following would seem iike a good candidate:

(MANIPULATION_ T: CAUSED_EVENT (INSTRUMENT) (CONSEQUENT (01 RECTED-ACTION (ACTOR) (PATIENT)))) Here

CAUSED_EVENT

is

a

constituent

of

CAUSER_SEQUENCE

MANIPULATION_T is thus a constituent of MANIPULATION. scene for instrumental with, AKO this scene and will

call it MANIPULATION-L,

and

The lexical

will simply be

reference reference MANIPULATION,

without

acquiring any structure not in MANIPULATION_T. Note that optional; scene.

this,

nothing in MANIPULATION_T has been designated as in fact,

is what we want for the

preposition's

Let us adopt the convention that any particular figure scene

may require optional material in its ground scene to be realized. is,

figure

in effect,

This

a lexical choice made with respect to a particular scene

description. The treatment of instrumental with is intended to

be

subcategorized-for

or

to

be

an

adjunct.

to allow it either As

an

adjunct

308 instrumental with modifies only

ACTIONS,

the actors of those

and

ACTIONS must be understood as the manipulators of the instrument (because of the ground

MANIPULATION scene).

I assume that most

ACTION verbs take instrumental with as an adjunct:

break, turn,

destroy, fix, operate, write, and on indefinitely. want to

argue,

however,

that there

is

a special

class of

ACTION verbs which subcategorizes for instrumental with, namely those that take the kind of instruments we called non-effective in 3.2.2: hit,

touch, beat, slam, and

beat.

Fillmore has

called these

verbs

of

impingement. The principal thing distinguishing the impingement verbs from the other ACTION

verbs

is

that the

verbs

of

impingement

allow

the

instrument to become direct object: (3i) John hit the stick against the fence

(38) Mary beat her fist on the door. (39) lla slammed her shoe against the table. Presumably any participant that can be realized as a nuclear term is a core participant. assume then

that

the

specifications of MANIPULATION figure

scenes,

compatible

instrument is optional.

with

verbs

of

impingement

have

as their ground scenes, MANIPULATION_L,

in

further

and have which

the

309

But what about the fact that all ACTION verbs allow instruments to be realized as subjects in the absence of the actor?

The following

example is par-adigmatic. (40) The hammer broke the vase. suggest that because of examples like these the notion actor must be extended beyond animates to embrace any of t;-,e effective agents that can be subjects for these verbs.

This means that nothing will block

examples like this one, discussed in Fillmore 1971: (41) ?The storm broke the window with a tr·ee. This will have a coherent frame interpretation,

but it will be one in

which the storm manipulates a tree and in so doing breaks a window. That is, it will amount to a per·sonification of the storm. This subject

analysis

position,

amounts

and

to

claiming

collapsing that

other

instrument with

actor

in

instruments can

only

be

understood as objects manipulated by an actor in a complex event.

The

immediate consequence for the Saliency Hierarchy is that the distinction between causer and actor is no longer necessary, and we can get by with one less statement. One of discussed

is

the an

problems with such an analysis that ambiguity

in

the

sentence,

disambiguated in the following two texts:

"He

hit

Fillmore has the window,"

310

(42a) I shoved John hard.

He hit the window.

The glass shattered.

(42b) John was furious with me. He hit the window. shattered.

The glass

In both cases I will call John an actor, but in one case the physical event is one of his bcdy, not under his control, striking the window; in the other, some part of his body, or perhaps even an instrument, makes contact under his control. event an agent.

Note that I have described the notion agent without

reference to volition. a

This is because volition, although often cited as

distinguishing

feature

Let us call an actor in control of an

agents

from

instruments,

seems

to

be

orthogonal: In fact, adverbs of volition typically apply only to actors in In "John accidentally hit the window," John is unquestionably

control. an agent.

Only in the case of an entity in control of some course of

events does the question of some accidental result arise.

This is why

it is peculiar to have: (43) I shoved John hard.

He accidentally hit the window.

What I have to say about these example will apply only to the verbs of impingement.

The right description of these facts can, I believe, be

had in the current analysis if we take a particular interpretation of what it means for a constituent of a scene to be optional. What it means linguistically has, clear. say

hopefully,

already been made

But what it means conceptually has not. What does it mean to

that

a

HITTING

:nstrument is wielded?

scene

has

an

optional

subscene in which an

Does it mean that the actions described in

311

sentences with and without instruments differ significantly? better not.

Rather,

what it ought to mean

It had

is that the instrument

component may be unspecified and unindividuated in an actual HITTING situation.

On the other hand,

in an

APPLYING

situation,

both the

applicant and the karman must be fully individuated and specific. natural HITTING

physical difference here is scene

may

actually

be an

that

the blunt instrument

undetached

The in

a

part of one of the

participants, a hand or an elbow, and an absolute line cleaving actor from instrument may be hard to draw. Suppose

we

say

that

having

an optional

subscene means two

things: ( 1)

The participants of that subscene may be unspecified.

(2)

The procedures connecting that subscene with the rest of the scene are to be viewed as constraints on its individuation. That is, they only apply when the subscene is fully specified.

Given (2) and a MANIPULATION scene with an unspecified antecedent, we are free to imagine any prior scene we like.

Consider "John hit the

window,·· on the reading where John is not an agent.

Here we may

imagine a prior ACTION in which a distinct actor has done something to John, or alternatively, a simple UNDERGOING: (44) John slipped. On the

He hit the window.

reading in which John is an agent,

he is the actor of the

antecedent scene, and the patient will fall within the usual range of

312

blunt instruments, hammer, stick, elbow, nose. What

I essentially propose,

is that the notion agent, or

then,

actor in control, has no place in either the Saliency Hierarchy or the lexical description of these verbs.

Whether it ever has a place in

lexical description is an open question. to

find

a

verb

that

took

What would help decide us is

instruments

instruments in subject position.

but

would

not

allow

those

In fact there do appear to be such

verbs: (45) John passed Mel the sample with his forceps. (46) *The forceps passed f..1el the sample. To talk about the interpretation of an unindividuated prior scene will not do, of course, for verbs which do not have such a prior scene. This will presumably be the case for verbs that do not subcategorize for instruments.

What I would like to suggest is the relevant ambiguity

are far more difficult to get with the other verbs, and that this

is

instruments

of

further

support

for

an

analysis

which

treats

the

ordinary ACTION verbs as adjuncts. (47) ? I picked John up and threw him across the room. broke the window. (48) ?I picked the hammer up and threw it across the room. broke the window. A

final

point

that

needs

to

be

touched

on.

The

He It

above

313

representation for the MANIPULATION scene would seem to require that

instruments occur only with transitive verbs, since both an actor and patient are in evidence.

This can be fixed by having a kind of

ACTION called REFLEXIVE action: (REFLEXIVE-ACTION: Dl RECTED-ACTION (ACTOR) (PATIENT (=actor))) This will let us handle examples like "Superman walks with a cane.·· At the same time,

it provides a general account for the bizarreness of

notional instruments as the subjects of intransitives: (49) The cane walks If the above frame is a constituent of WALKING, the interpretation of (3-!) would have to be that the cane, by acting on itself, engaged in walking. 3.4.3

Kajita's Quandary and Subcategorization features It may be useful to consider an "application" of the theory of

instruments offered in the last subsection to a small puzzle articulated in Kajita 1967.

The problem is the verb serve, which

Kajita claims

subcategorizes for transitive complements. (50) The ice served to chill the beer. (51) *The ice served to chill. To subcategorize complements for arbitrary

syntactic properties like

314

transitivity

is

an

unpleasant

extension

of

the

subcategorization embodied in most generative theories.

account

of

In particular it

raises the question of just how non-local the syntactic requirements that a head makes on its complements can be.

Structurally, we have gone

from the requirement [_ VP] to [_ [V NP)].

The theory of strict

subcategorization proposed in Chomsky 1965, for example, did not allow this kind of specification. What alternative account can we offer, then, to cover his facts? Suppose we propose the following structures to describe serve: (ACTOR_IN_ACTION (ACTOR) (ACTION (ACTION (ACTOR )))) (SERVING: ACTOR_IN_ACTION (ACTOR_NOT _IN_CONTROL ) (ACTION (EFFECTIVE_ACTION_WITH OUT _CONTROL: ACTION; CHANGE (ACTOR_NOT _I N_CONTROL ) (PATIENT )))) First,

as the basic

structure on which

to found

our description,

I

propose a frame called ACTOR_IN_ACTION, which simply entifies the relation between an actor and the action that actor performs. basic

structure

of

this

scene

is

thus

of

the

right

sort

subject-equi verb, a relation between an individual and a scene.

The for

a

Other

Equi verbs that would be natural further specifications of this scene:

bother,

hasten, hesitate, offer, threaten, and venture.

SERVING

315

itself is a specification of the ACTOR_IN_ACTION scene which makes two further requirements on the embedded ACTION; first, it involves a change in the patient; second, the actor must be "not in role

which

have

simply

called

control," a

actor-not-in-control.

Actor-not-in-control will always be a subject, since it is an elaboration of the actor slot. The requirement that the embedded action involves change rules out cases like the following: (52) *The stick served to hit the fence. Recall that we have hypothesized that hit does not involve the role

changed, in order to allow the instrument to become direct object. Chapter 4,

In

we will discuss a number of inter·esting predictions that

follow if the role changed constrains a participant to be realized as a nuclear argument rather than obliquely.

In the case where some other

element was ranked above changed, this would constrain the clause to be transitive.

The crucial data here involves some somewhat marginal

examples, which may challenge Kajita's claim that the downstairs clause must be transitive: (53) ? The needle served to pierce through the cushion Under the proposed analysis of serve, such an example would also be a challenge to the general claim that realized non-obliquely.

a

changed participant must be

The main point here is that complicating the

316

syntactic theory of subcategorization is unnecessary in either case.

If

the above example is accepted, then Kajita's claim that serve takes only clauses

transitive

complementation

wrong,

is

is

more

and

desirable.

semantic

a If

account example

above

the

of

its

is

not

acceptable, then we may still have recourse to a general semantically stated constraint on role realizations: changed participants may not be realized

obliquely.

Nothing

special

has

to

be

said

about

the

subcategorization of serve. Calling the slot occupied by the controller both

the

embedded

treatment proposed.

toplevel

scene

actor-not-in-control in

and

the

is

not

crucial

to

the

~lost

importantly, there is no intended claim that

an equi controller must bear the same role in both the matrix and the embedded clause.

The fact that many equi verbs allow both actors and

embedded passives rules this course out. simplest of the alternatives.

Here, however, it seemed the

Note that serve does not allow embedded

passives: (54) *The needle served to be pierced through the cushion. (55) *The beer served to be chilled. Both

these

examples

are

ruled

out

by

the

requirement

that

controller must bear the downstairs role of actor-not-in-control. suffices to guarantee it must

be

"deep"

subject

in

the

the This

downstairs

clause: there are no verbs which, after passivization, have an actor in subject position.

317

In sum,

we have proposed a treatment of serve which explains without

the behavior of the verb at the level of the lexical semantics, any recourse to special syntactic devices.

The fact that serve requires

a two-participant clause as its complement is simply stipulated semantic

description

with

devices

its

in

have motivated in a number of

we

cases involving preposition descriptions. A Note on Figure, Ground, and POS_ TRANS

3.4.4

What

I want to

figure/ground

ranking

in

argue for in

current

our

section

this

is

hierarchy,

and

the

given

that,

some

given

plausible assumptions about the underlying structure of our POS_ TRAr\S scene, the patient in a POS_ TRANS scene outranks the donor. To motivate this,

we will

need to return to the

components

of

POS_ TRANS proposed in Chapter 2. ( POS_ TRANS: CAUSER_SEQUENCE (ANTECEDENT:COMPONENT (TRANSFERENCE (SOURCE) (FIGURE))) (CONSEQUENT:COMPONENT (POSSESSION (POSSESSOR) (PATIENT)))) The crucial thing point here is that TRANSFERENCE is a component of POS_ TRANS and TRANSFERENCE has a slot called had

in

mind

in

using

that

slot

name

in

figure in it.

TRANSFERENCE

What I

was

that

TRANSFERENCE be an extension of ORIENTATION_IN_SPACE. I am thus

318

proposing the following unpacked figure scene for TRANSFERENCE: (TRANSFERENCE: ORIENTATION_IN_SPACE (FIGURE) (SOURCE )) As before this frame will also be the figure scene for the preposition from.

Now I have claimed that TRANSFERENCE is a meaning of from

appropriate ev£>'1 for non-spatial verbs like buy.

Therefore, we will

need some ver·y general interpretation of ORIENTATION_IN_SPACE that does not limit it to conventional space.

That is, there must be other

kinds of spaces one can be oriented in besides conventional three-D Euclidean Space. Let

us concentrate here on the meaning of from

verbs like buy.

relevant for

For from to make sense with buy, there must be such

a thing as a POSSESSION space, a space in which there is no very useful

notion

of

distance,

but

where

corresponding to distinct possessors.

there

are

distinct

points

A preposition like over is limited

to other kinds of spaces, so that the sentence John," has no ''possessive" interpretation.

"T~e

book is over

In contrast we have, ''The

book is from John." Suppose we inter·pret that as meaning "The book was once in John's possession." Then a reasonable interpretation of from is that figure is

oriented with respect to a ground such that at some previous

time, figure occupied the point in space defined by ground.

In the

319

example, we mean POSSESSION space, and therefore at some previous decription

Note that this

point in time, John possessed the book.

incorporates an asymmetry between possessor and possessed.

In this

space, possessors are locations; possessed things are objects that can occupy

representation

we

make

this

TRANSFERENCE_GROUND

of

More

TRA~SFERENCE)?

important

can

How

them.

asymmetry

in

important,

how

the

scene

for

we distinguish

this

ground

(the can

the

from

TRANSFERENCE

VERTICAL_ORIENTATION_ IN_SPACE,

par·t of

asymmetry

symmetry

of

which was the ground scene for

OVER and UNDER? I propose that TRANSFERENCE_GROUND make explicit reference to the notion of a point-in-a-space, as opposed to the objects occupying it.

Suppose that TRANSFERENCE_GROUND had two slots in addition to

figure

and

ground,

namely

figure- point

and

ground- point,

corresponding respectively to the point in space occupied by the figure and the point occupied by the ground. ORIENTATION_IN_SPACE,

but there is

individuate them into their own slots.

These notions are implicit in no

reason

The crucial

TRANSFERENCE is that one of the objects in the fixed in space.

in that

scene to

difference for

orientation scene is

That is, if we are to think of possession as a kind of

space with points corresponding to possessors, then "John" defines a fixed point in possession-space. John

(including,

perhaps,

Albert,

Any object at that point belongs to who defines a different point

in

possession-space).

The

difference with

locating objects

in

physical

space is that in most contexts we don't think of objects as

defining

fixed points in physical space.

With all this in mind,

propose the

following unpacked ground scene: (TRANSFERENCE: ORIENT ATION_IN_SPACE (SOURCE ) (FIGURE-POINT (P.•\ST (=))) (GRGL.:i-.ID) (GROUND- POl NT( FIXED- LOCATION ))) Here I have made the figure-point equal the ground-point at some time in the past.

have also claimed that there is a functional relation

FIXED-LOCATION, FIXED-LOCATION

between must,

of

and

ground

cour·se,

be

set

ground- point,

to

the

contextually

appropriate space. One of the chief virtues of this definition is that it constructs a role source within the TRANSFERENCE frame, without any reference to motion. The basic asymmetry of from-ness is attributed to the fact that the ground must determine a fixed point in some contextually relevant space.

But the TRANSFERENCE_GROUND scene does not

corresponding fixed-point for the figure. possibility that the figure

may

possibly

select

any

This does not rule out the have

such

a fixed

another context; possessors can possess other possessors.

point

in

It is simply

that in a sentence like "This slave is from Mary," we are interested in Mary as "This

a possessor,

table

is

from

not the slave.

Rome,"

we are

Similarly, interested

in

in a sentence like, Rome

as

a

fixed

321

location, even though, in another context, the table can be viewed as a location. \'iith all this given, our description of TRANSFERENCE predicts that

figure

ground.

outranks

source,

because scene,

POS_ TRANS

The

source is

using

an

elaboration

of

as

a

TRANSFERENCE

compr.nent, will inherit this ranking, and thus, unless a higher ranking role intervenes, patient will outrank donor.

In verbs like buy and

sell, a higher ranking role does intervene; the paticipant that is donor is also an actor, and that participant becomes subject. One might now wish to extend this concept of possession space to the POSSESSION scene itself, also a component in POS_TRANS, and call the possessor a ground and the patient a figure.

Such a move has a

certain appeal, but it cannot be made in the POS_TRANS scene per se, if we are to use POS_ TRANS as a component in figure scenes, and if We would then have a scene with the

we are to maintain uniqueness.

role ground assigned to both the donor and the recipient.

In shor·t,

only one thing at a time can be regarded as ground in a figure scene or a component of a figure scene. What, then, is the motivation for having an asymmetric treatment at

all,

for

claiming

TRANSFERENCE

scene,

that but

there not

in

is the

a

ranking

chosen

POSSESSION

scene?

in

the The

relevant evidence is in verbs involving possession and tranference of

322 possession. (56a) (56b)

\Ve have the following kinds of pairs: John has a book The book belongs to John

(57a) Alan gave the book to Sue. (57b) Alan gave Sue the book. Pairs like have and belong suggest that neither possessor and patient are preferred for subjecthood by

th~

Hierarchy;

the English

Dative

Movement alternation suggests the same thing for direct objects.

\Ve

will take a closer look at Dative-Movement in the next chapter. 3.5

The Semantics of Valence In this

section we do

two

things.

First,

we

present

the

procedure that implements the statements of the Saliency Hierarchy, and determines, as far as it is possible to determine them, possible subjects and direct objects for verbs.

Next we present the procedure that

actually marks verbs for possible prepositional complements, selecting only

complements

which

are

semantically

compatible

with

the

verb.

These things taken together constitute a semantic characterization of a verb"s possible valences;

henceforth, we will

refer to the particular

configuration of nuclear terms and subcategorized-for complements that a verb syntactically selects as its valence. the current framework more than one valence.

As we shall see, nothing in

prevents a single lexical entry from selecting

323

A Hierarchy for Verbs

3.5.1

subsection

the

In Hierarchy

down

by

one

instruments

on

statement.

we

Before

whittled

the

presenting

Saliency

the

actual

oper·ation of the hierarchy on lexical entries, we will need to set aside two more statements, at least provisionally. To begin with, I c..k> not propose at this time to try to represent Clearly the appeal

the phenomenon of definiteness.

in the seventh

statement in the Hierarchy is to something that is not a lexical semantic category.

That Fillmore allows himself this is due, first of all, to some

major differences of emphasis

1n

the model he is building; his '"scenes··

are textual entities built up quite apart from the lexicon, or indeed, any linguistic structures. status as well,

Although my scenes have a non-linguistic

my proposal for implementing the Saliency Hierarchy

involves refer·ence to logical predicates,

and these have little status

outside the lexicon. I thus propose simply to eliminate the claim about definiteness from consideration at present. handle

such

phenomena,

Independently of lacking the means to

have

some doubts

handled with same kinds of mechanisms.

that

they

should be

Some very rough correlations

between definiteness and grammatical function have often been noted in the literature (for example, subjects tend to be definite), but there are clearly parameters other than definiteness and subjecthood involved.

324

David Justice and

Carl

Pollard

have independently supplied me with

examples of predicates that simply seem to reject indefinite arguments: (58) ?A unicorn was upsetting. (59) ?A man was fat. There are of course, numerous predicates that take indefinite subjects quite comfonably: (60) A man 1s sitting along in his room. The

question

particular

here

lexical

seems

semantics.

not

so

much

The

one

cases

of

that

Fillmore·s Mystery by Peter Sgall (Sgall, et a!,

subjecthood have

been

as

of

dubbed

1973), but which are

actually due to Gruber (see Gruber 1976, p. 140), add another layer of difficulty, since what is presumably the same lexical item has different definiteness requirements in different valences. (61a) Roman (61b)*Roman (61c) Roman (61d) Roman

made made made made

the log into a canoe. a log into the canoe. a log out of the canoe. the log out of the canoe.

Minimally,

there seems to be an

interaction

here

among

the

lexical

semantics,

the nuclear term assignments, and what Halliday calls the

Informational Level, that level at which notions like given and new come into play.

I propose that the Saliency Hierarchy as such simply state

constraints on the mapping from participant roles to grammatical roles, much as the original 1968 Subject Hierarchy did, without reference to any facts about the particular realizations of those roles in a sentence,

325

such as definiteness. Moreover, for simplicity"s sake, I propose to also omit statement 5 in the hierarchy, "A complete or individuated element outranks a part of an element.·· ear.

This covers examples like "I picked him up by the

The technical problem

is

simply

one of

creating

slots

represent "parts of' the contents of some of some other slot. an exercise that will not be particularly illuminating here,

that

This is

nor is the

semantics of the scenes that should be involved particularly clear to me. Below I list the statements left, along with frames that provide the necessary slots: 1.

An Actor element outranks other element.

2.

An Experiencer outranks other elements.

(ACTION (ACTOR) (PATIENT))

( INTENTIONAL_ATTITUDE (EXPERIENCER))

3.

A changed element outranks other elements.

(CHANGE (CHANGED))

4.

A Figure outranks a ground.

5.

A Superior element outranks an Inferior.

(ORIENTATION_IN_SPACE: THING (FIGURE) (GROUND)) (RANKING

326

(SUPERIOR) (INFERIOR)) The only scenes which the above hierachy will have anything to say about will

be those which are extensions of one of the above five

frames. The

implementation

is

rather

simple,

although

the

reader

interested in a close exa'l1ination of i.he te.;:,,ical details might wish to review 1. 9, where the definitions for predicate denotations are given. Recall that the kind of lexicon assumed there had lexical entries that associated

thr·ee

morphosyntactic

things,

phonological

a

representation,

and an

representation,

a

interpretation for the lexical

item, part of which is always the lexical semantic representation proper, a figure frame. What

the

hierar·chy

does

interpretation of the predicate.

is

place

some

constraints

on

the

The predicate denotation of a lexical

head that takes arguments is a sequence of a frame and slots. The slots in

the denotation of the predicate correspond directly to particular

predicate

positions,

and

those

positions,

because

of

the

phrase

structure rules that build up the semantics of sentences, correspond in turn to particular grammatical functions.

2

Thus there is a mapping

2. Currently, the possible grammatical functions that can be signalled by the position of a slot in a denotation include only subject and object. In Chapter 5, I propose a new addition to the list, XCOMP.

327

from certain slots to

The hierarchy defines

grammatical function.

which slots get top priority. This can be modeled with the following sort of procedure.

We

will discuss only verbs here, though the general procedure carries over quite straightforwar-dly to other categories, given a Saliency Hierarchy for those categories.

Hecall that transitive verbs w'li be three-place

predicates, and intransitive verbs will be two-place predicates. now engaged in

We are

stating what is predictable about their denotations.

First we recognize a certain stipulation.

Each lexical item begins by

being associated with a particular frame in the frame representation for our

English-speaking

Furthermore,

culture,

that

frame

the denotation of each

from its figure frame.

being

its

figure

frame.

lexical predicate is constr·ucted

The figure scene, then, is always given.

Let

us think of predicate denotations as sequences, with one-place predicate denotations

being

one-place

sequences

(such

denotations

will

necessary for non-argument-taking lexical heads, say, lamp).

be

Then,

for all open-class lexical items, the first member of the sequence will always be its figure frame. constrain

the

The task of the Saliency Hierarchy is to

possible choices

of other

members

of the

denotation

sequence, given that they must be chosen from slots within the figure frame. Imagine

that the

"choice"

between

a two- and three-

place

predicate (i.e., intransitive and transitive) is made in advance, that is,

328

either by stipulation, or by factors independent of the operation of the hierarchy. It will be helpful to consider the operation of the Hierarchy as a kind of redundancy rule, which fills in the missing parts of predicate denotations wherever possible -

in much the same way as phonological

redundancy rules fill in pr·edictable features. Consider a typical

say, run.

intransitive verb,

Its denotation

will be a pair consisting of the RUNNING frame and a slot, unspecified until our redundancy rule applies.

Call that slot the subject slot, since

it must correspond to subject position.

We must now find the subJect

slot among the actor slots in the figure scene. slots

that

denotation.

haven't

been

used

up

yet

in

By active slots I mean building

predicate

We shall also refer to active participants, meaning those

participants still associated with active slots.

Since we're just getting

started in this example, all the slots and participants active.

the

in RUNNING are

The implementation of statement one in the hierarchy, then, is

to examine the RUNNING scene, and if it is an extension of ACTION, make the actor the subject slot (note: actor will always be a toplevel slot in a scene that is an extension of ACTION). it

becomes

the

second member of the denotation

If there is an actor sequence,

and we

simultaneously render all the slots associated with the same participant in RUNNING inactive (so as not to use them again later, if we happen to be looking for more denotation slots).

Let us refer to the slots that

329

mark the same participant as some slot S mentioned in the Hierarchy as Thus, if RUNNING is an extension of actor, we put actor

the S slots.

into the predicate denotation and render all the actor slots inactive. is not an extension of

RUNNING

ACTION,

second statement in the hierarchy. of INTENTIONAL_ATTITUDE. slot,

then

we move on

to the

\Ve ask if RUNNING is an extension

If so,

we make experiencer

the subject

if it is still active, and render all the experier.r:. , [+V, -N, TRAN, LOC]»> There

is

a gross oversimplification

in

this

entry.

As

written,

all

participants must be realized (even if they are sometimes realized with differing sets of roles).

But we also have:

(83) John loaded the truck. (84) John loaded the hay. This is clearly not a case of what we called syntactic omissibility in chapter 3.

Thus, in order to permit these possibilities, we shall have

to complicate the

LOADING

scene by adding two separate optionality

components, one for the patient, one for the goal (just in case it is not also totally-affected).

This

can

be done using the indexed square

brackets we introduced in the treatment of optional instruments: (LOADING: APPLICATION;ACTION (LOADER [actor]) [ -2 LOAD [material]] -2 (RESULT: COMPONENT

382

(MOVEMENT: ORIENTATION_IN_SPACE [-2 MOVED [figure] ]-2 [ -3 GROUND ] -3)) [-1 TOTALLY-AFFECTED ]-1) Now the convention is that optional material may be discontinuous, and that all the instances of material between square brackets of the same index const1tute single optional constituents.

But different optional

constituents are presumably independent of each other. representation thus gives us eight possibilities.

The above

We need to make sure

that this frame, taken together with the above lexical entry, gets the right results,

in particular, that it does not allow incoherent results.

In the following list, "+" before a number n indicates a frame realization where the optional

constituent

indexed

n

has

been

included;

indicates a realization where the optional constituent has been left out: (1)

[+1, •2, +3). This is as before. It gives us sentences like "He loaded the truck with hay."

(2)

[•1, -2, +3). Not possible because the totally-affected slot contains a procedure that requires position 2 to be filled. goal = totally-affected.

(3)

[•1, -2, -3).

(4)

[+1, +2, -3). truck."

(5)

[-1,

truck.

~~'

•3].

Not possible for the same reason. This give us sentences like "He loaded the Sentences like "He loaded hay onto the

(6)

[-1, -2, +3].

(7)

[ -1, •2, -3]. Not possible with either predicate. automatically fills totally-affected.

automatically fills moved.

Sentences like "He loaded hay."

383

(8) Thus,

[ -1, -2, -3]. the

allowed

Not possible for the same reason.

above frame,

in

the

lexical

taken entry,

together predicts

with the

the

syntactic options

right

configuration

of

valences. One possible objection to this lexical entry is also an objection to the kind of entry we had for datives and section.

It

is

imping~rnent

verbs in the last

associated with two different logical

therefore with two different meanings.

Therefore,

predicates and

what we have is

merely an abbreviation for two lexical entries, no matter how you look at it.

This point of view, however, depends on the traditional model

theoretic view of what a meaning is. here,

In the framework being developed

logical predicates are a way of encoding certain facts about the

grammatical

function

of a

lexical

item.

These

predicates

are

not

semantic objects; rather they are closer in spirit to the F-structures of LFG.

Of course they do not encode as much information as a lexical

F-structure; theory

as

features.

many of the properties of a lexical item stated in that grammatical

function

But logical predicates

nuclear terms;

are

captured

here

with

syntactic

do critically encode information about

in this more modest grammatical domain they function

just as F-structures do; they are pointers from the syntactic realization of a lexical item to the right place in the semantics.

The real semantics

is the lexical scene. It is nevertheless true that we could have written the above

384

APPLICATION frame without the optional slot,

and written a lexical

redundancy rule that said, when A is a lexical item with the following semantics, then A plus a zero affix is a lexical item with with same semantics plus the slot totally-affected.

We would in that case want to

add totally-affected to the figure scene of A (and perhaps to its ground scene). Two questions arise here.

One: what makes a lexical entry a

lexical entry, and when do we have two?

Two: what does it really

means to have optional constituents, and why do we bother to have them? want to address only because it is a little simpler. representational ontology.

second

question

in

this

section,

One answer can be stated in terms of our Sometimes we wish to say that there is a

single scene which may or may certain type.

the

not have an

individuated part of a

But with or without that part an instantiating situation is

still the same type of scene. Another answer is more practical, from the point of view of the working grammarian.

Optional constituents in abstract scenes let us

make important generalizations about lexical items.

To have a generic

scene with an optional part shared by a number of lexical items means that for all those lexical items, that part is optional. to

argue that

a

scene

with

an

optional

Suppose we tried

constituent

is

really

an

385

abbreviation for two scenes, and that we could always say the same thing with two scenes.

Scenes are semantic objects.

Once semantic

objects are distinct there are no guarantees about their realizations. Each of those new pairs of objects might very well get pronounced in different

ways;

if

such

pairs of

scenes

are

phonological bundle, then that is an accident.

realized by the same Under a single scene

representation we have made the claim thilt semantically defined cases of optionality are not accidents. Looked at the second way,

frames with optional

constituents

really make a kind of morphological claim; semantic objects with and without a certain piece of structure get pronounced the same.

Looked

at the other way, they constitute an ontological claim; objects with and without this piece of structure are still the same kinds of objects.

One

of the underlying assumptions of this dissertation is that these two different ways of looking at things are really reflections of each other. Particular

cultures

recognize

specific

Those regularities define scenes.

regularities

among

situations.

Scenes in turn may define words.

Thus the fact that two distinguishable pieces of semantic structure are pronounced the same is evidence

(although

not conclusive evidence)

that they participate in a single scene. What I have argued, then, is that treating load as a single lexical entry is completely consistent with the direction we have been pursuing all along.

This

is nothing more than a

plausibility consideration.

386

What it means is that if we want evidence that load should really be viewed

as

two

evidence.

separate

entries,

we

will

need

purely

grammatical

The nature of the lexical representations themselves does not

decide the issue. Some evidence of that form will be considered in the conclusion of this chapter, and in Chapter 5. 4.2.2

Other rules

4.2.2.1

Container promotion

There is a rule completely parallel to Goal Promotion which has, as it were, just the opposite semantics.

Some examples:

(85a) He emptied the glass of milk. (85b) He emptied milk out of the glass. (86a) He drained the corpse of blood. (86b) He drained blood out of the corpse. (87a) He cleared the room of the junk. (87b) He cleared the junk out of the room. Where

Goal

Promotion

involved

movement

of

material

to

a

two-

or

three-D Goal, this rule involves movement of the material out of three-D spaces.

Let us call the space the container and the rule Container-

promotion.

Again, there are holistic and non-holistic readings.

In (),

the glass must be completely empty, while () may describe an ongoing or incomplete process.

It is apparent, just as it was with with in Goal

Promotion, that a productive meaning of the preposition is involved:

387

(88) This book is devoid of interest. (89) Tom robbed Lefty of his dignity. (90) The man is bereft of any human feeling. (91) She relieved the adjutant of all his duties. Let us represent this meaning of of with the following simple scene: (EMPTYING: THING (MATERIAL) (CONTAINER)) And let us choose something quite parallel to the APPLICATION scene to represent the common source for the Container-promotion verbs. (REMOVAL: RESULT; EMPTYING; UNDERGOING (MATERIAL) (RESULT: COMPONENT (ORIENTATION_IN_SPACE (FIGURE) (GROUND))) [TOTALLY-AFFECTED ]) Lexical entries will parallel those of the Goal Promotion verbs. Note that REMOVAL is so parallel to APPLICATION that it is easy to construct a scene that abstracts over both. for both FILLING and EMPTYING.

Call it

First we need a source

MATERIAL-TRANSFERENCE:

(MATERIAL-TRANSFERENCE: THING (SPACE) (MATERIAL)) (EMPTYING: MATERIAL-TRANSFERENCE (CONTAINER ) (MATERIAL)) ( Fl LLING: MATERIAL- TRANSFERENCE

388

(GOAL ) (MATERIAL)) Taking this route, both preposition meanings are further specifications of

abstract

a single

space.

involving

scene

the

relationship

of

material

to

This is a very simple example of the general pattern known as

antonymy.

Antonymic

crucial feature;

words

have

identical

semantics

and in differing in this feature

th~y

except for

one

set up a simple

paradigm of two members. Having FILLING,

we can

and REMOVAL. a

certain

supplied

a

common

source

for

both

EMPTYING

and

now also supply a common source for APPLICATION

Call this scene MATERIAL-MANIPULATION, suggesting

conceptual

connection

with

the

MANIPULATION

scene for

instruments: (MATERIAL-MANIPULATION; MATERIAL- TRANSFERENCE; UNDERGOING; RESULT (RESULT: COMPONENT (OR I ENTATION_! N_SPACE (FIGURE ) (GROUND ))) [TOTALLY-AFFECTED ]) Suppose now that APPLICATION is AKO this scene and simply specifies

space to be goal, and that REMOVAL is AKO this scene and simply specifies space to be container. Then we have, in effect, collapsed the two "rules" of Goal Promotion and

Container-Promotion.

389

4.2.2.2

Symmetric predicates

There

are

numerous

difficult

handling of symmetric predicates.

problems

connected

with

the

The only thing that will concern us

here is the alternation exhibited in the following examples:

(92a)

(92b)

Fred and Bertha agreed Fred agreed with Bertha

(93a) (93b)

Fred and Bertha argued Fred argued with Bertha

(94a)

Fred and Bertha collided Fred collided with Bertha

(9-lb)

In each of these pairs the sentence with conjoined subjects is truly symmetric, but the sentence with with is not.

Dowty 1972 shows this

for the verb collide: (95a) The truck collided with the lamppost (95b)?The truck and the lamppost collided. Examples for the other verbs are easy to find: (96) Fred always agreed with Bertha; Bertha never agreed with Fred (97) Fred always argued with Bertha; Bertha never argued with Fred. Let us propose a scene for these verbs, leaving aside the question just how

it fits

in

with

the

developing: (COLLECTIVE-ACT: ? (COLLECTION)

general

scheme

of

scenes

we

have

bee

390

(INITIATOR ( =>)) [COMITATOR ( => )]) A new procedure has been added to our list of procedural attachments. It is called "=>" and is interpreted to mean "is included among.'' This is simply a generalization of the procedure "=" to which we have appealed all along.

Rather than requiring that a value of one slot A equal a

value of another

B we

may

"included in" the value of B.

now

require that

the

value of

A be

This will only be appropriate when the

value of B is a collection, that is, an object which consists of other objects in the ontology. entity corresponding

to

This assumes an ontology where there is an "John

and

Mary",

to

which

the

individual

"John" and "Mary" bear some specially defined relationship we will call "included-in". treatment,

I am not prepared to offer the details of any such

but

something

like

the

ontology

of

Link

1983

seems

approriate; all we need here is to assume that a satisfactory semantics of plurals will have recourse to some kind of plural entity.

Following

Link, I have called such an entity a collection. The collection slot is thus a slot which adds a role but not a participant.

The COLLECTIVE-ACT scene counts as having only two

participants,

one optional,

and the two

slots

associated

with

those

participants may jointly fill another non-particpant slot. We may also assume that the slot collection requires that it be filled by a collection.

Thus, if the optional comitator slot is unrealized,

391

the instigator slot alone will have to supply a collection. This will admit the first two of the following examples, and rule out the third: (98a)Fred and Bertha agreed. (98b) The committee agreed. (98c) *Fred agreed. Note that a number of similar verbs do appear to be fully symmetric in both valences: (99a) Fred and Bertha played tennis. (99b) Fred played tennis with Bertha (lOOa) Fred and Bertha talked (lOOb) Fred talked with Bertha. (lOla) Fred and Bertha fought ( 101 b) Fred fought with Bertha Although talk sometimes seems to exhibit slight asymmetries: (102a) ?Bertha talks with the walls. (102b) *Bertha and the walls talk. It appears that for fight and play, at least, we shall want to allow for complete symmetry.

Note that this can be done in a ground scene with

slots that take multiple values.

A possible ground scene for play:

(PLAYING-G: COLLECTIVE-ACT (COLLECTION) (INSTIGATORl (=> ) ()) (COMITATORl (=> ) ())) The figure scene for playing will reference this ground along PATHl (see 2.5.1).

392

4.2.2.3

Origin Promotion

At issue here is the valence alternation in the following pairs:

( 103a) ( 103b)

Fred made the log into a canoe Fred made a canoe out of the log.

( 104a) ( 104b)

Fred molded the clay into a statue. Fred molded a statue out of the clay.

( 1OS a) ( 105b)

A lowly fisherman developed into a champion A champion developed out of a lowly fisherman.

Suppose the relevant meanings for into and out of are: (STATE-ASSUMPTION: THING (NEW-STATE) (ORIGIN)) (STATE-DEPARTURE:THING (OLD-STATE) (FATE)) The simplest template scene for the relevant verbs is something like: (STATE-CHANGE: STATE-ASSUMPTION; STATE-DEPARTURE (OLD-STATE ) (NEW-STATE )) Now,

if the Saliency Hierarchy says nothing about the roles old-state

and new-state,

then either· role is an

allowable direct object.

make might have as a lexical entry: , «ACTOR, NEW-STATE>, [•V, -N, TRAN, (OUT-OF)]»

Thus

393

This accounts for the following valence judgements: ( 106) John made the log into a canoe. ( 107) John made a canoe out of the log. ( 108) John made a canoe. ( 109) *John made the log (where "log" is old- state) ( 11 0) *John made (111) *John made into a canoe.

( 112) *John made out of the log.

Note that there are a number of verbs that might be thought of as "exceptions" to this rule.

Shape, turn, transform, and metamorphose

all take only the into valence.

fashion, whittle, sculpt, knit, and cast

all take only the out of valence. semantic split.

Moreover,

there seems to be

a

The first group are all verbs focused on the fact of

state-transformation;

the

particular

means

ar·e

not

The

specified.

second group contains verbs which specify a manner of effecting the transformation, material

of

and constrain the medium;

their

old-state;

whittling

knitting with yarn, casting with metal.

is

some

select the type

ordinarily

done on

or

wood,

Fashion is a bit less specific,

but its most literal use still suggests work done by hand; the manner in which the transformation is affected is constrained. One way to capture this semantic split is to make it part of what distinguishes STATE-ASSUMPTION from STATE-DEPARTURE.

That is,

STATE-ASSUMPTION may be a scene which highlights the fact or result

394

of transformation; manner.

STATE-DEPARTURE a scene which focuses on the

I do not propose any details here, because they do not really

affect questions of valence alternation.

But the natural account that

would follow on such a distinction in scenes is to make the first group of

verbs

AKO

STATE-ASSUMPTION.

STATE-DEPARTURE,

the

second

AKO

Only verbs that exhibited both valences need be

AKO both scenes. Note that on this account the fact

that the o/d_state becomes

direct object with a verb like fashion could be due to two different things: (1) Chance. (2) the fact that there is no available means of marking old_state if it is oblique.

396 4 .j

Conclusion: semantic options versus lexical rule The chief point of this chapter has been to present a plausibility

argument

that

a

number of

regular,

semantically

governed

valence

alternations could be captured in frame representations that gave rise to various kinds of realization options: options left open by the Saliency Hierarchy, optional participants, and optional roles.

This possibility in

turn raises a theoretical problem. The theory seems to have two mechanisms for describing the same phenomena, single frames with realizational options, and morphological redundancy say, they

rules.

a computational are

anathema

Where such free choices are quite congenial to, linguist working on a practical to

the

theorist,

because

implementation,

they

indicate

an

overabundance of theoretical apparatus. \ve can make the

dilemma

clearer

by

outlining

three

possible

positions differing in how much of the burden of accounting for valence alternations is laid on the morphology.

All three positions will assume

that the theory needs morphological redundancy rules, both for rules that

change grammatical

classical derivation,

function,

like

like un- prefixation.

passive,

and

for

cases

of

They will also assume that

the theory needs some account of optional participants.

Here, then are

three possible theoretical courses sketched within those assumptions, with the version that lays the most burden on the morphology coming

397

first: (1) All possible semantically systematic valence alternations must be treated by morphological redundancy rule. (Only participants can be optional).

(2) Only constituents that do not contain material mentioned in the Saliency Hierarchy can be optional. The intuitive thrust of this claim is that material mentioned in the Hierarchy is somehow essential to the ontology of scenes; and that any situations that differ in some salient role must be different types of scenes. Enforcing this constraint means that rules like Goal and Container-Promotion must be morphological Promotion redundancy rules involving zer·o-derivation. (3) (3) There are no category-preserving redundancy rules involving zero-derivation.

morphological

Position (3) resolves the dilemma because then any systematic valence alternation must be treated as realizational options allowed by the lexical semantics. rule.

Position ( 1) resolves the dilemma by stipulation; do it by

It has the advantage of being a strong position, and, in at least

one version, an easily falsified one. appear to resolve the dilemma.

as it stands does not

It constrains some of the analyses in

this chapter to be morphological unaffected.

Position (2)

redundancy

rules,

but others

are

In particular, the analyses of Instrument Promotion, Dative

Movement, and

Origin-promotion,

stand as proposed.

and Symmetric Predicates can all

The reason is that in each of these cases,

either

a non salient participant was optional or there was no optionality at all; the valence alternation was explained solely in terms of a realization option allowed by the Saliency Hierarchy. Position (1) says that all possible semantically systematic valence

398

alternations would have to be treated by morphological redundancy rule. One way to acheive this result is simply to forbid single lexical entries to have more than one denotation.

The problem that immediately arises

is what to do in the cases where the Hierarchy has nothing to say, with, say, verbs like blame. verbs

yield

redundantly

lexical

One can, of course, maintain that such

entries,

storing

the

but

same

then

one

will

lexical-semantics

representation in two differ·ent places.

If,

be

charged

and

with

phonological

in answering this charge

one resorts to some kind of shared structure among different the lexical entries for blame, then it is hard to see how this really differs from a single lexical entry. What a proponent of (1) really needs, in order to have an inter·esting position, is some criterion for individuating lexical entries,

and some

reason for claiming that "semantically

systematic'"

alternations need to be governed by morphological rule. The interesting version of (1),

it seems to me, is the one that

claims that Saliency Hierarchy is a complete algorithm which imposes a strict ordering of roles on every head, and that the nuclear terms must be chosen according to this ordering.

Then all valence alternations

involving nuclear terms would have to be done by lexical rules, whether semantically governed or not.

That is, in any case where a single head

form had two valences, either it would have two different meanings in those valences, or one of the valences would violate the hierarchy and would have to be derived by relation-changing rule.

This version of

399

(1)

explains why every systematic alternation must be a morphological

rule; the

first version only stipulated it.

In this guise, (1) is a stronger hypothesis than (2), but I think it is arguably wrong. Consider one case where (1) differs from (2).

Under position (1)

1.1strument Promotion would have to be a lexical rule; under (2) have the treatment we gave. not change meaning;

Assume that Instrument Promotion does

in effect this means that the rule is merely a

relation-changing rule (like Passive). semantically

relevant

it may

verbs

(1) predicts that none of the

will exhibit only the "derived" valence.

because (1) stipulates that there is a unique basic valence

This is

licensed by the Hierarchy for each verb"s semantics.

In the case of

the verbs of impingement and Instrument Promotion this seems to be right;

there is

no impingement verb which takes the instrument as

direct object which does not also take the patient as direct-object. the

(1)

position

holds

But

that there will always be such a "winning

valence, and this does not seem to be the case. We noted, for example, that in the case of Origin-Promotion, both valences exist independently of the alternation: (113a) K. metamorphosed into some strange sort of vermin. ( 113b)*A strange vermin metamorphosed out of K. (114a) John whittled a statuette out of wood (114b)*John whittled wood into a statuette.

400

Of course Origin-promotion is not a problem for (1) if it is taken to be an alternation that involves a change in semantics, because then both

valences could be licensed by the hierarchy.

But among the

verbs that do take both valences, there seems to be no independent reason to posit a systematic semantic change. More generally, the problem with (1) is that there just seem to be too many cases like blame (a number of example were given in () - ()), where very local valence alternations are possible.

(1)

might be tuned

to allow for scattered, idiosyncratic exceptions to the complete ordering of the Saliency Hierarchy,

but each such exception that cannot be

linked

rule

to

a

morphological

challenges

the

assumption

that the

Hierarchy completely decides the semantics of nuclear terms. In sum, position.

(2),

although

weaker,

seems

a

much

more

plausible

Within the scope of (2), it is always possible to capture any

particular prediction made in (1) by strengthening the hierarchy and calling a particular alternation a lexical rule.

For example, to capture

the fact that verbs of impingement will never take only the valence on which instrument is direct object, we need only change the hierarchy so that instruments can not ordinarily become direct objects (let patient be ranked,

for example),

relation-changing rule.

and

then

license

Instrument

Promotion

as

a

A parallel account will capture the fact that

within the semantic domain of the Dative Movement verbs,

there are

401

none that take only the double object construction. Let

us

now

consider

category-preserving zero-derivation.

rules

(3)

(3).

1

out

cases

all

of

It thus amounts to the claim that

all cases of so-called zero-derivation within categories can be handled with

semantic

representations

that

countenance

different

realization

options. It will be worthwhile examining a that

presents

alternation

certain

between

problems for transitive

a celebrated case of this sort This

(3).

and

intransitive

sometimes been called the English Causative rule.

is

productive

the

valences

that

has

This involves verbs

like open, turn, break, fill, crack, melt, cool, stop, start, and move. One very simple proposal for handling this is really an echo of Fillmore 1968: (DIRECTED-ACTION-CHANGE: Dl RECTED-ACTION; CHANGE [ACTOR] ( PATIENT)) If

all

of

the

above

verbs

incorporated

Dl RECTED-ACTION-CHANGE scene into their

lexical

the structure,

above we

could have single lexical entries yielding both valences (note that this

1. The idea of semantically constrained relation-changing rules is This is how Jackendoff 1972 proposes to treat certainly not new. And Wasow 1980 proposes a whole class of such rules Passive. (including one of his two passive rules), which he calls minor rules (borrowing a term from G. Lakoff).

402

treatment violates position (2)). But there are two classes of verbs for which this treatment is not really satisfactory.

Halliday 1967 distinguishes among the cases

of

transitive/intransitive alternation three distinct types: Nuclear: (115a) John opened the door. (115b) The door opened. Descriptive: (116a) He marched the pr·isoners across the yard. (116b) The prisoners marched across the yard Effective: (117a) The clothes washed (117b) He washed the clothes idea

The

transitive some

of

Halliday"s

(effective), are

division

some

truly

Dl RECTED-ACTION-CHANGE

is

basically neutral frame

that

some

verbs

are

basically

intransitive

(descriptive),

(nuclear).

The

is

representation only for the last class, and,

intuitively

a

and above

satisfying

indeed, all our paradigm

examples of the Causative rule seem to be nuclear. It is always difficult to argue what the "basic" nature of a verb is, but the following fact suggests that the transitive valence of march has a special status: (118) ?The prisoners were marched. Some sort of complement makes the passive much more acceptable:

403

(119) The prisoners were mar·ched across the compound. The necessity of a complement with the passive of the Descriptive cases parallels a fact about the Effective cases, a number of which

sound

much more acceptable with a following adverb. ( 120) Cockroaches don •t kill easily. The

markedness

of the

Effective

cases

is much clearer.

It seems

uncontroversial that we want our basic entry for a verb like kill to be transitive. A nice argument that orthogonal

processes govern the Nuclear

and Effective cases is Halliday·s observation that some verbs seem to swing both ways. (121a) This silver shines. (121b) This silver shines easily. (b) has a reading paraphrasable as ··It rs easy to polish this silver." One way to account for such data would be

to

handle shine

with

something like the above Dl RECTED-ACTION-CHANGE frame and then, contrary to (3),

derive an

intransitive

version

from

the transitive

404

version by lexical rule.

2

Supposing for the sake of argument that a distinction did have to be made.

How would we make it under proposal

descriptive class.

(3)?

Consider the

Short of generating a new frame, there is no way in

the current frame notation to represent valence addition.

Note that

even a rule relating frames is unsatisfactory, since no relation stated on a purely semantic level guarantees that the related objects will be pronounced the same.

To capture the notion of valence addition, we

need to have our hands on some morphology. The

Effective

class

poses

much

the

problem;

same

detransitivization can not be represented in a single frame. If we rule out the possibility of a zero-derivation rule, then the

only recourse left open is a syntactic rule, and this

is in fact the

treatment counseled in Keyser and Roeper 1982 for the Effective cases. In their framework

it

would

be

difficult to

conceive of

a

parallel

treatment for the descriptive cases, but there is no principled obstacle to a valence increasing syntactic rule (in GPSG, such a rule could be

2. For those who control a distinction between intransitive lie and transitive lay in their dialect, the contrast between descriptive and effective readings will have morphological reflexes: (i) This rug lies well. (ii) This rug lays well/easily. I am indebted to Charles Fillmore for pointing out this example.

405

using

formulated

easily

semantics). difficult

to

metarules,

if

one

could

find

a

plausible

The disadvantage of such a recourse in GPSG is that it is what

determine

principles in that framework distinguish

processes treated by lexical rules from processes treated by metarule. We have only shifted the domain of the problem; we have not solved it. have alternations,

general

no

treatment

to

propcse

for

these

if indeed they represent three separate processes.

three t-.1y

purpose here is to sketch some of the problems which would arise in a fully developed version of (3), problems which could be avoided under (2).

Hypothesis

(3) takes on a very different complexion when we

consider it in the light of some cross linguistic facts. the Causative rule.

Consider again

It is well-known that a number of languages handle

the same semantic alternation with a morphological affix.

Turkish and

Japanese, for example, both have causative suffixes which attach to a verb stem meaning "open" to yield a new verb with a new participant that mean roughly "cause to open."

Other languages may relate the

same semantic material, but with the morphological markedness going in the other direction. "open ,

Thus, Polish has a basic transitive verb meaning

and creates an

clitic; this verb form,

intransitive version by attaching a reflexive called the middle, may also be viewed as the

output of a morphological process.

406

is that it keeps us from

A consequence of position (3), then,

treating causativization in English with the same grammatical device as either causativization

in Turkish and Japanese or middle-formation in

Clearly, where there is an affix there must be a morphological

Polish.

rule; position (3) entails that English, with no affix, has no such rule. What are the compensating

What are the benefits of such a position? benefits of some version of positiun (2)?

Essentially, there are two interesting questions here: first, what Second, what can the

can the semantics of morphological rule involve? semantics of a single lexical entry contain?

Position (3) makes some interesting predictions about what can fit given some fixed formalism for

into the semantics of a single head lexical

representations.

introduced thus far separate

optional

For example, the notational devices we have

do

not

allow

constituents.

there to

Imagine

a

be

dependencies

lexical

representation

which two participants were in complementary distribution; was

present,

among in

when one

the other had to be absent, and vice versa (we shall

encounter an occasion for using such complementary constituents when we

deal

with

predicative

PP's

in

chapter 5).

Given

the current

formalism, position (3) entails that definable classes of heads with such complementary valences do not exist. formalism,

Note that even with the same

such heads might exist under position

(2).

Position

(2)

would allow two such complementary valences to be related by a rule of

407

there

wi II

processes at all;

this

that

states

(3)

Position

zero-derivation.

non-category-changing zero-derivation

be

no

includes

processes that would introduce one participant and remove another. It is thus possible to conceive of predictions that would follow from position (3), although they appear to be of a negative character. The drawback of position (3) is that it can only say something about zero-derivation

processes;

it

say

can

about

nothing

morphological

processes in general. I want to argue that position (2) can make make semantic claims about what sorts of valence alternations are mediated by morphological moreover

processes; alternations

that

in doing

so it also defines a class of valence mediated

need

to

be

simple example

to

begin

not

do

by

morphological

processes. To

give

a

with,

(2)

rules

out the

Dl RECTED-ACTION-CHANGE scene we suggested above as an analysis of

the English Causative alternation.

is because the optional

This

constituent in that scene is actor, a role mentioned in the hierarchy. (2) says that DIRECTED-ACTION-CHANGE must really be two scenes. This

means

that

if

a

language

is

going

to

express

a

systematic

relationship between those two scenes, it must do so by a morphological rule.

Thus, under (2),

zero-derivation

is

just an

the fact that that rule in English involves accident.

It

is

thus

encouraging to a

408

proponent

of

(2)

to find that numerous

languages

use an

affix to

express the causative relation. Of the analyses presented in this chapter, (2) rules out only Goal Promotion and Container Promotion; were

captured

with

a

both of these valence alternations

single frame

that

involved

totally-affected, a role mentioned in tt·.e Hierarchy. that

situations

different

scenes.

that

differ In

a

in way,

a

salient

CD

slot

optional

The gist of

must

constitutes

an

a

be

(2) is

instances

claim

slot

about

of the

individuation of scenes; as a consequence, it is also a claim about the individuation of lexical items.

The two different valences of load must,

under (2) belong to different verbs. To relate such valences systematically, we are then compelled to resort to a morphological redundancy rule; I will propose such a rule in Chapter 5. The payoff of such a treatment is largely in its cross-linguistic prespective.

English is not a morphology-rich language; therefore the

existence of zero-derivation processes in English is not

surprising.

In

languages with more morphological apparatus, we would expect to find the similar relations between lexical items mediated by some kind of affix.

In the case of Goal Promotion, a modest cross-linguistic survey

involving Polish, Russian, and Hungarian, is promising. Hungarian examples are from Davison 1980:

The following

409

az agy-ra a takaro-t (122a) Ra-takartam onto-spreadl sPa the blanket-ace the bed-onto I spread the blanket on the bed. az agy-at

(122b) Be-takartam a takaro-val -instr.

-ace. I covered the bed with the blanket.

(123a)

viz-et a Be-toltom into-pour-1 sPa the water-ace

a

pohar-ba/medence-be.

the glass-into/pool-into. I poured water in the glass/the pool.

( 123b) -ace. instr.

Fel-toltom a medence-t/ a poha r-at upto -ace.

viz- zel.

I filled up the pool/the glass with water

In both these Hungarian pairs, the same stem with different prefixes takes valences that parallel the English Goal Promotion alternation.

The

semantic shift also parallels the English semantic shift; that is, in the goal valence, with an instrumental, the Goal is totally affected.

A parallel example in Polish: (124a) W-pycha.t'em into- -push-impfl sgPa.

do poduszk-i.

pierz-~

feathers-ace

to pillow-gen. I stuffed the feathers into the pillow. (125)

Wy-pcha~m

compl-pushpflsg. Pa.

poduszke

t

pillow-ace

pierzem.

feathers-instr. I stuffed the pi II ow with feathers.

410

Here not only is there a prefix change, but we also have a change in the stem corresponding to a change from imperfective to aspect.

perfective

In Polish, and apparently in Russian and Hungarian as well,

there is a considerable interaction between the holistic semantics and the often

grammatical marked

institution of aspect,

with

a

prefix

where aspectual contrasts are

(almost all

imperfectivizing suffix are perfective).

prefixed

verbs

without

Thus, in Polish we find

an

~o:e1sE::S

of a single perfective form that exhibits both valences, with a pai r·ed imperfective that allows only one of the valences: ( 126) ,.l::.'adowaYem Load-impf1sg. Pa

siano hay-ace

I loaded hay on the truck

( 128)

on truck-ace

' ciezarowk-e sian-em

( 127) *kadowa-li:!m f1sg. Pa I loaded the

,

na cieiarO'wk-e

truck-ace

NaYcldowat'"em Load-pf1 sg. Pa

na ciezarowk-e

siano

l

hay-ace

I loaded hay on the truck ( 129)

Clearly,

Promotion

on truck-ace

"l

Load-pf1sg. Pa

Goal

?

ciezarO'wk-e sian-em

Na~dowaYem

hay-inst I loaded the

hay-inst

truck with hay.

truck-ace truck with hay.

is

not

a unified morphological operation

Polish, but nothing about (2) leads us to expect it will be.

in

All we

411

totally-affected morphological

means

role

differing

in

that

clearly,

to

capture in

instrumental representation

are

these of

available,

will the

relationship we

will

case

to

examples

Instrumental

need be

the one

If

role the

Just

as

appearances

of

morphologically.

between

the

between scenes

the relationship

expressed

be

items.

lexical

different

involve

will

involving

alternations

semantic

the

that

is

expect

of

something

the

semantic

very

Lkt our

FILLING scene. An example parallel to the last one from Russian: probkoj butelku. *butel ku probkoj. She plugged the cork in the bottle (131) zatknula probku v butelku. butelku probkoj

(130) zatykala

That this semantic domain gives parallel languages

is

clear,

just

as

it

is

valence possibilities for four

clear that

there

are

significant

differences among the languages in expressing the relevant semantics. Perhaps the most striking parallel is one we have not captured in our lexical representations: all four languages have a way of marking the

material in the FILLING scene which is morphologically identical to their

412

way of marking the instrument in the MANIPULATION scene.l

Our

current representations of these scenes show no abstract connection, and this is surely a shortcoming. As far as (2) is concerned, though, matters are just as expected. Scenes differing in the role totally-affected have mapped to different lexical items. There are two sides to

(2),

however.

The

semantic

domain

directly affected is quite small, and thus there is quite a bit of leeway as to what can be packaged into a single entry.

Most of the rules we

have considered in this chapter involve fairly idiosyncratic features of English, but there is at least one rule, which is fairly well-represented cross-linguistically, whose treatment in this chapter is countenanced by (2). expect

This is Dative Movement. that

the

lexical entries.

semantics

As stated,

(2) does not lead us to

of Dative Movement will

involve separate

That is, we need not expect Dative Movement,

if

it

appears, to involve an affixation process.

It should be clear from the 3. I am speaking a little loosely here. approach we have taken that there is no guarantee that a lexical scene in one language will be a lexical scene in another. The exceptions must of course be the scenes alluded to in the Saliency Hierarchy, which are Beyond that, one does expect some semantic proposed as universal. trends in lexicalization across language, concepts that are frequently, if not always, marked, and there is always the chance that there is some particular inventory of scenes, outside the Hierarchy, which is universal.

To sum up,

in this chapter I have shown how the theoretical

apparatus we have developed for lexical semantic description is powerful enough

to

represent

valence-alternations as lexical

entry.

considerations

Even might

a

number

of

semantically

simple realizational options taken given lead

us

this treat

power, sorr•e

however, of

these

morphological processes involving zero-derivation.

As

governed by a single

cross-linguistic alternations We

as

shdl: see 1n

Chapter 5, abandoning a realizational options analysis of Goal Promotion allows us to strengthen our definition of semantic compatibility back to its force in Chapter 2, and adopt what is perhaps a more persuasive predicative PP analysis of the patient

valence.

5.

Predication, Control, and Lexical Rules Revisited In this chapter I investigate the interaction of a general account

of predication or control with the phenomena

in chapter 4,

discu:>~t!cl

and with our general framework of lexical representation.

The material

here is not intended as a complete theory of control; it merely sketches some of the lines such a theor·y might follow if it were to be consistent with the view of lexical representations adopted here.

First a simple

account

the

of

infinitives

is

presented in chapter 1.

sketched,

compatible

with

semantics

Then thrs account is extended to predicative

prepositional phrases, and the question of the lexical representation of heads

with

predicative

complements

is

addressed.

Finally

Goal-promotion is reconsidered in light of a predicative analysis; there are problems

which

again

point

rn

the

direction

of

a

lexical

rule

account. I use the term predicative here to refer to any complement which is controlled by another one of the head's complements.

The classic

examples are thus equi and raising VP complements, both infinitival and participial.

But there are numerous other examples recognized in the

literature involving all the other major categories, and adjuncts as well (see Bresnan 1982).

In GPSG (Gazdar,

Pullum, and Sag 1980),

the

term "predicative" is used to denote a syntactic class, basically those

414

complements that occur after the copula be.

In standard American

English this does not include infinitives (and there is some question whether one can give a compositional treatment of the British modal future construction with "is to'").

I do not intend the term to be taken

as this syntactic class, whatever it may be precisely, but merely as the semantic class of controlled complements.

G?SG has no convenient term

to refer to the class of subcategorized-for controlled compleo1enls; LFG, however, does.

In Bresnan 1980, the only syntactic

subcategorization

information that a governing head involving control needs to specify is the grammatical function XCOMP.

I will adopt this term.

1

1. The idea that XCOMPs are what control verbs subcategorize for is implicit in Bresnan 1982 and has a great deal of appeal. Taken together with LFG"s principle of functional uniqueness, it correctly predicts that no verb will have more than one controlled complement, even controlled complements of different categories (this prediction was not made in an older version of LFG which allowed functions like NCOMP, PCOMP, and so on). It also countenances verbs that allow controlled complements of more than one category: (i) (ii) (iii)

I saw John in trouble I saw John leave I saw John angry.

The catch is that this seems to leave the finer points of selection up to the semantics. This is of course a view I am sympathetic to, but some room for extra syntactic stipulation seems to be necessary in the face of: ( iv) (v)

I made John leave. I forced John to leave.

The existence of minimal pairs like (iv) and (v) suggests that semantics cannot always be relied on to determine the precise syntactic features of a controlled complement, and that subcategorization for XCOMP alone does not give a fine enough set of distinctions.

41S

5.1

Infinitival Complements I propose we treat the relation between matrix and controlled

complement situations as unmediated by a complement marker, that is, as a primitive grammatical We will

object.

use the

relation on a par with subject and direct LFG term and call this

relation XCO:-..tP.

Under this assumption the infinitive rule and the object control rule become: V[TO] Vl: Vl> V N2 Vl: (lambda sigma (lambda x (exists tau ( N2 (lambda y [(and (V sigma x y tau) (Vl tau y))])])))>

This

object control rule will be adopted for the remainder of this

chapter.

A simple generalization will be proposed in 5.3 to cover all

predicatives VP's.

under the assumption they all

A few

have the same types as

remarks are in order to explain

phenomena are treated. First, subject control is minimal variation: V Vl: (lambda sigma (lambda x (exists tau [(and (V sigma x tau) (Vl tau x))])))>

how some standard

416

Second,

the distinction between Equi and Raising should, in our

current framework, be a distinction in whether or not the controller has To make the principle clear, let us

a role with respect to the matrix.

propose provisional figure frames for believe and persuade: (BELIEVING (BELIEVER) (PROPOSITION)) (PERSUADING (PERSUADER) (PERSUADEE) (PROPOSITION)) Let us assume the above control rule works for both equi and raising verbs. Then the predicates BELIEVING and PERSUADING must have the same

PERSUADEE,

PERSUADER,

Each

PROPOSITION>.

entity-type arguments will have a slot to fit into. BELIEVE?



Note that there is a morphological operation (though a vacuous one) relating not frames but frames surrounded by angle brackets. to

capture

the

fact

that

these

are

imp Iicational statements about frames. about frames, nothing

lexical

statements,

not

This is mer·ely

If this was only a statement

would guarantee that the pairs of related frames

realized lexical items pronounced the same way.

Presumably the same

format could be extended in cases where the morphological operation was more substantive, although matters would be complicated somewhat if the same kind of rule is to be extended to derivational morphology in general, since we would then wish to say something about the frame representation of the affix. Note that the above lexical frame has no syntactic component, no new specification of syntactic features for the lexical

item.

This

is

because the syntactic features of both valences of the Goal-promotion lexical

rule are predictable from the lexical

strong constraint on

semantics.

semantically-governed lexical

A possible

rules is that they

435

never include syntactic statements. We can sum things up by comparing the conclusions in the last section of chapter 4 with the conclusions conclusion

of

constituents

chapter

not

to

4

we

include

cross-linguistic resu!ts.

of this

suggested

salient

that

roles

section. requ1r1ng

might

have

In

the

optional

interesting

In particular it characterized a set of valence

alternations

that,

if

redundancy

rules,

which

regular,

had

to

be

treated

as

morphological

in turned characterized a kind of semantic

relatedness that might be mediated by affixation.

Dative Movement and

Instrument Promotion were unaffected by those considerations, as were our treatments of most of the other valence alternations discussed in Chapter

4

(I

will

turn

to

some

exceptions

in

considerations brought to bear in this chapter, analysis of the Goal-promotion verbs, Goal-promotion should be a lexical

a

moment).

The

chiefly a predicative

point to the same conclusion.

rule.

Again,

most of the other

alternations treated in Chapter 4 are unaffected by this consideration, since they did not involve valences with a large choice of prepositions. The obvious exception is Container{Promotion, which also involves the role totally-affected and which we subsumed to Goal-Promotion at one

point.

Nothing

Goal-Promotion analysis

would,

and

said

here

excludes

Container-promotion

however,

have

to

in

treat

a

uniform English;

the

analysis a

Source

predicative, even though only two prepositions are possible.

of

predicative valence

as

-B6

verbs of

There is another exception and this is the treatment of impingement briefly sketched in 4.3. 1. 1.

There we posited an optional

result component Iin ked to the role totally-affected,

which correctly

predicted which participant must be direct object in "John hit the ball over

the

fence,"

correctly

and

instrument

blocked

promotion.

predicative analysis of this valence is attractive because it allows number

prepositions;

of

that

on

analysis

result

would

be

not

A a a

component, but a participant and the same sorts of problems that we noted for a predicative analysis of Goal-Promotion would arise; again, these problems could be solved by resorting to a zer-o-derivation rule that created the resu ltative valence for verbs of impingement.

Again,

the requirement that optional constituents not include salient roles also points in that direction. To sum up, a number of independent considerations point to a treatment of Goal Promotion as a genuine lexical rule.

This does not

mean, however, that all of the valence alternations discussed in Chapter 4 need to be lexical rules.

As we noted at the end of Chapter 4,

English allows a number of idiosyncratic valence alternations that have no

as

status

rules,

and

can

be quite

easily described as options

available within a single lexical entry. In the final analysis, some valence alternations seem to best be described

as different syntactic options for

participant.

realizing the same core

Perhaps the simplest case for such options is a kind of

437

alternation

valence

among

common

we

have

not yet discussed,

verbs:

English

the

but which

possibility

of

is quite

realizing

a

propositional ar-gument as an ordinary NP, as a controlled complement, or as an embedded

finite clause:

(25) John expected an earthquake. (26) John expected there to be an earthquake. (27) John expected that there would be an earthquake. Each

of

these

examples

exhibits

what

is

arguably

a

subcategorization, yet all arguably involve the same main verb.

different I leave

as a question for future research an account of these three valences which

could

relate them to a single lexical entry with one semantic

representation.

The

relationship

of

alternative

valences

for

propositional complements is, it seems to me, an important problem for any general semantic theory of valence.

REFERENCES Alexander, D. •nd W.J. Kunz. 1964. Some Classes of Verbs In English.. Linguistics Reseuch Project, lndi•n• University, Principal lnvestig•tor, F.W. Householder. lndi•n University Linguistics Club, Bloomington. Alexander, D. •nd W.J. Kunz. 1965. More Classes of Verbs In English.. Linguistics Research Project, Indiana University, Principal Investigator, F. W. Householder. Indian University Linguistics Club, Bluomington. Amritavalli, R. 1979. The Representation o/ Transitivity In the Lexicon. Linguistic Analysis 5, ·71-92.

Expressing Cross-Categorla/ Selec:tlonal Amritav.alli, R. 1980. Correspondences, An Alternative to the X-bar SyntaJt Approach. Linguistic Analaysis 6.3: 305-343. Anderson, John M. 1977. On Case Grammar: Prolegomena to a Theory of Grammatical Relations. Croom Helm, London. Anderson, Steven R. 1971. On the Role of' Deep Structure In Semantic Interpretation., Foundations of Language 7: 387-96. Aronoff, Mark. 1976. Word Formation and Gener•tlve Gr•~r .. MIT Press, Linguistic Inquiry Monographs, No. 1. Cambrigde. Barwise, Jon and John Perry. 1981. Situations and Attitudes. Tile Journal of Philosophy. LXXVII4(11): 668-691. Barwise, Jon and John Perry. 1982. manuscript, Stanford University. Bennett, Michael. 1974. Fragment of English..

Situations and Attitudes.

Some Extensions of a Montague Doctoral Dissertation. UCLA.

Bresnan. J. W. 1976. Towards a Re.allstlc Model of Transformational Grammar. paper presented at at the MIT-AT&T Convocation of Communications. March, 1976, MIT.

Bresnan, J. W.

Polyodlclty: Port I of a Th~ry of LeKicol Rtdes and Re,-.sentotlo ns. In T. Hoekstra, H. van der Hulst,

and M. Moortgat (Eds.), Lexical Publications, Dordrecht, Holland.

Bresnan, Joan. 1982. Inquiry 13: 3.

Grammar.

Foris

Control and Complementa tion. Linguistic

Chafe, Wallace L. 1970. Meaning and the Language. University of Chicago, Chicago.

Structure of

Chierchia,

Genna~. 1983. Outline of a Semontlc ThN•'Y of Obligatory Control. Proceedings of the Second West Coast

Conference on Formal Linguistics. Association, Stanford.

Stanford Linguistics

Chomsky, Noam. 1970. Remorks on Nomlnallzot/o n. In Readings in English Transformati onal Grammar., Jacobs, R. and Rosenbaum, P. (eds.) Ginn, Waltham, Mass. Clark, Eve, and H.H. Clark. 1979. Verbs. Language 55: 767-811.

When Nouns Surface as

Davidson, Donald. 1967. The Logical Form of Action Sentences. In N . Rescher ( ed. ) . The Logic of Action and Decision. 81-95. Pittsburgh. Davison, Debora. 1980. Verbal Prefixing. In Hungarian. University of California at Berkeley M.A. Thesis. Dilvio. 1972 Studies in the Logic of Verb Aspect and Time Reference in English. University of Texas Dissertation.

Oc;wty,

Dowty, David. 1978. Montague Grammar and Lexical Rules. In Papers from the Parassession on the Lexicon, Chicago Linguistics Society. Dowty, David. 1982. Grammatical Relations and Montague Grommor.ln Jacobson, P. and Geoffrey K. Pullum. The Nature of Syntactic Representatio n. Reidel, Dordrecht, Holland.

Fillmore, Chutes. 1968. Thft Cos• for CtJse. In Bach, E. and R. Harms (eds). Universals of Linguistic Theory. Holt, Rinehart and 'Winston, New York. Fillmore, Charles. 1971. Some Problems for Case Gl"tJJMttJr. In Richard O'Brien, S.J. (ed) Monograph Series on Languages and Linguistics. 22nd Annual Roundtable. Georgetown University. Fillmore, Charles. 1977a. The Case for Case Reopened. In Cole, Peter, eeL Syntax and Semantics 8. Academic Press, New York. Fillmore, Charles. 1977b. Topics In Lex/col s.trtontlcs. In Roger Cole, ed. Current Issues In Linguistic Theory. Indiana University• Press, Bloomington. Friedin, R. 1974. Transformations and lnterpr«lve Semantics. In R. Shuy and N. Baily (eds.). Towards Tomorrow's Linguistics., Georgetown University Press, Washington,

D.C.

Gawron, Jean Mark, Jonathan J. King, John Lamping, Egon E. Loeb~r, E. Anne Paulson, Geoffrey K. Pullum, Ivan A. Sag, tll'ld Thomas A. Wasow. 1982. Processing English with a Generalized. Phrase Structure Grammar. Proceedings of the 20th Annual Meeting of the Association for Computational Linguistics, Toronto. Gazdar, Gerald. 1981a. Unbounded dependtmcles coorpinate structure. Linguistic Inquiry 12: 155-184.

and

Gazdar, Gerald. 1981b On syntactic categories. Philosophical Transactions of the Royal Society of London B 295: 267-276 .. Gazdar, Gerald. 1982. Phrase structu~ grammar. In Jacobson, Pauline and Geoffrey K. Pullum (eds.). The Nature of Syntactic Represenbtion. D. Reidel, Oordrecht. Gazdar, Gerald, and Geoffrey K. Pullum. 1982. Generalized phrase structure grammar: a theo~t/cal synopsis. Indiana University Linguistics Club, Bloomington.

Gazdar, Gerald, Geoffrey K. Pullum, and Ivan A. Sag. 1982 .. Au~tlllo,.les ond Nllot«< phenomeno In t1 ,-est,.lctlve theo,.y of gl"tlmtrttJ,.. language 58: 591-638. Subccrtego,.lzcrtlon. ond CiNIIMifltlctll 1982. Paper deliverfd at the First Annual West Coast Conference of Formal Lingllistics, Stanford University.

Grimshaw,

Jane.

Relcrtlons..

Goldstein, Ira and R. B. R.b~rts. Memo 408. MIT. July, 1971.

1977. The FRL Prime,.. AI

o 1979. Nudge, Dieter Metzing, In Pt"Ogl"tlm. Kl'tOWiedge-btlsed Sch«

etc.

8 4

2

Figure 7.4 (The numerical expressions represented in the model are: '2418', '24.18', '18.24', and '81.42'.)

7.3.4

Economy and Semiotic Subsystems

We noted right from the start that a semiotic system is always 'compounded' of a cenologicallevel and a plerologicallevel, and have also mentioned, in connection with the possibility of a system being 'doubly complex', that both these levels may display their own measure of economy (that is, both levels may be capable of a systematic production of complex entities). When this is the case, each of the levels is said to constitute a subsystem of the semiotic system in question. The cenological subsystem and the plerological subsystem out of which a given semiotic system is 'compounded' are autonomous, mutually complementary 'compartments' of one and the same semiotic system. Subsystems of a semiotic system can be classified in two dimensions: (1) cenological subsystems versus plerological subsystems; (2) componential subsystems versus articulations. The term 'componential subsystem' refers to a subsystem producing complex entities of the simultaneous bundle type; 'an articulation' refers to a subsystem capable of producing complexes that are articulated constructions. A matrix can be formed from the two dimensions listed above, leading to the establishment of four types of subsystem (see Figure 7.5).

AXIOMATIC SEMIOLOGY

Componential subsystem

Articulation

Cenological subsystem

Cenematics

Cenotactics

Plerological subsystem

Plerematics

Plerotactics

197

Figure 7.5

7.3.5

'Interlocking' of Semiotic Subsystems

At this point, we are in possession of all the ingredients required for classifying semiotic systems according to the way they are 'compounded' of a cenologicallevel (that may be a simple inventory with no potential for producing complexes, or may, itself, be 'compounded' of cenological subsystems) and a plerologicallevel (that may be a simple inventory with no potential for producing complexes, or may, in turn, be 'compounded' of plerological subsystems). The compounding of a semiotic system out of 'interlocking' subsystems is particularly clearly illustrated by human languages: Cenematics: a small closed inventory of basic figurae- the so-called distinctive features (e.g. /labial/, /occlusive/, /voiced/, etc.); a cenematic subsystem operates on these distinctive features, combining them into simultaneous bundles, the so-called phonemes (e.g. /labial/ + /occlusive/ + /voiced/ = /b/, /labial/ + /nasal/ = /m/, /dorsal/+ /occlusive/+ /voiced/= /g/, etc.). (2) Cenotactics: the phoneme-output of the cenematics provides an inventory of basic figurae (this constitutes the interlock between cenematics and cenotactics: an output-input relation); a cenotactic subsystem operates on the phonemes, combining them into articulated constructions, the so-called phoneme-chains (e.g. /t-H~a~m~z/, /~t~r~a~p/, /l~e~Q~e ~n4:-z/, etc.).

(1)

The combined output of the two cenological subsystems provides the means whereby the forms of the signa of the human language in question (i.e. English) can be differentiated (this constitutes the interlock between the cenologicallevel and the plerologicallevel). Plerematics: a closed and relatively limited inventory of basic signathe so-called monemes or morphemes (e.g. 'man', 'boy', 'walk', 'quick', 'plural', 'past', '-ly', etc.); a plerematic subsystem operates on these monemes, combining them into simultaneous bundles, the so-called words (e.g. 'boy' + 'plural' = 'boys', 'walk' + 'past' = 'walked', 'quick'+ '-ly' ='quickly', etc.). (4) Plerotactics: the word-output of the plerematics provides an inventory of basic signa (this constitutes the interlock between plerematics and plerotactics: an output-input relation); a plerotactic subsystem operates on the words, combining them into articu(3)

198

SEMIOTIC PERSPECTIVES

lated constructions, the so-called syntactic combinations (e.g. (the~man)~(has~taught )~(many~boys), (that~girl ~(in~ (the~garden))), (((his~father)~s)~brother), etc.). The compounding of human languages out of subsystems that interlock in a fixed order can be sketched in the scheme shown in Figure 7.6. (1) Cenematics (2) Cenotactics

(3) Plerematics (4) Plerotactics

Figure 7.6

In comparison, to take yet another example of a different type of semiotic system, a written language such as English is compounded out of the interlocking of three subsystems (see Figure 7. 7). (1) Cenotactics (2) Plerematics (4) Plerotactics Figure 7.7

The lowest level of economy in this system is that subsystem of which the basic inventory is the graphemes (alphabetic letters). These, however, do not enter into a cenematics forming them into simultaneous bundles, but directly into a cenotactics producing articulated constructions (that is, chains of alphabetic letters). As for the signa-level of an alphabetically written language, this operates, analogously with a spoken language, with monemes combined into simultaneous bundles: the words; and with words combined into articulated constructions of a syntactic nature. (There is, in other words, a high degree of isomorphism between, say, the plerological subsystems of written English and of spoken English. Such an isomorphism presupposes, however, that we have two a priori separate systems to compare, and that we have well-established ways for comparing them. Looking upon the written system and the spoken system as separate semiotic systems, each compounded out of a certain number of subsystems, brings out both the differences and the correspondences between the two systems.) A systemology (theory of semiotic systems) distinguishing between (a) simple semiotic systems; (b) complex semiotic systems with only one economising subsystem (but in this category, there are already four types according to the type of subsystem); (c) complex semiotic systems with two economising subsystems (this time there are six logically possible types); (d) complex semiotic systems with three economising subsystems (three logically possible types), and (e) complex semiotic systems with four economising subsystems (one logically possible type), is a powerful way of classifying semiotic systems according to their structural/functional properties. (All the more so as the classification is directly geared to systematic ways of describing semiotic systems.) The overall classification, in which only some of the categories have been filled in by concrete examples, requires the matrix-type representation shown in Table 7.1.

Table 7.1: logically possible types of semiotic system

simple signa only

simple and simultaneous signa only

simple and articulated complex signa only

simple simultaneous and articulated complex signa

simple figurae only

simple and simultaneous complex figurae only

simple and articulated complex figurae only

simple, simultaneous and articulated complex figurae

Simple system

An inventory of signa plus a cenematics to provide some of the forms of signa, e.g. traffic lights, semaphore

An inventory of signa plus a cenotactics to provide some of the forms of signa, e.g. Morse code

An inventory of signa plus a cenematics and a cenotactics working together to provide some of the forms of signa

A plerematics with the forms of simple signa provided by an inventory of figurae, e.g. Highway code

A plerematics with the forms of simple signa provided by a cenematics

A plerematics with the forms of simple signa provided by a cenotactics

A plerematics with the forms of simple signa provided by a cenematics and cenotactics working together

A plerotactics with the forms of simple signa provided by an inventory of figurae, e.g. Arabic number writing

A plerotactics with the forms of simple signa provided at least partly by a cenematics

A plerotactics with the forms of simple signa provided at least partly by a cenotactics

A plerotactics with the forms of at least some simple signa provided by a cenematics and a cenotactics working together

A plerematics as well as a plerotactics with the forms of simple signa provided by an inventory of figurae, e.g. ideographic writing

A plerematics and a plerotactics with the forms of simple signa provided at least partly by a cenematics, e.g. classical written Chinese

A plerematics and a plerotactics with the forms of simple signa provided at least partly by a cenotactics, e.g. alphabetic writing

A plerematics and a plerotactics with the forms of at least some simple signa provided by a cenematics and a cenotactics working together, e.g. spoken languages

e.g. system of map 'symbols'

200

SEMIOTIC PERSPECTIVES

7.4.1

Theory of Indices

In 'systemology', an extensive use was made of the notion of signum, provisionally understood as a 'two-faced' entity constituted by a fusion of formal and semantic properties (reminiscent, that is, of the Saussurean 'signe'). A definition of this notion will be seen to emerge from a deductive classification in which various types of index (of which signum is one type) are compared. Once again, the working out of a deductive classification presupposes a search for a pertinent dimension in which different types of index are to be compared. As a preliminary to this search, we should note that, by an 'index' is meant any 'two-faced' entity (that is to say, strictly speaking, a relationship) in which a type of signifier (manifested as an observable 'form') acts in a capacity of conveying a certain type of information (registered in the form of a 'message'). Adopting the convention of representing the mediating relationship 'acts in a capacity to convey' by the symbol'--,)', the general formulaic representation of 'index' will be: signifier-,) information. There being three 'components' involved in an 'index' - the signifier manifested by observable forms, the information registered as messages of a certain kind, as well as the mediating relationship represented by the '--,)'-there are three logically possible ways of classifying indices. Of these, however, two can be immediately eliminated as being possible but relatively trivial and unilluminating: (1) the substance that happens to manifest the signifier and (2) the substance of the message registered. (It has already been argued sufficiently that 'accidents' of substance do not provide pertinent dimensions of functional classification.) The pertinent dimension we ar~ left with is that of the mediating relationship that holds together the two 'faces' of an 'index' (and without which there can be no such thing as an 'index', nor a 'signifier', nor, for that matter, any 'information'). The basis of classification in this dimension boils down to inquiring as to the factors by virtue of which the mediating relationship has come about between a particular signifier and a particular information in a particular 'index'. As we shall find, there are, essentially, two types of factor that can bring about such a mediating relationship: natural (basically causal) links and conventions. The classification of indices takes off from this starting point and culminates (in a way that will be outlined subsequently) in the scheme shown in Figure 7.8. In the case of a natural index, the mediating relationship between signifier and information can be said to hold by virtue of regular, essentially causal or co-occurrential, links in the 'physical' world between observable manifestations of the signifier and observable objects that 'give substance to' messages registered. Thus, for instance, when lightning acts as a signifier of the information: 'impending thunder', it does so by virtue of the regular 'physical' link that we have come to expect habitually between lightning and thunder. Similarly, 'indices' made use of in medical diagnosis operate by virtue of symptoms regularly associated (in our knowledge of 'nature') with particular diseases.

AXIOMATIC SEMIOLOGY 201

Index

Signum

Natural index

Symbol

Sign

Proper symbol

Nonce symbol

Figure 7.8

Natural indices can be put to artificial uses (without in any way changing their status as natural indices). Thus, artificial devices such as thermometers, pressure gauges, warning lights indicating the malfunction of machines, and so on, profit from natural links by virtue of which their indices convey the appropriate information. In the same way, photographic images must be classified under the category of natural indices, since they 'represent' their subjects by virtue of the photographic process that, link by link, establishes a causal chain whereby light reflected from that subject causes an image to appear on photographic paper. As long as a causal, or co-occurrence, link can be maintained- no matter how complex a chain this may require - the index in question is identified as a natural index. However, as soon as such a chain is brokenbe it only in one small detail - in such a way that some element of convention needs to be stipulated in order to 'explain' the connection between signifier and information, the index can no longer be called 'natural'. That is to say, the natural link by virtue of which a particular 'index' operates (as an index) must be conceived by analogy with an electric circuit; contact needs to be broken only at one point, and the lights go off. Thus, an index that is 'partially motivated' by natural links (for instance, by a natural similarity between signifier and information) is not a natural index - just as a partially connected electric circuit is not a functioning electric circuit. For instance, a pictorial representation drawn by hand, or reproduced in print, may be motivated, as an index of the object represented, by a strong natural resemblance. Yet, in the absence of a complete natural link justifying (motivation is always a case of partial justification) the mediating relationship between signifier and information, such an index must be removed from the category of natural indices. (It would be possible to establish a classification according to degree and

202 SEMIOTIC PERSPECTIVES

type of motivation of indices, but this classification, apart from introducing a large number of imponderable factors, for example, what is the motivation, ultimately, of the sign· for 'female', originally for 'Venus'?, would be applicable only to the category of 'conventional' indices. In short, motivation or no motivation, an index is 'conventional' if its 'explanation' requires the slightest reference to conventions.)

0

7.4.2 The Notion 'Signum'

The discussion so far has established a clearcut and uncompromising borderline between natural indices, and indices that operate (at least to some degree) by virtue of conventions. The fact that some of these latter indices may be guessed when we do not actually know for certain the conventions behind them (for example, the stylised drawing of a human figure wearing a skirt- or is it a Scotsman wearing a kilt?- displayed on a ladies' toilet) does not in any way alter the fact that, in the final analysis, such indices are mediated by virtue of conventions. (A category of visually, or audially, motivated index could, in principle, be set up- we may even call this category 'iconic'- but the fact remains that members of this category will be conventional indices, rather than belong to some third class intermediate between natural indices and conventional indices.) The term signum refers in axiomatic semiotics to the total set of conventional indices; that is to say, indices in which the mediating relation between signifier and information hold by virtue of a convention (an 'arbitrary' element as opposed to the causal links in natural indices). Thus, for instance, even apparently 'marginal' cases such as

(a) sunrise---?'time to attack' (b) red---?' danger' (c) [k'uk'uw]-'a particular species of bird' (d) [brett]---?' a particular kind of noise' etc. fall fully into the category of signa, as they are dependent on conventions for their operation as indices. 7.4.3

Subcategories of 'Signum'

The further subclassification of signa into signs and symbols is based on a distinction between conventions that are 'fixed' within the semiotic system to which a given signum belongs, and conventions that are 'occasional' and 'extra-systemic'. Signa that are used (and understood) entirely in terms of intra-systemic conventions that do not have to be supplemented by the stipulation of 'operational' conventions of an occasional and extra-systemic nature are designated as signs. Signa that depend (for their use and interpretation as indices) on particular ad hoc conventions are classified as symbols. Thus, for instance, the signum '[k'uk'uw]---?a particular species of bird' is used and understood according to the intra-systemic (linguistic) con-

AXIOMATIC SEMIOLOGY 203

ventions of English - once these conventions have been grasped, the information value (denotation) of this signum is fixed within certain limits (all and only the birds that belong to the particular species in question). In other words, 'cuckoo' is a (linguistic) sign in English. In contrast to this, the signum 'sunrise----:. time to attack' does not belong to any established semiotic system; consequently, there is no question of the convention that makes it operational (as an index) being an intrasystemic one. We are dealing here with an ad hoc convention established for particular purposes and on a particular occasion (thence the term 'occasional' convention); the signum in question is a symbol. For all that the example of 'sunrise----:. time to attack'- being dependent on an ad hoc convention explicitly agreed upon (by the attackers)- seems very different from the case of a 'nonsense word' such as 'jabberwock', the two instances both belong to the same category of symbols. The difference between them lies in the distinction between conventions that are explicitly stated, and conventions that are left tacit. The signum in Lewis Carroll's poem- '[dzreb~w~k]----:.a particular kind of fierce mythical beast' - has the context of the whole poem as the factor that tacitly establishes a kind of ad hoc convention, a convention that is 'occasional' and extra-systemic, because it does not belong to English, but to the peculiar 'world' of the poem itself. While within the category of signa wholly dependent for their mediating relationship on 'occasional', extrasystemic conventions one may further distinguish between explicit and tacit symbols, these symbols clearly belong, in the first place, to a single category. This category is designated in axiomatic semiotics by the term 'nonce symbol'. (Using the word 'nonce' acts both as a reminder of the 'occasional', 'one-time' nature of the conventions involved, and as a hint at the presence of so-called 'nonsense words' in this category of symbols.) While sign and nonce symbol represent two extremes of signa - those whose mediating relation holds entirely by virtue of intra-systemic conventions, and those whose mediating relation holds entirely by virtue of extra-systemic, 'occasional' conventions - there is also an intermediate logical possibility. In short, there is a category of symbols whose information values depend in part on certain general, conventional limitations of an intra-systemic nature, but whose use and interpretation depend also in part on supplementary 'occasional' conventions that need to be separately stipulated. Once again, the 'occasional' conventions may be given explicitly (in the form of a 'naming' process, or a 'definition'), or they may be tacit from a context or situation. (We shall, for the most part, ignore this explicit-tacit distinction.) For instance, conventions generally accepted within the system of algebraic symbols limit the information values that can (in algebraic indices) be connected to such signifiers as a, b, c, and so on- these are not variables over the whole universe, but only over numbers (the algebraic signifier a can stand for any number, but it cannot stand for, say, 'elephant'). The mediating relation in an index 'a~any particular number' is intra-systemically conventional; it is a matter of the conventions of algebra as such. On the other hand, in given algebraic operations, the constant informa-

204

SEMIOTIC PERSPECTIVES

tion value (constant, that is, for that operation only) is not just any particular number, but some number different from all others. There is, however, nothing in the systemic conventions of algebra that can determine the particular values that algebraic symbols have in particular operations. This we can only know by virtue of 'occasional' conventions set up for the particular operation in question. Thus, when we say 'let a equal235', we are explicitly establishing an ad hoc convention by virtue of which we create an 'occasional' index 'a = 235'. (Similarly, when we place x in an equation in which it is the only unknown value to be calculated, we are tacitly creating an index in which x has the predetermined and calculable value attributed to it by the context of the algebraic equation in question.) Superficial differences apart, there is a remarkable similarity between the case of 'a~any particular number' and that of 'Peter~any animate male creature'. The fact of the matter is that, wholly unconventional uses apart, a system of first names conventionally limits the use of 'Peter' as a 'name for' animate (mostly, but not necessarily, human) creatures of the male sex. Once again, 'Peter' is not a variable over the whole of the universe, but only over a well-defined subset of it, a subset determined by intra-systemic conventions. On the other hand, to put it bluntly, not just any male individual can be, indiscriminately, addressed as 'Peter', only those who have been so named. There is, however, nothing in the systemic conventions of first names by virtue of which it is possible to know which particular individuals within the range of male creatures rna y, and which rna y not, be referred to as 'Peter'. The only way we can know this is by virtue of 'occasional' conventions that, in an ad hoc manner, establish a link between the symbol 'Peter~any animate male creature' and some particular human being, or domestic pet. The establishing of such ad hoc conventions for humans is considered important enough to constitute a ceremonial or official procedure (baptism, registration of birth, deed poll, and so on) in order to regularise and advertise these conventions. The category we have been discussing includes, on the one hand, indices such as algebraic symbols, and, on the other hand, indices usually referred to as 'proper names'. This fact motivates the use of the term 'proper symbol' for designating the category in question. Another way of explaining the category of proper symbols is by pointing to a 'discrepancy' between the messages they may carry (as far as systemic conventions are concerned) and the messages they do carry (in particular instances). The messages a symbol such as 'a~any particular number' or 'Peter~any animate male creature' may carry are typically conceived as a range of variables. This, however, is also the case for signs such as 'rr~¥' or '[k'uk'uw ]~a particular species of bird'. Where signs and proper symbols differ is in 'getting from' the range of variables that they may denote to the range of actual messages that they do denote. For signs, the transition is automatic within the system to which they belong: all and only measurements that conform to the value¥ can be denoted by the sign 'rr'; all and only members of the appropriate species can be denoted by the sign 'cuckoo'. For proper symbols, however, the transi-

AXIOMATIC SEMIOLOGY

205

tion needs to overcome a 'discrepancy': 'a' may denote any particular number, but the numbers it does denote are specific to the occasions of its use; 'Peter' may denote any particular male creature, but the creatures it does denote are only those who have been explicitly named 'Peter'. Without extra ad hoc stipulations (which we have called extra-systemic, 'occasional' conventions) it is impossible to overcome the discrepancy between 'may' and 'does' in the case of proper symbols. This is a typical feature by which proper symbols can be identified. The distinctions drawn between the categories of sign, proper symbol and nonce symbol are an interesting and, I think, original, aspect of axiomatic semiotics. However, the category that is most important from the point of view of the integrated theory as a whole is that of signum, since it is in terms of this notion that the 'systemological' classification of semiotic systems is couched; semiotics is concerned with systems containing signa, not merely with systems containing signs. It is, accordingly, the more general notion of signum, and not the more restricted notion of sign, that is to be further analysed in 'signum-theory'. 7.5.1

'Signum-theory'- the 'Expression' Aspect

The raison d' etre of a signum is the mediating relationship it creates (as an index) between a 'signifier' and an 'information value'. This mediation is, however, not a simple union of two irreducible entities, but rather a whole chain of successive relationships, for it is, ultimately, between (physical) emissions as the means of conveying messages, and messages conveyed by these physical means, that signa mediate. 'Signum-theory' requires, therefore, to analyse step-by-step the complete 'circuit' of relations linking the messages conveyed by realisations of a given signum (on particular occasions) and the physical means that are used to convey these messages (these concrete means are referred to as emissions). The two extremes between which the signum mediates as a central and abstract entity are emissions and messages. Within the signum as such (an entity of a high degree of abstraction, it must be remembered) there is a fusion of two equally weighted facets: the aspect of expression (symbolically: E) facing in the direction of the physical means (that is, emissions) and the aspect of content (symbolically: C) facing in the direction of messages conveyed. The way to capture the relationship between these two facets of the signum is by insisting that the signum is the union of these facets, while these facets are each other's logical converses. (The analogy of monogamous marriage relationships may help to clarify this conception: a married couple, z is the logical union of two relationships that are each other's converses, that is to say, of 'husbandhood' = man x in the capacity of husband of woman y and 'wifehood' = woman y in the capacity of wife of man x.) In the case of signa, the two converses are: expression (E) = class of figurae x in its capacity of having the plerological value y, and content (C) = plerological value y being the (semiotic) capacity of class of figurae x. A signum is thus seen as the union of an expression and a content (like a bottle of wine is the union of a solid container and its liquid contents), but

206

SEMIOTIC PERSPECTIVES

with each of the united terms being reducible to a relationship between identifiable factors. The class of figurae (symbolically {p }')-in its turn, further reducible, ultimately to concrete realisations of figurae in actual communication (as we shall see below)- is an abstraction and generalisation representing the range of physical means from which concrete emissions of the signum may be drawn. The plerological value is a relative and negative value (in the Saussurean sense) created by the network of oppositions to which a signum belongs by virtue of its membership of (the plerological level of) a semiotic system. In short, plerological value is definable as the set of oppositions between a given signum and all the other signa from which it differs within a given semiotic system (for example, the plerological value of '/red and amber/ = prepare to go' is determined in the system of traffic-lights as: {-'/red/= stop', -'/green/= go', -'/amber/ = prepare to stop'}). A particular plerological value is symbolically represented by sY. Thus the formulaic representation of expression and content is: E = {p }' R sY and C = sY)~. {p }', where R stands for the relation 'in its capacity of having', and R for the converse relation 'being the (semiotic) capacity'. Converses mutually imply one another, as well as each of them being in mutual implication with the entity resulting from the union of the converses; that is to say, given that S(ignum) = E and C, and that E and C are converses, it is also the case that S, E and C mutually imply one another (see Figure 7.9).

s

E+-----------t>C Figure 7.9

Reducing the expression of a signum to the concrete means by which that signum conveys individual messages on particular occasions is a complex matter that takes us through a whole scale of successive steps of reduction: reduce {p }' RsY (i.e. expression) to its members p 1 R sY, p 2 RsY . .. p" RsY (where p 1 • • • p" = {p }'). Each of these members is a variant allomorph of the expression (for example, contextual variants 'liz/ R SpJur.', 'lzl Rsplur.', '/~n/ Rsplur.', etc. of the signum 'plural' in English, in the respective contexts of 'boxes', 'hills', 'oxen', and so on). (2) Reduce {p }' to its members (abstracting away from plerological value) {p\p2 .. . p"} = {p}' Each of these members is a figura (with a potential for differentiating signa, but now abstracted away from the realisation of that

(1)

AXIOMATIC SEMIOLOGY 207

(3)

(4)

(5)

(6)

potential (for example, liz/, lz/, /;}n/, etc. as abstracted from their plerological capacities, and viewed as mere means of systematic formal differentiation in English). Reduce each figura to a value in a cenological subsystem within which it is opposed to other figurae, and has its own range of variation over its appropriate domain of substance PI =:= {j}· Rtf' {j}• represents the class of units of substance that may have the same cenoiogicai value; tf' represents a particular cenological value, defined by the oppositions between a particular figura and all the other figurae which it differs from within the (cenological level of the) same semiotic system (for example, the cenological value, in the Morse code, of the figura /short/ is '-/long/'). R continues to represent the relationship stated as 'in its capacity of having'. Reduce {J}• Rtf' to its members {JI Rtf', f R db . .. f Rtf'} (where {JI ... /"} = {f}•) Each of these members is an allo-figura, a 'substantial' variant of the figura in question (for example, the Morse code figura /long/ can vary in substance from medium to medium, and can vary also in actual length according to the pace of transmission). Reduce {f}" to its members (abstracting away from cenological value) {f,f ... f} = {f}• Each of these members can be termed an 'etic form' (in the case of languages, a 'phonetic form'). While reducing figurae to observably different variants in substance (for example, 'tongue-trilled' [r] and 'uvular' [R] are variant substances of French /r/), 'etic forms' are still not irreducible, concrete phenomena- they must be conceived, in turn, as 'classes of approximations' (for example, 'uvular' [R] is produced in more or less subliminally different ways on each occasion- these pronunciations 'approximate' to one phonetic specification, and are, by courtesy, said to be 'the same'). Reduce each 'etic form' to a class of concrete phenomena = {i}' {i}' = {P, P . .. i"} where {i 1 • • • i"} represents the class of actual, individual emissions approximating to the specification of a particular 'etic form' (for example, all individual pronunciations, by any speaker of French, on any given occasion, of the phonetic form [R]).

r

As we have seen, the reduction of the expression of a signum to a class of concrete and individual emissions, that class whose members can potentially realise the message-bearing means implicated in the given signum, involves a highly complex chain, taking us through a whole scale of abstractions. (Note that the notion figura finds its definition on this scale.) A further scale of abstractions will be required for dealing with the

208

SEMIOTIC PERSPECTIVES

content aspect of the signum, and the relationship this bears to actual messages. However, before we explore this 'semantic' side of the signum, we shall sum up in Figure 7.10 the scale of abstractions that systematically relates expression to concrete emissions.

Signum

E +--------+ C

E={p}xRsY

Plerology

and

Allomorph

Figura

Cenology Allo- figura

Etic form

Substance Individual emissions

Figure 7.10

7.5.2

Signum-theory- the 'Content' Aspect

The content of a signum, while defined in terms of a plerological value (as the oppositional, negative property in terms of which the semiotic capacity of the signum is measured), has as its positive semiotic property

AX/OMA TIC SEMIOLOGY 209

the correspondence of the signum with a specific information value. Such a correspondence of a signum with its 'semantic' properties is most frequently represented as a 'triangular' relationship; for example, Ullmann's adaptation (Ullmann, 1957, 1962) of the famous 'Ogden and Richards triangle' (Ogden and Richards, 1923) (see Figure 7.11}. name

s e n s e - - - - - - - - - - - - - - referent Figure 7.11

In axiomatic semiotics (Hervey, 1979), a 'rectangular' model (henceforth referred to as the 'denotational rectangle') replaces the usual 'referential triangle'. The role of this 'denotational rectangle' is to relate, in clearly specified ways, the abstractions designated by the terms 'signum', 'information value', 'signum-token' and 'denotatum', both to one another, and to their concrete correlates in the realm of actual phenomena. In this exposition, the 'denotational rectangle' will be built up from a confrontation of the Peircean idea of type versus token with the Saussurean idea of form versus substance. In connection with the Peircean idea of type, we note that types are both abstract and generalising- for instance, a type such as '£1.00' is both a 'notional figure', and a value comprised of a potentially unlimited class of its tokens (it is both an abstraction and a generalisation of all manifestations of a particular monetary value). Tokens of a given type are members of the general category brought into one abstract entity by that type; that is to say, tokens are specific or particularising (as opposed to being generalisations). They stand for particular manifestations of a category. Since, however, it would be absurd to say that a category that is abstract consists in a class of members that are concrete, and since tokens, themselves, are descriptive of (stand for) concrete particulars, rather than being concrete particulars, it follows that tokens are also abstractions (for example, 'a£1.00 note', as a token of the type '£1.00', is a 'notional', though particular, representative of the type in question). Tokens, in short, are abstract and particularising: particular by virtue of being individual members of a general category, but abstract by virtue of representing certain relevant aspects of a given concrete phenomenon (for example, the token 'a £1.00 note' represents the inherent aspect of financial value attached to a particular piece of paper bearing the appropriate physical markings). In connection with the Saussurean notion 'form', we note also that forms are both abstract and generalising. A form is a value infinitely realisable in its (infinitely renewable) substance - for instance, the form

210

SEMIOTIC PERSPECTIVES

'£1.00' is both an abstraction and a generalisation over the whole range of substances (pieces of paper currency, sets of coins, and so on) that have the same value. Substance - and the agglomeration of concrete physical entities that constitute the variable substances of a given form- concretely manifests form. It is, in other words, concrete (as opposed to being abstract). Since concrete entities are individual and particular in 'reality', the phenomena that constitute the substance of a form are concrete particulars. When applied to the plerologicallevel of a semiotic system, the notion type covers signum and the information value corresponding to a signum (that is, these two abstractions are plerological types, both being abstract, generalising, categorial entities). Similarly, and for the same reasons, signum and information value are the forms (values) pertaining to the plerologicallevel of a semiotic system. Thus, in plerology, signum and information value belong to the highest level of abstraction, the level in which type and form coincide with one another. These abstractions - signum and information value - need to be correlated both to their appropriate tokens and to their appropriate substances. Token and substance cannot, however, be incorporated on a single level, for, although both involve particulars, tokens are abstract, while substances are concrete phenomena. Tokens of signa (designated, for lack of a better term, as 'signumtokens') are one step removed from signa- they stand to a signum as members to a class. Analogously, tokens of information values (designated by the term 'denotatum') are one step removed from information values, standing to an information value as members to a class. Substance is relegated to a further level- that of concrete phenomena. General ising (type):

Abstractions

Particularising (token):

Concrete phenomena:

signum I

I I I I I

signum-token

semiotic event

(emission carrying a message)

Figure 7.12

corresponds to

denotes

information value I

I I I I

denotatum

denotable

AXIOMATIC SEMIOLOGY 21 1

On this level, the particular phenomena corresponding to signa (realisations of signa) are semiotic events. Each such event is, itself, constituted of the interplay between an emission (cf. Section 7.5.1, p. 205) and the message it conveys. The concrete phenomena that are the substances of information values (that is, concretisations of denotata) are the 'objects' designated by the term 'denotable'. 'Denotables' are, on the one hand, the subject matter of messages and, on the other, entities susceptible to being experienced independently of communication (Hervey, 1979). The 'denotational rectangle' built on the basis of the foregoing discussion can be diagrammatically represented as in Figure 7 .12. The relationships between terms in the 'rectangle' in Figure 7.12 can now be precisely stated:

Signum to signum-token: a relationship of class to member (that is, a signum is a class of its signum-tokens). (2) Signum to semiotic event: a signum is a model for a particular class of concrete semiotic events; conversely, a particular class of semiotic events constitutes the potentially infinitely repeatable realisations of a given signum. (For example, every time a driver observes a circular design with a blue field enclosing a white arrow, interpreting its message by following the direction of the arrow, the resulting event is, regardless of all its material differences from other such events, classified as a realisation of one and the same Highway Code signum). (3) Signum-token to semiotic event: while signum is a model for a class of semiotic events, and signum-token is a member of the signum, a particular signum-token is a model for a particular semiotic event. That is to say, signum-token is an abstract model, but a model with an extreme specificity, standing in a one-to--Dne relation with a single semiotic event (for example, a cine-film recording of a driver observing a particular Highway Code signal on a particular occasion would be a model with the required degree of specificity). (4) Signum-token to denotatum: a particular signum-token conveys a particular piece of information (in this way, the model reflects the message necessarily conveyed in the course of the semiotic event represented by the signum-token); we shall say that the particular piece of information conveyed is denoted by the signum-token (this piece of information is a denotatum). Since what is essentially the same denotatum can be conveyed by an unlimited number of tokens of the same signum, though a given signum-token can only denote one single denotatum, the relationship between signum-token and denotatum is a relationship of one-way implication (for example, the grapheme 'a' of English can be denoted by a signum-token involving a 'short' followed by a 'long' flash with the torch I bought last week, or by another signum-token I have just tapped out on my desk). (5) Denotatum to information value: a relationship of member to class; an information value is the generalisation of a class of denotata, each of whose members can (according to the conventions gov(1)

212

SEMIOTIC PERSPECTIVES

erning a particular signum) be denoted by signum-tokens of the same signum; conversely, a particular denotatum instantiates (in the way that a specimen instantiates a species) a particular information value (for example, the grapheme 'A' of English, the grapheme 'a' of English, the grapheme' A' of French, the grapheme 'a' of French, the grapheme 'A' of German, the grapheme 'a' of German, and so on, are all members of the same class of denotata, the class generalised as the information value ('the alphabetic letter "a" ')of the Morse code signum ('/short+ long/~ the letter "a" '). (6) Signum to information value: a correspondence between two types; since (a) signum is conceived of as a determinate class of signumtokens, (b) information value as a determinate class of denotata, and (c) each signum-token denotes its own determinate denotatum, it follows that the correspondence of signum to information value is not only conceivable as a direct correspondence between two types, but also as the result of the fact that each individual token of a signum denotes its own denotatum. A given signum determines its class of signum-tokens and each such signum-token determines its denotatum; consequently, the class of signum-tokens ( = signum) determines a class of denotata: the information value of the signum, seen as the potential range of denotata that can be denoted by tokens of the signum. Once again, a one-way implication emerges between signum and information value: a signum implies the class of denotata designated as its information value. (This one-way implication allows synonymy between signa; two or more signa may have the same information value corresponding to them.) (7) Denotatum to denotable: a denotatum faces at once in two directions: in one sense, it is the information-bearing capacity of a signum-token (that is, it is a term necessarily linked via a token to a signum); in another, it is an entity endowed with properties that make it a part of practical experience (that is, it is an 'object' that can be reacted to in terms of practical, non-semiotic behaviour; it is not just something that can be 'referred to' or 'talked about', but also something that can be 'experienced' in other ways). Outside of its functioning as the information-bearing capacity of a signum-token, a denotatum, while it does not lose its potential for being denoted, becomes a mere denotable. We may say that a denotable is the 'substance' that gives a denotatum its basis in extra-semiotic reality. (The wide sense in which the term denotable is to be used- including 'objects', 'qualities', 'processes', 'relations' and 'complex circumstances' that are networks of the former- can only be hinted at in the context of this discussion. Reference may be made to Harre (1970a, 1970b) and to Hervey (1979). Similarly, I can only allude to the notion of 'functional ostension'- the idea that 'experiential reality' should be viewed, by analogy with any 'functional' system, as a network of opposed values - as the basis for endowing denotables with an extra-semiotic identity. On this point, reference may be made to Hervey (1979, 1980).)

AXIOMATIC SEMIOLOGY 213

What is important about the 'denotational rectangle' for our present purposes is that it locates content as that aspect of signum that faces in ~he direction of information value, and that it 'analyses' both signum and information value, relating them, ultimately, to their concrete correlates in the realm of actual phenomena (semiotic events and denotables, respectively). In this way, the 'denotational rectangle' fulfils its purpose by setting up a scale of abstractions that link signum to its 'semantic' aspect. 7.6 7.6.1

Some Examples of Axiomatic Semiotic Analyses Roman Number-writing

The system is analysed as the interlock of a simple cenological inventory with a plerologicallevel containing a (rudimentary) plerematics and a (more sophisticated) plerotactics. The cenological inventory contains the figurae: II/, IV/, IX/, ILl, /C/, ID/, /MI and rl. The basic inventory of the plerematics consists of the signs: 'I~ 1', 'V~5', 'X~10'. 'L~50', 'C~100', 'D~500', 'M~1000' and,-~ x1000' (e.g. ·v~s x 1000 = 5000'). For plerological purposes, this inventory forms a hierarchy, with 'I~ 1' having the lowest 'rank', and the hierarchy beyond 'M~ 1000' being continued in ·v~5000', ·x~ 10000', etc. Elements belonging to an odd 'rank' in the hierarchy behave differently (in constructions) from elements of an even 'rank' (odd 'rank': '1~1', ·x~10', ·c~100', etc., even 'rank': ·v~s', 'L~50', ·n~500', etc.). Rank1 Rank2 Rank3 Rank4 Rank5 Rank6 Rank? Rank8 Rank9 etc.

'1~1' ·v~s' ·x~1o' 'L~50' ·c~10o' ·n~soo'

Plerematic hierarchy (in ascending order)

'M~1000'

·v~sooo· 'X~10000'

Plerematic constructions are of two types: Type A: simultaneous bundles of two, or three identical signs (e.g. 'II~2', 'III~3', etc.). Such constructions only occur with signs belonging to an odd 'rank' in the hierarchy (e.g. 'XX~20', but not *'VV~10').

Type B: simultaneous bundles consisting of a sign from 'rank' 2 to 'rank' 7 (i.e. ·v~~·- 'M~1000') and the sign'-~ x 1000' (e.g. ·v~s x 1000 = 5000', 'X~10 X 1000 = 10000', etc.).

214 SEMIOTIC PERSPECTIVES

Plerotactic constructions operate with the subordination of a lower 'ranking' sign to the next higher 'ranking' sign with which it co-occurs in a given complex. There are three types of position (one nuclear and two subordinative types) in the plerotactics of Roman number writing:

Nuclear position (the semantic function of this position is simply to attribute the face value to the sign standing in that position; e.g. ·x~10' has its face value in 'XVII~17', where it stands in nuclear position with respect to 'VII-~ 7'). (2) Pre-nuclear position (the semantic function of this position is to subtract the face value of the sign standing in that position from the face value of the sign standing in nuclear position; e.g. in 'XC~90', ·x~10' has the actual value '-10'). (3) Post-nuclear position (the semantic function of this position is to add the face value of the sign standing in that position to the face value of the sign in nuclear position; e.g. in 'XV~15', where it stands in post-nuclear position, ·v~5' has the actual value '+5').

(1)

Plerotactic conventions: (a) (b)

Only simple signs, and plerematic complexes of Type B (e.g. ·v~5000') can stand in nuclear position. If the pre-nuclear position is filled, the post-nuclear position can only contain signs of a 'rank' lower than that of the sign filling the pre-nuclear position (e.g. 'XCV~95', but not *'XCXI~101, nor *'XCL~140).

Pre-nuclear position can only be filled by simple signs or plerematic complexes of Type B (e.g. ·v~5000'). (d) Pre-nuclear position can only be filled by a sign one 'rank' below the 'rank' of the nuclear sign if that nuclear sign is of even 'rank'; if the nuclear sign is of odd 'rank', pre-nuclear position can only be filled by a sign two 'ranks' below that of the nuclear sign (e.g. 'XL~ - 10 +50= 40', but not *'VL~- 5 +50= 45'; ·xc~- 10 + 100 = 90', but not *'LC~ - 50+ 100 =50', nor *'VC~ - 5 + 100 = 95'). (e) Post nuclear position can only be filled by signs of a lower 'rank' than that of the sign filling nuclear position (e.g. 'VI~ 5 + 1 = 6', 'LX~50 + 10 = 60', but not *'VX~5 + 10 = 15', nor *'XC~ 10 + 100 = 110'). (f) Post-nuclear position may be filled by (and is the only position that may be filled by) plerematic complexes of Type A. It may also be filled by plerotactic complexes whose highest 'ranking' constituent is (i) of lower 'rank' than the filler of nuclear position, if pre-nuclear position is not filled (e.g. 'CXL~100 + ( -10 + 50) = 140', 'CLXV~ 100 + (50 + (10 + 5)) = 165', but not *'LDXV~50 + (500 + (10 + 5)) = 565'); (ii) of lower 'rank' than the filler of pre-nuclear position, if the latter is filled (e.g. 'XCVI~ - 10 + 100 + (5 + 1) = 96', but not *'XCXI~- 10 + 100 + (10 + 1) = 101').

(c)

AXIOMATIC SEMIOLOGY 215

In terms of the foregoing inventories and constructional conventions, any Roman numeral expression can be described or predicted - the system is exhaustively described by these means. (I am indebted to Sheena Gardner, University of St Andrews, Scotland, for her collaboration in producing this semiotic description of the system of Roman number writing.)

7.6.2

Ideographic writing: Classical Chinese

There is, of course, no question of attempting here a full-fledged description of a complete writing system with all its inventory and conventions. We can only indicate the 'skeletal' structure of ideographic writing, as exemplified by classical Chinese, in terms of its constitutive subsystems, and of the partial isomorphism between these subsystems and subsystems of the corresponding spoken language. These indications will, however, be found to be revealing of the differences between types of writing system. Classical Chinese has a fully fledged plerology that is the interlocking of a plerematics (forming written 'words' as simultaneous bundles of simple signs of the writing system) and a plerotactics (forming written complexes of a syntactic type from the 'words' produced by the plerematics). On the plerological level, therefore, classical (written) Chinese is structurally analogous with semiotic systems of the spoken language type. Furthermore, there is a close one-to-one isomorphism whereby simple signs of the writing system correspond to those of a spoken language, 'words' of the writing system with 'words' of a spoken language, syntactic complexes of the writing system with syntactic complexes of a corresponding spoken language. Both the inventories, and the structures of classical (written) Chinese and of a corresponding spoken language run closely parallel with one another on the plerologicallevel (reading or writing can, on this level, take the form of a word-for-word transition from one system to the other). On the cenologicallevel, however, classical (written) Chinese shows only the most rudimentary degree of structural economy. Characters, as figurae, may be analysable into so-called 'radicals' (for example, I~/ into 111 and 1-ft! I, !t:fl into/~ I and I I;: I) that are 'systematically' recurring cenological elements. These complex characters (figurae) can, consequently, be regarded as (economically constructed) simultaneous bundles of figurae (radicals). Accordingly, we recognise a cenematic subsystem in classical (written) Chinese; this subsystem, however, exhausts the cenological economy of the writing system. While the presence of a cenematics in classical (written) Chinese argues for a degree of structural analogy between the cenologicallevel of this system, and that of a spoken language (compounded of a cenematics and a cenotactics), there is no isomorphism whatever between the inventory and constructions of the written cenematics, and the cenematics of a corresponding spoken language. 'Radicals' cannot be transliterated into units of the spoken language (they cannot be 'read aloud' at all). As for the cenematics of the corresponding spoken language, culminating in the

216

SEMIOTIC PERSPECTIVES

cenematic complexes that linguists call 'phonemes', its elements are not 'represented' in classical (written) Chinese in any way (this is our way of saying what, in common parlance, is often expressed as the 'non-phonetic nature of ideographic writing'). Classical Chinese

Writing system syntactic /complexes

Plerotactics

Plerematics

I

""

Spoken language syntactic complexes Plerotactics

words

words Plerematics

basic signs

signs

Cenematics ('radicals')

Cenotactics Cenematics

Figure 7.13

7.6.3

Syllabic Writing: Japanese Kana

The 'skeletal' structure of Japanese Kana is, as we shall see, somewhat different on the cenological level from that of an ideographic writing system. On the plerological level, however, such a system is both analogous with spoken languages (constituted by an interlocking plerematics and plerotactics) and displays a close one-to-one isomorphism (in terms of inventory and constructions of signs) with spoken Japanese. (Reading and writing take the form, on the plerologicallevel, of a word-for-word transition from one system to the other.) On the cenologicallevel, Japanese Kana displays a highly systematic cenotactic economy: the basic figurae, for example,~lh I, 1+-tl, I? I, I r I, ancl so on, can be combined into articulated constructions, for example, lbh, lfhl, 1-+-tb I, and so on. Thus the system of Kana is actually a semiotic system with two articulations (both a plerological and a cenological articulation are instanced). Where the system differs from alphabetical writing, however, is the point of correspondence between written figurae and figurae of spoken Japanese: this point of correspondence is through the one-to-one isomorphism of basic figurae of the writing system with 'phoneme-chains' (cenotactic constructions) of spoken Japanese. (These 'phoneme-chains' are in fact the 'syllables' of

AXIOMATIC SEMIOLOGY 217

spoken Japanese; thence the term 'syllabic writing' in token of the 'syllable-to-basic-written-figura' correspondence between the two systems.) Japanese Kana and Spoken Japanese

Writing system

Spoken language

syntactic +--+--+--+syntactic complexes complexes Plerotactics

Plerotactics words +----+--+--+WOrds

Plerematics

Plerematics basic signS+-+--+-.basic signs articulated constructions

'syllables' Cenotactics 'phonemes'""-._

Cenotactics basic figurae

'1-,-C-en_e_m_a-ti-cs-.,

/

basic figurae

Figure 7.14

7.6.4 Alphabetic Writing: English

The 'skeletal' structure of written English (in terms of constitutive subsystems) is analogous with that of Japanese Kana: we find in it the same three subsystems (plerotactics, plerematics and cenotactics). Furthermore, there is a high degree of isomorphism between (a) basic signs of the writing system and those of the spoken language; (b) words of the writing system and those of the spoken language; (c) syntactic constructions of the writing system and those of the spoken language. (It is only relatively rarely that a written sign can be 'read' in alternate ways- e.g. 'e-i-t-h-e-r' as [ail'i~J or as [i:l'i~]; or two written signs 'read' in identical ways- e.g. 'C-h-r-i-s-t-m-a-s' and 'X-m-a-s' as [krism~s]·) On the cenological level, there is no qualitative difference between Japanese Kana and alphabetic writing (both have a cenotactic economy). Where there is a divergence is in the point of correspondence between figurae of written English and figurae of spoken English: this point of correspondence is through a degree of isomorphism (which we know to

218 SEMIOTIC PERSPECTIVES

Alphabetic Writing Spoken language

Writing system syntactic complexes

syntactic complexes Plerotactics

Plerotactics words

words Plerematics

Plerematics

I

basic signs

basic signs

articulated constructions

articulated constructions

r Cenotactics

I.__ basic figurae

'phonom")

Cenotactics

I

Cenematics

I

basic figurae

Figure7.15

be far from perfect because of the vagaries of English spelling, a spelling that reflects too much of historically earlier stages of English to be able to stand in close isomorphism with present-day spoken English) between basic figurae (graphemes) of the writing-system and 'phonemes' of spoken English.

8 Semiotics as a Stylistic Theory: Bureau and Riffaterre General Remarks Attempts to fathom the mysteries of literary style span the whole history of Western thought. It is only recently, however, that style has been studied as a form of 'communication', a form sufficiently analogous with other communicational codes for it to be a potential object of semiotic study. Since it is, currently, by no means uncommon to subscribe to the view that semiotics holds the key to matters of style, we might have chosen any of a number of semiotic approaches to stylistics (Granger, 1969; Guiraud and Kuentz, 1970; Jakobson, 1979; Kristeva, 1974; Todorov, 1967; and so on). Of course, a full discussion even of the salient points that unite or divide the most influential semiotic-stylistic theories of recent times is out of the question in this volume. At the same time, leaving the entire area of these theories completely unexplored would have an unfortunate effect on the balance of its contents; the fact of the matter is that, to many scholars, 'semiotics' connotes the stylistic-semiotic study of literary works. The present chapter - in fact, the conjoining of two antithetical 'sketches'- is in the nature of a compromise by means of which I hope to convey some idea of the kind of theoretical and descriptive activity one might expect to encounter in a semiotic approach to literary style. Choice of the two authors selected is, in part, motivated by the very antithesis they offer between a 'text-based' (Bureau) and a 'reader-based' (Riffaterre) approach to stylistics. In addition, both theories recommend themselves by their relative accessibility, their up-to-dateness, and by the intrinsic merits of the ideas they encompass.

8

Bureau

8.1.1

Literature as a Semiotic Object; Background to Bureau's Styli sties

The identification of literature as an object of semiotic study is conditional, as pointed out by Bureau (1976), on the identification (both in principle and in practice) of various semiological levels that intermingle in constituting a literary work. A text, taken purely on the level of its sentences (that is, its units of

220 SEMIOTIC PERSPECTIVES

discourse), is a linguistic object- it does not become a literary object by mere 'osmosis', but only by the superimposition of further, and in that sense, 'higher', literary properties. The properties constitutive of the 'literariness' of a work should not be automatically assumed to be semiotic; what makes them so, however, is the fact that they -like other 'codes'- operate with recurrences and contrasts, and that- also like other 'codes'- these operations have communicative ends. The functioning of 'superimposed' codes (surcodages, cf. Granger, 1969) over a linguistic text is made possible by two general factors: (a) the substance in which linguistic messages are necessarily realised (whether phonic or graphic) can, at the same time, be put to other uses (for example, in uttering the sentence 'I am keen', the phonetic substance can be so modulated as to convey reluctance - creating thereby a contrast between linguistic message and 'superposed' message; a poem about the Crucifixion can be typographically set out in the form of a cross- creating a recurrence between the linguistic message and the 'superposed' message); (b) the form (in so far as that form is a matter of choice) in which a linguistic message is couched can, at the same time, be put to other uses (for example, the particular form in which the linguistic message of 'Excuse me, but you make me sick' is couched, conveys, over and above the purely offensive linguistic message, a conciliatory message of polite apology - creating a contrast between the linguistic message and the 'superposed' message; choice of 'May I say how much I admire your work?' conveys, over and above the linguistic message, an expression of respectful admiration carried by the respectful manner of address, creating a recurrence between the linguistic message and the 'superposed' message). It should be obvious that the 'non-primary' use of linguistic substance or linguistic form for the operation of a superposed code is conditional on the substance or form in question being primarily that of an underlying linguistic code. Equating literary style with the operation of superposed codes on a necessary foundation of a linguistic text makes it, furthermore, apparent that the key word of stylis tics is 'strategy'; to say something is to avail oneself of a basic linguistic code- to say it in a strategically apt way is to avail oneself of a 'superposed' stylistic code. Methodologically, as Bureau (1976) points out, a knowledge of the underlying linguistic code, and a full analysis of the text as a linguistic object, are a necessary condition for the subsequent identification and analysis of superposed codes operating over the linguistic text. Admittedly, any level of linguistic form (or of the corresponding substance) may become available for 'strategic' use by a stylistic code. Thus, for instance, the structure of a text in terms of syllables, or in terms of 'graphemes', may give rise to contrasts and recurrences (rhymes, assonances, metric variation, and so on), each level of structure contributing thereby a level of stylistic coding. However, Bureau's attention is specifically focused on one particular linguistic level (though he is well aware of others), that of syntactic structure, and the strategic use of syntactic structure for stylistic purposes. It is in this area that Bureau's stylistic-semiotic theory is extensively elaborated and exemplified.

STYLISTIC SEMIOTICS

8.1.2

221

Syntactic Analysis and the Stylistic Aspects of Syntax

Bureau has a great advantage in being able to avail himself of a detailed functional analysis of syntactic structure (Bureau, 1978). We are, unfortunately, unable here to cover the principles and practices of Bureau's syntactic model, but must instead imagine that sentences analysed according to this model serve as a background to identifying the interplay, in a text, of syntactic forms and substances. This interplay constitutes the 'strategic' operation of syntactic features in the realisation of elements of a stylistic code. (It must also be understood that the text has been prepared for syntactic analysis by a segmentation into sentences - not necessarily in a way corresponding to the 'sentences' indicated by punctuation, but those that answer Bureau's syntactic definition of selfcontained linguistic units; Bureau, 1976, 1978.) On the basis of properties precisely specifiable as a result of syntactic analysis, there emerge three types of potentially stylistically relevant characteristic: (a) length (quantifiable from 1 to n); (b) complexity (quantifiable from (1) ton) and (c) structure. These three characteristics will be discussed under separate headings below. 8.1.3

Syntactic 'Length' as a Potential Carrier of Style

The 'length' of a sentence is the numerical index of the number of its constitutive syntactically minimal elements (that is, the so-called syntaxemes). Thus, while the sentence 'Look!' contains only one syntaxeme (index of length = 1), some sentences in literary texts may run into hundreds (for example, the Proustian sentence analysed in Bureau (1976) has an index of length of 222). It is fairly obvious that the extent of maximum, minimum and average sentence length can be a salient 'strategic' feature of the style of a literary text. While individuation of a literary style (that of a text, or, even more generally, that of an author) is, of course, never solely a matter of sentence length, sentence length is clearly one of the features that may play a significant role in characterising an individual style. An exact measure of sentence length contributes, for instance, to a specification of the syntactic aspect of Proust's literary style- a specification that is highly relevant to appreciating the particularities of that style. Given an exact measurement of sentence length, a number of important statistical operations become possible: (a) computation of the average sentence length in a literary text; (b) comparison of maximum, minimum and average sentence length from one text to another, or from one author to another; (c) plotting of the distribution of shorter and longer sentences in a given text. The latter may be of particular interest as a 'strategic' device establishing a kind of 'rhythmic' pattern in a text. In particular, the juxtaposition of a relatively short and a relatively long sentence creates a contrast in that pattern, whereas recurrences will be found in the distribution of sentences of (approximately) the same length at different points in the text. Where the contrasts and recurrences that make up this 'rhythmic'

222 SEMIOTIC PERSPECTIVES

textual pattern can, furthermore, be put into correlation with a textual or thematic meaning that they serve to bring into relief, they constitute fully fledged carriers of stylistic coding. To take a crude example (the genuine examples furnished in Bureau (1976) are too extensive and complex to include here), in a text where a number of relatively long sentences are 'sandwiched' between two minimally short sentences, 'Birth.' and 'Death.', the patterning of sentence length could be correlated with a stylistic meaning paraphraseable as: 'the seemingly long-drawn-out efforts of Life are bounded by the decisive incidents of Birth and Death'. Syntactic length may create a 'rhythmic' patterning between sentences. It may also lend a sometimes unexpected balance to the strategic construction of a text through the recurrence of structurally isolable groups of comparable length within sentences. On Bureau's analysis, for instance, Proustian sentences consisting of over two hundred syntaxemes- sentences that, on the surface, appear to be merely 'rambling'- turn out to be remarkably balanced in terms of a patterned distribution of internal syntactic groups of comparable length. Again, where such patterns can be put into correlation with thematic features (for example, they give stylistic emphasis to a particular 'privileged' sign in the text) these patterns are carriers of a superposed stylistic coding. 8.1.4

Syntactic 'Complexity' as a Potential Carrier of Style

Syntactic 'complexity' is a matter of 'depth' in the hierarchical syntactic structure of a sentence. In a sentence whose syntaxemes all belong to the nucleus, the 'complexity index' is said to be zero. In sentences that are 'expanded' by the addition of constituents subordinated to the nucleus as a whole, these subordinated elements constitute Ievell 'expansions'. In an increasing hierarchical 'depth', part of a secondary expansion may, in turn, have a further 'expansion' appended to it, and so forth. Since an expansion subordinated to the nucleus is labelled level 1, the level of 'secondary expansions' (subordinative expansions of the second degree: SE 2) is given the index 2. 'Third level' elements of the sentence, which are subordinative expansions of the third degree (SE 3), are given the index 3, and so on with each increasing level of 'depth' in syntactic structure. Sentences are given a complexity index according to the 'deepest' level of subordinative expansion found in them, thus a sentence containing one or more subordinative expansions of, say, the fourth degree (SE 4), would have a complexity index of 4. It may be argued that a complexity index of(!) is only appropriate to sentences such as 'Run!', 'Here!', 'Mary!', etc. while sentences like 'I ran.', 'I shot Mary.', which contain already one level of subordination ('I' is subordinated to the predicate 'ran'; both 'I' and 'Mary' are subordinated to the predicate 'shot'), should have a complexity index of 1. In that case, 'secondary expansions'- that is to say, elements that are expansions of the nucleus of a sentence - would receive a numerical index of 2, rather than 1. Be that as it may, Bureau counts the level of other nuclear syntactic units clustering round a predicate as level (!). Appendages to the nucleus as a whole are level 1 expansions. Subsequent appendages of (!) level constituents, appendages to parts of these appendages, and so forth, consti-

STYLISTIC SEMIOTICS

223

tute 'deeper' levels of subordinative expansion that are numbered from 2 ton. According to Bureau's procedures, the sentence 'This is the house.' (containing no subordinative expansions) is a sentence with a complexity index of 0. (This may seem somewhat odd, since the sentence is obviously complex to some degree, unlike, say the sentence 'Run!' which is not.) In 'This is the house that Jack built.', a subordinative expansion has been appended to the nuclear element 'the house'. This expansion is, in Bureau's terms, an expansion of the first degree of 'depth' (SE 1). Accordingly, the sentence has a complexity index of 1. In 'This is the malt that lay in the house that Jack built.', the unit 'that lay' is a first degree subordinative expansion of the unit 'the malt'. In turn, however, this subordinative expansion (SE 1) is further expanded by the unit 'in the house' (SE 2), to which 'that Jack built' is attached as a further, 'deeper' level of subordinative expansion (SE 3). The sentence, containing 3 subsequent levels of subordinative expansion, has a complexity index of 3.

IThisi[G]I the maiden I - /all (SE2)/-- forlorn (SEl) - that milked (SEl)

-I the cow I(SE2)

I

-with the /crumpled (SE4)/---.I horn (SE3) - that tossed (SE4)

-I the dog I(SES) - that worried (SE6)

I

-I the cat (SE7) - that chased (SE8)

-I the rat I(SE9) - that ate (SElO)

-hhe maid (SEll) - that lay (SE12)

I

- in I the house (SE13) -that Jack built (SE14).

Figure 8.1

224

SEMIOTIC PERSPECTIVES

The complexity index of 'This is the maiden all forlorn that milked the cow with the crumpled horn that tossed the dog that worried the cat that chased the rat that ate the malt that lay in the house that Jack built.' is measurable in terms of the number of successive 'depths' of subordinative expansion that characterise the hierarchical structure of this sentence. Since the deepest level of 'embedded' subordinative expansion is of the fourteenth degree (SE 14), the complexity index of the whole sentence is 14. (We should note, incidentally, that there is a closer connection between the intellectual effort required for the comprehension of a sentence and the complexity of that sentence, than there is between comprehensibility and sentence length. For instance, the more complex, though shorter, sentence 'He is my father's brother's wife's son.' is conspicuously more opaque than the longer, but less complex, sentence 'The fat old man must have drunk a few large gins on the way home.') The nursery rhyme to which belongs the sentence represented in Figure 8.1 provides a crude example of a strategic-stylistic play on sentence complexity. The stylistic character of the nursery rhyme - in a sense its whole point is a piece of 'fun' - derives from the game it makes of its pattern of cumulatively increasing sentence complexity: Sentence Sentence Sentence Sentence etc. 8.1.5

1: 2: 3: 4:

Complexity Complexity Complexity Complexity

index index index index

= 1 = 3

= 5

= 7

Structure as a Potential Carrier of Style

There are two properties of sentences that, together, add up to what Bureau refers to as syntactic 'structure': (a) the syntactic functions (relationships) responsible for the particular cohesion and organisation of the constituent units in a given sentence; (b) the linear succession in which these units are disposed. The most immediately striking contribution of syntactic 'structure' to the level of 'strategic' stylistic functions is, of course, in terms of variety. Common sense alone tells us that substantially equivalent linguistic messages can be couched in sentences of varying length and complexity, sentences that are different in internal organisation and/or the linear disposition of their elements (compare, for instance, 'Death is our end.' with 'In the end, we die.' or with 'Our end is death.'). Structural variety (the spice of style) in a text is quantitatively measurable as a function of the number of sentences over the number of syntactic classes (or types) to which these sentences can be assigned. That is to say, the greater the number of sentence types, the greater the variety - the syntactic richness- of the text. More importantly still, 'external' symmetry between sentences, and 'internal' symmetry within sentences- both matters of the manifestation

STYLISTIC SEMIOTICS 225

of equivalent syntactic structures that 'mirror' each other in the text impose a pattern of recurrences (and, of course, contrasts) on a given text. Our rather crude example of the nursery rhyme 'The house that Jack built' can again be used to illustrate such a patterning. Each sentence of the rhyme belongs to the same type ('This is the ... '),from which comes the particular symmetry between sentences that is so characteristic of the text as a whole. There is, however, a more detailed structural patterning that emerges when the nursery rhyme is tabulated as in Figure 8.2. The (admittedly rather childish) playfulness of the pattern of symmetries and contrasts in the example should not mislead us into denying that we are dealing with a basically serious stylistic device. In so far as Sentence 1 of the nursery rhyme contains the sign 'house', this sentence can be seen as the 'foundation stone' on which the whole nursery rhyme is built by successive extensions. Thus, stylistically speaking, 'house' becomes a privileged sign in the text. It may not even be too fanciful to suggest that the 'construction' of the nursery rhyme is in correlation with the theme of 'building a house'; in this way, we might view the whole nursery rhyme as 'miming' the process of building a house, by means of an analogy that shows up in the textual pattern represented in Figure 8.2. There is certainly some basis to the suggestion that this throws an interesting light on the playful stylistic strategy of the nursery rhyme. Within sentences of this nursery rhyme, we find cases of total symmetry (designated by Bureau as isomorphism). For instance, in the sentence tabulated in Figure 8.1, there is isomorphism between several successive subordinative expansions: that tossed the dog that worried the cat that chased the rat that ate the malt which are exact syntactic equivalents of one another. Not only does this isomorphism constitute an obvious part of the stylistic 'game' in which the nursery rhyme seems to indulge, but it may not be too fanciful to correlate the isomorphism of these sentence units with the uniformity of building materials in the construction of a house. If such a correlation is not too far-fetched, then we have here another stylistic instance of the way the nursery rhyme plays on an analogy with the theme of 'building a house'. The nursery rhyme example, albeit crude when compared with the serious literary examples analysed by Bureau, has the advantage of illustrating the stylistic observations and generalisations that may be suggested by analysing the syntactic properties of a text. It also has the advantage of showing up the weaker side of this type of stylistic exercise, which is that, while patterns such as those in Figures 8.1 and 8.2 can, indeed, be objectively established, the placing of these patterns in correlation with thematic features or privileged signs of the text remains a matter of speculation, subject only to vague considerations of 'plausi-

I

-

that ate

-... I

-

SE 2--.........

--.

--........_SE 1

I that at;1

3

--.

'
is that they are about QUANTITIES of inventors, defeats or toes, rather than being about the inventors, defeats or toes themselves. Similar considerations hold for examples where conjoined atomic noun phrases impose singular rather than plural agreement, such as the well-known example (30): (30)

Ham and eggs is my favorite breakfast.

It is quite clear that singular agreement occurs here because ham and eggs is taken as referring to a particular dish, and not to a group in the normal sense. Likewise the differing agreement in examples (31)a. and b., from

A Semantics for Groups and Events

134

van Eijck (1981), corresponds directly to the fact that his aged servant and the subsequent editor of his papers refers to a single individual in (31)a. but a group in (31)b. (31)a.

His aged servant and the subsequent editor of his papers was with him at his deathbed.

b.

His aged servant and the subsequent editor of his papers were with him at his deathbed.

A final argument against semantic agreement comes from Krifka (1987). The numbers zero and one point zero impose plural morphology, but do not seem to imply semantic plurality: (28)a.

b.

Zero tables have been built since we arrived. One point zero liters of water flow through this valve per minute.

A natural treatment of plural zero is available in the agreement system developed above: let [zero] (X)( e) map a plural VP denotation f onto 1 if X is a subset of the exclusion set off at e. Other plural VP denotations should be mapped onto 0; the function should be undefmed for singular VP arguments.

One point zero is more problematic, however. A similar treatment to the case of zero or fewer than two is technically possible, but seems intuitively wrong; one point zero does not "feel" at all like an exclusive quantifier and should not have access to the exclusion sets of its VP arguments. I have no suggestions at this point as to how a semantic treatment of one point zero should work. Any such treatment ought to part of a more general study of the semantics of fractional parts of objects, however, and perhaps such a study will uncover some parallel with other plural determiners.

References Austin, John L. (1950) 'Truth'. Proceedings of the Aristotelian Society, Supplementary Volume xxiv. Reprinted in Philosophical Papers, John L. Austin. Clarendon Press, Oxford (1961). Bartsch, Renate (1972) 'The Proper Treatment of Number and Numbers in Montague Grammar'. In: Papers in Montague Grammar. UCLA Occasional Papers in Linguistics 2, R. Rodman (ed.) Bartsch, Renate (1973) 'The Semantics and Syntax of Number and Numbers'. In: Syntax and Semantics 2, J.P. Kimball (ed.) Seminar Press, New York. Barwise, Jon and John Perry (1983) Situations and Attitudes. Cambridge.

MIT Press,

Bennett, Michael (1972) 'Accomodating the Plural in Montague's Fragment of English'. In: Papers in Montague Grammar. UCLA Occasional Papers in Linguistics 2, R. Rodman (ed.) Bennett, Michael (1975) Some Extensions of a Montague Fragment of English. Indiana University Linguistics Club, Bloomington. (Corrected version of 1974 University of California at Los Angeles dissertation.) Benthem, Johan van (1986) Essays in Logical Semantics. Linguistics and Philosophy 29. D. Reidel, Boston.

Studies in

Blau, Ulrich (1981) 'Collective Objects'. Theoretical Linguistics 8.101-130. Carlson, Gregory N. (1977) Reference to Kinds in English. University of Massachusetts dissertation. Distributed (1977) by the Indiana University Linguistics Club, Bloomington; published (1980) by Garland Press, New York. Chwistek, Leon (1924-1925) 'The Theory of Constructive Types'. Annales de Ia Societe Polonaise de Mathematique 2.9-48,3.92-141. Cooper, Robin (1983) Quantification and Syntactic Theory. Language Library 21. D. Reidel, Boston.

135

Synthese

136

A Semantics for Groups and Events

Cooper, Robin and Terence Parsons (1976) 'Montague Grammar, Generative Semantics, and Interpretive Semantics'. In: Montague Grammar, B. Partee (ed.) Academic Press, New York. Dowty, David (1981) 'Quantification and the Lexicon: A Reply to Fodor and Fodor'. In: The Scope of Lexical Rules, M. Moortgat, H. van der Hulst and T. Hoekstra (eds.) Foris, Dordrecht. Dowty, David (1986) 'Collective Predicates, Distributive Predicates, and All'. In: ESCOL86. Proceedings of the Third Eastern States Conference on Linguistics, F. Marshall (ed.) Ohio State University, Columbus. Dowty, David, and Belinda Brodie (1984) 'A Semantic Analysis of "Floated" Quantifiers in a Transformationless Grammar'. In: Proceedings of the West Coast Conference on Formal Linguistics 3, M. Cobler, S. MacKaye and M.T. Wescoat (eds.) Stanford Linguistics Association, Stanford. Eijck, Jan van (1983) 'Discourse Representation Theory and Plurality'. In: Studies in Mode/theoretic Semantics. Groningen-Amsterdam Studies in Semantics I, A. ter Meulen (ed.) Foris, Dordrecht. Francis, W. Nelson (1986) 'Proximity Concord in English'. English Linguistics 19.2.309-318.

Journal of

Gillon, Brendan S. (1987) 'The Readings of Plural Noun Phrases in English'. Linguistics and Philosophy 10.2.199-219. Gillon, Brendan S. (1990) 'Plural Noun Phrases and Their Readings: A Reply to Lasersohn'. Lingusitics and Philosophy 13.4.477-485. Hamish, Robert (1976) 'Logical Form and Implicature'. In: An Integrated Theory of Linguistic Ability, T. Bever, J. Katz and T. Langendoen (eds.) Thomas Crowell, New York. Hausser, Roland (1974a) Quantification in an Extended Grammar. University of Texas at Austin dissertation.

Montague

Hausser, Roland (1974b) 'Syntax and Semantics of Plural'. In: Proceedings of the lOth Regional Meeting, M. LaGaly, R. Fox and A. Bruck (eds.) Chicago Linguistic Society, Chicago. Heim, Irene (1982) The Semantics ofDefinite and Indefinite Noun Phrases. University of Massachusetts at Amherst dissertation.

References

137

Hoeksema, Jack (1983) 'Plurality and Conjunction'. In: Studies in Modeltheoretic Semantics. Groningen-Amsterdam Studies in Semantics 1, A. ter Meulen (ed.) Foris, Dordrecht. Hoeksema, Jack (1989) 'The Semantics of Non-boolean "And"'. Journal of Semantics 6.1.19-40. Juul, A. (1975) On Concord of Number in Modern English. Copenhagen.

Nova,

Kamp, Hans (1981) 'A Theory of Truth and Semantic Representation'. In: Fonnal Methods in the Study of Language, J. Groenendijk, et al. (eds.) Mathematical Centre, Amsterdam. Reprinted (1984) in Truth, Interpretation and /nfonnation. Groningen-Amsterdam Studies in Semantics 2, J. Groenendijk et al. (eds.) Foris, Dordrecht. Katz, Jerrold (1977) Propositional Structure Thomas Crowell, New York.

and Illocutionary Force.

Keenan, Edward L., and Leonard M. Faltz (1985) Boolean Semantics for Natural Language. Synthese Language Library 23. D. Reidel, Boston. Krifka, Manfred (1987) 'Nominal Reference and Temporal Constitution: Towards a Semantics of Quantity'. FNS-Bericht 17. Forschungsstelle fiir natiir-sprachliche Systeme, Tiibingen.

Ladusaw, William (1982) 'Semantic Constraints on the English Partitive Construction'. In: Proceedings of the West Coast Conference on Fonnal Linguistics 1, D. Flickinger, M. Macken and N. Wiegand (eds.) Stanford Linguistics Association, Stanford. Landman, Fred (1986) 'Pegs and Alecs'. In: Towards a Theory of Infonnation: The Status of Partial Objects in Semantics. GroningenAmsterdam Studies in Semantics 6, Fred Landman. Foris, Dordrecht. Landman, Fred (1989) 'Groups'. Linguistics and Philosophy 12.5.559-606, 12.6.723-744. Lasersohn, Peter (1986) 'The Semantics of Appositive and PseudoAppositive NP's'. In: ESCOL86. Proceedings of the Third Eastern States Conference on Linguistics, F. Marshall (ed.) Ohio State University, Columbus. Lasersohn, Peter (1987) 'Collective Nouns and Distributive Determiners'.

In: Papers from the 23rd Regional Meeting of the Chicago Linguistic

Society, B. Need, et al. (eds.) Chicago Linguistic Society, Chicago.

138

A Semantics for Groups and Events

Lasersohn, Peter (1989) 'On the Readings of Plural Noun Phrases'. Linguistic Inquiry 20.1.130-134. Leonard, Henry S. and Nelson Goodman (1940) 'The Calculus of Individuals and Its Uses'. Journal of Symbolic Logic 5.2.45-55. Link, Godehard (1983) 'The Logical Analysis of Plurals and Mass Terms: A Lattice-theoretical Approach'. In: Meaning, Use and Interpretation of Language, R. Bauerle, et al. (eds.) De Gruyter, Berlin. Link, Godehard (1984a) 'Hydras. On the Logic of Relative Clauses with Multiple Heads'. In: Varieties of Fonnal Semantics. GroningenAmsterdam Studies in Semantics 3, F. Landman and F. Veltman (eds.) Foris, Dordrecht. Link, Godehard (1984b) 'Plural'. To appear in: Handbook of Semantics, D. Wunderlich and A. von Stechow (eds.) Link, Godehard (1987a) 'Generalized Quantifiers and Plurals'. In: Generalized Quantifiers: Linguistic and Logical Approaches. Studies in Linguistics and Philosophy 31, P. Giirdenfors (ed.) D. Reidel, Boston. Link, Godehard (1987b) 'Algebraic Semantics for Event Structures'. University of Munich manuscript. Umning, Jan Tore (1987) 'Collective Readings of Definite and Indefinite Noun Phrases'. In: Generalized Quantifiers: Linguistic and Logical Approaches. Studies in Linguistics and Philosophy 31, P. Giirdenfors (ed.) D. Reidel, Boston. Massey, Gerald (1976) 'Tom, Dick, Harry and All the King's Men'. American Philosophical Quarterly 13.2.89-107. McConnell-Ginet, Sally (1982) 'Adverbs and Logical Form: A Linguistically Realistic Theory'. Language 58.144-184. Morgan, J.L. (1972) 'Verb Agreement as a Rule of English'. In: Papers from the Eighth Regional Meeting of the Chicago Linguistic Society. Chicago Linguistic Society, Chicago. Partee, Barabara and Mats Rooth (1983) 'Generalized Conjunction and Type Ambiguity'. In: Meaning, Use and Interpretation of Language, R. Bauerle, et al. (eds.) De Gruyter, Berlin.

139

References

Perlmutter, David (1970) 'On the Article in English'. In: Progress in Linguistics, M. Bierwisch and K. Heidolph (eds.) Mouton, the Hague. Postal, Paul M. (1966) 'On So-Called "Pronouns" in English'. In: 19th Monograph on Languages and Linguistics, F. Dineen (ed.) Georgetown University Press, Washington. Reprinted (1969) in Modern Studies in English, D. Reibel and S. Schane (eds.) Prentice-Hall, Englewood Cliffs. Quine, Willard V.O. (1969) Set Theory and Its Logic (revised edition). Harvard University Press, Cambridge. Reid, Wallis (1984) 'Verb Agreement as a Case of Semantic Redundancy'. In: ESCOL84. Proceedings of the First Eastern States Conference on Linguistics. G. Alvarez et al. (eds.) Ohio State University, Columbus.

Anaphora, Roberts, Craige (1987a) Modal Subordination, Distributivity. University of Massachusetts at Amherst dissertation.

and

Roberts, Craige (1987b) 'Distributivity'. Center for the Study of Language and Information manuscript. Root, Rebecca (1986) 'Quantifiers and Plural Anaphora'. Texas at Austin manuscript.

University of

Scha, Remko J.H. (1981) 'Distributive, Collective and Cumulative In: Formal Methods in the Study of Language, J. Quantification'. Groenendijk, et al. (eds.) Mathematical Centre, Amsterdam. Reprinted (1984) in Truth, Interpretation and Information. Groningen-Amsterdam Studies in Semantics 2, J. Groenendijk et al. (eds.) Foris, Dordrecht. Schwartzschild, Roger (1989) 'A Call for Simplicity in the Interpretation of Plurals'. Presented at the 1989 annual meeting of the Linguistic Society of America. Stockwell, Robert P., Paul Schachter and Barbara Partee (1973) The Major Syntactic Structures of English. Holt, Rinehart and Winston, New York. Wald, Jan David (1977) Stuff and Words: A Semantic and Linguistic Analysis of Non-Singular Reference. Brandeis University dissertation. Williams, Edwin (1986) 'A Reassignment Linguistic Inquiry 17.2.265-299.

of the Functions of LF'.

Williams, Edwin (1988) 'Is LF Distinct from S-Structure? A Reply to May'. Linguistic Inquiry 19.1.135-146.

140

A Semantics for Groups and Events

Willis, Penny (1986) 'Is the Welsh Verbal Noun a Verb or a Noun?' Unpublished manuscript. Zwicky, Arnold and Jerry Sadock (1975) 'Ambiguity Tests and How to Fail Them'. In: Syntax and Semantics 4, J. Kimball (ed.) Academic Press, New York.

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 9

A SEMANTICS FOR THE ENGLISH EXISTENTIAL CONSTRUCTION

A SEMANTICS FOR THE ENGLISH EXISTENTIAL CONSTRUCTION

LOUISE MCNALLY

First published in 1997 by Garland Publishing, Inc. This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1997 Louise McNally All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69085-1 978-1-315-53649-1

(Set) (Set) (ebk) (Volume 9) (hbk) (Volume 9) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

A SEMANTICS FOR THE ENGLISH EXISTENTIAL CONSTRUCTION

LOUISE McNALLY

GARLAND PUBLISHING, INc. NEw YoRK & LoNDON I 1997

Copyright© 1997 Louise McNally All rights reserved

Library of Congress Cataloging-in-Publication Data McNally, Louise. A semantics for the English existential construction I Louise McNally. p. em.- (Outstanding dissertations in linguistics) Revision of thesis (Ph.D.)-University of California, Santa Cruz, 1992. Includes bibliographical references and index. ISBN 0-8153-2557-6 (alk. paper) 1. English language-Semantics. 2. English languageSyntax I. Title. II. Series. PE1585.M38 1997 96-46189 420'.143-dc20

Printed on acid-free, 250-year-life paper Manufactured in the United States of America

For Josep M.

Contents 1

2

The Problem 1.1 Introduction . . . . . . . . . . . 1.2 Facts to be Accounted For . . . 1.2.1 The Definiteness Effect 1.2.2 The Predicate Restriction 1.3 Previous Analyses . . . . . . . . 1. 3.1 There- Insertion and Its Descendants 1.3.2 Previous Characterizations of the DE 1.3.3 The Predicate Restriction 1.4 Outline of Remaining Chapters Notes . . . . . . . . . Existential Syntax 2.1 Introduction . . . 2.2 Identifying VP-Adjuncts 2.2.1 Description . . . 2.2.2 A Proposal for Depictive Adjuncts 2.3 Why a DP-External XP Must Be Posited 2.3.1 Arguments for Independence 2.3.2 Williams' Counterarguments .. . 2.4 Supporting the Adjunct Analysis . . . . . 2.4.1 Stowell's Small Clause Diagnostics 2.4.2 Extraction . 205 Chapter Summary Notes .. 0 0 ..• 0 •• Vll

3 3 7 8 12 14 14 24

36 38 40 45 45 46 46 49 50

51 55 57 58

66 72

74

Contents

3 The Existential, Descriptions, and Instantiation 3.1 Introduction . . . . . . . . . . 3.2 Data to Be Accounted For . . 3.2.1 Quantificational DPs . 3.2.2 Scope . . . . . . 3.2.3 Contact Clauses 3.2.4 Relative Clauses 3.3 Property Theory . . . . 3.3.1 The Syntax of PT 3.3.2 The Interpretation of PT 3.4 Interpreting the Existential . . . 3.4.1 A Dynamic PT Fragment of English 3.4.2 Interpreting Existential Sentences 3.5 Definites and the Other Half of the DE. . . 3.5.1 Augmenting the Fragment . . . . . . 3.5.2 A Felicity Condition on Existentials 3.5.3 Summary . . . . . . . . . . . . . . . 3.6 Advantages of the Analysis . . . . . . . . . 3.6.1 Acceptability of Quantified Kind DPs 3.6.2 Quantification and Scope 3.6.3 Contact Clauses 3.6.4 Relativization . 3.6.5 Summary Notes ..

77 77 82 82 82 84 85 86 87 89 91 91 103 116 116 120 125 125 130 133 135 136 143 144

4 Adjunct Predicates and the Predicate Restriction 4.1 Introduction. . . . . . . . . . . . 4.2 Interpreting Depictive Adjuncts . . . 4.2.1 The Adjunct Rule . . . . . . 4.2.2 Individual/Stage Sensitivity . 4.3 Extending the Analysis . . . . . . . 4.3.1 Nominalized Functions as Controllers 4.3.2 The Predicate Restriction Revisited 4.4 Eventive Existentials . . . . . . . . . . . . . . 4.4.1 The Problem . . . . . . . . . . . . . . 4.4.2 Eventive Participles Are Not Adjuncts 4.4.3 Why These Existentials Are Not Passives 4.4.4 A Proposal 4.5 Chapter Summary . . . . . . . . . . . . . . . . .

151 151 152 154 157 167 167 170 171 171 173 176 177 185

Contents

Notes 5

......... .

Some Final Remarks 5.1 Introduction . . . . . . . . . 5.2 Definites in the Existential . 5.2.1 List Existentials .. 5.2.2 Focus and the Existential 5.3 Other Expectations . . . . . . . . 5.3.1 Other Expletive There Sentences 5.3.2 The Existential and Other Indexicals . Notes ..

. 187

191

191 193 193 197 199 199 . 201 . 203

References

205

Index

217

Preface This book is a slightly revised version of my 1992 University of California, Santa Cruz, dissertation, entitled An Interpretation for the Engli8h Ezistential Construction (the title of the present work has been modified for bibliographic reasons). In addition to editing and correcting the text, I have added an index and updated the references. However, there have been no substantive changes in the contents for two reasons. First, some revisions have taken the form of subsequent papers (McNally 1994, 1995a, and 1995b); and second, substantive changes would have implied writing a completely new book, something which did not seem appropriate for this series. However, I would like to mention briefly one issue that received very little attention in the original dissertation. The reader is referred to the works cited below for further details. That issue is the proper role of location or locativity in the analysis of existential sentences. There is a persistent intuition in the linguistics literature that existential sentences have something in common with sentences that ascribe location to an individual (see e.g. Lyons 1967, Kuno 1971, Clark 1978, Lakoff 1987, Freeze 1992, among others); on some of these analyses, a sentence like There were people on the dock has the same underlying syntax and argument structure as People were on the dock. While I share the basic intuition, at the time I wrote the dissertation I could not find convincing empirical evidence to support an analysis which treated the existential predicate as a relation between an individual and a location or which ascribed existential and locative sentences the same underlying structure. I still find such evidence lacking; 1 however, I now believe it is important to reconsider the locative nature of existential sentences because, when properly understood, it may Xl

Preface

shed light on two phenomena: 1) the acceptability in existential sentences of DPs containing monotonic decreasing and nonmonotonic determiners (such as at most three and ezactly three, respectively; see Barwise and Cooper 1981 for definitions); and 2) the so-called predicate restriction, that is, the exclusion in existential sentences on certain final predicative phrases (e.g. *There was a man talQ. The analysis developed in what follows does not adequately treat DPs containing monotonic decreasing and nonmonotonic determiners. The problem is that it is not possible to combine an analysis of such determiners as essentially adjectival with an analysis of the existential predicate as either existence asserting or instantiating, as I propose here. This combination results in incorrect truth conditions. For example, if the semantics of a sentence such as There were ezactly two books on the table is paraphrasable as "There exists a set X of individuals such that X is a set of books and the cardinality of X is exactly two and X are on the table," the sentence can be true even when there are more books on the table-it suffices to set apart a set of exactly two books out of a larger set on the table. Intuitively, though, in such a situation the sentence should be false. The same is true for sentences containing monotone decreasing determiners. In McNally 1995a I propose a means of relativizing the truth conditions on existential sentences to a locational parameter in order to resolve this problem. However, no independent motivation is given in that paper for the particular way in which location is integrated into the semantics of the construction. I argue in Chapter 1 against Milsark's claim that the predicate restriction should be related to a similar restriction in copular sentences (cp. ?? A man was talQ, that is, that the unacceptability of this and the existential There was a man tall is due to some characteristic of a man. However, I have since concluded that Milsark was correct, although a convincing treatment of the predicate restriction remains to be worked out (though see Ladusaw 1994 and McNally 1995b for related proposals). In the present work, I argue that the temporal characteristics of the existential construction and of the final predicative phrase lie at the source of the predicate restriction; in subsequent work (McNally 1994, 1995b), I have argued that their locative characteristics are more likely to

Preface

ziii

be crucial. I remain skeptical that the semantics of the English existential construction should be directly related to that oflocative sentences. Nonetheless, I hope these comments, while very brief, will stimulate further research into the semantics of location which, in turn, may deepen our understanding of the existential construction and its relation to locative and copular constructions. Barcelona September, 1996

Acknowledgments I am especially grateful to Bill Ladusaw, Donka Farkas, and Sandy Chung for their guidance and support throughout this project. I would also like to thank the following people for comments on and discussion of earlier versions of this material: Judith Aissen, Chris Barker, Betty Birner, Giulia Centineo, Cleo Condoravdi, Cathal Doherty, Josep M. Fontana, Caroline Heycock, Ed Keenan, Peter Lasersohn, Bob Levine, Jim McCloskey, Geoff Nunberg, Geoff Pullum, Ivan Sag, Peter Svenonius, Kari Swingle, Annie Zaenen, and Sandro Zucchi. Finally, thanks to Chris Barker and Caroline Heycock for help with the formatting, and to Josep M. Fontana for keeping little Marta entertained. The research contained in this work was supported by an NSF Graduate Fellowship and a UCSC Chancellor's Dissertation Year Fellowship.

XV

A Semantics for the English Existential Construction

Chapter 1

The Problelll 1.1

Introduction

This work addresses a simple question: What is the interpretation of English there-existential construction? By there-existential, I mean the familiar construction exemplified in ( 1): (1)

a. There are students who failed Syntax I enrolled in Syntax II. b. There is a painting by Goya on the classroom wall. c. There are many solutions to that problem.

This construction has attracted a lot of attention in the linguistics literature, perhaps the most attention being devoted to the definiteness restriction on the postcopular DP 2 (see below) and to the observed syntactic and semantic similarities between the existential and other constructions, including (in the case of English) copular, passive, progressive, locative, and possessive constructions. However, one of the principal goals of this work is to focus attention away from these two aspects of the construction, and to develop an interpretation for the construction that will specifically address other properties of the postcopular DP. Informally stated, the proposal to be developed is the following:

3

4 (2)

The English Ezistential Construction

The existential predicate in English is interpreted as a property of a description of an entity, specifically the property that the description is instantiated by some entity at some index. The addition of a (non-negative, nonmodalized) existential sentence to a context entails the introduction of a discourse referent into the domain of the context that corresponds to the instantiation of the description-argument. An additional felicity condition requires this referent to be novel.

The principal point of interest in (2) is the claim that the postcopular DP should be interpreted a description of an entity, by which I intend an object that corresponds most closely to Chierchia and Turner's notion of a nominalized function (Chierchia and Turner 1988; it is equivalent to Chierchia's 1984 notion of the entity correlate of a property). This claim is of interest for a couple of reasons. For one thing, it commits me to a domain of entities that includes (inter alia) both "ordinary" individuals (like you or Jane or Jane's bicycle) and abstract entities that constitute the descriptions of those individuals (hereafter, I will refer to these as nominalized functions or "nfs" ), and to the position that a DP can be interpreted as an entity of either sort. While this type of enriched domain of entities is not new to model-theoretic semantics (consider notably Carlson 1977a), comparatively little attention has been devoted to investigating the consequences of positing a subdomain of nominalized functions. In the present case, positing a nominalized function as the argument of the existential predicate will lead to reflection on how the notion of "(non)referentiality" should be understood. It has been observed that there is something "nonreferential" about the postcopular DP in the existential construction (see e.g. Fodor and Sag 1982, Safir 1987, Higginbotham 1987); however, this observation has never been seriously investigated. Indeed, at first blush it is a rather curious observation, since the ostensibly conflicting intuition that the existential construction is used to introduce a (persistent) discourse referent is equally strong. These quite different intuitions have each driven distinct lines of analysis of the definiteness restriction. The position I take here is that both are correct, and that the way to reconcile (though of course this is not to say unify) them is to make precise what lies behind

Chapter 1.

The Problem

5

this intuition concerning the nonreferentiality of the postcopular position. Specifically, I will argue that what lies behind this intuition is simply the fact that the postcopular DP in the existential is interpreted as a nominalized function, rather than as an entity of the "ordinary" sort. I will assign the same kind of interpretation to predicate nominals. The term "referential" can then be reserved for a very restricted class of DPs, viz. those that are interpreted with respect to the actual world and identify objects in the ordinary entity domain (which I will refer to, following Chierchia and Turner 1988, as the set U). 3 This sortally based classification of DPs as referential vs. nonreferential crosscuts two other classifications in which the notion of referentiality has been given a role in the past. On the one hand, it is distinct from the "referential/necessarily quantificational" classification, implicit in the Discourse Representation Theory (DRT, Kamp 1981; see Roberts 1987 for detailed discussion) and File Change Semantics (Heim 1982) literature: As Williams 1983 observed, and as we will see below, necessarily quantificational DPs can quantify over nominalized functions (members of the set I designate as NF) as easily as they can quantify over objects in U. On the other, (non)referentiality has been used to classify individuals in actual vs. nonactual worlds, "nonreferentiality" characterizing nonspecific indefinite DPs within the scope of modal operators and intensional predicates. Taking these three classifications together, one sees that, while the referential DPs are a homogeneous class-just those that are interpreted as u-sort entities with respect to the actual world-the nonreferential DPs form a heterogeneous class. Consequently, I will avoid using the term "nonreferential" for specific subclasses of nonreferential expressions, preferring instead expressions such as "nf-sort DP," "necessarily quantificational DP," or "DP interpreted with respect to some nonactual world." We thus have three dimensions along which DPs (or more precisely, their interpretations) can be classified: whether or not the DP is quantificational, whether it identifies/quantifies over ordinary individuals vs. nominalized functions, and whether it identifies an actual vs. non-actual entity. The claim that the postcopular DP in the existential construction is nonreferential in the sense of "necessarily quantificational"

6

The English Ezistential Construction

or "identifying a nonactual entity" is incompatible with the intuition that a persistent discourse referent is introduced with the addition of nonnegated, nonmodalized existentials in extensional contexts; in contrast, the claim that it is nonreferential because of a sortal condition placed by the existential predicate is not. Indeed, assuming (2), it should be unsurprising that the assertion of an existential sentence has the effect of introducing an additional discourse referent instantiating the argument of the existential predicate, since the presence of that discourse referent will support the truth of the existential assertion. What lacks a full explanation is why this discourse referent must in general be novel. 4 But perhaps this is as it should be, since there is a certain amount of crosslinguistic variation in the class of DPs licensed in the existential construction. 5 The proposal in (2) is also of interest because of its particular empirical predictions. By claiming that the argument of the existential predicate is a nominalized function, I am claiming that nothing in the argument structure of the existential predicate directly corresponds to the individual whose existence is ostensibly asserted, or to the predicate denoted by the optional final phrase. This, in turn, has a variety of consequences that will become apparent as we proceed. One is that the definiteness restriction cannot be treated as a unitary phenomenon: some of the facts will follow from the interpretation assigned the postcopular DP, while the rest will follow from the felicity condition mentioned in (2). The advantages of a nonunified account of the restriction will be discussed in Chapter 3. A second consequence of positing a nominalized function as the complement to the existential predicate is that certain differences (beyond the definiteness restriction) between the postcopular DP in the existential construction and DPs that appear in referential argument positions are clarified. Previous analyses have had little to say about these differences. It further predicts that whenever we find a linguistic phenomenon sensitive to the semantic sort of an argument as I have characterized it, the postcopular DP will contrast with DPs in referential positions. A few such cases will be discussed in Chapter 3; whether there are more, and what they might be, is a question to be explored in future research.

Chapter 1.

The Problem

7

Finally, I will argue in Chapter 2 that the optional predicative phrase (XP) that can appear in existential sentences is a semantic adjunct of the sort found in (3); and, in Chapter 4, that the restriction on the sorts of predicative phrases licensed in the construction will follow from the adjuncthood of the XP: ( 3)

Sally is drinking the tea hot.

Thus, on the view to be defended, the similarity between existential and (truth-conditionally similar) copular, passive, and progressive sentences in English is accidental rather than necessary-a position rather different from what is generally assumed for English in the syntax literature since Milsark 1974 (a notable exception being Williams 1984; see also Chung 1987 and Williams 1994). 6 The rest of this chapter will introduce the two central phenomena that previous analyses have attempted to account for, followed by a discussion of some of those analyses. Since the literature on the existential construction is relatively large and has been summarized in other works (e.g. Lumsden 1988), I will mainly be concerned with examining and evaluating the principal analytic strategies that have been adopted in the past.

1.2

Facts to be Accounted For

At the very least, a successful analysis of the existential construction should account for the following: • The fact that, under certain circumstances, DPs with determiners like the, every, both, most, as well as proper names and pronouns, are excluded from the postcopular position in the construction, as in *There is every friend of mine at this party. Following common practice, I will to refer to this phenomenon as the definiteness effect or definiteness restriction. • The prohibition against certain kinds of predicates serving as the interpretations of the optional sentence-final XP, viz. those Carlson 1977b characterized as kind- and individuallevel predicates (e.g. widespread, intelligent). Again following common practice, I will refer to this prohibition as the predicate restriction.

8

The English Ezistential Construction

However, there are other facts one might expect the analysis to explain, including certain constraints on relativization out of the DP position and the fact that, with one set of exceptions, the postcopular DP must take narrowest scope with respect to other quantifiers and operators in its clause. These issues have received less attention in the literature in comparison with the definiteness effect and predicate restriction; consequently, I will postpone discussion of them until Chapter 3. Finally, one might hope that an analysis of the existential construction might provide some insight into related phenomena, including the "list" interpretation which often arises when a definite or quantificational DP appears in the construction (Milsark 1974, Rando and Napoli 1978), as in (4), and the "presentational-there" construction (Milsark 1974, Aissen 1975), exemplified in (5): (4)

A: What can I eat for dinner? B: Well, there's that leftover macaroni and cheese, Kent's chicken concoction, or Mary's meatloaf.

(5)

a. There appeared a masked man from behind the counter. b. There ran into the room a reporter with news about the president's assassination.

The data in (4) and (5) will come up again in Chapter 5, once the analysis of basic existential sentences is fully developed.

1.2.1

The Definiteness Effect

The definiteness effect (DE) is undoubtedly the most salient and most discussed feature of the construction. Since summaries of the relevant data are not uncommon (Lumsden 1988 is a useful source) I will introduce these facts with a minimum of commentary.

DPs Excluded from the Construction For taxonomic reasons, it is useful to divide the DPs generally excluded from the construction into two categories. One category includes proper names, personal and demonstrative pronouns, DPs headed by definite possessives such as John's and a subset of DPs

Chapter 1. The Problem

9

headed by the definite determiners the, these, those, this, that, and all. Representative exam plea appear in (6): 7 (6)

a. #There was Margaret at the party. b. #There were them/those waiting outside. c. #There was Monica's sister available to help.

The other category comprises a subset of the DPs headed by the necessarily quantificational determiners every, each, both, and most; the non-negative exceptive determiners every/all ... but John; and "free choice" any: (7)

a. *There was every /each participant upset with the arrangements. b. *There were both/most ambassadors housed at that hotel. c. *There was anybody happy about the election results. d. *There was everyone but Michelle taking Spanish.

In addition, Boolean combinations involving one or more of these are excluded (Keenan 1987). 8 DPs Appearing in the Construction with No Restrictions All of the classes of DPs to be mentioned here have been identified previously in the literature; relevant references will be provided along the way. DPs Headed by Intersective Determiners The postcopular DP may be an existentially interpreted bare plural or mass term, or it may be headed by any of the following sorts of intersective determiners (see Barwise and Cooper 1981 and below for definitions): the indefinite article, all cardinal determiners, cardinal comparatives such as as many male as female, vague non-proportional determiners such as many, few, the negative indefinite, and negative exceptives (e.g. no ... but John); see Keenan 1987 for a comprehensive list:

10 (8)

The English Ezistential Construction a. There were flies swarming around the cow's head. b. There was a tarantula crawling across the rock. c. There were three pianos in the salon. d. There were more red than blue flags flying above the castle. e. There were many /few explanations for his behavior. f. There were no replies. g. There was no child but Martha playing outside.

These DPs may be interpreted as speaker specifical or nonspecific, as evidenced in particular by the appearance of the necessarily specific "indefinite this" in the construction (cf. Prince 1981a; see also Farkas 1996 on the notion of speaker specificity): (9)

a. There was this one dog that kept chasing the cars on our street. b. There was a woman you know quite well on the panel. c. There was a specific book on every professor's shelf.

In order for use of an indefinite this NP to be felicitous, the existence of a referent for the NP must be presumed by the speaker (though not necessarily by the hearer). In addition, for many speakers, partitives headed by determiners drawn from among this class are also licensed in the construction (though this is not universally agreed upon in the literature, I have heard many naturally occurring examples of this kind): 9 (10)

a. This time, there were none of the objections they had encountered on other occasions. b. There are two of us on the selection committee. c. There were many of the same people at both events.

I take partitives to be acceptable and assume they differ from non-partitives only in the felicity conditions on the familiarity or anaphoricity of the referents associated with the complement to of (see Ladusaw 1982 for details).

Chapter 1.

The Problem

11

Morphologically Definite DPs That Do Not Refer Uniquely The postcopular DP may be (and sometimes must be) headed by a definite determiner if the complement noun denotes a 2- or moreplace relation (cf. Woisetschlaeger 1983, Holm back 1984 for these data; see e.g. Barker 1991 for discussion of relational nouns): (11)

a. There was the smell of liquor on her breath. b. There was the air of a soldier about him. c. There was the lid to a jar on the counter. d. There was the mother of a student in the office.

However, when the internal argument of the noun in ( 11 )c and d is modified by a definite determiner, the sentence becomes anomalous: (12)

a. #There was the lid to the jar on the counter. b. #There was the mother of the student in the office.

Similarly, possessives with prenominal possessors that are independently licensed in the existential may also appear in the construction, although possessives with other kinds of prenominal possessors are impossible: (13)

a. There was some woman'sjacket left behind. b. There were many student's papers unclaimed. c. #There were the woman's sisters running down the street.

Note that we get the same effect with relational nouns as we see with the prenominal possessors. A final set of examples in this category includes the following (Milsark 1974): (14)

a. There were the same people at both parties. b. There was the usual crowd at the bar.

Definite and Quantificational DPs Describing Kinds Nominals that denote the name of a kind, or that are headed by a noun

1£

The English Ezistential Construction

like kind, sort, flavor, etc., may appear with any otherwise prohibited determiners, whether definite or quantificational (Lumsden 1988, Wilkinson 1988): (15)

a. There were those kinds of books at the library. b. There was every flavor of ice cream for sale. c. There were both wines available for tasting.

This concludes the basic inventory of the types of DPs licensed in the existential construction. 10

1.2.2

The Predicate Restriction

I now turn to the issue of the predicate restriction. As Milsark 1974 and Keenan 1987 have convincingly shown (see Chapter 2 for a recapitulation), in the existential construction a predicative phrase may follow the postcopular DP. That it must be a predicative phrase is evidenced by the prohibition in the existential construction on anything that cannot follow the verb be in a copular construction ( cf. Siegel 1976): (16)

a. *The soldier is veteran. b. *There is a soldier veteran. c. *The president is former. d. *There is a president former. e. *The tree is fallen. f. *There was a tree fallen.

However, the restriction is stronger than that. A number of XPs which are ostensibly predicative in virtue of the fact that they can be nominal postmodifiers are also excluded:

Chapter 1.

(17)

The Problem

13

a. A book belonging to me has been lost. b. *There were two books that Fred had bought belonging to me. c. A woman knowing the answer has arrived. d. *There was a woman who came late knowing the answer.

In addition, full DP predicate nominals (which cannot be a nominal postmodifier, but which of course can appear in the complement to be position), are absolutely excluded, although my judgment is that determinerless predicate nominals are slightly more acceptable: 11 (18)

a. *There was a woman a contestant on the game show. b. *There are many people I know students of linguistics. c. *There was a radical student the president of the club.

Carlson 1977a suggests that the predicate position is restricted to just those predicates which he labeled "stage-level;" "individuallevel and "kind-level" predicates are excluded. 12 Thus, APs, PPs, and VPs which satisfy tests for stage-levelhood (such as allowing non-generic, non-specific indefinite and bare plural subjects, as in (19)) can appear in the construction (see (21)), while those that fail the stage-level tests, as in (20) cannot appear in the construction (see (22)): (19)

a. A student is sick with the flu. b. A new student is enrolled in the course. c. Dogs were chasing cars down the street. d. A teacher is out of work.

14 (20)

The English Ezistential Construction a. #A student is able to answer that question. b. Students are able to answer that question. c. #A dinosaur is widespread. d. Dinosaurs are widespread. e. #Some tires are in short supply. f. Tires are in short supply.

(21)

a. There are three people who know you sick with the flu. b. There were students who object to that enrolled in the course. c. There was a dog that was missing a leg chasing cars down the street. d. There are many educated people we know out of work.

(22)

a. *There are three people who know you able to answer that question. b. *There is a kind of dinosaur widespread. c. *There are tires in short supply.

This concludes the introduction to the principal range of data that previous analyses have focused on. We may now consider what those analyses have had to say about these data.

1.3 1.3.1

Previous Analyses There-Insertion and Its Descendants

The approach to the existential construction which assimilates it to the copular construction goes back at least as far as the "thereinsertion" transformation of Burt 1971, and has persisted in various guises in Milsark 1974, Jenkins 1975, Stowell1978, Safir 1982,

Chapter 1.

The Problem

15

Keenan 1987, and Pollard and Sag 1994, and elsewhere. As noted above, this analysis has its roots in the observation that existential sentences appear to have copular, progressive, and passive counterparts: the truth-conditional equivalence of the pairs in e.g. (23) suggested a common underlying structure (or alternatively, the postulation of an identical semantic relation, viz. predication, between the DP and XP in each pair): {23)

a. A child was ill. b. There was a child ill. c. A woman was playing the guitar. d. There was a woman playing the guitar. e. A passenger was killed in the accident. f. There was a passenger killed in the accident.

Moreover, the link has been bolstered by the observation that many Indo-European languages show the same Subject-Verb/VerbSubject alternation, often (putatively) with the requirement that the postverbal subject be indefinite. 13 While the analogy of existential vs. non-existential counterparts in English to personal vs. impersonal counterparts in other languages is tempting, in this section I point out a couple of problems with attributing common syntactic structures to the pairs in (23), using a version of the standard Government-Binding (GB) analysis to illustrate. 14 In so doing, I do not mean to imply that an account of the putative definiteness effect in impersonal constructions in other languages cannot ultimately be related to that observed in the existential in English, though I will not undertake such a project here. My point is simply that such a relation could (and if it is pursued, should) be developed without committing oneself to deriving existential and copular/passive/progressive sentences from the same syntactic or semantic representation.

The Stowell/Safir Analysis The work of Stowell 1978 and Safir 1982, 1985, 1987 has been influential in promoting the "small clause'' analysis of existential and

16

The Engli11h Ezi11tential Con11truction

copular sentences in English. This analysis takes the D-structure complement of (predicational) be to be a so- called small clause, as in (24): 15

(24)

IP

1\ A I VP

DP I'

~VP

V

~e~ DP

VP

D6

a woman laughing

Be is hypothesized to be a raising verb, like seem. In order for the Extended Projection Principle to be satisfied- that is, for there to be a surface subject of the sentence-and, on some analyses, in order for the postcopular DP to receive Case, one of two things has to happen: either the postcopular DP has to raise to Spec(IP), or else an expletive has to be introduced at S-structure. In the former case, the result is a passive, progressive, or copular sentence, depending on the small clause predicate:

Chapter 1.

Tke Problem

(25)

17

IP

~I'

DP,

D/\ A

A woman I VP V

~e

VP

/\ DP VP

t',

D

laughing

In the latter, the result is an existential sentence:

(26)

IP

A I' D/\ There I DP

VP

~VP

V

~e~ DP

VP

D~

a woman laughing I will not dispute the plausibility of the small clause analysis of copular sentences here; however, extending this analysis to the existential construction raises three problems in particular: ( 1) it incorrectly predicts that other copular constructions will have existential-like counterparts; (2) it cannot account simultaneously for the definiteness effect and the predicate restriction; and (3) it fails to predict the extraction facts associated with the construction. I now consider the first two of these problems in turn; the third will be addressed in Chapter 2.

The English Ezistential Construction

18

The Existential and Non-be Copular Constructions It has been observed (e.g. by Burzio 1986) that the small clause analysis of the existential, at least as formulated within GB, predicts that sentences such as those in (27)d,f should be grammatical, on analogy with (27)b: 16

(27)

a. A woman is in the yard. b. There is a woman in the yard. c. A woman seemed tired. d. *There seemed a woman tired. e. A woman appeared tired. f. *There appeared a woman tired.

Under the assumption that predicative seem and appear have the same argument structure as copular be, they should have a D-Structure like that in (28): (28)

IP

1\I' 1\ I VP

DP

~AP

V

I

seem

/ DP

A

""" AP

DD

a woman

tired

Since the expletive can satisfy the Extended Projection Principle and transmit Case when the verb is be, it should be able to do so with similar verbs as well-thus, the ungrammaticality of (27)d,f remains unexplained. This asymmetry between seem and appear, on the one hand, and be, on the other, has lacked an elegant explanation. We might adopt the general proposal of Raposo and

19

Chapter 1. The Problem

Uriagereka 1990 for similar data in Portuguese: they argue that the difference between paracer, 'seem,' in Portuguese, which does not allow an expletive subject with a small clause complement (like English), and verbs that do have this configuration of arguments, is that the small clause complement to paracer is actually dominated by a projection of a functional head (e.g. AgrP), while the small clause complement to the expletive-licensing verbs lacks a functional head. Extending this idea to English, seem and appear would license the structure in (29), rather than that in (28) (which we continue to assume for be): (29)

IP

1\I' 1\ I VP

DP

~

AgrP V I~ "'seem / Agr' DP

D/\ D

a woman Agr AP

tired

Raposo and Uriagereka suggest that the presence of this AgrP renders it impossible for the subject of the small clause to get Case: They make AgrP a barrier both to inherent Case marking by the main verb and to expletive-argument chain formation (see next section); and the small clause predicate is not a Case assigner. Failure of the small clause subject to get Case would thus be the reason that expletive subject structures are not possible with seem when it takes a small clause complement. This explanation of the asymmetry between be and other copular predicates may be viable, but it is unclear why be should have a small clause complement headed by a lexical projection, while seem and appear do not. The thrust of the analysis (intended or not) is

20

The English Ezistential Construction

clearly to make the postverbal DP as much like an argument of be as possible, while making a clause the argument of seem. 17 One therefore wonders what is gained by maintaining the type of small clause analysis for existential be sketched here, especially if the semantic similarity between existentials and copular I passive I progressive sentences has an alternative explanation. 18 If we reject the small clause analysis, that is, if we take the position that the similarity between the existential and copular constructions is not a reflex of a common syntactic origin, then we no longer predict that (27)d,f should be grammatical, and we are free to pursue other accounts of the distribution of the expletive there in English.

The Small Clause Analysis and the Definiteness Effect A second problem with the small clause-type analysis is its inability to account without stipulation for both the definiteness effect and the predicate restriction; it can account for at most one of the two. The problem is the following: On the small clause analysis, the postverbal DP is not a complement to be, but rather a subconstituent of the complement to be. Consequently, any semantic or discourse-functional condition imposed directly on the DP on this analysis would violate commonly accepted locality constraints: predicates are generally not assumed to be allowed to place semantic restrictions on subconstituents of their arguments. Similarly, a restrictive theory of discourse should not allow for statements such as: "The subject of the complement to the existential predicate must be able to introduce a novel discourse referent," which make reference to a position nonlocal to the lexical item or construction triggering the condition. Hence, neither semantics nor discourse can be appealed to to explain the DE, given the small clause analysis. However, since the small clause complement to be is headed by the final predicative phrase, it is possible to impose some sort of condition to capture the predicate restriction. This leaves us with two options: a syntactic account of the definiteness effect, or an account on which the effect follows from some aspect of the semantics of the small clause predicate. The latter strategy was pursued by Milsark 1974, 1977, who argued that the definiteness effect was intimately connected to the predicate

Chapter 1.

21

The Problem

restriction; however, in Section 3.3 I argue that this strategy fails. The best-developed syntactic account of the DE is that advocated in Safir 1982, 1985, 1987, in which it is argued that the definiteness effect is a consequence of the postcopular DP forming a chain with the expletive, specifically, an unbalanced 8-chain, in order to satisfy the Case Filter. That is, Safir posits that the postcopular DP is not in a Case-marked position at D-Structure. The fact that the DP gets Case parasitically from the expletive accounts for the fact that it controls verb agreement: The DP's agreement features are inherited by the chain, and those features are in turn shared by the chain's Case assigner, lnfl-Agr: (30)

IP

A DA DP

I'

There

I VP

~VP

V

~e ~ DP

VP

6D

a woman laughing Safir observes that the definiteness effect would follow from this analysis if the restriction of expletive-argument chains to indefinite DPs were justified. However, any such justification, including Safir's own proposal, entails appeal to semantics-exactly what we are trying to avoid. Safir 1987 suggests the following account. He begins by adopting the principle in (31): (31)

Predicate Principle: A potential referring expression is a predicate or else free (i.e. subject to Condition C of the Binding Theory-L.McN.].

This principle is motivated on the grounds that predicate nominals might reasonably be excused from the Binding Theory on the basis

22

The English Ezistential Construction

of their nonreferentiality and (concomitantly) because they are 8role assigners. Now, if the expletive and postcopular DP form a chain headed by the expletive, then the expletive will necessarily bind the postcopular DP. If the DP is a nonpronominal, (31) will require that the it be interpreted as a predicate in order to avoid a Principle C violation. The intuition lying behind this proposal is essentially that developed in this work. It amounts to a concession that the interpretations of the DPs in existential vs. copular /passive/progressive sentences are quite different. Observe that the small clause predicate ceases to play any role at all in determining the interpretation of the DP as referential or not (in my terms, as an ordinary individual vs. a nominalized function). This is a very curious state of affairs if the DP is the complement to the small clause predicate and only to that predicate. Further sign that Safir's proposal is essentially semantic is apparent in this remark made in subsequent discussion ofthe definiteness effect and its connection to the "existence assertion" character of the construction (Safir 1987:93£.): It seems that the natural language interpretation of a bare predicate [i.e. a predicate apparently missing an argument in the syntax; here Safir is referring to a class of which he takes the postcopular DP to be a memberL.McN .] is as an existential "event" assertion about the denotation of the predicate. But then it is no surprise that with a few exceptions, the typical limitation on unbalanced 8-chains is that they require an existential interpretation .

. . . the existential [e.g. There is a child sleeping] involves two bare predicates, one of them formed by the unbalanced chain (a child), the other by the whole VP is a child sleeping. The first predicate should have the default interpretation, namely that there is an event involving a child, and the second predicate should be interpreted to mean that there is a "(child- )sleeping event." This passage reveals that Safir in no way appeals to the small clause in his attempt to explain the DE, and his justification for

Chapter 1.

The Problem

23

expletive-argument chains makes crucial use of semantics. Thus, a nonstipulative, purely syntactic account of the DE does not seem to be forthcoming. In sum, the small clause only serves to hamper an idea that I will argue in subsequent chapters to be quite interesting and successful. There is a final alternative one could pursue in order to maintain the small clause analysis, namely, to develop a notion of something like an "indefinite small clause" which would require an indefinite subject, perhaps for semantic reasons. Be might then require its complement, be it a DP (in bare-DP existentials) or a small clause, to be indefinite. The problem lies in motivating such a notion. Whether we take "indefinite" to mean "capable of introducing a new discourse referent" or "not uniquely referring," (to use two well-supported characterizations of the term indefinite), it is impossible to see how the definiteness of the subject could be relevant. The "new discourse referent" characterization is not of obvious help because, in the case of a clause, the relevant discourse referent for evaluating the putative (in)definiteness of that clause should correspond to the denotation of the clause itself, and not to the denotation of one of its subconstituents (e.g. its subject). The "non-unique reference" characterization of an indefinite small clause would be of even less use, because a clause does not have to have and indefinite subject in order to have non-unique reference. Consequently, I do not see much hope for the utility of any conception of "indefinite event" in an explanation of the DE. To summarize, I have argued that analyses in the tradition of the there-insertion transformation, which try to derive existential and copular constructions from a common source, 19 should be avoided because they do not offer hope for simultaneous explanation of the definiteness effect and predicate restriction. The small clause analysis has the further disadvantage of predicting that other copular constructions should have existential counterparts. Additional syntactic arguments against these analyses will be provided in the next chapter. Having examined a part of the previous literature on the construction as a whole, I now consider two sorts of proposals that have been advanced to account specifically for the definiteness effect.

24

1.3.2

The English Ezistential Construction

Previous Characterizations of the DE

There are two recurring themes in characterizations of the DPs licensed in existential sentences: one is the idea that some formal property of the interpretation of the postcopular DP in involved (cf. Milsark 1974, Higginbotham 1987, Safir 1987, and I take it to be implicit in the characterizations of Barwise and Cooper 1981 and Keenan 1987); the other, the idea that the DP must be associated with a new discourse referent (cf. Prince 1981b, Prince 1988, Lumsden 1988, Zucchi 1995 ). Both reflect correct observations. Indeed, previous accounts of the DE have failed to be fully successful only to the extent that they have not put these two observations together.

Theme 1: The DE and Formal Properties of DP For some historical perspective, I begin with a brief look at the final version of Milsark's 1974 interpretation rule for the construction: (32)

E Rule: there AUX (have-en) be Q NP X is interpreted: the class C denoted by NP has at least one member c such that P(c) is true, where P is a predicate and P is the reading of X and the set of such members c is of cardinality Q.

The "Q NP" in the rule are the subconstituents of the postcopular DP node, Q standing for the determiner, when there is one, and the NP standing for the nominal sister to the determiner. The insight of this rule lies in the phrase the class C denoted by NP has at least one member c. The obvious problem lies in calculating the contribution of the material denoted by Q to the interpretation of the construction, and its relevance to the DE. Unsurprisingly, then, the existential construction has served as a testing ground for the formal theories of determiner and DP interpretation proposed by Barwise and Cooper 1981 (hereafter, B&C) and Keenan and Stavi 1986 (K&S)/Keenan 1987. Their proposals concerning the DE are presented in turn, following by discussion of a version of the proposal in Milsark 1977.

The Triviality Explanation B&C begin with the assumption that all DPs denote generalized quantifiers, i.e. sets of sets. Among

Chapter 1.

The Problem

25

the various classifications of quantifiers they propose is the by now well-known 8trong/weak distinction, inspired by the strong/weak classification suggested in Milsark 1977. Strong DPs fall into two groups: positive strong and negative strong. Simplifying considerably, a DP is positive strong iff it is entailed in all models that the universal set (viz. the set of all individuals in the model) is an element of the generalized quantifier the DP denotes; negative strong DPs are those for which it is entailed in all models that the universal set is not an element of their denotation. The weak DPs are those that are neither positive nor negative strong. (See B&C for complete definitions and explication.) B&C, observing that existential sentences of the form "There i8 X" (where X is a DP) are often paraphrasable by sentences of the form "X ezi8t8," propose the semantics in (33) for the existential construction, the universal property being the suggested denotation for the existential predicate: (33)

[There i8 DP]M,g is true iff 1 E [DP]M,g, where 1 is the

"universal property", i.e. a set which consists of all of the individuals in the domain of the model.

They then propose that the existential construction sounds acceptable only when the postcopular DP is weak; on their view, sentences with strong DPs, such as (34), sound odd because, given the semantics they propose and the definition of strong DP, such sentences will be either tautologies or contradictions: (34)

*There is every riot on this block.

( 34) is a tautology because 1 (the universal set) is an element of the generalized quantifier every riot in all models. Were (34) negated, the result would be a contradiction. In other words, B&C's account of the DE is pragmatic: the offending sentences sound anomalous because they are uninformative. However, Keenan 1987, Milsark 1990, and others have observed that this account is problematic for two reasons. First, the assumption that the copula is followed by a single DP is demonstrably wrong (cf. Chapter 2). Barwise and Cooper's analysis thus predicts that if a strong DP appears in the construction with additional (DP-external) modification, the anomaly should disappear. However, this is not the case:

26

(35)

The English Ezistential Construction

*On this block, there was every riot that caused major damage.

(35) is anomalous despite the fact that it is not a tautology (assuming that the locative phrase makes a non-trivial contribution to the truth conditions of the sentence). Second, as Keenan 1987 points out, there are plenty of nonanomalous sentences that are clearly tautologies or contradictions: (36)

a. Every thing is not a thing. b. Red is red. c. There are either zero or more than zero people sitting in that chair.

On the B&C account, we have no explanation for why the putatively tautologous or contradictory existential sentences sound as bad as they do. Consequently, I follow Keenan and Milsark in concluding that lack of informativeness cannot be the explanation for the DE. Keenan's Definition of Existential DP In order to make the Keenan/Keenan and Stavi account of the DE clear, a little more formal background is necessary. Keenan and Stavi propose that all natural language determiners denote conservative functions. 2 For our purposes, conservativity (a property of functions relating sets) can be defined as in (37): 21

°

(37)

A function /is conservative iff f(A, B)

=f(A, An B)

The observation underlying conservativity is that to decide whether a sentence of the form "Det A's are B's" is true, it is necessary to look only at things which are A's (and, therefore, not at the B 's which are not A's). For example, to evaluate the truth of the sentence Every child swam we need not know anything about any individual who was not a child. Notice that it may be necessary to look at the A's which are not B's in order to decide whether a sentence is true. For example, in this case it is crucial to know whether there were any children who did not swim in order to know whether Every child 8Wam is true.

Chapter 1.

The Problem

27

Conservativity characterizes all determiners. 22 What more do we need to say about the determiners which appear in existential sentences? Keenan, p.c. proposes that the existential predicate licenses those DPs which are headed by determiners denoting what he defines as co-conservative functions: 23 (38)

A function /is co-conservative iff f(A, B)

= f(A n B, B)

Since all determiners are conservative, those which appear in the existential construction are the intersection of the conservative and the co-conservative ones; hence, he refers to them as intersective: 24

(39)

A function f is intersective iff f is both conservative and co-conservative. Consequently, f is intersective iff f(A, B) =:f(AnB,AnB).

The intuition behind (39) is that an intersective determiner is one for which, in order to decide the truth of "Det A's are B's," it is necessary to look only at the A's which are also B's. For example, to evaluate Two children swam, we need only look at the children who were swimming, and verify that there were at least two of them, to declare the sentence true. Any children who were not swimming, and any swimmers who were not children, are irrelevant. Some examples of intersective determiners include all of the cardinal determiners (one, two, several, many), no, the exceptive determiner no ... but John, and cardinal comparatives such as as many male as female. The non-intersective determiners include the, every, both, each and all proportional determiners (e.g. most, more than half, ten percent of). It is important to note that the class of intersective determiners and intersective DPs is defined recursively, as follows: ( 40)

a. A basic determiner is intersective iff it is always interpreted by an intersective function. b. A determiner is intersective (simpliciter) iff either it is a basic intersective determiner or it is built up from basic intersective determiners by Boolean combinations.

28

(41)

The English Ezistential Construction

a. A basic intersective DP is one formed from an intersective determiner and the appropriate number of

N's.

b. The intersective DPs are the basic intersective ones together with those formed from them by Boolean combinations. A recursive definition is necessary because it is possible to form logically equivalent complex determiners using Boolean operations that behave differently with respect to the existential construction: only those formed from just basic intersective determiners or their Boolean combinations will be licensed, as seen in the contrast in (42): (42)

a. There are either zero or else more than zero even prime numbers.

b. *There are either all or else not all even prime numbers. Since either zero or else more than zero and either all or else not all are both trivial, there is no difference between them other than that the former is composed of two intersective determiners, while the latter is composed of two non-intersective determiners. Consequently, it appears that it is not merely the denotation of the determiner, but the elements out of which it is composed, that is relevant. With these definitions in hand, we are ready to consider the explanation of the DE advanced by Keenan/K&S. Keenan's 1897 explanation is similar to B&C's in that he does not take the DE violation to be necessarily ungrammatical or anomalous. Instead, he simply asserts the following: (43)

Existential sentences of the form in (a) are logically equivalent to the existence assertion reading (b), iff the determiner expression Det is always interpreted by an intersective function. a. there [be [Det student]DP [in the garden]xP] b. 1 E [Det]([student] n [in the garden])

In other words, Keenan claims that when the postcopular DP is headed by an intersective determiner, the interpretation of the sen-

Chapter 1. The Problem tence is (presumably appropriately) logically equivalent to an assertion of the form Det &tudents that are in the garden ezi&t, which he translates using the universal property in ( 44)b. In contrast, when the DP is headed by a determiner which does not denote an intersective function, the existential sentence is not equivalent to an assertion involving ezi&t (compare (44 )c and d): (44)

a. There are two students in the garden. b. Two students that are in the garden exist. c. #There is the student in the garden. d. The student that is in the garden exists.

Keenan does not claim that DPs headed by non-intersective determiners will be systematically good or bad in the construction, but rather that when they are interpretable, they will not be logically equivalent to the existence assertion reading. As a result, his proposal is, like B&C's, immediately subject to the objection that it does not explain the anomaly of some existentials with nonintersective determiners, but not others. Moreover, while I have not found any counterexample to ( 43), it is not completely explanatory for the following reason. Consider examples such as the following: (45)

a. There is the lid to a jar on the counter. b. The lid to a jar which is on the counter exists. c. #There is the lid to the jar on the counter. d. There are both varieties of that plant in the garden. e. Both varieties of that plant in the garden exist.

My intuition is that while (45)a and b, and (45)d and e, are (as Keenan predicts) not logically equivalent, ( 45 )a and d are not strange in the way that existentials with non-intersective determiners are usually strange (compare e.g. (45 )c)-a contrast that needs to be explained. The difference clearly has to do with the internal semantics of the DP: In the case of the DP in (45)a, licensing is conditioned by properties of the complement to the noun; in (45 )d,

30

The English Ezistential Construction

it is not the complement of the noun but the noun itself that renders the DP acceptable in the existential. Consequently, Keenan's account of the DE is incomplete insofar as it focuses entirely on the semantics of the determiner. Nonetheless, as long as we confine ourselves to a restricted class of nominals, Keenan's generalization is compelling. What is it about intersective determiners that the existential construction is sensitive to? In order to answer that question, I will address another: Does the complement to the set of intersective determiners constitute a single natural class? I take the answer to this second question to be negative Inspection of this class reveals that it consists exactly of all of the determiners that give rise to necessarily quantificational DPs (e.g. every), with the exception of no 25 and the exceptives built from no (e.g. no ... but Suzanne); plus all of those determiners that give rise to DPs that must refer uniquely (e.g. the). This suggests that the existential is sensitive to whether or not its complement DP must refer uniquely, and indeed this conclusion is supported by the fact that DPs in the acceptable (45 )a and the unacceptable ( 45 )c differ only in that the former need not refer uniquely while the latter must. But is the existential sensitive to quantification? It would be very odd for quantification into an argument position to be idiosyncratically excluded, and indeed ( 45 )d shows that quantification into the postcopular position is not always excluded. The position I will take in Chapter 3 is that the sensitivity associated with necessarily quantificational DPs is actually to the sort of individual being quantified over. The Existential and Existential Quantification As a conclusion to the discussion of previous explanations the DE that have been driven by consideration of formal properties of DP, I will argue that accounts that try to derive the DE by positing a necessary binding relation between a special existential quantifier associated with the construction and the postcopular DP are undesirable. Milsark 1977, Williams 1984, Higginbotham 1987, and Safir 1987 all suggest something along these lines. For purposes of exposition, I will present Milsark's version of the proposal. Consider ( 46): ( 46)

There is a tree in my yard

Chapter 1.

The Problem

31

Following Milsark, we could introduce the existential operator via a stipulated part of the interpretation rule for the construction. In addition, assume that the indefinite is interpreted in situ as an open formula at L(ogical) F(orm):

(47)

s

~VP

DP

6~ is a tree(x) in my yard

There=3

This proposal would account of the definiteness effect if, as Milsark suggests, one could justify the stipulation that the operator must bind a variable in the denotation of the DP, since neither quantificational DPs such as every tree nor definites such as the tree will make any such variable available - the former because every will bind any bindable variables associated with the DP; the latter, because the felicity conditions on the use of definites entail that the value of the variable associated with the definite DP be antecedently determined. There are two reasons to disfavor this account. First, if Abbott 1992 is correct in arguing that so-called "list" existential sentences (and existentials containing definite DPs, more generally) should have the same semantics as ordinary existentials, then it is not clear how the former sort of existentials could be interpreted. Consider a sentence like There's Mary. If Mary provides no variable for the posited existential operator to bind, this sentence should be semantically anomalous. And yet it is not, as seen in the fact that there are many contexts in which it can not only be interpreted but also felicitously used. A second problem is raised by indefinite DPs that contain within them other quantifiers, such as a member of every committee. These DPs are licensed in the construction with an inverse linking reading (sketched in (48)b,d for (48)a,c, respectively; cf. May 1985), where the quantificational DP takes scope over the DP containing it: 26

3~

{48)

The English Ezistential Construction a. Contrary to expectation, there was a short blurb about every course in the Institute handbook.

b. ('Vz: course(z)) [(3y: blurb(y))[about(y, z) 1\ in(y, 1-h)]] c. There was a representative from every department appointed to the steering committee.

d. (Vz: department(z)) [(3y: rep(y))[from(y, z) 1\ appointed(y, s-c)]] It should be clear that a construction-specific existential quantifier cannot have wide scope with respect to the universal quantifier on the inverse linking reading. The problem is that it cannot have narrow scope with respect to that quantifier, either. May 1985 shows that a quantificational DP embedded within another DP cannot Quantifier Raise higher than the DP in which it is contained; rather, it must take immediate scope over that DP. Thus, there is no way for a construction-specific existential operator to intervene between the universal quantifier and the variable corresponding to the indefinite-the structural integrity of the DP would have to be violated in order for this to happen. But if this is the case, then an account of the DE built on the stipulation that the postcopular DP must be bound by a special existential quantifier cannot explain the grammaticality of sentences like those in (48), since there is no way for such a quantifier to bind the variable corresponding to the indefinite if an inverse linking interpretation is to result. The problems encountered by Milsark's account of the DE are clearly due to the way in which the indefinite is provided with existential force: Appeal to a necessary binding relation between an existential operator and a variable in the translation of the indefinite fails to account for the possibility of definites in the construction and the inverse-linking reading.

Summary of Theme 1 We have seen that previous accounts of the DE grounded in formal properties of DP fall short in various ways. The purpose of considering the Barwise and Cooper account was to make the point that the DE does not correlate with the triviality or contradictoriness of an existential sentence. I concluded

Chapter 1.

The Problem

33

on the basis of this that we must look for a linguistic, as opposed to extralinguistic, explanation for the DE. However, I showed that trying to make the DE follow by positing a special existential operator of the sort suggested by e.g. Milsark 1977, that targets the denotation of the postcopular DP, had undesirable consequences. In addition, we have considered Keenan's characterization of the determiners that appear in the existential construction. This characterization is insufficient insofar as the licensing of a DP in the construction depends not just on the determiner, but on the denotation of the DP as a whole. However, it has pointed us towards some properties of DP to which the construction might be sensitive, viz. necessarily unique reference and, when the DP is quantificational, the sort of individual being over. Theme 2: Discourse/Presupposition-Based Explanations

A different view of the DE and DE violations runs through the more functionally oriented literature (e.g. Prince 1988, Hannay 1985, Lumsden 1988, Ward and Birner 1995), and has been taken up recently in a modified form by Zucchi 1995. On this view, the condition on the postcopular DP is, roughly, that it introduce an individual whose existence is not presupposed in the context of utterance. Consider this paraphrase of the characterization of the condition according to Prince 1988: ( 49)

The post verbal DP in the English there-existential construction must identify a hearer new discourse referent.

Prince's notion of hearer new largely corresponds to Heim 's 1982 notion of novelty: A hearer new discourse referent (DR) is one which, at the time of utterance, is not an element of the domain of the discourse model (be it in virtue of explicit mention, inference, or shared knowledge), and whose descriptive content is not presupposed to be satisfied by any individual in that domain. DPs such as a cat identify hearer new DRs by convention and are, consequently, licensed in the existential construction. Definite DPs such as the cat or Fluff must identify a DR which is already part of the common ground (i.e. not hearer new); hence, they are excluded from the construction.

The English Ezistential Construction (49) is an extremely successful characterization of the condition on appearance in existentials; however, it has one drawback: it is not designed to handle cases where the DP is not associated with a persistent discourse referent at all. Consider the DPs no student and every student. Neither of these DPs introduces a persistent discourse referent (though we might say that both introduce temporary discourse referents, see Heim 1982:249ff.), and yet the former is licensed in the existential construction, while the latter is not: (50)

a. There is no assistant available. b. *There is every assistant available.

Consequently, (49) must be generalized to encompass quantificational DPs. One way to do this is to a notion of novelty that makes reference to descriptive content in addition to a notion of discourse referent. For example, following Condoravdi 1992b we could adopt the following: (51)

A DP is strongly novel (in a context) iff it bears a novel referential index and its descriptive content is not presupposed to be satisfied by any individual in the domain of the common ground of the context.

The referential index mentioned in (51) is used not only to index persistent referents in the discourse model, but also to keep track of antecedent-pronoun relations confined to the scope of quantifiers. The descriptive content is contributed by the complement to the head determiner in a DP. The condition on the postcopular DP could thus be restated as in (52): (52)

A DP may appear in the postcopular position of the thereexistential construction iff it is strongly novel.

Strong novelty distinguishes no student and every student as follows. Both DPs are assigned a novel referential index (because the index they receive simply serves the purpose of keeping track of the variable(s) bound by an operator; quantificational DPs are

Chapter 1.

The Problem

35

never anaphoric). However, the D Ps differ with respect to presupposition of descriptive content: no aui8tant carries no presupposition that there is any assistant-individual in the domain of the discourse model, while every a88i8tant does carry such a presupposition. 27 In other words, no 8tudent is strongly novel, while every 8tudent is not, and the prediction is correctly made that only the former will appear in the existential construction. However, although (52) covers a wider range of data than does (49), it raises the following question: 28 How does it extend to examples involving quantified kind expressions such as (53)? (53)

There were both kinds of apples in the sauce.

That is, what do we take the descriptive content of both kinds of apple8 to be? In order for the DP in (53) to count as strongly novel, its descriptive content must not be presupposed to be satisfied by any individual in the domain of discourse. But since (53) involves quantification over kinds, it must be presupposed that something is instantiating the descriptive content associated with the nominal kind8 of apple8. Consequently, if (52) is going to account for the acceptability of (53), a more sophisticated characterization of the notion of satisfaction of the descriptive content of a DP will have to be forthcoming. Perhaps we could avoid rethinking our assumptions about what constitutes the satisfaction of descriptive content by extending the notion of hearer newness explicitly to cover cases of quantification, building on an observation made by Lumsden 1988. Lumsden asserts that (53) is acceptable because it predicates existence of instances of kinds, and the existence of those instances is not presupposed, even if existence of the kinds is. Reformulating his hypothesis in terms of ( 49), and expanding it to account for cases of quantification, we could assert that: (54)

The postcopular DP in the English there-existential construction must introduce a (persistent or temporary) hearer new discourse referent.

But is (54), or any analysis that tries to account for the distribution of both definites and necessarily quantificational DPs in terms of the presuppositions associated with their descriptive contents the best solution? One of the goals of Chapter 3 is to argue

36

The Engli&k Ezi&tential Con&truction

that it is not. Rather, sentences like (53) point to the need to compare closely the behavior of (both quantified and nonquantified) kind terms and ordinary DPs in existential sentences. Consequently, a clarification of the different behavior of kind terms and ordinary DPs in the existential is one of the principal objectives of the present work. This clarification will both allow for an extremely simple statement of the novelty condition associated with the postcopular DP, and will also shed light on other properties of the postcopular DP not immediately related to the novelty condition. I have not discussed the relative merits of those characterizations of the DE that are based on formal properties of the determiner or DP vs. those that are discourse-functional. The reason that I have not is that I do not take what has been labeled "the" DE to be a unitary phenomenon. Rather, the two sorts of explanations target distinct properties of the postcopular DP: the former is sensitive to a semantic condition on the DP; the latter, to an independent felicity condition conventionally associated with the construction. I will provide arguments for this nonunified approach to the definiteness effect in Chapter 3.

1.3.3

The Predicate Restriction

In contrast to the definiteness effect, the predicate restriction has received scant attention. Three strategies have been pursued to explain it: (1) Claim that the predicative material is actually a part of the postcopular DP and try to derive the restriction from conditions on postnominal modification; (2) Assume an independent XP, and try to derive the restriction from the condition governing the postcopular DP; and (3) Assume an independent XP, and build the explanation on its semantic properties. We saw in Section 3.1.3 that pursuing at least one version of this last strategy left us with no way to account for the definiteness effect. Barwise and Cooper 1981 and Williams 1984 pursue the first strategy; that is, they treat what we have referred to as the XP as postnominal modifiers and attempt to derive the predicate restriction from independent conditions governing such modifiers. For example, they propose that an AP such as intelligent is ruled out in the existential construction because it cannot be a postnominal

Chapter 1.

The Problem

37

modifier: (55)

*Many professors intelligent are on their faculty.

However, since convincing evidence has been offered that there in fact can be an independent XP at the end of an existential sentence (see Milsark 1974, Keenan 1987, and Chapter 2), and since the predicate restriction demonstrably persists there, this sort of explanation cannot be the whole story. Milsark 1977 adopts the following version of the second strategy to explain the predicate restriction: Assume that the definiteness restriction is an independent fact about the construction. Demonstrate that DPs which appear in the existential construction are incompatible in copular sentences with the predicates excluded from existential construction. Then the predicate restriction will follow from the definiteness effect plus an independently motivated incompatibility between the licensed DPs and the excluded predicates. However, the facts do not support this explanation; specifically, the construction licenses a whole range of DPs which are not incompatible with the excluded predicates.

(56)

a. This friend of mine is very clever. b. A woman I know is a member of the City Council. c. No solution I know of is permanent. d. At least two people on the research team are doctors.

(57)

a. There was this friend of mine there, watching the fire. b. There was a woman I know on the City Council. c. There is no solution to the last homework problem. d. There are at least two people on the research team.

38 (58)

The English Ezistential Construction a. *There was this friend of mine very clever. b. *There was a woman I know a member of the City Council. c. *There was no solution I know of permanent. d. *There were at least two people on the research team doctors.

Consequently, we must find an explanation for the predicate restriction which does not depend on the nature of the postcopular DP, or else we have to come up with some non-superficial difference between the DPs in (56) vs. (58). An additional criticism which can be brought against Milsark's analysis is that it precludes any unified account of the predicate restriction in the existential and the very similar restrictions that apply to "reduced" clefts (Jenkins 1975), and to adjunct XPs (see e.g. Rapoport 1991 and Chapter 4): (59)

a. That's John sick. b. *That's John a doctor. c. *Here's Margaret tall.

(60)

a. Max played tennis barefoot. b. *Max played tennis intelligent. c. *Max played tennis an amateur.

It is obvious that the nature of the DP that provides the subject for the secondary predication has nothing to do with the restriction in (59)-(60). Any analysis of the predicate restriction in the existential construction that depends on the DE is not going to account for these facts and will thereby quite possibly miss a generalization.

1.4

Outline of Remaining Chapters

Chapter 2 presents syntactic motivation for the claim that the postcopular DP is the sole complement to the existential predicate, as

Chapter 1.

The Problem

39

well as for the claim that the optional final phrase is a predicative adjunct. In addition, I will argue that none of the usual tests for small clausehood offer any positive indication that the DP-XP string in the existential is a small clause. In Chapter 3 the interpretation for the basic existential construction (without the final XP) is developed and then compared to analyses that take the postcopular DP to denote an ordinary individual or a generalized quantifier over such individuals. Chapter 4 augments the analysis to account for the contribution of the final XP and shows how the predicate restriction can be derived from a more general condition on depictivefcircumstantial VP-adjuncts. Finally, Chapter 5 contains some rather speculative discussion of the broader implications of the proposal in the context of data such as "list" existential and "presentational- there" sentences.

40

The English Ezistential Construction

Notes 1 For

example, I believe that too much is made of the fact that the expletive in the existential is phonologically identical to the locative pronoun (see Lakoff 1987, for example). Historically, there has been variation in the choice of expletive used in existential sentences (see Ball1991), and in African American Vernacular English, expletive it is still used. 2 Following Abney 1987, I use "DP" rather than "NP" to refer to phrases such as students who failed Syntaz I. 3 0f course, DPs interpreted as nominalized functions can be used to refer insofar that they name nominalized functions; thus, "referential" as used here should be taken as a technical term, as characterized in the text. 4 The relevant sense of "novel" will be taken up in Section 3.3.2. 5 For example, the Catalan existential is less restrictive in allowing definites in non-list contexts); see Chapter 3. 6 0utside of the syntax literature, the inclination to link the existential construction to these other sorts of sentences is much weaker. See e.g. Barwise and Cooper 1981, Zucchi 1995 for analyses that explicitly preclude such a connection. Within the syntax literature, see Lyons 1967, Kuno 1971, Kimball1973, Lakoff 1987 for the rather different view that the existential construction forms a class with locative constructions, a view that is in a sense preserved in the present analysis. 7 Anticipating my analysis, I will use an asterisk (*) to mark sentences that I take to be ill-formed. Sentences marked with a cross-hatch ( #) are infelicitous in the relevant context. 8 Certain examples involving the definite article and every, in which the NP contains an infinitival or modalized relative clause modifier, seem to be exceptions to these generalizations. Examples appear in (i) (pointed out to me by Jim McCloskey), (ii) (from Pollard and Sag 1994), and (iii) (inspired by Moltmann 1995): (i) There weren't the doctors to staff the clinic. (ii) There was every reason to leave. (iii) There was everything you ever could have wanted. Although I do not have an account for these facts and will not discuss them in this work, note that the presence of such a relative clause does not improve DPs headed by determiners such as most, which is not universal, and each, which unlike every, always has a

Chapter 1. The Problem

41

distributive, and never a "maximizing," effect: (iv) *There were most reasons to leave. (v) *There was each reason to leave. 9 There seems to be some cross-linguistic variation with respect to the felicity of partitives in the existential. See e.g. de Hoop 1990, En~ 1991, Hoeksema 1989 for relevant discussion. 10 See Ward and Birner 1995 for a purely pragmatically-based classification of the quantificational DPs licensed in existential sentences which attempts to subsume them under the same generalization used to account for definite DPs. See Section 3.2.2 and Chapter 3, Section 6 for further discussion. 11 Cf. Nathan 1978 and Hannay 1985. These authors claim that the DPs are not universally prohibited by the predicate restriction, but rather that any state-descriptive DP will be licensed and cite examples such as There wa6 a woman the pre6ident in support of this claim. My own judgment is that this sentence is bad. 12 This amounts to a refined version of the claim Milsark 1974 made, viz. that those XPs excluded from the construction were those which denoted more "permanent" properties of individuals. 13 See e.g. Burzio 1986, Belletti 1988, Safir 1987, Perlmutter and Zaenen 1984, Legendre 1990. 14 1 will presuppose familiarity with GB throughout. See e.g. Chomsky 1986b for an introduction. 15 ! assume for the sake of exposition that the subject of the small clause is adjoined to the maximal projection associated with the predicate (see e.g. Chomsky 1986a for such a treatment); this assumption is not crucial. 16 Since this prediction depends partly on the assumed G B analysis of 6eem, it is not necessarily made by all analyses of the existential which relate it to copular/passivefprogressive sentences. For example, Milsark 1974 and Pollard and Sag 1994, do not make this prediction, because they stipulate the licensing of the expletive with be by syntactic and lexical rules, respectively. 17 Lasnik's (1992) analysis, on which be can exceptionally Case mark the subject of the small clause, while 6eem and appear are stipulated not to be able to, is similar in this respect. 18 An analysis for existential be on which the expletive was the subject, and the postcopular DP the predicate, of a small clause complement to be would be more in the spirit of the conception

The English Ezistential Construction

of copular and existential sentences defended in Chapters 2-4, although I will not pursue such an alternative here. See Moro 1991 for yet another small clause analysis, which takes the expletive to be the small clause predicate, and the postcopular DP, the subject. 19 1 have focused on approaches positing a common syntactic source, but the same point could be made for those positing a common lexical source (cf. Pollard and Sag 1994). If, like Pollard and Sag, we take both the DP and the XP to be independent semantic arguments of be, we regain the possibility of a semantic or discourse functional account of the definiteness effect; it could be stipulated as part of the semantics or conventional pragmatics of be. However, if we continue to maintain that the existential and copular constructions are syntactically and semantically essentially the same, we fail to explain why there is a predicate restriction in the existential, but not in the copular, progressive, or passive constructions (see Section 3.3). While the predicate restriction could of course also be determined lexically by be, an analysis on which the DE and the predicate restriction are related is, prima facie, more elegant and interesting. 20 The sole putative counterexample to the conservativity claim is only; however, only's status as a determiner has been disputed. See Keenan and Stavi 1986, Rooth 1985. 21 Actually, (37), (38) and (39) below are theorems of the definitions of conservativity, co-conservativity, and intersectivity, but no harm will arise from treating them as definitions. Throughout this discussion, in cases where no confusion will arise, I may use determiner or DP rather than the denotation of a determiner/DP. 22 Keenan and Stavi's notion of determiner is much broader than that of most linguists-it includes e.g. Det+Adjective combinations. The conservativity theorem holds for this generous class of basic determiners and its closure under Boolean operations. In addition, the definitions and theorems generalize to what they term "n > 1place" determiners (viz. more ... than ... , as in more Democrats than Republicans). 23 See below for a more precise statement of Keenan's proposal concerning the appearance of these DPs in the construction. 24 Keenan's intersectivity is more general than B&C's intersectivity; it constitutes a refinement of his notion ezistential from Keenan 1987.

Chapter 1. 25 But

The Problem

see Chapter 3, Section 4.2.1 for a nonquantificational treatment of no following Ladusaw 1992. 26 0f course these sentences also admit a narrow scope reading for the quantifier. 27 The presupposition of descriptive content for every and other quantificational DPs corresponds to the intuition that the domain of quantification should not be empty; since the complement to every always supplies the domain of quantification, that complement should have a non-null extension. In contrast, DPs with negative determiners have properties which suggest that, at least in some cases, they should be treated not as generalized quantifiers but rather as indefinites constrained to appear within the scope of a semantically potent feature [NEG] (cf. Ladusaw 1992). I will follow Ladusaw in assuming that DPs such as no student need not be analyzed as generalized quantifiers. If cne does not make this assumption, then it must somehow be stipulated that DPs marked with negative determiners need not carry a presupposition that their descriptive content is satisfied. 28 The same sort of question is raised by the condition proposed in Zucchi 1995 (his Felicity Condition 1): Felicity Condition (FC) 1: Existential sentences are felicitous only if the context c (defined in terms of a set of propositions (common ground), a variable assignment function, and a domain of individuals) in which they are uttered entails neither [N'] n [XP]([V]) = 0, nor that [N'] n [XP]([V]) # 0 (N' taken from the postcopular DP in the construction; XP, the phrase which optionally follows the postcopular DP; V, be, which Zucchi takes to denote the universal property). Simply put, this condition requires that at the time of utterance, the discourse model be neutral with respect to the question of whether the descriptive content of the postcopular DP is satisfied by some individual in the domain of that model, where the relevant context for the felicity condition is stipulated to be circumscribed by the XP (The appeal to domain circumscription is not crucial to the present point.) Zucchi's felicity condition is superior to the statement in (52) to the extent that it is more than a simple descriptive generalization concerning the postcopular DP; however, the comments that I make in the text concerning (52) will apply to it as well.

Chapter 2

Existential Syntax 2.1

Introduction

In this chapter, I present the D-structure syntax for the existential construction that I will be using for the rest of this dissertation. The structure to be advocated for a sentence such as {61) appears in {62): 1

{61) (62)

There was a student taking notes. IP

A I' DA There I VP DP

I V'

~ XP=VP

V'

~~ DP taking notes

V

lA

was~

a student 45

The English Ezistential Construction The proposal embodied in (62) has two crucial properties: First, the DP and XP do not form a constituent. Second, I take the XP to be a semantic adjunct similar in status to the depictive adjuncts in (63), for which a VP-internal position has also been defended, with minor differences (Rapoport 1991; see also Carrier and Randall1992 for a similar proposal for resultatives): (63)

a. She gave me the towel wet. b. They carried him screaming.

That is, on the view developed here, the XP in the existential construction is neither a complement nor a constituent of a complement to be. This position has also been defended by Zucchi 1995, though on different grounds from those to be articulated here. The proposal thus stands in contrast both to the "small clause" analysis that has been commonly assumed since Stowell1978, 2 and to analyses that take the XP to be a complement to be (e.g. Keenan 1987, Pollard and Sag 1994). After elaborating further on the relevant notion of VP-adjunct and what I take its syntactic structure to be, I will justify (62), in particular the treatment of the XP as a non-complement to be, in two steps. First, I will show why an independent DP and XP are needed in the first place. Second, I will show that neither of applicable diagnostics for small clausehood proposed by Stowell 1991 is reason to argue against the proposed analysis for the existential construction, and that extraction facts in particular offer positive evidence for an analysis taking the XP to be an adjunct rather than a complement.

2.2 2.2.1

Identifying VP-Adjuncts Description

What I am referring to as predicative VP-adjuncts includes what have been referred to in the literature as "depictives" and "circumstantials," depending on whether they modify an object/theme or subject, respectively (cf. Halliday 1967, Green 1973, Roberts 1988, Rapoport 1991). 3 The discussion here will be confined to

Chapter 2. Ezistential Syntaz

the object/theme-controlled depictives. Some examples appear in (64): (64)

a. Max drank the lemonade unsweetened. b. Kim handed the book unopened to her sister. c. That guy always plays his guitar untuned.

The term "VP-adjunct" as used here should not be taken to mean "adjoined to VP in the syntax." Rather, I refer to the XPs in (64) as "VP-adjuncts" because they are not semantic arguments of the verbal projections with which they combine, in contrast to the predicate complement of e.g. consider: the absence of the predicate in (65)b radically changes the possible interpretation of consider, while no such change is effected by the absence of the predicate in (65)d (compare (65)c): (65)

a. We consider Mary enthusiastic. b. We consider Mary. c. Kim handed it unopened to her sister. d. Kim handed it to her sister.

Nonetheless, these predicates have the effect of restricting the state of affairs denoted by the verb in the clause in which they appear, much in the way that an adjective modifies a noun. Moreover, although the adjunct is not semantically selected for by the verb, not every verb licenses an object-controlled depictive adjunct: (66)

a. *Terry kissed her motheri tiredi. b. *Joan kicked the hal~ weti.

In contrast, IP-level predicative adjuncts (or "free" adjuncts, Stump 1985), are not restricted in this way: (67)

a. Sad to be leaving, Terry kissed her mother. b. Hoping for a goal, Joan kicked the ball.

Restrictions such as those exemplified in (66) are the hallmark of structurally local relations.

48

The English Ezistential Construction

Green 1973, Andrews 1982, Roberts 1988, and others provide additional, syntactic evidence for the claim that these predicates appear within some projection of VP. For example, the adjuncts in question can accompany the verb in VP- and thoughr-preposing and in wh-clefting: (68)

a. Steve warned his brother that he would throw the towel to me wet, and throw it wet, he did. b. Play the piano out of tune though you may, you will not succeed in annoying the neighbors. c. What Margaret did was hand the book to Joyce unopened.

Further evidence for the VP-hood of at least object/themecontrolled VP-adjuncts is the fact that they cannot be left behind when the VP is reduced to do so (examples from Green 1973, 261): (69)

a. She'll bring them to me alive if you don't do so. b. *She'll bring them to me alive if she doesn't do so dead.

Finally, to the extent that object/theme-controlled predicative adjuncts can appear sentence-initially, they are interpreted as focus- preposed: (70)

a. Untouched,, they left it,. b. New,, they bought them,.

The intonation and interpretation associated with these sentences clearly contrasts with that of sentence-level free adjuncts: (71)

Raw, the eggs will taste awful.

We now need to decide on a D-Structure for VPs containing depictive adjuncts; this is the structure we will subsequently use for existential sentences.

49

Chapter 2. Ezistential Syntaz

2.2.2

A Proposal for Depictive Adjuncts

The literature offers us a variety of proposals to choose from for the analysis of the depictives, including a triple branch structure ((72)a; see e.g. Rapoport 1991), sister to V' ((72)b, e.g. Rizzi 1990), and adjoined to VP ((72)c, Rothstein (1983)): (72)

a.

VP

~

V DP XP

b. VP

1\ 1\ V DP

V' XP

c.

VP

1\ 1\ V DP

VP XP

The differences between these structures are minor for our purposes. I follow Rizzi in choosing a version of (72)b; however, I differ in taking the node dominating the depictive and V' to be another V'; that is, I assume base-generated adjunction to V' is possible, as in (73): (73)

VP

I V'

1\ 1\DP

V' XP

V

This assumption is not uncontroversial, but it has the advantage of allowing us to reserve Spec of VP for D-structure subjects (see e.g. Kitagawa 1986, Kuroda 1988), should it be necessary or desirable (see e.g. McNally 1992b). I follow Rizzi principally because, as he shows (1990:46ff. ), the adjunction structure allows for the cleanest analysis of extraction facts such as those to be discussed in Section 4.2. For example, the VP-adjunction option does not allow for a structural distinction between predicative adjuncts and manner adverbials, a distinction that seems motivated on the basis of contrasts in extraction such as the following:

50 (74)

The English Ezistential Construction a. How quickly did they eat dinner? b. *How raw did they eat the meat?

The VP-adjunction structure makes other incorrect predictions (e.g. that the depictive XP can be stranded by VP-preposing) and has been argued against in Rapoport 1991 and references cited there. The decision to choose a version of (72)b over the triple branch structure is dictated entirely by the definition of proper head government and the version of the ECP that I will adopt (again, see the discussion of extraction in Section 4.2. ); if an explanation for the non-extractability of the predicative XP is possible for the triple branch analysis, then I would have no reason not to adopt it. What is crucial is that depictive XPs are treated as semantic adjuncts, and not as complements or parts of complements. The next step in our defense of the proposal in (62) is to demonstrate that the XP in the existential construction is not a subconstituent of the postcopular DP.

2.3

Why a DP-External XP Must Be Posited

The idea that the optional XP in existential sentences is a postnominal modifier goes back at least to Jenkins (see references in Jenkins 1975); its most recent proponent has been Williams 1984. On this analysis, a sentence like (61) has only the structure in (75):

Chapter 2. Ezistential Syntaz

(75)

51

IP

A I' D/\ There I VP DP

~DP

V

w~s~ D

!

NP

J,

D

XP~VP

~ taking notes

student I first present arguments due to Milsark and others that show (75) not to be the only available analysis; I then consider Williams' arguments for (75) and show them not to be compelling when the alternative is (62).

2.3.1

Arguments for Independence

There are at least two convincing arguments in the literature for locating the XP outside of DP and yet internal to VP in the existential construction. First, the simplest and most commonly cited argument is that strings appear in the postcopular position which cannot occupy other positions typically held by DP, e.g. subject position, and which are not amenable to an analysis on which the XP is VP-external. Some examples appear in (76) (cf. Barwise and Cooper 1981, Keenan 1987, from whom I take (76)a): (76)

a. There are two students who object to that enrolled in the course. b. There is a woman who knows you working at the Pub these days. c. There was nobody but Mitzi interested in that movie.

52

The English Ezistential Construction

As both Keenan and Barwise and Cooper point out, and as the data in (77) confirm, the postcopular material in (76) cannot replace a DP in subject position: 4 (77)

a. *Two students who object to that enrolled in the course just came in. (Keenan 1987, (28b), ?* is his judgment) b. *A woman who knows you working at the Pub these days just won the lottery. c. *Nobody but Mitzi interested in that movie will be at the theater.

There are at least two reasons to believe that the XPs in (77) cannot be adjoined to IP. First, IP-level modifiers are generally preposable, but the XPs under discussion are not: 5 (78)

a. *Enrolled in the course, there are two students who object to that. b. *Working at the Pub these days, there is a woman who knows you. c. *Interested in that mov1e, there was nobody but John. d. *To blame, there was only myself.

Note that depictive adjuncts likewise fail to prepose: (79)

a. *Cooked, they ate the vegetables. b. *Unopened, she handed her sister the book.

Second, an IF-adjunction analysis violates the locality condition generally assumed to govern controller-controllee relations. A version of this locality violation can be seen in the contrast in (80). IFadjoined predicative XPs (which, when postposed, can be identified by a preceding intonation break; cf. {80)a), unlike VP-adjoined XPs, cannot modify a DP that is interpreted as a bound variable {compare (80)b and c; I am assuming following Ladusaw 1992 that nothing is so interpreted). The unacceptability of (80)b is probably

Chapter 2. Ezi&tential Syntaz

53

due to a requirement that an adjunct with an operator-bound controller be within the scope of the operator binding the controller. The fact that the DP in the existential sentence in (80)d can control a sentence-final XP thus argues against IP-adjunction: (80)

a. She ate the meat with her bare hands, uncooked. b. *She ate nothing that could not be identified, uncooked. c. She ate nothing that could not be identified uncooked. d. There was nobody who knew French available to guide the visitors.

Since we find postcopular material that is neither analyzable as DP nor amenable to an analysis where the DP-external material is IP-adjoined, on distributional grounds we are forced to posit an independent XP which is somewhere within the sentence projection. My claim is that it is adjoined to V1 • A second argument for locating the XP external to DP is that extraction and comparative deletion can target the latter without the former (Keenan 1987): (81)

a. Who is there performing at the Academy this week? b. The new mall ruined the few businesses there were still functioning downtown. c. There are more students voting for the environmental initiative than there are voting against it.

Compare the ungrammatical results obtained when extraction or comparative deletion targets the head of DP while leaving a postmodifier untouched:

54

(82)

The English Ezistential Construction

a. *Who do the musicians admire performing at the Academy this week? b. *The businesses to which the city has given a tax break still functioning downtown are in greater danger than ever. c. ??The lobby endorsed more candidates supporting gun control than they did opposing it.

A final argument concerns those cases where DPs headed by every can appear in existential sentences (see Lumsden 1988 and Chapters 1 and 3), such as (83): (83)

a. There is every breed of dog with any chance of winning competing in the competition. b. There is every kind of music anyone can 1magme available on that jukebox.

Ladusaw 1979 showed that the semantic properties of every, in concert with the conditions governing the distribution of negative polarity items (NPis) such as any, entail that every licenses NPis only within the DP it heads, and not external to that DP (compare (84)a and b): (84)

a. Everyone with any money has bought a VCR; why haven't you? b. *Everyone has any money.

If the final XPs in (83) were part of the postcopular DP, we would expect to find NPis licensed within them as well. However, the examples in (85) show that they are not: (85)a is ungrammatical, and (85)b allows only the quasi-universal "free choice" reading for any jukeboz, and not the existential reading it would have in the scope of an NPI licenser: (85)

a. *There is every breed of dog with any chance of winning competing in any competition. b. There is every kind of music that anyone can imagine available on any jukebox.

Chapter 2. Ezistential Syntaz

55

I believe the above facts are sufficient to show that the postcopular string cannot always be just a DP.

2.3.2

Williams' Counterarguments

Williams 1984 offers several arguments for the DP-only structure in (75), but none of them are compelling, largely because they are based on the assumption that the XP is a complement just like the postcopular complement in other copular constructions-an assumption that I do not make. First, he points out that the XP is optional, which is surprising if the XP is structurally related to that in other copular sentences, but expected if the XP is a postnominal modifier. On the adjunct analysis advocated here, the optionality of the XP is predicted. Williams bases a second argument on the failure of the XP to extract, citing the following example: (86)

*How happy was there someone?

Observe again, however, that failure of extraction is not an argument against positing an independent DP and XP; it only argues against analyses on which the XP is treated as a complement. In Section 4.2 I will show how the structure I advocate predicts (86). A third argument is that only the DP analysis predicts the prohibition on DP codas, as in (87): {87)

*There is a senator a CIA agent.

Again, this is a valid criticism of an analysis that takes the XP to be just like the complement of a copular sentence, but I will show in Chapter 4 that this prohibition follows from the treatment of the XP as an adjunct. Sketching the argument briefly, it has been claimed that adjunct XPs must be stage level predicates in the sense of Carlson 1977b (see e.g. Rapoport 1991 for defense of this view). Since DPs quite universally pattern as if they were individual-level predicates, they cannot function as predicative VPadjuncts:6 (88)

a. *Martha handed me the towel a mess. b. *Felix was walking home a drunk. c. *Anna played tennis a child.

56

The English Ezistential Construction

Fourth, Williams observes that VP-deletion can target both the DP and XP ((89)a), a fact that is ostensibly unexplained if the two do not form a constituent; a similar point can be made with coordination ((89)b): (89)

a. There was a band playing in the park last Sunday, and there will be this week, too. b. Today there is a band playing in the park and a theater group performing in Rittenhouse Square.

But this argument is undercut by the fact that these kinds of phenomena are reveal more about S-structure than D-Structure. For example, cases of coordination have been attested (as in (90), adapted from Johnson 1991) involving strings that, given the Projection Principle, could not form a constituent at D-Structure: (90)

Bob ate the vegetables quickly but the meat slowly.

The present analysis could account for this kind of coordination in the existential construction as well as for the VP-deletion in (89)a by motivating the raising of be from V to I (or some other functional projection in an exploded INFL). Different versions of this general sort of strategy have been advocated recently by e.g. Larson 1990 and Johnson 1991. Verb raising would give the effect of making a surface constituent of the DP and XP to the exclusion of be, while allowing us to maintain the desired D-structure in (62). 7 A final argument of Williams' involves the failure of Heavy-NP Shift in the construction, as in his (22b), repeated here in (91)a: (91)

a. *There are sick several of George's recent acquamtances. b. They considered crazy several of George's recent acquaintances.

Williams observes that if the XP were independent of the DP, Heavy-NP Shift should be possible, just as it is in cases like (91)b. The explanation for the failure of Heavy-NP Shift in this case is not fully clear to me. However, the force of this argument is weakened by the fact that Heavy-NP Shift is sometimes blocked by factors other than syntactic constituency, as in the well-known case of double-object constructions:

Chapter 2. Ezistential Syntaz

(92)

57

a. Terry gave her sister a present. b. *Terry gave a present her favorite person m the world.

It is thus entirely possible that some alternative explanation for the contrast in ( 91) can be found.

Thus, with the exception of this last one, none of Williams' arguments gives us any strong reason not to pursue the analysis presented at the beginning of this chapter; indeed, in concert with the positive arguments we have seen for assigning the XP a DPexternal position, some of his arguments actually support (62), to the extent that they constitute valid criticisms of analyses on which the XP is treated as an independent complement.

2.4

Supporting the Adjunct Analysis

Having established that the postcopular material does not have to form a DP, we may now turn to the specific proposal in (62), repeated below in (93)a. In this section I discuss the data and other considerations that lead me to adopt {93)a over the small clause-type structure in (93)b, and more generally over a structure in which XP is an independent complement of be: 8 (93)

a.

IP

A I' DA DP

There

I VP

I V'

1\ A

V' XP

V DP

I

be

b.

IP

A I' DA There I DP

VP

1\ I 1\

v

be

XP

DP XP

The English Ezistential Construction

58

I focus on the small clause analysis because it has been so prevalently assumed in the GB literature; I begin by applying Stowell's 1991 diagnostics for small clausehood to the existential construction.

2.4.1

Stowell's Small Clause Diagnostics

A number of arguments have been advanced for the existence of small clauses in English, most notably in the work of Stowell (e.g. 1983, 1991). 9 These arguments are based on the following facts: • The appearance of DP-XP strings as the complement to prepositions in absolute constructions (With Mary gone, we cannot make the decision) • The possibility of that string in preverbal position in a copular sentence: Workers angry about the pay is the sort of situation that the ad campaign was supposed to avoid (example due to Safir). • The fact that the DP-XP constitutes a binding domain for reflexives, as seen in the failure of the matrix subject to be a possible reflexive binder in e.g. *Mary considers John proud of herself • The inability of main clause "floating" constituents such as adverbs to intervene between the DP and XP (see below for examples). • Subject condition effects on the DP, i.e. contrasts in the acceptability of extraction (see below for examples). These arguments are largely theory-independent. In addition, certain theory internal considerations motivate the small clause, most strongly, the requirement that all predicates-main, subordinate or adjunct-have a (local) syntactic subject, be it overt or null; and the requirement that all arguments of a predicate must be projected within the maximal projection of that predicate. Since these theory-internal considerations are bound up with assumptions about interpretation that I do not share, and since it is

Chapter 2. Ezistential Syntaz

59

certainly possible to construct appropriate interpretations for sentences without committing oneself to these assumptions (see in particular Chierchia 1984, 1985 for discussion and a view to which I am sympathetic), I will not consider them in evaluating the small clause analysis. Of the five arguments based on these facts, only two are potentially useful in the case of the existential construction, viz. those based on the floating constituent data and the subject condition data. Arguments sensitive to the first two bulleted facts are obviously of no use because we are evaluating the status of a post verbal string, and an argument from the binding facts (third bullet) cannot be made because there is only one DP in existential sentences that can serve as a (relevant) potential binder, given that the subject is an expletive.

The "Floating" Constituent Argument Postal 1974:146ff., citing Kuno 1972, makes the claim that adverbs with matrix construal cannot be inserted in a clausal complement to the matrix predicate, on the basis of contrasts such as that in (94): (94)

a. I have found that Bob recently has been morose. b. I have found Bob recently to be morose.

The adverb in (94)a is construed exclusively with the downstairs clause, while that in (94)b has either matrix or downstairs construal. Postal takes these facts to support the hypothesis that Bob in (94)b does not form a surface constituent with the infinitive. Stowell 1991 takes up this argument and uses the impossibility of a matrix construal in (95)a, in contrast to the acceptability of such a construal in (95)b and c, as evidence that the postverbal material in (95 )a is a small clause constituent: (95)

a. John considers Bill sincerely foolish(* on matrix construal) b. John promised repeatedly to leave. c. John ordered Mary repeatedly to leave.

60

The English Ezistential Construction

Now, an adverb interpolated between the DP and XP in an existential sentence is generally not construable as modifying the main clause, as pairwise comparison of the examples in (96) shows: (96)

a. There was often a student interrogated by the dean. b. There was a student often interrogated by the dean. c. There was obviously nobody getting shot in that part of town. d. There was nobody obviously getting shot in that part of town.

Though this might be taken prima facie as an argument for the small clause analysis, the argument depends on demonstrating that Postal's claim is a biconditional: that is, we have good evidence that if the adverb appears in a downstairs clause, matrix construal is impossible; but is it the case that if matrix construal is impossible, the adverb must be in a downstairs clause? The facts in (97) suggest that this second half of the biconditional fails, and therefore render the adverb interpolation facts inconclusive as a diagnostic for the relation between the DP and XP in the existential. The sentences in (97) contain two complements and a depictive adjunct controlled by the DP a book. This adjunct cannot possibly form a small clause with its controller, since the two are not adjacent, nor can it form a small clause with e.g. to Mary, since that would result in the wrong interpretation. Nonetheless, when the adverb appears immediately prior to the adjunct, it can only be construed as modifying the adjunct predicate: (97)

a. Bob handed boxes to Mary often unopened. b. Bob put boxes on the counter repeatedly unopened.

The facts do not change when a normal transitive verb is used: the adverb in (98) cannot be construed as modifying the main verb: 10 (98)

The child ate vegetables rarely uncooked.

A complete explanation for these facts would entail more investigation than can be done here, since the semantics of adverbials is

Chapter 2. Ezistential Syntaz

61

quite complex and must be controlled for. However, I venture the following hypothesis: Assume the adverbs under consideration here are generated right-adjoined to VP and that the effect of adverb interpolation in the VP is really a case of attempted extraposition of VP-internal material past the adverb. Some support for this position comes from the fact that adverb interpolation, though generally quite acceptable between DP and PP, is quite strange if the DP following it is light, but acceptable if the DP following it is heavy: (99)

a. ??I tossed Joan carelessly the book. b. I tossed Joan carelessly the beautiful book she had loaned me last week.

The VP for the sentence in (100) will then have the structure in (101) (I adopt the tripartite structure for the sake of simplicity; it is not crucial): (100)

Bob handed the boxes carefully to Mary.

(101)

VP

~PP,

VP

~D AdvP to Mary

VP

~

V

ha~ded

D

DP

PP

the boxes

ti

careLlly

~

If this is the right structure, we can account for the facts by showing that rightward movement of certain complements, but not adjuncts, is licensed. To determine whether and how the relevant structures are licensed, some theoretical notions will have to be introduced. The basis for the analysis will be Rizzi's Relativized Minimality (Rizzi 1990) approach to the licensing of extraction. Before proceeding, let me lay out the definitions of the relevant theoretical notions to which I will be appealing. 11 I assume the

The English Ezistential Construction

62

following definitions of head and (A-bar) antecedent government, adapted from Rizzi 1990, Chapter 1: (102)

Head Government: X head-governs Y iff (i) X is a head and X m-commands Y (ii) X is A, N, P, V (iii) no barrier intervenes (iv) Relativized Minimality is respected.

(103)

(A-bar) Antecedent Government: X (A-bar) antecedentgoverns Y iff (i) X andY are co-indexed (i) X c-commands Y (iii) no barrier intervenes (iv) Relativized Minimality is respected.

Relativized Minimality is defined as follows: (104)

Relativized Minimality: X a-governs Y only if there is no Z such that: (i) Z is Typical Potential a-governor of Y (ii) Z c-commands Y and does not c-command X.

In addition, I adopt a conjunctive definition of the Empty Category Principle (ECP, Rizzi 1990:32): (105)

ECP: A non-pronominal empty category must be (i) properly head-governed, and (ii) theta-governed or antecedent-governed.

I further follow Rizzi in defining proper head government as government by X0 under sisterhood. Theta government is understood to exclude inter alia lexically selected adverbs, measure phrases, and idiom chunk complements. Thus, none of the italicized constituents in the examples in (106) are theta-governed: 12 (106)

a. The dog behaved badly. b. Those boxes weigh thirty pounds. c. You're making little headway.

Chapter £. Ezistential Syntaz

63

With these theoretical preliminaries in hand, we can proceed. Consider (101) again, repeated in (107): {107)

VP

~PPi

VP

~D AdvP to Mary

VP

~

V

hadded

D

DP

PP

the boxes

ti

careLlly

~

The trace of the dative PP is properly head-governed under the sisterhood definition of proper head government adopted here. It is also theta-governed, since it receives a theta-role from the verb. Consequently, both clauses of the ECP will be satisfied, and the structure is licensed. Now consider the VP structure for (98): (108)

VP

~

VP

/\

VP

AdvP

V'

rarely

I

APi

I

uncooked

I

~AP

V'

~~

V DP ti I~

ate~

vegetables

This structure will not be licensed because the trace of the extraposed AP is not properly head-governed. As failure of proper head

The English Ezistential Construction government is a sufficient condition for violation of the ECP, the structure is ruled out and the corresponding sentence is predicted to be ungrammatical. Since the adverb cannot be construed as if it were adjoined to VP (i.e. as modifying the main verb) without engendering an ECP violation, the only possibility is to construe it as a modifier of the adjunct predicate. Since, on the analysis advocated in this chapter, the XP in the existential construction is adjoined to V', just like the depictive adjunct in (108), the explanation sketched here for the interpretation of (98) will carry over to the facts in (96). Due to the availability of such an alternative account of the facts, the failure of main clause construal for interpolated adverbs in the existential construction ceases to be an argument for the small clause structure. As it stands, this proposal leaves unexplained the contrast between the oddness of adverb interpolation in Raising to Object/ECM ((94)a) and its acceptability in object control cases such as (95)c, as well as the unacceptability of extraposition of complement predicative XPs vs. PPs and infinitives, and a serious attempt at explaining these phenomena would take us too far afield. However, even in the absence of an account of the effect of adverb interpolation between DP and infinitival VP, we may conclude that the facts do not favor the small clause analysis of the existential over the analysis proposed for sentences containing depictive adjuncts in general and the existential construction in particular. The Subject Condition Argument

The remaining small clause diagnostic involves extraction out of the postcopular DP. Stowell, citing Kayne, observes that we might expect small clause subjects to be like sentential subjects in resisting extraction, then notes that this prediction is not very consistently borne out (cf. his example in (109)c). My own intuitions are that extraction from the subject of a putative small clause depends in part on the DP being extracted from; (109)d seems much better than (109)c:

Chapter 2. Ezistential Syntaz

(109)

65

a. *Who would for John to visit _bother you? b. cp. Who would it bother you for John to visit _? c. ?*Which books did you find the authors of _ very eloquent? d. Which city did you find pictures of _ enticing?

However, Stowell makes the point that the relevance of subject condition effects depends on the formulation of extraction conditions; thus, to the extent that sentences such as (109}c,d are acceptable, it could be because the DP being extracted from is governed by the matrix verb rather than by the predicate complement. Consequently, the possibility of extraction in cases like (109)c,d does not preclude a small clause structure, though it certainly gives us no positive reason to adopt the small clause structure over some other structure that equally well accounts for the facts. As it turns out, extraction is consistently acceptable from the postcopular DP of the existential: (110)

a. Which book was there an author of_ on TV? b. The politician about whom there was an article _ published in the Voice could not be reached for comment. c. It was Elizabeth Taylor that there was a feature on _ in this week's People. d. What there was no solution to _ was the candidate's image problem. e. To this problem, there is only one solution _.

This result is perfectly consistent with the structure proposed at the beginning of the chapter, since the postcopular DP is a direct argument of be. Therefore, the absence of subject condition effects gives us no reason to favor a small clause analysis over the analysis in (93)a (or, again, a triple branch analysis on which the XP is treated as a complement). If anything, the fact that (llO)a is fully acceptable in contrast to (109)c raises doubt about the appropriateness of the small clause structure for the existential.

66

The English Ezistential Construction

We have seen that neither of the viable tests suggested by Stowell for identifying small clauses supports a small clause-type analysis over the XP-as-adjunct analysis in (93)a. I now turn to evidence that lends positive support to the latter analysis.

2.4.2

Extraction

Two kinds of extraction support the position that the XP is an adjunct and not a complement. First, if the XP were a complement, then we would predict it to be extractable, just as the XP complement to consider is extractable. But, as Williams 1984 observed, it is not: (111)

a. How clever do you consider Angela? b. How happy did the award make him? c. *How available was there a man? d. *How sick were there children?

Second, adjuncts and complements differ in that the extraction out of the latter is more restricted than is extraction out of the former (Huang 1982). Though theta-governed complements can be extracted from an adjunct, as in {112)b (where the adjunct is an object-controlled depictive), the same is not possible for material that is not theta-governed: {112)d,f,h, which respectively show the extraction of an adverbial, idiom chunk complement, and measure phrase from an adjunct, are all bad:

Chapter 2. Ezistential Syntaz

(112)

67

a. We cornered her stealing the boxes. b. ?What did you corner her stealing? c. We cornered them behaving badly in the yard. d. *How badly did you corner them behaving in the yard? e. We ran into them trying to make headway on the problem. f. *How much headway did you run into them trying to make?

g. We found the box weighing more than a pound less than it was supposed to. h. *How many pounds did you find the box weighing? The same point can be made with subject-controlled XP-adjuncts: We can extract a complement from within the adjunct ((113)b) but not any material that is not theta-governed (e.g. (113)d,f): (113)

a. Frank arrived wearing two shirts. b. ?How many shirts did Frank arrive wearing? c. Frank arrived behaving quite badly. d. *How badly did Frank arrive behaving? e. Frank arrived weighing only 95 pounds. f. *How many pounds did Frank arrive weighing?

In contrast, it is possible to extract both theta-governed and non-theta-governed material from a complement, as the examples in (114) demonstrate:

68

(114)

The English Ezistential Construction a. Of what do you consider Hillary capable? b. How badly do you consider her capable of behaving? c. How much headway do you consider her capable of making? d. How many pounds do you consider the horse capable of carrying?

Given this contrast, the prediction is clear: if non-theta-governed material such as an adverbial or an idiom-chunk cannot be extracted out of the XP in the existential, we have good reason to conclude that the XP is an adjunct. If the extraction is grammatical, we have support for the claim that the XP is a complement. The facts support the adjunct analysis. Though we see from the examples in (115) that extraction oftheta-governed constituents from within the XP yields a grammatical result, the comparable examples in (116) showing extraction of measure phrases, adverbials, and idiom chunks are quite bad: {115)

a. To whom has there just been a celebrity introduced? b. ?How many cookies have there been children baking?

(116)

a. *How many miles a day are there people running? b. *How badly has there been a man shot? c. *How much headway could there be people making?

Let us now see how these contrasts support the analysis in (93)a over that in (93)b. I first show how only the V' adjunction structure accounts for the contrast between ( 111 )a and b, focusing on the predictions concerning the (b) sentence. A small clause structure for (111 )b appears in ( 117):

69

Chapter 2. Ezistential Syntaz

(117)

CP

~C'

AP·

~~ Ci IP

How available

I./"-..

was

/ DP

""-

I'

DA A v

there

VP

I

t1i

I

AP

1\

DP

AP

a man

1

D

~t·

The discussion will presuppose the definitions introduced in the treatment of adverb interpolation in Section 4.1. The small clause (here, AP) is both properly head-governed and theta-governed, since it is semantically selected by the verb, and a sister to it as well; and these properties of the small clause are inherited by the maximal projection that is its head, viz. the AP sister to DP (cf. Rizzi 1990:49). The AP's trace is thus properly head-governed and theta-governed, with no intervening potential head-governor to violate Relativized Minimality. Finally, no barriers are crossed by the chain connecting how available to the trace. The structure in ( 117) should therefore be licensed, and the sentence is incorrectly predicted to be grammatical on the small clause analysis. The situation would be in all relevant respects the same if the AP were not part of a small clause, but were instead an independent complement to the verb. However, if we adopt the analysis of the existential construction advocated in this chapter, taking the XP to be an adjunct, we have the following for (111)b:

70

The English Ezistential Construction CP

(118)

~C'

AP·

~~ Ci IP

How available

I/\

was

DP

I'

D/\ there I VP I

ti

I

V'

A

V'

AP

A~ DP ti

V

ii

D

a man

Adjoined to V', the XP is not properly head-governed, since proper head government is defined as government under sisterhood to V 0 . The absence of proper head government results in an automatic violation of the ECP, under the conjunctive definition adopted here. The structure in (118) is, therefore, illformed, in accord with the ungrammaticality of (lll)b. The small clause analysis also incorrectly predicts that nontheta-marked material can extract from within the XP, as in the case of the idiom chunk complement in (116)c. On this analysis, (116)c has the structure in {119):

71

Chapter 2. Ezistential Syntaz

(119)

CP

~C'

DP·

~~IP

How much headway Ci

co~ld ~

I'

DP

D~ VP1 I

there

ii

A

V

VP2

~e ~

VPa

DP

DA people

V

matng

DP

~ t;

The trace of the idiom chunk is both head-governed by the verb and antecedent-governed by the phrase in Spec(CP), satisfying both clauses of the ECP. Crucial to the antecedent government of the trace is the fact that the small clause(= VP 2 in (119)) is semantically selected by be; as such, it will not be a barrier for extraction of the idiom chunk. 13 As before, the analysis works in exactly analogous fashion for a non-small clause structure in which the XP is treated as a complement. In contrast, on the V'-adjunction analysis advocated here, (116)c has the structure in (120):

72

The English Ezistential Construction

(120)

CP

~C'

DPj

~~IP

How much headway Ci

co~ld

A

DP

I'

~I\

there

I

VP1

ti

V'

I

I

~VP2

V'

A

V

Je

DP

6

people

A

V

DP

ma~ing ~

t1

Since it is not semantically selected by the verb, the XP ( = VP2) constitutes a barrier to government. Though the trace is properly head-governed by making, it is neither theta-governed (since idiom chunks are not theta-governed) nor antecedent-governed (since VP2, as a barrier, will block antecedent government). The structure is therefore not licensed, and the sentence is correctly ruled out. We see, then, that extraction affords a particularly clear reason to prefer an analysis on which the XP is treated as an adjunct to one on which it is treated as a complement, the small clause analysis being a variety of the latter.

2.5

Chapter Summary

In this chapter I have argued for a syntactic analysis of the existential on which the sentence-final XP is an adjunct, and not a complement, of the copula. In addition to the motivation provided here, the analysis has at least two positive consequences.

Chapter 2. Ezistential Syntaz

13

First, the optionality of the XP follows directly. Though of course it has always been possible to account for bare-DP existentials by independent stipulation on analyses taking the XP to be a complement, on the analysis defended here that stipulation is no longer needed. We may take the DP to be the sole complement to be and license the XP by the independently needed rule that licenses predicative VP-adjuncts. Second, in taking the XP to be an adjunct and not a complement, the proposed analysis dissociates the existential from the copular construction. This dissociation has two positive consequences which were discussed in Chapter 1: It opens the door for an explanation of the predicate restriction, which was argued to be inexplicable on an analysis that linked the existential and copular constructions; and it eliminates the incorrect prediction that other varieties of copular construction (specifically, those built around seem and appear) should have existential counterparts. In addition, the analysis developed in this chapter makes a cross-linguistic prediction, namely that the range of predicative XPs found in the existential in a language will correlate with the range of predicative adjuncts in the language. Analyses on which the XP is treated as a complement fail to make any such prediction.

74

The English Ezistential Construction

Notes 1 This

sentence has an alternative analysis on which all of the postcopular material is within DP; however, this alternative will not be relevant for the present discussion. 2 See also Jenkins 1975 for a proposal very similar to the small clause proposal. 3 Though there are interesting similarities between depictives and the other species of predicative VP-adjunct such as resultative predicates (Green 1973, Carrier and Randall 1992, Fernald 1991, Goldberg 1994) and purpose clauses (Bach 1982, Jones 1985), the remarks made in the rest of this chapter should not be construed as claims about these other classes. 4 Barwise and Cooper claim that the postcopular material can, counter to expectation, appear as the complement to certain prepositions, as in (i ): (i) Among the people sick were Bob and his wife. However, they provide only two examples, neither of which involve relative clauses. In contrast to (i), (ii) sounds bad: (ii) *Among the students who hate homework enrolled in the course were Martha and Jay. A further weakness in this argument is the fact that some prepositions arguably can take small clause complements, an clear example being with: (iv) With Gates no longer police chief, we will all feel safer. 5 There is a systematic set of exceptions to this generalization involving participles headed by verbs of motion or location, as in (i): (i) Sitting outside in the yard, there were three children. However, this does not weaken the argument in the text insofar as the data in the text still must be accounted for. 6 The claim that VP-adjunct predicates must be stage-level is something of an oversimplification. There is systematic set of exceptions to this claim, as in (i): (i) Martha returned from Belize a new woman. The licensing of this kind of modification depends on special factors that will be elaborated on in Chapter 4; for the moment it is sufficient to observe that relevant properties of the existential construction will preclude modification by an adjunct DP. 7 For alternative analyses of putative nonconstituent coordina-

Chapter 2. Ezistential Syntaz

75

tion, see e.g. Dowty 1988 and Steedman 1990. 8 I show here an adjunction analysis of small clauses, as in e.g. Chomsky 1986a. The arguments discussed here would not significantly change under the other analyses of small clauses with which I am familiar. 9 Evidence for small clauses has been adduced in other languages, e.g. Irish (Chung and McCloskey 1987). Since my purpose is only to evaluate the appropriateness of the small clause analysis for the existential construction, and not to evaluate the viability of small clause analyses for other constructions in English or other languages, I will discuss only evidence that is relevant to English. 10 Similar facts also hold for the resultative construction, which Carrier and Randall 1992 have argued not to be amenable to a small clause analysis: (i) They painted the barn carelessly red. 11 The reader is referred to Rizzi 1990 for additional details, and to Chomsky 1986a:10ff. for the relevant definition of barrier. 12 Rizzi points out certain problems with this notion of theta government. The present proposal could be reformulated in terms of his Chapter 3 treatment of the ECP, but not without technical complications that are not of particular relevance here. 13 It must be the case that the head of the small clause (VP 3 ) does not constitute a barrier either.

Chapter 3

The Existential, Descriptions, and Instantiation 3.1

Introduction

This chapter addresses the interpretation of the postcopular DP and the issue of the definiteness effect, defending the following proposal, introduced briefly in Chapter 1: (121)

The existential predicate in English is interpreted as a property of a description of an entity, specifically the property that the description is instantiated by some entity at some index. The addition of an existential sentence to a context entails the introduction of a discourse referent that corresponds to the instantiation of the descriptionargument into the domain of the discourse model. An additional felicity condition requires this referent to be novel.

(121) consists of two logically distinct claims: one about the argument structure of the existential predicate (viz. that it is interpreted as a 1-place property of a special sort of entity whose nature will be elaborated on shortly), and one about the conditions on and effect of an assertively uttered existential sentence with respect to 77

78

The English Ezistential Construction

the domain of discourse. From the first claim it will follow that certain necessarily quantificational DPs, though not definites, are excluded from the construction. The felicity condition in the last sentence of (121) is needed to exclude the definites. That is, (121) entails the claim that what we have been calling the definiteness effect consists of two independent phenomena. It will become clear below that the first claim in (121) can be recast as the assertion that the argument of the existential predicate is in some sense intensional; however, it differs notably from other intensional arguments inter alia in that the existence of an individual fitting the description contributed by the postcopular DP is entailed. The idea that the postcopular position in the existential construction is in some sense intensional is perhaps surprising, but it has surfaced before under the guise of the claim that the postcopular DP is somehow "nonreferential" (Fodor and Sag 1982, Safir 1987; indeed, I believe Lyons attributes a similar observation to Frege). My own motivation for exploring this possibility grew out of the observation that in several respects the postcopular DP in the existential construction resembles a predicate nominal-perhaps the prototypical exemplar of nonreferentiality. Some of these similarities will be introduced in the next section. The task of interest is to develop an interpretation for the existential construction that will clarify the sense in which its argument is intensional or nonreferential, and more generally, which will clarify these phenomena themselves. Specifically, our interpretation should allow us to capture appropriate similarities among the argument of the existential predicate, the predicate nominal, and more prototypical examples of intensional arguments such as the complement to seek, as well as their differences. I will ground the interpretation to be developed here in a version of the property theoretic semantics of Chierchia and Turner 1988, which grows out of independent work by both authors. Specifically, I will argue that the argument of the existential predicate must be interpreted as an entity of a special sort, viz. the entity correlate of a property or nominalized function, rather than as an "ordinary" entity. This proposal is close in spirit to the claim that the existential predicate is a predicate of something like a Carlsonian kind (Carlson 1977b), 1 and indeed Chierchia 1985 sug-

Chapter 3.

The Ezistential, Descriptions, and Instantiation

79

gests that nominalized functions might be appropriate interpretations for kind terms. I will posit a similar interpretation for the predicate nominal, following proposals by Farkas 1981 and Partee 1987, although I will not argue for this at any length. Note that the predicate nominal indeed manifests one of the hallmarks of an entity-denoting expression, namely, that it can be the target of quantification under certain circumstances (Williams 1983): (122)

Fred has been every kind of doctor.

We can paraphrase (122) as: "Every kind of doctor is such that Fred has been that kind of doctor." If we adhere to the commonly accepted position that the predicate nominal is interpreted as either a set of individuals or a functor of some sort, we are forced to abandon the position that quantification in English is strictly firstorder. Since, as Chierchia has argued, restricting quantification to the entity domain is desirable for both theoretical and empirical reasons, it is worth exploring the position that the predicate nominal is interpreted as some sort of entity, albeit a different sort of entity than that identified by e.g. Fred. Property theory offers the sortal distinctions within the entity domain that we need to do this. Taking the argument of the existential predicate to be a nominalized function will lay the foundation for establishing a link between it and the predicate nominal (though certain obvious differences will have to be accounted for), and will hopefully lead to insight into the intuition that the argument of the existential has something in common with more familiar intensional arguments. The following comment from the introduction to Chierchia et al. 1989:7 suggests why: Turner notes further that positions where properties and propositions can be referred to quantificationally are precisely those that, according to Frege, create intensional contexts. Frege proposed that in oblique positions, expressions have their ordinary sense as extension. This suggests that we might want to embed intensional creatures (i.e. properties and propositions) in the domain of individuals.

80

The English Ezistential Construction

Perhaps, then, the similarities between the argument of the existential predicate and familiar intensional arguments reduce to their interpretations. One of the aims of this chapter is to provide independent motivation for interpreting the postcopular DP in the existential construction as a nominalized function ( n/). The availability of independent means for identifying nfs will open the door for future evaluation of the correlation mentioned in this quote, by enabling us to decide on a case by case basis whether DPs in the full range of classically intensional argument positions can or must be interpreted as nfs. This, in turn, should eventually result in a better understanding of intensional phenomena more generally. The second part of the claim in ( 121) concerns the effect of uttering an existential sentence on the discourse model, i.e. what I will call its contezt change potential, following Heim 1982. 2 I will take a discourse model to be defined in terms of a set of worldassignment function pairs (which satisfy the formulas that constitute the context) and a domain of discourse referents. Although the domain of a discourse model is largely determined by the set of DPs explicitly introduced in the course of the conversation, it is not exclusively so determined. To give just one example, it has been argued (e.g. Prince 1992, Condoravdi 1992b, though see Barker 1993 for a different view) that the certain DPs license inferences that other individuals besides those to which they refer are part of the domain. Such inferencing is ostensibly seen in the acceptability of the italicized definite in the following discourse: (123)

A ghost appeared at a house on Wellington Street. The residents were frightened.

Though definites must typically be anaphoric, the use of the definite in (123) is arguably rendered felicitous by the introduction of the DP a hou8e on Reed St., plus the background information that houses generally have residents. The conditions determining the licensing of additional discourse referents are nontrivial and heterogeneous, and I will not have anything to say about them here. The point is that this sort of licensing is possible under certain circumstances. My claim, then, is that the existential construction always gives rise to a similar kind of licensing: Under the hypothesis that the existential predicate ascribes the property of being instantiated to a description of an

Chapter 3.

The Ezistential, Descriptions, and Instantiation

81

individual (formalized as a nominalized function), it is unsurprising that addition of an existential sentence to the context licenses, in addition to a discourse referent corresponding to the nf itself, a second discourse referent corresponding to the instantiation of the nf that supports the truth of the existential assertion, since every true nonnegated, nonmodalized existential of the form in (124)a will entail a sentence of the form in (124)b: (124)

a. There is DP XP. b. An individual which 1s an instantiation of DP has property XP.

Though future research will perhaps render it possible to derive the licensing of this extra discourse referent from some general principle, for the present, I will assume that it is simply stipulated as part of the context change potential of existential sentences. (121) further stipulates that this extra discourse referent introduced via the existential construction must be novel. This clause is necessary because both definites and indefinites will be interpretable as nominalized functions on the analysis advocated here, and yet definites are prohibited from the construction except under certain circumstances. Though I will show in this chapter how this novelty condition takes effect, I will postpone discussing its status in the overall interpretation of the construction until Chapter 5. The rest of this chapter proceeds as follows. I begin by discussing some properties of the argument of the existential predicate and of the predicate nominal that distinguish them from the arguments of other predicates. I will then introduce Chierchia and Turner's property theory, incorporating enough modifications to allow for dynamic interpretation. An interpretation for a fragment of English will subsequently be presented which will be large enough to illustrate some of the consequences of the claim that the argument of the existential predicate is a nominalized function. Once this much of the analysis is worked out, I will introduce definites into the fragment and discuss examples such as (125): (125)

There was the lid to a jar on the counter.

The chapter closes with a comparison of the present analysis with analyses that interpret the existential predicate as a property of

82

The English Ezistential Construction

"ordinary" entities, focusing on the data to be presented in Section 2.

3.2

Data to Be Accounted For

3.2.1

Quantificational DPs

We have already seen the definiteness effect facts associated with the existential construction. The copular construction is similar in the following respect: DPs that quantify over ordinary individuals are prohibited from the predicate nominal position, while those that quantify over kinds of entities are licensed ((126)a,b). Recall that we find an identical kind of contrast in the existential construction ( ( 126)c,d): (126)

a. *Fred has been every doctor. b. Fred has been every kind of doctor. c. *There was every doctor. d. There was every kind of doctor.

Of course, typically there is no such restriction on the use of quantificational DPs. 3

3.2.2

Scope

In general, VP-internal DPs such as many people may take either wide or narrow scope with respect to clausal negation. Consider (127): (127)

Andy didn't see many people at the party.

On the narrow scope for negation reading, (127) could be true in a situation where Andy saw 100 people she knew at the party, but failed to see 150 others she knew; on the wide scope for negation reading, (127) would be false in such a situation, though it would be true if Andy had seen only 5 people there. It has long been held on the basis of judgments concerning examples such as (128)a,b that the postcopular DP in the existential construction must always take narrowest scope with respect to

Chapter 3.

The Ezistential, Descriptions, and Instantiation

83

negation and any other clause-level operators. For example, (128)a cannot describe a situation in which 100 pictures were hanging on the wall, even if 150 were not. Similarly, (128)b is unacceptable because some must be interpreted with wider scope than negation (Ladusaw 1979), in violation of the observed generalization: (128)

a. There weren't many pictures hanging on the wall. b. *There hasn't been some student sick at school.

However, it turns out that this narrowest scope requirement is not absolute: The postcopular DP can take wide scope with respect to negation if it is headed by a noun such as kind, sort, type, varietythe same nouns that license quantificational DPs in the construction. For example, compare the the acceptability of ( 129) to the unacceptability of ( 128)b: (129)

The food critic was annoyed because there wasn't some type of wine on the list.

(129) could be true in a situation where the food critic was annoyed because Mer lot was missing, even though Pinot Noir was available. With determiners other than some, the availability of the wide scope interpretation for the DP often requires intonational prominence for the DP, facilitating context, or both (see (130)a). While I do not have an explanation for this, observe that nothing can be done to interpret the DP wide scope when it is not headed by a kind nominal, as in (130)b: (130)

a. The food critic is annoyed because there will not be one variety of wine available, namely Zinfandel. b. *The food critic is annoyed because there will not be one competent waiter available, namely George.

The postcopular DP in the copular construction manifests the same scopal behavior: (131)a can only be interpreted with negation wide, but (131)b is scopally ambiguous: (131)

a. Terry and Mary aren't two students. b. Terry isn't one kind of doctor I hoped she would be.

84

3.2.3

The English Ezistential Construction

Contact Clauses

A third characteristic shared by the existential and copular constructions involves the distribution of DPs containing a particular sort of postnominal modifier. In several dialects of English, including Hiberno-English and African American Vernacular English, tensed VPs 4 can be used as postmodifiers of certain nominals, in a construction which resembles a subject relative which is missing a relative pronoun or complementizer. Following Jespersen 1927, I will refer to this kind of post modifier as a contact clause. J espersen noted that DPs containing contact clauses typically show up in sentences with "meaningless existential" subjects, such as those in (132), existential sentences being a paradigm example, as well as in the complement to "existential" have and after the copula in some-though not all-contexts: 5 (132)

a. There's a student in my class went to America. b. This is the girl wants to see you. c. Here's the boy'll fix it for you. d. I have a brother works in Dublin. e. I'm the only one knows how to act.

However, Doherty 1993 shows that contact clauses are generally prohibited from typical extensional contexts, as seen in the (133): 6 (133)

a. *I married the man could do that. b. *I didn't hit a fella said that. c. *Anyone does that is arrested.

Interestingly, the only non-copular/non-existential contexts in which contact clauses can appear are modal or intensional: (134)

a. I want to marry a man always pays his debts. b. I wouldn't hit a fella said that. c. Anyone does that should be arrested.

Chapter 3.

The Existential, Descriptions, and Instantiation

85

Given the ungrammaticality of the examples in (133), the licensing of contact clauses in the existential construction is surprising if the postcopular DP is interpreted as an ordinary entity in an extensional context.

3.2.4

Relative Clauses

A fourth special property of the postcopular DP in the existential construction and the predicate nominal involves relativization. First, for many (though not all) speakers, the DP in the existential construction cannot be relativized with wh-relative pronouns, especially who ((135)a), although it can be relativized with that or a null complementizer ((135)b; "%" indicates a dialect split): (135)

a. %The people who there were at the party were drunk. b. The people (that) there were at the party were drunk.

This is true of the predicate nominal as well: ( 136)

a. *They dressed like the eccentric women who they were. b. They dressed like the eccentric women (that) they were.

Moreover, the relativized nominal in both cases can follow only a limited class of determiners, all of which have universal force. The range of determiners available for DPs modified by relatives formed from the existential are illustrated in (137)a,b. A sample of those excluded appears in (137)c,d: (137)

a. The/Every/ All the beer there is is cold. b. Any /Whatever beer there is is cold. c. *A/Some/ John's beer there is is cold. d. *Most/Many /Three/Several people there were at the party were drunk.

Nominals modified by relatives formed out of the predicate nominal are subject to an even more stringent condition: they can be

86

The English Ezistential Construction

preceded only by the, with the exception of examples such as those in (139): (138)

a. I tell this story not to the child you are now, but to the man you are going to be. (Jespersen 1927, citing H.G. Wells) b. *I tell this story not to every/whatever/any child you are now, but to the man you are going to be. c. *I tell this story not to a/some/John's child you are now, but to the man you are going to be.

(139)

a. They respected everything she was. b. all the things you are

Of course, neither restriction holds for relativization in ordinary contexts: (140)

They talked to some/many/a few people who had voted for Brown.

See Carlson 1977a for additional details concerning relativization involving the existential construction. To summarize, we have seen four sorts of cases in which the postcopular DP in the existential construction patterns with the predicate nominal to the exclusion of ordinary referential DPs in extensional contexts. Any analysis of the construction should have something to say about this. Since it is not obvious how a theory that interprets the postcopular DP as an ordinary individual or generalized quantifier over such individuals could account elegantly for this disparate range of facts, we have incentive to pursue a different line of analysis. Later we will look at these facts again, using them to compare the present proposal with analyses based on more standard assumptions about DP interpretation.

3.3

Property Theory

In this section I introduce a version of the PT1 property theory of Chierchia and Turner 1988, modifying it enough to allow for a dynamic interpretation of utterances.

Chapter 3. The Ezistential, Descriptions, and Instantiation

87

One of the principal advantages of a property theory is its richer ontology. As has been discussed in the recent linguistics literature, most notably by Chierchia 1984, natural languages arguably have the expressive powe._ to treat propositions and properties as entities, as in the following examples: (141)

a. The following proposition is true. b. Clarity is a property of any good piece of writing.

Consequently, the ontology of the interpretive theory we use should be rich enough to include them. The positing of entities corresponding to properties has been largely motivated by the desire to provide a semantics for statements involving self-predication and to resolve the associated paradoxes; however, our interest in properties will mainly concern their potential contribution to our understanding of intensionality and (non)referentiality (see Farkas 1981, Zimmermann 1992, Van Geenhoven 1992, for appeals to properties or property-like objects in the analysis of intensional phenomena).

3.3.1

The Syntax of PT

Building on the PT1 of Chierchia and Turner (hereafter, C&T), I begin by taking the ontology of the property theory to consist of four basic sorts of entities and a set of complex sorts as specified in (142), below. 7 The entities (E) are divided into two subsets, the set U of "urelements," or what I will sometimes refer to as the "ordinary" individuals, and the set NF of nominalized functions, which are the entity correlates of properties. Properties will be construed as functions from entities to entities-that is, they are not entities themselves, but rather objects of complex sorts. 8 The set U contains as a subset the set I of information units, which correspond to a pre-theoretic notion of propositions; the term "proposition" will have a technical use below. Thus note that though U consists of what I refer to as the "ordinary" individuals, it does not have to consist exclusively of entities that are "concrete": the information units, for instance, are not concrete in any intuitive sense. The sorts of PT are summarized in (142):

88

The English Ezistential Construction

(142)

a. Basic Sorts: E, the entities, U ~ E, the "urelements," I ~ U, the information units, N F ~ E, the entity correlates of properties. b. Complex Sorts: Complex Sorts.

< a,/3 >

where

a,/3

are Basic or

We may want to place other conditions on the structure of the entity domain (e.g. to allow for an analysis of plurals), but I will not do so here as it will not be necessary for our purposes. The language PT may have constants of any sort (basic or complex), but it will have variables only of the basic sorts. That lS:

(143)

a. For any sort r, Consr is a constant of sort r b. For r

= e, i, u, nf, V arr

is a variable of sort r

The meaningful expressions of PT (ME PT) will be defined recursively (where "MEa" stands for "a meaningful expression of sort a"). The base of the recursive definition appears in (i); (ii) simply states that the meaningful expression of PT are ordered with respect to each other so as to reflect the structure of the entity domain:

The next two rules ((iii) and (iv)) give the syntax of pred and ent (typographical variants of the "cup" and "cap" operators, respectively), which will be interpreted as operators relating nominalized functions to their property correlates:

m. 1ft E MEnf, pred(t) E ME tv. Iff E ME, ent(f) E MEnf

PT has rules for lambda abstraction and function-argument application; these appear in ( v) and (vi): v. If t E MEe, and :c E Vare, ~:c[t) E ME

Chapter 9.

The Ezistential, Descriptions, and Instantiation

89

vi. Iff E ME and t E MEa, f(t) E MEb PT also contains a truth predicate, t, which C&T define to turn entities of all sorts into information units. The application of the truth predicate to any sort of entity other than one that independently identifies an information unit yields, on their analysis, a necessarily false result. 9 vii. If t E MEe, then tt E MEi Application of the dagger predicate to a formula is a crucial step in making it combinable with the connectives and negation, which require expressions in MEi as input. Finally, PT has the usual logical expressions: vm. If t, t' E MEe, then (t = t') E MEi. 1x. If 1/J E MEi, then

•1/J E MEi.

x. If 1/J, ¢ E MEi, then are all in ME,.

(1/J V ¢), (1/J 1\ ¢), (1/J--+

¢),and

(1/J

+-+

¢)

x1. If 1/J, ¢ E ME,, and :z: E Vara, for any basic sort o, then 3:z:('r/J) and V:z:('r/J) are in MEi. C&T provide PT with an axiomatic theory which, among other things, is designed to avoid Russell's paradox. However, since selfpredication is not an issue of immediate concern here, I will not elaborate on this axiomatic system other than to point out that lambda conversion is assumed to hold. The reader is referred to C&T and references cited therein for further details. We may now consider how these PT-expressions are interpreted.

3.3.2

The Interpretation of PT

The C&T model for PT is a pair < F, i >, where F is a model frame and i is an interpretation function. The frame F is a tuple < 0, I, P, S, .6., T >, whose constituents are defined as follows: 0 = < E, [E --+ E], ent, pred >, where E is a set of entities, and [E --+ E] is a set of functions from E into E-the properties. Nominalized functions, are related to their {functional) property correlates by the operators ent and pred as follows:

90

The English Ezistential Construction

(144)

If a is of sort < e, e >, ent(a) 1s of sort nf, and pred(ent(a)) =a

I is an algebra of information units:

(145)

I=< I,l\,•,n,.,=:>

The basic connectives and operators include conjunction (/\,a function from [[I xI _... I]), negation ( --,, [I _... I]), and the universal operator (n, relativized to a non-complex sort r, [[E _... I] - I]). The identity relation (=:) takes two entities of any sort to yield an information unit ([[E x E] - I]). As these connectives are unremarkable, I will not discuss them here. In PT11 P is the Boolean algebra generated by a set of propositions, where the nature of a proposition is not specified; in PT2 , it is the Boolean algebra generated by the power set of a set of possible world-time pairs (P(W x J))). Chierchia and Turner then define a homomorphism T from the information unit algebra I into P (however Pis characterized):

(146)

T(i 1\ i') = T(i) 1\ T(i') T( •i) = •T( i) T(e = e') = T(e) = T(e') T(nr f)= >.e.T(f(e))

nr

I will follow PT2 here (though leaving the temporal index aside) and take P to be the Boolean algebra generated by the power set of a set of worlds P(W). Finally, the frame contains a truth operator, L\, and its correspondent on the homomorphism from I to P, namely S:

(147)

T(L\(e)) = S(e)

I will leave open the question of what exactly the value of S( e) is, as it will not be crucial in what follows. C&T relativize interpretation in a model to a choice of assignment function g. In order to make the semantics here dynamic, I will not relativize interpretation in this way but will rather take all interpretations to be functions whose domain is the set of assignment functions. Assuming an interpretation function i, we get the following interpretations to expressions of PT, where for all g: 10

Chapter 3. The Ezistential, Descriptions, and Instantiation (148)

91

i. If :z: E Varr 1 [:z:](g) = g(:z:); if a E Consr 1 [a](g) =

i(a)

ii. [ent(t)](g) = ent([t](g))

m. [pred(t)](g) = pred([t](g)) 1v. [>.:z:[t]](g) = >.e.[t](g(e/:z:))

v. [f(t)](g) Vl.

[tt](g) =

= [f](g)([t](g)) ~([t](g))

vu. [t = t'](g) = [t](g) Vlll.

=[t'](g)

[•'!fl](g) = •['!fl](g)

1x. ['¢~A .e E Er['!fl](g(e/:z:r)) This characterization of ME PT does not incorporate tense or modality. Aside from the treatment of P, the broader definition of what can be a complex sort, and the restatement of the interpretation rules in terms of functions whose domain is G, PT is exactly like C&T's PT1.

3.4 3.4.1

Interpreting the Existential A Dynamic PT Fragment of English

D P Interpretations in PT Perhaps the most important aspect of the fragment for our purposes is the availability of nominalized functions 11 as the interpretations of nonquantificational DPs. Indeed, I will take the nf-sort interpretation as basic in the sense that I will define the other possible DP interpretations in terms of it. It will also have to be possible for a DP like a child to be interpreted referentially (a term I will reserve for DPs interpreted as entities of sort u). Consequently, following Partee and Rooth 1983 and Partee 1987 (though differing in detail from these proposals),

92

The English Ezistential Construction

I take nonquantificational DPs to be associated with a family of interpretations. 12 To arrive at this family of interpretations it is first necessary to introduce some assumptions about the interpretation of nominal expressions and to define the notion of eztension of a property. The fragment of English will be interpreted indirectly, via translation into PT. In general, I will indicate the translation of English expressions whose values are constants of PT by simply bold-facing the English expression (e.g. dog for F( dog), F the translation function); this same representation will also generally stand for the interpretations of those constants (i.e. their value on the interpretation function i). Where no confusion will arise, I will usually skip explicit mention of the translation step, using e.g. [dog] for what is more precisely [ F(dog)]. Intransitive nouns and their projections are going to be translated as constants of one of the complex sorts < e, e >, < u, e >, < nf, e >, or < i, e > and are, correspondingly, interpreted as properties; the specific sort is determined by the lexical semantics of the predicate. Thus, we have:

(149)

a. F( dog) = dog b. For all g, [dog](g) = i(dog), a function in [U--+ E]

In assigning property- rather than nominalized function-sort interpretations to nominal projections (NP), 13 I follow Chierchia 1985, rather than C&T, who take bare nominals (though not DPs) to be interpreted as the entity correlates of properties, rather than as properties themselves. Although the decision in favor of properties rather than nominalized functions is perhaps not crucial in this case, it may allow us to capture certain differences between DP on the one hand, and bare nominals (and VP and AP, see below) on the other. For example, modal copular sentences such as (150)a are ambiguous with a predicate nominal, but unambiguous with an bare nominal, AP, or VP predicate: 14

Chapter 3.

(150)

The Ezistential, Descriptions, and Instantiation

93

a. I could have been a president. b. I could have been president. c. I could have been famous. d. I could have been living in the White House.

(150)a shares a reading with the rest of the examples in (150), which is true just in case there is some counterfactual possible world in which the speaker, say Joan, is a president rather than e.g. a teacher. The second reading of (150)a manifests itself in a scenario such as the following: someone arrives at the Bellevue hotel late at night, and the desk clerk is sleeping on the job. It's just the hotel manager checking up on her employees, but she admonishes the clerk by uttering (150)a. On this reading, the sentence would be true just in case there is a counterfactual world in which, rather than seeing the manager at the desk, the clerk sees George Bush. None of the other sentences in (150) (including, notably, (150)b) have this second kind of interpretation. I take this as some evidence that we want to maintain a distinction between the interpretation of DP vs. bare nominals such as president, VP, and AP. Since by hypothesis DPs are interpretable as nominalized functions, the other categories will not be so interpreted. It will be useful, in addition, to define an eztension function, ezt, for each property, viz. the set of (n-tuples of) individuals that, for any given model, world, and time, have that property. We can say that ann-tuple < z 1 , ... , Zn >is in the extension of ann-place property pn at some index (a world-time-location triple) just in case the property combined with that n-tuple, at that index, yields a true proposition (i.e. just in case that world is among those determined by homomorphism T applied to the information unit corresponding to the assertion that pn(z 1 , ... , zn) is true). That is, for all g: (151)

ezt(Pn(g)) =

{ < :llt, ... , :lln > lw

< t, l >}

~

T(6.((Pn(zt, ... , Zn))(g)) at

So imagine we want to know what individuals have the "dog" property in some world at some time/place. We apply ezt to [dog] and

94

The English Ezistential Construction

we get a set of individuals z (since [dog] will be a 1-place property) for which it is the case that the relevant world is included in the set of worlds determined by the assertion that [dog] (:r:) at < t, l > is true. I will suppress the subscript to ezt and reference spatia-temporal indices where they are not crucial. With a basic interpretation assigned to the category N and this notion of extension in hand, we may now turn to the interpretations for DP. Given that expressions projected from N are interpreted as properties, to yield an nf-sort interpretation for the DP a dol 5 we can interpret the indefinite article as the function ent, which maps properties onto their of-correlates. (152)

For all g, [(a NP]DP ](g)= [a]([NP](g)) = ent([NP](g))

That is, the DP a dog will be interpreted as the nominalized function corresponding to the property dog. This is what I will refer to as the "nf-sort" interpretation of a dog; I claim that it is the interpretation selected by the verb be both for the predicate nominal position and in the existential construction. The of-sort-interpreted DP should be understood as naming the nf, in a way similar to the way that the name Sally names an individual. However, DPs also have referential uses. Intuitively, when we use a dog referentially, we are designating an individual that has the property dog at the relevant index, that is, we are identifying an individual in the extension of dog; 16 for reasons that will become clear when we consider kind terms more closely, this individual must be of a particular sort, viz. a member of U as opposed to NF. Adapting from Heim and the DRT literature, a referential DP will be translated into PT as an variable of sort u, such that the value of that variable under an assignment function g is in the extension of the of-interpretation the DP (e:ct'([DP{+nf}])), where I take the extension of a DP to be identical to the extension of its internal nominal. (For bookkeeping purposes, I will sometimes mark DPs as (+u], mnemonic for "u-sort," or as (+nf].): (153)

If a is a DP indexed with variable :r:, then for all g, [a[+u}](g) = g(:r:u) E e:ct'([a{+nf}](g))

(154)

For all g, e:r:t'([(D NP]DP[+nf]])(g) = e:r:t([NP](g))

Thus, [a dog[+u]z](g) will be the entity g(:r:), g(:c) subject to the condition that it must be a dog.

Chapter 3.

The Existential, Descriptions, and Instantiation

95

We have assigned two interpretations to the DP a dog; definite DPs, names and pronouns, and DPs with other, not necessarily quantificational determiners (e.g. two dogs) will be introduced in subsequent sections. Once quantification is incorporated into the semantics, we will assign additional, lifted interpretations to a dog. Logical Form, Quantification, and Dynamic Interpretation The interpretation of an English sentence is going to be mediated by a logical form (lf) 17 derived from the S-Structure of the sentence; the constituents of this If are translated into PT. The purpose of If-construction in this system is largely to simplify the analysis of quantification and negation; nothing crucial depends on the adoption of this particular method of interpreting the surface syntax. As we will be adapting the dynamic approach to quantification proposed in Heim 1982, the If-construction rules largely resemble those she employs. Specifically, we posit the following (where "DP[Q]" stands for "necessarily quantificational DP" ):

(155)

Logical Form Formation Rules a. Indexation: Superscript all nonexpletive DPs with an alphabetic variable. b. Quantifier Raising (QR): Adjoin each DP[Q] to the minimal IP that most immediately dominates it, replacing the DP[Q] with a variable that exactly matches the variable-index of the DP[Q] it replaces. c. Quantifier Construal (QC): Adjoin every quantificational determiner (and no other determiner) to be a sister to the DP from which it originates.

The Indexation rule assigns each nonexpletive DP a unique alphabetic subscript (za, Ya, Za, ••. , where a: codes the sort of the semantic argument associated with the position in which the DP appears). Since the fragment for the moment will contain only indefinite and necessarily quantificational DPs, it is safe to assume that every DP gets a novel index. We can illustrate logical form construction using the sentences in (156):

96 (156)

The English Ezistential Construction

a. A dog barked. b. Every dog barked.

Assume that the verb bark18 is interpreted as a property of individuals of sort u, on the intuition that barking is a property that only certain sorts of "ordinary" individuals have; 19 its subject DP receives an index of sort u. Since that DP in (156)a is nonquantificational, QR (and, consequently, Quantifier Construal), do not apply, and the If shown in (157) is only trivially different from the S-Structure from which it is derived: (157)

IP

A D/\ A dog I VP D barked DP:c,.

I'

Now consider (156)b. Again, since the argument of [bark] is of sort u, the subject DP will receive the index :Z:u· However, since that DP is quantificational, QR will apply with the result in (158): (158)

IP

~

DP[QJ:c

A

Every dog

IP

A I' ~ "1\ VP

DP[Q]:c :z:u

I

D

barked The output of QR then undergoes Quantifier Construal, yielding (159):

Chapter 3. The Ezistential, Descriptions, and Instantiation

(159)

97

IP IP

Every

A

DP[QJ:c

~ ~ :z:u

I'

1\ I VP D barked

These lfs will be interpretable once translations for each of their parts is provided. We have already seen how nonquantificational DPs will be interpreted; verbs, like nouns, will be interpreted as properties: (160)

a. F{bark) = bark b. For all g, [bark](g) is a function in [U-+ E]

Since bark requires a referential argument, in order for A dog barks to be interpretable, we will need to make use of the referential interpretation of a dog in ( 161), rather than the nf-sort interpretation. (161)

=

F{a dog{+u]:c) :Z:ui for all g, [ent(dog)](g) = ent(dog(g)) [a dog{+u}](g) == g(:z:u) E e:z:t'([a dog{+nfJ](g))

The syntactic and semantic combination ofF{a dog) and F(barked} appear in (162)a and b, respectively: (162)

a. F(bark)(F(a dog))= bark(:z:u) b. For all g, [bark](g)([a dog{+u}](g)) = bark(g(:z:u)), where g(:z:u) E e:z:t'([a dog{+nf}](g))

In order for bark(:z:u) in (162)a to be able to serve as input to various sorts of logical predicates, we must first combine it with the information-unit-creating predicate, t. What part of an If, if anything, might contribute t, and therefore, the truth operator that is its interpretation? One possibility

98

The English Ezistential Construction

is some part of the projection of INFL. I will make the technically simplest assumption, namely that INFL projects a feature up to the IP most immediately dominating it, and that this feature is translated at the IP-level as t. Thus, a more precise representation of the lf in (157) is as in (163): (163)

IP[f]

~I'[f]

DP.,,.

D I[f]1\VP

A dog

D

barked This lf now translates as a true information unit, one which, when interpreted with respect to some assignment function, corresponds to the assertion of the truth of [A dog barked]: (164)

a. F(f)(F{A dog barked)) = thark(zu) b. For all g, [f + A dog barked](g) = ~bark(g(zu)), where g(zu) E ezt'([a dog](g))

Since [/ + A dog barked] is really a function from assignment functions to information units, it can be equivalently characterized as a set of ordered pairs of assignment functions and information units: {165)

[!+A dog barked]= { < g, i

> I~[ A

dog barked](g) = i}

Given (165), it will be completely straightforward to define a context change potential for any formula. In a dynamic system, the extension of a sentence is a function (for example, from discourse models to discourse models), rather than (for example) a truth value. As noted above, I will define a discourse model as consisting of a set D of entities which constitute its domain and a set G of world-assignment function pairs< w, g > that satisfy the utterances that make up the discourse. If we allow our assignment functions to be partial, we can define D in terms of them. Specifically:

Chapter 3.

(166)

The Existential, Descriptions, and Instantiation

99

D(c) ::: {:ciVg such that there is some< w, g >E G, g(:c) is defined}

Thus, the addition of a sentence (translated as a formula 4>) to a set of sentences, or context, c will take us from the pair < D, G >, where G satisfies c, to a new < D, G >, where G satisfies c + ¢. I will refer to this function associated with 4> as its context change potential and follow Heim in using "Sat( c + 4> )" to stand for the pair < D, G > which is the value of the context change potential of 4> applied to c. 20 We can straightforwardly define a set of world-assignment function pairs that satisfy 4> in terms of the set of assignment functioninformation unit pairs we associate with 4> via the PT homomorphism T. Recall that T maps information units into the power set of a set of worlds. The context change potential for an atomic formula 4>, then, can be defined as in (167), which is based on the definition in Heim 1982:363 (a: ranges over variables and constants of sorts i,u,nf,e): (167)

If 4> is a formula of the form t pn(o: 1 , ... , o:n ): Sat( c + 4>) ::: < D, G >, D a set of entities, G a set of ordered pairs < w, g > such that: 1. D::: D(c) U {o: 1 , ... ,o:n} 2. G = { < w, g >E Sat(c)lw ~ T(~([t/>](g)))}

(Note that G could be equivalently defined as in (168), given the definition of ezt in (151), above: (168)

G::: { < w, g >E Sat(c)i < [o: 1](g), ... , [o:n](g) >E e:ct([Pn](g)) at some

< t, l >})

Applying this to the formula derived from the lf in (157), we get (169), which says that the effect of adding A dog barb to a context c is (a) that a discourse referent :Cu will be added to the domain of the discourse and (b) that the resulting context will be satisfied by that set of world-assignment function pairs < w, g > satisfying c in which it is true that g(:cu), a dog, barked: (169)

Sat(c

+ bark(:cu)) =< D, G >,such

that:

1. D::: D(c) U {:cu} 2. G::: {< w,g >E Sat(c)ig(zu) E e:ct~([a dog{+nf}](g)) and w ~ T(~(bark(g(:cu))))}

100

The English Ezistential Construction

Now, consider again the quantified lf from (159), repeated in (170): (170)

IP Every

IP[f]

DP[Q]z

6~ _dog

~

DP[Q]z

~. :z:,.

I'[f]

1\ I[f] VP 6 barked

To interpret this lf, we first need translations for every and the material that gets left behind when Quantifier Construal applies. I will assume the following: a. F(every) =every, a constant of sort

(171)

< i, < i, i >>

b. Where the DP containing a bears the index {3, F(_a)

= t F(a)(/3)

Plugging these into our example, we get (172): (172)

every(tdog(:z:,.), tbark(:z:,.)) E MEi

The interpretation of this formula appears in (173): (173)

[every(tcf>',tcf>")] = {< g,i >I [A([every(tcf>', tcf>"J](g))]/\ [[every(t¢', tcf>")](g) = i]}

A[every(tt/>', tcf>")](g), in turn, is subject to the following condition (where an assignment function h is an extension of another assignment function g iff for all z such that g assigns a value to z, g(:z:) = h(:z:)): ( 17 4)

If 4> is a formula of the form every( ttl>', tel>"), then for all A[c/>](g) iff for all han extension of g such that [t¢'](h), there is some h' an extension of h such that [t4>"](h').

g,

Chapter 3.

The Ezistential, Descriptions, and Instantiation

101

In order to characterize the context change potential for this quantified formula, we must first define the context change potential of a complex formula consisting strictly of a string of atomic formulas; this definition, adapted from Heim 1982:363, appears in (175). It says that the context change potential of such a complex formula is arrived at by computing sequentially the context change potential of each of its atomic parts: (175)

If¢ is a formula with immediate constituents t/>1, ... , t/>n, then:

Sat(c + ¢) = < D, G >,such that: 1. D=D( ... (c+t/>l) ... +t/>n) 2. G= Sat( ... (c+t/>l) ... +t/>n) Appealing to (175), we can define Sat(c +every(¢',¢")) as in ( 176), again adapting from Heim. In prose, ( 176) says that the addition of the quantified formula to a context has no effect on the domain of the context, and that the resulting context will be satisfied by those world-assignment function pairs < w, g > for which the following is the case: for every g' that assigns the same value as g to every variable for which g( :z:) is defined, such that < w, g' > satisfies the result of adding ¢' to the context, there is some g" that assigns the same value as g' to all variables for which g'(:z:) is defined, such that < w, g" > satisfies the result of adding ¢" to the original context plus ¢': ( 176)

If ¢ is a formula of the form every(¢',¢"), then: Sat(c + ¢) = < D, G >,such that:

1. D = D(c) 2. G = { < w, g >E Sat(c)l for every < w, g' > (where g' agrees with g on every element in D(c)) in Sat(c + ¢'), there is some < w, g" > (where g" agrees with g' on every element in D(c + ¢')) in Sat((c + ¢') + ¢")}.

Thus, Sat(c + every(dog(:z:u), barked(:z:u))) is as in (177). In prose, it says that Every dog barked can be consistently added to a context c just in case, for every way one can imagine satisfying a (temporary) context consisting of the original context plus the sentence :Z:u is a dog, there is some way of satisfying that temporary context plus the sentence :Z:u barked:

102

(177)

The English Ezistential Construction 1. D = D(c) 2. G = { < w, g >E Sat(c)l for every < w, g' (where g' agrees with g on every element in D(c)) Sat(c + dog(:z:,.), there is some < w, g" >, (where agrees with g' on every element in D(c + dog(:z::r:)))

Sat((c + dog(:z:,.)) + barked(:z:,.))}.

> in g"

in

To round out the fragment, I incorporate an analysis for negated sentences such as those in (178): (178)

a. No dog barked. b. Fido didn't bark.

I follow Ladusaw 1992 in taking both clausal negative elements such as not, never, n't and negative DPs such as no dog to be subject to a condition that they be governed at lf by an expression of negation (representable as a univalued feature [NEG]). 21 On this analysis, no dog in the scope of [NEG] is assigned the same interpretation as the indefinite a dog, 22 and not, n't will appear to make no semantic contribution to the lfs in which they appear, as the semantic negation will be associated with [NEG). For example, the lf for (178)a appears in (179):

(179)

IP

A

NEG IP[f]

~

DP:r:,.

D

No dog

I'[f]

1\VP

I[f]

D

barked Though no will be subject to whatever conditions govern its cooccurrence with [NEG], it will otherwise be assigned the same interpretation as the indefinite article, viz. ent; F([NEG]) will be the standard negation predicate:

Chapter 3. The Ezistential, Descriptions, and Instantiation

(180)

103

a. F(no) = F(a) = ent

b. F([NEG]) = • Negated formulas will be interpreted as specified m PT ((148), clause (viii)): (181)

For all g, [•1/l](g) = •([,P](g))

The context change potential for a negated formula appears in (182). It says that the addition of a negated sentence to a context will have no net effect on the domain of the context and, moreover, that the set < w, g > satisfying a context plus the negated sentence will consist of just those which cannot be extended in such a way that they could satisfy the original context plus the sentence's affirmative counterpart: (182)

If¢ is a formula of the form •( ¢'), then: Sat(c + ¢) =< D, G >such that: 1. D = D(c)

2. G = { < w, g >E Sat( c)\ there is no < w, g' > (where g' agrees with g on all elements in D(c)) such that< w, g' >E Sat(c + ¢')} We now have enough of a fragment to consider the interpretation of the existential construction. Every sentence in the fragment (whose elements consist simply of an intransitive verb, an intransitive noun, a, no, and every) has been assigned an interpretation in PT as well as a context change potential. In the rest of what follows, we will be interested principally in context change potentials.

3.4.2

Interpreting Existential Sentences

Our next step is to add a lexical entry for the existential predicate, call it beexi 1 t, which has exactly one internal argument, whose semantic role might be described as "the instantiated," and no external argument. I assume that this internal argument receives inherent Case from beexi 1 t: (183)

beexi 1 t:

The English Ezistential Construction

104

By hypothesis, [be~:z:i•t] is a property of nominalized functions; therefore, its argument must be a nominalized function. I assume that a clause containing this beni•t picks up an expletive subject 23 to satisfy the syntactic requirement that it have a subject, and that the expletive does not contribute a semantic argument. As an example, consider (184): ( 184)

There is a dog.

The S-Structure for (184) in (185) has the logical form in (186): (185)

IP

A I' D/\ There I VP /\DP DP

V

1\

I

/__j

lS

(186)

a dog

IP[f]

~I'[f)

DP

D 1\ A V DP:z:,.,

There

l[f]

VP

I

lS

u1\

a dog

The verb selects for the nf-sort interpretation of a dog; as I mentioned above, [a dog] on this interpretation will be the nf that corresponds to [dog]. Using the method from the previous section, we get the translation in (187) for this lf; its interpretation is given in (187)b:

Chapter 3. The Ezistential, Descriptions, and Instantiation 105

(187)

a. tbeexist(ent(dog)) b. [tbeexist(ent(dog))] = { < g, i > ~beexist(ent(dog))(g) = i}

I

(187)b is subject to the condition in (188), which effectively states that an existential sentence will be true at some index iff its extension at that index is non-empty: (188)

For all g, ZnJ 1 ~(beexist(ZnJ ))(g) at such that y E ezt~w,t,l> (znf)


iff 3y

Existentials, like other sentences, can be assigned a a context change potential. However, this context change potential will be special in that it will involve the addition of an extra discourse referent into the domain, namely a referent corresponding to the instantiation of the nf; consequently, any set of world-assignment function pairs satisfying the context resulting from addition of an existential sentence will have to fix the value of the variable corresponding to that extra discourse referent in such a way that it is in the extension of the nf. That is, the context change potential of an existential sentence is as in (189): (189)

Sat(c + tbeexist(ent(P)) =< D, G >,such that: 1. D(c) U {z,..f, y}

=

2. G {< w,g >E Sat(c)l w ~ T(~([be~.:i&t](ent[P]))(g))

and g(y) E ezt'(ent([P])(g))}

Nothing in (189) guarantees that y be novel, that is, that y E Sat(c)lw 1

~ T(~be(ent(dog))(g))

g(y) E ezt'(ent(dog)(g))} >

and

The English Ezistential Construction

106

We may now examine the source of the existential force that is typically associated with the existential construction.

Negated Existentials and the Narrowest Scope Effect As mentioned in Chapter 1, Milsark 1977 and others have suggested that the existential construction contains as part of its interpretation a special existential operator. On Milsark's view, the definiteness effect could be attributed to a condition that this operator bind a variable corresponding to the entity whose existence is intuitively asserted; we also saw in Chapter 1, Section 3.2.1, that positing such an existential operator was problematic. In the interpretation presented here, no such operator is posited at all; nor, indeed, could it be, because the existential predicate is not even a property of the individual identified by that existentially bound variable in a Milsark-style analysis. That is, there is nothing in the translation of an existential sentence on the present analysis ((191)a) to correspond to the variable z in the traditional translation in (191)b: (191)

a. theexist(ent(dog)) b. 3z[tdog(z)]

The information embodied in the traditional translation corresponds to a rather different information unit than that embodied in ( 191 )a, one whose principal predicate expresses a property of an individual such as Wes (my friend Clare's dog), rather than a property of a kind of thing Wes is. However, (191)b does characterize a necessary and sufficient condition on the truth of (191)a (cf. (188), above). That is, we can think of the existential force Milsark was trying to capture as being hidden in the satisfaction conditions associated with the existential predicate, and as being manifested in its context change potential in the addition of the extra discourse referent y to the domain. Note crucially that (191)a itself is completely nonquantificational; this fact will account for some of peculiar scope facts associated with the construction. For example, consider the negated existential in (192); as mentioned in Section 2, the absence of a wide scope construal for the indefinite in this sort of example has lacked a good explanation: 24

Chapter 3. (192)

The Ezistential, Descriptions, and Instantiation

107

There wasn't a sound.

Since the indefinite in (192) must have narrowest scope, I will refer to this phenomenon as the "narrowest scope effect." On the interpretation proposed above for existential sentences, the narrowest scope effect ceases to be mysterious. To see why, one need only see that (192) has just one lf-not two. This lf appears in (193); its translation, in (194): (193)

IP

A IP[f)

[NEG)

~I'[f]

DP

D 1\

There

I[f]

VP

~

V

wa!n't (194)

DPz ...,

6

a sound

•(tbeexist( ent(sound)))

The context change potential for (194) is determined by the rule for negation, (182) above; applying that rule to (194) we get: (195)

Sat(c+•(tbeexist(ent(sound)))) =< D,G >such that: 1. D = D(c) 2. G = { < w, g >E Sat(c)l there is no < w, g' > (where g'

agrees with g on all elements in D(c)) such that< w, g' >E Sat(c + theexist(ent(sound)))} Spelling out Sat(c + tbeexist(ent(sound))) further, G becomes (196). It states that the satisfaction set for There wasn't a sound will be that set < w, g > for which there is no < w, g' > (meeting the usual conditions) satisfying the statement that [ent(sound)] is instantiated:

108 (196)

The English Ezistential Construction G = { < w, g >E Sat( c)\ there is no < w, g' > (where g' agrees with g on all elements in D(c)) such that [w :S; T(~beexist(ent(sound))(g')) and g'(y) E ezt' (ent( sound) (g'))]}

The existence of any particular entity instantiating [ent(sound}] would be sufficient to render the set Gin (196) empty and therefore render (192) false. We thus get a narrow scope effect; note that "effect" really is the word, since the postcopular DP in (192) does not bear any interesting scope relation at all to the negation, beyond being the argument of a predicate that is itself interpreted within the scope of the negation. This example easily generalizes to cases involving more than one quantifier; narrow thereby generalizes to narrowest. I have presented a basic interpretation and context change potential for simple affirmative and negated existentials. Its novelty lies in th'! claim that [bee:r:iat] is a property of a nominalized function; already we have seen one benefit of this claim-insight into the narrowest scope effect. One Half of the DE: A Sort Mismatch We are now ready to examine one of the two phenomena that fall under the definiteness effect, namely the exclusion of certain necessarily quantificational DPs. Note that in order to appear in the postcopular position at all, a candidate DP must be interpretable as an nf. Necessarily quantificational DPs will thus never appear in the postcopular position of the existential unless they quantify over nfs. To see this, consider (197): (197)

*There is every dog.

The logical form for (197), constructed according to the algorithm introduced in Section 4.1.:;, appears in (198):

Chapter 3.

The Existential, Descriptions, and Instantiation

109

IP

(198)

~ IP[f] DP '1!..,.,

every

D~ I'[f] _dog DP

D 1\ A

There

VP

I[f]

V

DP'I!..,.,

lS

f_j

I

1\

In order for (198) to be interpretable, the variable z must range over individuals of the same sort in the restriction (determined by _ dog) and nuclear scope (determined by There is Xnf) of the quantifier. But this is impossible since the domain of the function [dog](g) is drawn from U, while that of [beu:i•t](g) is drawn from NF. Consequently, the sentence cannot be assigned an interpretation and is predicted to be unacceptable. 25 In contrast, if a DP quantifying over entities in NF appears in the construction, we expect it to be acceptable, since it will not give rise to a sort mismatch of the kind discussed in the previous paragraph. This expectation is in fact realized.

Quantified Kind Expressions in the Existential I pointed out in Chapter 1 that previous accounts of the definiteness effect, with the exception of Lumsden 1988, have had little to say about the systematic acceptability in the existential construction of quantificational DPs whose nominal projection is headed by a kind noun 26 , such as the example in (199)b: (199)

a. There was a kind of wine that Chris disliked. b. There was every kind of local wine.

Note, that we have not only the sheer acceptability of (199)b to explain, but also the following fact: (199)a is ambiguous: on one

110

The English Ezistential Construction

reading it asserts the existence of a kind itself, with no commitment to the existence of instances; on the other, it asserts the existence of instances of a kind (e.g. bottles of wine at a contextually determined location). In contrast, (199)b is unambiguous: it can only be felicitously used to assert the existence of instances of the kinds in question, not the existence of the kinds themselves. Given the observation in the previous section that necessarily quantificational DPs are predicted to be acceptable in the existential construction when they quantify over entities in NF, an obvious explanation for (199)b presents itself: take every kind of local wine to quantify over nfs, rather than entities in U. That is, construct a translation for kind of local wine as in (200), whose interpretation is a property of nfs: 27 (200)

a. F{kind {of]} = kind[ of], a constant of sort < nf, < nf,e >> b. F(local wine)

= local wine, a constant of sort nf

c. [kind [of](local wine)]= a function in [N F-E]. This treatment of kind is very much in the spirit of Carlson's 1977b analysis and those following it (e.g. Wilkinson 1988). It has some intuitive plausibility, since kind of local wine describes entities such as Pinot Noir, Zinfandel, etc., that are themselves descriptions. Consider, then, the S-Structure for (199)b in (201): (201)

IP

A I' D/\ There I VP DP

~DP

V

I~

was~

every kind of local wine

The existential predicate requires its argument to be an nf; consequently, the index assigned to the DP will be coded for the nf sort. QR and QC will apply to yield the If in (202):

Chapter 3.

The Ezistential, Descriptions, and Instantiation

111

IP

(202) every

DP:r: ..,

~ _k. of l. wine

IP[f]

~I'[f]

DP

D /\

There

I[f]

VP

A DP:r: ..,

V

I

was

A

LJ.

;r,nf

Since, by hypothesis, kind of local wine is interpreted as a property of nfs, (202), unlike the If for There is every dog in (198), will be interpretable: the variable in both the restriction and nuclear scope of the quantifier can range over the same sort. The translation for (202) is unremarkable; it appears (simplified) in (203): (203)

every(tkind(of](wine)(;r,nf ), theexist(;r,nf ))

(203) has the context change potential in (204). (204) entails that ( 199) b will be true just in case every nf that is in the extension of kind of local wine is such that it is instantiated by some entity at the relevant index: (204)

Sat(c + every(tkind(of](wine)(;r,nf ), theexist(;r,nf ))) = < D, G >,such that: 1. D = D(c) 2. G = { < w, g >E Sat(c)i for every < w, g' > (where g' agrees with g on every element in D(c)) such that < w, g' >E Sat(c + tkind [of](wine)(;r,nf ), there is some < w, g" > (where g" agrees with g' on every element in D(c + tkind [of](wine)(;r,nf))) such that < w, g" >E Sat((c + tkind [of](wine)(;r,nf ))) + theexist(;r,nf ))}

The context change potential assigned to (199)b precludes a net increment to the domain of discourse; the introduction of the instances of each kind takes place subordinate to the quantification. 28 This is the right result, given the failure of quantified kind expressions in the existential to license discourse anaphora:

112 (205)

The English Ezistential Construction There was every kind of local wine at the festival. #It was very good./#They were reds.

Note also that, since QR must apply, (199)b can be assigned only one lf, as is consistent with its observed lack of ambiguity. In contrast, if we allow (as may be independently necessary) for a traditional interpretation of the indefinite article as an existential quantifier, alongside the interpretation advocated above (as ent), we can derive two lfs for (199)a and thus account for its two interpretations. I will provide only the lfs, rather than full interpretations, as the lfs will suffice to make the point. The lf for the quantificational interpretation of the indefinite article will look just like that for (199)b, and its interpretation will proceed in the same fashion:

IP

(206) a

DPz,.,

~

_kind of wine ...

IP[f]

~I'[f]

DP

D I[f]1\ A V DPz,.,

There

VP

I

was

I\ LJ

The only difference between this interpretation of (199)a and that of (199)b is that the former involves existential, rather than universal, quantification. Thus, the interpretation associated with the lf in (206) is the "existence of instances of a kind" interpretation. If we interpret the indefinite article as ent, the postcopular DP will be nonquantificational, and so the lf for the sentence will be as in (207):

Chapter 3.

The Ezistential, Descriptions, and Instantiation

(207)

113

IP[f]

~I'[f]

DP

D 1\

There

I[f)

VP

~

v DPZ,.J I~ was~

a kind of wine ...

On this interpretation, a kind of wine that Chris di!liked is interpreted as an nf-an nf whose extension will consist of other nfs, namely, those that are kinds of wine that Chris disliked (perhaps he doesn't like Sauvignon Blanc). This If will therefore correspond to the "existence of a kind" interpretation, since nf associated with a kind of wine that Chris disliked is the direct argument of the existential predicate. The last task before we turn to the issue of definites in the existential is to introduce additional intersective determiners into the fragment. Adding Intersective Determiners to Fragment

We know from the work of Keenan 1987 and Keenan and Stavi 1986, discussed in Chapter 1, that any DP headed by an intersective determiner is licensed in the existential construction. The examples in (208) illustrate: (208)

a. There were three bears living in the cabin. b. There were many arguments. c. There were exactly two pieces of cake remaining.

Assuming that the nouns in (208) are interpreted as properties of u-sort individuals, the DPs in (208) will have to be interpretable nonquantificationally, as nominalized functions. Two things must be added to the property interpretations of the nominal projections in the postcopular DPs in (208) (e.g.

114

The English Ezistential Construction

[bear]) to derive the nominalized functions that are the interpretations of those DPs (e.g. [three bears]): the content contributed by the determiner and the nominalizing function ent. They can be added straightforwardly if it is possible to associate each intersective determiner with a property. As mentioned in Chapter 1, it is a theorem of Keenan's intersectivity definition that:

(209)

f(A, B) =: f(A

n B, An B)

Recall that (209) tells us that in order to decide whether "Det A's are B," it is necessary to look only at the A's that are B. But if the determination of the truth of Det A 1s are B depends only on inspection of a single set (viz. An B) and not on comparison of two sets, then the truth of Det A's are B must depend on the presence or absence of some characteristic that A n B can have. That characteristic is the contribution of the intersective determiner. For example, the interpretation of the determiner two requires that, for Two A 's are B to be true, A n B must have cardinality two. But once we have identified such a characteristic (or property) of sets, it should be adaptable to the semantic framework in use here. For example, the property of having cardinality two is a property any plural entity can have, including entities of the sorts being posited here. Thus we should be able to associate some property with every intersective determiner. We need only to determine how the ent operator is contributed. There are two options: introduce ent as part of the interpretation of the intersective determiner, thus making even nonquantificational interpretations of the intersective determiners truly determinerlike, that is, semantically like a; or introduce it separately, e.g. via a null determiner, effectively reducing nonquantificational intersective determiners to adjectives. Perhaps there is empirical evidence that would point towards the superiority of one option over the other, but since the choice is not crucial for our purposes, I will simply adopt the position that intersective determiners contribute only a property and not ent. That is, I will assign the determiner two the following interpretation:

Chapter 3.

The Ezistential, Descriptions, and Instantiation

115

a. F{two) = two, a constant of sort < e, e > with a counterpart in sort , < e, e > >.

(210)

b. For all g, :z:, P, .6-two(P)(:c)(g) iff .6-(two(:z:) 1\ P(:c))(g), where for all g, .6-two(:z:)(g) iff jg(:z:)i ~ 2. Similar interpretations can be constructed for the others. The interpretation for existential sentences whose DPs are headed by such determiners will be just like those we saw earlier involving DPs headed by the indefinite determiner. In addition, I will assume that each intersective determiner can also be interpreted as a quantificational determiner in the usual fashion. For example:

(211)

Sat(c +two(¢, 1/J)) =< D, G >,such that 1. D = D(c) 2. G = {< w,g >I there are at least two g' (where g' agrees with g on all elements in D(c)) such that: < w, g' >E Sat( c + ¢) and there is at least one g" (where g" agrees with g' on all elements in D( c + ¢)) such that < w, g" >E Sat((c + ¢) + 1/J)}.

The interpretation of existential sentences whose DPs are headed by quantificationally interpreted intersective determiners will proceed, mutatis mutandis, just as did the interpretation of existentials in which the postcopular DP was headed by every; those DPs will consequently be subject to the same sortal conditions as the universally quantified DPs. Summary

To summarize the discussion so far, I have made the following claims: that the existential predicate is a predicate of nominalized functions; that DPs headed by the indefinite article and intersective determiners in general can be interpreted as nominalized functions, while necessarily quantificational DPs cannot be; and that [kind] in combination with its argument is a property of nominalized functions. Though these claims are cast in property-theoretic terms, they could be recast in some other fashion, provided that a semantic distinction is made between the argument of the existential predicate and the arguments of predicates like bark, walk, eat,

116

The English Ezistential Construction

etc.; and provided that the analysis allows for quantification over the sort of object that serves as the argument of the existential predicate. These claims have yielded two attractive results already: the contrast between the acceptability of quantified kind expressions and other necessarily quantificational DPs follows automatically, along with the failure of quantified kind existentials to license discourse anaphora; as does the narrowest scope restriction.

3.5

Definites and the Other Half of the

DE

It is now time to consider the remainder of the definiteness effect, viz. the exclusion in unmarked contexts of certain definites, names, and pronouns:

(212)

a. #There was the dinner on the table. b. #There was Frank. c. #There was him.

I begin by adding definites, names and pronouns to the fragment.

3.5.1

Augmenting the Fragment

Consider first the definites. Since we have adopted a nonquantificational analysis of indefinites, I will do the same for the definites. Specifically, the, like a, will be interpreted essentially as ent, the function that turns properties into their entity correlates; the two will differ only in that [the] will be associated with a uniqueness condition (see e.g. Prince 1981a, Kadmon 1987). (213)b introduces the translation and interpretation for the: [the] is a partial function that can be applied to a nominal projection [NP] iff under every assignment function, [NP] has a unique extension. For every argument on which [the] returns a value, that value is identical to that of ent on that argument (I use a prime (') to distinguish the interpretation of the definite determiner from that of the indefinite determiner): 29

Chapter 3.

(213)

The Ezistential, Descriptions, and Instantiation

117

a. F{the} = ent' b. [ent'] =a partial function ent' such that ent'([NP]) is defined iff for all g, le:z:t ( [NP] (g)) I = 1. For all a for which ent'(a) is defined, ent'(a) = ent(a).

Thus, when [the] is defined, [the NP] will be a nominalized function identical to [a NP]. Note that (213) requires relative, but not absolute, uniqueness. 30 To see this, consider a property such as [lid to a jar]. Lid {to} will be interpreted as a two-place property of individuals whose extension will be a set of ordered pairs of entities of sort u that stand in the "lid to" relation. When it combines with an argument, the result will be interpreted as in (214), which carries the further condition that :Z:u be in the extension of [jar]: (214)

For all g, [lid [to](:z:u)](g) = [lid [to]](g)(g(:z:u))

But given (214), [lid to a jar], unlike the other properties we have so far seen, will potentially yield a different value for every assignment function it takes as its argument. Its extension at any index (relative to any assignment function) will be the set of entities that has the property of standing in the "lid" relation to some jar (whose identity may vary with the choice of g). The definite determiner requires that for every g, that extension is a singleton set, but there is no requirement that the member of that singleton set be identical across all g. The absolute uniqueness (as I will refer to it) associated with a definite like the dog is, given (214), a consequence of the fact that [dog] is a constant function: for all g, [ dog](g) will be identical; consequently, for all g, e:z:t([dog](g)) will be identical. Of course, in addition to a uniqueness condition, definites are typically associated with a condition requiring that the referent of the definite be familiar, i.e. already in the domain of the discourse at the time of utterance. Whether this familiarity condition can or should be made to follow from other properties of definites is not clear; 31 what is clear is that the familiarity condition is associated only with those definites whose extensions are absolutely unique: for example, the lid to a jar, which does not refer absolutely uniquely, can be used to introduce a novel discourse referent. Thus,

118

The English Ezistential Construction

since all DPs headed by a definite determiner satisfy the uniqueness condition embodied in (213), while only a subset of those DPs are subject to a familiarity condition, it is reasonable to have the determiner contribute the uniqueness condition, but not a familiarity condition. For our purposes, it will suffice to characterize the familiarity condition as in (215): (215)

The extension of a DP must be familiar iff it must be absolutely unique.

At this point we should consider the implications of this characterization of familiarity for the indexation rule of our If-formation algorithm. Currently our indexation rule assigns every DP a new index; if a new vs. old index is going to correlate with the novelty vs. familiarity of a discourse referent, that will have to change. In Heim's system, all indefinite DPs received a novel index, while definites, names, and pronouns all received old indices; however, since we have given up the perfect correlation between familiarity and morphosyntactic definiteness, our indexation algorithm will have to be somewhat more complicated where definites are concerned. We have a couple of choices. If we want to keep the indexation dependent strictly on morphosyntactic properties of DP, we will need to adopt a rule such as the following: (216)

Indexation: 1. All morphologically indefinite DPs receive a new index. 2. All necessarily quantificational DPs receive a new index. 3. All names and pronouns receive an old index. 4. A morphologically definite DP receives an old index iff all of its complement DPs receive an old index; otherwise, it receives a new index.

Clause 4 in (216) applies recursively: A complement to DP will receive an old index iff all of its complements receive an old index, and so on. If, in contrast, we are willing to let indexation depend on interpretative properties of the DP, we can adopt a rule such as that in (217):

Chapter 3.

(217)

The Existential, Descriptions, and Instantiation

119

Indexation: A DP receives an old index iff its extension (given a choice of < w, t, l >) is absolutely unique; otherwise, it receives a new index.

Since (217) is simpler than (216) and, all things being equal, therefore preferable, I will adopt it. Finally, we incorporate names and pronouns. Because they pattern with definites in being licensed in the same restricted set of contexts (e.g. the "list" context), they should, like definites, not be excluded categorically from the existential construction. Consequently, they must be interpretable as nominalized functions. Like complementless definites, their nf-interpretations will have absolutely unique extensions and thus (215) will mandate that those extensions be familiar; by (217), names and pronouns will therefore bear an old index. Proper names will be translated as constants of sort nf. To capture their unique reference and rigid designation properties, we can subject them a condition such as {218)b, which requires that a proper name refer to the same unique individual across worlds and times: (218)

a. If a is a proper name, then F(a[+nf]) E Consnf b. If a

is a proper name,

then for all w, t, l,

e:z:t~w,t,l> ([a]) is a constant function whose value

is a singleton set.

Pronouns, in contrast, will be interpreted as variables of sort nf. The uniqueness condition on their extensions appears in (219)b: (219)

a. If a is a pronoun, then F(a[+nf]) E Varni b. If a is a pronoun, then for all g, ie:z:t'(a(g))l = 1.

Of course, proper names and pronouns, like other DPs, will have corresponding u-sort interpretations as well, defined as in Section 4.1.1. The interpretations we have assigned to definites, names, and pronouns are unremarkable given the treatment of indefinites we have adopted; we have preserved the essence of an analysis such as Heim's insofar as definites and indefinites are distinguished only presu pposi tionally.

120

3.5.2

The English Ezistential Construction

A Felicity Condition on Existentials

Given the analysis of definites, names, and pronouns just proposed, we know that they cannot be excluded from the existential construction for the same reason as are the necessarily quantificational DPs. What, then, excludes them? The answer I offer is the one proposed in one formulation or another by Prince 1988, Lumsden 1988, and Zucchi 1995: (220)

Felicity Condition: The discourse referent corresponding to the instantiation of the nf-argument of the existential predicate must be novel.

Because I believe that the existential predicate occurs with expletives other than beuist in English without sensitivity to definiteness (see Chapter 5), I will associate the felicity condition in (220) with the expletive there, restated as in (221): (221)

y(E ezt'([DP])) fJ. D(c) (i.e. is not familiar)

Since (221) is pragmatic in nature, we might expect in some contexts that it can be exploited for communicative purposes. I will discuss such possible instances in Chapter 5; among them is the context that licenses "list" existentials. This felicity condition is quite straightforward, but I will work through some examples to illustrate its effect. Consider There was Frank. As mentioned in the previous section, Frank is interpretable as an nf whose extension must be absolutely unique and, therefore, familiar. Thus, while [Frank] is acceptable input to the existential predicate, the attempt to instantiate it will go awry because the felicity condition in (221) requiring that the instantiation (to be drawn from ezt'([Frank])) be novel will directly conflict with the condition that the unique entity in ezt'([Frank]) be familiar. Should the condition in (221) cease to apply, we predict There was Frank to be both interpretable and potentially felicitous. Now consider the minimal pair in (222): (222)

a. There was the top to a box floating in the stream. b. #There was the top to the box floating in the stream.

Chapter 3.

The Existential, Descriptions, and Instantiation 121

The difference in the acceptability of these examples results from the fact that the top to the boz carries a familiarity condition on its extension, while the top to a boz does not. The presence vs. absence of the familiarity condition is directly related to the definiteness vs. indefiniteness of the DP complement to top [to]. Recall that if the complement to top is indefinite, the property resulting from combining it with top will vary with the choice of assignment function. Consequently, though the definite may demand a unique extension for each choice of assignment function, there is no requirement of what I have referred to as absolute uniqueness, viz. that for all g, ezt([top to a boz](g)) be identical. But since the extension of [the top to a boz] need not be absolutely unique, it need not be familiar either. The context change potential for (222)a appears in (223): (223)

< D(c) U {znf, y, z,J, { < w, g >E Sat(c)i w ~ T(.6.beexist(ent'(top [to](zu)))(g)) and g(y) E ezt'(ent'(top [to](zu))(g)) and Z14 E ezt'(ent(box))} >

For any given g, (223) will result in the introduction of a lid to some box that, while unique with respect to that box, is nonetheless novel. To see why the extension of the top to the boz must be familiar, assume [top {to]] is a function in [U --+ [U --+ E]]. It will combine with the referential (u-sort) interpretation of the boz. But since the boz is definite, the variable that is its u-sort interpretation will be subject to the condition that it be unique for any choice of g; since for all g, box(g) will return the same value, for all g, ezt(box(g)) will be the same-that is, it will be absolutely unique. In combination with the definite article, the absolute uniqueness of the extension of [the top to the boz] will follow. Given the familiarity condition adopted in the previous section, the unique member of the extension of the top to the boz will be familiar. Consequently, (222)b is unacceptable for the same reason as is There was Frank. Incidentally, the acceptability of (222)a in contrast to (224) lends support to the nonquantificational treatment of the definite determiner: (224)

*There was every top to a box floating in the stream.

The English Ezistential Construction

122

Given that [top to a boz] is a property of entities in U, (224) is bad for the same reason as is *There was every dog. Were the definite determiner to be assigned only a quantificational interpretation, we would expect (222)a to be bad for the same reason. The fact that it is not supports our decision to interpret the definite determiner as a partial function corresponding to ent. Finally, consider (225)a, which makes a slightly different point from that made by the previous examples: (225)

a. There was that kind of book. b. There was every kind of book. c. There are two kinds of books.

Like (225)b, and unlike (225)c, (225)a has only a reading that intuitively corresponds to the assertion of the existence of instances of the kinds in question (see Section 4.2.3, above); it lacks a reading corresponding to the assertion of the existence of the kinds themselves. The absence of the latter reading follows from the novelty condition in an obvious way: It could only arise if we were drawing from the extension of an nf that itself corresponded to a property of nfs-that is, if the nf being instantiated by the existential predicate was ent'([kind [of](book)]). 32 The lf corresponding to this interpretation appears in (226): (226)

IP[f]

~I'[f]

DP

D /\

There

I[f]

VP

~

V

DPz . . ,

w~s~

that kind of book

But given what we have said about the interpretation of definite determiners, ent'([kind [of](book)]) will have a unique and familiar extension; consequently, combining it with [beezi 8 t] will produce the same conflict we saw with There was Frank: The inherent

Chapter 3.

The Ezistential, Descriptions, and Instantiation

123

properties of the postcopular DP dictate that its extension is familiar, while the felicity condition associated with the existential construction require that it not be. Consequently, the If in (226) will not yield an acceptable context change potential, and the "existence of kinds" interpretation is blocked. Consider now the derivation of the reading (225)a does have. This interpretation is exactly analogous to the quantificational example in (225)b, suggesting that that is being interpreted as a quantifier in (225)a. Assume that the definite determiners (like the numerals and, as I suggested above, the indefinite article) have quantificational interpretations alongside the interpretations we have assigned. This assumption allows us to assign the If in (227) to (225)a: (227)

IP that

DP'3!.,.1

~ kind of book

IP[f)

~I'[f)

DP

D I(f)1\ ADP'3!.,.

There

VP

V

I

was

1\

1

L_j

This If will be translated into PT, interpreted, and assigned a context change potential in exactly the same way as was done earlier for lfs containing the quantifier every. The only difference will involve the particular contributions of the two quantifiers. Thus, we assign (226) the translation in (228) and the context change potential in (229): (228)

that(tkind[of](book)(xnf ), tbeexist(Xnf ))

1£4

(229)

The English Ezistential Construction Sat(c + that(tkind[of](book)(:z:nf ), tbeexist(Znf ))) < D, G >, such that: 1. D = D(c) 2. G = { < w, g >E Sat(c)i there is some< w, g' > (where g' agrees with g on every element in D(c)) such that: (a) < w, g' >E Sat(c + tkind [of](book)(:z:nf ); (b) For all < w,g" >,< w,gm >E Sat(c + tkind[of](book)(:z:n/ ), g"(:z:nf) = g111 (ZnJ) in w; and (c) there is some < w, g"" > (where g"" agrees with g 1 on every element in D(c + tkind[of](book)(:z:nf ))) such that < w, g"" >E Sat((c + tkind[of](book)(:z:n/ )) + tbeexist(Znf ))}

The fact that our analysis has forced a quantificational interpretation of that kind of book in the existential construction predicts that, as in the quantification existentials we have seen above, discourse anaphora to the instantiated entity should be blocked. The examples in (230) confirm this prediction: 33 (230)

a. There was that kind of book listed in the library card catalog. #However, it was checked out. b. There is that type of student in the class. #She does better than most. c. There was the sort of person you don't usually find at a corporate party. #But I didn't get to meet him.

The failure of discourse anaphora in the examples in (230) poses a problem for one kind of explanation that has been offered for the acceptability in existential sentences of definite kind DPs, namely that they are licensed in virtue of being "covertly" indefinite or having an interpretation equivalent to that of an indefinite (e.g. that kind of book equivalent to a book of that kind; see most recently e.g. Wilkinson 1988). For if they had an existential indefinite interpretation, we would expect discourse anaphora to be possible, as it is possible in (231):

Chapter 3.

(231)

The Ezistential, Descriptions, and Instantiation

125

a. There was a book of that kind listed in the library card catalog. However, it was checked out. b. There is a student of that type in the class. She does better than most. c. There was a person of the sort you don't usually find at a corporate party. But I didn't get to meet him.

I will return to this point in Section 6.

3.5.3

Summary

I have completed the exposition of our PT-fragment of English, incorporating an analysis of the existential construction and showing how the facts encompassed under the definiteness effect are accounted for. Definites, names, and pronouns were our final addition and were interpreted in fundamentally the same way as indefinites, differing only in that their extensions must be unique (and in most cases, familiar). We have already seen some attractive consequences of the analysis developed above; we should now subject it to a closer comparison with previous analyses. The principal focus of comparison will be on the decision to treat the argument of the existential predicate as a nominalized function as opposed to an ordinary (u-sort) entity; the discussion will take us back to the data introduced in Section 2.

3.6

Advantages of the Analysis

The difference between the analysis advocated here, which I will refer to as the property-argument analysis, and an entity-argument analysis (as I will refer to any analysis on which the existential predicate expresses a property of an ordinary entity) 34 simply involves the semantic sort attributed to the argument associated with the postcopular DP. Since we know in any case that predicates may be sensitive to the sort of their arguments (witness the "kind-level" predicates such as widespread (Carlson 1977a)), our only innovation has been to allow all nonquantificational DPs (as opposed to

126

The English Ezistential Construction

just e.g. nominalizations or mass terms, see e.g. Chierchia 1985) to be interpretable as nominalized functions. A property-argument analysis makes at least three general predictions that an entity-argument analysis does not make. First, it predicts no cross-linguistic variation in the range of necessarily quantificational DPs licensed in the construction, because the licensing and exclusion of such DPs depends on an essential semantic characteristic of the existential predicate. If there is a true counterpart to the existential predicate in other languages, we expect that predicate never to license DPs that quantify over entities in U-if it did, it would not express the property I take the existential predicate in English to express-but always to license DPs that quantify over nfs. However, since the restriction on definites in existential sentences is attributed to a relatively superficial felicity condition rather than to the nature of the property expressed by existential predicate, we might expect cross-linguistic variation in the range of the definite and indefinite DPs licensed in the construction-that is, we might expect the felicity condition associated with the construction to vary. Moreover, any such variation should be sensitive to independently motivated discourse-functional classifications of nonquantificational DPs. For example, we might find a language which, rather than requiring the instantiated referent contributed by the utterance of an existential sentence to be hearer new (which is effectively what I have claimed; cf. Prince 1992 and Chapter 1), requires it merely to be discourse new; 35 or, we might find a language in which there is no definiteness effect at all-only a "quantification effect." The predictions of an entity-argument analysis concerning crosslinguistic variation in the range of DPs licensed in the construction will depend on the account of the DE associated with that analysis. However, in general (and as reflected in previous works) such analyses have nothing to gain by positing a nonunified account of the DE. For what would such an account be like? To rule out definites, something like the novelty condition adopted above would be necessary; but to rule out the right subclass of quantificational DPs on grounds meaningfully different from those ruling out definites, some stipulation would have to be made, for a condition on DP[Q)s such as (232) (the distillation of Lumsden's proposal suggested in Chapter 1) encompasses the definites in any case, undercutting the

Chapter 3. The Ezistential, Descriptions, and Instantiation 127 point of a nonunified account: 36 (232)

A DP is licensed in the existential construction iff it can be used to introduce a (persistent or temporary) hearer new discourse referent.

To avoid undercutting a nonunified account of the DE in this way, the only choice is a completely unmotivated condition such as: "necessarily quantificational DPs are excluded unless their nominal is a kind expression." Consequently, advocates of an entityargument analysis will try (and have tried) to capture the exclusion of definites and the relevant class of quantificational DPs under a single generalization. However, any analysis positing a unified generalization covering the DE makes rather different predictions about cross-linguistic variation; in particular, it offers no reason to expect asymmetric degrees of variation in the licensing of quantificational vs. nonquantificational DPs. So what kind of cross-linguistic differences does a generalization like (232) lead us to expect? Since (232) is a felicity condition on use, the claim underlying it (in conjunction with the assumption that the argument of the existential predicate is of sort u) is that the DE is entirely a discourse-functional phenomenon. Ultimately, a purely discourse-functional account of the DE predicts not only the kind of variation predicted by the property-argument analysis (see the preceding paragraphs), but also variation in the licensing of necessarily quantificational DPs. The reason it predicts the latter sort of variation is that quantificational DPs are not naturally covered by discourse-functional classifications such as "introduces a hearer new discourse referent"; consequently, languages are likely to classify DP[Q]s arbitrarily with respect to discourse-functional categories. Thus, the analysis advocated in this chapter, with a nonunified account of the DE, predicts a more limited range of cross-linguistic variation than does an entity-argument analysis with a unified account of the DE cast as a felicity condition; and it is superior to an entity- argument analysis with a nonunified account of the DE in that it need not separately stipulate the condition governing the distribution of necessarily quantificational DPs. In addition, the property-argument analysis, unlike such an entity-argument analysis, predicts that, should there be any context in English in which

128

The English Ezistential Construction

the novelty condition can be defeated, only definites and not quantificational DPs, should be affected. This second prediction is realized. In the next Chapter, I will offer evidence that the appearance of definites in list existentials such as (233)b exemplifies the possibility of violating the existential construction's novelty condition (Hannay 1985, Lumsden 1988, and Abbott 1992 have all suggested a discourse-functional explanation of this sort for list existentials). Notice that the necessarily quantificational DPs generally excluded from the existential construction do not sound any better in list contexts (e.g. (233)c,d): (233)

a. Who can help us? b. Well, there's the plummer in that shop. c. *Well, there's each plummer in that shop. d. *Well, there's every plummer in that shop.

The contrast between (233)b vs. c,d must be stipulated on a unified analysis of the DE. The prediction concerning cross-linguistic variation remains to be tested. But looking at just one language that manifests a weaker DE in the existential construction than English, namely Catalan, we find nothing to contradict the prediction of the analyses we have advocated. Proper names and definites are quite acceptable in Catalan existential sentences in contexts where they would not be allowed in English, such in (234) (note that the question is understood to be rhetorical, and both question and answer are uttered by the same speaker; compare the oddness of the English translation): (234)

Vaig anar a la festa i saps qui hi havia? AUX-1sg go to the party and know-2sg who there-was Hi havia la Joana! there-was the Joan I went to the party and do you know who there was? There was Joan!

In contrast, DPs quantifying over u-sort entities are bad ( (235) ), even though quantified kind expressions are licensed ((236)): 37

Chapter 3.

The Ezistential, Descriptions, and Instantiation

(235)

*Hi havia cada un dels cotxes a la cursa. there-was each one of-the cars at the race.

(236)

Hi havia tota classe de cotxes a Ia cursa. there-was every class of cars at the race. There was every type of car at the race.

129

This array of facts has no straightforward explanation if there is a single, nondisjunctive felicity condition associated with the postcopular position in existential sentences. Though much more cross-linguistic investigation needs to be done, we see preliminary evidence from the distribution of DPs in list existentials supporting a nonunified account of the DE and, along with it, the general type of analysis developed in this chapter. The second prediction made by a property-argument analysis, but not an entity-argument analysis, is that any linguistic generalizations sensitive to the semantic sort or referentiality of an argument will likely distinguish the argument of the existential predicate from the argument of a verb such as bark. In contrast, if we adopt an entity-argument analysis, we expect, if anything, that linguistic generalizations sensitive to the semantic sort of an argument or its referentiality will group them together. One such generalization will be discussed shortly; we will again see that the facts support the property-argument analysis. Third, the property-argument analysis predicts there should be a correlation between the licensing of necessarily quantificational DPs in the existential construction and the behavior of the DP with respect to "sort-sensitive" generalizations, because the licensing of those DPs is determined by the sort of entity over which they quantify. For example, suppose, contrary to the expectation raised by this analysis, we do find one or more languages that license quantificational DPs such as each dog in their equivalent of the existential construction. In any such languages we should find that any phenomena that target u-sort (as opposed to nf-sort) arguments should target the argument of that existential predicate as well. An entity-argument analysis of the construction makes no such prediction. These, then, are some general differences between the two kinds of analyses. I now turn to the data introduced in Section 2. We will

130

The English Ezistential Construction

see that, while an entity-argument analysis can account successfully for some of the data, the property-argument analysis accounts more successfully for more of the existential construction's peculiarities.

3.6.1

Acceptability of Quantified Kind DPs

The first special property of the construction noted in Section 2 was the contrast in acceptability of quantified kind DPs and other necessarily quantificational DPs. In order to account for the acceptability of quantified kind expressions under an entity-argument analysis of the construction, two strategies can (and have been) be pursued: ( 1) find a way to analyze quantified kind terms as covert indefinites of some sort (cf. e.g. remarks in Milsark and Wilkinson 1988); or (2), assuming a treatment of the DE that is entirely grounded in presupposition, find a way to distinguish the presuppositions associated with quantified kind terms vs. other quantificational DPs (cf. Lumsden 1988 and the discussion in Chapter 1). The latter strategy is reasonably successful; the statement in (237), a version of which was introduced in Chapter 1, will do the job: (237)

A DP is licensed in the existential iff it can be used to introduce a (persistent or temporary) hearer new discourse referent.

Assuming that all quantificational DPs carry the presupposition that their domains are non-null, the temporary discourse referents they introduce will not generally correspond to hearer new individuals; consequently, such DPs will not be licensed in the existential construction. The exceptions are the quantified kind terms, which can introduce hearer new referents corresponding to instances of those kinds. In contrast, the covert indefinite strategy faces empirical problems, one of which was alluded to at the end of Section 5.2. As mentioned there, Wilkinson 1988, 1991 hypothesizes that DPs of the form Det kind of X are licensed in the existential construction because they have an existential indefinite interpretation equivalent to that of a(n) X of Det kind. 38 But we have already seen facts such as the contrast in (238), which shows that Det kind of X cannot have an existential indefinite reading, since it is unable

Chapter 3.

The Ezistential, Descriptions, and Instantiation

131

to license discourse anaphora to instances of the kind, suggesting that, when Det is definite or quantificational, it must be interpreted quantificationally: 39 (238)

a. There is that type of student in the class. #She does better than most. b. There is a student of that type in the class. She does better than most.

One might try to account for the data in (238), while maintaining that the argument of the existential predicate is of sort u, by adopting the hypothesis of Heim 1987 that the existential construction is subject to the following condition at the relevant level of logical representation: (239)

*There be z, where z is an individual variable.

Heim takes (239) to follow from what is essentially a presuppositionbased account of the DE; in particular, it is clear from her discussion (especially her footnote 4) that {239) is meant to exclude pronouns and wh-traces and is orthogonal to the theory of indefinites as "variables" developed in her dissertation. 40 Heim points out that, given the right assumptions, (239) correctly blocks binding of the full postcopular position by a quantifier (e.g. after QR has applied) but does not itself block binding of some part of the postcopular position. She then suggests that some cases of apparent full binding of the postcopular position, such as in the questions in (240)a and b, can be accounted for by positing reconstruction to derive interpretations as in (240)c,d, respectively (where questions are interpreted as sets of propositions): {240)

a. What is there in Austin? b. How many people were there at the party? c. {p: There is a kind z such that p is true and p is the proposition that there are (is) things (a thing/stuff) of kind z in Austin} d. {p: There is a degree d such that p is true and p is the proposition that there were d-many people at the party}

132

The English Ezistential Construction

Although Heim does not discuss definite/quantified kind terms, the interpretation she gives for (240)c suggests strongly that the licensing of those kind expressions in the existential construction should fall under the same generalization. Combining her proposal with some version of Wilkinson's, the sentence in (241)a could be assigned the interpretation in (241 )b, paraphrasable as (241 )c: (241)

a. There was that kind of wine.

b. (THAT:z:: kind of wine(:z:)) (3y[wine(y) 1\ kind(y, x)] c. That kind z of wine is such that there was wine of kind z. The fact that the quantifier denoted by That kind takes scope over the variable z in the quantifier denoted by wine of kind z in (241 )b could account for the failure of discourse anaphora in (238)a and like examples. The idea underlying this analysis is essentially the same as that underlying the analysis proposed in this dissertation. But its implementation raises some questions. First, why should that have to be interpreted quantificationally in {241 )b? This has no explanation on the reconstruction analysis, though we saw in Section 5.2 that it does have an explanation on the property-argument analysis. A second question is more general. What conditions govern the reconstruction needed to derive e.g. (241)b above? Are they the same as those governing other putative cases where reconstruction is posited? To answer these questions, we will need to consider some of the scope and quantification facts to be discussed in the next section. To summarize, an entity-argument analysis of the existential predicate can account for the distribution of definite and quantified kind expressions without trouble if they are licensed in virtue of their presuppositional properties, but we have seen some reason to think that a "covert indefinite" analysis of their distribution will run into difficulties. However, the former type of entity-argument analysis offers no insight into the similarity between the argument of the existential predicate and the predicate nominal with respect to the distribution of quantificational DPs, since the predicate nominal is not a referring expression. In contrast, as I have shown

Chapter 3.

The Ezistential, Descriptions, and Instantiation

133

above, the similarities between the two sorts of arguments follow directly if we take them both to be nominalized functions.

3.6.2

Quantification and Scope

In Section 2.2, I observed that though the postcopular DP, like the predicate nominal but unlike DPs in other argument positions, generally manifests a narrowest scope effect (compare (242)a vs. b), this effect goes away when the postcopular position is filled by a quantified kind DP ((242)c): (242)

a. Andy didn't invite some student to the party. b. *There wasn't some student at the meeting. c. The food critic was annoyed because there wasn't some variety of wine on the list.

On the property-argument analysis, this array of data is fully expected: The fact that some must be interpreted wide with respect to negation entails that whenever it heads the postcopular DP in a negated existential, that DP will have to undergo QR-that is, that it will have to be interpreted quantificationally. Since this analysis prohibits from the existential construction DPs that quantify over u-sort entities, (242)b is bad; (242)c is acceptable because its postcopular DP quantifies over an acceptable sort. And since the DP in (242)c can be interpreted quantificationally, we expect it to be able to interact scopally with other material in its clause. An entity-argument analysis that posits (237) has nothing to say about these facts, since there is no requirement that DPs introducing hearer new referents be interpreted as if they had narrow scope with respect to other operators in the sentence. In contrast, though the reconstruction analysis does predict the narrowest scope effect, it is less clear that it accounts successfully for the scopal properties of quantified kind expressions. On this analysis, the narrowest scope effect results from the fact that the postcopular DP must in general be interpreted fully in situ: if QR takes place with no reconstruction, a variable is left behind in the postcopular position, in violation of (239), and the requisite kind of partial reconstruction is not an option for DPs such as some student. Thus, the only possibility is interpretation in situ, which

The English Ezistential Construction

134

will either yield consistently narrowest scope for the DP or, in the case of (242)b, correctly result in an uninterpretable formula. However, it is unclear that the reconstruction analysis should predict sentences such as (242)c to have an interpretation on which the DP does not receive narrowest scope. To see why, consider the following schematic post-reconstruction representation of the relevant portion: (243)

(Some z: z is a variety of wine) -, [there wasn't wine of variety z available]

In the scope configuration corresponding to (243), negation intervenes between the quantifier contributed from within the DP and the DP itself. But Roberts 1987, citing earlier work by Larson, observes that when a DP-internal quantifier takes scope over the DP containing it (that is, on the inverse linking reading), it must have immediate scope over that DP; another quantifier (such as the negation in (243)) cannot intervene. For example, consider (244): (244)

Two politicians spied on someone from every city.

Of 6 logical possibilities, (244) can have only the scope configurations in (245)a; those in (245)b, in which two takes scope between the universal and existential operators contributed by the object DP, are impossible: (245)

a. V32, 2V3, 3V2, 23V b. *32V, *V23

In light of this immediate scope constraint, the interpretation assigned to (242)c is surprising; we might have expected the sentence to lack this interpretation, since some does not take immediate scope over the DP from which it originates. Perhaps we could try to derive the exceptional relaxing of the immediate scope restriction from the fact that reconstruction is being posited. But since reconstruction is an operation that can apply to quantificational DPs in all kinds of positions, we would have no explanation for the fact that partial reconstruction cannot provide a sentence like (244) with interpretations that violate the immediate scope restrictionunless there was something fundamentally different between a DP like a senator from every state (or an animal of every kind) and one like every kind of animal. 41

Chapter 3. The Ezistential, Descriptions, and Instantiation

3.6.3

135

Contact Clauses

The contact clause facts introduced in Section 2.3 exemplify a phenomenon that distinguishes the argument of the existential predicate from typical (extensional) u-sort arguments. Recall that contact clauses are restricted to certain copular and existential contexts such as those in (246), along with modal and intensional contexts ((247)a,b); they are prohibited from typical extensional contexts ((247)c,d): (246)

a. There's a student in my class went to America. b. This is the girl wants to see you. c. Here's the boy'll fix it for you. d. I have a brother works in Dublin. e. I'm the only one knows how to act.

(247)

a. I'd like to marry a man always pays his debts. b. I wouldn't hit a fella said that. c. *I married a man always pays his debts. d. *I didn't hit a fell a said that.

What do the copular/existential and intensional contexts have in common? On the basis of these data, Doherty 1993 proposes that contact clauses are possible only in contexts where the DP may be construed "nonreferentially." Recasting this hypothesis more precisely in the terms developed in this chapter, we can say: 42 (248)

A DP modified by a contact clause cannot have a u-sort interpretation when evaluated with respect to the actual world.

That is, a DP modified by a contact clause cannot be referential, as the term is defined in Chapter 1. This condition allows DPs containing contact clauses to appear in nf-sort argument positions when those positions are interpreted with respect to the actual world (e.g. in the existential or copular construction), and in usort positions when interpreted with respect to nonactual worlds.

The English Ezistential Construction

136

The fact that DPs modified by contact clauses sometimes receive a generic or kind interpretation (Doherty (p.c.)) lends further support to this proposal. Neither variety of entity-argument analysis we have considered has anything to say about the distribution of contact clauses. Their distribution cannot be related to the felicity condition associated with the use of the construction, since no similar felicity condition holds for argument positions within the scope of modals or intensional predicates to the exclusion of extensional argument positions; nor does their distribution correlate in any obvious with the condition that the postcopular position may not contain a variable at the relevant level of representation. The reader may object that this is not a fair criticism of these analyses. Could we not propose that existential be is a kind of intensional predicate and capture the distribution of the contact clauses while preserving the essence of the entity- argument analysis? For example, suppose we took the existential predicate to express a property of an intensional generalized quantifier, that is, a function from possible worlds to generalized quantifiers, building on Montague's proposal that the predicate nominal (in the relevant cases) and arguments of intensional predicates have such an interpretation. 43 We would then have the semantic distinctions necessary to account for the contact clause facts. But this change in the analysis is an admission that there is something special about the interpretation of the postcopular DP that goes beyond the novelty condition. And one of the points of this chapter is to argue that, once this admission is made, it is worth trying to make as many properties of the postcopular DP as possible follow from its basic interpretation. I have shown one way of imputing a kind of intensionality to the postcopular position that allows for a much simpler statement of the definiteness restriction than either of the alternative accounts, one which may extend more successfully to other languages as well.

3.6.4

Relativization

Finally, consider again the relativization facts introduced in Section 2.4. First, when the postcopular DP is relativized out of the existential or copular constructions, many speakers will accept only

Chapter 3. The Ezistential, Descriptions, and Instantiation

137

that or 0 as the relative pronoun: (249)

a. %The people who there were at the party were drunk. b. The people (that) there were at the party were drunk. c. *They dressed like the eccentric women who they were. d. They dressed like the eccentric women (that) they were.

On the analysis we have developed, the selective distribution of the relative pronoun reduces to a case of sensitivity to semantic sort: we can say that just as relative pronouns must match the case marking that would be assigned to a constituent in the gap position, they must be able to match the semantic sort associated with the gap position. If who can identify only entities of sort u, then it should be prohibited from marking the postcopular gaps in (249). An entity-argument analysis with the presupposition-based account of the DE represented in (237) has nothing to say about the distribution of relative pronouns. Consider an example like (250): {250)

A doctor to whom I spoke yesterday recommended surgery.

There is no reason to suppose that either the doctor identified in (250) or the fact that the speaker spoke to anyone at any time are familiar to the hearer of the sentence. Consequently, we have no reason to suppose that there is any connection between the impossibility of who in (249)a and the condition licensing DPs in the postcopular position. Thus, the analysis fails to shed light on yet another similarity between the existential and copular constructions. In contrast, relative pronoun distribution could be accounted for on the reconstruction analysis with the condition that who be licensed only when the gap in the relative clause is filled by an individual variable at LF. Such a condition would plausibly extend to the predicate nominal case as well. The second peculiarity of relativization out of the postcopular position is the limited range of determiners that can cooccur with the noun modified by the relativization:

138 (251)

The English Ezistential Construction

a. All the /What /The people there were at the party were drunk. b. *Some people there were at the party were drunk.

These facts are less simple to explain. Carlson 1977b claimed that the only kind of relativization possible out of the existential construction was what he called "amount" relativization, a special form of restrictive relativization. Amount relatives are so called because the DPs in which they appear denote quantities (or degrees) or else quantify over quantities or degrees. For example, the DP in (252)a can be paraphrased as the amount of champagne they spilled that evening, and as such does not require identity of champagne, but rather only identity of quantity, as Heim 1987 (38) observed: (252)

a. It will take us the rest of our lives to drink the champagne they spilled that evening. (Heim 1987) b. Max put everything he could in his pocket. (Carlson 1977b)

Similarly, the preferred interpretation of (252)b can be represented along the lines of (253)a, as opposed to (253)b: (253)

a. (MAX d: Max could put d-many things in his pocket) [Max put d-many things in his pocket] b. (Vz: z is a thing that Max could put m his pocket)[Max put z in his pocket]

These examples indicate that one of the defining properties of amount relatives is that the head of the relative contributes a different kind of condition on the referent of the DP as a whole than does the relative clause. Intuitively, the relative clause provides a measure or quantity while the head contributes information about what is being measured, or in some cases, the unit of measure (as in: It would take me a year to earn the dollars Kent owes the bank.). This intuition apparently lay behind Carlson's decision to treat the determiner+relative clause as a complex quantificational determiner taking the head noun as its argument.

Chapter 3.

The Ezistential, Descriptions, and Instantiation

139

As with relative pronoun distribution, an entity-argument analysis with a presupposition-based account of the DE has nothing to say about this constraint. However, Heim 1987 addresses these facts directly. Taking up Carlson's analysis, Heim suggests that amount relatives are licensed in the existential construction because, at the relevant level of representation, the postcopular position in a relativized existential is filled by a reconstructed expression such as d-many books; for example, the relative clause in (254)a (adapted from Safir 1982) would have a representation like (254)b: (254)

a. The books there were on his shelf b. (THE d: There were d-many books on his shelf)

The idea is that amount relativization should involve binding of a variable over amounts or degrees, as opposed to a variable over individuals. We saw in previous sections that Heim posits the condition in (239), repeated in (255): (255)

*There be z, where z is an individual variable.

Since (254)b does not violate the constraint in (255), the amount relative is licensed. In contrast, in order for an example like (254)a to have an ordinary restrictive relative interpretation, it would have to have a representation such as (256): (256)

(THE x: :z: is a book and there was :z: on his shelf)

Since this representation violates (255), the restrictive relative interpretation is blocked. However, this line of analysis results in insufficiently strong satisfaction conditions for sentences containing relatives formed from existentials. Given the representation in (254)b, the truth of the sentence in which an amount relative appears should depend only on the quantity identified by the DP, and not on the particular individuals or matter identified by that DP. Consequently, if it is the case that relative clauses formed from existentials can only have the type of representation shown in (254)b, the truth of any sentence containing a relative clause constructed from an existential sentence should only ever depend on a quantity of individuals

The English Ezistential Construction or matter, and never on the identity of particular individuals or matter. But this is not the case. The truth of (257)a depends on the speaker's having talked to a set of particular people, and not merely to a particular cardinality of people. That is, it does not have the interpretation schematized in (257)b: (257)

a. I talked to the people there were at that party. b. (MAX d: There were drmany people at the party) [I talked to drmany people at the party]

Though (257)a entails (257)b, satisfaction of the weaker truth conditions associated with (257)b is not sufficient to guarantee the truth of (257)a. This indicates that the DP containing the relative clause in (257) is not interpreted simply as quantifying over degrees and, therefore, that the amount relative analysis cannot stand as is. But if the interpretation of relatives formed from existential sentences does not in general follow from the reconstruction-based analysis, we lose one of the more interesting pieces of motivation for that analysis. Though the relativization facts also fail to follow in any direct way from the interpretation proposed in this chapter for the existential construction, we can consider how the property-argument analysis might illuminate them. Clearly there is a similarity between the relativization of quantity expressions and the relativization of the postcopular DP in existential sentences that led Carlson to give them similar analyses in the first place. Notably, a relativized quantity expression must be headed by a determiner with universal force, as the contrast between (258)b and c shows: (258)

a. The cargo weighed 5 tons. b. The tons the cargo weighs make little difference. c. *Some ton/two tons/most tons the cargo weighs make little difference.

We might begin to get at the source of the similarity by asking what the relata of the predicate weigh are. Does weigh express a relation between an entity and a quantity, or a relation between an entity and a quantity description?

Chapter 3.

The Ezistential, Descriptions, and Instantiation

141

Suppose, for the moment, that weigh expresses a relation between an entity and a quantity description. This relation must be mediated by another entity not explicitly a part of the argument structure of the predicate, namely the actual weight of the cargo. The weight of the cargo (at any given world-time-location index) is presumably determined by a weight-measure function, call it W. The weigh relation between some entity z and some quantity description Q is mediated by W ( z) in that the weigh relation will hold iff W(z) is in the extension of Q (as stated in (259)a), where W( z) is in the extension of Q iff the value of W( z) on the measure function corresponding to the unit constituent of Q is equal to the numeric constituent of Q ((259)b): (259)

a. For all z, Q, weighs(z, Q) iff W(:z:) E e:z:t(Q) b. For all :z:, n a number, U a unit, W(:z:) E e:z:t([n(U)]Q) iff U(W(:z:)) = n

The weigh relation needs to be mediated in this way for two reasons: (1) despite the fact that for any z, there will be exactly one value for W( :z:) at any given index, the weigh relation may hold between :z: and any number of quantity descriptions at the same time; and (2) the quantity descriptions to which :z: bears the weigh relation at any given time are systematically related. That is, our cargo may weigh five tons, but if it weighs five tons, it also weighs 10000 pounds, 160000 ounces, some number of kilograms, etc. Since there is exactly one value for W(:z:), it will be describable by exactly one quantity description for any given unit in which the W ( :z:) is measured. 44 This fact could give the beginning of an explanation for the contrast between (258)b and c. The DP the tons the cargo weighs intuitively identifies the value for the weight function on the cargo, measured in ton-units: (260)

a. Aw[w E e:z:t([tons]) 1\ W([the cargo])= w] b. For all :z:, U a unit, W(:z:) E e:z:t(U) iff 3!1n[U(W(:z:)) n]

=

If we take (260)a to be a reasonable representation of the interpretation of the nominal tons the cargo weighs, then the fact that the value of W on the cargo is necessarily unique will require that

142

The English Existential Construction

it combine with a determiner entailing uniqueness or maximality. Observe that other nominals characterizing unique or maximal individuals sound anomalous with determiners that do not guarantee uniqueness or (where relevant) maximality: (261)

a. *Some weather of the day was lovely. b. The weather of the day was lovely. c. *Some grandparents of Sally raised her well. d. The grandparents of Sally raised her well.

An analysis similar to that in (260)a could be assigned to a DP containing a relative clause built from an existential sentence if we could motivate some function whose value was the (perhaps plural) individual introduced via that sentence. An example of such a function would be one that gives a "population" value for a given index in terms of the description associated with the postcopular DP. That is, we might assign a relative clause such as The people there were at the party the representation in (262): 45 (262)

Au[u E ext([person]) 1\ Pop(< w, t, l >) = u]

Though the representations in (260)b and (262) leave much to be explained, they suggest what is similar to DPs modified by relative clauses headed by expressions of quantity-units and those modified by relative clauses built from the existential construction: in each case, the DP as a whole expresses the value of a function: the weight-function in the former case, and the "population" function in the latter. Moreover, the head of the relative clause in each case provides the description by which the function value is identified. The distinguishing feature of these relatives is that the gap does not transparently identify the sort of entity in the extension of the DP containing the relative clause: the sort of the gap in each case is an abstract object, that is, a description, while the DP as a whole determines a group of entities of a different sort, viz. the sort in the extension of the sort corresponding to the gap. Finally, since the "population" function posited for the existential construction returns a (perhaps plural) individual and not an amount, the truth of a sentence such as (257)a will depend on the identity of those individuals, and not merely on their quantity. 46

Chapter 3. The Ezistential, Descriptions, and Instantiation

3.6.5

143

Summary

In the preceding subsections I have compared the analysis of the existential construction developed in this chapter to analyses that take the argument of the existential predicate to be an "ordinary" entity, and pointed out some of its advantages. Specifically, the property-argument analysis, as I have called it, both makes stronger predictions about cross-linguistic variation in the range of DPs found in the existential construction and stands a chance of accounting for a wider range of phenomena than do what I have called entity-argument analyses. The discussion is in many ways incomplete. At a descriptive level, I have not considered, for example, what other sorts of extraction facts might reveal about the construction, nor have I looked much beyond English. At a theoretical level, I have not talked about the possible costs of allowing nonquantificational DPs quite generally to have both nf- and u-sort interpretations. Clearly, an important issue concerns the implications of this proposal for the analysis of natural kind expressions and genericity, since the range of DPs that give rise to natural kind and generic readings is only a very small subset of those that I assign nf-interpretations. It is clear that what is called for is a better understanding of the contribution made by the cardinal and other intersective determiners, whose nature remains somewhat opaque and which I have treated with a certain amount of equivocation. Finally, comparison of the existential predicate with intensional predicates remains largely a promissory note. 47 Nonetheless, the present proposal is explicit enough to allow for serious consideration of these issues in the future. The next task is to provide a semantics for the optional predicative phrase. This is the topic of Chapter 4.

144

The English Ezistential Construction

Notes 1 Let

me emphasize that I am not claiming that the argument of the existential predicate is identical to a Carlsonian kind. The postcopular DP in the existential construction can take many forms that cannot be used to express natural kinds; cf. the anomaly of (i), for example: (i) *Two cats are widespread. When I use the phrase "kind expression" or "kind DP" throughout this chapter, it should be understood to refer only to certain bare plurals and those expressions containing a noun such as kind in the relevant position. Nonetheless, the special behavior of kind expressions in existential sentences clearly points to a similarity between the Carlsonian kind and the argument of the existential predicate, though the exact nature of that similarity remains to be explored. 2 Technically, however, what I call a context change potential will differ somewhat from Heim, as will become clear below. See e.g. Kamp 1981 and Groenendijk and Stockhof 1991 for other implementations of dynamic interpretation. 3 Whether it is interesting or useful to extend the analogy between copular and existential sentences to cases involving definite descriptions is not entirely clear at first glance. Definite descriptions are licensed in predicate nominal position, though they generally result in a "specificational" interpretation: (i) Fred is the doctor. Definite DPs in the existential construction will be discussed in Chapter 5. 4 0n a GB analysis, some functional node (e.g. TenseP) would presumably have to dominate VP in order for tense to be licensed. Since the syntactic analysis of these clauses is not our concern here, I will stick with the more descriptive label. See Doherty 1993 for discussion. 5 Specifically, contact clauses are excluded from predicational copular sentences: (i) *Bill is a man owns one. In addition to the context mentioned in the text, contact clauses show up frequently in it-clefts, for example: (ii) It's Bill owns one. I will have nothing to say about it-clefts here.

Chapter 3. 6 There

The Ezistential, Descriptions, and Instantiation

14 5

are a couple of exceptions to this prohibition, viz. DPs containing contact clauses have been attested as the complements to the verbs know and meet. I do not have an explanation for why these verbs are exceptional. Note, incidentally, that (134)c, below, confirms that the nouncontact clause string does form a constituent: since does that is not the sort of unit that could intervene between subject and verb, it must be a part of the subject DP. 7 The definition of complex sorts is taken from C&T's PT2. 8 There is variation in the literature as to whether properties themselves, rather than or in addition to information units (or propositions), should be treated as primitives. I will not address this issue here. See Chierchia et al. 1989 for various perspectives. 9 I will leave this characterization of the truth predicate as is in the interest of remaining as faithful as possible to C&T's system; it seems more appropriate for natural language applications to redefine t so as to combine only with information units. 10 [a] stands for "the interpretation of a." It is assumed to be relativizable to different models. g( ef :r,) is a function just like g except perhaps in that g(ef:r.)(:r.) =e. 11 More precisely, functions from assignment functions to nominalized functions, but I will generally leave out reference to assignment functions when talking about interpretations in the text. 12 I will further assume below that definite and indefinite DPs also have quantificational interpretations; I take these to be due to ambiguity in the interpretation of the determiner, i.e. that the articles may also be interpreted as existential quantifiers (the definite carrying an additional uniqueness condition). Note, however, that necessarily quantificational DPs like each student will have only one interpretation-the quantificational one. See McNally 1995a for further justification for these assumptions. 13 Following Abney 1987, I take NP to be the complement to D. In systems that do not posit DP, what I refer to as NP is equivalent toN'. 14 I am grateful to Geoff Nunberg for bringing this contrast to my attention. 15 For now I will consider only the indefinite article. 16 With other conditions applying, e.g. that the individual is designated as novel to the discourse model.

The English Ezistential Construction 17 Which

should not be identified with the GB stratum of LF. this discussion I will abstract away from tense; therefore, [bark] will be taken as equivalent to [barked]. 19 I will ignore the fact that properties such as barking can systematically be construed as generic properties of kinds of things on the basis of generalization from instances of the kinds. See Carlson 1977b. 20 Note, however, that for Heim a context change potential consists of a domain and a set of sequences, not a set of world-assignment function pairs. 21 The precise nature of the relation between the appearance of [NEG) and a negative element or DP in an If will not concern us here, and no attempt will be made to account for sentences containing more than one negative element. See Ladusaw 1992 for further discussion of these issues. 22 No dog may also have a necessarily quantificational interpretation, but that will not concern us here. 23 Not necessarily exclusively there; see Chapter 5. 24 Though see Heim 1987. I will address her suggestion concerning this fact in Section 6. 25 Note that whether or not a DP such as every dog can be taken to quantify over the appropriate sort of entity is partly a function of the context. That is, if we can construct a context in which it is possible to interpret every dog as meaning every kind of dog, then it should be possible to use the DP in an existential sentence (see below). See Carlson 1977b, Chapter 6, for discussion of this issue. 26 Where by "kind noun" I intend nouns like kind, sort, type, flavor, variety, etc. 271 have not attempted to account for the limited range of nominals that appear in the complement position to kind [of}, in particular, singular bare nouns (see Carlson 1977b for examples and discussion), since the peculiarities of this internal argument will not be relevant in what follows. 28 See e.g. Roberts 1987 for a variety of examples and for discussion of the relation between quantification and discourse anaphora. 29 (213) abstracts away from such well-known exceptions to uniqueness such as the leg of a table. 30 In this sense it closely resembles the characterization of uniqueness proposed in Kadmon 1987. 18 Throughout

Chapter 3. 31 See

The

E~istential,

Descriptions, and Instantiation

147

Kadmon 1987:293ff. and Roberts 1993 for pertinent discussion of the relation between familiarity and uniqueness. 32 For the sake of discussion, I assign that the same interpretation as the. 33 However, discourse anaphora may be licensed by the nf argument itself, e.g. (i) There were those kinds of beer at the bar. And we tried them all. 34 Although technically, on the analysis presented here, the existential predicate takes an entity as its argument (specifically, the entity correlate of a property), the term "property-argument analysis" will serve as a useful way of describing the essence of the proposal. 35 According to Prince 1992, proper names, but not e.g. pronouns, can identify a discourse new referent, viz. cne that may be in the common ground of the discourse, but not salient. 36 The same can be said for a Keenan-style account of the DE (viz. "a DP will be licensed in the existential construction iff its determiner is existential"). 37 The contrast between (235) and (236) does not depend on the simple difference between the determiners cada un and tota. I have used different determiners here in order to obtain the most natural sounding DPs in each sentence, but both are necessarily quantificational. I am grateful to J osep M. Fontana for discussion of these examples. 38 Wilkinson 1991 does not explicitly suggest that this equivalence holds for any determiners other than the definites; however, for the sake of argument I will take the proposal to extend to all determiners. 39 Similar problems arise for this proposal in the other two contexts where it has been claimed that DPs such as that kind of animal can have an existential indefinite interpretation, viz. in what could be termed "vanilla" stage-level argument positions (such as in (i), the example is Wilkinson's) and in the scope of an adverb of quantification (as in (ii)): (i) That kind of book is on the shelf. (ii) That kind of animal usually eats oats. Like the DP in (238)a, (i) does not license discourse anaphora to instances of the book (see (iii)), even though its putative coun-

148

The English Ezistential Construction

terpart (in (iii)) does: (iii) #It has a broken spine. (iv) A book of that kind is on the shelf. It has a broken spine. Similarly, if the kind DP has an existential indefinite reading, the unavailability of wide scope for that interpretation of that kind of animal in (ii) goes unexplained (compare ( v), which is ambiguous): (v) An animal of that kind usually ea.ts oats. 40 To emphasize this point, she assigns indefinites generalized quantifier-type denotations that are interpreted in situ, rather than raised. 41 The fragment presented in this chapter could be extended with minor modifications to include inverse linking. For example, Quantifier Raising could be reformulated so that DP-internal quantificational expressions adjoin to DP rather than IP, and so that DPs to which quantifiers have been adjoined obligatorily undergo QR; Quantifier Construal could be reworked to allow determiners in those Df'-adjoined expressions to take scope beyond DP (see e.g. Barker 1991 for a similar proposal involving possessive DPs). With these revisions, the lf for e.g. (i) would be as in (ii): (i) There was a top to every bottle. (ii) [rP every [DP.,,. 1 [DP!h.- bottle] [DP.,,., a top to Yu]] [I P[f] There was Znf ]] The only question raised by (ii) concerns the translation of the string a top to Yu, which corresponds to a function from entities to nfs. Since the phrase contributes to the restriction on the quantifier and as such must contribute information conjoinable with the formula that is the translation of _ bottle, a plausible candidate for its translation is (iii): (iii) Znf = >.zu(ent(top(to](zu))](yu) The translation for the entire sentence appears in (iv ): (iv) every(tbottle(yu) 1\ Znf = >.zu[ent(top[to](zu))](Yu), tbeexist(Znf )) The interpretation of (iv) is straightforward and the reader can verify that it matches intuition. 42 (248) is not complete in that it does not account for the impossibility of contact clauses in predicational copular clauses, but this does not affect the present point. 43 Pace the criticisms that have been leveled against this analysis

Chapter 3.

The Ezistential, Descriptions, and Instantiation

14 9

of the former. 44 Abstracting away from the fact that if the cargo weighs five tons, by entailment it (that is, some portion of it) also weighs e.g. four tons. 45 Similarly, if we take have to express the possession function from the entity identified by the subject of have to the individuals instantiating the post-have DP, we can explain the contrast between (i) and (ii) (cf. Carlson 1977a) in the same way as we explain the existential facts in the text: (i) *a daughter he has (ii) the daughters he has 46 This discussion of relativization would have been strengthened by consideration of the additional extraction facts discussed in Postal 1992 and his analysis of them. Unfortunately, I received that paper too late to include examination of its findings. 47 See Zimmermann 1992, Moltmann 1995 for recent work in this area.

Chapter 4

Adjunct Predicates and the Predicate Restriction 4.1

Introduction

This chapter is concerned with the rest of the semantics of the existential construction. We found motivation in Chapter 2 for assigning the XP to the same syntactic position as a VP-adjunct such as the depictive, rather than treating it as a complement of be. We might expect, then, that whatever interpretation rule governs the semantic contribution of such adjuncts will cover the existential's XP as well. Since it has been independently argued (e.g. in Rapoport 1991) that circumstantial and depictive adjuncts manifest the same kind of predicate restriction we have seen in existential sentences, a very attractive consequence of assimilating the XP to such adjuncts is that the existence of the predicate restriction in the existential will follow from its membership in that adjunct class. However, in order for this account to work, two issues must be addressed. First, since the interpretation of the XP's controller in the existential construction (viz. the postcopular DP) is different than that of the usual depictive controller-it is interpreted as a 151

152

The English Ezistential Construction

nominalized function rather than an ordinary entity-, we have to look carefully for any effects this difference might have and ensure that they are not problematic. Second, Milsark used examples such as (263)a to argue for an independent XP in the existential, but depictive and circumstantial adjuncts cannot have an eventive interpretation, as (263)b,c show; these latter two sentences express impossible states of affairs in which some pig is simultaneously live and roasted: (263)

a. There has been a live pig roasted. b. We have eaten a live pig roasted. c. A live pig lay there roasted.

This contrast must be explained, and the eventive interpretation of (263)a must be accounted for. The rest of this chapter is organized as follows: I begin by presenting a semantics for predicative VP-adjuncts of the sort we have been concerned with. I then extend this analysis to the existential construction. Once it has been shown that the proposed analysis works for adjectival and present participle XPs, I turn to existentials with past participle XPs. I resolve the contrast in (263) by arguing that Milsark incorrectly concluded that a live pig roasted must be treated as a DP-XP sequence; I argue instead that it is a single DP.

4.2

Interpreting Depictive Adjuncts

The literature on depictives (inter alia, Halliday 1967, Green 1973, Jackendoff 1990 and Rapoport 1991) is quite consistent in its characterization of the adjunct: the depictive contributes a property that the interpretation of its controller must have while participating in the state of affairs described by the main predicate. That is, the addition ofthe depictive in (264) contributes the condition that the tea Margaret is drinking must be cold while she is drinking it on the occasion in question: (264)

Margaret is drinking the tea cold.

Chapter

4. Adjunct Predicates and the Predicate Restriction 153

This condition contrasts with (and in some sense is simpler than) the semantic contribution of other kinds of predicative VP-adjuncts, such as the resultative or the purpose clause. There are at least two ways one could set up the semantics for the depictive. One way would be to take some notion of state of affairs or event as primitive, redefine predicative expressions to be interpreted as properties of states of affairs (or relations between individuals and states of affairs), and assign the depictive a semantics in which it functions as a secondary predicate on the state of affairs variable which is the target of the main predication: (265)

3e[drink(m, t, e)

A

cold(t, e)]

Jackendoff's 1990 treatment of depictives is in the spirit of this approach. A slightly different position would be to claim that the depictive constrains the spatia-temporal parameters over which the state of affairs associated with main predicate holds. This latter position would keep the tea drinking event and the tea being cold (in (264)) logically independent; however, since the same tea must simultaneously have the property of being in a cold state and being drunk, the net result is quite similar to (265 ). Brief remarks in Gawron 1986a suggest this view . 1 I will adopt the latter of these approaches because thinking about states of affairs and property ascription relations in terms of the intervals over which they hold will be more helpful when we examine the predicate restriction associated with the depictive. Consequently, we need to add a set of intervals, T and a set of locations L, along with the partial orders ~T and ~L to our model. The partial order is taken to give each of these domains the structure of a complete join semilattice (see e.g. Link 1987, Lasersohn 1988 for similar proposals). We will also want to adopt a standard overlap relation over objects in each of these domains, as in (266), defined for elements of L (A stands here for the meet relation): 2 (266)

For any l, l' E L, lol' iff there is some l" such that lAl' = l"

The overlap definition for intervals is exactly analogous. We can now augment our system to include the temporal interval and location over which properties hold of entities. These

154

The English Ezistential Construction

intervals will be the values of the functions int and lac, respectively. Int ("hold time") and lac ("location") are defined on ordered pairs of 1-place propositional functions and entities, that is, they are functions of type [[E-+ E] -+ [E-+ T]] and [[E-+ E] -+ [E-+ L]], respectively. :3 A word on how int and lac should be understood: Although I will not modify the system adopted in the previous chapter beyond the addition of these two functions, the notion of a delimitable state of affairs is obviously implicit in the thinking behind these functions, and they should return values that match one's intuitions about what constitutes the relevant boundaries for the hold time and location of such states of affairs. For instance, int should be understood as picking out a maximal interval for atelic states of affairs, and a minimal interval for telic states of affairs: if Martha runs for a total of 20 minutes, it is necessarily true that she runs for 10 minutes, but we want int to return the value 20 minutes for (run, m). If she runs from her house to my house in 20 minutes, it is also true that she does so in 30 minutes; however, the relevant interval is again the former, and not the latter. The value for lac should depend on similar criteria: we want its value to be no bigger or smaller than necessary. For instance, the location of Martha's running from her house to my house should include a region of space large enough to include her house, my house, and the path between-no more, no less. See Lasersohn 1988:100ff. for related comments.

4.2.1

The Adjunct Rule

With these modifications, we can now propose a depictive rule and incorporate it into our semantics. The sentence in (267) will illustrate; I will consider only its object-controlled reading: (267)

Fido ate the meat raw.

(267) has the S-structure in (268), where [+PRED) indicates a predicative XP:

Chapter

(268)

4. Adjunct Predicates and the Predicate Restriction 155 IP

A

DP

I'

D IAVP

Fido

I V'

~ AP[+PRED]

V'

~

V

j

DP raw

1/\.

ate~

some meat

The interpretation of this structure requires the designation of a controller for the adjunct predicate; Williams 1980 shows that the requisite controller is what can for present purposes be referred to as the "theme" argument of the main predicate. 4 There is no controlled PRO in (268); I will take the control relation to be represented in the If of a sentence via coindexation of the adjunct with the designated controller, according to the following rule: 5 (269)

Controller-Controllee Coindexation: If XP[+PRED] in the configuration: VP

~

V ... XP(+PRED] is not an argument of V, then label XP[+PRED] with the index of the theme argument of [V]. This rule, combined with the other If-construction rules from Chapter 3, yields the lf in (270):

156

The English Ezistential Construction

(270)

IP[f]

~I'[f]

D

DP,

Fido

1\VP

I[f]

I V'

~

V'

AP[+PRED] 11

~DP raw ~

V

11

lA

ate~

some meat

We need to augment our interpretation rules from the previous chapter to allow for the incorporation of the information added by the adjunct. I will use the following: (271)

Adjunct Rule: [[V~ XP[+PRED],Jv;] is a function f of the same sort as [Vi], such that for all y E e;z;t ([V~]), y E e;z;t (f) iff int([V~], y) ~T int([XP[+PRED]], ;z;i) and there is some l such that loc([V~], y) 1\ loc([XP[+PRED]], ;z;i) = l.

The rule in (271) treats the predicative VP-adjunct essentially as a function from properties to properties that serves to constrain the interval over which the property identified by the main predicate holds. It further requires that there be some overlap between the spatial regions in which the main and adjunct properties hold. As an example, consider (270). [ate] (I abstract away from tense) will combine with [the meat], and the result is a 1-place propositional function whose extension at some index is a set of entities that ate the meat. The combination of this propositional function with the function identified by the adjunct yields a new 1-place propositional function [ate some meat raw], whose extension will be a subset of the extension of [ate some meat]: the subset of meat-eaters that ate some meat in question when it was raw. If the only meat Fido ate was cooked, Fido will not be

Chapter

4.

Adjunct Predicates and the Predicate Restriction 157

in the extension of ate some meat raw, since, for the relevant choice of meat, int([ate some meat], [Fido]) is not a subinterval of int([raw], [some meat]). What will the values for int(f, y) and loc(f, y) be for a complex propositional function /formed via the Adjunct Rule? Given what the rule says, int(f, y) will never be larger than the interval over which the ascription relation holds between the adjunct and its controller, and it can conceivably be less than that interval: The meat Fido ate could have been raw long before he ate it (and, should any of it remain when Fido is done, it will surely still be raw). Consequently, it is reasonable to suppose that int(f, y) will be the meet of int([V'], y) and the interval [XP[+PRED]] holds of its controller. Things will be slightly different for loc(f, y) because the Adjunct Rule places a weaker condition on spatial regions, 6 requiring that merely an overlap relation, rather than an inclusion relation, hold between the location associated with the main property ascription and that associated with the adjunct ascription. However, we want to preserve the intuition that the location associated with the complex property ascription should be no larger than necessary. Thus, we may take loc(f, y) to be the join of loc([V'], y) and the location over which [XP[+PRED]] holds of its controller. This concludes the exposition of the basic depictive Adjunct Rule. I now turn to the issue of the individual/stage sensitivity of the adjunct.

4.2.2

Individual/Stage Sensitivity

Both subject- and object/theme-controlled VP-adjuncts appear to manifest a sensitivity to individual-level predicates quite similar to that found in the existential construction. The examples in (272) and (273) are anomalous: (272)

a. *Margaret is drinking her tea green. b. *Sue played the piano a grand. c. *Fido is eating a shoe leather.

158

(273)

The English Ezistential Construction a. *Jay built the house naive. b. *Kent was cooking dinner a student of Hegel. c. *Mary was flying the plane altruistic.

Rapoport's 1991 analysis of this contrast builds on the assumption that individual- and stage-level predicates differ in that the latter, but not the former, have an event argument in their semantic representation. 7 She then adopts the licensing principle in (274): (274)

Licensing Principle (Rapoport 1991:170): Every phrase in a syntactic structure must be licensed through the direct linking of a position in its theta-structure [= argument structure + event structure, LMcN .] to a position in the theta-structure of the head of its clause, within the government domain of that head. From this principle it follows that only the combination of a stagelevel main predicate with a stage-level depictive predicate will be licensed: the licensing of the adjunct predicate depends on its having a position in its event structure that can link to some position in the main predicate's event structure; and that main predicate must in turn have a position to which the adjunct predicate can link an argument. Rapoport proposes that an event argument is the necessary mediator of this linking: without an event argument in both the main and adjunct predicates' theta-structures, the appropriate linking will not be possible. While the intuition behind this proposal is one I share, its specific formulation can be improved upon once we take a closer look at the facts. Rapoport takes the prohibition on individuallevel adjunct predicates to be absolute. However, it is not. First, a typically individual-level main predicate can sometimes co-occur with a stage-level adjunct predicate: (275)

a. Sam disliked Paul angry. b. Dan never believed Mary drunk.

On the intended readings, Sam disliked Paul when Paul was angry (either on a particular occasion, or in general); Dan never believed Mary when she was drunk. Observe that these sentences are interpreted in such a way that the main predicate holds of its subject

Chapter

4. Adjunct Predicates and the Predicate Restriction 159

argument only during the time when the adjunct predicate holds of the object argument. In effect, the presence of a stage-level adjunct induces "stage-level-like" behavior in the main predicate. Second, individual-level predicates do show up as controlled adjuncts under certain circumstances. Subject-controlled individuallevel predicates of various sorts co-occur principally (though not exclusively) with main predicates associated with states of affairs classifiable as achievements (i.e. telic and punctual; Dowty 1979): 8

(276)

a. Yesterday, Matt went to work nihilistic; today, he is a changed person. b. Today, Joe is sitting in church a convert to CatholiClsm.

c. Martha went to Europe naive and excited and came back a jaded pseudo-intellectual. For reasons I do not understand, similar sentences with non-subject controllers sound much less natural:

(277)

a. ?Five years ago they came upon Fido scrawny and illtempered, but they quickly turned him into a prizewinning show dog. b. ?We got to know Max young and naive, but ended our friendship once he became jaded and cynical.

The issue of available controllers is orthogonal to the present point; only the following two observations are relevant. First, the adjuncts in these sentences make the same kind of semantic contribution as do the other adjuncts we have seen above: it is entailed that the controller have the property identified by the adjunct while participating in the state of affairs associated with the main predicate. Second, in addition to this simultaneity, these sentences strongly implicate that the controller either has undergone a change of character shortly before the hold time of the main ascription begins, or is about to undergo such a change soon after the hold time culminates. Observe further that individual-level predicates are never modified by individual level predicative adjuncts:

160

(278)

The English Ezistential Construction

a. *Everyone admired the rescuers courageous. b. *Felix knows methods effective.

When individual-level secondary predicates appear in VP with individual-level main predicates (as in (279)), they can do so only as (parts of) complements to those predicates, as can be seen from at least two facts. (279)

a. Sara believes Ann sincere. b. Betty doesn't like Fred so naive.

First, subject control in the examples in (279) is absolutely impossible; object control is obligatory. Since, this is exactly the opposite of what we just saw above, positing adjuncthood in (279) would render such examples exceptions to the rule that if anything in a given sentence can be an adjunct predicate controller, it is the subject. Second, the main predicates of the acceptable examples in (279) differ in meaning depending on the presence vs. absence of the secondary predicate, as can be seen by considering certain entailments of these sentences. For example, (279)a does not entail that Sara believes Ann; rather, it means something paraphrasable as "She believes that Ann is sincere." Similarly, (279)b doesn't entail that Betty doesn't like Fred (though in fact she may not like him)-its truth only depends on her not liking a character trait of his. Recall that adjuncts, whether individual- or stage-level, do not affect the interpretation of the main predicate in this way: (276)a, for example, still entails that Max went to work. Rapoport's analysis could be augmented to account for the first kind of exception to her generalization, viz. the stage-level behavior of individual-level main predicates, by positing a second, event argument-containing, lexical entry for each of those predicates; however, the second sort of exception is a real problem. The adjunct predicates in (276) do not have a stage-level flavor, so positing an event argument in their theta-structure fails to preserve the standard intuition about the information that the event argument is supposed to encode; moreover, we are left with no insight into the implicature associated with those examples. Clearly, the way we choose to characterize the individual/stagelevel distinction is the crucial ingredient in a successful account of

Chapter

4.

Adjunct Predicates and the Predicate Restriction 161

these exceptions. I will draw on the proposal in Condoravdi 1992a that the distinction rides on the presence vs. absence of an inference of temporal persistence both into the past and into the future. Individual-level properties carry an inference that they hold continuously of the entities to which they are ascribed (and generally over some extended period of time), unless information is provided to defeat that inference; that is, int( P, :z: ), where Pis an individuallevel property, will be assumed to be unbounded. Stage-level properties carry no such inference. For example, if someone is a child or is naive, it is a fact about childhood and naivete that that person cannot alternately have and not have the property of being a child or na1ve. 9 What about the location ( loc) value for individual- and stagelevel predicates? Intuitively, an individual-level predicate holds of that individual wherever s/he happens to be: If Mother Teresa is altruistic, she is going to have that property whether she is in Calcutta or California. In this sense, any location external to the individual is irrelevant, which is surely the source of the frequent claim (e.g. Carlson 1977b, Gawron 1986b, Kratzer 1989) that individuallevel predicates are "unlocated." But arguably certain predicates commonly classified as stage-level can have this characteristic as well. If I get sick with the flu, I can carry the flu around with me for quite awhile. However, the difference between a stage-level property like sickness and an individual-level property like altruism is that it is not necessary that a property like sickness "follow around" the individual to whom it is ascribed, precisely because sickness is not a property guaranteed to hold of me or anyone else for any length of time. In contrast, it is inconsistent with the inference of temporal persistence carried by individual-level predicates that they not "follow around" their bearers. Of course, certain stage-level predicates are necessarily fixed in location: If I am sitting at the computer, as soon as I leave I'm no longer going to have the property of sitting at the computer. Consequently, I suggest the following: For certain predicates P 1 (all of which will come from the standard stage-level category), loc(P 1 , :z:) will be a simple value from L, the set of locations. However, for the remainder of the stage-level and all of the individuallevel predi :ates, loc( P 1 , :z:) will return another function, call it zloc, whose value when applied to an entity-interval pair gives the

162

The English Ezistential Construction

location of that entity at that interval: (280)

z-loc(z, t) = l

z-loc will help capture the fact that some properties "follow around" their bearers, without making it impossible to assign, at any given time, some real region of space at which an individual has a property such as sickness or altruism. Note that it is indeed possible to assign a real spatial parameter to an individual-level property ascription at some interval: It is not necessarily untrue or anomalous to say that Mother Teresa is altruistic in Calcutta-it's just rather strange, altruism being the kind of property one is likely to have irrespective of where one is. But generally, if the value of the location function associated with the individual-level property ascription is itself a function (viz. z-loc) whose value depends on a choice of temporal interval, while ascription of the property generally does not depend on or make reference to any particular choice of interval, we get the effect that the property holds of its bearer at no location in particular or, more precisely, wherever its bearer happens to be at the time. The power to assign a specific location to an individual-level property ascription at some interval (in terms of z-loc) is crucial in order for those predicates to be licensed by the Adjunct Rule, since the rule demands that there be some spatial overlap between the loc value of the main predicate-argument pair and that of the secondary predicate-argument pair (an example will be discussed shortly). This view of the individual/stage distinction is slightly different from the recently prevalent view that the distinction rides on the absence vs. presence of an event or spatia-temporal argument in the argument structure or semantic representation of the predicate (see Kratzer 1989, Diesing 1992, Parsons 1990, inter alia). The two views differ principally in that the approach advocated here attributes a spatia-temporal parameter to all predicates, individuallevel and stage-level alike; however, for most purposes, the spatial and temporal parameters associated with individual-level predicates is simply irrelevant. In contrast, the event-argument view holds that individual-level properties are essentially unlocated in space-time and that, should spatio-temporal parameters be assigned to an individual-level property, it will "turn into" a stage-

Chapter

4. Adjunct Predicates and the Predicate Restriction 163

level property. There are notable advantages to appealing to a temporal persistence inference rather than the presence or absence of an event argument. 1° First, it obviates the need to appeal to lexical ambiguity in order to account for the fact that sometimes individual-level predicates behave as if they were stage-level; moreover, it makes predictions about the contexts in which we are likely to find this behavior, and which predicates will allow it. The lexical ambiguity analysis of stage-individual alternations makes no such predictions without ancillary assumptions. For example, locative predicates such as in the room are paradigm examples of stage-level predicates. Like stage-level predicates, locatives license existential readings for bare plural subjects in the copular construction (Carlson 1977a), and they license universal quantification in when-clauses (Kratzer 1989): (281)

a. People were in the room. b. When Mark is in the room, he makes a lot of noise.

However, there is a class of nouns for which locative predicates behave as if they were individual-level. These are nouns such as dent, hole, space, which are interpreted as relations between some entity and a location (Kuno 1972, Kimball1973, McNally 1992a). These license neither existential readings for bare plurals nor sensibly interpretable universal quantification in a when clause: (282)

a. #Holes were in the wall. b. #When the hole is in the wall, it is easy to look through.

I have suggested elsewhere (McNally 1992a) that these facts are explainable if locative predicates express individual-level properties for the entities identified by this class of nouns. Holes, dents, and spaces are special in that they are defined by their location: A dent that is in my car door cannot be moved to the rear fender. This kind of relativization of the notion "individual-level" is understandable, and indeed expected, when the notion is characterized in terms of a temporal persistence inference. In contrast, taking the individual/stage distinction to consist in the presence vs. absence of an event or spatio-temporal variable leaves us with the

164

The English Ezistential Construction

task of explaining why a non-specific indefinite DP such as dents could not be the argument of the stage-level version of in the car door. Though one could develop such an explanation while maintaining the ambiguity, it is difficult to see what the appeal to an event argument contrast really buys when it is possible to account for facts such as these without it. Another advantage to characterizing the individual/stage distinction in terms of a temporal persistence inference is the insight it offers into the exceptional licensing of individual-level predicative adjuncts introduced in (276) and their prohibition in examples like (278). If those individual-level predicates lacking an inherent temporal dimension license an inference of unbounded temporal persistence both back into the past and forward into the future, they will not be very useful as temporal modifiers, that is, as restrictors of the interval over which the principal property ascription associated with the clause holds. However, we predict that if it is possible to establish such a relevant connection, the predicate should be a licit adjunct. This is what happens in the examples in (276). I mentioned above that sentences such as these (another example is provided in (283)) carry the implicature either that the adjunct property has begun to hold of the referent of its controller within some relevant interval just prior to the hold time of the principal ascription (which is what happens in (283)), or else that it will cease to hold shortly thereafter: (283)

After a year of working with Mother Teresa, the teenager came home generous and altruistic.

The relevant connection is established in this case by defeating in one direction the inference of temporal persistence for the individuallevel property, rendering the hold-time of the property bounded at one end (though leaving it potentially unbounded at the other): we infer that the teenager was neither generous nor altruistic before his experience with Mother Teresa. Observe that this really is just an inference: nothing in (283) entails that there is any particular relation between the coming home and the onset of generosity and altruism, other than that the interval over which the coming home takes place must be contained within the interval that the generosity and altruism hold of the teenager. Nor is there any causal

Chapter

4.

Adjunct Predicates and the Predicate Restriction 165

connection between the two: the most likely cause of the onset of generosity and altruism is not the coming home, but rather some part of the experience for which the coming home marks an end point. That is, the interpretation for the main clause in (283) is just like what we saw for the example worked through in the previous section. Consider its If, in (284) (I leave out the coordination for simplicity): (284)

IP(f]

~I'[f]

DPz

~

The teenager

1\VP

I[f]

I V'

~ AP(+PRED]z

V'

A~

came home generous According to the Adjunct Rule, the ascription of the property identified by came home generous to the entity identified by the teenager will be subject to the following condition: 11 (285)

[the teenager] E ezt([came home generous] iff int( [came home], [the teenager]) :ST int( [generous], [the teenager]) and there is some l such that l = loc( [came home], [the teenager]) 1\ loc( [generous], [the teenager])

In order to determine whether the second clause of the biconditional in (285) is satisfied, a real-space value for the loc function on [generous] and [the teenager]) is going to have to be fixed via the function z-loc; consequently, we have to choose an appropriate interval as input. It is reasonable to suppose that this interval is going to be the interval during which the teenager came home. The value that will become the value for loc([generous],[the teenager]) at the relevant interval will be z-loc(int([came home], [the teenager]), [the teenager]), which will be a subpart of the interval during which

166

The English Ezistential Construction

the teenager came home, and the condition that there be some overlap between the location parameter of the main and adjunct predications will be satisfied. So the interpretation for (283) is just the same as for a sentence like Fido ate some meat raw (in fact, raw is even the kind of adjective whose location parameter should have been treated as we have treated that of generous here). The only difference is that the inference of temporal persistence associated with the individuallevel predicate gives rise to certain implicatures in order to make sensible the use of that predicate as a temporal modifier. In the case of (283), the fact that the teenager's coming home in all likelihood marks the endpoint of the experience that induces the new personality traits renders it a natural and useful a reference point for the onset of the hold-time of these traits, because the coming home presumably takes place at a relevantly minimal interval after that onset; and we therefore infer that the teenager's coming home marks a point after which we may consider him to be generous and altruistic. This pragmatic sort of explanation could extend to the failure of individual-level secondary predicates to modify individual-level main predicates, observed in (278). Here we are faced with the task of using one temporally persistent property ascription to restrict the interval over which another holds. But if both ascriptions are taken to persist indefinitely into the past and future, there would not appear to be any useful sense in which the ascription relation associated with the adjunct could provide a temporal restriction on that associated with the main predicate, nor will the property associated with the main predicate have the aspectual characteristics that facilitated the interpretation of (283). Finally, an individual/stage distinction grounded in the presence vs. absence of temporal persistence accounts naturally for the coercion of certain individual-level main predicates into stage-levellike behavior. If the interval over which the adjunct predicate holds of its controller determines the bounds within which the principal property ascription must hold, the boundedness of the interval associated with the adjunct predication will obviously defeat any inference of temporal persistence associated with the property denoted by the main predicate.

Chapter

4. Adjunct Predicates and the Predicate Restriction 167

With an interpretation rule for VPs containing depictive adjuncts and a more fully-developed notion of the individual/stage contrast, we can now return to the existential construction. Our first task is to work through any consequences the special semantic properties of the postcopular DP might have for the depictive adjunct interpretation rule. We will then examine in detail the interpretation of "eventive" existentials with past participle XPs.

4.3

Extending the Analysis

4.3.1

Nominalized Functions as Controllers

Consider the following existential with a final predicative phrase: (286)

There was a dog barking.

Based on the discussion in Chapter 2, this sentence has the DStructure in (286): (287)

IP

A

DP

I'

There

I VP

D. A

I V'

~VP

V'

ADPD barking

V

w~s

D

a dog

Applying the Controller-Controllee Coindexation rule in (269) along with the other relevant lf construction rules, we derive (288):

168

The English Ezistential Construction

(288)

IP[f]

~I'(f]

DP

D

/\

There

I[f]

VP

I V'

~ VP:z:

V'

A D DP:z: barking

V

w~

D

a dog

Notice that the coindexation rule has assigned the index of postcopular DP to the adjunct. Recall from Chapter 3 that, as a complement to the existential be~:z:i 1 t, this index corresponds to a nominalized function, not an individual in U. However, the correct interpretation of this If depends on the adjunct predicate holding of the (singular or plural) entity instantiated by the existential. Consequently, we need to be careful about how the effect of control is worked out. In addition, we must work around the fact that, even though we have a VP-internal controller, that controller is the argument of a monovalent predicate on the analysis in Chapter 3. Thus, if we keep our adjunct rule the same, the interpretation of the XP will have to combine with that of be~:z:i•t before the latter combines with the interpretation of the postcopular DP. Given the syntactic structure we have adopted, this order of composition will not be strictly compositional. There are a variety of ways to work around this problem. For example, we could revise the adjunct rule, allowing the adjunct to combine with fully saturated properties (i.e. objects of sort e that become information units in combination with the t predicate) in addition to unsaturated properties. Another option would be to recharacterize bee:z:i 1 t as a bivalent predicate whose interpretation is stipulated to be identical to that we have been assuming (cf. Dowty 1985). However, since it is not crucial for

Chapter

4. Adjunct Predicate8 and the Predicate Re8triction 169

our immediate purposes which of these solutions (or any other) is adopted, I will simply ignore this problem and assume that the adjunct predicate can combine with verb before the verb combines with the its DP complement, as doing so will not require us to adjust any existing rules or definitions. According to the Adjunct Rule, the result of combining [bee:ci 1 t] with the interpretation of the XP barking is going to be a 1-place propositional function whose extension is a set of nfs. An nf a will be in this extension iff: (1) An individual {3 E U that is in ezt(a) is in e:.ct([barking]); and (2) int([bee:ci 1 t],a) ~T int([barking], {3) and there is an l such that l = loc([bee:ci 1 t], a) 1\loc( [barking] ,/3) .12 We must first calculate int and loc for bare existential sentences. Consider the sentences in (289): (289)

a. There is exactly one even prime.

b. There was a woman. If nothing in the context explicitly precludes it, such sentences are interpreted as holding indefinitely, within the limits pragmatically admitted by the nature of the individual instantiating the postcopular DP. For example, int([bee:ci 1 t], [ ezactly one even prime]) will be a maximally large interval; in contrast, the interval over which (289)b holds will not be infinite, but in the absence of other information will be assumed to hold over the natural lifespan of a woman. To the degree that the interval is taken to be persistent, the existential predicate is reminiscent of individual-level predicates as discussed in Section 2.2.2. loc([bee:ci 1 t]), nf), for any nf EN F, is simply the space occupied by the instantiation of the individual that supports the existential assertion. In contrast, a existential sentence with a final predicative phrase like There wa8 a dog barking in the yard describes a state of affairs that most likely holds for a delimited period of time, rather than indefinitely, and, depending on the semantics of the final predicative phrase, it might well describe one whose location is larger than the individual supporting the truth of the main predication. That is, it appears that the final predicative phrase or other information in the context is what provides interesting values for the temporal and spatial parameters of the states of affairs described by an existential sentence.

170

The English Ezistential Construction

Now, we must ensure that the individual that is instantiated in order to meet the satisfaction conditions associated with beu;ist is the same individual that has the property contributed by the XP. For instance, consider a sentence like (290): (290)

There was exactly one dog barking in the yard.

What prevents this sentence being satisfied by a situation in which Fido is instantiated as the only dog barking in one part of the yard; and in which, simultaneously, Fifi is the only dog instantiated in a nearby part of the yard (perhaps doing something else)? The answer lies in the condition that there must be some overlap in the location values for the main and adjunct predications, coupled with the general principle we have adopted that the spatial parameter associated with any given ascription should be no larger than necessary. [bee:r:i 1 t](nf), for any nf E N F, will always hold within the location associated with the adjunct predication (here, loc([barking], u), u E e:z:t ([ ezactly one dog])), since whatever individual guarantees the satisfaction of the adjunct predication will automatically guarantee the satisfaction of the existential predication. Moreover, the minimal region in which the location values for the main and adjunct predications overlap is one in which the individual supporting the existential assertion is a participant in the adjunct state of affairs. The exactly one dog we have barking in the yard (290) must therefore be the exactly one dog that supports the bare existential assertion.

4.3.2

The Predicate Restriction Revisited

It is now time to see how the predicate restriction in the existential construction follows from what we have laid out. Recall from Section 2.2 that individual-level predicates are acceptable predicative adjuncts when certain pragmatic conditions are met, for example, if the main predicate contributes information useful in demarcating a boundary at which the individual-level property begins or ceases to hold. Consequently, if the analysis advocated here is correct, there must be no such pragmatically determined connection available between the hold time of the individual-level property ascription and that associated with the existential property ascription.

Chapter

4. Adjunct Predicates and the Predicate Restriction 171

One case in which no such connection could be established was when both the main and adjunct predicates were inherently associated with temporally unbounded intervals, as in (278)a, repeated here in (291): (291)

*Everyone admired the rescuers courageous.

We may therefore hypothesize that the prohibition on individuallevel predicates in the case of the existential construction arises for the same reason: Bare DP existentials express property ascriptions that are associated with an inference of temporal persistence. One need only compare sentences such as the following to see that this 1S so: (292)

a. There are women. b. There are women cycling.

The effect cycling has on (292)a is just like the effect we found stagelevel adjunct predicates to have on individual-level main predicates in Section 2.2. (292), in the absence of any contextualization, describes a state of affairs that holds as long as instantiations of [women] exist; in contrast, (292)b describes a state of affairs circumscribed by the temporal and spatial parameters associated with the women's cycling. In this sense, the adjunct induces "stage-levellike" behavior for the existential predicate. 13 This concludes the extension of the Adjunct Rule to the existential construction, and our account of the predicate restriction. The final task is to construct an interpretation for eventive existentials.

4.4

Eventive Existentials

4.4.1

The Problem

As mentioned in the introduction to this chapter, one very obvious difference between the XP in existential sentences and the depictive adjuncts we have looked at involves the interpretation of past participles. This difference is exemplified in the contrast between (293)a and b: A past participle in an existential sentence licenses

172

The English Ezistential Construction

an "eventive" interpretation ((293)a), whereas the same participle in a standard depictive context licenses only a result interpretation, as evidenced by the anomaly of (293)b: (293)

a. There has been a live pig roasted. b. #We have eaten a live pig roasted.

Examples such as (293)a are part of the traditional motivation for relating existentials and passives derivationally, since existentials usually have synonymous passive counterparts (cf. A live pig has been roasted). However, I will argue that, to the extent that such a relation entails that the XP is the main predicate of the sentence (as opposed to a bare predicative complement or adjunct predicate), it should not be maintained. More recently, Hannay 1985 has used such examples to argue that the object of assertion of an existential utterance is not typically an individual but rather a state of affairs. While the intuition that {293)a is about a roasting event rather than about a pig is quite strong, such an analysis has to confront the fact that the XP cannot be analyzed as a complement to be,_zi•t· Observe that extraction of a measure phrase, shown in Chapter 2 to be acceptable out of complements but not adjuncts, is impossible out of the past participle in the existential construction (compare {294)b): (294)

a. Ann wanted the car driven 1000 miles. b. How many miles did Ann want the car driven? c. There was a new plane flown 10000 miles last week. d. *How many miles was there a new plane flown last week?

An additional problem is that an analysis that takes (293)a to assert the existence of a roasting state of affairs raises questions about why the definiteness restriction manifests itself in such cases, since such an analysis requires the postcopular DP to combine semantically with the XP before it combines with beui•t, and the DP should therefore not be subject to influence by that verb (see the

Chapter

4. Adjunct Predicates and the Predicate Restriction 113

discussion in Chapter 1, Section 3.1 ). Thus, if we can avoid treating the XP as a complement while still accounting for the facts, we will have a more successful analysis than we would have otherwise, since on any other analysis the failure of extraction in (294)d remains unexplained, and the DE is not easily accounted for. The solution I defend here is to treat the postcopular string in examples like (293)a and (294)c as a DP, rather than a DP-XP string-just as Williams 1984 suggested. This move is perhaps surprising, given that (293)a was a central part of Milsark's argument for the independence of the XP. Nonetheless, it is less problematic than an analysis taking the XP to be either a complement or an adjunct to bee:z:i~t, and it accounts for the failure of extraction in examples like (294)d as cases offailed extraction from DP. I begin this section begins by showing why an adjunct analysis of the past participle in (293)a is unlikely. I then turn to the evidence against treating examples such as (293)a as some mutated form of the passive. Finally, I suggest how an analysis that takes the postcopular material in (293)a to be a DP can account for the facts.

4.4.2

Eventive Participles Are Not Adjuncts

The adjuncts we have seen so far have been mainly adjectives and present participles, including subject- and object-controlled present participles which are identifiably VP (in virtue of taking a direct DP complement): (295)

a. Harry ran into her lecturing the students. b. We recognized him wearing those sunglasses.

Let me justify briefly the claim that the adjunct XP can be VP, by showing that VP secondary predicates in (295) are adjuncts. The fact that the postverbal DP is a pronoun rules out the possibility that the present participle is a postnominal modifier in these examples, as does the unacceptability of it-clefting:

The English Ezistential Construction (296)

a. *It was Mary lecturing the students (that) Harry ran into. b. It was Mary Harry ran into. c. *It was him wearing those sunglasses (that) we recognized. d. It was him we recognized.

To confirm that these have a complement-adjunct structure rather than a small clause complement structure, note that expletivesubject present participle VPs are unacceptable substituted int.o these sentences (compare (297)c): (297)

a. *Harry ran into it raining. b. *We recognized it raining. c. We noticed it raining.

The unacceptability of the expletive subject is predicted if the verb requires a contentful DP-type complement. Further evidence for the adjuncthood of these VPs comes from the fact that, as the reader may verify, their omission does not significantly change the verb's interpretation or render the sentence ungrammatical, as it does with a XP complement-taking verb such as consider. And a final piece of support for treating them as adjuncts is that it is impossible to extract a selected adverbial out of them: (298)

a. We encountered him behaving quite badly. b. *How badly did you encounter him behaving? c. We recognized her behaving badly. d. *How badly did you recognize her behaving?

Now, though we have just seen that present participle-headed VPs are acceptable predicative adjuncts, past participle-headed VPs are not:

Chapter

(299)

4. Adjunct Predicates and the Predicate Restriction 175 a. *We recognized Harry served a steak. b. We recognized Harry being served a steak. c. We recognized Harry after he had been served a steak. d. *They ran into Sally sold a new car. e. *They brought the baby to the party given a bath.

The problem in (299)a is clearly linguistic and not conceptual: Two sorts of states of affairs the sentence might be trying to characterize are described felicitously in (299)b and c. There must therefore be a clash between the semantics of the past participle and the linguistic conditions governing the possible relations that may obtain between the intervals over which the main and adjunct property ascription relations hold. We can begin to understand this clash by considering the fact that past participles entail culmination (to borrow Parsons' term), while present participles do not. That is, the participial morphology on served a steak entails that the interval during which Harry was served a steak ends prior to the relevant reference time, while the present participle counterpart being served a steak, carries no entailment of culmination-int([serve a steak], [Harry]) may overlap with the relevant reference time. If the reference time to which the past participle morphology is sensitive is the onset of the interval of the principal property ascription, then the clash becomes apparent: the use of the past participle in (299)a entails that the interval during which Harry was served a steak precedes the interval during which we ran into him. This conflicts with the requirements of the Adjunct Rule, and the result is unacceptability. We might wonder whether an individual can have the property of having had a property. That is, can Harry have the property associated with [serve a steak] once the steak has been served to him, or alternatively, is there a property corresponding to the denotation of having been served a steak? McCoard 1978 claims not; Parsons 1990 disagrees, but suggests that such a property is a strange one to ascribe to the extent that it must a property that an individual necessarily has forever. Some support for McCoard 's view comes from the unacceptability of subject- or object-controlled ing-form

176

The English Ezistential Construction

perfect participle VP-adjuncts, whether active or passive: 14 (300)

a. *We recognized Harry having been served a steak. b. *Nobody encountered the VIP having properly dressed. c. (cf. Nobody encountered the VIP properly dressed.)

The problem, then, is not merely the morphosyntactic form of the VP, but its temporal and aspectual properties: though we can felicitously use the result state of dressing as an adjunct property (as in (300)c), we cannot so use the putative property of having dressed. We therefore should not be able to maintain the position that past participial phrases in existential sentences are predicative adjuncts without predicting that a similar kind of clash should emerge in sentences such as There was a live pig roasted. But if we give up the connection between predicative adjuncts and the final predicative phrase in existential sentences, we lose any explanation of the predicate restriction. Consequently, rather than pursue this avenue, I will take the position that past participle VPs that give rise to eventive readings in the existential construction are not adjuncts of the kind we have been looking at.

4.4.3

Why These Existentials Are Not Passives

Let us now consider why these participial phrases are not directly related to passives. Contrary to what one might predict given the view that passives and existentials are structurally related, adverbs that can appear post-participially in ordinary passives are downgraded for many speakers, and ungrammatical for some, in existential sentences: (301)

a. A child who lives down the street was taunted cruelly by that bully. b. Last night a man was beaten mercilessly by the police. c. A proposal to encourage development has been opposed vociferously by the city council.

Chapter

(302)

4. Adjunct Predicates and the Predicate Restriction 177 a. %There was a child who lives down the street taunted cruelly by that bully. b. %Last night there was a man who was carrying a knife beaten mercilessly by the police. c. %There has been a proposal to encourage development opposed vociferously by the city council.

Moreover, on the original" There-insertion" analysis and its descendants, one would expect any past participle to be licensed in the construction. However, this does not appear to be the case. The native speakers I consulted found past participles of dative-shifted verbs in the XP position ungrammatical, or very awkward at best: (303)

a. *There has just been a student who likes syntax (assigned that article for class presentation]. b. *Recently there have been two professors who study muons (awarded a Nobel prize in Physics]. c. *Just now there was someone who represents Century 21 (shown the new house].

This array of facts is unexpected if the existential and passive constructions are directly related, e.g. by derivation. Consequently, I reject such a relationship here.

4.4.4

A Proposal

Having found reasons to reject analyses on which an eventive past participle VP is a complement or part of a complement to existential be, on which it is an adjunct, and on which it is an analog of the passive, I now want to explore the consequences of treating it as an adjunct to the postcopular nominal, as in (304):

178 (304)

The English Ezistential Construction IP

A I' D/\ There I VP DP

I

V'

A

V

II\

was

DP

Det NP

I

a

I

N'

A /JU N'

AP

pig

roasted

Note that I am not claiming that all existentials have this structure; my claim is simply that eventive existentials are associated with this structure. The immediate appeal of the DP analysis is that it dissociates the reference point for the culmination of hold time of the ascription associated with the past participle from the interval during which the main predication holds. The examples in (305)a,b show that the state of affairs associated with the past participle may culminate after that associated with the main predicate (cp. (305)c, where the past participial phrase is not a part of the postverbal DP and such a construal is impossible): (305)

a. [DP The people telephoned yesterday] had all voted for Brown. b. Farmer Jones sold us [nP the pig roasted yesterday]. c. They bought it roasted.

This dissociation will prevent the kind of clash discussed in earlier in this section and open the door for the eventive interpretation

Chapter

4.

Adjunct Predicates and the Predicate Restriction 179

of existential sentences. Crucially, among the possible reference points for the culmination of the ascription associated with the past participle is the moment of utterance, as is made explicit by the use of yesterday in (305)a,b. At least two facts support the structure in (304). First, when postcopular material is extracted, leaving behind the past participle, as in (306), the result is either bad, as in (306)a, or if acceptable, has only a reading on which the participle is interpreted adjectivally, as in (306)b: (306)

a. *What has there been shot? b. What has there been roasted?

These data are unexpected if the DP and past participial phrase in eventive existentials are distinct constituents, but unsurprising if they must form a single constituent in order to have an eventive reading. Second, it is possible to coordinate a phrase like a live pig roasted with appropriate DPs, something that should not be possible if the DP and XP are distinct constituents (and do not form e.g. a small clause): (307)

a. At the picnic, there was a volleyball game and a live pig roasted. b. In the last month, there have been riots and peasants murdered in the countryside.

This kind of coordination sounds much worse in cases where the XP is demonstrably external to the DP, as in (308)c and (309)c: (308)

a. There was constant protesting. b. There was a political leader who supported revolution calling for the cabinet to resign. c. ??There was constant protesting and a political leader who supported revolution calling for the cabinet to resign.

180 (309)

The English Ezistential Construction a. At the picnic, there were volleyball games. b. At the picnic, there were people that played well showing off for the crowd. c. ?? At the picnic, there were volleyball games and people that played well showing off for the crowd.

In addition, one of the main diagnostics for the eventive existential, viz. the licensing of expressions such as just now or the present perfect with just, is not as reliable as one might expect. Milsark based one of his arguments for DP-XP independence on the following contrast: (310)

a. There has just been a man shot. b. #There has just been a man.

The argument was that if a man shot had the interpretation normally associated with DPs, (310)a should be as anomalous as (310)b. However, this argument is weak because any DP that carries implicit reference to a temporally bounded interval can appear in this context, not just those DPs that are strictly speaking eventdescribing ((311)b): 15 (311)

a. There has just been a riot. b. There has just been a visitor. c. There has just been a visit by the inspectors.

We can see that the DP in (311)b picks out an individual, and not a covert state of affairs (as one might be tempted to argue), by comparing the felicity of anaphoric pronouns in subsequent discourse. (312)a may be continued by a sentence containing pronoun referring to a person (e.g. (312)b), but not by one referring to a state of affairs (e.g. (312)c; compare with (311)c, for which the opposite is true): (312)

a. There has just been a visitor. b. She stayed 10 minutes. c. #It lasted 10 minutes.

Chapter

4. Adjunct Predicate! and the Predicate Restriction 181

That is, reference to short periods of time via the existential construction is not inconsistent with the instantiation of ordinary individuals as opposed to states of affairs. Consequently, the structure in {304) cannot be dismissed out of hand on the basis of examples involving just now or the present perfect tense. 16 Indeed, the facts we have just laid out lend support to the structure proposed in (304). The discourse in (313)a is as strange as that comprised by (312)a and c: {313)

a. Yesterday there was a live pig roasted. ??It was over in an hour. b. Yesterday there was a visit by the authorities. It was over in an hour.

I conclude that there is no strong reason not to pursue the structure in (304). Let us see, then, how it might be interpreted. Consider the pig-roasting example once again, comparing it to (314)b and c: {314)

a. Yesterday, there was a live pig roasted. b. #Yesterday, there was a pig. c. Yesterday, there was a visitor.

(314)b is, of course, perfectly felicitous in certain contexts, e.g. as a response when I ask what animals were put up for sale this week at the farmers' market. The problem with (314)b out of context is that the use of yesterday implicates that the state of affairs described by the clause it modifies no longer holds. This implicature of temporal boundedness is, in turn, incompatible with the intuition that the interval over which There was a pig holds is inferred to persist, since we infer that any individual that instantiates a pig will be a pig for a persisting interval (see Section 3, above). Apparently no such implicature is associated with the main clause in (314)a or c. Why not? Consider the case involving visitor first. On the basis of the contrasts in {315), I take visitor to be interpreted as a relation between a temporal interval and an individual:

182 (315)

The English Ezistential Construction

a. the table's top/#the top's table b. the day's man/the man's day c. the day's visitor/ #the visitor's day

Barker 1991:59ff. observes that relational nominals cannot form prenominal possessives with DPs intended to supply a value for the nominal's internal argument (hence the asymmetry in (315)a). The cross-hatched DP in (315)a can be interpreted with an arbitrary possession relation ( cf. Barker 1991), just as the relation between man and day is arbitrary in (315)b, but it cannot be assigned an interpretation on which the denotation of the possessor is a part of the table. A similar asymmetry emerges with visitor in (315)c, suggesting that it too, unlike e.g. man, is relational and, moreover, that its internal argument is an interval: the interval during which the visitor pays his or her visit. 17 But if visitor is interpreted as a relation between an interval and an individual, then it is reasonable to suppose that for any pair t,z, int(([visitor](t)), :z:) = t and, moreover, that this value could be quite short. This has implications for the existential construction because the interval over which the existential predicate holds of some nominalized function is parasitic on the interval over which an individual is taken to instantiate that nominalized function. Since int([pred(a visitor(t))], :z:), for any z,t, can be quite short, int([beu:iat], [a vi.!itor]) can be similarly short. Consequently, (314)c is acceptable with a modifier like ye,terday, while (314)b is not. If we can extend this sort of explanation to the pig-roasting cases, we will have no reason not to adopt the DP structure proposed at the beginning of the section. Consequently, we must ask under what circumstances, and for what interval, an individual z will be in the extension of the interpretation of an N' like live pig roasted, and whether that interval is inferred to persist (as with [pig]) or not (as with [ vi.!itor]). I will assume for the purposes of this discussion that a live pig roa.!ted has the syntactic structure in (316):

Chapter 4. Adjunct Predicates and the Predicate Restriction 183 (316)

DP

A

D NP

I

I

N'

a

~VP

N'

/\ D

AP

N'

li~e ~

roasted

pig

I adopt this structure because the DP will not describe anything sensible unless [live] modifies [pig] directly, with the combination in turn modified by [roasted]: Our target of description is a pig that was alive at the time it underwent a now-finished roasting and that (presumably) is no longer alive now that the roasting has finished. I will take (317), a version of the standard treatment of adjectival modification, to determine [live pig]: (317)

=

a function f of the same sort as for any choice of < w, t, l >, Vy E e:z:t([N']),y E e:z:t(/) iff y E e:z:t ([AP]). [[AP N')N']

[N'], such that,

All this rule says is that an individual is in the extension of e.g. [live pig] at some index iff it is both live and a pig at that index. We have no reason to suppose that the inference of temporal persistence associated with ascription of the property [pig] is defeated by the addition of live: (318) sounds just as odd as (314)b: (318)

#Yesterday, there was a live pig.

We must therefore look to roasted. I maintained in Section 4.1, above, that an individual could have the property of participating in a state of affairs only while that state of affairs is in progress, not before or after. What the past participle morphology tells us is that the property the participle identifies is one that its controller

184

The English Ezistential Construction

had at an interval prior to some reference point. Therefore, we can maintain the same rule as in (317) for the combination of N' with a past participle postmodifier, assuming the participle carries with it the condition that the interval at which the individual has the property identified by the past participle terminate prior to some contextually determined reference point. If we maintain (317) for the interpretation of [N' VP] strings as well, then by what we have said about the interval over which an individual can have the property identified by a past participle, any inference of temporal persistence will be defeated. An individual z instantiating live pig roasted will do so only while it has the property of being live, being a pig, and being roasted. The inflection of roasted indicates that the interval associated with this latter property has culminated with respect to the relevant reference time. This temporal restriction associated with [roasted] will be inherited by the description as a whole; that is, for any :z:, int( [live pig roasted], :z:) will at most equal int( [roasted], :z:). Since it was suggested above that for any choice of :Z:nJ, int([beu:i 6 t], :Z:nf) is determined by the interval over which the requisite individual(s) instantiate :Z:nf• we may conclude that the interval over which [be~:z:i 6 t] holds of [a live pig roasted] can be quite short; short enough to accept modification by a temporal adverb such as yesterday. Since the interval over which the existential ascription holds is, on this view, entirely parasitic on the interval over which the past participle property holds of its controller, it is unsurprising that existential sentences such as There was a live pig roasted seem to be about a state of affairs rather than about an individual. This section has not by any means done full justice to the facts concerning past participle postmodifiers. For example, I have not investigated the differences between bare participle postmodifiers such as those under discussion and full relative clauses. DPs with full relative clauses differ from the strings I am calling DP here in at least two ways. First, indefinites with full relative clauses take wide scope quite successfully, as in the ambiguous (319)a, while we saw in the text above, and it can be seen in (319)b as well, that indefinites with reduced past participle postmodifiers strongly prefer narrow scope:

Chapter

(319)

4. Adjunct Predicates and the Predicate Restriction 185 a. Everyone admired a scientist who was awarded the Nobel Prize. b. Everyone admired a scientist awarded the Nobel Prize.

Second, DPs with full relative clauses do not give rise to "eventive" existentials:

(320)

a. #There has just been a man who was shot. b. #Yesterday there was a man who was shot. c. #Yesterday, there was a live pig that was roasted.

I suspect that these facts are related; however, I must leave their analysis for future research. Nor have I attempted to explore any generalizations concerning the behavior of predicative postmodifiers of N vs. V.

4.5

Chapter Summary

In this chapter I have presented an interpretation for predicative adjuncts along with an extension of that interpretation to existential sentences containing final predicative phrases. The requirement that, in most cases, the adjunct be a stage-level predicate was argued to follow from the fact that the adjunct imposes a kind of temporal restriction on the interpretation of the VP in which it appears, in concert with pragmatic considerations. I then suggested how eventive existentials, which are not amenable to either the DP-complement, XP-adjunct analysis or to a small clause analysis, could be interpreted appropriately if all of the postcopular material was contained within DP. We now have a complete picture of the existential construction and its interpretation within a property-theoretic semantics based on that presented in Chierchia and Turner 1988. As mentioned previously, it is likely that the facts discussed in this chapter could be handled more elegantly if an algebra of events was incorporated directly into the system along with a more sophisticated treatment of aspect. I have avoided incorporating events directly into the

186

The English Ezistential Construction

semantics in order to keep things simple, but I see no reason why they could not be added if necessary.

Chapter

4. Adjunct Predicates and the Predicate Restriction 187

Notes 1 Rapoport

1991 has an analysis that looks very much like ( 265); however, it is unclear whether she takes the event argument to stand for a primitive notion of state of affairs or a discourse-anchorable spatio-temporal parameter. 2 A precedence relation should also be defined over the intervals in T, but I will not do this here as it will play no role in the analysis. 3 Link 1987 uses the terms "temporal trace" and "spatial trace" to characterize the analogous functions in his system; Lasersohn 1988, "running time" and "running space." 4 "Theme" here includes participants that are transferred (We handed her the boz unopened); that are consumed or destroyed They ate the meat raw, They blew up the cave unezcavated); that appear (They found him crying); that are simply present (They waited embarrassed). It excludes those that are created (*They made the boz wooden), as well as those which might be called "patients": *She kissed him feeling sorry for himself, *She kicked the boz unopened. Though the precise descriptive generalization governing the class of possible nonsubject controllers remains to be determined, it clearly does not reduce to either "direct object" (as Williams 1980 observed) or "direct internal argument." 5 See e.g. Carrier and Randall 1992 for arguments against positing PRO in these structures. Like Jackendoff 1990:207ff., I consider the issue of whether or not PRO belongs in these structures to be largely orthogonal to the issue of principal concern here. 6 The rule treats the locative and temporal parameters differently in virtue of the fact that, while full temporal simultaneity is quite possible for distinct states of affairs, fulllocational overlap is not the norm. 7 In locating the event argument in the semantics (her "event structure") but not argument structure, Rapoport's proposal differs from the treatment of event (or spatia-temporal) arguments in e.g. Kratzer 1989 and Diesing 1992. 8 See below and also McNally 1994 for more on why this might be so. 9 Unless we redefine what we take that property to mean (as in Maz is being a child, where the child property corresponds to a particular kind of behavior); or unless the property is relativized in some way (as in Terry has been naive in his view of politics before,

188

The English Ezistential Construction

and he will be again, where the naivete is not necessarily a general property of Terry but rather characterizes an attitude or reaction of Terry along a particular dimension). 10 See Condoravdi 1992a for additional remarks to this effect. 11 (285) presupposes the existence of some mechanism that I have not specified in the text that allows the subject to be an adjunct predicate controller; however, the Adjunct Rule is exactly the same whether the controller is the subject or a non-subject theme. 12 Some question might arise concerning the fact that barking selects an argument of a different sort than does beu:i•t· Note, however, that this happens in other control contexts as well, as in (i), in which the sort of the subject argument selected by the verb try is an ordinary individual, while that selected by the controlled predicate eztinct is a kind: (i) It is as if human beings are trying to become extinct. 13 0n this analysis, it is surprising that examples such as (ii) are bad, even though the hold-time of a riot is imaginably quite circumscribed, and so could conceivably license an individual-level adjunct as in Martha returned a new person: (i) There was a riot. (ii) *There was a riot horrible. I do not have a good explanation for the unacceptability of (ii). However, the fact that (iii) is also bad suggests that the problem involves the fact that events such as a riot have no existence independent of when and where they occur (cf. Kimball1973): (iii) *Yesterday, a riot occurred horrible. That is, the riot could never have the property of being horrible except while it was occurring. Perhaps this fact is the source of the problem. 14 However, subject-controlled progressive perfect participles may be, and often are, used as free adjuncts (Stump 1985). Sentencefinal free adjuncts are identifiable by an intonation break, by obligatory subject control, and by the fact that they cannot be controlled by a necessarily quantificational DP. 15 The class of non-eventive nouns that show up in the existential in this context appears to be limited. Other examples I have found are: burglar, speaker, guest, band, and perhaps noise, as well. 16 There is a contrast between these facts and those associated with the wh-cleft examples used by Hannay 1985 to argue that

Chapter

4.

Adjunct Predicates and the Predicate Restriction 18 9

existential sentences are used to assert something about a state of affairs rather than about an individual. (ii) is just as acceptable as (i), even though (iii) is bad, a fact that cannot be a consequence of a visitor describing a state of affairs: (i) What happened yesterday was (that) there was a live pig roasted. (ii) What happened yesterday was (that) there was a visitor. (iii) #What happened yesterday was (that) there was a woman. I do not have an explanation for this contrast. 17 0f course, visitor may also be interpreted as a 1-place property of an individual (perhaps one who has the habit of being a visitor), and the second possessive in (315)c is interpretable as identifying, for example, a particular day assigned to that visitor. However, the phrase lacks an interpretation on which it refers to the day on which the visitor arrived, which is what is relevant.

Chapter 5

Some Final Remarks 5.1

Introduction

I have argued in the preceding chapters that the existential predicate is interpreted as a one-place property of a nominalized function. However, felicitous use of the existential construction also requires, in all of the cases we have seen so far, that a novel discourse referent (fitting the description associated with the nominalized function) be introduced, in effect to support the truth of the existential predication. The first goal of this chapter is to show how the attested examples of true definites in existentials fit into and support the overall structure of the proposal. The specifics of the proposal raises certain expectations about what we might find elsewhere in English. For example, adding the novel discourse referent condition on to the construction independently of bee~i 5 t (via the expletive there) makes two predictions: first, that we might find the condition in other places where we find expletive there; and second, that we should not find the condition associated with bee~i 4 t if bee~i 5 t occurs with other sorts of expletive subjects (as I will suggest it does). The second goal of this chapter is to offer some programmatic evidence that both of these predictions are borne out in English. Of course, this way of parceling out the informational contribution associated with the existential construction also opens itself to the criticism of insufficient cross-linguistic generality: For 191

192

The English Ezistential Construction

instance, isn't there the equivalent of a novel discourse referent condition associated with the equivalent construction in languages such as Spanish, which lack an overt expletive? The answer is yes. However, this sort of objection is not so strong in the larger scheme of things. The key to understanding why it is not is to remember that the basic interpretation of the existential construction, the interpretation I take to be cross-linguistically consistent, is the ascription of the property "is instantiated" to a description. As interpretations go, this is extremely general, and since the novelty condition is logically independent of it, we might expect cross-linguistic variation to start there. For example, it is perfectly possible to imagine a language that does not require the instantiation of a novel discourse referent to support the truth of the existential assertion. Such a language is Catalan. 1 It may be the case that languages are overwhelmingly likely to conventionalize use of existential sentences in such a way as to exclude definites because of the presuppositions generally associated with them. However, that does not necessarily mean that the definiteness restriction should be built overly deeply into the semantics of the construction. Similarly, since the basic interpretation of the existential is so general, it is also possible to imagine a language that uses the same basic predicate to express different, and fairly specific, sorts of instantiation, distinguishing them by supplementing the basic predicate with a paradigm of particles or phrases; I will suggest below that such a language is English, and that existential sentences in English forms a paradigm with sentences such as these (see Carlson 1991 for arguments that sentences of the form in (321 )b,c are not equative statements): (321)

a. It is Mary. b. This is my mother. c. That was the mail carrier just now.

If the informational contribution of the existential construction is not divided up in English as I suggest, the similarities between the examples in (321) will go unaccounted for. In other words, the positing of very specific locations for the pieces of the interpretation is not inconsistent with the capturing

Chapter 5. Some Final Remarks

193

of broader generalizations.

5.2

Definites in the Existential

5.2.1

List Existentials

It was observed at least as early as Milsark 1974 that one or more definite DPs in the postcopular position are acceptable with the interpretation that the DP(s) identify individuals in a list. An example appears in (322): 2

(322)

a. Who is there available to fix the computer? b. Well, there's Alice.

List existentials are often identified in the literature by the context in which they occur (often after questions, as in the example here) and by their intonation-they tend to have rising intonation at the end, rather than falling intonation (Rando and Napoli 1978). However, as Abbott 1992 observes, it is important to note right away that the issue of whether or not a sentence is properly classified as a list existential is independent of whether or not its DP is definite. (323) would be an equally felicitous response to (322)a, and could be given the same rising intonation contour standardly associated with the list existentials: (323)

Well, there's a guy at Gunther.

The distinction is important because several interesting properties have been ascribed to list existentials that are more appropriately described as properties of existentials containing definite DPs. For example, it has also been claimed 3 that list existentials cannot contain a secondary predicate (Hankamer 1973, Safir 1985), as seen in the oddness of (324)b and (324)c: 4 (324)

a. Who is there available to fix the computer? b. ??Well, there's Alice free right now. c. ??Well, in the back room, there is the guy. d. Well, in the back room, there is a guy who's competent.

194

The English Ezistential Construction

However, the fact that (324)d is an appropriate response to (324)a (with the list intonation) indicates that the badness of an adjunct phrase is associated with the definite, and not the context in which the existential is being used. Another putative property of list existentials is that they cannot be negated (Hankamer 1973): (325)

a. Who is there available to help me? b. ??Well, there isn't Alice. c. Well, there isn't a technician, a sales rep, or a manager anywhere to be found.

To the extent that (325)b is interpretable, it is with metalinguistic negation. Again, we can see from the acceptability of (325)c that the definite is responsible. An additional notable property of existentials containing definites is that they make poor non-negated questions ( (326)b ), though they make very good negated questions ((326)c): 5 (326)

a. Who can help us? b. ??Is there Alice? c. Isn't there Alice? d. Is( n 't) there a technician around?

Again, indefinites in the same context behave differently. The fact that definites and indefinites behave differently in the same list-inducing contexts supports the position taken by Hannay 1985, Lumsden 1988, and Abbott 1992 that these peculiarities are not the result of a semantically distinct "list interpretation," as has been suggested by e.g. Milsark 1974 and Safir 1985. 6 Rather, they have all argued that ordinary and list existentials differ only pragmatically, the presuppositional properties of the definite DP being responsible for the differences between the two sorts of sentences. Indeed, there is a bias in the analysis developed in the preceding chapters towards a pragmatic explanation of the facts, since the analysis does not exclude definites from the construction for any reason other than the familiarity presuppositions they carry.

Chapter 5. Some Final Remarks

195

Now consider the role of the individual instantiated when an existential sentence is added to a context: that individual is necessary to support the truth of the existential predicate as applied to its argument. But since that argument is not the individual itself, but rather the description of the individual, there is no reason to suppose that the truth of a sentence such as There is Alice could not be supported by individuals already instantiated in the discourse. We thus might expect existential sentences containing definite DPs to be licensed just in case the conflict between the act of instantiation associated with the utterance of an existential sentence and the definite's familiarity presupposition can in one way or another be resolved. Since it is a fact that the familiarity condition associated with definites cannot be shut off, the only possibility must be that in some cases we can shut off the condition that a novel discourse referent be introduced when the existential construction is used. For example, imagine that the novelty condition could be blocked or defeated when the truth of an existential sentence was presupposed, the rationale being that conversation participants should be willing to acknowledge that the individuals needed to support the truth of the utterance have already been instantiated and, consequently, that no further individuals should be instantiated upon repeat utterance. That is, assume the following (language particular) principle: (327)

If an existential sentence is presupposed to be true in the context in which it is uttered, its ( re- )introduction into the context is not accompanied by the instantiation of any discourse referent.

With (327), the door is open for the felicitous use of existentials containing definites: Bare-DP existentials will be licensed, since the (effective) familiarity condition on the use of a definite DP entails the truth of There is DP in any context where the use of the DP is felicitous (as Barwise and Cooper 1981 observed). However, (327) correctly predicts that the facts will be different when the definite appears with a secondary predication. Secondary predicates introduce information that is potentially informative; for example, nothing independent of the particular context guarantees that Alice will have any particular property at any given time.

196

The English Ezistential Construction

Consequently, we expect that sentences such as There is Alice free will be bad if their truth is not already entailed by the context; likewise, if their truth is entailed, such sentences should be acceptable. The unacceptability of the examples in (324)b,c show that the former expectation is realized. Whether the latter is realized as well is somewhat less clear; while the third sentence in the sequence in (328) is reasonably acceptable as an echo of the answer to the question in (328)a, that is, in the wake of explicit introduction of information that will guarantee the truth of the existential clause, it is still seems a little strange. (328)

a. Who is available to help? b. Alice is free right now. c. ?That's right, there's Alice free.

Perhaps (327) should be modified on the basis of the facts in (328); for now, however, I will move on, leaving (327) as it is. Presumably the presupposed truth of a sentence such as There is Alice can be exploited by the speaker, as a flouting the Gricean (sub )maxims that one's utterances should be informative and relevant. If someone asks a question such as Who can help me?, it is reasonable to suppose that a trivially true response that makes salient a particular individual's name or description would license the inference on the part of the hearer that the speaker believes the individual fitting the description can help him/her. Trivially true assertions may also be useful as affirmations of previous comments in the discourse, as in sequences such as the following, with the unstressed anaphoric pronoun that, which I have found quite common in naturally occurring conversation: (329)

a. I couldn't vote for Smith; how can you trust anyone who was considered a potential donor to the Contras? b. Yeah, there's that, and let's not even mention his apparently uncompromising personality.

Further support for the position that speakers might exploit the triviality of an existential sentence by appealing to something like (327) comes from the contrast in acceptability between negated

Chapter 5. Some Final Remarks

197

questions, on the one hand, and non-negated questions and negated assertions, on the other. The utterer of the negated question Isn't there Alice? implicates that s/he believes that There is Alice is true; the negated question is simply a slightly hedged version of the positive assertion (as are negated yes-no questions and tag questions in general) and, given (327), can be exploited in similar contexts as the non-hedged assertion. In contrast, the utterer of the non-negated counterpart Is there Alice? explicitly does not assume that There is Alice is true, and whoever utters There isn't Alice explicitly denies it. Both of these positions conflict with the felicity condition associated with the use of the name. Consequently, only the negated question is acceptable. A final source of support for (327) could be adduced from the behavior of entailed existentials with indefinite DPs, such as B's remark uttered in a context immediately following A's comment: (330)

A: There were [15 people]i at the meeting. B: So that means there were [10 people]j at the meeting, which is enough for a quorum. B: Did theyi/•j agree to vote on the measure?

The existential sentence uttered by B is necessarily true in this context; consequently, (327) should apply to it just as it applies to existentials with definites, and no new discourse referents should be introduced. In fact, none are: The use of the pronoun they in subsequent discourse may pick out the 15 people who were at the meeting, but not the 10 people B mentions, as B's continuation in (330) shows. 7 I have pointed out that one way to guarantee that an existential containing a definite is entailed by the context is to exclude any secondary predication. However, there is at least one other way to guarantee the entailment and, hence, to license definites in the construction.

5.2.2

Focus and the Existential

For many speakers, the presence of only can improve the acceptability of definites in the existential, as in the examples in (331):

198 (331)

The English Ezistential Construction

a. There was only Kent available. b. Among those in the cage, there was only the brown dog barking.

These examples are of interest because, contrary to the generalization introduced above, they have an independent final phrase. The obvious difference is the contribution of only. One look at the semantics for only (see e.g. Rooth 1985) reveals the crucial property: 8 (332)

[only](a) (Rooth 1985:120): Assertion: Vp[C(p) 1\ p-+ p a] Conventional Implicature: a is true

=

One of the effects of only in (331) is to introduce the presupposition that the propositions associated with the sentences in (333) are true: (333)

a. *There was Kent available. b. *Among those in the cage, there was the brown dog barking.

Since they are not asserted but rather are presupposed, by the principle in (327) there will be no discourse referent instantiated and consequently, no problem. That is, the acceptability of these sentences in spite of this final phrase is completely in keeping with the general pragmatic approach we have been taking. Consistent with this is the fact that replacement of only with even renders the examples in (331) as bad as the examples in (333) for a number of speakers: (334)

a. %There was even Kent available. b. %Among those in the cage, there was even the brown dog barking.

The difference is that the addition of even does not change the fact that the main assertion of the sentence is [beu:i•t]([DP]), where the DP is definite, as inspection of Rooth's semantics for even shows: (335)

[even](a) (Rooth 1985:120): Assertion: a is true Conventional Implicature: 3p[C(p) unlikely(p) J

1\ p 1\

[p -::j::. a]/\

Chapter 5. Some Final Remarks

199

To summarize, the purpose of this section has been to suggest a way in which the licensing of definites under certain circumstances can be accounted for given the analysis proposed in the previous chapters. An appeal to Gricean principles, rather than to a distinct semantic interpretation, has been made: specifically, I have suggested that definites in are licensed in the existential in English iff the truth of the existential assertion is presupposed.

5.3

Other Expectations

5.3.1

Other Expletive There Sentences

As I mentioned at the beginning of this chapter, the association of the novel discourse referent condition with the expletive in English predicts that such a condition might show up in other contexts where the expletive appears. Of course, expletive there does cooccur with certain other predicates in English. These predicates have in the past been divided into various classes (see e.g. Ross 1974, Milsark 1974, Aissen 1975), a familiar classification distinguishing predicates such as ezist, occur, follow, ensue, which one might be tempted to classify as essentially "existential"; the nonagentive verbs of motion and verbs of location and appearance (e.g. fall, hang, appear); and the agentive verbs of motion (e.g. run, jump). Sentences containing predicates from the latter two classes are often referred to in the literature as "presentationalthere" sentences (Milsark 1974, Aissen 1975, hereafter abbreviated PT). These three classes of verbs are exemplified in (336): (336)

a. There exists an z such that z left at noon. b. There hung a portrait of the Queen behind the counter. c. There roared out from behind the bushes a huge white Continental with suicide doors.

It is quite commonly agreed that when the DP occurs immediately after the verb in PT sentences, 9 it is subject to the same definiteness restriction as we have seen in the existential:

200

(337)

The English Ezistential Construction

a. *There exists the z such that z left at noon. b. *There hung the portrait of the queen behind the counter. c. *There appeared Jane at my door.

This is exactly what we expect if expletive there has come to be associated with the condition that a new discourse referent must be introduced. In addition, Milsark observed that definites can appear in these sentences when the DP is right dislocated (typically, to the right of a locative phrase), an observation that seems to conflict with the claim that expletive there in English is consistently associated with a novelty condition. Some examples supporting Milsark's claim appear in (338): (338)

a. Suddenly, there stood before me Michelangelo's David. b. All at once, there appeared from nowhere my best friends from high school days long past.

c. There hangs in that museum the largest of the Calder mobiles. But the licensing of definites in this context is not systematic. Many definites do not sound good at all in these sentences, suggesting that it is an overstatement to say that there is no novelty condition whatsoever associated with the right dislocated position in PT. Consider, for example, the rather odd-sounding sentences in (339): (339)

a. ??There hung behind the counter the portrait of the Queen. b. ??There roared out from behind the bushes the huge white Continental with suicide doors.

c. ??There ran into the chamber the lawyer for the defense. Though I have not attempted a thorough investigation of just what makes some definites sound better than others in PT, a comparison of these examples suggests that definites that do not have to

Chapter 5. Some Final Remarks

201

be used anaphorically (or, alternatively put, that can introduce discourse new referents (Prince 1992) or that are very easily "accommodatable" (see e.g. Heim 1982 and references there)) appear felicitously in PT, while those definites that must be used anaphorically (or: cannot introduce discourse new referents, are difficult to accommodate) are infelicitous. Proper names such as Michelangelo's David, 10 definite possessives, and superlatives all fall into the former category. The vanilla definite DPs in (339) fall into the latter. It is important to observe that the oddness of definites in these sentences cannot be explained away as a general property of presentational focus or inversion constructions. The inverted DP in inversion constructions lacking there are subject to no such definiteness restriction (see Birner 1992 for a comprehensive discussion of inversion and relevant references): 11 (340)

a. In walked Bob. b. Out in the corridor stood the candidate. c. Voting in favor of a strike were the Spanish and French students.

Thus, despite the fact some definites are licensed in the rightdislocated position of the PT construction, it appears that they fall into a restricted category of DP that obeys a weakened version of the novelty condition we expected. In addition, it remains the case that PT sentences in which the DP immediately follows the verb are subject to the same novelty condition found in ordinary existential sentences. The facts therefore generally support the first of the predictions mentioned at the beginning of the chapter, namely that sentences containing the expletive there should consistently be associated with a novelty condition of some sort. 12

5.3.2

The Existential and Other Indexicals

The second prediction made by an analysis that associates the novelty condition in the existential construction with the expletive is that, should bee:z:i&t appear with other expletive subjects in English, no novelty condition should manifest itself. Arguably, the sentences in (341 ), which discussion in Jenkins 1975 suggests should form a

202

The English Ezistential Construction

natural class with existentials despite their lack of definiteness effect, bear this prediction out: (341)

a. It's Mary at the door. b. This is my mother playing the piano. c. That was the girls next door.

Jenkins observed that these constructions manifest a predicate restriction just like that found in the existential: When the final XPs in (341)a are replaced with XPs interpreted as individual-level predicates, the result is bad: (342)

a. *It's Mary intelligent. b. *This is my mother devoted. c. *That was the girls out of their minds.

These constructions further resemble the existential in that necessarily quantificational DPs over u-sort entities are not licensed in postcopular position: (343)

a. *It's each child at the door. b. *This is most piano students performing. c. *That was every girl next door.

Note that the lack of a novelty condition is expected if that condition is associated with there and not with the verb beezi•t· Though each of the expletives in (341) contributes something different, the similar conditions on the postcopular material suggest that there is something common to all of them. An account that assigned these various constructions essentially the same interpretation, varying only in the felicity conditions or other conventional information contributed by the different expletives, would be parsimonious in its unified interpretation of expletive-taking be; it might also allow us eventually to capture those similarities that have been observed (see e.g. Lakoff 1987) between deictic locative there sentences (which manifest e.g. the predicate restriction shown here) and the existential construction.

Chapter 5. Some Final Remarks

203

Notes 1 See

Chapter 3, Section 6. term "definites" in this section covers proper names in addition to DPs headed by a definite determiner. See Chapter 3, Section 5 on the status of personal pronouns. 3 Though such claims have occasionally been disputed (e.g. the negation claim, discussed below, by Rando and Napoli 1978; the claim concerning the restriction on the appearance of XPs, also discussed below, by Hannay 1985). 4 An apparent counterexample is presented by the acceptability of infinitival XPs in examples such as (i)-(iii) (we can see that these are not simply DP in light of examples such as (iv)): (i) There is her future to consider. (ii) There is Martha to blame for this. (iii) There is the drought to worry about. (iv) *Her future to consider could be a bright one. However, it appears that infinitival XPs of this sort are not part of the existential assertion: Question and negation examples involving indefinite DPs indicate that the infinitival phrase must escape the scope of negation and question: (i) and (ii) presuppose that someone must be considered; (iii) and (iv), that someone must be blamed. (i) Is there anyone to consider? (ii) There isn't anyone to consider. (iii) Is there anyone to blame for this? (iv) There isn't anyone to blame for this. As will become clear in the text below, the fact that the infinitival phrase is not part of the basic existential assertion in these sentences must be crucial. However, exactly how and why infinitival phrases are different from other XPs in this context is a matter that must be left for future research. 5 Abbott credits this observation to Bill Ladusaw. 6 More precisely, Milsark's suggestion was that the special property of list existentials was that what was asserted to exist in the list case was the list itself, rather than the individuals in the list. 7 Of course, this does not rule out the possibility of introducing additional discourse referents that in fact constitute one or more subgroups of this group of 15, e.g. via (i): (i) There were 2 people from the Physics department. 2 The

204

The English Existential Construction

Note that (i) is not entailed by A's comment or by anything else in the present context. 8 The details of Roath's proposal concerning the assertion associated with [only S] and the conventional implicature associated with [even S] (see below) are not important here; the reader is referred to Rooth 1985 for explication. 9 Which appears to be possible only when the verb is non-agentive. 10 Since right dislocation in PT also appears to carry a heaviness restriction, many proper names will be excluded on those grounds. 11 1 am grateful to Bob Levine for emphasizing to me the differences between the DPs licensed in PT sentences and locative mverston. 12 See Ward 1995 for a similar analysis of the discourse conditions on PT sentences.

References Abbott, B. 1992. Definiteness, Existentials, and the 'List' Interpretation. In Proceedings of SALT II, ed. C. Barker and D. Dowty, 1-16. Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. Doctoral dissertation, MIT. Aissen, J. 1975. Presentational There-Insertion: A Cyclic Root Transformation. In Proceedings of CLS 11, 1-14. Andrews, A. 1982. A Note on the Constituent Structure of Adverbials and Auxiliaries. Linguistic Inquiry 13:313-317. Bach, E. 1982. Purpose Clauses and Control. In The Nature of Syntactic Representation, ed. P. Jacobson and G.K. Pullum. 35-57. Dordrecht: Reidel. Ball, C. 1991. The Historical Development of the 'It'-Cleft. Doctoral dissertation, Univeristy of Pennsylvania. Barker, C. 1991. Possessive Descriptions. Doctoral dissertation, University of California, Santa Cruz. Revised version published 1995 by University of Chicago Press/CSLI Publications. Barker, C. 1993. Definite Possessives and Discourse Novelty. In Proceedings of CLS 28, ed. C. Canakis et al., 26-41. Barwise, J ., and R. Cooper. 1981. Generalized Quantifiers and Natural Language. Linguistics & Philosophy 4:159-219. Belletti, A. 1988. The Case of Unaccusatives. Linguistic Inquiry 19:1-34. 205

206

References

Birner, B. 1992. The Discourse Function of Inversion in English. Doctoral dissertation, Northwestern University. Bolinger, D. 1977. Meaning and Form. London. Burt, M. 1971. From Deep to Surface Structure. New York. Burzio, L. 1986. Italian Syntaz. Dordrecht. Carlson, G. 1977a. Amount Relatives. Language 53:520-542. Carlson, G. 1977b. Reference to Kinds in English. Doctoral dissertation, University of Massachusetts at Amherst. Published 1980 by Garland Press, New York. Carlson, G. 1991. Cases of Really Direct Reference: Perception and Ostension? Paper delivered at SALT I. Carrier, J., and J. Randall. 1992. The Argument Structure and Syntactic Structure of Resultatives. Linguistic Inquiry 23:173234. Chierchia, G. 1984. The Syntaz and Semantics of Infinitives and Gerunds. Doctoral dissertation, University of Massachusetts at Amherst. Published 1989 by Garland Press, New York. Chierchia, G. 1985. Formal Semantics and the Grammar of Predication. Linguistic Inquiry 16:417-443. Chierchia, G., B. Partee, and R. Turner. 1989. Properties, Types, and Meaning, Volume 1. Dordrecht. Chierchia, G., and R. Turner. 1988. Semantics and Property Theory. Linguistics £J Philosophy 11:261-302. Chomsky, N. 1986a. Barriers. Cambridge. Chomsky, N. 1986b. Knowledge of Language. New York. Chung, S. 1987. The Syntax of Chamorro Existential Sentences. In The Representation of (In)definiteness, ed. E. Reuland and A.G.B. ter Meulen. 191-225. Cambridge: MIT Press. Chung, S, and J. McCloskey. 1987. Government, Barriers, and Small Clauses in Modern Irish. Linguistic Inquiry 18:173-237.

References

207

Clark, E. 1978. Locationals: Existential, Locative, and Possessive Constructions. In Universals of Human Language 1 Volume IV, ed. J. Greenberg. 85-126. Stanford, CA: Stanford University Press. Comarovski, I. 1991. Partitives and the Definiteness Effect. In Proceedings of WCCFL 10, ed. D. Bates, 91-102. Condoravdi, C. 1992a. Individual-Level Predicates in Conditional Clauses. Paper delivered at the LSA Annual Meeting, Philadelphia, PA. Condoravdi, C. 1992b. Strong and Weak Novelty and Familiarity. In Proceedings of SALT II, ed. C. Barker and D. Dowty, 17-37. de Hoop, H. 1990. Restrictions on Existential Sentences and Object-Scrambling: Some Facts from Dutch. In Proceedings of WCCFL 9, ed. A. Halpern, 277-288. Diesing, M. 1992. Indefinites. Cambridge. Doherty, C. 1993. Clauses Without That: The Case for Bare Sentential Complementation in English. Doctoral dissertation, University of California, Santa Cruz. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht. Dowty, D. 1985. On Recent Analyses of the Semantics of Control. Linguistics & Philosophy 8:291-331. Dowty, D. 1988. Type-Raising, Function Composition, and Nonconstituent Coordination. In Categorial Grammars and Natural Language Structures, ed. R. Oehrle. 153-198. Dordrecht: Reidel. Enc;, M. 1991. The Semantics of Specificity. Linguistic Inquiry 18:1-25. Farkas, D. 1981. Intensional Descriptions and the Romance Subjunctive Mood. Doctoral dissertation, University of Chicago. Published 1985 by Garland Press, New York.

208

References

Farkas, D. 1996. Specificity and Scope. In Langue et Grammaire J, ed. L. Nash and G. Tsoulas. Fernald, T. 1991. Resultification. Ms., University of California, Santa Cruz. Fodor, J.D., and I. Sag. 1982. Referential and Quantificational Indefinites. Linguistics & Philosophy 5:255-298. Freeze, R. 1992. Existentials and Other Locatives. 68:553-595.

Language

Gawron, J.-M. 1986a. Situations and Prepositions. Linguistics & Philosophy 9:327-382. Gawron, J .-M. 1986b. Types, Contents, and Semantic Categories. Linguistics & Philosophy 9:427-476. Goldberg, A. 1994. Constructions: A Construction Grammar Approach to Argument Structure. Chicago. Green, G. 1973. A Syntactic Syncretism in English and French. In Issues in Linguistics, ed. B. Kachru. 257-278. Urbana: University of Illinois Press. Groenendijk, J., and M. Stockhof. 1991. Dynamic Predicate Logic. Linguistics & Philosophy 14:39-100. Halliday, M.A.K. 1967. Notes on Transitivity and Theme in English, Part 1. Journal of Linguistics 3:37-81. Hankamer, J. 1973. There-Insertion and the Definiteness of Deletion Targets. Paper presented at NELS 4. Hannay, M. 1985. English Ezistentials in Functional Grammar. Dordrecht. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation, University of Massachusetts at Amherst. Published 1989 by Garland Press, New York. Heim, I. 1987. Where Does the Definiteness Effect Apply? In The Representation of (In)definiteness, ed. E. Reuland and A.G.B. ter Meulen. 21-42. Cambridge: MIT Press.

References

209

Higginbotham, J. 1987. Indefinites and Predication. In The Representation of {In)definiteness, ed. E. Reuland and A.G.B. ter Meulen. 43-70. Cambridge: MIT Press. Hoeksema, J. 1989. Review of Reuland and ter Meulen 1987. Language 65:115-125. Holmback, H. 1984. An Interpretive Solution to the Definiteness Effect Problem. Linguistic Analysis 3:195-215. Huang, J. 1982. Logical Relations in Chinese and the Theory of Grammar. Doctoral dissertation, MIT. Jackendoff, R. 1990. Semantic Structures. Cambridge. Jenkins, L. 1975. The English Ezistential. Tiibingen. Jespersen, 0. 1927. A Modern English Grammar on Historical Principles. London. Johnson, K. 1991. Object Positions. Natural Language and Linguistic Theory 9:577-633. Jones, C. 1985. The Syntaz and Thematics of Infinitival Adjuncts. Doctoral dissertation, University of Massachusetts at Amherst. Kadmon, N. 1987. On Unique and Non- Unique Reference and Asymmetric Quantification. Doctoral dissertation, University of Massachusetts at Amherst. Published 1992 by Garland Press, New York. Kamp, H. 1981. A Theory of Truth and Interpretation. In Truth, Interpretation, and Information, ed. J. Groenendijk et al. Dordrecht: Foris. Reprinted in 1984. Keenan, E. 1987. A Semantic Definition of "Indefinite NP". In The Representation of (In)definiteneu, ed. E. Reuland and A.G.B. ter Meulen. 286-317. Cambridge: MIT Press. Keenan, E., and J. Stavi. 1986. A Semantic Characterization of Natural Language Determiners. Linguistics & Philosophy 9:253-326.

210

References

Kimball, J. 1973. The Grammar of Existence. In Proceedings of CLS 9, ed. C. Corum et al., 262-270. Kitagawa, Y. 1986. Subjects in Japanese and English. Doctoral dissertation, University of Massachusetts at Amherst. Kratzer, A. 1989. Stage-Level and Individual-Level Predicates. Reprinted 1995 in Carlson, G. and F .J. Pelletier, eds., The Generic Book, University of Chicago Press, Chicago. Kuno, S. 1971. The Position of Locatives in Existential Sentences. Linguistic Inquiry 2:333-378. Kuno, S. 1972. Subject Raising in Japanese. Papers in Japanese Linguistics 1. Kuroda, S.-Y. 1988. Whether We Agree or Not. Investigationes 12:1-7.

Lingvisticae

Ladusaw, W.A. 1979. Negative Polarity as Inherent Scope Relations. Doctoral dissertation, The University of Texas at Austin. Published 1980 by Garland Press, New York. Ladusaw, W.A. 1982. Semantic Constraints on the English Partitive Construction. In Proceedings of WCCFL 1. Ladusaw, W.A. 1992. Expressing Negation. In Proceedings of SALT II, ed. C. Barker and D. Dowty, 237-259. Ladusaw, W.A. 1994. Thetic and Categorical, Stage and Individual, Weak and Strong. In Proceedings of SALT IV, ed. M. Harvey and L. Santelmann, 220-229. Lakoff, G. 1987. Women, Fire, and Dangerous Things. Chicago. Larson, R. 1990. Double Objects Revisited: Reply to Jackendoff. Linguistic Inquiry 21:589-632. Lasersohn, P. 1988. A Semantics for Groups and Events. Doctoral dissertation, Ohio State University. Lasnik, H. 1992. Case and Expletives: Notes Toward a Parametric Account. Linguistic Inquiry 23:381-405.

References

211

Legendre, G. 1990. French Impersonal Constructions. Natural Language and Linguistic Theory 8:81-128. Link, G. 1987. Algebraic Semantics for Event Structure. University of Munich.

Ms.,

Lumsden, M. 1988. Ezistential Sentences: Their Structure and Meaning. London. Lyons, J. 1967. A Note on Possessive, Existential, and Locative Sentences. Foundations of Language 3:390-396. May, R. 1985. Logical Form. Cambridge. McCawley, J. 1981. Everything that Linguists have Always Wanted to Know About Logic* *but were ashamed to ask. Chicago. McCoard, R. 1978. The English Perfect: Tense-Choice and Pragmatic Inferences. Amsterdam. McNally, L. 1992a. Locative Relational Nouns in Copular and Existential Contexts. Paper delivered at the LSA Annual Meeting, Philadelphia, PA. McNally, L. 1992b. VP-Coordination and the VP-Internal Subject Hypothesis. Linguistic Inquiry 23:336-341. McNally, L. 1994. Adjunct Predicates and the Individual/Stage Distinction. In Proceedings of WCCFL 12, ed. E. Duncan et al., 561-576. McNally, L. 1995a. Existential Sentences Without Existential Quantification. Ms., Universitat Pompeu Fabra, Barcelona. McNally, L. 1995b. Stativity and Theticity. To appear inS. Rothstein, ed., volume of papers from the 1993 Bar-Ilan Conference on Events and Grammar, Kluwer, Dordrecht. Milsark, G. 1974. Ezistential Sentences in English. Doctoral dissertation, MIT. Published 1979 by Garland Press, New York. Milsark, G. 1977. Peculiarities of the Existential Construction in English. Linguistic Analysis 3:1-29.

212

References

Milsark, G. 1990. Review of Lumsden 1988. Language 66:850-854. Moltmann, F. 1995. Intensional Verbs and Quantifiers. To appear in Nat ural Language Semantics. Moro, A. 1991. The Raising of Predicates: Copula, Expletives, and Existence. MIT Working Papers in Linguistics 1:119-181. Nathan, G. 1978. The Syntaz and Semantics of the English Ezistential Construction. Doctoral dissertation, University of Hawaii. Parsons, T. 1990. Events in the Semantics of English. Cambridge. Partee, B.H. 1987. Noun Phrase Interpretation and Type-Shifting Principles. In Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, ed. J. Groenendijk et al. 115-43. Dordrecht: Faris. Partee, B.H., and M. Roath. 1983. Generalized Conjunction and Type Ambiguity. In Meaning, Use, and Interpretation of Language, ed. R. Bauerle et al. 361-383. Berlin: de Gruyter. Perlmutter, D., and A. Zaenen. 1984. The Indefinite Extraposition Construction in Dutch and German. In Studies in Relational Grammar 2, ed. D. Perlmutter and C. Rosen. Chicago: University of Chicago Press. Pollard, C., and I. Sag. 1994. Head-Driven Phrase Structure Grammar. Chicago. Postal, P. 1974. On Raising. Cambridge. Postal, P. 1992. Overlooked Extraction Distinctions. Ms., IBM T.J. Watson Research Center. Prince, E. 1981a. On the Inferencing of Indefinite-this NPs. In Elements of Discourse Understanding, ed. A. Joshi et al. 231250. New York: Cambridge. Prince, E. 1981b. Toward a Taxonomy of Given/New Information. In Radical Pragmatics, ed. P. Cole. 223-55. New York: Academic Press.

References

213

Prince, E. 1988. The Discourse Function of Yiddish Expletive ES and Subject-Postposing. IPrA Papers in Pragmatics 2:176194. Prince, E. 1992. The ZPG Letter: Subjects, Definiteness, and Information-Status. In Discourse Description: Diverse Linguistic Analyses of a Fund Raising Tezt, ed. S. Thompson and W. Mann. Philadelphia: John Benjamins. Rando, E., and D.J. Napoli. 1978. Definites in There-Sentences. Language 54:300-313. Rapoport, T. 1991. Adjunct Predicate Licensing and D-Structure. In Syntaz and Semantics 25: Perspectives on Phrase Structure, ed. S. Rothstein. New York: Academic Press. Raposo, E., and J. Uriagereka. 1990. Long Distance Case Assignment. Linguistic Inquiry 21:505-537. Roberts, C. 1987. Modal Subordination, Anaphora, and Distributivity. Doctoral dissertation, University of Massachusetts at Amherst. Published 1990 by Garland Press, New York. Roberts, C. 1993. Uniqueness Presuppositions in Definite Noun Phrases. Ms., Ohio State University. Roberts, I. 1988. Predicative APs. Linguistic Inquiry 19:703-710. Rooth, M. 1985. Association with Focus. Doctoral dissertation, University of Massachusetts at Amherst. Ross, J .R. 1974. There, There, (There, (There, (There))). Proceedings of CLS 10, ed. M. LaGaly et al., 569-587.

In

Rothstein, S. 1983. The Syntactic Forms of Predication. Doctoral dissertation, MIT. Safir, K. 1982. Syntactic Chains and the Definiteness Effect. Doctoral dissertation, MIT. Safir, K. 1985. Syntactic Chains. Cambridge: Cambridge University Press.

214

References

Safir, K. 1987. What Explains the Definiteness Effect? In The Representation of (In)definiteness, ed. E. Reuland and A.G.B. ter Meulen. 71-97. Cambridge: MIT Press. Siegel, M.E.A. 1976. Capturing the Adjective. Doctoral dissertation, University of Massachusetts at Amherst. Steedman, M. 1990. Gapping as Constituent Coordination. Linguistics & Philosophy 13:207-263. Stowell, T. 1978. What Was There Before There Was There. In Proceedings of CLS 14, ed. D. Farkas et al., 458-471. Stowell, T. 1983. Subjects Across Categories. Review 2:285-312.

The Linguistic

Stowell, T. 1991. Small Clause Restructuring. In Principles and Parameters in Comparative Grammar, ed. R. Freidin. 182-218. Cambridge: MIT Press. Stump, G. 1985. The Semantic Variability of Absolute Constructions. Dordrecht: Kluwer. Van Geenhoven, V. 1992. Noun Incorporation from a Semantic Point of View. In Proceedings of BLS 18: Parasession on the Place of Morphology in a Grammar, ed. L. Buszard et al., 453467. Ward, G. 1995. A Comparison of Postposed Subjects in English and Italian. Paper delivered at the LSA Annual Meeting, San Diego, CA. Ward, G., and B. Birner. 1995. Definiteness and the English Existential. Language 71:722-742. Wilkinson, K. 1988. The Semantics of the Common Noun Kind. In Proceedings of the Tubingen Conference on Generics, ed. M. Krifka. Reprinted 1995 in Carlson, G. and F.J. Pelletier, eds., The Generic Book, University of Chicago Press, Chicago. Wilkinson, K. 1991. Studies in the Semantics of Generic Noun Phrases. Doctoral dissertation, University of Massachusetts at Amherst.

References

215

Williams, E. 1980. Predication. Linguistic Inquiry 11:203-238. Williams, E. 1983a. Against Small Clauses. Linguistic Inquiry 14:287-308. Williams, E. 1983b. Semantic vs. Syntactic Categories. Linguistics f1 Philosophy 6:423-446. Williams, E. 1984. There Insertion. Linguistic Inquiry 15:131-153. Williams, E. 1994. MIT Press.

Thematic Structure in Syntaz.

Cambridge:

Woisetschlaeger, E. 1983. On the Question of Definiteness in 'An Old Man's Book. Linguistic Inquiry 14:137-154. Zimmermann, T.E. 1992. On the Proper Treatment of Opacity in Certain Verbs. Natural Language Semantics 1:149-179. Zucchi, A. 1995. The Ingredients of Definiteness and the Definiteness Effect. Natural Language Semantics 3:33-78.

Index Abbott, B. 31, 128, 193-194 Abney, S. 40 African American Vernacular English 84 agreement 21 Aissen, J. 8, 199 Andrews, A. 48 antecedent government 62, 7172 Bach, E. 74 bare nouns, interpretation of 92 Barker, C. 11, 80, 148 Barwise, J. 9, 24-25, 32, 36, 40, 42, 51, 195 Belletti, A. 41 Binding Theory 21 Birner, B. 33, 41, 201 Burt, M. 14 Burzio, L. 18, 41 Carlson, G. 4, 7, 13, 55, 78, 86, 110, 125, 138-140, 144, 146, 149, 161, 163, 192 Carrier, J. 46, 74-75, 187 Case Filter 21 Catalan 40, 128, 192 Chierchia, G. 4, 5, 59, 78-79, 81, 86-87, 92, 126, 145, 185

Chomsky, N. 41, 75 Chung, S. 7, 75 co-conservative function 27, 42 Condoravdi, C. 34, 80, 161, 188 conservative function 26 contact clause 84, 135-136, 144 context change potential 8081, 98, 99 of atomic sentences 99 of complex sentences 101 of existential sentences 105 of negated sentences 103 of quantificational sentences 101 control 155, 168 Cooper, R. 9, 24-25, 32, 36, 40, 42, 51, 195 coordination 179 copular construction 5, 7, 12, 14-15,17,20,22, 73, 82, 84, 92, 135-137, 144, 192, 202 predicate restriction in 38 de Hoop, H. 41 definite DPs 116 and familiarity 117-118 interpretation of 116 uniqueness presupposition of 118 217

218

definite DPs (cont'd) vs. necessarily quantificational DPs 121 definiteness restriction 4, 68, 17, 20, 42, 77-78, 106, 116, 126, 172 and quantificational DPs 108 Barwise and Cooper's account of 25 Keenan's account of 28 Milsark's account of 30 non unified account of 36, 126-127, 129 pragmatic accounts of20, 33, 127 semantic accounts of 20 syntactic accounts of 20, 21 Diesing, M. 162, 187 discourse model 80, 98 discourse new discourse referent 126 Discourse Representation Theory 5 Doherty, C. 84, 135-136, 144 Dowt~ D. 75, 159, 168 dynamic interpretation 86, 90 Empty Category Principle 50, 62, 64, 70-71, 75 Enc;, M. 41 event argument 158, 160, 162 event structure 158 existence assertion 22 existential construction adjunct predicate in 195, 202 and passives 176 bare existentials 169

Indez

Barwise and Cooper's semantics for 25 basic interpretation for 104 context change potential for 105 coordination in 179 cross-linguistic variation in 6, 41, 73, 126-127, 191 definite DP in 193-194 event reading 152, 172, 177, 179-180 extraction from 17, 53, 55, 65-66, 68, 172, 179 felicity condition on 120 Heavy-NP Shift in 56 kind terms in 35 list existentials 8, 30, 128, 193 Milsark's semantics for 24 necessarily quantificational DPs in 29-30 novelty condition on 201 postverbal DP in 5 predicative phrase in 7, 12, 20, 46, 66, 69, 71-72, 151, 168-169, 194 DP as 55 optionality of 55 proposed semantics for 3 scopal properties of 8283 small clause analysis of 15, 20, 23, 46, 57, 65, 68, 70, 72 tests for 58 VP-deletion in 56 existential force 106

Indez

existential function 42 existential predicate 103 expletive-argument chain 21 Farkas, D. 10, 79, 87 Fernald, T. 74 File Change Semantics 5 Fodor, J.D. 4, 78 free adjuncts 47-48, 188 Frege, G. 78-79 Gawron, J.-M. 153, 161 Goldberg, A. 74 Green, G. 46, 48, 152 Grice, H.P. 196 Groenendijk, J. 144 Halliday, M.A.K. 46, 152 Hankamer, J. 193-194 Hannay, M. 33, 41, 128, 172, 188, 194, 203 head government 63, 69-72 definition of 62 hearer new discourse referent 33, 126, 133 Heim, I. 5, 33-34, 80, 94-95, 99, 101, 118-119, 131132,138-139,146,201 Hiberno-English 84 Higginbotham, J. 4, 24, 30 Hoeksema, J. 41 hold time function 154 Holmback, H. 11 Huang, J. 66 indefinite DPs 81, 94 individual level predicates 7, 13,157-159,169,202 Condoravdi's analysis of 160, 163, 166 Kratzer's analysis of 162 locative characteristics 161162

219 stage level behavior 160 temporal characteristics 161 information unit 87-90, 98 intensional predicates 5 intensionality 78-79, 87, 136 intersective determiners 113 recursive definition of 27 interpretation for 114 intersective function 27, 42 intersectivity 114 inverse linking 31, 134, 148 ent operator 89 pred operator 89 there 146 it-clefta 173 Jackendoff, R. 152-153, 187 Jenkins, L. 14,38, 50, 74,201 Jespersen, 0. 84 Johnson, K. 56 Jones, C. 74 Kadmon, N. 116, 146 Kamp, H. 5, 144 Kayne, R. 64 Keenan, E. 9, 12, 15, 24-26, 29, 37, 42, 46, 51, 53, 113 Kimball, J. 40, 188 kind level predicates 125 kind terms 79, 109, 122, 130, 132 and discourse anaphora 124, 131 and scope 133 covert indefinite analysis of 124, 130 kind-denoting expressions 11 kinds, natural 78 Kitagawa, Y. 49 Kratzer, A. 161-163, 187

220

Kuno, S. 40, 59 Kuroda, S.-Y. 49 Ladusaw, W.A. 10, 43, 52, 54, 80, 83, 146, 203 Lakoff, G. 40, 202 Larson, R. 56, 134 Lasersohn, P. 153-154, 187 Lasnik, H. 41 Legendre, G. 41 Link, G. 153, 187 location function 154 locative predicates 163 locative there 202 logical form formation rules 95 logical form 96, 118 Lumsden, M. 7, 8, 12, 24, 33, 35,54, 109,120,126, 128, 130, 194 Lyons, J. 40, 78 May, R. 31 McCloskey, J. 75 McCoard, R. 175 McNally, L. 49, 145, 187 Milsark, G. 7-8, 11, 12, 14, 20, 24-26, 30, 33, 3738, 41, 51, 106, 130, 152, 173, 180, 193194, 199 modal operators 5 Moltmann, F. 40, 149 Moro, A. 42 Napoli, D.J. 8, 193 narrowest scope effect 107, 133 Nathan, G. 41 necessarily quantificational DP 5, 78, 108, 107, 110, 115,120,126-128,130, 145

Indez

negation 102 negative polarity items 54 nominalized function 4-5, 7880, 104 nonreferential DPs 4-5 nonreferentiality 78, 135 Parsons, T. 162, 175 Par~ee, B. 79, 91 partitive DPs 10 past participles 175 as DP modifiers 178 entailments of 175 in existentials 176 Perlmutter, D. 41 Pollard, C. 15, 40-42, 46 possessive DPs 11 possible world 90 Postal, P. 59, 149 predicate nominal 78-79, 82, 85-86, 92, 94, 132133, 136-137, 144 predicate restriction 7, 12, 17, 20, 36, 42, 73, 153, 170, 176, 202 predicative adjuncts 47, 4950, 151 adverbial modifiers of 60, 64 and past participles 172, 174 as event predicates 153 extraction from 66 individual level 159 implicature of 159-160, 164, 166, 170 Rapoport's rule for 158 semantic restrictions on 158 semantics for 152, 156

Indez

VP 173 predicative complements, 47, 50, 160 presentational there 199-200, 204 presupposition 130-131, 195 Prince, E. 10, 24, 33, 80, 116, 120, 126, 147, 201 pronouns 119 interpretation of 119 proper names 119, 128 interpretation of 119 property theory 78-79, 86-87 property 79 entity correlate of 4, 87 extension of, defined 93 proposition 79, 87, 90, 145 PT 87 interpretation for 90 meaningful expressions of 88 model for 89 ontology of 87 quantifiers 24 strong 25 weak 24 Randall, J. 46, 74-75, 187 Rando, E. 8, 193 Rapoport, T. 38, 46, 49, 50, 55,151-152,158,160, 187 Raposo, E. 18-19 reconstruction 130-134, 137, 140 reduced clefts 38 relational nouns 11, 121, 182 relative clauses 85, 137 amount relatives 138-140

221

reconstruction analysis of 139 relative pronouns, and semantic sort 137 Relativized Minimality, 61 resultatives 46, 74-75, 153 Rizzi, L. 49, 61, 69 Roberts, C. 5, 134 Roberts, I. 46, 48 Rooth, M. 42, 91, 198 Ross, J .R. 199 Rothstein, S. 49 Safir, K. 4, 14-15, 21-22, 24, 30, 41, 78, 139, 193194 Sag, I. 4, 15, 40-41, 46, 78 scope 108 Siegel, M.E.A. 12 spatial overlap 153, 187 specificity 10 stage level predicates 13, 55, 157-159, 161 Stavi, J. 24, 26, 42, 113 Steedman, M. 75 Stockhof, M. 144 Stowell, T. 14-15, 46, 58-59, 64 strong novelty 34-35 Stump, G. 47, 188 temporal overlap 153, 187 temporally sensitive nominals 180 theta government 69, 72 truth predicate 89 Turner, R. 4-5, 78-79, 81, 8687, 185 uniqueness, 117 Uriagereka, J. 19 Van Geenhoven, V. 87

222 Ward, G. 33, 41, 204 wh-clefts 188 Wilkinson, K. 12, 110, 124, 130, 132, 147 Williams, E. 5, 7, 30, 36, 5051,55-56,66,79,155, 173, 187 Woisetschlaeger, E. 11 Zaenen, A. 41 Zimmermann, T. 87, 149 Zucchi, A. 24, 33, 40, 43, 46, 120

Inde-z

ROUTLEDGE LIBRARY EDITIONS: SEMANTICS AND SEMIOLOGY

Volume 10

GERMAN TEMPORAL SEMANTICS

C\ Taylor & Francis ~Taylor & Francis Group

http://tayl ora ndfra n ci s.com

GERMAN TEMPORAL SEMANTICS Three-Dimensional Tense Logic and a GPSG Fragment

JOHN A. NERBONNE

First published in 1985 by Garland Publishing, Inc. This edition first published in 2017 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1985 John A. Nerbonne All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-69750-8 978-1-315-52029-2 978-1-138-69031-8 978-1-315-53702-3

(Set) (Set) (ebk) (Volume 10) (hbk) (Volume 10) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

German Temporal Semantics Three-dimensional Tense Logic and a GPSG Fragment

John A. Nerbonne

Garland Publishing, Inc. • New York & London

1985

Library of Congress Cataloging in Publication Data Nerbonne, John A., 1951German temporal semantics. (Outstanding dissertations in linguistics) Thesis (Ph.D.)-Ohio State University, 1983. Bibliography: p. 1. German language-Temporal constructions. 2. German language-Semantics. 3. German language-Grammar, Generative. 4. Tense (Logic) I. Title. II. Series. PF3301. N47 1985 435 85-11610 ISBN 0-8240-5436-9 (alk. paper)

© 1985 by John A. Nerbonne All rights reserved

The volumes in this series are printed on acid-free, 250-year-life paper. Printed in the United States of America

to my grandmother, Margaret O'Grady Joyce and to the memory of my grandfather, John Joseph Joyce

Q

~-

Taylor & Francis Taylor & Francis Group

http://taylorandfrancis.com

ACKNOWLEDGEMENTS All art is a collaboration, and there is little doubt that in the happy ages of literature, striking and beautiful phrases were as ready to the story-teller's hand as the rich cloaks and dresses of his time. - J.M.Synge, Preface to Playboy of the Western World Alfred Schopf fi rst interested me in the structure of temporal expressions and the range and scope of issues involved in their analysis. My advisor, David Dowty, first demonstrated the subtlety and explanatory power of model-theoretic semantics as a tool in this analysis. My approach, as well as innumerable specific improvements in this work, are due to him. 11 se Lehi ste and Arnol d Zwicky al so critici zed my work productively. Erhard Hinrichs was sharp about theoretical COJl1l1ents and generous in providing German examples and judgements about their meaning. Robin Cooper, Ronald Hendricks, Brian Joseph, Barbara Partee and Greg Stump commented benefi ci ally on one or another aspect of the analysis. The local intellectual infrastructure is infusive and must have been the source of most of whatever might be "striking and beautiful" below; I particularly appreciated Bill Boslego, Doug Fuller, Mike Geis, Jean Godby, and Rex Wallace.

Q

~-

Taylor & Francis Taylor & Francis Group

http://taylorandfrancis.com

TABLE OF CONTENTS INTRODUCTION ••..•.•••••.••.•.•••••••••.••..••••.•••

Page 1

Chapter 1. A Reichenbachian Tense Logic...............

2

Introduction........................... Triple Dependence...................... What is Reference Time ................ The Logic.............................. The Preterite and the Indexical Interpretation of Reichenbach............ 1.5.1 The Motivation for Indexical Treatments of Tense...................... 1.5.2 Vagueness and Indexicality........... 1.5.3 Indefinite Reference to Time......... 1.6 Temporal Reference in Connected Discourse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 The Treatment of Aktionsarten........ 1.6.2 Time in Connected Discourse.......... 1.6.3 Some Special Uses of the Preterite... 1.7 The Need for Three Indices............. 1.7.1 The Need for at Least Three Indices.. 1.7.2 More than Three Indices .........•... Notes .............•.......................

2 3 4 14

32 35 36 38 40 40 44 49

2. A Tense Logical Sketch of German...........

56

2.1 Frame Adverbials....................... 2.2 Baeuerle and Stechow's Analysis........ 2.3 The German Present Tense............... 2.3.1 Semantics of the German Present Tense 2.3.2 Sample Derivation of Complex Truth Conditions.......................... 2.3.3 Why Atelics are Presumed to Refer to Speech Time......................... 2.3.4 Kratzer's Speech Time Pragmatics..... 2.3.5 The Nonambiguity of the Present Tense

56 64 68 68

1.1 1.2 1.3 1.4 1.5

16 16 24 29

71 73 79 81

2.3.6 Conclusion........................... 2.4 Duratives.............................. 2.5 Frist Adverbials....................... 2.6 (Temporal) schon....................... 2.6.1 PreliminarTeS:'". .•.• ... .•. ...•. ••..... 2.6.2 Other Uses of schon.................. 2.6.3 The Truth Conditions of Temporal schon............................... 2.6.4 schon with Telic Aktionsarten........ 2.7 Summary of Semantic Rules •••.....••.••.

86 86 93 96 96 97 102 108

Notes ...•..•..........•.........•.•.......

3. A Fragment of German.......................

124

3.1 GPSG: Formalism and Notation........... 3.2 German Syntax.......................... 3.2.1 Constituents of the Sentence......... 3.2.2 Fronting (of Several Kinds).......... 3.2.3 Phantoms and Some Recalcitrant Sorts of Fronting......................... 3.3 Basic Rules............................ 3.3.1 Features for Complements............. 3.3.2 Separable Prefix Verbs............... 3.4 Fronting Formalized.................... 3.5 The Analyses of Jean Fourquet.......... 3.6 Two Strategies for the Treatment of Temporalia (in GPSG)................ 3.6.1 Tense as a Verb Operator............. Excursus: On Duratives (and Frequentatives) as Complements to the Verb.......... 3.6.2 Tenses as Phrasal Operators.......... 3.7 Metarules for Temporalia............... 3.7.1 Duratives and Frist Adverbials....... 3.7.2 Frame Adverbials (that Modify Reference Time...................... 3.8 Some Derivations.................... ... 3.9 The Syntax of Temporal Schon........... Notes ...................... ~...........

125 130 130 136 139 146 146 159 169 175 175 177 180 183 187 187 190 199 210 218

4. Extending the Fragment..........................

224

4.1 The Perfect Tenses..................... 4.1.1 The Forms of the Perfect............. 4.1.2 The Meaning of the Perfect Tenses.... 4.1.3 The Syntax of the Perfect............ 4.1.4 A Sample Derivation.................. 4.2 Adverbials which Modify Event Time..... 4.3 Noch................................... 4.3.r-NOntemporal Noch..................... 4.3.2 Immer Noch .. -:-:-:-:-.....................

224 224 225 236 240 242 254 254 257

4.3.3 The Syntax of Noch................... 4.4 Passives •...•...--:-::-:................... 4.4.1 The Subjectlessness of Impersonal Passives............................ 4.4.2 The Lexical Nature of the Passive.... 4.4.3 A Formulation of the Rule............ Notes. . . . . • . • . . . . • . • . . • . . . . . • . . . . . . . . . . • . • .

259 261 262 266 271 288

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

290

Q

~-

Taylor & Francis Taylor & Francis Group

http://taylorandfrancis.com

Introduction This dissertation analyzes temporal meaning in German. The framework is that of a model-theoretic semantics. more specifically one incorporating a multi-dimensional tense logic. Chapter 1 presents this framework and is sufficient for those interested only in the general theory of temporal meaning. It argues that three dimensions are optimal for the description of natural language temporalia. giving rise to a Reichenbachian system for temporal description. Special attention is paid to the definite interpretations of tense noted in Partee (1973). Although it is not the purpose of the investigation. it turns out that the interpretation of Reichenbach's speech. event, and reference times as indices within model theory explains several otherwise unmotivated aspects of Reichenbach's remarks on tense. Chapter 2 applies this theory to the analysis of temporal meaning in German. Frame adverbials, the Present and Past tenses, duratives, aspectual adverbials using in, and the adverbial particle schon are examined. None of the last three were included in Baeuerle's (1979) tense logical analysis of German, the most extensive (and best) to-date, and both of the first two are given novel analyses. The section on schon uncovers data which has escaped previous notice. Chapter 3 provides a formal syntax to bear the semantic analysis proposed in 2. This is of some purely syntactic interest because General i zed Phrase Structure Grammar hasn't yet confronted German extensively and because it suggests one innovation, the use of complement features, to treat VP fronting. The chapter may also be of interest because it demonstrates how temporal semantics may be incorporated within GPSG with essentially no new grammatical apparatus. Chapter 4 explores syntactic and semantic extensions of the fragment, showing how the Perfect, the particle noch, the Passive, and a distinct reading of frame adverbials may be accommodated. - 1 -

A Reichenbachian Tense Logic 1.1 Introduction The semantics of tense and other temporal expressions, involving as

it does modification,

recursion,

contextual

dependence,

lexical

variety, crucial scope relationships, and the interaction of elements in several grammatical categories is perhaps as rich and problematic as any in the field of natural language semantics.

Model-theoretic

semantics allows precise investigation using fairly simple mathematical

techniques, and there is,

work upon which to build.

finally,

no lack of very competent

This is, in short, a most attractive field

of study. This work proposes a semantics for the description of temporal expressions inspired largely by Hans Reichenbach's brief remarks on the Engl ish

tenses,

and

the

ins i ghts of a number of contemporary

researchers, including Partee (1973), Kuhn (1979), Baeuerle (1979), and Enc (1981), that tenses behave semantically rather like definitely referring (nominal) expressions.

In spite of the attention paid to

it, the parallel between tense and definite nominal reference, it is argued, has been i nsuffi c i ently appreci ated--both with respect to its extent, and with respect to its consequences. The semantic theory presented in this first chapter is inspired by Reichenbach (1947), and it employs his three-way distinction among times

relevant

to

semantic

interpretation--the well-known

event and reference times introduced by Rei chenbach.

speech,

The semanti cs

doesn't simply assume Reichenbach's system, but interprets it (and is somewhat selective about certain inexplicit aspects of his temporal descriptions).

In

the

present interpretation

- 2 -

speech,

event,

and

3

reference times are viewed as times to which deictic reference rmy be made--effec t i ng the pa ra 11 eli sm to defi ni te nomi na 1 reference mentioned above. The proposed semantics is illustrated in Chapter 2 by an extended semantical sketch of German temporal reference. The proposed system for temporal semantics will be tested on an extensive, but necessarily 1 imited range of temporal phenomena--incl uding tense, temporal adverbials and particles, and the inherent temporal structure of verbs (Aktionsarten). All of these expressions are incorporated into a forma 1 fragment (i n General i zed Phrase Structure Gramma r) in Chapter 3. Chapter 4 presents some semantico-syntactic extensions of the system developed in the first three chapters. Ultimately, if it is to be adopted, the semantics proposed here must allow cogent analyses of all temporal reference, including not only the phenomena named above, but also temporal clauses, sequence of tense restri cti ons, and aspect. The system hasn t been tested on these phenomena to- da te, thou gh they do not seem to present speci a 1 difficulties. I

1.2 Triple Dependence Reichenbach is to be credited for introducing the idea that the meani ng of some tenses and temporal express ions depends not only on the time of speech, and the time at which an event takes place (or is reported to take place), but also on a third time, the reference time. In this chapter, I suggest a semantical formalization of Reichenbach's triple dependence and outline some further crucial background assumptions to a system using this formalization. Chapter 2 then argues that the semantics of temporal reference in German, in particular that of adverbs, and that of adverbial particles such as schon depends on the employment of reference time as a theoretical tool. (In the treatment proposed, reference time functions as one of three dimensions in a tense logic; it is otherwise the same concept introduced by Reichenbach.)

4

1.3 What is Reference Timew The concept of reference time has puzzled some researchers. Rei chenbach di sti ngu i shed speech time s, event time e and reference Let us examine these as Reichenbach applied them to the time r. following example: In 1678 the whole face of things had changed ... eighteen years of misgovernment had made the ... majority desirous to obtain security at any risk. The fury of their returning loyalty had spent itself in its first outbreak. In a very few months they had hanged and halT-hanged, quartered and emboweled, enough to satisfy them. The Roundhead party seemed to be not merely overcome, but too much broken and scattered ever to rally again. Then commenced the reflux of public opinion. The nation began to find out to what a man it had entrusted without conditions all its dearest interests, on what a man it had lavished all its fondest affection. (Reichenbach, 1947:288f) Speech and event time are easily recognizable. Speech time is simply the time of utterance (read here: writing), while the time of the various episodes described constitutes event time. As to reference time, let us note Reichenbach's remarks: The point of reference is here the year 1678. Events of this year are related in the simple past, such as the commencing of the reflux of public opinion, and the beginning of the discovery concerning the character of the king. The events preceding this time point are given in the past perfect, such as the change in the face of things, the outbreaks of cruelty, the nation's trust in the king. (Reichenbach, 1947:289) An event is thus seen not only from the vantage poi nt of the speech time: it is al so seen from time of reference. I) It is the time of reference which distinguishes the simple past from the past perfect. Each recounts episodes which are prior to speech time, but the episodes re 1ayed in the past perfect are additi ona lly pri or to the time of reference. (We will accept Reichenbach's characterization of this distinction, and we try to provide additional support for it in principles for analyzing contextual dependence in 1.6.2.) A reference time may be explicitly identified, e.g. as 1678 in the

5

passage above, or it may be provided e.g. by a superordinate clause, as in the sentence below: After he had eaten everything, he said good-bye. The event time of the subordinate clause is the time at which he ate. The reference time of this same clause is provided by the event time of the main clause: it is the time of his saying good-bye. Note that event time is prior to reference time here, and that the past perfect his used, just as it was in main c1uases in Reichenbach's example. The configuration of speech, event and reference times is crucial, not syntactic structure. (Cf. 1.7.2.) Reference time may be neither explicit nor provided by superordinate clauses, but given only by the context, as Reichenbach noted. He commented that in the sentence Peter had gone: ... it is not clear which time point is used as the point of reference. This determination is rather given by the context of speech. In a story, for instance, the series of events recounted determines the point of reference, which in this case is in the past, seen from the point of speech; some individual events lying outside this point are then referred, not directly to the point of speech, but to this point of reference determined by the story. (Reichenbach (1947:288)) Two aspects of Reichenbach's proposal will be exploited below. First, reference time is subject to pragmatic influence. Second, and more specifically, reference time may be given by the previous discourse. Given this rough characterization of the notions of speech, event and reference times, we note that it was Reichenbach's strategy to ascribe one configuration of these times to each tense. For example, he lists the following (p.297):

6

Past Perfect Simple Past Present Perfect Present Simple Future Future Perfect

E - R- S R,E - S E - S,R S,E,R S,R - E S - E- R

E, R, and S stand for speech, event and reference times. A comma between two times stands for simultaneity, while the hyphen means that the left time temporally precedes the right. Impl icit in Reichenbach is surely the position that no more than three times are involved in the interpretations of any tense. I shall accept a slightly more general version of this position: Maximally Triple Dependence: No more than three times are involved in the interpretation of any temporal expression. The generalization is from "tense" to "temporal expression." The notion "times" is admittedly still vague above. It may be made precise in 1.4 through the notion "temporal index," and Maximally Triple Dependence will be seen to follow as a trivial consequence of the position that tense 1ogi c for natural 1anguage are three dimensional. There is some reason, however, to reject other positions which also seem impl icit in Reichenbach's analyses. Returning to the tense schemata above, it is perhaps remarkable that every tense specifies a linear configuration of all three times: in no case is a tense regarded as specifying a relation among less than three times, and never does it appear to have seemed necessary' to Rei chenbach to resort to non 1i near confi gurati ons of S, E, and R. On the contrary, however, the Perfect infinitive seems to require only that E precede R, and is indifferent to speech time, as the sentences below might suggest:

7

He seems to have left He seemed to have left He will seem to have left

She believes him to have left She believed him to have left She'll believe him to have left

This isn't the point at which one even could argue for any semantic rule in detail; we've simply developed too little of the overall apparatus for any rule to be justified in detail. But if we accepted Reichenbach's specification of the Present, Simple Past and Simple Future tenses (for the purposes of this illustration), then it might be seen that the only relationship with which the Perfect Infinitive may consistently be associated is that of the event time (of the VP to which it is attached) preceding reference time. The following schemata illustrate how the Perfect Infinitive specifies its times: seems believes S,R,E

1

E' - R' leave

seemed believed ER- S

t' -

E' R' leave

wi 11 seem will believe S,R - E

,

E' - R' leave

At least in the complements of the verbs seem and believe, the event time of the matrix clause is used as reference time in the complement. The Perfect Infinitive then marks the time at which the episode reported in the compl ement cl ause takes pl ace--regardl ess of speech time. The remarks above cannot be construed as defended analysis of the temporal import of the Perfect Infinitive--but only as an indication of the possible wisdom of allowing temporal elements to specify less than an exhaustive relation among speech, event, and reference times. Similarly, there are tenses which seem to specify a nonlinear relation among the speech, event and reference times. This is perhaps Relations are linear, of course, iff they are a bit surprising. transitive, irreflexive, and connex. Clearly the points of time are ordered 1inearly under '~,' so that it may be surprising that some

8

tense specify times in a nonlinear fashion. The key is connexity. Recall that a relation R is connex in a set S iff

The connexity axiom disallows then situations such as the following:

where i1 and i3 are distinct, but unordered with respect to each other. Note the following use of the Future Perfect: I don't know whether he's left. tomorrow, however.

He certainly will have left by

The pair of sentences is sensible enough, but this indicates that the event time of the Future Perfect may either foll ow or precede speech time. Either of the following configurations is thus compatible with the Future Perfect: S - E - R (Reichenbach's configuration) E - S - R

(Cf. Comrie, 1981:28.) The Future Perfect requires then that speech, event and reference time be ordered thus:

This relation is nonlinear. Let us avoid one potential confusion: the time of utterance and

9

the time at which he left (in the sentence above) will certainly be ordered with respect to each other (i n any part i cu 1a r use of tha t sentence). This isn't the point, which is rather that the Future Perfect cannot specify that speech and event time always stand in one or the other relation. The tense cannot require a particular linear relationship among the three times. We therefore will not follow Reichenbach in having each tense exhaustively specify a linear relation among the three times. It is not explicit in Reichenbach in any case that one ought to do so, though it certainly was his practice. I would like now to turn to an area where Reichenbach will be followed most exactly; this concerns his conception of reference time. Let us be careful to note the nature of the influence of context on reference time (noted above): reference time may be given by the previous discourse. This seems to have been Reichenbach's conception as well. Notice that his remark about how the events recounted may determine reference time is likewise qualified: this is so "in a story." This suggests that reference time isn't always provided in previous discourse, and that we have, in effect, two sorts of discourse--that in which reference time is fixed by previous discourse, and that in which it isn't. Let us call the first sort (tetlporally) connected discourse and the second (tellporally) free discourse (or temporally non connected discourse), and let us contrast examples of these: (I) Temporally connected discourse Al went to N.Y. The others were there, too. Temporally free discourse Al went to N.Y. The others were there once, too. The temporally connected discourse continues talking about the "same" time, while the temporally free discourse does not. In connected discourse, times may not be out of order, while in free discourse, this is possible. (We will examine which times these are presently.)

10

There will obviously be different principles of temporal reference in force in these two different types of temporal discourse. Let us attempt a first formulation of these, however rough. Reichenbach claims that "the series of events recounted determines the point of reference" (in connected discourse), and the example above bears him out. The ti me spoken of in the second sentence in the connected example in (1) seems to be identical to the time at which Al arrived in N.Y., the event time. Some examples are different, however.

(2) Al went to N.Y.

Bo had found him a room.

He went directly to

it. Here it is clear that the time spoken of in the final sentence is not the event time of the previous sentence, i.e. the time at which Bo found the room. It is also clear that events have not been recounted in order, and therefore that event times are not ordered properly. Still, this has the feel of a temporally connected discourse--a story. The not overly elusive principle of organization is based on reference time. The second sentence in (2) has an event time prior to the first's, but its reference time is fixed and non-prior to the first's. And it is again the reference time of the second sentence which is used in the third. This suggests the following codification of Reichenbach's pragmatics. In moving toward a formalization, we note that we shall employ an "interval" semantics, following most notably Bennett and Partee (1972), Cresswe 11 (1977) and Dowty (1979). Van Benthem (1983) i nvestigates the model theory of tense logic based on both points and "periods" (objecting to the boundaries impl ied by "interval"). We retain the linguistically familiar term "interval." In this semantics propositions are evaluated as true or false not relative to points of time, but rather relative to intervals. We first need to define some subsidiary notions. Since the times we will be dealing with may be intervals, the notion of precedence is somewhat vague. Consider the time line below:

11

It is clear that i1 precedes both i2 and i 3 , since every point of time in i1 precedes every point of time in both i2 and i 3 ,

We shall

symbolize this relation as ' -TEMP

,xTEMp>.y>.zPRET(vT'(x)(y)(z»>

-TEMP -NPacc +pret

Crucial is the feature [-TEMP] here, which ensures that some temporal adverbial--perhaps null in realization--must be added to (2) and (3), and (given the semantics) that it must be quantified in within the scope of tense. The complement-adding MRs may then be applied to derive the rules in (8): (8) .yPRES(vT'(TEMP')(y»> +TEMP +pres' =

.yPRES(TEMP'(V'(y»), semantically

TEMP, V ], >.x>.yPRET(vT'(TEMP)(x)(y»>

+TEMP +pret -NPacc

=

>.x>.yPRET(TEMP'(V'(x)(y»), semantically

184

The required scope relations have been realized in (8). This has been achieved, however, at the cost of the introduction of new semantical apparatus, the predicates IVp I and new syntactic mise en scene, the null realization of indefinite temporal adverbials (to account for the opti onal i ty of these adverbi al s once they have been given the status of complements). In addition, since we analyze optional adverbials as complements, we can no longer regard optionality as proof of modifier status. Thus an interesting empirical claim has been relinquished as well. If this were not enough to bias one against the analysis of tense as a verb operator, rather 1ess tractable problems arise when one considers the iterability of these temporal adverbials. Duratives and frequentatives are not limited to a single occurrence per clause as BR's 2 and 3 above might suggest: (9) Schon zwei Jahre besucht er uns jede Woche eine Stunde already 2 years visit he us every week an hour 'He has visited us an hour every week for two years I

Er hat uns zweimal eine Woche lang jeden Tag vier mal angerufen he AUX us two-time a week long every day four time call 'He's called us four times a day for a week twice ' This is particularly inappropriate under the view that duratives and frequentatives function as members of the same category (cf. 2.4-2.5). In that case the sentences in (9) represent three and four iterations of the single category TEMP. But even a single iteration is an irreparable embarrassment for the view that these are complements. Iteration is not only regarded as uncharacteristic for complements, it presents formidable technical problems as well. Since complements do not in general iterate, the admission of one iterable category would require a split in syntactic

185

treatments--one for standard complements, one for iterables. We work ourselves into a semantic cul de sac as well because the addition of a temporal adverb to a verb should, under the complement view, yield a unique predicate--just as every function should yield a unique value when applied to a particular argument. But applying the function represented by the verb besuch- to its putative durative argument seems to yie14 two functions--one which takes two NP arguments to form a proposition, and one which takes the NP arguments and a frequentative argument (and possibly another durative and then possibly still another frequentative argument). There isn't a recognizable sense in which this could be regarded as functional application. There does not seem to be a plausible way to maintain an analysis of duratives (and frequentatives) as complements to the verb, and therefore to save the analysis of tense as a verb operator.

3.6.2 Tenses as Phrasal Operators The analysis of tenses as verb operators looks even worse when one considers that there is a straightforward syntactic treatment of the scope relations between tense and duratives. We write the BR's so that no tensed elements can be introduced, i.e the BR's admit only constituents whose verbs are not semantically tensed. Duratives, and other elements with scope narrower than tense are admitted via MR's which operate exclusively on untensed rules (in this case, simply those marked [-fin]). Tense-introducing MR's change this feature to [ +fi n], ensuri ng that duratives etc. cannot subsequently be i ntroduced. Finally, those elements whose temporal scope seems wider than tense, e.g. frame adverbials, are introduced via MR's which operate on tensed rules. In each case, the semantics associated with the rule simply attaches an appropri ate operator, so that the meani ng of the derived rule is obtained compositionally.

186 If we e.g. introduced tense as a VP operator, following Bach, the meaning would simply be AxOPERATOR'(VP'(x», where VP' is the meaning of the constituent which would have been introduced by the input rule. Bach (1980) suggested the use of the 1ambda operator in the above fashion as a means of introducing tense and aspect in order to reconcile the apparent sentence scope of tense (and aspect) operators with the fact that they regul arly appear on verb phrases. Hi s remarks apply equally well to Gennan with the minor adjustment (within the present framework) that we introduce tense not on VP's, but on CVP's, in order to be able to introduce tense even where we find no VP (=PVP-NPnom), i.e. on impersonal constructions such as the impersonal passive (discussed in 3.2 above). As semantically elegant as the VP (or CVP) analysis might be, it appears nonetheless to run into a syntactic problem. There are subconstituents within the VP which bear tense marking--which woul d seem to i ndi cate that tense ought to be introduced at the 1eve 1 of these constituents (or lower). In order to demonstrate the difficulty, we , shall state the tense-introducing MR on CVP's: (1)

Tense MR --> (TENSE-MR)

(2) -fin +agr

( TEMP-MR)

NPn, TEMP, PRED, V+ agr ], TEMP'(V'(PRED'}(NPn'}» - f'1n +agr --> (TENSE-MR)

(FRAME-MR)

--> r and A . s,l,r P).

F

(We ignore the imperfectivity issue here.) For the most part, this extra constraint res.Jlting from ...!!och is inconsequential, since the exact delineation of r is often left up to Pragmatics. But note that if r is specified elsewhere in the sentence, then there may be no s,e,r satisfying the above conditions. consider in this connection (1): (1) * Morgen ist er noch zwei Jahre da tomorrow is he yet 2 years there Other rules guarantee that this is assigned the meaning: (2) morgen'(PRES(zwei Jahre'(noch'(p))))

and (2)'s truth conditions are straightforwardly derived: (3)

As,e,r

r (2) iff.. As,e,r

f

[morgen']A and e=r-1 Surely we need not attribute this to an actual third person subject, since we can equally well regard the third person as the unmaru:n-V is determined by LP rule (2) in 3.3.1 above, repeated here for convenience:

X-verb < V- f'1 n The rul es responsible for the expansion of PPum do not concern oJ,) here. Several other aspects of the tree above wi 11 recei ve CO!llonell t after we have examined an appl ication of the passive rule to a verb which is not subcategorized [-NPacc], i.e. an impersonal passive. For the sake of variety, we '~xamine a separable prefix verb fro:n r;lass 15 in this application. BR 15 is first repeated.

275

~:::F + 'I, which, as 3.3.2 noted, may be requi red anyway (guaranteei ng the exi stence of the constituent is trivial, should it definitely be required--(5) would do this). CAC MR is repeated here for convenience: ~Q.n_!.Q.l!re~_~r!~i n~_Q.~_~~npl eollents

-->

(CAC)

and intentionality imp 1i ca tu re -fin -fin +pass +pass +prt +prt +PPauf -PPauf

276 This admits the following subtree:

(6)

CVP -fin +pass +prt +PPauf +15 PPauf

auf Einzelheiten on detail s

PVP -cl itic -fin +pass +prt -PPauf +15

I

PVP +pass +prt +15

J elngegangen go in (prt) 'gone into details' (Redundant features have been suppressed in the lower PVP node in (6). We shall turn directly to the introduction of the passive auxi1 i ary werd-, but these exampl es may have suffi ci ently cl arifi ed the workings of the rules, in particular the passive rule, so that a discussion of their details and motivation may be fruitful. Let us first note that in making the type of passive dependent on the need for an accusative complement, this proposal reflects the conditioning of the passive rule by this factor and thus satisfies the desideratum established in the introduction to this section. With reference to 4.4.1, we may note that impersonal passives have no subjects, and no provi s i on for the 1ater i ntroduct i on of subj ects accord i ng to til is passive rule. Cf. the tree immediately above.

277 The generation of passive sentences has been broken down into two stages, the introduction of the passive auxil iary, to be presented below, and the above passive rule, which creates participial phrases. This was done for two reasons. First, there are passive participial phrases which appear adnominally without the passive alJxiliary, werd-. For example:

das vor kurzem gebaute Haus the recently built house Although more must be said about tense in their generation, it SI~I~;n') most economi ca 1 to concei ve of these phrases as crea ted by tl1e sa·~·~ passive rule responsible for (1) and (2). But in this case the passive rule must be separated from the rule introducing the passive auxiliary werd-. Second, there are conjunction facts which indicate that the participial phrases created by this passive rule are constituents to the e~clusion of the passive auxiliary. Thus the (standard) VP without werd- is subject to conjunction (7a), as is the CVP without werd- (7b), and the PVP without werd- (7c): (7a) Die Kinder wurden ins Haus g~~~hj~!t und dem Ga~~~~~Q~~~e~~~ the chil dren AUX into house send( prt) a the guest introduce (prt) 'The children were sent into the house and introduced to the guest' (b)

Es wurde getanzt und gefeiert it AUX dance(prt) and celebrate(prt) 'People danced and celebrated'

(c)

Ihm wurde 9.~~~hmeichelt und ~~9.~laechelt him AUX flatter(prt) and at-smile(prt) 'He got flattered and smiled at'

Several of the points made in 4.4.2 about the lexical nature of

278

the German passive are reflected in the present rul!:. First, note that this metarule applies to rules to which no syntactic complements have been added. This is the significance of the rule's requiring that all such features be marked [-comPi]--essentially requiring that verbs be marked [+pass] before syntacti c rul es d;Jplf to t:le'TI. The rule thus applies only to (rules for) individual lexical items, and not to (rules for) phrases which the syntax has constructed. Second, the system allows for lexical exceptions. We noted earl i er that the verbs introduced by BR 9, repeated for conveni ence below, are apparently split vis-a-vis passivizability.

schlafen, lachen, existieren, ... ,sich ~~h)_~~~~

279

The derivation of impersonal passive sentences using these verbs is quite straightforward. Syntactic reflexives could not have been specifically provided for before the passive rule applies, since the passive rule req'Jires that all syntactic complements be yet missing. We may pl ausibly assume that the attempt to add refl exives after the passive ::1etar:Jlc has deformed the original will be successful just in case a suitable nominative antecedent is available. Since nominative anteced~nts are never available in impersonal passives, riO syntactic reflexives may be found there. This explains the ungraillnatical ity of t!l(~ :xanple Ijsed if) 4.4.2 above, repeated here: Es wurde von einer Geschichte it AUX of a story

(* ueber sich) geredet about sel f

Spl~dK:(;l"t)

A final point regarding the lexical status of rule may be nad:~ before we turn to the introduction of the passive auxiliary. We noted in 4.4.2 that one normally frontable item, the emphatic reflexive sich selbst, is not frontable in impersonal passive sentences, even though it may appear there. Again, given the assumption that sich sel~s! may appear in impersonal passives by virtue of its abil ity to function within the verb as part of a lexical unit, the fronting behavior is predicted. To see this, suppose that ~~~l__s_e_l~)$t he1 Fen, like sich schlage_n_, is part of the class of verbs introduced by BR 2 (above). Then the passive MR applies to it to derive:

-->

+pass -prt +PPum +NPnom This might be used, as is, in conjunction with BR 301, to derive such sentences as es wurde Herr Schmidt urn einen Gefallen gebeten. Or the fronting metarule may apply, yielding:

+fin -fin + intentionality implicature +pass +pass +prt -prt +PPauf -PPauf with which the es-introduction rule combines nicely:

287

S

+me

es it

wurde

1

cVP +me +fin +pass -prt +15

PPauf

AUX

auf Einzelheiten on detail s

PVP -el itie -fin +pass +prt -PPauf +15

I

PVP +pass +prt +15

.

\

eingegangen go in (prt) '[They] went into details'

Notes--Chapter Four 1. To a certain extent the choice between haben and sein is temporally determined, of course. All unprefixed intransitives which denote telic Aktionsarten use the auxiliary sein, and, with the exception of bleiben and sein, all other intransitives and all transitives use haben. Thus telic intransitive sterben, gestorben sein; atelic intransitive schlafen, geschlafen haben; and transitive essen, with gegessen haben. Verbs which are ambiguously telic or atelic may have Perfects with both auxiliaries: (in den Fluss) schwimmen, geschwommen sein, but (im Fluss) schwimmen, geschwommen haben. Prefi xed verbs use the auxil i ary of thei r unprefi xed stem, even if this contradicts the semantic indicatio; thus herumgehen 'to walk around, walk about' clearly has an atelic sense (and consequently may be used with duratives). But the expected * herumgegangen haben is wrong due to the telic stem gehen 'to go,' which has the expected gegangen sein. Thus: herumgegangen sein. The determination of transitivity (for this purpose) is also complicated. Cf. er hat mir geholfen 'he helped me' vs. er ist mir entgegengekommen 'he accommodated me.' Again, the complicating factor may be the unprefixed stem (but it may al so be whether the transitive form is basic or derived from a basic intransitive). The entire picture is slightly more complicated in the South, where the atelic intransitives liegen, stehen, and sitzen unexpectedly form Perfects using sein. 2. It is worth noting that the Perfect Infinitive (as it is used above), like tenses in subordinate clauses (cf. 1.7.2), may indexically refer to an event time not among the speech, event and reference times of its matrix clause. 3. What follows in the text is not simply a presentation of Baeuerle's semantics for the Perfect, but rather my sketch of what any semantics with similar ambitions must be 1ike. Baeuerle's rules are - 288 -

289

flawed in not showing how the Perfect index is affected by the tenses. The rules amount to requiring that e and therefore nothing blocks coreference in (45). (The same is trivially true for sentences like (44a) with 'forward' rather than 'backward' pronominalisation.) Next, let us consider the relations of NP 3 and the object NP2, in trees

Coreference of Definite NPs

45

(41) and (42). This relation is, again, identical in both trees- NP2 does not c-command NP 3, since it is immediately domina ted by the VP, which does not dominate NP 3• Hence NP 3 is not in the domain of NP 2 , the coreference restriction does not apply, and coreference is possible, even if NP 3 is a full NP (and NP 2 is a pronoun). This means that 'backward pronominalisation' is permitted in the sentences of (I 9), which have the structure (42), e.g.: (46) We had to fire him since Mcintosh's weird habits had reached an intolerable stage.

(Mcintosh (NP 3) is not in the domain of the pronoun {NP2);hence coreference is not blocked.) In structures like (41 ), this means that 'forward pronominalisation', which is not permitted with subjects, is permitted with non-subjects, as in the cases of (23)-(26), e.g.: {47) For Ben's car I'm willing to give him 2 grand. We saw that the asymmetry between subjects and objects with respect to coreference - the difference between {44a) and {46) and between {43a) and {47) - was a major problem for the precede-and-command domain. For the- c-command domain, this is just the predicted result of the fact that subjects have the whole sentence in their domain while objects have only the VP in their domain. The c-command domain, thus, naturally distinguishes between subjects and objects. We saw, further, that there·is no way simply to mention the grammatical relations of the NP in the coreference restriction, so that it will apply differently to subjects and objects, since the mere fact that a given NP is an object does not permit free coreference between that NP and any NP to its right, as can be seen in the comparison between {19), or (46), above, and {48). {And compare also (48) with {47).) {48) *I'm willing to give him 2 grand for Ben's car. {49) See diagram on next page. This again indicates that what really determines coreference is the ccommand domains of NPs. In {48), the noun Ben is NP 3 of tree {49), which is in a verb-phrasal PP (unlike NP 3 in structures like (42) which is in a sentential PP). Hence, Ben, in (48), is in the domain of him and the coreference restriction blocks the sentence. 6

Coreference of Definite NPs

46

s

~ NP

1

V~ VP

V~P

~

(NP)

N~

2.4.2. A systematic empirical evaluation of the two alternative definitions of domain depends on noting that c-command entails command, i.e. if node A c-commands node B, then A also commands B and if A does not command B, then A also does not c-command B. When the linear order is also taken into consideration, the relations between the two definitions are as illustrated in (SO): (50) B is in the domain of A by the c-command definition

B is in the domain of A by the precede-and-command definition:

Ill

A precedes and c-commands B (-. A precedes

A precedes, commands,

In a right-branching language it is often the case that the c-commanding node precedes the c-commanded node. The intersection, in which both definitions include B in the domain of A, is, therefore, quite large. This may account for how it has been possible for the precede-and-command rule to yield the right prediction in such an amazing number of cases

Coreference of Definite NPs

47

and nevertheless be the wrong rule. It may also help explain why the relation of precede is believed to play such a crucial role in the grammatical restrictions on coreference. Obviously, in an overwhelmingly large body of the language, forward pronominalisation is the only grammatical option. Given the c-command relation, this fact is just an obvious result of the application of the co reference rule to right-branching trees. This is not true for the relation command, where all the nodes dominated by the same S equally command each other. A rule stated in terms of command must therefore introduce the relation precede into the rule of coreference. The failure to distinguish between grammatical and pragmatic constraints has also contributed to the belief that precede is a major factor in the grammar of coreference. It is clear that the large correlation between the domains defined by precede-and-command and by c-command holds only for right-branching languages. The sharpest discrepancy between the domains picked up by the two definitions will show up in VOS languages (assuming that these languages have a VP). In such languages a preceding node would often be in the domain of a following node. I have not studied cross-language restrictions on anaphora, and, consequently, I cannot argue for the hypothesis that the c-command restriction on anaphora is universal. However, the following examples from Malagasy (aVOS language with some evidence for a VP) suggest that this hypothesis should be considered. (The examples are from Ed Keenan, personal communication.) (51 a) namono azy ny anadahin-dRakoto hit/killed him the sister-of-Rakoto Rakoto's sister killed him. (b) *namono ny anadahin-dRakoto izy hit/killed the sister-of-Rakoto he he killed Rakoto's sister. In (Sla) the pronoun precedes and commands the antecedent, hence, by the precede-and-command restriction, the sentence should have been blocked. However, since the pronoun is in the VP, and, thus, does not c-command the antecedent (i.e. it is not in the c-command domain of the pronoun), the c-command restriction correctly permits coreference. The sentence in (51 b), on the other hand, does not violate the requirement of precede-and-command (since the antecedent precedes the pronoun), but coreference is, nevertheless, blocked. This is precisely the prediction of the c-command restriction, since the pronoun c-commands the antecedent, although the antecedent precedes.

48

Coreference of Definite NPs

Since the two definitions of the domain differ empirically only in the relatively small number of structures of the types I and III mentioned in (50), all that is left for the evaluation of the predictions made by the two definitions is to check structures of these types. It is not an accident that problems for the precede-and-command rule arose in trees like (41) and (42). The structure in (41) is an example of type I, since the subject c-commands but follows the NP in the PP; the structure in (42) is an example of type III - the object precedes and commands but does not c-command the NP in the PP. In fact, structures with PPs attached to S like (41) and (42) provide a major source for types I and III, and I will, therefore, devote the next chapter to a more detailed study of such cases. In a right-branching language there are almost no further examples (apart from the cases with preposed PPs) of the type I situation in (50), namely cases where the antecedent precedes the pronoun, but the pronoun c-commands the antecedent. 7 The situation of type III is more common, and I will, therefore, continue to exemplify further this type of case.

2.5 Coreference in Sentences with Extraposed Clauses Sentences with extra posed clauses provide another test for the alternative definitions of syntactic domain discussed above. Suppose we have a pronoun in object position and an antecedent in the extraposed clause to its right. If the extraposed clause is attached to the VP, it is in the domain of the pronoun by both definitions of domain and coreference should be blocked. If, on the other hand, the extraposed clause is attached to S, we have an example of type III in (50) where the pronoun precedes and commands but does not c-command the antecedent. Hence, by the precede-and-command definition the antecedent is in the domain of the pronoun and coreference should be blocked, while by the c-command definition the antecedent is not in the domain of the pronoun and coreference is permitted. Starting with Rosenbaum (1967) it has often been argued that the transformation of extra position which derives sentences like (52b) from the structure underlying (52a) moves the that-clause in (52a) to the VPfinal position, yeilding the (simplified) structure (52c) 8 (which can be, alternatively, base-generated). (32a) That Rosa has failed (should have) bothered her. 9 (b) *It (should have) bothered her that Rosa has failed.

Coreference of Definite NPs

49

(c)

sl NP 52

it

bothered

her

~ that Rosa has failed

Given this analysis of the position of extraposed clauses, the c-command definition of domain yields the right coreference result: in sentence (52b) coreference between her and Rosa is correctly blocked since Rosa, in S2, is in the domain of the pronoun her (the pronoun is immediately dominated by the VP, which dominates S2). However, if we look now at other types of extraposition we see that the situation is different. Thus (53b) is derived from (53a) by extra position from NP (namely, extraposition of S2) but, unlike (52b ), the sentence in (53b) permits 'backward pronominalisation'. (53a) [s Nobody [s who knows anything about Rosa's weird sleeping habits] woJld ever call her before noon] (b) [s Nobody would ever call her before noon [s who knows 2 anYthing about Rosa's weird sleeping habits]] (c) [s 1 So many people wrote to him [s 2 that Branda couldn't answer them all] ] Similarly 'backward pronominalisation' is permitted in (53c) which, as argued in Williams (1974),is derived by extra position of the result clause {S2) from its initial position in the Determiner of the subject NP. 10 If all extraposed clauses are attached to the VP-final position, then the objects in {52b), (53b) and (53c) have identical domains, and there is no way to account for the difference in their coreference options. {Coreference in {53b) and in {53c) would be equally blocked in this case by the precede-and-command rule and by the c-command rule.) An alternative analysis for extraposition was suggested in Williams (1974, 197 5). He argued that all extra posed clauses are attached to the matrix S node, rather than to the VP. Under this analysis the acceptability of co reference in ( 53b) and ( 53c) would be accounted for by the c-command rule, since the object in this case does not c-command S2 • However, for the same reason coreferen.ce should be permitted in (52b) as well. The precede-and-command rule will still successfully block coreference in

SO

Coreference of Definite NPs

(52b ), but it would also incorrectly block coreference in (53b) and (53c). In view of this difficulty, it is appropriate to check the common assumption that extraposition is a unified phenomenon and all extraposed clauses are attached to the same position. We saw that there are two views: one, that all extraposed clauses are attached to the VP, and the other, that all such clauses are attached to S. Emonds (1976), who elaborates the first view (arguing, further, that extra position is structurepreserving, namely that the extra posed clause is moved into the S position that is independently present in the VP) supports his. proposal with the fact that extraposition is impossible when the VP contains a filled S position, as in (54). (54a) That she smokes proves (that) she is nervous (b) *It proves (that) she is nervous that she smokes However, the same test does not hold for extraposition from NP, as indicated in the following sentences from Williams (1975). 11 (55) Many people said they were sick who weren't sick. Extraposition of result clauses is also possible in such contexts, as seen in (56). (56) So many people told hiin he is a genius that he started believing it. So, in fact, Emonds's strongest argument for his analysis of extraposition holds only for the extraposition of 'sentential subject' of the type illustrated in (52), but not for all types of extraposition. A similar difficulty shows up in Williams's (197 5) argument for the alternative view that all extra posed clauses are attached to the S position. In fact, his arguments hold only for extraposition from NP and Resultclause extraposition. His main argument is that extraposed clauses cannot show up before sentential prepositional clauses (if they were to be attached to the VP, they should have been permitted in this position). As we see in (57), this is indeed true for the two types of extraposition in ( 57b) and (57 c), where the although clause cannot be construed as modifying the matrix S, but the sentence in (57a) is perfect, in violcdon of Williams's prediction. 12

Coreference of Definite NPs

51

(57a) It shocked Rosa that she lost the case, although she had no reason to believe she would win. (b)* A man came in who looked very threatening, although the office was officially closed. (c) *So many people wrote to Bran do that he couldn't answer them all, although they did not know him. These facts suggest that regardless of whether they are derived transformationally or not, extraposed clauses are not always attached to the same position. Further, there is also a more decisive argument which shows that this is indeed t~ue: if the extra posed clause is attached to the VP it should be possible to prepose it along with the VP, when the VP is preposed. This is indeed true for extraposition of sentential subjects, as illustrated in (58a). However, the same movement in the case of the other two types results in the nonsensical (58b) and (58c). (The sentence (58a) is perhaps not a most natural one, but it is obviously grammatical. ) 13 (58a) I warned you that it would upset Rosa that you smoke, and upset her that you smoked it certainly did. (b) It was predicted that many people would resign who disagreed with the management's policy, *and resign who disagreed with the management's policy many people did. (c) I was afraid that so many people would show up that we wouldn't have enough room *and show up that we didn't have enough room so many people did. The same point can be illustrated by 'though movement' (which derives the (b) sentences below from the structures underlying the (a) sentences): the extraposed S2 in (59) can be dragged along with the VP, as in (59b), but the extraposed s2 in (60) cannot. (59a) Though it was unlikely [s 2 that she would pass], she still decided to take the exam. (b) Unlikely that she would pass though it was, she still decided to take the exam. (60a) Though many people are unhappy (S 2 who live in New York], nobody thinks of moving. (b) *Unhappy who live in New York though many people are, nobody thinks of moving.

52

Coreference of Definite NPs

We can conclude, therefore, that while extraposed 'sentential subjects' are attached to the VP, the other types of extra posed clauses are attached to S (hence they cannot be preposed with the VP). Consequently the coreference facts in (52)-(53), repeated in (61)-(63), are no longer a mystery. (61) * [s It should have [yp bothered her [s that Rosa has failed] ] ] (62) [s Nobody would ever [yp call her\efore noon] [s who 2 kn~ws anything about Rosa's weird sleeping habits]] (63) [So many people [yp wrote to him] [s 2 that Branda couldn't answer them all]] Given the c-command definition of domain, Rosa is in the domain of the pronoun in (61), hence coreference is blocked. In (62) and (63), on the other hand, the antecedents (Rosa and Branda) are not in the domain of the pronoun since their clauses are outside the VP. Hence the restriction on coreference does not apply to block coreference in these cases. By the precede-and-command definition of domain the object pronouns have identical domains in (61 )-(63) since in all three cases the pronoun precedes and commands the antecedent. Hence the precede-and-command restriction blocks coreference in all three sentences. It could perhaps have been argued that coreference in (62)-(63) would be permitted by the precede-and-command rule if the extraposed clauses are attached to a higher S (S), and, thus, are not commanded by the objects. Note, however, that the same subject-object asymmetry which posed a problem to the precede-and-command rule in the examples discussed in Section 1 .4 shows up in the case of extra position: coreference is impossible in (64) and (65). (64) *He met a woman in Chicago who went to school with Dan's mother. (65) *She was approached by so many people in Rome that Rosa couldn't do any work. If the object does not command the extraposed clause, the subject does not either. Hence if the suggested solution could be true, coreference should have been permitted in (64) and ( 65). The c-command restriction correctly blocks coreference in _these sentences, since, given the full definition of c-command, in (36), the subject, unlike the object, ccommands nodes which are attached either to S or S. Extraposition from NP and extraposition of result clauses thus provide one more argument for the preference of the c-command definition of domain.

Coreference of Definite NPs

53

2.6 PPs and Indirect Objects Another instance of type III in (50) above arises whenever an NP dominated by a PP (or a branching NP) precedes another NP. The examples in (66) illustrate such cases, where the c-command condition yields the right results: co reference is possible although the antecedent precedes and commands the pronoun. While ( 66c-e) can be handled by a modified definition of command that considers NPs as the relevant cyclic nodes, (66a-b) cannot. (66a) You won't believe who I saw near him in Ben's car. (b) Rosa never fails to think about them on her lover's birthday. (c) Rumours about her please Rosa. (d) Details of her death fill the day for family of the latest victim (Headline in NYT, quoted in Carden, 1978). (e) ... and father Wolf taught him his business ... till every rattle in the grass ... meant just as much to him as the work of his office means to a businessman (Kipling, 'Mowgli brothers', quoted in Carden, 1978). However, problems for the c-command restriction on anaphora arise when we consider anaphora options of indirect objects. If indirect objects are dominated by a PP, the c-command definition of domain assigns different domains to direct and indirect objects, since, in this case, the domain of indirect objects consists of the PP alone. However, the following examples show that there is no difference in anaphora options of direct and indirect objects: (67a) *It didn't surprise her that Rosa has failed the exam. (b) *It didn't occur to her that Rosa has failed the exam. (68a) *I met him in Ben's office. (b) *I spoke to him in Ben's office. (69a) *Someone should tell her that Rosa's driving is dangerous. (b) *Someone should point out to her that Rosa's driving is dangerous. Coreference in the (a) sentences above is blocked since the pronoun ccommands the antecedent. But coreference in the (b) sentences is just as impossible although, by the formal defintion, the indirect-object pronoun does not c-command the antecedent. Note, further, that the similarity between anaphora options of direct and indirect objects is preserved

54

Coreference of Definite NPs

when they are preposed (topicalised). Thus, coreference is equally blocked in the (a) and (b) sentences of (70) and (71 ). (70a) *Him, I met in Ben's office. (b) *To him, I spoke in Ben's office. (71a) *Him, Max's mother gave a book. (b) *To him, Max's mother gave a book. (72a) *Him, Don's mother found a gun near. (b) Near him, Don 's mother found a gun. The c-command restriction has no problems in blocking the (a) sentences, where the pronoun c-commands the antecedent. It also, correctly, permits coreference in (72b ), since the pronoun is dominated by a (locative) PP, and, therefore, it does not c-command the antecedent. The problem is, however, the (b) sentences in (70) and (71). If the indirect object in these sentences is dominated by a PP, the c-command restriction predicts that they should behave like the PP in (72b ), and coreference should be permitted. To handle these cases we may assume that indirect objects are distinguished syntactically from such PPs as locative and instrumentals. Rather than being dominated by a PP they are dominated by an NP with a case marker which is lexically realised in English with a preposition, but it can be realised by other means in case-marked languages. If this is assumed, indirect objects c-command everything in the VP, and when they are preposed they c-command everything in S, which will capture the facts in (70) and (71 ). 14 However, the problems for c-command arise in a few other PP types as well, which suggests that this solution might not be sufficient and I shall return to these problems in Chapter 8.

2.7 Coordinate Structures I have argued that the linear order of a pronoun and a full NP plays no role in the sentence-level restrictions on coreference, which are sensitive only to properties of constituent structure. One case where the linear order may seem to play a role is coordinate structures, such as (73), where anaphora seems impossible. (73) She entered the room and Rosa collapsed. It is not clear that the precede-and-command restriction on anaphora can

Coreference of Definite NPs

55

handle such cases, since, given the standard definition, the preceding pronoun does not command the antecedent. In any case, the c-command restriction clearly does not apply in such cases. In fact, however, there is no reason to expect that the unavailability of anaphora in such cases should be attributed to sentence-level consideration. The same problem arises 'across sentences' as in {74) where, excluding literary narrative contexts, anaphora is unlikely to be obtained and where, obviously, no sentence-level account is possible. {74) She entered the room proudly with her new hat on. A few minutes later Rosa collapsed. The linear order requirement on anaphora in such cases is attributed to discourse considerations. In normal, rational discourse, new referents are first introduced by their proper name or by a description which enables the hearer to identify them, and once a reference is unmistakably established, it may be referred back to with a pronoun. It is the same standard communication norm that explains the difficulties in obtaining coreference in {73). As we noted already in Section 2.2, the use of backward anaphora, when the grammar permits it, is in any case restricted by various discourse and stylistic considerations. Although the precise conditions permitting such use are not fully understood, it is sufficient to note here that (as observed, e.g. in Bolinger, 1979; McCray, 1980; and Mittwoch, 1979) when they are met, backward anaphora in conjoined sentences is possible as in (75). Usually it is easy to get coreference if the first conjunct is pragmatically subordinated (Mittwoch, 1979) to the second. {7 Sa) She has the whole city at her disposal and Rosa just sits at home. 15 {b )He has sent Melinda dozens of roses as testaments of his love, and yet never has Walter been so absolutely sure of failure. {McCray, 1980) (c)He hasn't contacted me, but I'm sure John is back. (Mittwoch, 1979)

Obviously, the linear order of the pronoun and the full NP may affect the actual referential choice for the pronoun in a given context, just as many other discourse considerations may. Our task here, however, is not to account for this actual choice of reference, but rather to determine under what conditions the choice is restricted by purely sentence-level

56

Coreference of Definite NPs

considerations. If the grammar allows a pronoun to corefer with a given NP, this of course, does not mean that they would actually corefer in a given discourse.

Notes

t

Section 2.5 of this chapter was published in part in Reinhart (1980) and is reprinted with permission of MIT press. A summary version of the material of Chapters 2 and 3 was published in Reinhart (1981a). 1. Following the conventional notation in studies of anaphora, marking two NPs with italics means that the sentence is considered under its coreference interpretation only. A star before such sentences means that the sentence is considered ungrammatical in this interpretation. 2. One consequence of this statement of the coreference restriction is that the terms antecedent and anaphor are superfluous. As we have seen, the problem is not defining the required structural relations between antecedents and anaphors, but rather defining the structural conditions which affect the coreference options of any two NPs. 3. Given Jackendoff and Lasnik's modification of the definition of command which we observed in Chapter I - namely, that it mentions cyclic nodes rather than just S-nodes - the sentences in (18b-e) are not counterexamples to the precede-and-command relation, since the pronoun does not command the antecedent by this defmition of command. However, this modification does not help to account for (18a), nor does it have anything to say about the sentences in (19) where the pronoun is not in a possessive NP, and, thus, commands the antecedent under this definition as well. 4. It could, perhaps, be argued, as proposed in Kuno (1971, 1975b), that in (21b), unlike (21a), there is no preposing and that the PP originates in initial position. Thus, since the sentence is not derived from (22b), 'forward pronominalisation' is not blocked. But the question will then be, what permits backward pronominalisation in a case such as (i). (i) In John's picture of her, Mary looks sick.

since the acceptability of (i) cannot, in this case, be explained by the ordering of coreference prior to the pre posing of the PP. This solution will also fail to hold in the next example, (23), where there is no reason to argue that the PP originates in initial position. 5. We also have to add to (40) the requirement that the non-pronoun be distinct, since by the definition of c-command, nodes c-command themselves, so without this addition (40) would have the contradictory effect of requiring all full NPs to be non-coreferential with themselves. 6. Of the examples mentioned in Section 2.2, I have not accounted yet for the pairs in (21) and (22) and for the difference between the (a) and (b) sentences in (32). These cases will be discussed in Chapter 3. 7. A possible example of a situation of type I is provided by coordinate NPs. For many speakers, coreference is impossible in sentences like (ia) and (ib). (ia) ?Cavallo's wife and he are getting on my nerves. (b) ?I met Cavallo's wife and him in the office.

Coreference of Definite NPs

57

NP

(c)

NP

~

NP

I

Cavallo's

and

NP

NP

\

wife

he

If the coordinate NPs in (a) and (b) have the structure in (c), the impossibility of coreference is due to the fact that the pronoun c-commands the antecedent, although the pronoun follows it. However, it is not clear that the NPs in (i) have the structure in (i). Alternatively and might be attached to a higher NP dominating he, in which case he does not c-command Cavallo. Furthermore, the awkwardness of these sentences seems to have an independent semantic (or pragmatic) account in terms of 'empathy' (a notion developed in Kuno, 197Sb, and Kuno and Kaburaki, 1975): Cavallo is the center of empathy in these sentences (which is indicated by the fact that his wife is not introduced independently, but rather identified as a function of Cavallo). Their rule stating that the center of empathy should not be pronominalised intra-sententially accounts, therefore, for the inappropriateness of (ia) and (ib). Another potential example of a situation of type I is a double object construction as in (ii).

(ii) *I sent the book's owner it. In (ii), the pronoun it c-commands the antecedent the book and coreference is indeed blocked. However, this type of example carries only little force, since for most dialects of English a pronoun cannot occur in this position anyway, as illustrated in (iii). (iii) ??I sent Bill it. After the revision of the co reference restriction I propose in Chapter 7, it would turn out that in neither of the last two cases should non-coreference be expected. 8. In fact, as argued, e.g. in Chomsky and Lasnik (1977), it is more likely that the extraposed clause is attached to a VP higher than the one dominating the object. Such structural differences do not affect my subsequent argument, since given the full definition of c-command, as e.g. in (36), the object nevertheless c-commands nodes dominated by the higher VP. Note, incidentally, that such cases serve as an argument for preferring this definition of c-command. Under the simplified definition adopted in Chomsky and Lasnik (1977), or Chomsky (1980) (which we surveyed in Section 1.2.2 of Chapter 1), the coreference facts in this section cannot be handled, assuming this position of the extraposed clauses. 9. I included the modal should have in (52) to assure that the inappropriateness of coreference is not merely due to a conflict in point of view. It was argued in Kuno (1972b) and in Reinhart (197Sa) that when an embedded clause can be interpreted as representing the point of view of a person designated by a noun in the matrix sentence, a noun in the embedded clause can refer to this person only if this noun is a pronoun. However, this restriction does not apply in cases involving a modal like should have, as indicated by the fact that (5 2a) is acceptable. (The point-of-view convention applies equally forwards and backwards, and would have otherwise blocked coreference in (52a) as well). 10. It is likely that sentences like (53c), just as (52b), are not, in fact, derived transformationally. This, however, is irrelevant for the present discussion. My

58

Coreference of Definite NPs

question is where the that clause attached, regardless of whether it originates in this position or moved into it. 11. Emonds (1976, n. 19, p.146) is aware of the fact that judgments concerning extraposition from NP in the case where the S position in the VP is filled do not support his analysis of extraposition, and he suggests some performance account for this fact. However, this account is based on the assumption that a judgment of such cases as acceptable is only possible if the extraposed clause is exceptionally long, which is certainly not the case in (55). 12. In his dissertation, Williams (1974) is aware of such difficulties and he concludes that 'sentential subject' extraposed clauses are attached to S while the other types are attached to S. The arguments below will show that this solution is not sufficient and extraposed sentential subjects are attached to the VP. 13. In Emonds's (1976) framework, VP preposing is a root transformation, hence it should be possible to apply it to the output of the structure-preserving transformation of extraposition. Extraposition should in any case be ordered before VP-preposing to allow for sentences like (58a). 14. A possible argument for this proposal is the fact that the indirect object marker to cannot have an independent stress while locative prepositions can. Thus, in (i) the normal intonation would put the stress on near, while this is impossible in (ii). (i) Near him, Rosa found a book.

(ii) *To him, Rosa gave a book.

On the other hand, if this solution is adopted, a substantial problem to be solved is the fact that, in English, unlike in the languages with clearer case marking, the to marker can be left behind while the indirect object is preposed, as in (ilib). (iiia) To Bill, Rosa gave a book.

(b) Bill, Rosa gave a book to.

15. This example is from Mark Liberman, personal communication.

3

PREPOSITIONAL PHRASES AND PREPOSED CONSTITUENTS

Sentences with prepositional phrases (hereafter PPs) provide several tests for the coreference restriction I proposed in Chapter 2 and for the relevance of c-command syntactic domains in general. We saw that in these cases, the precede-and-command definition of domain has failed most impressively, precisely because in these structures the domains defined by the alternative definitions (precede-and-command vs. c-command) are clearly distinct. In the previous chapter, I mentioned some clear examples of sentential and verb-phrasal PPs. However, sentences with PPs are particularly interesting, since many of them provide what looks like coreference mysteries: sentences which look identical syntactically differ in their coreference options, and, furthermore, the same PP in the same sentence may behave differently with respect to coreference if it is lengthened. Such mysteries have led several scholars to believe that there can be no grammatical account of coreference in sentences with PPs (or in general), and that they should be faced with performance constraints (Wasow, 1972) or purely semantic restrictions in terms of presuppositions (Bickerton, 1975). We will see that, in fact, all these seemingly mysterious coreference differences correlate with syntactic differences among sentences with PPs which are captured naturally by the c-command definition of domain. Thus, these cases provide further support for this notion of domain. A partial functional (theme-rheme) account for these cases may seem possible because, as we will see, there are certain correlations between syntactic domains and functional relations in these cases. In fact, in many cases the syntactic position of a PP depends crucially upon the interpretation given to the sentence, and it may vary from speaker to speaker. This provides a systematic way of accounting for disagreement on judgments among speakers. The point is that there should be a correlation between the interpretation given to the sentence by a given speaker and his judgments concerning both coreference and the syntactic tests which distinguish the PPs structurally. The main purpose of this chapter is to make clearer the notion of domain and to establish the correlation between the co reference options of NPs and their syntactic domains. However, the further striking correlations between syntactic domain and semantic or functional relations

60

Prepositional Phrases and Preposed Constituents

in sentences with PPs also provide an example of the potential power of the notion of syntactic domain, and thus anticipate the discussion of Chapter 9, which deals with other interpretative rules restricted by c-command domains. I will therefore include some detailed examples of such correlations.

3.1 Sentential and Verb-phrasal Prepositional Phrases

like adverbs, certain PPs are inherently sentential or verb-phrasal, i.e. their position in the tree is fixed. Thus, as Williams (1974) points out, in order to-phrases can only be attached to S, while infinitive to-phrases, as in We sent Rosa home to please her father, can be attached either to S or to VP. 1 Other sentential PPs discussed by Williams (1974, 1975) are although ... , (causal) since ... , whether or not ... , and withphrases of the type illustrated in (leV (I am following Emonds, 1976, in his analysis of although clauses etc. as PPs.) PPs which are always verbphrasal (according to Williams and many others) are instrumental (with ...) and manner (by ... ) PPs. In certain other cases, the PP itself is not inherently marked as to its position, yet its occurrence in a sentence with a verb which is strictly subcategorised to require a PP determines its obligatory position in the VP. Thus, locatives (e.g. in NP) are obligatory verb-phrasal following verbs like dwell, reside, or put. Similarly the verb flirt requires a with-phrase, which will then be verb-phrasal. In cases where the position of the PP is clear-cut, we get clear coreference judgments. While the sentences of (1 a)-(le ), some of which were mentioned in section 2.2, are possible in a proper discourse, no discourse will permit coreference in sentences like (2a)-(2e). (la) We sent him to West Point in order to please Ben's mother. (b) We 11 just have to fire him whether Mcintosh likes it or not. (c) Rosa won't like him anymore, withBen'smother hanging around all the time. (d) *Rosa tickled him with Ben's feather. (e) *It's time to put him in the baby's bed.

As we saw, given the c-command definition of domain, sentential PPs are not in the domain of the object. Hence the coreference restriction does not apply to block coreference in (1). Verb-phrasal PPs, on the other hand, are in the domain of the object (c-commanded by the object). In the sentences in (2), then, the antecedent is in the domain of the pronoun, which violates the coreference requirement.

Prepositional Phrases and Preposed Constituents

61

However, in many cases, the PP itself is not inherently marked (nor do selectional restrictions force its occurrence in only one position) and, as was first observed in Kuno (1975a), its position in the sentence seems to depend upon subtle semantic and discourse considerations. 3 I will illustrate this situation with the following pairs of sentences with locatives in (2)-( 4 ). 4 (2a)Ben is an absolute dictator in his office. (b) Ben placed his new brass bed in his office. )·m Be n ,s picture. . (3 a) Rosa ( .looks .d. sick h IS n mg a orse (b) Rosa found a scratch in Ben's picture. (4a)People worship Kissinger in Washington. (b) The gangsters killed Hoffa in Detroit. Several syntactic tests indicate that the same PP is sentential in the (a)sentences and verb-phrasal in the (b)-sentences. It should be kept in mind, however, that the position of the PP is dependent upon various aspects of the interpretation of the sentence, and thus it can be relative to speakers and to contexts. In particular, I have found several speakers who can interpret the PP in (4b) as sentential. What I want to illustrate is the correlation between the semantic interpretation of the PP, judgments of the tests indicating its syntactic position, and judgments of its coreference options. If the interpretation given to the PP is different from mine (whlch would be indicated by a disagreement in judgments concerning the syntactic tests), so should the coreference judgments be. 5 3.1.1 Syntactic Tests a. Two Tests with Pseudo-clefts (from Ross, 1973). The predicate part of pseudo-cleft sentences can contain only VP material. The what-clause, on the other hand, can contain only non-VP material. We thus get the following two tests, where the subscript indicates the node that immediately dominates the PP: (5) I (a) *[what ... did] is VP + PPs (b) [what ... did] is[ ... + PPvp·. -1yp II (a) [what ... did+ PPs] is .. . (b) *[what ... did+ PPvp1 is .. . Subjecting the sentences in (2)-(4) to these two tests shows that the PP in the (b)-sentences is part of the VP. Therefore, its occurrence is

62

Prepositional Phrases and Preposed Constituents

permitted in constructions like (5) I (b) and is blocked in (5) II (b). But the PP in the (a)-sentences is not in the VP. Hence its occurrence is permitted in the complementary environment. (In judging the sentences below, their intended reading should be kept in mind. For example (6) II (b) is acceptable if Rosa is depicted in Ben's picture as finding a scratch, which is not the intended reading.) (6) I (a) (b) II (a) (b) (7) I (a) {b) II (a) (b)

*What Rosa did was ride a horse in Ben's picture. What Rosa did was find a scratch in Ben's picture. What Rosa did in Ben's picture was ride a horse. *What Rosa did in Ben's picture was find a scratch. *What Ben is is an absolute dictator in his office. What Ben did is place his new brass bed in his office. What Ben is in his office is an absolute dictator. *What Ben did in his office is place his new brass bed.

b. VP Preposing (from Ross, 1973). The various types of VP fronting operations must apply to the whole VP. For example, such transformations can operate on the VP in (8), but not on parts of the VP, as in (9). As we see in (10)-(12) such operations can 'leave behind' the PP of the (a)-sentences in (2)-(4), which indicates that it is not part of the VP, but the same is impossible for the (b)-sentences. (8) ~NP]PP .. . (9)

... NP[ypVNPPP] .. .

~~

'Though' movement: (lOa) An absolute dictator though Ben is in his office, he is a sweetheart at home. (b) *Place his new brass bed though Ben did in his office, he is not always so whimsical. VP preposing: (lla) I wanted Rosa to ride a horse in Ben's picture, and ride a horse she did, in Ben's picture. (b) They wanted Rosa to fmd a scratch in Ben's picture, *and find a scratch she did in Ben's picture. (12a) It was predicted that people would worship Kissinger in Washington, and worship Kissinger they did in Washington. (b) It was predicted that the gangsters would attack Hoffa in Detroit, *and attack Hoffa they did in Detroit.

Prepositioned Phrases and Preposed Constituents

63

c. PP Preposing. Jackendoff (1972) noted that there is a difference in

the pre posing options of sentential and verb-phrasal adverbs. Verb-phrasal adverbs cannot be attached to S inside the sentence, e.g. between the subject and the VP, while sentential adverbs can, e.g. . probably) (13) John ( *slowly was eatmg a carrot. The same difference holds for sentential and verb-phrasal PPs, and it clearly distinguishes the (a)- and (b)-sentences of (2)-(4). (To avoid the reading on which people in Washington in (I Sa) is a constituent, I have added anyway to the PPs in these examples.) (14a) Rosa, in Ben's picture (anyway), looks sick. (b) *Rosa, in Ben's picture (anyway), found a scratch. (I Sa) People, in Washington anyway, worship Kissinger. (b) *The gangsters, in Detroit (anyway), killed Hoffa.

3.1.2. Some Semantic and Pragmatic Differences between Sentential and Verb-phrasal PPs An ordinary intuitive description of the semantic difference between sentential and verb-phrasal adverbs which can be extended to the two types of PPs, is that the first modifies the sentence, while the second modifies only the VP. However, what this means exactly, or what could count as semantic indications of this difference in the range of 'modification', needs clarification. It should be noted that the relevance of this question to the notion of syntactic domain goes beyond the establishment of the coreference options in the pairs under consideration. So far, we have concentrated on the domain relations of NPs (questions like is the NP in a PP in the domain of the object, etc.). However, the definition of the c-command domains (in (I 2) of Chapter I) does not mention NPs. It defines the domains of all the nodes in a sentence, stating that the domain of a given node is everything c-commanded by it. If we check now the domains of PPs (or adverbs), it is obvious that sentential PPs (as in (16a)) and verb-phrasal PPs (as in (' 6b )) have different domains. (Again, I use, for simplicity, the trees of (16), rather than the precise trees, containing two VPs.) The PP in (16a) c-commands all the nodes dominated by S (since it is immediately dominated by S). Hence the domain of a sentential PP (or adverb) consists of all the nodes in S or of the whole sentence. The PP in (I 6b) c-commands only (and all) the nodes in the VP; hence the domain of a verb-phrasal PP (or

64

Prepositioned Phrases and Preposed Constituents (16a)

A {::v} s

~ NP

V

NP

(l6b)

s

~VP ~ V NP {pp}

NP

Adv

adverb) consists only of the VP. Once the correspondence between domain relations and semantic relations is established, the difference in the semantics of the two types of PP will be merely a consequence of these general correspondence rules. I will return to these rules in Section 9.1, and here I will only list some of the semantic indications for a PPs being sentential or verb-phrasal, and show that they apply to the pairs in (2)-(4). a. Quantifier Scope. Ioup (1975) has observed that a quantifier in a PP cannot have a wider scope than a quantifier in subject position. Although she does not distinguish between sentential and verb-phrasal PPs, all her examples are from verb-phrasal PPs. In accordance with this observation, in the pairs under consideration here, the (b)-sentences (with verb-phrasal PPs) have only one possible scope reading, namely the one in which the quantifier in the PP is in the scope of the subject quantifier. Thus, the sentence in (l7b) has only the reading in which there is someone such that he found a scratch in all of Ben's pictures, and similarly, (18b) does not have the reading on which some is inside the scope of every. (17a) Someone is riding a horse in all of Ben's pictures. (b) Someone found a scratch in all of Ben's pictures. (18a) Some reporters worship Kissinger in every capital he visits. (b) Some gangsters ambushed Marcello in every town he visited. The (a )-sentences (with sentential PPs ), on the other hand, are ambiguous with respect to quantifier scope. (17a) can have, like (17b), the reading in which the PP quantifier is inside the scope of the subject quantifier, but also the reading in which the subject quantifier is inside the scope of the PP quantifier, namely that all the paintings of Ben are such that someone is riding a horse in them (not necessarily the same person). The same is true for (l8a). (The reading with wider scope for the subject quantifier requires imagining a situation in which a certain group of

Prepositioned Phrases and Preposed Constituents

65

reporters follow Kissinger from capital to capital and worship him wherever he is.) This situation is consistent with Ioup's framework, since she observed also that topics have scope options similar to subjects, and, as we will see shortly, sentential PPs, but not verb-phrasal PPs, can be considered to be topics.

b. Entailments. It has been observed in semantic studies of adverbs (e.g., Bartsch and Vennemann, 1972, and Cooper, 1974) that an affirmative sentence with a manner (VP) adverb always entails the same sentence without the adverb. But a negative sentence with a manner adverb does not have such an entailment. The same holds, in the pairs under consideration, for the sentences with verb-phrasal PPs: {19a) Rosa found a scratch in Ben's picture. ~ Rosa found a scratch. (b) Rosa didn't find a scratch in Ben's picture. "'~Rosa didn't find a scratch. (20a) The gangsters killed Hoffa in Detroit. ~ The gangsters killed Hoffa. (b) The gangsters didn't kill Hoffa in Detroit. "'~The gangsters didn't kill Hoffa. In sentences with sentential adverbs or PP, no predictable entailments hold (though entailment may be forced by particular adverbs). In our pairs, Rosa rides a horse in Ben's picture does not en tail Rosa rides a horse, nor does the negation of this sentence entail Rosa doesn't ride a horse. 6

c. Theme-rheme Relations. The two types of PPs differ also in their thematic (theme-rheme) functions in the sentence. Kuno (1975a) has argued that while sentential PPs are part of the old information of a sentence, verb-phrasal PPs usually provide new information. As suggested by Kuno and many others, this difference in function can be checked · with questions. A questions (with 'normal' question intonation) usually questions only the new information part of the sentence. (Below, the symbol '#'stands for pragmatic or semantic awkwardness.) (21a) A: Is Rosa riding a horse in Ben's picture? B: #No, in Max's picture. (b) A: Did Rosa find a scratch in Ben's picture? B: No, in Max's picture.

66

Prepositioned Phrases and Preposed Constituents

(22a) A: Is Ben an absolute dictator in his office? B: #No, in his kitchen. (b) A: Did Ben place his new brass bed in his office? B: No, in his kitchen. In the (b)-sentences, a response denying the PP is appropriate, which indicates that it is the PP which is questioned in this case (or that the PP is the new information); in the (a)-sentences, such a response is inappropriate, indicating that the PP is not new information. Another test that can be suggested for the theme-rheme distinction is the occurrence of expressions like anyway or at least. Such expressions indicate topichood, or afterthought, and they cannot modify new information. (Thus, they can never occur in clefted phrases, which always convey new information; compare Ben's father, anyway, will never allow it to #It's Ben 'sfather, anyway, who will never allow it.) Consequently, anyway can occur in our examples only with the sentential PPs as in (a) of (23) and (24 ). (23a) Rosa looks sick, in Ben's picture anyway. (b) #Rosa found a scratch, in Ben's picture anyway. (24a) People worship Kissinger, in Washington anyway. (b) #The gangsters killed Hoffa, in Detroit anyway. 3.1.3 Coreference facts

The discussion above has established the differences in the status of the PP in the pairs under consideration. The structure of the (a)-sentences in (2}-(4) is the one given in (16a), while the (b)-sentences have the structure (16b ). If we focus now on the domain relations of the NPs involved in these structures, we see first that the NP in the PP is in the domain of the subject in both cases. The coreference restriction ((40) of Chapter 2) requires therefore that the NP in the PP must be a pronoun in order for coreference to hold. It correctly predicts, then, that there will be no difference in the co reference options of the subjects and the NP in the PP in the two types of sentences, and indeed, the (a)- and (b)sentences in (25) and (26) are equally impossible. (25a) *He is an absolute dictator in Ben's office. (b) *He placed his new brass bed in Ben's office. (26a) *He is considered a genius in Kissinger's home town. (b) *He was killed in Hoffa's home town.

Prepositioned Phrases and Preposed Constituents

67

However, the two types of sentences differ in the domain relations of the objects. While the verb-phrasal PP is in the domain of the object, the sentential PP is not. The coreference restriction, therefore, puts restrictions only on the coreference options of the object and the NP in the verb-phrasal PP, while in the case of sentential PP, coreference is free (which means that coreference is possible even if the object is a pronoun and the NP in the PP is not). Thus we get the following differences in coreference options: (27a) Rosa is kissing him passionately in Ben's high school picture. (b) *I can't even find hif!l in your picture of Ben. (28a) People worship him in Kissinger's native country. (b) *The gangsters killed him in Hoffa's home town. The (a)-sentences have the same status as the sentences in (la-c) (and in {19) of Chapter 2): they may need some discourse justification for their usage, due to discourse constraints on backward anaphora. ((28a), for example, would seem more natural in a discourse like: Although people still worship him in Kissinger's native country, I can assure you that his glory won't last for long.) But (for speakers who agreed to the judgments in the tests mentioned above), the (b)-sentences are ungrammatical even in a proper context. ((28b) is still bad in a discourse like: *Although the gangsters killed him in Hoffa's hometown. I can assure you that he won't be forgotten for some time.) 3.2 Preposed PPs In Section 2.2 of Chapter 2 we saw that pairs like those in (29) show that the coreference restriction cannot apply to deep structure, before preposing the PP. (29a) (b) (c) (d)

*She is riding a horse in Ben's picture of Rosa. *She found a scratch in Ben's picture of Rosa. In Ben's picture of Rosa, she is riding a horse. *In Ben's picture of Rosa: she found a scratch.

Although the sources (29a and b) of sentences (29c and d) are equally bad, when the PP is preposed, (29c) is possible, but (29d) is still bad. At the same time, the examples in (29) show that marking the grammatical relations will not help, since in both cases the pronoun is the subject

68

Prepositioned Phrases and Preposed Constituents

and the antecedent is in a PP. Obviously, the linear order does not matter either, since the antecedent precedes the pronoun in both (29c) and (29d). Jackendoff ( 1975) attempted a semantic account for these pairs which has to do with the fact that the sentences involve pictures (orimages). However, as observed in Kuno (l975b ), this fact is accidental, since as we see in (30)-(31 ), the same distinction appears in the other pairs, which were discussed here (as well as in pairs like (32) and (33)) which do not involve images. Sentences like (33a) were noted in Koster (1979). (30a) In Kissinger's home town, he is considered a genius. (b) *In Hoffa's home town, he was killed by the gangsters. (31 a) In Ben's office, he is an absolute dictator. (b) *In Ben's office, he placed his new brass bed. (32a) With Rosa's new job, she'll end up in the hospital. (b) *With Rosa's new boss, she doesn't argue. (33a) According to Felix, he won the race. (b) *In front of Felix, he held a candle. In all these cases, the PP in the (a)-sentences originates as a sentential PP, while in the (b)-sentences, it originates as a verb-phrasal PP. (If the PP is not preposed, coreference is equally blocked in the (a)- and the (b)-sentences.) If sentential and verb-phrasal PPs are preposed to the same position, the (a)- and (b)-sentences in (30)-(33) have identical surface structures, and hence identical domain relations, and the difference in their co reference options remains a mystery. 7 However, we will see now that there are syntactic differences between the (a)- and (b)sentences, which suggests that they do not have the same surface structure.8 3.2.1 The Syntax ofPreposed PPs

Within the COMP theory it would be assumed that if there is a transformation of PP-preposing, the PP is moved into COMP position. (This analysis of PP preposing is developed in Emonds, 1976.) Since it is currently-believed that there is only one COMP position in the sentence, this means that if a PP is preposed, the COMP is filled and no other constituent, such as a wh-word, can be moved into·it, or, alternatively, that if the COMP position is filled either by a moved wh-word, or by a Q(uestion) or Imp(erative) marker, PP-preposing cannot take place. If we look now at the sentences below, we will see that this is irtdeed true in the case of preposed verb-phrasal PPs, but the sentential PPs do not

Prepositioned Phrases and Preposed Constituents

69

obey any such restrictions. Thus, wh-questions are possible after the sentential PP in (34a ), though not after the verb-phrasal PP in ( 34b ). (34a) In Ben's picture of her, how does she look? (b) *In Ben's picture of her, what did she find? (35a) In Washington, who do they worship? (b) *In Detroit, who did the gangsters kill? (36a) According to Felix, who won the race? (b) *In front of Felix, who holds a candle? (37a) With her new job, why can't Rosa be more cheerful? (b) *With her new boss, why does(n't) Rosa argue? Yes/no questions are also possible in the case of pre posed sentential PPs, as in (38a) and (39a), but not in the case of preposed verb-phrasal PPs, as in (38b) and (39b ). (38a) In Ben's picture, does Rosa look her best? (b) *In Ben's picture, did Rosa find a scratch? (39a) With her new job, can she spend more money? (b) *With her boss, does she argue? The same is true for imperatives: (40a) In my next picture, look more cheerful, please! (b) *In my next picture, find a scratch, if you can! (41a) With your new job, go spend more money! (b) *With your boss, stop arguing! These facts show that preposed verb-phrasal PPs are indeed moved into COMP position, but that sentential PPs require different treatment. They occur in a position preceding the COMP, and since COMP is the leftmost constituent of a sentence, this suggests that they are attached to a higher node than the one dominating COMP. (Alternatively it is possible that these PPs are not moved at all and that they can be generated both in final and in higher initial positions.) Banfield (1973) has suggested that certain presentential elements like exclamations or conjoined predicates or NPs (e.g. the first conjunct in the sentence One more glass o[beerand/or I'm leaving) are to be generated under a higher category E(xpression), which can expand to a presentential constituent and S. This proposal was adopted in van Riemsdijk and Zwartz (1974) for the analysis ofleft-dislocated (as opposed to topicalised) constituents.

Prepositioned Phrases and Preposed Constituents

70

Preposed sentential PPs share some properties with left-disbcated and presentential constituents - they are not semantically dependent, on the sentence (as we saw in Section 3.1.2), and they can be described as 'setting the scene' (in Kuno's, 1975b, terminology) for the rest of the sentence. (I will return to their semantic function in the sentence in Section 3.2.3 .) They seem, therefore, possible candidates for generation under E. This decision, however, is not crucial, and I will not elaborate on it. 9 What is crucial, though, is that there is a syntactic difference between the two types of sentence, which is illustrated in (42). (42a) (preposed) sentential PPs

/s~~

(b) preposed verb-phrasal PPs -

s

E

pp

COMP NP

VP

J's~

(COMP)

I

NP

VP

pp

Another possible indication of the structural difference is provided by quantified sentences of the two types. We saw in (17), repeated here as (43), that there are scope differences between the two types when the PP appears S-finally: (43a) Someone is riding a horse in all of Ben's pictures. (b) Someone found scratches in all of Ben's pictures. Roughly, the difference was that in (43a) it can be either the same person in all of Ben's pictures or a different person, while in (43b) it must be the same person who found scratches in all the pictures (which means that someone must have wider scope). Now let us see what happens if we prepose the PP: (44a) In all of Ben's pictures someone is riding a horse. (b) In all of Ben's pictures someone found scratches. The (a)-sentence loses its ambiguity: (44a) has only the reading with wider scope for all (i.e. it need not be the same rider in each picture). The {b)-sentence, on the other hand, becomes ambiguous. Some speakers cannot assign any reading to (44b ), probably, for pragmatic reasons that will be discussed in Section 3.2.3, but those who get the sentence (and a stress on all should help) agree that it is ambiguous: it can be one

?repositioned Phrases and Preposed Constituents

71

person or different people who found scratches in all the pictures. These facts do not yet supply us with evidence for the structural differences between ( 44a) and (44b ), since I have not yet established the correlation between domain relations of quantifiers and their scope, which I will do in Chapter 9. However, note that there is very little hope of finding a non-syntactic account of this difference. Ioup (1975) has argued that topics and subjects have identical status with respect to scope (with a slight preference for a wide scope for topics). However, the major function of preposing is to make the preposed constituent a topic (which means that preposed constituents are always topics, unless they have a special intonation marking). Consequently, in (44), both PPs are topics, so by this criterion they should have identical scope options. If we look now at the syntax of the sentences, we will see that when the verb-phrasal PP of (43b) is preposed, its domain relations become identical to those in (43a), with a sentential PP in final position- the only difference being in linear order of the quantifiers. A comparison of (45b) and (45c) makes this clear. It would be recalled that for the full definition of c-command in (I 7) of Chapter I, there is no difference between the domains of nodes dominated by S and S.

s

(45a)

~---------------VP

NP

~

~

someone

'pp

~

found scratches

L~

in all of Ben's pictures

{b)

~~ (Cr NP

V~

~

in ali of Ben's pictures

found scratches someone

72

Prepositioned Phrases and Preposed Constituents (c)

s

D 6

NP

VP

I

is riding a horse

someone

pp

in all of Ben's picture

In both (45b) and (45c), the NP and the PP are in each other's domain. In both of these cases, we get scope ambiguity. Other things being equal, if the sentence in (44a) had a structure identical with that of (44b)namely (45b) - the two sentences would have identical scope options. It is plausible, therefore, to conclude that they do not have identical structures. 3.2.2 Coreference in Preposed PPs Given the structures in (42), the coreference difference between the (a)and (b)-sentences in (30)-(33) is no longer a mystery. The (a)-sentences (one of which is repeated in (47)) have the structure of (42a), repeated in (46), in which the subject NP 2 does not c-command NP 1 in the PP, since the definition of c-command in (17) of Chapter 1 allows it to ccommand only nodes dominated by S- see footnote 9. In other words, NP 1 in the PP is not in the domain of the subject NP2 • Also, the subject NP2 is not in the domain of NP 1 (since the latter c-commands only the nodes in the PP). The coreference rule, therefore, puts no restrictions on the coreference options of these two NPs, and we can get 'forward pronominalisation', as in (47a), 'backward pronominalisation', as in (47b), or no pronominalisation at all, as in (47c). (46)

E (or S)

~s

P{

~ p

NP,

(\s

I :\----____

COMP

NP2

A

V

NP

Prepositional Phrases and Preposed Constituents 73 (47a) In Ben's office, he is an absolute dictator. (b) In his office, Ben is an absolute dictator. (c) In Ben's office, Ben is an absolute dictator. The (b)-sentences of (30)-(33), one of which is repeated in (49a), have the structure (42b), repeated in (48). In this structure, the subject NP2 does c-command the PP (by (17) of Chapter 1), which means that NP 1 in the PP is in the domain of the subject. The coreference restriction, therefore, requires that NP 1 must be a pronoun in order to be coreferential with NP2 , hence blocking coreference in (49a) and (49c). As in tree (46), however, the subject is not in the domain ofNP 1 (whose domain is only the PP). Hence the rule does not block (49b ). In sentences with the structure (48), then, the only grammatical way to express coreference is as in (49b) (unless, of course, both NP 1 and NP2 are pronouns). (48)

/"s~

(COMP)

NP 2

I

PP

~NP

P

VP

~

V

NP

1

(49a) *In Ben's office, he placed his new brass bed. (b) In his office, Ben placed his new brass bed. (c) *In Ben's office, Ben placed his new brass bed. It should be mentioned that the decision concerning where the PP is attached, when preposed, does not depend on any semantic properties of the PP itself. The rule is purely structural: constituents preposed from the VP are attached to the COMP of S, while sentential PPs are attached to a higher node if they are pre posed, or, more plausibly, they can be generated both in final position under S and in initial position higher than S. For example, in sentence (50), the PP in his next picture is sentential in its own clause (S 2).

(50) [s 1 Ben promised Rosa [s 2 that she would look more attractive in his next picture] ]

74

Prepositional Phrases and Preposed Constituents

s2

However, itself is in the VP of St; hence when the pp is preposed, it behaves like a VP constituent and is attached to COMP of S1• This is confirmed by the fact that the COMP position in this case cannot be filled by a wh-word, as in (51). (51) *In his next picture, how did he promise Rosa that she would look? Consequently, the preposed PP in such cases is in the domain of the subject; hence the NP in the PPmust be a pronoun, ahd (52a) is blocked. (52a) *In Ben's next picture, he promised Rosa that she would look more attractive. (b) In his next picture, Ben promised Rosa that she would look more attractive. The last point is, perhaps, not too convincing since pre posing a PP out of an embedded clause is not free, and, for some speakers, it is impossible. 10 Hence, for these speakers the sentences in (52) are unacceptable regardless of coreference. The argument holds, therefore, only for speakers who can prepose in such cases. Speakers who can get the (a)-sentences below (which do not involve coreference problems) clearly distinguish between the (b)- and (c)-sentences. (53a) In her wedding picture, she [hopes [that she will look attractive.]] VP S2 (b) In her wedding picture, Rosa hopes that she will look attractive. (c) *In Rosa's wedding picture, she hopes that she will look attractive. (d) In Rosa's wedding picture, she looks attractive. (54a) In his family, he [told me [that he is considered a genius.]] VP

s2

(b) In his family, Ben told me that he is considered a genius. (c) *In Ben's family, he told me that he is considered a genius. (d) In Ben's family he is considered a genius. The PP in (53) and (54) is sentential in its clause (and consequently, if this clause is not embedded, 'forward pronominalisation' is possible when the PP is pre posed, as in the (d)-sentences). However,since in the (a)-(c)sentences the S node which dominated the PP before preposing (S2) is dominated by theVP,thePPbehaves as a V-phrasal PPwhen it is preposed, and 'forward pronominalisation' as in (53c) and (54c) is impossible.

Prepositional Phrases and Preposed Constituents 75 PPs that consist of P and S (such as after-, when- or because-clauses) are always sentential (see Williams, 1974, 1975), which explains why coreference can go both ways when they are in initial position (as in (55)). However, in cases where it is clear that these PPs must have originated in a clause inside the VP, no 'forward pronominalisation' is possible, as illustrated in (56). (55a) When Rosa finishes school, she will go to London. (b) When she finishes school, Rosa will go to London. (56a) *When Rosa finishes school,she has promised Ben that she will go to London. (b) When she finishes school, Rosa has promised Ben that she will to to London. Here again, questions are possible after the PP in (55), as in (57a), but not after the one in (56) (as in (57b)), which indicates that only the PP in (56) is in COMP position, or, in otber words, that in (56), but not in (55), the PP is in the domain of the subject. (57 a) When she finishes school, will Rosa go to London? (b) *When she finishes school, has Rosa promised Ben that she will go to London? These facts further support the claim made here to the effect that the coreference options of NPs are determined on the basis of their surface structure domains. Needless to say, neither the relation of precede-andcommand nor the distinction between grammatical relations (subject, object, etc.) provides an adequate account of these facts. Preposed PPs still provide other coreference mysteries, such as the change in coreference options when the PP is lengthened. Since such cases depend even more crucially on the interpretation given to the sentence, before attempting an account of them, we should have a brief look into the interpretation of preposed PPs.

3.2. 3 Some Aspects of the Semantics and Pragmatics of Preposed PPs The semantics of preposed PPs is quite puzzling. Of the semantic criteria that distinguished the two types of PPs in final position (see Section 2.1.2), only the one concerning entailments still distinguishes them in initial position: the sentence In Ben's picture Rosa found scratches, still entails Rosa found scratches, and its negation still does not entail Rosa didn't find scratches. Similarly, the sentence In Ben's picture, Rosa looks

76

Prepositional Phrases and Preposed Constituents

attractive behaves with respect to entailment in the same way that it did with the PP in final position: namely, there is no entailment. We have already seen, in Section 3.2.1, that quantifier scope does change with preposing, and we will see now that theme-rheme relations change as well. Preposed (or topicalised) constituents are known to be the topics of the sentence -or to express old information. 11 (This holds for 'normal intonation': as noted by Gundel, 1974, topicalised constituents can also serve as contrastive focus with the proper contrastive intonation.) So, when verb-phrasal PPs are preposed, they function as themes (or topics), and theme-rheme tests no longer distinguish them from sentential PPs. Thus, anyway, which is impossible in final verb-phrasal PPs, as in (c) below, is possible both in the case of (a) and of {b).

(58a) In Ben's picture anyway, Rosa looks sick. {b) In Ben's picture anyway, Rosa found a scratch. (c) *Rosa found a scratch, in Ben's picture anyway. [construe anyway in (c) as attaching to the PP] {59a) In his office anyway, Ben is an absolute dictator. {b) In his office anyway, Ben placed his new brass bed. (c) *Ben placed his new brass bed, in his office anyway. [construe anyway in (c) as attaching to the PP] So, a PP in initial position is a thematic element, or a topic, regardless of whether it originates as sentential or verb-phrasal. Still, intuitively, we feel that there is some difference in the function of the two types of PP in these sentences. Kuno (1975b) has described, without much detail, sentential PPs like those in the (a)-sentences above as 'setting the scene' for the rest of the sentence. This correlates also to Dik's (1978) distinction between topics and themes. Although this seems intuitively right, it is worthwhile to try and make somewhat more explicit the intuitions which underlie such descriptions as 'setting the scene'. We have already seen that the explanation for the different functions of the two PPs in the sentence cannot come from examining the information status of the PPs themselves, since both are topics. The answer must lie, therefore, in the information status of the rest of the sentence, or in the exact relation of the PP to the sentence. To see this, let us focus on the subjects of sentences with preposed PPs. As we see in (60), sentential PPs can be followed by indefinite subjects. But indefinite subjects are much worse in sentences with preposed verb-phrasal PPs, as in the sentences of {61 ). (Contrastive intonation of the subject, or a 'list reading', in the sentences of (61) should be excluded.Y 2

Prepositional Phrases and Preposed Constituents 77 (60a) In Ben's picture, a fat woman is riding a horse. (b) In Ben's family, a cousin always did the dishes. (c) With such poor security arrangements, a thief managed to walk off with the office football pool. (d) In spite of the efforts of the police, a bomb exploded yesterday in the courthouse. (61a) #In Ben's picture, a fat woman found a scratch. (b)# In Ben's office, a stranger spent the night. (c) #With the boss, a client has been arguing bitterly. (d) #With a loud noise, a bomb exploded yesterday in the courthouse. (e)# In a great huff, someone just left the boss's office. Since indefinite nouns carry no existential presupposition, they usually convey new information and they cannot easily serve as topics. The fact that their occurrence in subject position in the sentences of (61) results in somewhat awkward sentences suggests, therefore, that constituents in this position tend to be interpreted as old information. In other words, it is not only the preposed PP which must be the theme in such structures, but the subject as well. The function of PP-preposing in the case of verbphrasal PPs is not just to make the PP a theme, but mainly to make the VPthe rheme (new information). Hence, given some notion of pragmatic 'aboutness' (see e.g. Reinhart, 1981b) sentences with preposed verbphrasal PPs state something 'about' the PP and 'about' the subject. The fact that, as we saw in the sentences of (60), subjects in sentences with initial sentential PPs do not have to be definite, suggests that they do not function as topics. In these cases, the whole main clause is the new information, and the PP alone is the topic. The sentence In Ben's picture, Rosa is riding a horse can be interpreted as stating that Rosa's riding a horse is a member of the set of things that took place in Ben's picture (or the set of properties of Ben's picture). The sentence with the verb-phrasal PP- In Ben's picture, Rosa found a scratch - on the other hand, is interpreted as stating that finding a scratch is a member of the intersection of the set of the properties of Ben's picture (things that happened in or to it) and Rosa's properties (things that Rosa did), or, in other words, that it is a member of the set of things that Rosa did to Ben's pictureY The difference in the iniormation ( theme-rheme) relations in sentences with preposed sentential and V-phrasal PPs suggests another type of correlation between syntactic domains and semantic properties. The structures underlying the two types of sentences are repeated in (62h) "0

and Xncn