Linguistic Analysis: From Data to Theory 9783110222517, 9783110222500

This book reconsiders the classic topics of linguistic analysis and reflects on universal aspects of language from a typ

226 25 3MB

English Pages 412 Year 2011

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Linguistic Analysis: From Data to Theory
 9783110222517, 9783110222500

Table of contents :
Table of contents
Introduction
1. Categories and functions
1.1. Different levels of analysis
1.2. The theoretical framework: Generative Grammar
1.2.1. X-bar (X') Theory
1.2.2. The head-complement parameter
1.2.3. Levels of analysis and X' structure
1.2.3.1. The Syntax-Semantics interface: Verb Phrase
1.2.3.2. The Syntax-Morphology interface: Inflectional Phrase
1.2.3.3. The Syntax-Discourse grammar interface: Complementizer Phrase
1.3. Parts-of-speech classification: The concept of category
1.3.1. A formal approach to classification
1.3.2. The concept of grammaticality
1.4. Categories: Typology and formal properties
1.5. Categories and syntactic functions
1.5.1. Subject function
1.5.2. Predicate function
1.5.3. Modifier function
2. The structure of the verb phrase
2.1. Predicates, arguments and structural categorization
2.2. Argument structure
2.2.1. Macro-roles and event structure
2.2.2. Mono-argument verbs
2.2.3. Two-argument verbs
2.2.4. Three-argument verbs
2.2.5. Macro-roles and VP structure
2.3. Argument roles, event structure and overt syntax
2.3.1. Event structure and macro-role of the subject in inflecting and agglutinative languages
2.3.2. Event structure and macro-role of the subject in incor-porating languages
2.3.3. Event structure and macro-role of the subject in predicate-initial languages
2.4. Arguments realized by clauses
2.5. Clauses as nouns
3. Syntactic functions
3.1. The relationship between deep roles and surface structure
3.2. Deep and structural Cases
3.2.1. Correlations and tendencies
3.2.2. When structural function is not associated with deep Case: Expletive subjects
4. The structure of the noun phrase
4.1. The internal structure of the NP
4.1.1. Elements in Spec,NP
4.1.2. Elements in Compl,NP and restrictive noun modifiers
4.1.3. When the head is a nominalization
4.2. NPs as arguments
4.3. The functional area of NP: The determiner phrase (DP)
4.4. Noun head modification
4.4.1. Adjectival modification
4.4.2. The relative clause
4.4.2.1. The internal structure of restrictive relative clauses
4.4.2.2. Relative clauses from a typological perspective
4.4.2.3. Types of relative clauses and their meaning
5. Adverbial modification
5.1. Adverbs and adverbials: Preliminary considerations
5.2. Properties and distribution of adverbial modifiers
5.3. Structural insertion of adverb phrases
5.4. Noun and prepositional phrases with adverbial function
5.5. Clausal structures with adverbial function
5.5.1. The internal syntax of adverbial clauses
5.5.2. The external syntax of adverbial clauses
6. The sentence as utterance
6.1. Discourse grammar and information structure
6.2. Speech acts
6.3. Information structure in declarative clauses
6.3.1. The notion of markedness
6.3.2. Given and new in marked and unmarked structures
6.4. Focus
6.4.1. Focus strategies in typologically different languages
6.4.2. Focus as a specificational predicate
6.4.3. Focus-prominent languages
6.4.4. Syntactic interface: The complex structure of CP
6.4.5. Prosodic interface
6.5. Topicalization
6.5.1. The functions of Topic in discourse
6.5.2. Phrasal restrictions on the topicalized constituent
6.5.3. Topicalization strategies in typologically different languages
6.5.4. Topic-prominent languages
6.6. Contrast
7. Illocutionary force
7.1. Types of illocutionary acts
7.1.1. Overt performatives
7.1.2. Covert performatives
7.1.2.1. Formal markers of the covert performative
7.1.2.2. Formal markers in the presence of a performative
7.2. Illocutionary force and performative structure
7.2.1. The distribution of information in different types of sentences
7.2.1.1. Requests for information
7.2.1.2. Requests for action
7.3. Illocutionary force and subordinate clauses
7.3.1. Declarative sentences
7.3.2. Interrogative sentences
7.3.3. Imperative sentences
7.4. More on illocutionary acts
7.4.1. Performative structure modification
7.4.2. Indirect illocutionary acts
Conclusions
List of languages
List of abbreviations
Notes
References
Subject index
Language index

Citation preview

Linguistic Analysis

Trends in Linguistics Studies and Monographs 220

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume

Walter Bisang

De Gruyter Mouton

Linguistic Analysis From Data to Theory

by

Annarita Puglielli Mara Frascarelli

De Gruyter Mouton

ISBN 978-3-11-022250-0 e-ISBN 978-3-11-022251-7 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data Puglielli, Annarita. Linguistic analysis : from data to theory / by Annarita Puglielli, Mara Frascarelli. p. cm. ⫺ (Trends in linguistics. studies and monographs ; 220) Includes bibliographical references and index. ISBN 978-3-11-022250-0 (hardcover : alk. paper) 1. Linguistic analysis (Linguistics) I. Frascarelli, Mara. II. Title. P126.P84 2011 410⫺dc22 2010050359

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/New York Originally published as L’analisi linguistica. Dai dati alla teoria ” Caissa Italia editore, Cesena/Roma, 2008 Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Table of contents Introduction…………………………………….……………………

1

1.

Categories and functions………………………………………

5

1.1. Different levels of analysis……………………………………... 1.2. The theoretical framework: Generative Grammar……………… 1.2.1. X-bar (X') Theory…………………………………… 1.2.2. The head-complement parameter…………………… 1.2.3. Levels of analysis and X' structure…………………. 1.2.3.1. The Syntax-Semantics interface: Verb Phrase……………………………… 1.2.3.2. The Syntax-Morphology interface: Inflectional Phrase……………………….. 1.2.3.3. The Syntax-Discourse grammar interface: Complementizer Phrase……...... 1.3. Parts-of-speech classification: The concept of category………... 1.3.1. A formal approach to classification………………… 1.3.2. The concept of grammaticality……………………... 1.4. Categories: Typology and formal properties ………………...… 1.5. Categories and syntactic functions……………………………... 1.5.1. Subject function…………………………………….. 1.5.2. Predicate function…………………………………… 1.5.3. Modifier function…………………………………….

5 6 8 11 15

2.

16 17 19 20 22 25 28 32 33 38 41

The structure of the verb phrase…………………………….. 43

2.1. Predicates, arguments and structural categorization……………. 2.2. Argument structure……………………………………………... 2.2.1. Macro-roles and event structure…………………..... 2.2.2. Mono-argument verbs………………………………. 2.2.3. Two-argument verbs………………………………… 2.2.4. Three-argument verbs………………………………. 2.2.5. Macro-roles and VP structure………………………. 2.3. Argument roles, event structure and overt syntax……………… 2.3.1. Event structure and macro-role of the subject in inflecting and agglutinative languages……………

43 45 49 52 55 61 62 66 66

vi

Table of contents

2.3.2. Event structure and macro-role of the subject in incor-porating languages…………………………. 2.3.3. Event structure and macro-role of the subject in predicate-initial languages……………………...... 2.4. Arguments realized by clauses………………………………...... 2.5. Clauses as nouns………………………………………………… 3.

73 77 79 83

Syntactic functions…………………………………………...... 89

3.1. The relationship between deep roles and surface structure…....... 89 3.2. Deep and structural Cases………………………………………. 90 3.2.1. Correlations and tendencies…………………………. 91 3.2.2 When structural function is not associated with deep Case: Expletive subjects………………………. 98 3.3. Deep roles in surface structure: Ergative languages……………. 100 3.4. The morpho-syntactic realization of structural Cases………....... 104 3.4.1. Subject………………………………………………. 104 3.4.2. Direct object………………………………………… 112 3.5. Morphological Case: Cross-linguistic strategies……………….. 116 4.

The structure of the noun phrase……………………………. 129

4.1. The internal structure of the NP………………………………… 4.1.1. Elements in Spec,NP……………………………….. 4.1.2. Elements in Compl,NP and restrictive noun modifiers………………………………....…… 4.1.3. When the head is a nominalization…………….…… 4.2. NPs as arguments………………………………………….……. 4.3. The functional area of NP: The determiner phrase (DP)……..... 4.4. Noun head modification…………………………………….….. 4.4.1. Adjectival modification…………………………….. 4.4.2. The relative clause………………………………….. 4.4.2.1. The internal structure of restrictive relative clauses…………………………… 4.4.2.2. Relative clauses from a typological perspective……………………………….. 4.4.2.3. Types of relative clauses and their meaning………………………………......

129 130 133 136 139 142 148 149 161 161 166 172

Table of contents vii 5.

Adverbial modification……………………………………….. 177

5.1. 5.2. 5.3. 5.4. 5.5.

Adverbs and adverbials: Preliminary considerations…………… 177 Properties and distribution of adverbial modifiers……………… 178 Structural insertion of adverb phrases………………………….. 181 Noun and prepositional phrases with adverbial function……… 194 Clausal structures with adverbial function……………………… 201 5.5.1. The internal syntax of adverbial clauses…………..... 201 5.5.2. The external syntax of adverbial clauses…………… 211

6.

The sentence as utterance…………………………………….. 219

6.1. Discourse grammar and information structure………………..... 219 6.2. Speech acts……………………………………………………… 219 6.3. Information structure in declarative clauses…………………..... 223 6.3.1. The notion of markedness………………………….. 224 6.3.2. Given and new in marked and unmarked structures……………………………………………. 226 6.4. Focus……………………………………………………………. 234 6.4.1. Focus strategies in typologically different languages……………………………………………. 236 6.4.2. Focus as a specificational predicate………………… 239 6.4.3. Focus-prominent languages…………………………. 243 6.4.4. Syntactic interface: The complex structure of CP….. 244 6.4.5. Prosodic interface…………………………………… 246 6.5. Topicalization…………………………………………………... 247 6.5.1. The functions of Topic in discourse………………… 248 6.5.2. Phrasal restrictions on the topicalized constituent...... 255 6.5.3. Topicalization strategies in typologically different languages………………………………..... 258 6.5.4. Topic-prominent languages…………………………. 261 6.6. Contrast…………………………………………………………. 264 7.

Illocutionary force…………………………………………….. 269

7.1. Types of illocutionary acts……………………………………… 269 7.1.1. Overt performatives…………………………………. 270 7.1.2. Covert performatives……………………………….. 275 7.1.2.1. Formal markers of the covert performative……………………………… 276

viii Table of contents 7.1.2.2. Formal markers in the presence of a performative……………………………. 276 7.2. Illocutionary force and performative structure…………………. 281 7.2.1. The distribution of information in different types of sentences……………………………….….. 283 7.2.1.1. Requests for information…………….…… 284 7.2.1.2. Requests for action………………….……. 290 7.3. Illocutionary force and subordinate clauses……………….……. 293 7.3.1. Declarative sentences………………………….……. 294 7.3.2. Interrogative sentences……………………………… 297 7.3.3. Imperative sentences……………………………….. 302 7.4. More on illocutionary acts……………………………………… 305 7.4.1. Performative structure modification………………… 305 7.4.2. Indirect illocutionary acts ………………………...... 307 Conclusions………………………………………………………..... 311 List of languages…………………………………………………..... 317 List of abbreviations………………………………………………… 321 Notes………………………………………………………………..... 323 References…………………………………………………………… 359 Subject index………………………………………………………… 393 Language index……………………………………………………… 401

Introduction

Writing a textbook invariably entails, on the one hand, the wish to explore different aspects within a specific discipline which appear to be significant and, on the other, the opinion that those aspects have not been investigated before, or at least not in a totally satisfactory way. With this in mind, this book is intended to offer (non-beginner) university students, researchers and scholars a type of linguistic analysis which accommodates typological and comparative considerations within a formal framework where the interaction between the various levels of grammar is taken into account. These components are reflected in the very title of the book and form the basis for each of its seven chapters. Combining formal analysis with typological considerations has clearly entailed specific methodological choices as to how our data are presented. A completely inductive approach would certainly have been most appropriate, as it would have enabled the reader to rely on data alone to arrive at those structural generalizations which we believe to be central and universal in the language system. However, as we wanted to illustrate specific (and sometimes complex) phenomena in typologically different languages, we decided that a totally free approach might be confusing (especially for students, the main intended reader of this book), and our comparative perspective would be lost. For this reason, we opted for a semi-guided presentation: The description of each phenomenon is preceded by a short introduction, where the relevant literature is presented and commented upon. Subsequently, data are described and discussed, also in the light of structural considerations. Finally, generalizations are proposed which attentive readers may have already predicted. Our objective is to show that, although individual grammars are different on the surface, the language system is based on just a few universal principles which determine the formal (morpho-syntactic and phonological) properties of utterances and various aspects of their semantic-pragmatic content. These principles operate on the system as a whole, thereby producing different effects at the various levels of grammar, ensuring interpretation and avoiding redundancies (based on a general economy principle). Hence there is the need for an integrated type of linguistic analysis, where data are not considered as exclusively belonging

2

Introduction

to a certain domain. Indeed, their understanding relies on the notion of interface to account for the relationship between identical contents at a deep level and diversified structure realizations. For this reason, while single chapters are dedicated to specific aspects of linguistic analysis, none of them should be thought of as self-contained and separate from the rest of the book. The description of phenomena will always entail backward and forward references to notions or data, showing how the various levels interact and proving that apparently independent facts can be accounted for by using a common generalization. The order of chapters reflects the unitary and gradual progression of our research, and reading them in sequence aids comprehension. Regarding individual topics, the first three chapters provide the conceptual basis for the subjects discussed in the remainder of the book. Chapter 1 (Categories and Functions) first presents a brief introduction of Generative Grammar and the theoretical framework of reference (though we do not go into formal details when discussing our data). Then, the concepts of category and function are discussed, both individually and in how they interact. Chapter 2 (The structure of the verb phrase) explores the notion of argument structure in depth, both semantically and syntactically. In particular, as the interface between these two levels is considered, the concept of macro-role is of central importance, as is the (dynamic vs. stative) event structure of predicates. In Chapter 3 (Syntactic functions), the surface functions of constituents are examined. The distinctions made in the preceding chapter are immediately shown to account for a number of phenomena relating to the difference between deep case and structural case. Finally, the two major functions (subject and object) are analyzed, along with their realizations in typologically different languages. Chapter 4 (The structure of the noun phrase) presents the analogy between noun and verb phrases, a discussion of the notion of nominalization and a description of the properties of the major noun head modifiers, i.e. adjectives and relative clauses. Chapter 5 (Adverbial modification) is also dedicated to constituents functioning as modifiers, in this case at the level of predication: The object of our analysis is the distribution of adverbials, their internal composition and the (syntacticpragmatic) relationship with the sentences to which they belong. The two final chapters deal with the interfaces between morphosyntax and intonation and between discourse grammar and illocutionary force respectively. In particular, Chapter 6 (The sentence as utterance) is intended to analyze the order of constituents on the basis of two informational pairs: Focus-Presupposition and Topic-Comment. Their

Introduction

3

realizations and interactions at the various levels of analysis are taken into account and cross-linguistic Focus and Topic strategies are examined. Particular attention is devoted to the notion of Contrast. Finally, in Chapter 7 (Illocutionary force) illocutionary acts are described, along with the covert and overt realization of performative verbs. Semantic-pragmatic considerations and cross-linguistic comparisons show that illocutionary force should also be included within a formal framework, in which (possibly null) performatives are assumed to be present in all utterances, thus determining their properties and interpretation. The data presented in this book, an essential element in our discussion, represent 74 languages from different language families and types, spoken in different parts of the world (see the List of languages). Our data were partly taken from previous descriptions (sources are cited before each relevant example), and partly obtained with the help of a number of informants, whose contribution was truly invaluable.1 We would like to thank Abdalla Omar Mansur for Arabic and Somali, Ahmed Abdullahi Ahmed for Somali, Mohammed Ali Mahmoud for Afar, Marites Navarro Alcaria for Tagalog and Eveliina Pohjola and Elina Di Santo for Finnish. Many scholars and language experts also provided opinions and clarifications on examples from existing works. Our heartfelt thanks go to Roland Hinterhölzl, Rosa Lombardi, Carlo Serra, Giuliano Lancioni, Rita Liberatori and Peter Douglas. Special thanks also go to Francesca Masini, Anna Pompei, Francesca Ramaglia, Giorgio Testa and Alessandra Trecci for carefully reading previous versions of this book and providing us with relevant and detailed comments which were evaluated and integrated into our work and proved a useful contribution to the final product. Annarita Puglielli Mara Frascarelli

1. Categories and functions

1.1. Different levels of analysis Language is by its nature a complex system and its comprehension may prove extremely difficult for anyone trying to describe and study it from a general and global perspective. On the other hand, the structure and functioning of this system reveal a coherent and harmonic organization if linguistic investigation focuses on the various levels of analysis that form such a system (i.e. phonology, morphology, syntax, semantics, pragmatics) and their basic interactions. The different levels must clearly remain distinct throughout the analysis, but it should be kept in mind that all of them jointly make up a language system. Only by assuming this approach will it be possible to conduct a successful study of language, conducive to better comprehension of its nature and working processes. The linguistic datum should therefore be examined and compared in its multiple aspects: Sounds identified as significant elements (phonology), segments of meaning that form their lexical basis (lexicon and morphology), argument structure (semantics), structural and distributional properties (syntax), co-reference relationships within a text and the illocutive force of a sentence (pragmatics). From a methodological viewpoint, this type of research is radically different from traditional analyses. Those analyses have usually assumed an inventory of (only partly defined) basic categories, and have mainly been based on a classification exercise, which is nothing more than an uninteresting and not particularly useful labelling exercise. A fruitful analysis should rely on the observation and systematic description of linguistic data in order to provide explanatory hypotheses. These hypotheses should enable the researcher to relate the various phenomena to a finite (and limited) number of principles, and to make predictions about parts of the system which have not yet been observed (later to be verified by using a larger amount of data). Such analysis may well be performed on a single language. However, it will be more useful and valuable if it involves an interlinguistic dimension, i.e. (typologically) different languages.

6

Categories and functions

Interlinguistic comparison and the interfaces between different levels of analysis will thus be viewed as the main axes of linguistic research for the purposes of this book.

1.2. The theoretical framework: Generative Grammar The theoretical framework used in this work is Generative Grammar. This syntactic model, developed by the American linguist Noam Chomsky in the late Fifties (Chomsky 1957), later underwent a number of revisions, giving rise to Generative-Transformational Grammar (Chomsky 1965, 1970), the Government and Binding Model (Chomsky 1981, 1982, 1986a, 1986b) and, finally, the Minimalist Program (Chomsky, 1993, 1995, 2001, 2002, 2004, 2005). Each revision was aimed at reducing the formal machinery, formulating the principles at the basis of Universal Grammar (see below) and going “beyond explanatory adequacy” (Chomsky 2004).1 Throughout its evolution, Generative Grammar has always adhered to innatism and remained true to its final aim, i.e. providing a solution to what is commonly known as Plato’s problem (Chomsky 1986a): How do children acquire language?2 Its objective is to identify the few, extremely productive linguistic principles which are part of the genetic makeup of an individual and are activated through interaction with the environment. A native speaker’s ability to distinguish between grammatical and ungrammatical sentences is ascribable to such innate capacity. A native speaker does not identify a sentence as correct because s/he has heard it before. Even sentences that are heard for the first time can be accepted or refused; this happens because speakers have an operative competence in their own language, which makes it possible for them to decide whether a sentence is syntactically correct (grammatical) or incorrect (see § 1.3.2. for a discussion of grammaticality). This capacity shows that humans must possess an innate mechanism that can generate (and judge) an unlimited number of sentences, an “underlying biological matrix” providing a framework in which language growth develops (see Chomsky 1980: 178). Such innate mechanism (as well as the principles that it comprises) defines Universal Grammar. Universal Grammar, as such, determines the idea of a “possible natural language”. This means that all languages must have some common properties. In particular, the theoretical model of Generative Grammar assumes two fundamental components in human language:

The theoretical framework

7

a) A LEXICON, i.e. a set of basic linguistic items with phonological and semantic features; b) A COMPUTATIONAL SYSTEM, whereby each linguistic expression (projected onto a linear sequence) must obey a hierarchical and recursive structure that determines linear order. These two components establish a mutual relationship through the operation of Lexical Insertion. As the classic Y-model in (1) shows, the grammar model presents a fixed derivational path, involving four levels of representation:3

(1)

Lexical Insertion

D-STRUCTURE

Syntax S-STRUCTURE

PF

LF

The process of Lexical Insertion results in the D-structure (D standing for ‘deep’), defined as “a pure representation of thematic relations” (Chomsky 1986a). From this basic structure, a number of movement operations (referred to by the general term Move Į) take place, whose aim is the realization of specific morpho-syntactic requirements. Once all the necessary operations are completed, the S-Structure (S standing for ‘surface’)4 is transferred for interpretation to the interface levels: Phonetic Form5 (PF) and Logical Form6 (LF). Interface levels are defined by Chomsky (1995) as the only “conceptually necessary” levels, because they represent what speakers interpret, both from an articulatory-perceptual point of view, and in terms of logical-semantic factors. Research should therefore focus on the set of syntactic derivations which enable words to combine into phrases and phrases, in turn, to form structures that are interpreted in specific ways and shared by a community of speakers.

8

Categories and functions

In this context, syntactic investigation plays a crucial role. In Chomsky’s theory, the syntactic relations of a language cannot be reduced to a mere sequence of words, but are founded on a more abstract structure, namely, the X-bar (X') structure.

1.2.1. X-bar (X') theory The primary objective of syntactic theory called X' theory is to explicitly represent the mechanisms which operate unconsciously as an individual is speaking (see Graffi 1994: 159). The theory is aimed at representing, with universal validity and scientific criteria, the internal structure of the various types of constituents and their different combinations producing sentences. Following this theory, the constituents in a sentence are arranged in a recursive structure,7 which is subject to a hierarchical order: A constituent with a specific internal structure contains another constituent with an identical structure. The main function of the different representation levels is to express speaker intuitions explicitly and to give a non-ambiguous graphical representation to their mutual links. The diagram proposed in the X' theory can be represented as follows:8 (2)

XP Spec(ifier) X° (Head)

X’

Compl(ement)

This schema is called a phrase marker (or tree diagram): The elements belonging to the Lexicon are merged into the structure (as heads, see below) via Lexical Insertion and are classified into different categories only when the phrase is complete (see § 1.3.). In this perspective, linguistic items assume a categorial status in a phrasal environment. The HEAD corresponds to the X° node, which is the founding and characteristic element in a phrase, typically not exceeding one word.9 The head is the only element that must necessarily be realized in this position; at the same time, it defines the type of phrases which can be placed to its left (in the Specifier position), to its right (in the Complement position) or outside its maximal projection X-Phrase (in an Adjunct position, see

The theoretical framework

9

below). Such elements integrate the information expressed by the head in different ways. The COMPLEMENT (which will be referred to as Compl) may be defined as the internal argument of the head (Graffi 1994: 160). A head and its complement are sister nodes, as they are at the same level in the phrase hierarchy, and together form the X' level, i.e. the intermediate projection level. The Compl position may be iterated. In other words, based on its semantic subcategorization (see § 1.2.3.1.), a head may select more than one phrase as its internal argument (as in the case of trivalent verbs; see Chapter 2). In this case, the X' structure presents a mechanism known as adjunction: The intermediate projection is replicated so as to reflect the existence of two constituents (Y-Phrase and Z-Phrase in [3]) which are located at the same hierarchical level as compared to the head: (3) XP Spec(ifier)

X’ X’

X° (Head)

ZP (Compl,2)

YP (Compl,1)

The SPECIFIER (which will be referred to as Spec) is so called because it qualifies or further specifies the head of a phrase (that follows it). It is the sister node to the X' level: Together they form the maximal projection (XPhrase), which includes the representation of the whole phrase and signals its completion. The Spec cannot be iterated and is always found to the left of the head. Finally, a head can be modified by a (potentially unlimited) number of constituents which are not part of its semantic subcategorization, but contribute additional, circumstantial information.10 Such elements are therefore called ADJUNCTS and, as such, are also represented in the tree by means of the adjunction mechanism. In this case, however, the projection undergoing duplication is not found within the phrase (as was shown in [3] in the case of Compl): the adjunction of these constituents to the maximal projection indicates their more external relationship with the head. It should finally be noted that the adjunction mechanism has no fixed direction. The

10

Categories and functions

adjunct node can develop both to the right (as in the W-Phrase in [4]) and to the left (as in the Z-Phrase): (4) XP XP

ZP (ADJUNCT)

WP (ADJUNCT)

XP

Spec X°

Compl

Evidently, the structure of X' theory does not only provide hierarchical information. As well as indicating that the head and its Compl are at a lower level than Spec, it also shows that the Spec precedes the head, which in turn precedes the complement (but see below for the head-complement parameter). The tree structure thus makes it possible to represent the fundamental properties of syntax: hierarchy and recursivity. The locations in the tree diagram where the different branches originate are called branching nodes, whereas the nodes with no branches departing from them are called terminal nodes. The highest node, from which the whole tree stems, is the root node. When a phrase structure is represented by drawing the corresponding tree diagram, it is important to remember that not more than two branches may originate from a single branching node (Binary branching principle; see Kayne 1984). Another way of representing the X' structure graphically, often used as an alternative, is by means of labeled brackets. In this case, a square bracket is opened at each branching node, bearing the name of that node as a subscript index. Each bracket is closed as soon as all the elements contained in that node have been included. The two representations of the Noun Phrase (NP)11 in (5) provided below are therefore completely equivalent (the position of the article, in brackets in [5], will be discussed in Chapter 4): (5) (The) book by Chomsky.

The theoretical framework (5')

11

NP N' PP

N book

P' P by

NP Chomsky

(5'') The [NP [N’ book [PP [P’ by [NP Chomsky]]]]] 1.2.2. The head-complement parameter Although the hierarchical structure is a fundamental property of human language, the diagram in (2) cannot be considered a linguistic universal proper, as variation can be found in the mutual order of head and complement. A syntactic parameter related to the linear order of a head and its complement (i.e. their precedence relationship) is needed in order to understand how languages work and to classify them into language types. Head-initial and head-final languages must be distinguished. In the languages known as head-initial (such as English, Italian and Arabic), a head linearly precedes its complement (as shown in diagram [2]). Head-final languages (such as Turkish, Japanese and Afar), on the other hand, present the opposite order, shown in (6): (6)

XP Spec(ifier)

X’

Compl(ement)

X° (Head)

12

Categories and functions

Due to the central role of verbs in sentence structure (see Chapter 2), head-initial languages are also called VO languages. In these languages, the verbal head (V) precedes its complement (i.e. its O[bject]). Similarly, headfinal languages are known as OV languages. The X' structure and the head-complement parameter provide an immediate structural interpretation of the typological principle of tendency, as formulated by Greenberg (1966) on the basis of cross-linguistic comparison. As Greenberg noted, regardless of the subject position, VO languages tend to share some characteristics, whereas OV languages share other, diametrically opposed characteristics: (7) a. head-initial: VO; typological tendencies: PrepN/NAdj/NRelC/NGen b. head-final: OV; typological tendencies: NPostp/AdjN/RelCN/GenN These systematic correlations result directly from the abstract structure shown in (2) for head-initial languages and in (6) for head-final languages: As the X' structure is universal, in a VO language every type of head is expected to precede its modifers (complements and adjuncts), whereas the opposite order is expected in OV languages, regardless of the family those languages belong to.12 In other words, as a consequence of the headcomplement parameter predictions can be made on linear order, and a preposition can be expected to precede its NP complement simply because a verbal head precedes its object. This is borne out by languages belonging to different families, such as Russian (SVO), Arabic (VSO) and Malagasy (VOS), respectively shown in the following examples:13, 14 (8)

Ona brosila kamni v ozero. (RUSSIAN) PRO.3SGF.NOM throw.PST.SGF stone.PL.ACC in lake.SG.ACC S V O Prep N (IO) ‘She threw the stones into the lake.’

(9)

‘akala al-talibu eat.PERF.3SG DET-student .NOM V S fi al-baiti. at DET-home.OBL Prep N (IO) ‘The student ate couscous at home.’

al-cuscusa DET-couscous.ACC

O (ARABIC)

The theoretical framework

13

(10) Nividy vonikazo hoan’i Maria i Joana. (MALAGASY) PST.ACT.buy flowers for DET Maria DET Joana V O Prep N (IO) S ‘Joana bought flowers for Maria.’ Similarly, as the object precedes the verbal head in languages such as Turkish and Korean (SOV languages),15 the presence of postpositions (as opposed to prepositions) in these languages is not surprising. Let us observe the postposition within the subordinate clause (functioning as an object) in example (11) from Turkish (Kornfilt 1997) and in sentence (12) from Korean (Ho-Min 1999): (11) Herkes [ben-im-le sinema-ya gel-esin] isti-yor. S.PRO.1SG-GEN-with cinema-DAT come-2SG.OPT want-PROG all N Postp (IO) (TURKISH) S O ‘Everyone wants you to come to the cinema with me.’

V

(12) Apeci-ka kay-hanthey mwuli-li-si-ess-ta. (KOREAN) father-NOM dog-by bite-PASS-HON-PST-DECL S N Postp (IO) V ‘My father was bitten by a dog.’ These tendencies clearly show how crucial the notion of head is, and how successfully the X' structure can account for language acquisition. Assuming that such parameter exists as part of Universal Grammar makes it possible to understand how a child can build her/his own grammar and produce previously unheard sentences, starting from simple and often repetitive sentences. In other words: A child in an English-speaking environment, exposed to simple sentences such as Want num nums? or Kiss mummy, will not only understand that her/his language is a VO language, but also (more importantly) that that particular language is right-branching, and consequently that every complement follows its head (noun head, preposition head etc.) in linear order. For that reason, an English-speaking child will immediately know that her/his language has prepositions (I go to school), postnominal relative clauses (the child that runs), complementizers preceding subordinate clauses (I think that I will leave) and so on. Similarly, a speaker of Afar (an OV language), when confronted with sentences such as (13), will know that his or her language is left-branching,

14

Categories and functions

and will productively apply this parameter to every type of head. S/he will not need to be taught this explicitly, nor to have heard sentences of this kind (see the postposition in [14], the prenominal relative clause in [15] and the complementizer at the end of the subordinate clause in [16]): (13) Ahmed lee yorobé. Ahmed water drink.PST.3SGM ‘Ahmed has drunk water.’ (14) A’nu ‘kaa ‘ko-h S.PRO.1SG O.PRO.3SGM O.PRO.2SG-to ‘I sent him to you.’

(AFAR)

ruu’be. send.PST.1SG

(15) [A’nu a’mo ‘kaa-k oogo’re] ‘too ‘num S.PRO.1SG head O.PRO.3SGM-from hit.PST.1SG that man yer’de. escape.PST.3SGM ‘That man [whom I hit on the head] has run away.’ (16) Yab’loon-u-h gedaa‘na. see.3PL-CONJ-for go.PRES.1PL ‘We are going to watch.’ Of course, the head-complement parameter allows for exceptions. Thus, languages with linear orders deviating from the head-complement parameter established by the V-O relationship are not infrequent. Adjectives, for example, are prenominal in VO English and admit some variation in Italian and French (both VO languages). On the other hand, in an OV language such as German16 prepositions should follow nouns and nouns should follow their modifiers (adjectives, relative clauses, phrases expressing possession). In fact, this head-final order is not always maintained. There are prepositions that necessarily precede their complements (such as mit ‘with’ or von ‘of’), as well as mixed situations, with some prepositions accepting both orders (as in the case of nach, which is prenominal in its directional meaning [17a] and postnominal in the meaning ‘after/according to’ [17b]); besides, relative clauses follow the head noun (17a):

The theoretical framework

15

(17) a. Wir fahren nach Berlin/ *Berlin nach. (GERMAN) S.PRO.1PL travel.1PL to Berlin ‘We are going to Berlin.’ b. Meiner Meinung nach / *Nach meiner Meinung. POSS.1.SG.F.DAT opinion according to ‘In my opinion.’ c. Der Mann [den Marie liebt] DET man REL.PRO.ACC Maria love.PRES.3SG heiȕt Paul. be called.PRES.3SG Paul ‘The man Maria loves is called Paul.’ Exceptions to the head-complement parameter may be ascribed to language change (diachronic variation), interaction with other levels of analysis (phonology, semantics) or language-specific parameters. Such complex interaction will be discussed in more detail in the following chapters. It should be observed, however, that these exceptions do not affect the predictive power and the explanatory capacity of the head-complement parameter in terms of language acquisition. In other words, by assuming innatism and postulating a Universal Grammar which contains a limited number of principles and parameters applied to a hierarchical and infinitely recursive structure, it is possible to conclude that the child learns only exceptions.

1.2.3. Levels of analysis and X' structure Following Generative Grammar principles, syntax is the generative level of grammar,17 which constructs the structure and interpretation of a sentence, by correlating with the other levels of analysis. As we saw in (1), in the standard model the logical-semantic (Logical Form) and the articulatoryperceptual (Phonetic Form) interpretations are defined by syntactic structure once all necessary operations are completed and the syntactic structure is formed (the Spell-out point, in Generative Grammar terms). The relationship between syntax and the other levels of grammar, however, begins from the moment of Lexical Insertion and continues throughout the whole derivation. In other words, the X' structure provides a representation in terms of syntactic areas. Within these successive areas, a number of interfaces are verified: With lexical semantics (Verb Phrase or

16

Categories and functions

VP area), with inflectional morphology for the grammatical categories of verb (tense, aspect and mood; see § 1.2.3.2.), agreement and Case assignment (Inflectional Phrase or IP area), and with pragmatics, defined as discourse grammar, information distribution and illocutive force of a sentence (Complementizer Phrase or CP area): (18) CP

Syntax-Discourse grammar interface C’



IP

Syntax-Morphology interface I’



VP

Syntax-Semantics interface V’



Such interactions, as well as the internal subdivision of each of the three areas, will be the topic of Chapters 2, 3 and 6 respectively. A preliminary description is given below.

1.2.3.1. The Syntax-Semantics interface: Verb Phrase Following the Extension Condition (Chomsky 1995, 2004), syntactic derivation proceeds bottom-up. This means that the Verb Phrase (VP) is the first syntactic object determined by Lexical Insertion. The VP is defined in Chomsky (1981: 43) as “the direct representation of thematic functions”, and the correlation between Deep structure and argument role assignment is still maintained in the most recent works (albeit with different terminology). The VP is therefore considered as the generative level, the syntactic locus where sentence constituents are merged, based on the argument selection of heads. When a lexical root is merged as the head of an X' structure, and that X' structure is completed in the Spec and Compl positions by arguments defined according to the thematic subcategorization of the head (Chomsky

The theoretical framework

17

1965), the resulting phrasal object is a VP (see Chapter 2 for a definition of argument roles). In particular, the insertion of arguments within the X' structure follows a specific hierarchy, which has to do with their degree of agentivity in the event described by the verb (see Givón 1984, Grimshaw 1990, Williams 1994). Argument roles with greater control are merged in Spec position: In particular, agent or experiencer (as selected by the verb). On the other hand, the Compl position (possibly iterated, depending on the number of arguments selected) hosts the argument roles with a lower degree of agentivity (such as patient, theme, beneficiary and locative):18 (19)

lexical projection related to argument structure

VP V’





1.2.3.2. The Syntax-Morphology interface: Inflectional Phrase As a universally valid model, the X' structure extends to the whole clause. The sentence is thus analyzed as the effect of recursivity of phrase structure (i.e. as a complex phrase), presenting the VP in Compl position. In particular, the head of the sentence-phrase is the syntactic locus where the features related to grammatical information – such as agreement (phifeatures), and tense, aspect and mood (generally known as TAM features) – are merged. The Spec of the Inflectional Phrase (IP), on the other hand, is typically associated to the realization of the subject, and thus to the assignment of Nominative (NOM) Case and features related to the Extended Projection Principle (EPP).19 The sentence is thus seen as the maximal projection of verbal inflection, and is assigned the IP label. In this description, the head of the sentential phrase is therefore a functional category (see § 1.3.), as its meaning is not denotative (also called “descriptive” or “referential”; cf. Geeraerts 2002), but grammatical. The sentential head contains the functional features needed to complete the lexical information of the verbal head.

Categories and functions

18

That information may be realized by means of a morphemic head (incorporated by the verb via movement from V°, where it is generated, to I°), or through an auxiliary (or modal) element, directly merged in I°. Similarly, the Spec of the IP is the syntactic locus of the functional features needed so that one of the arguments merged in the VP can take the role of syntactic subject (for a detailed discussion, see § 3.4.1.): (20) IP [NOM Case] [EPP]

Functional projection related to inflectional features

I’

VP I° [phi-features] [TAM features]

V’ V°



As illustrated in (20), following the Extension Condition, the VP structure examined in (19) must be thought of as the Compl of the new syntactic object (the IP), the place where the interface with inflectional morphology is realized. Assigning the role of sentential phrase head to the functional category of inflection is fully justified, as the information related to the VP argument structure is independent of categories such as tense, mood and agreement. For example, the participants in an event such as a meeting cannot be altered by the moment at which the event takes place (also see Graffi 1994: 188). Syntactic tests such as ellipsis (see Lobeck 1995, Merchant 2001) also show that the information conveyed by auxiliaries and modals is distinct from and independent of that of the verbal head. Examples of this are given in (21) and (22): (21) Mario can and should get better. (22) John said he would go and he will.

The theoretical framework

19

1.2.3.3. The Syntax-Discourse grammar interface: Complementizer Phrase Once the argument structure (in the VP) and the syntax-morphology interface (in the IP) have been realized, sentence interpretation is completed with information on discourse grammar and illocutive force. Such information, also functional, is realized in the functional projection of the Complementizer Phrase (CP). The Complementizer head (C°) usually includes a grammatical element – free in many languages –20 which typically introduces a subordinate clause, inducing a particular interpretation regarding the clause type (declarative, interrogative, etc.), as in the following example: (23) a. John doesn’t know that Mark has left. b. John doesn’t know if Mark has left. The C° position may, however, also be realized in root clauses, as in the case of (yes-no) interrogative clauses which require the movement of the auxiliary (from I° to C°) , in languages such as English, or the insertion of a functional head directly in C° in languages such as Arabic (see Chapter 6 for more details): (24) Are you going home? (25) Hal ikhtƗrat Maryam Q choose.PERF.3SGF Maria ‘Has Mary chosen her dress?’

al-libƗs? DET-dress

(ARABIC)

The Spec position of the CP node hosts elements such as interrogative phrases (26), focused constituents (see the Hungarian example in [27], from Horvath 1986) or topicalized constituents (see the Greek example in [28], from Tsimpli 1995). Here and in the rest of the volume, new information appears in small capitals, as is standard use: (26) Where has John gone?

20

Categories and functions

(27) JULISKÁT akarja hogy hívjuk meg Juliska.ACC want.3SG that invite.1PL PV vacsorára. (HUNGARIAN) dinner.SUBL ‘S/he wants us to invite JULISKA for dinner.’ (28) To

vivlio to-edhose i Maria sto book.ACC OCL.3SG-give.PST.3SG DET Maria to.DET Yani. (MODERN GREEK) Yani ‘The book, Maria gave it to Yani.’

DET

The CP is therefore the functional projection that represents the interface between syntax and pragmatics, the latter being redefined as discourse grammar. Discourse grammar information is then transferred onto the sentential phrase, which the C° head selects as its complement:21 (29) functional projection related to discourse features

CP C’ C° [Force] [+Q, +foc, etc.]

IP [NOM] [EPP]

I’

I° VP [phi-features] [TAM features]



V’



1.3. Parts-of speech classification: The concept of category The classification of the Lexicon into parts of speech, or word classes, has been a central topic for research since ancient times.22 More recently, the

Parts-of-speech classification

21

research on parts-of-speech systems has focused on the issue of their universality (see Croft 1991, Puglielli 2004, Simone 2006). Various typological studies23 have shown that while the Lexicon is divided into word classes universally, there are differences regarding the type and number of classes identified in different languages. Of the four open lexical classes (nouns, verbs, adjectives and adverbs), “nouns and verbs are major lexical classes in all languages” (Givón 2001: 49). Indeed, nouns and verbs are considered by most scholars as universal categories. On the contrary, adjectives are not usually thought of as universal, although scholars of different theoretical frameworks (for example Croft 1991, Baker 2003, Dixon 2004) have asserted that a lexical class of adjectives can be identified in all languages. Different criteria have been set forth in the literature to identify and distinguish word classes. Some of these have been proposed as universal criteria for classification within the Lexicon (and therefore crosslinguistically valid); others are more specifically morpho-syntactic and identify within an individual language a number of formal properties shared by words belonging to a specific lexical class. Croft (1991: 37) defines the former criteria as language-external (“also known as functional, or semantic/pragmatic in the realm of morphosyntax”) and the latter as language-internal (“also known as structural”). Generally, many authors support semantic criteria for identifying word classes (see Robins 1997, Givón 2001). According to these criteria, nouns usually designate concrete referents (people, things, places), verbs are words which refer to actions or processes, and adjectives express properties.24 Of course, this type of analysis is only useful in some cases, as it cannot account for either cross-linguistic differences (a specific concept may be expressed with words from distinct lexical classes in different languages) or the fact that distinct semantic features may be marked within an individual language (for an extensive discussion, see Ramaglia 2008). Similarly, morphological differences cannot be assumed as a discriminating factor: In many cases, elements belonging to word classes considered different show the same morphological behaviors (Schachter 1985, Anward 2001). Furthermore, the validity of a morphological criterion is denied by the fact that it simply does not hold true in a large number of languages (isolating languages; see § 1.4.). We therefore believe that a parts-of-speech classification based on semantic or morphological criteria is not feasible as it has limited validity and is only partly true. Following the principles applied in our theoretical framework, we propose that the distinction into word classes, and thus the

22

Categories and functions

idea of category, should be built within a phrasal frame and based on formal and distributional criteria. In other words, we propose that words do not contain any information related to classification within the memory storage defined by the Lexicon (and are therefore underspecified for that characteristic). On the contrary, that specification is defined by: (a) (b)

The type of elements that qualify as Spec, Compl, and, in general, as modifiers of the phrases headed by each of these elements; and The syntactic context where each phrase is merged in the tree.

Hence, if the memory storage is the paradigmatic axis (Saussure [1916] 1988), it can be concluded that it simply contains semantic roots, whose classes are in turn defined on the syntagmatic axis.25 According to this hypothesis, defining a word as belonging to a specific class means comparing it, within the object language, to words of different types, and grouping it with other words showing common morpho-syntactic properties. This is, in sum, the position supported by Schachter (1985: 3): “It is assumed that the primary criteria of parts-of-speech classification are grammatical, not semantic.”

1.3.1. A formal approach to classification Empirical data can help to illustrate the validity of a formal approach to word classification more clearly. Let us take a word like mass. If we were to identify just one category for this word, we would most probably describe it as a noun; a sentence such as (30a) confirms this. The same word, however, may be used in other contexts, such as those in (30b-c). This shows that the association between the word mass and the category noun is just one of the possibilities available in English: (30) a. The mass of Jupiter is 318 times that of the Earth. b. Three thousand students are going to mass in the plaza. c. Mass media are a powerful tool of propaganda. In particular, it can be observed that mass is (or rather displays the properties of) a verb in (30b) and qualifies as an adjective in (30c). Similar considerations are true for a host of lexical entries in English. Apart from the most obvious examples, where the choice is between a noun and a verb (such as in play or water), we can think of more nuanced cases:

Parts-of-speech classification

23

For example, the category of morphologically invariable words such as no or before, whose categorial association is definitely not as immediate for an English native speaker: (31) a. I won't take no for an answer. b. No answer is completely true. (32) a. He called me before. b. She's coming home before lunch. In these cases, the identification of the relevant category clearly depends on the function of the words in question. Only by resorting to syntactic functions and distributional criteria can no be defined as a noun in (31a) – where it is an object of the verb take – and as a determiner in (31b), where it is followed by a noun within a DP. Similarly, before in (32a) functions as an adverb, whereas its prenominal position qualifies it as a preposition in (32b). The same is observed in an isolating language such as Chinese. For example, what class does a word like qù belong to? Let us consider the sentences in (33a-b) and their respective translations: (33) a. TƗ qù

knj-zhè cry-IMPF le.

DIRECT

PERF

PRO.3SG

păo huí jiƗ run return home (CHINESE)

‘He/she ran back home crying.’ b. Wǂ xiànzài yào qù yóujú. PRO.1SG now must go post office ‘I have to go to the post office now.’ c. Ruguo n΃ qù wǂ ye qù. if PRO.2SG go PRO.1SG also go ‘If you go, I’ll go too.’ The two occurrences of the word qù are evidently connected in meaning: They both express the idea of motion, sharing a common semantic-functional root. However, if the word follows a noun (such as ‘home’ in [33a]), it assumes the role of a preposition (which is realized as a postposition, consistently with the final-head order of the language); conversely, if it follows a modal (33b), a focusing adverb (‘too’) or a

24

Categories and functions

subject (33c), it has to be considered as a verbal head (with predicative function). These considerations still hold true in languages with a richer morphology, such as Italian: a word like medico, for example, may be a noun (‘doctor’), an adjective (‘medical’) or a verb (‘I medicate’). Let us return to the English example in (30). What determines the spontaneous intuition that the word mass is a noun, rather than a verb or an adjective (which are still possible alternatives)? We believe that the answer lies in the semantic notion of categorial prototype (also see note 24 to Chapter 1). Speakers tend to assign a particular word to the category in which that word is used most frequently in their everyday spoken language. This association is completely natural, and has nothing to do with the scientific idea of categorization, but only with the organization of experience. The limitations of this type of association become evident in the case of lexical entries which are not prototypically assigned to any specific category, such as back. Which class does this word belong to? An immediate answer to this question is difficult to provide, which goes to prove that an English language speaker does not normally have a categorial prototype for this word. It is therefore interesting to note that the semantic root back can be merged as the head of any type of phrase structure: (34) a. b. c. d. e.

I have a pain in my back. The bank decided to back his initiative. He stood up and ran back through the house. Back issues cost 3 dollars each. I went to Paris and came back.

(noun) (verb) (preposition) (adjective) (adverb)

We therefore assume a strictly syntactic criterion for classification: Distributional analysis will then allow us to determine the various lexical classes and define the syntactic relationship between them, and will make it possible to view the sentence as a system of syntactic dependencies between different categories.26 This conclusion leads us to a brief discussion of the idea of grammaticality, which should be intended in purely structural terms, thus remaining independent of any semantic considerations.

Parts-of-speech classification

25

1.3.2. The concept of grammaticality Under the formal approach assumed here, the grammar of a language generates word classes and sentence structure regardless of word meanings. From this programmatic perspective, Chomsky (1986a: 56) defines language as: “A rule system of some sort, a specific realization of the options permitted by UG... The rule system assigns to each expression a structure, which we may take to be a set of representations, one on each linguistic level... The rule system expresses the relations among the various levels in the language in question and determines the elements and properties of each level.”

According to Universal Grammar principles and the parameters set on the basis of experience, native speakers can recognize sentences in their own language and express grammaticality judgments, even when the sentences in question are meaningless. A sentence such as (35a) must therefore be considered to be grammatical in English, as it follows the rules of the system. For example, the subcategorization of the verbal head observe, which requires the presence of two arguments (see § 1.2.3.1.), and the head-complement parameter (see § 1.2.2.), whereby sentence structure proceeds from left to right: (35) a. Green boys furiously observe freedom on carrot trees. On the other hand, a sentence like (35b) cannot be considered grammatical in English because it lacks the formal properties of a grammatical sentence. The subject is placed between verb and complement, the article is in post-nominal position, prepositions are not adjacent to their complements and the adjective little is not adjacent to its relevant head noun. Sentence (35b) is therefore ruled out by the English grammar, although it is still possible to infer its meaning:27 (35) b. *Plays boy the with football friends his little. We can conclude that syntax functions independent of semantics: Sentence structure is constructed regardless of the meaning of the words that make up a sentence. The linear order and the relations between constituents expressed in sentence (35b), however, are not totally impossible. Such options are

26

Categories and functions

admitted by Universal Grammar and, as such, appear in other language systems. The verb-initial order, for example, is the unmarked option in many languages of the world (see the examples in [36a-c]),28 and postnominal articles are found in languages such as Albanian and Bulgarian (cf. [37a-b]): (36) a. Qara’a al-talibu al-kitaba. (ARABIC) read.PERF.3SGM DET-student.NOM DET-book.ACC ‘The student read the book.’ b. Gwelet en deus Yann Ann er gegin. (BRETON) see.PRT have.PRES.3SG Yann Ann in the kitchen ‘Yann saw Ann in the kitchen.’ c. Bumabasa ang titser ng diyaryo. (TAGALOG) 29 ACT.read.PST TRIG teacher DIR newspaper ‘The teacher read the newspaper.’ (37) a. Klara mbaroi shkoll-ën. Klara finish.PST.3SG school-DET ‘Klara finished school.’ b. Namerich kniga-ta na Marija find.PST.1SG book-DET of Mary pod masa-ta. under table-DET ‘I found Mary’s book under the table.’

(ALBANIAN)

(BULGARIAN)

In a similar way, some languages admit the possibility of realizing prepositions in clitic clusters, thus separating them from their respective noun complements. Let us take the Somali example (38) below (from Puglielli 1981): The preposition ka (‘from’) forms a clitic cluster with the preposition u (‘for’) and the first person plural object clitic (‘us’). It is thus separated from its noun complement (gurigii), which precedes it in the linear order: (38) Axmed baa guri-gii nooga (= na+u+ka) qaaday. (SOMALI) Ahmed FM home-DET.AN OCL.1PL.for.from take.PST.RED ‘Axmed took it from home for us.’ Finally, discontinuous constituents appear in a number of languages. A head and its modifier may be separated while still being interpreted as structurally dependent. Examples (39a-b) show this:

Parts-of-speech classification

27

(39) a. Firmo in Treveris ad pontem in Treveri.PL.ABL at bridge.SG.ACC solid.M.ABL praesidio relicto […] reliquas rest.PL.ACC garrison.M.ABL leave.PRT.PST.ABL.SGM copias equitatum-que traducit. (LATIN) troup.PL.ACC cavalry.ACC-and lead.PRES.3SG ‘Having left a solid garrison at the bridge among the Treveri, he leads over the rest of the troops and the cavalry.’ b. Na-marngorl gagarme na-gimuk. (MAYALI) CLASS-barramundi 3SG.catch.PST CLASS-large ‘He was catching a large barramundi.’ In particular, in the Latin example in (39a) (from Caesar’s De Bello Gallico, VI.9), the dependency is ensured by the morphology expressing the syntactic function (ablative Case), which links firmo and praesidio unambiguously. In the Mayali example in (39b) (Evans 1991), on the other hand, the relationship between the noun head and its adjectival modifier is established by the particle na (which may be defined as a classifier).30 The ungrammaticality of a sentence may alternatively be determined by the violation of a Universal Grammar principle. In this case, the sentence can be predicted as ungrammatical in any language. For example, according to the Thematic Criterion (see § 2.2.) “each argument is assigned one and only one ș-role and each ș-role is assigned to one and only one argument” (Chomsky 1986a: 106); as a consequence, sentences such as those in (40a-b) below are ungrammatical in any language. The former is ungrammatical because a constituent (Mark) is realized without an argument role, the latter because the sub-categorization of the verb complete is not saturated. Similarly, Case Theory assumes that each phonetically realized NP is assigned a Case (i.e. a syntactic function in the sentence; see Chapter 3). Sentences like the following will therefore be ruled out in any language: in (40c) Jim cannot receive NOM Case from the uninflected verb and in (40d) Jim cannot receive accusative (ACC) Case directly from the noun: (40) a. b. c. d.

*Jim has invited a friend Mark. *Jim has completed. *Andy believes Jim play very well. *I appreciated the description Jim.

28

Categories and functions

In each of the ungrammatical cases presented, it is always possible to grasp the gist of the sentence; the relevant sentences, however, are unacceptable. In sum, grammaticality merely relies on the properties of the system: In some cases it is related to specific options provided by a language (parameter violations), in others in absolute terms (Universal Grammar principle violations). Having clarified the concept of grammaticality, which is crucial in our formal approach, we can go back to analyzing the behavior of the categories from a morpho-syntactic perspective. We will be able to examine the functions a category can assume depending on its position within the phrase structure.31 Although some functions are typically associated with specific categories, a rigid association is by no means necessary.

1.4. Categories: Typology and formal properties According to the tripartite distinction shown in the tree diagrams in (19), (20) and (29), two basic types of heads can be identified with the position X°: Heads for insertion of lexical categories and heads available for insertion of functional categories. The latter may in turn be divided into sentential (IP) and discourse (CP) functional categories.32 On this basis, the traditional morphological typology of isolating, agglutinative, inflecting and polysynthetic languages (see Sapir 1921, Bloomfield 1933, Comrie 1981, Croft 1991) may be redefined based on the way lexical and functional categories combine (or fail to combine) within the sentential structure. In short: a) In isolating languages, both lexical and functional categories are realized as free (unbound) elements in the phrases they head. Lexical and grammatical information thus creates independent words. Let us consider examples (41a-b), (42) and (43), respectively from Li & Thompson (1981), Bickerton (1993) and Everett & Kern (1997): (41) a. Wǂ

cóng Zhǀngguó from China ‘I come from China.’ b. Wǂ-de yƱzi. PRO.1SG-ASSOC chair ‘My chair.’ PRO.1SG

lái come

le. (CHINESE) PERF

Categories: Typology and formal properties (42) Shi a aalweez de a sing. S.PRO.3SGF HAB always DUR PROG sing ‘She usually keeps singing all the time.’

29

(GUYANA CREOLE)

(WARI)

(43) Tototoc nana com com. drink 3PL.PERF water water ‘They have drunk a lot of water.’

Clearly, no morphological inflection or incorporation phenomena take place among categories in these languages. As the examples show, information on verbal categories, such as aspect (perfective in [41], progressive and habitual in [42]), or nominal categories, such as number (in [43]), is not realized as an affix on the relevant lexical category, but consists of independent elements, either functional (as in [41]-[42]) or lexical (as in [43], where quantification is expressed by means of noun reduplication). Furthermore, categories assume distinct functions according to the position they occupy. The word wǂ is a (first person singular) pronoun in pre-verbal position (41a), but is a (first person singular) possessive when followed by the associative particle de in prenominal position (41b). The relationship between categories and functions will be analysed in § 1.5. Morpho-syntactic independence extends to discourse-functional categories. Focus markers such as we in Saramaccan Creole (Veenstra & den Besten 1994) or Topic markers such as ya in Gungbe (Aboh 2004) will always appear as separate heads (unlike in agglutinative or polysynthetic languages, as shown below): (44) DI

BUKU we book FM ‘I read A BOOK.’

DET

mi

bi

PRO.1SG PST

lesi. read

(45) VƱ lò yà è yì wéxò mȑ. child DET TM 3SG.NOM go.PERF school in ‘(As regards) the child, (he) went to school.’

(SARAMACCAN)

(GUNGBE)

b) In agglutinative languages, a lexical category combines with one or more functional categories, depending on the amount of grammatical information needed for interpretation. Each head bears one and only one piece of grammatical information and the boundaries between categories are clearly preserved. Grammatical categories are thus bound elements in

30

Categories and functions

languages such as Turkish, Swahili and Korean (examples from Kornfilt 1997, Vitale 1981 and Ho-Min 1999 respectively): (46) Denis kitap-lar-im-i Denis book-PL-POSS.1.SG-ACC ‘Denis saw my books.’ (47) Clay Clay

marafiki

(TURKISH)

gör-d-ü. see-PST-3SG

a-na-wa-pig-i-a SCL.M-PRES-OCL.3PL-hit-DAT-IND

simu telephone

zake.

(SWAHILI)

friends POSS ‘Clay is calling up his friends.’ (48) Minca-nun hayngpohka-n tusi happy-INTENS as Minca-TM palapo-ass-ta. look-PST-DECL ‘Minca looked at me with great happiness.’

na-lul OCL.1SG-ACC (KOREAN)

It can be noted that each word comprises a string of elements: only one of them is a lexical head, all the others being heads of functional projections. The Korean example also shows that discourse categories behave just like any other functional element. The Topic marker nun, for example, is joined to the lexical category which plays the Topic role in the sentence (Minca). c) In inflecting languages, a lexical category is combined with one or more functional heads. Again, grammatical categories appear as bound elements. Unlike in agglutinative languages, however, here one functional morpheme may realize several pieces of grammatical information. Consequently, in these languages words are less complex than in agglutinative languages, but their internal structure is less clear. Let us consider category realization in languages such as Arabic, Russian and German: (49) ‘Asbaha akhnjhu brother.O.PRO.3SGM.NOM become.PERF.3SGM mumaththil-an. actor.ACC-INDEF ‘His brother has become an actor.’

(ARABIC)

Categories: Typology and formal properties (50) Ona brosila S.PRO.3SGF.NOM throw.PST.SGF v ozero. in lake.SG.ACC ‘She threw the stones in the lake.’

31

kamni stone.PL.ACC (RUSSIAN)

(51) In meiner Familie haben beide in my.SGF.DAT family.SGF.DAT have.PRES.3PL both Eltern immer gearbeitet. (GERMAN) parents.PL.NOM always work.PRT ‘In my family both my parents have always worked.’ The suffix -a of the Arabic verb ‘asbaha and the suffix -la of the Russian verb brosila, on their own, provide information on tense, aspect, number and gender. It is impossible to separate and distinguish each piece of information within an individual segment. In these languages, the realization of the nominal category of Case is also inseparable from that of gender and number. From a morpho-syntactic derivation perspective, it may be assumed that these functionally complex morphemes result from a process of functional head incorporation, which took place in a prior stage of the language and whose present outcome can no longer be analyzed analogically. d) In polysynthetic languages, the combination between categories is the most productive. The specific characteristic of these languages is that several lexical and functional categories can combine to form a single, extremely complex word. The morpho-syntactic realization of a word thus implies a complex process of incorporation between lexical and functional heads along the syntactic tree.33 For this reason, such languages are also called incorporating. Let us consider the following examples, the first taken from Baker (1996) and the other two from Evans & Sasse (eds.) (2002): (52) Angya-ghll-ng-yug-tuq. boat-AUG-buy-DESID-3SG ‘He wants to buy a large boat.’ (53) Wa-hi-’sleht-a-hni:nu-:.34 FACT-SCL.1SG.OCL.3M-car-buy-PERF ‘I bought a car from him/I bought his car.’

(GREENLANDIC)

(ONEIDA)

32

Categories and functions

(54) E-s-kakha-hona’tá-yethw-ahs. FUT-ITER-SCL.1SG.OCL.3PL-potato-plant-BEN.PERF ‘I will plant potatoes for them again.’

(CAYUGA)

A word such as Greenlandic angyaghllngyugtuq in (52) comprises two lexical categories, corresponding to the verbal head ‘buy’ and its argument ‘boat’. This shows that what is realized by means of independent phrases in a sentence in isolating, agglutinative or inflecting languages is expressed by a single word in polysynthetic languages.35

1.5. Categories and syntactic functions In our morpho-syntactic and distributional approach, the syntactic function of a category depends on the structural collocation of the phrase that hosts it. In particular, first-level and second-level functions can be identified. First-level functions involve the concepts of SUBJECT (S), PREDICATE (Pr) and OBJECT (O). These functions are typically used in typological studies to refer to constituent order. In typology, however, the predicate function has traditionally been defined as verb (as in SVO order or SOV order). This has generated confusion between categories (the verb) and syntactic functions (the predicate),36 which is especially undesirable as nonverbal categories (nouns and adjectives; see below) may, too, function syntactically as predicates. The relationship between categories and functions is the point that we will try to clarify in this section. Second-level functions include MODIFIERS, i.e. phrases which are embedded in other phrases and specify some of the properties of the hosting phrases, providing extra information about their heads. An adverb, for instance, is typically the modifier of a VP, whereas an adjective functions as a (restrictive or appositive) modifier when embedded in an NP. Resuming our previous analysis, we can summarize and conclude the discussion as follows. Semantic roots are inserted from the Lexicon into the computational component of syntax, where they assume a categorial status depending on their position in the phrasal structure. The position of the phrase within the sentential structure thus determines the distinction among predicates, arguments and non-argument (or circumstantial) constituents. These deep positions may then undergo derivations giving rise to the (surface) realization of specific syntactic functions. According to our theory, deep structure is universal; consequently, the argument selection related to participants in an event must be the same, or

Categories and syntactic functions

33

at least overlap almost completely, in the world’s languages. This means that differences between languages occur at surface structure level, i.e. where syntactic functions are realized. In other words, our hypothesis is that the argument structure of a verb such as ‘to like’ is the same in English and, for example, Italian (see [55]). The difference between (56) and (57) simply reflects the different arguments realized as syntactic subjects in the two languages: the experiencer in English and the patient in Italian: (55) like: (56) I like music. (57) Mi

piace la IOCL.1SG please.PRES.3SG DET ‘I like music.’

musica. music

(ITALIAN)

It is important to define syntactic functions in more detail so that they can be characterized in a clearer and univocal way, avoiding confusion and overlap with lexical categories or, more importantly, with deep roles. Chapter 3 will be devoted to this purpose. We believe, however, that a preliminary discussion of the issue is in order so that some distinctions needed for the analysis become immediately evident.

1.5.1. Subject function The definition of subject is not as obvious as it may appear. The subject has been traditionally defined as “the person who performs the action”: an acceptable definition for the examples in (58), but a completely inappropriate one for those in (59): (58) a. Jim eats an apple. b. Mark puts the book on the table. (59) a. The movie starts at five. b. The pen is on the table. c. John was hit.

34

Categories and functions

In (59a), the movie does not perform any action, nor does the pen that someone has put on the table in (59b), or John in (59c), who has clearly had no active role in the action expressed by the predicate. We might be tempted to define the subject as “the person/thing being talked about”. This definition would account for both the subject in (58) and those in (59). But it would not work in cases such as the following: (60) a. Mark, you always have to tell him things twice. b. Care for a coffee? In (60a) Mark is being talked about although this is not the sentential subject; in (60b) the speaker is not saying anything about their interlocutor, although the (unpronounced) subject is a second person singular (you). Hence, the adoption of semantic and/or pragmatic criteria does not appear to be conducive to an unambiguous definition of the subject function. On the other hand, by adopting formal criteria and examining the morpho-syntactic characteristics of the previous sentences, it can be observed that in a language with no (or little) inflection such as English or Chinese the subject is always defined unambiguously by its position with respect to the verb. In these languages the subject is the nominal constituent that precedes the verb, and its position is blocked by grammar for that specific function: (61) a. You dance very well. b. *Dance you very well. (62) a. Zài TáibƟi tƗ chƯ le in Taipei PRO.3SG eat PERF ‘S/he ate very well in Taipei.’ b. *TƗ zài TáibƟi chƯ le hƟn hăo.

hƟn very

hăo. good

(CHINESE)

Conversely, in (morphology-rich) inflecting, agglutinative and incorporating languages, the subject is marked by the specific property of being the constituent which determines agreement with the verb. In particular the position of the subject within the clause can vary in inflecting and agglutinative languages, as verbal inflection allows unambiguous identification. This is shown in the following Russian examples:

Categories and syntactic functions (63) a. Anja videla Anja.NOM see.PST.3SGF ‘Anna saw Ivan.’ b. Videla Anja Ivana. c. Videla Ivana Anja.

Ivana. Ivan.ACC

35

(RUSSIAN)

Although predicates follow subjects in unmarked sentences in Russian, reversing that order does not result in ungrammaticality. This, of course, does not mean that word order is free in inflecting languages, as each variation in the order of constituents corresponds to a different and specific pragmatic meaning (see Chapter 6). In these cases, the subject is the phrase which, departing from deep structure, goes to form a structural configuration of agreement with the verbal head, i.e. occupies the Spec,IP position (see [20]). The syntactic role of subject is thus totally unrelated to its (deep) argument role, which may vary broadly in languages such as Italian. Let us consider the following examples: (64) a. Leo ha aperto la porta. (ITALIAN) Leo have.PRES.3SG open.PRT DET door ‘Leo has opened the door.’ b. La porta è stata aperta DET door be.PRES.3SG be.PRT.F.SG open.PRT.F.SG (da Leo). by Leo ‘The door has been opened (by Leo).’ c. Il vetro si rompe. DET glass MIDD break.PRES.3SG ‘The glass breaks.’ d. La porta si apre. DET door MIDD open.PRES.3SG ‘The door opens.’ e. Leo ha ricevuto molti libri. Leo have.PRES.3SG receive.PRT many book.PL ‘Leo has received a lot of books.’ f. Leo possiede molti libri. many book.PL Leo own.PRES.3SG ‘Leo owns a lot of books.’

36

Categories and functions g. Leo ama leggere. Leo love.PRES.3SG read.INF ‘Leo loves reading.’

Only in sentence (a) is the (syntactic) subject also an agent (the participant performing the action). In (b) door is clearly a patient (the participant undergoing the action with a dynamic predicate), whereas glass and door in (c) and (d) are themes (the first participant of a stative or motion predicate). In (e) Leo is a goal, or the participant towards which the action is directed; in (f) it is a locative as it embodies the figurative location of the books; in (g) it is an experiencer, i.e. a participant that has a certain feeling or is in a certain psychological state. In sum, the concepts of syntactic subject and agent refer to distinct levels of analysis and should not be confused.37 It is important to stress, however, that not just any argument role may be promoted to subject in all languages. For example in Warlpiri, an inanimate object which is the instrument of an action may not appear as a subject. Thus, a sentence such as the boomerang hit me will have to be expressed as follows (from Simpson 1991): (65) Karli-kirli-rli-ji luwa-rnu. (WARLPIRI) boomerang-ASSOC-ERG-1SG.ACC shoot-PST ‘The one with the [who possesses the] boomerang shot me.’ Contrary to appearances, karli (‘boomerang’) is not the subject of this sentence. This is confirmed by the presence of the ergative (ERG) Case morpheme (-rli), which can only refer to subjects with an agent ș-role (see below and § 3.3. on ergative languages). On a closer analysis, it can be noted that the lexical root karli (‘boomerang’) is followed by the associative element kirli. This element creates a relationship between the instrument and the agent, which is expressed by the ERG Case morpheme rli. Kirli, meaning ‘x possesses y’, is a way of saying that a certain ‘he’ (unpronounced but present as the functional head of ERG Case) ‘hit me because he had a boomerang’. Malagasy applies a similar logic: only the agent and patient argument roles are admitted as subjects in that language. A special syntactic structure is needed to express a beneficiary subject, as shown in the following sentence:

Categories and syntactic functions

37

(66) Nividianan’ i Bao ny akoho i Soa. (MALAGASY) PST.buy.BEN DET Bao INDEF chicken DET Soa ‘Bao bought a chicken for Soa.’ In Malagasy (as in other Austronesian languages), the element with the syntactic function of subject takes final position and determines the presence of a special morpheme on the verb, corresponding to its argument role. The BEN agreement morpheme in (66) thus clearly shows that i Soa (the subject) is not the agent of the action expressed by the verb, but its beneficiary. This sentence could be translated with a passive structure, such as Soa was bought a chicken by Bao. However, (66) cannot be considered a passive sentence as the verbal head displays active morphology and the agent (Bao) lacks any prepositions indicating its status of adjunct complement (similar to English by) in the event of buying. Evidently, this structure cannot be equated with a passive, and the promotion of a beneficiary to subject appears to require a special syntactic structure. We will not dwell upon this issue here;38 see Givón (1984), Payne (1997), Potsdam (2006) for a thorough discussion. As we have seen, there is no necessary correlation between the function of subject and the agent argument role. Typologically though, it is interesting to note that the agent ș-role may be associated with the subject function in all languages, whereas a host of morpho-syntactic restrictions or effects apply to other argument roles. Finally, in some languages semantic and syntactic criteria join up to mark subjects. This is the case in ergative languages (including Basque, Eskimo and Dyirbal), illustrated below (Avar examples quoted from Comrie 1981):39 (67) a. Vas-ass jas j-ecc-ula. boy-ERG girl.ABS SGF.ABS-praise-PRES ‘The boy praises the girl.’ b. Vas v-eker-ula. boy.ABS SGM.ABS-run-PRES ‘The boy runs.’ c. Jas j-eker-ula. girl.ABS SGF.ABS-run-PRES ‘The girl runs.’

(AVAR)

As we can see, the subject of a transitive verb in (67a) receives a special Case, the ergative, whereas the subject of an intransitive motion verb (‘run’

38

Categories and functions

in [67b] and [67c]), bears a different Case mark (called absolutive), which is also used for the objects of transitive verbs (for example, jas in [67a]). Such a Case system clearly depends on the semantic role associated with the subject. In particular, ERG Case marks agent-subjects alone, whereas a subject with a patient ș-role always bears absolutive Case.40

1.5.2. Predicate function As in the case of the subject, a definition of predicate requires a premise and some distinctions. The idea of predicate is traditionally associated with the lexical class of verbs, and rightly so, since the category of verb is by definition related to the expression of a predication (see note 36 to Chapter 1). However, it is not the only one. The world’s languages show that other elements from the Lexicon may also be inserted into structures where they can function as predicates: For example, nouns and adjectives. Let us consider some extremely common sentences, such as the following: (68) Jim is a doctor. (69) Jim is my manager. (70) Jim is nice. We can see that an NP and an adjectival phrase (AP) may both predicate a state, an identity, a property etc. about a subject. A nominal predicate can also be used to specify the reference of the subject noun, as in the following case:41 (71) The person I was talking about is Jim. Sentences of this kind are also called copular in the literature, as the nominal (or adjectival) predicate is preceded by the auxiliary be in numerous languages. The copula, however, is not strictly needed in a nonverbal predicate, as shown in the following examples (sentence [73] is quoted from Bickerton 1993):

Categories and syntactic functions

(ARABIC) (RUSSIAN)

(72) a. Huwa talib. b. On student. S.PRO.3SGM student ‘He is a student.’ (73) Di

kaafi kuol. coffee cold ‘The coffee is cold.’

39

(GUYANA CREOLE)

DET

It is clear that the realization of a noun in a predicate function does not entail verbal inflection. The noun retains the morphological characteristics typical of the phrasal category it belongs to. It can thus be preceded by an article (as in [68]), a possessive (as in [69]) or any other modifiers (adjectives, relative clauses) available for a noun head. The behavior of predicative adjectives, for example, varies according to the language. In Italian and French, the predicative adjective retains the formal properties of an attributive adjective, assuming a masculine, feminine or plural form according to the features of the subject: (74) Leo è contento / Leo COP.PRES.3SGM happy.SGM Maria è contenta / Maria COP.PRES.3SGF happy.SGF Loro sono contenti. S.PRO.3PL COP.PRES.3PLM happy.PLM ‘Leo is happy / Maria is happy / They are happy.’

(ITALIAN)

(75) Leo est heureux / Maria est heureuse / Ils sont heureux. (FRENCH) (=[74]) However, this is not true in all language systems. Russian and German, for example, present a dedicated form of the predicative adjective for this function (a short form in Russian,42 an invariable form in German): (76) a. Kostjum velik /*velikij. dress large.M (SHORT/*LONG FORM) ‘The dress is large.’

(RUSSIAN)

40

Categories and functions b. Bþera devuška byla yesterday girl be.PST.SGF sþastliva /*sþastlivaja. happy.SGF (SHORT/*LONG FORM) ‘Yesterday the girl was very happy.’

(77) Die

Bluse / Der Rock / shirt.F DET.M skirt.M ist neu. be.PRES.3SG new (INVAR) ‘The shirt / skirt / dress is new.’

DET.F

Das DET.N

oþen very

Kleid dress.N (GERMAN)

In other languages, the predicative adjective assumes morpho-syntactic properties that cause it to behave like a verb. One example is the Frenchbased Haiti creole (see Lefebvre 1986). In this language, predicative adjectives may combine with past tense markers (78a), future tense markers (78b) and the irrealis modality marker (78c), all of which are typically found (as independent particles) on the verb: (78) a. Kay li te bèl. (HAITI CREOLE) his house PST beautiful ‘His house was beautiful.’ b. Kay li ap bèl. his house FUT beautiful ‘His house will be beautiful.’ c. Kay li ta bèl si li te pentire. paint his house IRR beautiful if house PST ‘His house would be beautiful if it were painted.’ Another option for the predicative adjective is to present an inflection very similar to that of verbs: this is observed in Schachter’s (1985) data and analysis (contained in the glosses) for Bemba (a Bantu language) and Mojave (an Amerindian language). In other cases, predicative adjectives display a reduced verb inflection, as shown by Hopper & Thompson (1984) in Chukchi, a Paleo-Siberian language whose adjectives provide aspectual rather than temporal information:

Categories and syntactic functions (79) a. Umuuntu á-ashipa /á-akosa person PRES-courageous /PRES-strong /á-aceenjela. /PRES-wise ‘The person is courageous/strong/wise.’ b. Umuuntu á-alemba. person PRES-write ‘The person is writing.’ (80) a.

?

I:pa-þ homi:-k man-CONJ tall-PRES ‘The man is tall.’ su:paw-k b. ?I:pa-þ know-PRES man-CONJ ‘The man knows.’

n-erme-Ȗg.m. (81) Ȗ.m 1SG.ABS IMPF-strong-1SG ‘I am/used to be strong.’

(iðu:m). (AUX)

41

(BEMBA)

(MOJAVE)

(iðu:m). (AUX)

(CHUKCHI)

In conclusion, the predicate function is not restricted to verbs but may be fulfilled by other categories as well. These categories can either retain their formal properties or assume, to various degrees, those of the verbal category (which is prototypically associated with the predicate role).

1.5.3. Modifier function Modifiers are second-level functions: they are inserted from the Lexicon as complements or adjuncts to other (first-level) phrases, and as such contribute information about the hosting phrases. With the subject and predicate as primary functions, modifiers must be divided into nominal (typically determiners, adjectives and relative clauses) and verbal (adverbs and adverbial expressions). In particular, nominal modifiers can be NP complements or adjuncts, whereas verbal modifiers are invariably VP adjuncts. Specific chapters of the book (Chapters 2, 4 and 5) will be devoted to the analysis of these elements.

2. The structure of the verb phrase

2.1. Predicates, arguments and structural categorization The aim of this chapter is to provide a detailed description of the portion of sentence structure called VP within an X' representation. We will examine the lexical elements that can occur as predicates, and the number and type of thematic relations they define. As will become clear, this level of syntactic representation is related to semantics and, at the same time, determines a number of syntactic functions which interface with surface morpho-syntactic phenomena, e.g. constituent order and morphological agreement mechanisms (see § 1.5.). In this line of analysis, the term predication will be used to refer, at a semantic level of description, to the pivot element in the sentence, whose meaning determines a rich and diverse set of thematic relations with arguments. The VP, as previously noted, is the locus of the syntax-semantics interface, insofar as the core of a sentence (i.e. a predication and its arguments) is formed by lexical selection and insertion within the VP. The basic structure of a sentence consists of one predication and one or more arguments (or referential elements; see Lyons 1977), the number of which is determined by the meaning of the predication. The question now centers on which lexical entities can function as predicates in a sentence. Let us consider some examples: (1)

Mark leaves.

(2)

Joan burns newspapers.

(3)

Mike is an architect.

(4)

He is intelligent.

(5)

The boys are in the attic.

(6)

Paula is here.

44

The structure of the verb phrase

The predicate is a verb in (1) and (2); a noun in (3); an adjective in (4); a PP in (5) and an adverb in (6). Furthermore, in (3) to (6), an element is added which is traditionally called copula and is required for inserting tense, aspect and mood (TAM) features. In a language like English, TAMrelated information is conveyed by verbal morphology. Consequently, the copula (merged in I°) is needed to realize these features; what follows the copula is the actual predication. A sentence whose predicate is a non-verbal form is usually called a Small Clause.1 A cross-linguistic analysis, however, immediately shows that the copula is not always strictly necessary (also see Chapter 1, § 1.5.2.). In a number of typologically distinct languages, any phrase (be it nominal, adjectival or adverbial) may be merged as a predicate into the syntactic structure to obtain a(n unmarked) present-tense interpretation:2 (7)

Ivan chorošij. Ivan good ‘Ivan is good.’

(8)

Cali waa3 macallin. Cali DECL teacher ‘Cali is a teacher.’

(SOMALI)

(9)

Ali mu’allim. Ali teacher ‘Ali is a teacher.’

(ARABIC)

(10) TƗ PRO.3SG

pàng. fat

(RUSSIAN)

(CHINESE)

‘S/he is fat.’ (11) Maganda Linda.4 pretty Linda ‘Linda is pretty.’

(TAGALOG)

This confirms what was assumed in Chapter 1 (cf. § 1.3., especially note 25): At a semantic level, words can be considered as sets of features which assume categorization and a syntactic function upon lexical insertion, based on the syntagmatic context in which they are merged. As such, words receive both derivational and inflectional morphological elements5 relating to their specific category.

Predicates, arguments and structural categorization

45

Let us consider a few more examples: (12) a. The archives are complete. b. I am at the archive. c. An archive and a library are two different things. (13) a. Jim archives each document carefully. b. I was hoping you would archive those files. c. Archive the documents as I have taught you to. (14) The archiving of the data is tiring. (15) We need an archive room. (16) That is an archive. If the root archive is merged in a node that projects an NP, it can cooccur with an article (as in 12a), follow a preposition (12b) or be coordinated to another noun (12c); furthermore, it can be used in the plural (archives) and be modified (complete archive). If the same root is merged as the head of a VP, it can co-occur with different verbal tenses, persons and moods (13a-c). In (14), the root archive is accompanied by the derivational morpheme -ing, which makes it possible to merge it in N° with the meaning ‘action of archiving’. In (15), the entity archive occurs as the modifier of a noun head (room). Finally, in (16), archive is merged as a predicate in a identificational context, and as such requires a copula (within a Small Clause structure). It can be concluded that categorization and any consequent adjustments undergone by a lexical entity can vary within an individual language and cross-linguistically, depending on the morpho-syntactic characteristics of the language (inflecting vs. isolating, etc.).

2.2. Argument structure No matter what lexical element takes on the syntactic function of predicate, the number and type of arguments that become part of its deep semantic structure are determined by the lexical and semantic properties of the predicate. In other words, for a sentence to be well-formed, the valencies6 activated by the predication all need to be saturated, i.e. filled with

46

The structure of the verb phrase

elements hierarchically arranged with respect to the predicate. Each one of these elements has, in turn, a specific semantic role which will be called thematic role (or ș-role), following Chomsky’s terminology. At a semantic level, every predication assigns a role (by definition semantic) to its arguments. Semantic roles have been variously analyzed: They have been analyzed according to features,7 or alternatively interpreted as macro-roles,8 each macro-role being divided into more specific argument roles closely connected with the semantic composition of the verb (see below). One example of macro-role is the ACTOR. Syntactically, the ACTOR is the first argument of many multiple-argument predications; semantically, it includes the θ-role of 9 (characterized by the feature [+animate]), but also that of ([-animate]). An ACTOR can be the of an action, the of an action performed by another participant in the event or the in a sensory or psychological event (see Van Valin & LaPolla 1997: 141). Macro-roles therefore qualify as sets of argument roles. This concept will be especially important for our analysis as syntactic hierarchy does not interface specifically with individual ș-roles. Macroroles, on the other hand, do interact with syntax, building relationships with what are called deep Cases (Fillmore, 1968, 1970) in the VP and with surface grammatical functions (see § 3.2.). Our analysis will therefore focus on the latter as our interest is mainly centered on the interaction between the various levels of grammar. As thematic roles are connected to the semantic properties of predication, they are arguably universal. This amounts to saying that a certain set of semantic features corresponds to a specific argument structure shared by all languages.10 The natural discrepancies between typologically different languages are related to surface projection, where the syntaxsemantics interface is decisively influenced by the morpho-syntactic type of the language in question (head-initial vs. head-final, inflecting vs. isolating, agglutinative or polysynthetic; see § 1.4.). Let us now examine the predication-argument structure in more detail and see which and how many ș-roles are made necessary by the different meaning of predicates. As this is an extremely vast topic, our analysis will focus mainly on predicates lexicalized from roots that can directly combine with verbal morphology, i.e. verbs. As the predicative core of a sentence, the verb activates a structure which can include one or several arguments (up to three; see below) with a specific ș-role. These roles are assigned by the predicate, and each argument must receive one and only one role.11 Arguments are thus

Argument structure

47

selected, i.e. sub-categorized by the verb; as a consequence, they are required for a sentence to be grammatical. Let us compare (17) and (18): (17) a. *They have seen. b. *They collected. (18) a. Paul and Sally have seen that movie. b. The boys collected the garbage. The verbs in (17), without all the selected arguments and an appropriate communication context, produce incomplete and therefore ungrammatical sentences; on the other hand, the same verbs in (18) are accompanied by the arguments required by their meanings, and consequently form grammatical sentences. A sentence may of course include other elements not required by the verb. These are called adjuncts (or circumstantial extra-nuclear elements; see § 1.2.1.) and appear in sentences such as (19): (19) a. The boys collected the garbage this morning. b. The boys collected the garbage at school this morning. This morning in (19a-b) and at school in (19b) are adjoined constituents. These phrases contribute extra information about the event described by the predicate, hence there can be a potentially unlimited number of adjuncts in a sentence, and their presence is not indispensable. Only the two argument NPs (the boys and the garbage) are necessary for comprehension. In particular, the adjuncts this morning and at school in (19a-b) locate the event in space and time. However, it is worth noting that this type of element may, in other contexts, play the role of argument: (20) I left my diary at school. (21) The class lasted three hours. (22) A similar earthquake took place 150 years ago. In (20) leave, a three-argument predicate, requires a third argument; in (21) and (22) the second argument expresses a and a respectively, as required by verb semantics. It is

48

The structure of the verb phrase

therefore important to notice that no specific meanings are necessarily associated with a circumstantial role. What has so far been shown for English is also confirmed in other languages belonging to different morphological types. Let us take polysynthetic languages as an example. In these languages, as we have seen, arguments are incorporated into the verbal head, either as lexical heads or by means of clitic pronouns (i.e. unstressed pronouns that need to attach to a head in order to form a phonological word; see Nespor 1993). It is therefore expected that adjoined constituents outside the VP area may not be incorporated or doubled by clitics on the verbal head,12 which is exactly what the data show. Let us consider for instance how adverbs are realized in Gunwinjgu and Rembarunga, in (23) and (24), from Evans (2002) and Nordlinger & Saulwick (2002) respectively: (23) Balay Ø-ngan-waralh-miny already SCL.3SG-OCL.1SG-ask-PST.PERF nga-Ø-bak-rniya-Ø. SCL.1SG-OCL.3SG-BEN-cook-PRES ‘She has already asked me to cook for her.’ (24) Ka-birrih-nalkbu-n NONPST-SCL.3PL-cry-HAB ‘They always cry.’

munguyh. always

guwa purpose (GUNWINJGU)

(REMBARUNGA)

In both sentences, the verbal head incorporates all the arguments, whereas the adverbs (already and always) are located in either periphery. Other evidence for a VP-external syntactic status of adverbial NPs is provided by Bantu languages such as Chichewa (Bresnan & Mchombo 1987). In Chichewa, the verbal head presents a final associative morpheme (-e). If this morpheme is followed by adjoined NPs, it bears prominent stress (clearly indicating a prosodic boundary); if, on the other hand, the verb is followed by an NP linked to an argument role (and, consequently, to a clitic pronoun on the verbal head), the stress falls on the preceding syllable. This shows that pang’óno pang’óno (‘a bit a bit’) in (25) is an adjunct, while phunziro (‘class’) in (26) is not:

Argument structure

49

(25) Ndi-kufúná kutì áná a-pitiriz-è SCL.1SG-want that kid.COLL SCL.3SG-continue-ASSOC pang’óno pang’óno. (CHICHEWA) a bit a bit ‘I want the kids to continue slowly.’ (26) Ndi-kufúná kutì áná a-li-pitirîz-e SCL.1SG-want that kid.COLL SCL.3SG-OCL.3SG-continue-ASSOC phunziro. class ‘I want the kids to continue the class.’ Similarly, in agglutinative languages such as Swahili, arguments are realized as obligatory clitics on the verbal head, whereas adjuncts occupy peripheral positions as they are optionally preceded by prepositions that signal their syntactic role (from Vitale 1981): (27) Juma Juma wa

a-na-wa-pig-i-a

simu marafiki zake phone friends POSS Roma siku zote kwa upendo. (SWAHILI) LOC Rome day each with love ‘Juma calls his friends in Rome every day with love.’ SCL-PRES-OCL.3PL-hit-DAT-IND

The topic of adjuncts and their syntactic status will be developed in Chapter 5. Let us now see how many and what types of arguments can occur with different types of verbs. Firstly, however, a brief description of macro-roles and the θ-roles they select is in order. 2.2.1. Macro-roles and event structure As said earlier, macro-roles are sets of argument roles (or θ-roles). Within these sets, argument roles are merged hierarchically into the VP (the interface between semantics and syntax). Therefore, different θ-roles are associated with the same syntactic position (Spec,VP or Compl,VP) as they belong to the same macro-role.13 Different groupings have been proposed by different authors, based on specific semantic considerations (cf., among others, Foley & VanValin 1984, VanValin & LaPolla 1997, Lehmann 2004 among others). In this work, a subdivision will be proposed which can

50

The structure of the verb phrase

account for some formal distinctions shown by the grammars of different languages (see below). Let us examine these groups in detail. The ACTOR macro-role primarily includes the θ-role, i.e. the argument with such features as [+animate] and [+human], which is the first participant in dynamic and formally transitive actions (such as eat, kill, watch, etc.). Secondly, we have the , i.e. the first (or sole) participant, usually [+human], in psychological events (such as love and tremble), and the , i.e. the ([+animate]) referent that causes an action (as in Jim let the cat go out). This macro-role, however, also includes argument roles with a [-animate] feature, such as the (an inanimate argument similar to the , as in The hurricane caused many casualties) and the , i.e. the object used to perform an action (as in This key opened that door). The ACTOR macro-role is therefore independent of the animateness feature.14 As for the second macro-role, VanValin & LaPolla (1997) include the and θ-roles in a single macro-role (the UNDERGOER) for semantic reasons. Conversely, based on formal considerations related to the syntax-semantics interface (which are crucial to our analysis), we assume that these two θ-roles should be kept apart and assigned to two distinct macro-roles. Thus, the θ-role alone will form the PATIENT macro-role, indicating the second participant in a transitive (or ditransitive) structure, which is modified by or undergoes the consequences of a dynamic action. As this macro-role is indifferent to the animateness feature, it may be [+animate] and [+human] (as in John visited Anne), but also [-animate] (as in John visited a lot of museums). On the other hand, the θ-role alone forms the THEME macrorole. The main difference between the two is the following: a PATIENT invariably indicates the second participant in a dynamic event, the first participant being an ACTOR; THEME, on the other hand, indicates the first participant in a stative event (as in Jim lives in Rome), or in a resultative event whose first participant undergoes a change of state (the ship is sinking), location (Jim arrived at five) or position (the book fell on the floor). Semantically, THEME occupies a category halfway between ACTOR and PATIENT:15 like ACTOR, it is the first participant in an action (thus excluding the presence of an ACTOR); like PATIENT, it is affected to some degree by that action.16 In stative events, THEME can also express the (i.e. the thing owned, as in This book belongs to me). This macro-role, like the previous ones, is indifferent to animateness features. Since the cooccurrence with an ACTOR is ruled out, in two-argument

Argument structure

51

predicates the correlates with participants that are semantically secondary in the event: LOCATIVES (as shown in some of the above examples), and arguments related to DURATION and QUANTITY (see below). The GOAL macro-role includes θ-roles like , and , i.e. the [+animate] participant or the [-animate] location towards which an action is directed (as in I sent a letter to Mike, I sent a letter to Paris). These argument roles typically indicate the third participant in a three-argument predicate (such as give, tell, entrust, lend, etc.). As a consequence, they depend on dynamic events where an ACTOR is present. The LOCATIVE macro-role includes all the arguments that express the where the first participant in the event (specifically a THEME) is located (in terms of state or motion), as in Jim goes to Berlin, Mike has arrived in Rome, The plane leaves from Fiumicino. This macro-role also includes the as an abstract representation of location: The possessor of x is considered ‘the location of x’ (in other words, the place where a [±animate] object is found), as in Jim has a lot of friends, Jim owns many books. The DURATION macro-role refers to those arguments that express the needed for an event to be completed. This type of constituent should not be confused with accessory time expressions (adjuncts), as they are required to complete the argument structure of predicates with verbs such as happen, last and go on. Finally, the QUANTITY macro-role expresses the or of an event. Typical examples are verbs such as weigh (as in Pat weighs 150 pounds) and behave (John always behaves in a pleasant way). Before we go on to discuss how macro-roles are realized according to verb valencies, a final observation on the event structure of predicates is in order. As we have noted, the difference between dynamic and stative verbs (see Vendler 1967) plays a crucial role in determining which macro-roles may occur with specific types of predicates. In particular, ACTOR and PATIENT are necessarily linked to dynamic events, whereas THEME relates to the predication of a state, either defining its conditions and properties (stative verbs) or describing a change and its result (resultative verbs). The correlation between macro-roles and event structure will prove extremely useful and important for understanding a number of subtle semanticsyntactic differences between apparently similar constructions (e.g. passives and middles, see § 2.3.).

52

The structure of the verb phrase

2.2.2. Mono-argument verbs Verbs have been traditionally described as having a zero- to three-argument valency (see Chomsky 1981: Chapter 2, Williams 1994, Grimshaw 1990, Hale & Keyser 1993). As a consequence, there seem to be predicates with no participants in the event described. This possibility will be addressed (and refuted) in the following discussion. Normally, weather verbs are considered zero-valent on the basis of verb-only sentences such as (28)-(30) in typologically different languages: (28) Pioveva. rain.IMPF.3SG ‘It was raining.’ (29) Es-ett. fall-PST.3SG ‘It rained/It was raining.’ (30) Um-ulan.17 ACT/TH-rain.PST ‘It rained.’

(ITALIAN)

(HUNGARIAN)

(TAGALOG)

The observation of other languages, however, reveals a more complex picture. The same sentence, equivalent to It was raining, can be realized in extremely different ways: (31) Roob baa da’ajay. rain FM drop.PST.RED ‘It was raining.’

(SOMALI)

(32) As-samƗu tamtir. DET-sky.NOM rain.PST.3SG ‘It was raining.’

(ARABIC)

(33) Xia yu le. PERF fall rain ‘It was raining.’

(CHINESE)

Languages such as Somali, Arabic and Chinese clearly show that a verb like rain actually has a single valency and, in particular, selects a ș-role

Argument structure

53

within the THEME macro-role. As a matter of fact, this verb expresses an action whose first (and only, in this case) participant (i.e. ‘rain’ or ‘sky’, as in the Arabic metaphor) is affected by the action, and consequently may not be considered an ACTOR. As we assume a universal deep structure, our deduction is that the mono-argument structure of these verbs is valid in all other languages too. Let us test this assumption. In languages like Italian, the THEME is lexicalized in the verb.18 This is proved by the fact that whenever the THEME does not realize the default referent (water) the argument of the verb piovere (‘rain’) can be filled lexically by an element with that argument role. This is projected as the subject: (34) a. Quel DEM

giorno day

pioveva rain.IMPF.3SG

acqua water

sabbia. sand ‘It was raining water and sand that day.’ b. È piovuta pioggia acida. be.PRES.3SG rain.PRT.SGF rain acid ‘It rained acid rain.’

e and (ITALIAN)

The same holds true for the languages seen in (29) and (30), repeated here as (35)-(36): (35) HomokesĘ es-ett. sand-rain fall-PST.3SG ‘It rained sand.’ (36) Um-ulan ACT/TH-rain.PST ‘It rained sand.’

ang TRIG

(HUNGARIAN)

buhangin. sand

(TAGALOG)

Languages like English and French, on the other hand, behave differently: (37) Il

pleuvait. rain.PST.3SG ‘It was raining.’

EXPL

(FRENCH)

54

The structure of the verb phrase

(38) It was raining. Here, the syntactic subject is not a lexicalization of the shadow argument, but an expletive pronoun, i.e. a constituent with no argument role which is merged in subject position (Spec,IP) as required by the syntax of these languages (on expletives, see § 3.2.2.).19 As the subject position is occupied by the expletive, if a non-prototypical THEME is realized in these languages for the verb rain, it will have to be merged in object position (compatible with the deep Case of THEME): (39) Il

pleuvait rain.PST.3SG ‘It was raining insults.’

EXPL

des INDEF

injures. insult.PL

(FRENCH)

(40) It is raining troubles. We therefore assume that a verb like rain can be considered monovalent in every language. This means that there is only one lexico-semantic representation for the VP and that the differences between languages depend on the surface realization of the syntax-semantics interface. We can therefore conclude that the possibility of a zero-valent verb is not verified in any of the world’s languages. This is also expected in terms of meaning, since a predication must be about something. Consequently, on a semantic level, a predication must have at least one argument with various possible surface realizations. Other monovalent predicates include: (41) a. sleep, tremble, be cold, be hungry, be sleepy, laugh, abort, etc. b. be born, die, exist, ring, simmer down, skid, trip, fall, (be) smart, (be) sincere, etc. The verbs in (41a) select one argument with the ș-role (and is thus included in the ACTOR macro-role), while for those in (41b) the argument falls within the THEME macro-role. All these verbs require just one argument to obtain a well-formed sentence. The presence of an vs. a depends on the event structure of the verb: Dynamic mono-argument verbs select the ; stative verbs select the . One remaining question is whether a monovalent verb may present a single argument with the θ-role, as it would appear in cases such as the following:

Argument structure

55

(42) a. Jane sang. b. Jim ate. These verbs, however, should be considered falsely monovalent. It was assumed earlier that a verb such as the Italian piovere contains a shadowargument lexicalized in the verbal head. Similarly, we assume that there is a shadow PATIENT in the predicates in (42).20 Indeed, the answer to a question like (43A) may include a quantifier which clearly refers to the covert object of the (transitive) action expressed by the verb. Conversely, this possibility is ruled out in the case of a truly monovalent verb like sleep (see [44B]): (43) A: Has John eaten? B: Yes, everything. (44) A: Has John slept? B: *Yes, everything.21 Besides, in many languages it is impossible not to lexicalize the second argument of this type of verb. Consider one example from Somali: (45) Isag-aa S.PRO.3SGM-FM ‘HE has eaten.’

*(wax) thing

cunay. eat.PST.RED

(SOMALI)

As we can see, the two-argument verb strictly needs both participants to be overt, although the second argument is a generic, semantically uninformative noun like thing. There is thus isomorphism between the syntactic and semantic structures, which is not always the case in languages like English.

2.2.3. Two-argument verbs Let us consider two-argument verbs and examine the possible combinations of macro-roles: (46) Jim read a book.

ACTOR-PATIENT

(47) The cat ate the mouse.

ACTOR-PATIENT

56

The structure of the verb phrase

(48) A stone broke the window.

ACTOR-PATIENT

(49) Mary knew the truth.

ACTOR-PATIENT

(50) I have a lot of books.

LOCATIVE-THEME

(51) The bag contained a lot of books.

LOCATIVE-THEME

In (46)-(49), the macro-role of the argument that assumes the subject function is the ACTOR. In particular, we have a [+human] in the first case, a [+animate] in the second, a [-animate] in the third and an in the fourth. The second argument in all these cases is a PATIENT. Evidently, grouping together θ-roles into macroroles makes it possible to formulate the following generalization: The presence of a PATIENT is related to the presence of an ACTOR (and vice versa) with verbs expressing a dynamic action. The other two predicates select a LOCATIVE instead; the former is intended as a , the latter as the where a thing is found. Both express a stative event. Although THEME is the first participant in these types of predicates (see § 2.2.1.), the second participant (LOCATIVE) is projected syntactically as the sentential subject, contrary to the norm. This, in our view, is a surface difference that depends on the semantic properties of the verb and, possibly, on the distribution of information in terms of foreground and background (see § 3.2.1. and note 17 to Chapter 2). In verbs like belong, whose meaning is close to have or own, the THEME is projected as the subject and is normally a definite NP (e.g.: The book belongs to me vs. I have a book). Furthermore, in many languages of the world, such as Arabic and Russian, predicates like have or own are realized through nominal constructions (a Small Clause in structural terms; see note 1 to Chapter 2). In such constructions, the THEME functions as the subject and the (predicative) LOCATIVE is realized with a PP: (52) a. Al-kitƗb li-Marco. (ARABIC) DET-book.NOM to-Marco.OBL ‘Marco has the book.’ [lit.: ‘The book (is) of Marco.’] b. Al-kitƗb ’inda al-walad. DET-book.NOM at DET-boy.OBL ‘The boy has the book.’ [lit.: ‘The book (is) at the boy.’]

Argument structure c. U menja kniga. at O.PRO.1SG book.NOM ‘I have the book.’ [lit.: ‘The book (is) at me.’]

57

(RUSSIAN)

All the verbs in (46)-(51) are bivalent and transitive. They should not be equated, however, because valency is a semantic property, while transitivity has to do with syntactic structure, in particular with projecting the second argument of a predicate as a direct object (see Chapter 1, § 1.5.). Besides, not all bivalent verbs are transitive: (53) Everything was clear to Jim.

THEME-GOAL

(54) I went to Milan.

THEME-LOCATIVE

(55) The flight departs from Frankfurt.

THEME-LOCATIVE

(56) The show lasts three hours.

THEME-DURATION

(57) The timber weighs a ton.

THEME-QUANTITY

22

These verbs all have two arguments23 but are intransitive, regardless of whether the second argument is accompanied by a preposition. Furthermore, the first argument is always a THEME, which combines with different macro-roles. The presence of a THEME as a first participant is a simple consequence of the previous generalization: Since a transitive verb requires a PATIENT and a PATIENT depends on the presence of an ACTOR, a nontransitive predicate will rule out both macro-roles. In particular, it can be easily proved that the argument formally presented as a direct object in sentences such as (56) and (57) is not actually a direct object since it cannot function as the subject in a passive sentence: (58) *Three hours are lasted by the show. (59) *A ton is weighed by the timber. Let us now consider the argument structures of the verbs in the following pairs: (60) a. Jim kissed Louise.

58

The structure of the verb phrase b. Jim gave a kiss to Louise.

(61) a. A kid punched one of his classmates in the face. b. A kid gave one of his classmates a punch in the face. (62) a. He helped the victims. b. He brought help to the victims. The (a) sentences each have two participants: The first is an ACTOR, the second a (GOAL macro-role), although the lack of a preposition (and the consequent presence of a structural ACC Case) might lead us to think otherwise. These cases resemble those examined in § 2.2.2. (with the verbs rain and sing) where one of the two argument roles – the – is contained in the verb root. As a matter of fact, if the PATIENT is not lexicalized in the predicate and is realized as an overt argument, the 24 GOAL must be preceded by a preposition (as in [60b]-[62b]). Notice that a different lexicalization may determine a different composition of the argument structure and a different syntactic structure. In a language like Italian, in sentences equivalent to (a), the GOAL can be projected as the subject of a passive sentence: (63) Luisa è stata Luisa be.PRES.3SG be.PRT.SGF da Mario. by Mario ‘Luisa was kissed by Mario.’

baciata kiss.PRT.SGF (ITALIAN)

However, in the corresponding (b) sentences, where the GOAL is not realized as a direct object, the subject of the corresponding passive sentence can only be the PATIENT (on the other hand, this is possible in languages like English; see discussion below regarding example [89]): (64) a. Un

bacio è kiss be.PRES.3SG dato a Luisa da by give.PRT.SGM to Luisa ‘A kiss was given to Luisa by Mario.’

INDEF.DET.SGM

stato be.PRT.SG Mario. Mario

Argument structure b. *Luisa è Luisa be.PRES.3SG un bacio INDEF.DET.SGM kiss

stata be.PRT.SGF da Mario. by Mario

59

data give.PRT.SGF

This shows that there is a hierarchy between macro-roles and the structural position they may occupy, with the latter determining the possibility of realizing specific syntactic structures (this topic will be dealt with in detail in § 2.2.5.). Finally, let us consider the argument structure of verbs like fill and break by analyzing the following English sentences: (65) I filled the barrel with this year’s wine. (66) Jim broke the glass with his elbow. These verbs seem to select three arguments, namely , and . However, both and belong to the ACTOR macro-role. Although from a semantic point of view the two arguments may be considered as participants in the actions described, the interface with syntax cannot admit the presence of both arguments within the VP, as they have to be merged in the same structural position, i.e. Spec,VP. Since is the higher argument role (its feature [+animate] giving it more control on the action), the conclusion is that the PP must be an adjunct. Therefore, verbs of this type must be considered syntactically bivalent.25 The need for the to be assigned a circumstantial role in the presence of a [+animate] argument belonging to the same macro-role is also clearly shown in polysynthetic languages. As we know, in these languages arguments are realized as clitics on the verbal head. Let us take a sentence from Warlpiri (from Simpson 1991), analyzed in Chapter 1: (67) Karli-kirli-rli-ji luwa-rnu. boomerang-ASSOC-ERG-1SG.ACC shoot-PST ‘The one with the boomerang shot me.’

(WARLPIRI)

In this sentence, karli (‘boomerang’) cannot be the subject, since the ERG Case morpheme entails a [+human] referent. In addition, it is a subject-modifying adjunct, as the presence of the associative particle -kirlishows.

60

The structure of the verb phrase

Similarly, if knife is a circumstantial in (68), one may ask how (69) should be analyzed: (68) Mary cuts the bread with a knife. (69) That knife cuts well. In (68), the participants to the action of cutting are an ACTOR and a In (69), on the other hand, there is just one argument, whose macro-role is THEME. The two different argument structures depend on different interpretations of the event. In the former case, the verb refers to an activity which requires an agent and an object to which the action applies; indication of the instrument used for performing the activity is optional. The evidence that the is optional is that it may always be omitted (as in [70a]), unlike the : PATIENT.

(70) a. Mary cuts the bread. b. *Mary cuts.26 In (69), conversely, the predicate cut does not refer to an activity but rather to a property attributed to its argument. It is therefore a stative verb. In other words, a sentence such as (69) may be paraphrased as a copular structure in which a quality of the THEME-subject is predicated (i.e. ‘the knife is sharp’). Evidence for this analysis comes from classical languages like Latin where such predications were expressed as follows: (71) Ensis acutus est. knife.SGM sharp.SGM be.PRES.3SG ‘The blade cuts.’ [lit.: ‘The blade is sharp.’]

(LATIN)

We conclude that verbs like cut entail two lexical entries in English (and presumably in other languages too), with different semantic features according to the event structure of the verb (stative vs. dynamic): (72) cut 1: [+dynamic] two-argument predicate [ACTOR, PATIENT] mono-argument predicate [THEME]27 cut 2: [+stative] Evidence of the fact that some verbs need a double lexical entry depending on the event structure of the verb also comes from Tagalog. In

Argument structure

61

this language, verb morphology signals both the argument role of the trigger28 and the type of event. Let us consider the following sentences: (73) a. G-um-aling ACT.PST-heal na ito. ASSOC

si TRIG

Maria Maria

dahil cause

gamot OBL medicine (TAGALOG)

DEM

‘Maria was healed by this medicine.’ b. Ikin-a-galing ni Maria ang STAT.ACT-PST-heal DIR Maria TRIG na ito. ASSOC

sa

gamot medicine

DEM

‘This medicine healed Maria.’ In sentence (73a), Maria is the trigger, and the instrument for her healing is realized by a complex PP (dahil+sa) which signals its syntactic role of adjunct. In (73b), on the other hand, the verb presents a stativizing prefix, which points to a different interpretation, and consequently a different argument structure, of the event expressed by the semantic root galing: Maria is no longer an ACTOR but a PATIENT, and gamot becomes the ACTOR, thus assuming an argument role. As a consequence, gamot is no longer preceded by dahil sa, but by the marker ang. Verbal morphology indicates that gamot, an , is itself an ACTOR, although the form of the infix does not match that in (73a). These data show that a single verbal root may correspond to several lexical entries. The argument structure of a predicate cannot therefore be analyzed independently from its event structure.

2.2.4. Three-argument verbs The three arguments in the thematic grid of trivalent verbs may belong to different macro-roles. a)

ACTOR-PATIENT-GOAL

for verbs such as give, donate, send:

(74) a. He gave Anna a book. b. We have sent a letter to the Chancellor. b)

ACTOR-PATIENT-LOCATIVE for

verbs such as put, lay, place:

62

The structure of the verb phrase

(75) a. John put the wood in the fireplace. b. Anna placed the glasses on the table. c) ACTOR-PATIENT-COMITATIVE for verbs such as arrange, associate, reconcile: (76) a. I arranged everything with Paul. b. Anna reconciled herself with her sister. In these cases, the function of the third argument appears to be rather different from examples (74)-(75). Sentences (76) may alternatively be formulated as in (77), where the COMITATIVE (a macro-role expressing an argument of , which we have not analyzed in detail in section 2.2.1.), is not the third argument, but is included in the ACTOR and projected with the surface function of subject: (77) a. Paul and I arranged everything. b. Anna and her sister reconciled. In this case, variation is possible between a three-argument predicate (with a COMITATIVE third argument) and a two-argument predicate whose subject-ACTOR needs to consist of two coordinated elements or a plural: (78) The guys arranged an intervention.

2.2.5. Macro-roles and VP structure We have so far used a set of macro-roles to describe nuclear sentences in different languages: ACTOR, PATIENT, THEME, GOAL, LOCATIVE, COMITATIVE, DURATION and QUANTITY. From a semantic point of view, the argument roles included in these macro-roles can be classified in a more specific way. A distinction can be made, for instance, between ACTOR and PATIENT as “first order elements” (Lyons 1977: 495) involving humans, as opposed to CAUSE and EFFECT as “second order elements” involving inanimate referents too. Also, different types of LOCATIVE can be distinguished. What we aim at, however, is not a semantically complete description, but the understanding of interaction phenomena among the various levels of analysis needed to describe languages. It will therefore be interesting to see how the broad range of semantic roles is arranged within

Argument structure

63

the set of macro-roles and deep Cases, which in turn interface with grammatical functions and the morpho-syntactic surface realization. There is a hierarchy between argument roles which is related to their number and determined by the semantics of predication. Clearly, there cannot be a hierarchy in the case of mono-argument verbs (as a single macro-role is found in the VP); nevertheless, we have noticed that also these verbs present restrictions on the types of realizable ș-roles (§ 2.2.2.). If more than one argument is present, not only is there a hierarchy between the different macro-roles, but there are also restrictions on their cooccurrence. When the first argument is an ACTOR, a PATIENT must be present in the predication; when the first argument is a THEME, it is usually linked to a LOCATIVE (§ 2.2.3.). Finally, some combinations between macro-roles appear to be completely ruled out by grammar. For example, no bivalent verb has *ACTOR-THEME or *ACTOR-LOCATIVE as its arguments; similarly, no trivalent verbs have an *ACTOR-GOAL-LOCATIVE argument structure (§ 2.2.4.). Furthermore, two constituents belonging to the same macro-role cannot be arguments within the same predication. This restriction may be considered a consequence of the ș-Criterion (see note 11 to Chapter 2). The structural assignment of ș-roles (within macro-roles) takes place in the VP as a consequence of the structural condition verified at sister-node level. Thus, the verbal head (in V°) determines ș-role assignment to Compl, and the verb and complement (i.e. the V' projection) together select the șrole for the sister node of the Spec. In Chapter 1 (§ 1.2.3.1.) we proposed an illustration of the structural position of argument roles in the VP. A more detailed picture, based on macro-roles, can now be given: (79)

VP V’

ACTOR

V’ ș-role V°

GOAL LOCATIVE QUANTITY MEASURE COMITATIVE PATIENT THEME

ș-role

64

The structure of the verb phrase

Our assumption is therefore that the VP structure is more complex than shown in Chapter 1, and V' is doubled so as to create two distinct, yet structurally identical, positions (both being V' nodes).29 The two groups of macro-roles found in Compl,VP both qualify as complements of V° and, as such, receive their ș-role from the verbal head (as shown in [79]). Notice, however, that the position of PATIENT/THEME establishes a special relationship with the verb since it is structurally the closest to it. This is also shown by the fact that these two macro-roles do not need a preposition to receive their Case (see Chapter 3). The structure in (79) helps us to understand the rationale behind the hierarchy in macro-roles and the reasons for their possible combinations, partly based on their syntactic functions in the sentence. If we consider the positions at which macro-roles are generated, the following structural hierarchy clearly emerges: (80) HIERARCHY OF MACRO-ROLES ACTOR > PATIENT/THEME > GOAL/LOCATIVE/QUANTITY/MEASURE/ COMITATIVE

Since each syntactic position can only host one constituent (as a direct consequence of the Thematic Criterion; see note 11 to Chapter 2), the generation of a PATIENT clearly rules out the presence of a THEME. Similarly, if a GOAL is merged, no other macro-role can be merged in the same position. As a consequence, unless a verb selects an ACTOR, its argument grid can be expected to be bivalent at most. Based on the syntaxsemantics interface (i.e. the insertion of arguments into the VP), strong predictions are possible concerning the combinatorial possibilities of macro-roles and the complementary distribution of some of them. Further predictions are possible if (80) is combined with the hierarchy among syntactic functions (in [81]): (81) HIERARCHY OF SYNTACTIC FUNCTIONS SUBJ > DO > IO What (81) shows is that the syntactic function of the subject is the first that requires realization, followed by the direct object (DO) and all the others. This means that with a mono-argument verb, the only macro-role selected will necessarily fulfill the function of subject. If there are two macro-roles where one is an ACTOR, the ACTOR will be the subject, because it fulfills the first function in the hierarchy in (81). With a bivalent verb

Argument structure

65

whose second macro-role is a PATIENT, the PATIENT will have the role of DO; if no PATIENT is present, a GOAL (the third macro-role in the hierarchy in [80]) will be able to fulfill this syntactic function as it precedes the indirect object (IO) function in the hierarchy in (81). Let us go back to the implications of the structure in (79): It can be observed that the union of a verbal head and its Compl (of any type) determines the argument role of the argument in Spec,VP. This assumption is proved by sentences like the following: (82) Mick weighs the flour. (83) Mick weighs 150 pounds. In (82), the argument merged in Compl position is a PATIENT and the union of [V°+PATIENT] assigns the role of ACTOR to the constituent (in Spec,VP) projected with a subject function. Conversely, in (83) the Compl is a QUANTITY argument: As a consequence, the argument with subject function cannot be an ACTOR, but has to be a THEME. In conclusion, the semantic hierarchy that exists between argument roles is reflected by the syntactic structure (in the VP) into which macro-roles are merged based on a specific order. This order accounts for a number of effects, including what is known as the superiority of the ACTOR over other macro-roles, shown in the behavior of reflexives. Let us consider the following sentences: (84) a. Johnk meets hisk friend. b. *His ownk friend meets Johnk. For a reflexive to receive interpretation, its antecedent must be merged at a higher point in the structure. For this reason, the sentence in (84a) can be correctly interpreted, whereas that in (84b) is ungrammatical as the reflexive his own is merged higher in the structure than its antecedent.30 The hierarchical order between arguments is also evident in the restrictions on the relative order of agreement morphemes in polysynthetic and agglutinative languages (see Chapter 3, § 3.4.).

66

The structure of the verb phrase

2.3. Argument roles, event structure and overt syntax We will now focus on the interface between the event structure of a predicate (dynamic vs. stative), the argument selection and the syntactic structure in an attempt to describe the relationship between macro-roles, deep Cases and syntactic functions. These will be dealt with more extensively in Chapter 3. Here we will concentrate on the structural implications of the fact that arguments with different macro-roles may assume subject function. To this end, we will examine typologically different languages, which present interesting and important similarities at a deep level in spite of surface differences.

2.3.1. Event structure and macro-role of the subject in inflecting and agglutinative languages Let us begin our analysis by considering data from English, a (partially) inflecting SVO language: (85) a. John broke the window with a stone. b. The storm broke the window. c. The stone broke the window. Because the subject in the three examples belongs to the ACTOR macrorole, we know for sure that a passive structure will always be possible, although the adjunct will be realized in slightly different ways: (86) a. The window was broken by John. b. The window was broken by the storm. c. The window was broken with/ ??by the stone. As we notice, the selected preposition depends on the ș-role of the adjunct-ACTOR: (unlike and ) is preferably introduced by the preposition with, although by is not considered completely unacceptable. Hence, preposition selection is sensitive to macro-role-internal semantic distinctions. In stative-type events, on the other hand, the ACTOR macro-role cannot be realized. This is completely expected since the syntactic subject is a THEME:

Argument roles, event structure and overt syntax

67

(87) The window broke *by John/ *by the storm/*by the stone/*with the stone. A typical construction of English is what is traditionally known as a ditransitive (or double-object) construction: A three-argument predicate (ACTOR-PATIENT-GOAL) may have a canonical realization with the PATIENT receiving the syntactic function of direct object and the GOAL functioning as an indirect object, as in (88a); however, the GOAL may alternatively be realized in second position (with no preposition), thus preceding the direct object (88b):31 (88) a. John gave a kiss to Mary. b. John gave Mary a kiss. A corresponding passive sentence is possible for both (88a) and (88b): (89) a. A kiss was given by John to Mary. b. Mary was given a kiss by John. On the other hand, the prepositional object cannot be made passive: (90) *To Mary was given a kiss by John. This means that the element that functions syntactically as a direct object in the active construction can become the subject of a passive construction regardless of its argument role, provided the functional category Voice (cf. Speas 1991, Fox & Hopper [eds.] 1994) is activated (i.e. if the ACTOR is outside the VP).32 What was observed in English is also true in many other languages, albeit with some slight differences, both at structural level and in prepositional use (overt markers of deep macro-roles). In Italian, for example, we can find: (91) Carlo ha bruciato Carlo have.PRES.3SG burn.PRT ‘Carlo burned the stubble.’ (92) Le

le DET

stoppie. stubble.PL

stoppie hanno bruciato stubble.PL have.PRES.3PL burn.PRT ‘The stubble burned for a long time.’

DET

(ITALIAN)

(a lungo). for long

68

The structure of the verb phrase

In (91), the verb bruciare has two arguments belonging to ACTOR and macro-roles: Here, the ACTOR is projected as the subject and determines the agreement with verb inflection. In (92), the verb is again bruciare, but occurs with a single argument – stoppie – with THEME macrorole: This argument, projected as a subject, determines third-person plural agreement with the verb. There is a difference between the two bruciare: In (91) it is a causative (syntactically transitive) verb which denotes an action; in (92) the verb refers to a change of state, and involves the only argument selected (which obviously cannot be an ). We can conclude that the two occurrences of bruciare in (91) and (92) reflect two distinct lexical entries (as in the case of cut in [68]-[69]). Let us now consider the passive sentence in (93): PATIENT

(93) Le

stoppie sono stubble.PL be.PRES.3PL da Carlo. by Carlo ‘The stubble was burned by Carlo.’

DET

state be.PRT.PLF

bruciate burn.PRT.PLF

Again, bruciare has an ACTOR and a PATIENT. In this case, however, the is projected as a subject (as shown by the third-person plural agreement).33 While semantically present in the verb, the ACTOR is realized syntactically as an adjunct, and cooccurs with the preposition da (‘by’), which signals its role of outside the VP. Realizing the as an adjunct is structurally necessary for the PATIENT to be projected as a subject. This is because the projection of the subject depends on the combination between the hierarchies in (80) and (81). Therefore, whenever an ACTOR is realized in Spec,VP, it is necessarily projected as the sentential subject.34 A non-ACTOR subject thus entails an empty Spec,VP position, whereby an argument generated in Compl,VP will be projected as a subject (based on the hierarchy in [80]). Since an adjunct-ACTOR can be present, the event expressed by a verb with passive morphology is still dynamic. Indeed, a passive does not express a state, because it can be modified by duration adverbials, as in The stubble was burned for hours. However, in a language like Italian the auxiliary essere (‘be’) is used in compound tenses. Essere is not only a tense marker, but also indicates that a functional category is activated in the inflection area (i.e. the IP area), namely Voice, whose function is to make the sentence passive. PATIENT

Argument roles, event structure and overt syntax

69

In a sentence like (94), on the other hand, the predicate event is not an action, but simply expresses a final result, i.e. a state. The argument selected by the verb is therefore a THEME (with the synctactic function of subject) and recovery by means of an adjunct is completely ruled out: (94) Le

stoppie si sono stubble.PL MIDD be.PRES.3PL (*da Carlo). by Carlo ‘The stubble burned (*by Carlo).’ DET

bruciate burn.PRT.PLF

Both in (94) and in (92), the subject is a THEME (and as such rules out any ACTOR). The event in (92), however, is dynamic, whereas (94) expresses a state, as shown by the fact that it cannot be modified by duration adverbials: (95) Le

stoppie si sono bruciate stubble.PL MIDD be.PRES.3PL burn.PRT.PLF (*a lungo). to long (intended: ‘The stubble burned [for a long time].’)

DET

For this reason, the sentence in (94) is also distinct from the passive sentence in (93). In spite of superficial similarities (such as the presence of the auxiliary essere), the passive sentence expresses a dynamic event, where the affected argument (PATIENT) is assumed as the starting point. Conversely, the sentence in (95) expresses a state, a condition. Such constructions are called middles in the literature (see Keyser & Roeper 1984, Cinque 1990a, Shibatani 2004) and the presence of the particle si (not to be confused with a reflexive) in Italian is a typical sign of this. If the auxiliary essere is used in both middles and passives, it is not because of the event structure of the predicates in question, but because an ACTOR is lacking in Spec,VP, and consequently a non-ACTOR is projected as subject (a PATIENT-subject in passives and a THEME-subject in middles). The passive construction therefore expresses a dynamic-type event where a PATIENT is projected as the subject, regardless of the ș-role of the ACTOR. Given the interaction between the hierarchy on macro-roles (80) and that on syntactic functions (81), it is clear that the realization of a

70

The structure of the verb phrase

PATIENT-subject implies that an ACTOR, if present, is an adjunct. Let us consider the following Italian sentences:

(96) a. Tutti temono l’ esame di stato. exam of state all fear.PRES.3PL DET ‘Everybody is afraid of the final exam.’ b. L’ esame di stato preoccupa tutti. DET exam of state worry.PRES.3SG all ‘The final exam worries everybody.’ The ș-role of the ACTOR is in (96a) and in (96b). Both sentences, however, have a corresponding passive sentence, which goes to prove that this option merely depends on the ACTOR-PATIENT combination: (97) a. L’

esame di stato è temuto da exam of state be.PRES.3SG fear.PRT.SGM by tutti (*per tutti). all for all ‘The final exam is feared by everyone.’ b. Tutti sono preoccupati dall’ /per l’ all be.PRES.3PL worry.PRT.PLM by.DET / for DET esame di stato. exam of state ‘Everybody is worried about the final exam.’ DET

As (97a-b) show, however, the ș-role of the ACTOR is crucial for its realization as an adjoined PP: The preposition da may be used for both the and the , while per is only used for the .35 Let us also consider the following sentences, where the ACTOR is realized as an adjunct: (98) a. I DET

tuoi POSS.2.PL

comportamenti behavior.PL

sorprendono. surprise.PRES.3PL ‘Your behavior surprises me.’

mi OCL.1SGM

Argument roles, event structure and overt syntax

71

b. (Io) sono sorpreso dai/ I be.PRES.1SG surprise.PRT.SGM by.DET/ dei /per i tuoi comportamenti. for DET POSS.2.PL behavior.PL of.DET/ ‘I am surprised by/at your behavior.’ c. (Io) mi sono sorpreso dei/ I OCL.1SG be.PRES.1SG surprise.PRT.SGM of.DET per i/ *dai tuoi comportamenti. for DET by.DET POSS.2.PL behavior.PL ‘I was surprised by/at your behavior.’ (98a) and (98b) mirror (96b) and (97b) respectively. (98c) includes an extra argument, mi (a reflexive pronoun, co-referent with the -subject). In this case, the third argument (, realized as an adjunct) cannot cooccur with the preposition da. We conclude that the preposition da can only be used with a second argument.36 Finally, let us consider a sentence like (99), where the argument with subject function is the of the event: (99) Un INDEF.DET

sasso stone

ha have.PRES.3SG

rotto break.PRT

il

vetro. glass ‘A stone broke the glass.’

DET

As in the case of cut in (69), this break is a distinct lexical entry from the causative break (as in I broke the glass). Moreover, since the belongs to the ACTOR macro-role, no other can be added, either as an argument or as an adjunct: (100)

a. Un

ha have.PRES.3SG il vetro (*da Leo). DET glass by Leo ‘A stone broke the glass (*by Leo).’ b. *Un sasso ha INDEF.DET stone have.PRES.3SG da Leo. by Leo INDEF.DET

sasso stone

rotto break.PRT

rotto break.PRT.SGM

72

The structure of the verb phrase

Let us now see how argument roles and the predicate event structure interact in an agglutinative language such as Swahili (from Vitale 1981).37 In Swahili, subject and object clitic pronouns are realized on the verbal head, following a strict order (the subject invariably preceding the object, and the two clitics being separated by the temporal morpheme): (101)

Juma a-li-wa-pig-a Juma SCL.3SG-PST-OCL.3PL-hit-IND ‘Juma hit the kids.’

(SWAHILI)

watoto. kids

The corresponding NPs are arranged as shown in (101), i.e. in the unmarked SVO order, although their order may change according to pragmatic requirements. In the case of trivalent verbs, the third macro-role is also realized on the verbal head, and the corresponding constituent is not preceded by the preposition kwa (a multifunctional preposition; see [27]):38 (102)

Juma a-li-m-tum-i-a Juma SCL.3SG-PST-OCL.3PL-send-DAT-IND dada yake. sister POSS ‘Juma sent the letter to her sister.’

barua letter

In passive sentences, the morpheme -w- is used in Swahili. In this case, the PATIENT assumes subject role and is signalled on the verbal head by the same clitic pronoun found in active sentences. An ACTOR may be present, and is subsequently realized by means of an adjoined PP: (103)

Mlango a-li-fung-w-a door SCL.3SG-PST-open-PASS-IND ‘The door was opened (by Juma).’

(na by

Juma). Juma

Interestingly, Swahili has a particular morpheme which is both intransitivizing, as defined by Vitale (1981), and stative: -ik-/-uk-, based on vowel harmony. This morpheme is used in sentences like the following, where, crucially, the ACTOR cannot be realized: (104)

Mlango u-me-fung-uk-a SCL.3SG-PERF-open-STAT-IND door ‘The door opened (*by Juma).’

(*na by

Juma). Juma

Argument roles, event structure and overt syntax

73

In this sentence, the tense conveys a perfective, i.e. a completed action. This is in line with our previous analysis of English and Italian data: Unlike the passive in (103) which expresses a dynamic action, the event marked by –uk– in (104) is a stative one. The predicate here expresses a condition of the subject-THEME; consequently, the realization of an ACTOR is ruled out. Finally, the form of the third person subject clitic also changes according to event type: Clitic a- denotes the subject of a dynamic action (as in [101][103]), whereas u- is used with a THEME-subject in stative events. This formal difference gives cross-linguistic support to the idea (proposed in section 2.2.3.) that some verbs present two distinct lexical entries depending on their different event structures (see [72]). In conclusion, we can say that both inflecting and agglutinative languages show a clear and important correlation between the argument structure of the predicate and the type of event expressed. This correlation is manifested morpho-syntactically by different means: Free morphemes (i.e. Italian si), bound morphemes (as in Swahili), auxiliary selection (as in Italian), preposition selection for the ACTOR to be realized as an adjunct in passive sentences (as in English and Italian), and complete exclusion of the ACTOR in stative predicates (in all three languages). We have also seen that in a nuclear sentence with a mono-argument predicate, the only argument merged in the VP needs to assume the subject function (thus determining agreement with the verb). When there are two arguments, different cases are possible depending on other aspects of the structure. In an active sentence, the hierarchically highest argument will assume subject function; on the other hand, the PATIENT (merged in Compl,NP) will only be able to assume subject role if the ACTOR is merged outside the VP as an adjunct. Finally, it has been noted that the hierarchical promotion of an argument determines the need for the vacant structural position to be filled.

2.3.2. Event structure and macro-role of the subject in incorporating languages In this section, the relationship between macro-roles, syntactic functions and the event structure of the verb will be analyzed in languages where arguments are realized as clitics incorporated on the verbal head. Examples from Somali will be used for this discussion. Let us begin our analysis of the relationship between the argument structure and semantic property of the predicate by looking at a monovalent

74

The structure of the verb phrase

intransitive-stative verb, buux (‘be full’) in Somali. This verb has a THEME as its only argument, projected as the subject: (105)

Koob-ku waa glass-DET.NOM DECL ‘The glass was full.’

buuxay. fill.PST.3SGM

(SOMALI)

A causative infix, -i(s)- can be added to the root buux just like many similar verbs, making it bivalent and transitive and consequently dynamic. The two selected arguments therefore belong to the ACTOR and PATIENT macro-roles:39 (106)

Calí koob-ka waa Cali.NOM glass-DET DECL ‘Cali filled the glass.’

buux-i-yay.40 fill-CAUS-PST.3SGM

As expected, the ACTOR assumes subject function in (106). An additional constituent may also be inserted in the sentence:41 (107)

Calí koob-ka waa ku buux-i-yay Cali.NOM glass-DET DECL from fill-CAUS-PST.3SGM ‘Cali filled the glass with water.’

(108)

Calí koob-ka Axmed waa Cali.NOM glass-DET Ahmed DECL ‘Cali filled the glass for Ahmed.’

biyo. water

u buux-i-yay. for fill-CAUS-PST.3SGM

In (107), ku biyo has an argument role; in (108), u Ahmed is a . Both constituents are syntactically adjuncts. If, however, the third argument is coreferent with the ACTOR, the verb assumes an autobenefactive meaning, signalled by a specific moprheme (-ad-) on the verbal head: (109)

Calí koob-ka waa Cali.NOM glass-DET DECL ‘Cali filled his own glass.’

buuxsaday (buux-is-ad-ay). be full-CAUS-BEN-PST.3SGM

In this case, no third argument is possible:

Argument roles, event structure and overt syntax (110)

75

*Calí koob-ka waa is-u buuxsaday (buux-is-ad-ay). Cali.NOM glass-DET DECL REFL-for be full-CAUS-BEN-PST.3SGM

If the morpheme -am- is added to the causative-transitive verb buuxis (as in [108]), the verb receives a different interpretation, traditionally described as passive. However, since no ACTOR can be expressed (agentive by-phrases are non-existent in Somali), we assume that this construction is actually not a passive. In fact, the suffix -am- forces a stative interpretation of the predicate (whose subject is therefore a THEME): (111)

Koob-ku waa glass-DET.NOM DECL ‘The glass filled.’

buuxsamay (buux-is-am-ay). be full-CAUS-STAT-PST.3SGM

In sum, Somali morphology clearly signals event structure. From this perspective, the infix -am- might be analyzed in the same way as Italian si. Let us compare two sentences such as (112) and (113): (112)

Calí guri-ga waa Cali.NOM house-DET DECL ‘Cali burned the house.’

gubay. burn.PST.3SGM

(113)

Calí albaab-ka waa Cali.NOM door-DET DECL ‘Cali opened the door.’

furay. open.PST.3SGM

In both sentences there is an ACTOR and a PATIENT; yet, when the is projected with subject function, the two verbs assume different forms: PATIENT

(114)

Guri-gu waa gubtay (gub-ad-ay). house-DET.NOM DECL burn-CHAN-PST.3SGM ‘The house burned down.’

(115)

Albaab-ku waa door-DET.NOM DECL ‘The door opened.’

furmay (fur-am-ay). open-STAT-PST.3SGM

The suffix –ad in (114) makes the verb gub intransitive and semantically denotes a change of state, i.e. a dynamic-type action. In (115), on the other

76

The structure of the verb phrase

hand, the suffix -am- changes fur into an intransitive stative verb (as in the case in [111]). This difference must be ascribed to the specific semantics of the individual verbs. When a verb like burn, which has a duration in time, presents an intransitive structure, the only option seems to be a dynamic event with a THEME subject. Conversely, the intransitive verb open in (115) states a condition of the subject-THEME. This difference is confirmed by the fact that the verb fur, unlike gub, may cooccur with -ad and assume autobenefactive value in a transitive structure (with ACTOR and PATIENT):42 (116)

Cali albaab-ka waa furtay (fur-ad-ay). Cali.NOM door-DET DECL open-CHAN-PST.3SGM ‘Cali opened the door for himself.’

This means that the semantic structure of predication in Somali, including its arguments and ș-roles, is overtly projected through the same grammatical functions used in Italian. Here, too, whenever an ACTOR is present, it must assume the subject function. However, under no circumstances can the ACTOR appear as an adjoined PP in Somali, unlike in Italian. This supports the assumption that there is no passive construction proper in Somali.43 As we have seen, in languages like Somali morphemes can convey information on the event structure of the verb, thus determining important variations in the number and type of arguments selected. A monoargumental verb, for instance, can have a corresponding causative (transitive) or autobenefactive verb (see [107]-[109]). Similarly, when the argument projected as subject does not have the ș-role, a diathesis marker is required on the verb (traditionally described as a root extension; see Andrzejewski 1968, 1969, Puglielli [ed.] 1985). This marker determines a middle or stative value for the verb and may entail a reduction in the number of arguments. In conclusion, information regarding the semantic structure of the sentence is not only conveyed by overt syntactic structure, but also by morphological elements that are usually considered derivational. This functional exchange between inflection and derivation, noted by authors such as Booij (2005), is yet another argument in favor of the hypothesis that derivational morphology also takes place within the syntactic component of grammar, following the principles of Distributed Morphology (see note 26 to Chapter 1).

Argument roles, event structure and overt syntax

77

2.3.3. Event structure and macro-role of the subject in predicate-initial languages Having examined languages from different morphological types, we conclude this section by focusing on syntactically distinct language types, i.e. predicate-initial languages. Although these languages have specific formal properties (cf. Pearson 2000, Rackowsky & Travis 2000, Frascarelli in press), the relationship between macro-roles and event structure applies here in the same way as in the previously examined SVO and SOV languages. Let us take Tagalog (a VSO language) as our sample language.44 In Tagalog, much like in Somali and Swahili, the verbal root may take on various infixes that modify the semantic structure of the predicate, enlarge its argument structure and alter the type of macro-roles selected by the verb. For example, a monovalent verb like akyat (‘go up’) can be preceded by a morpheme which makes it causative and, consequently, bivalent and transitive: (117)

Umakyat

si

ACT.PST.go up TRIG

Juan sa puno ng Juan OBL tree ASSOC

bayabas. guava ‘Juan climbed the guava tree.’ (118)

(TAGALOG)

Nag (um-pag45)-akyat si Juan ng tubig ACT-CAUS.PST-go up TRIG Juan DIR water ‘Juan carried the water up to his house.’ [lit.: ‘Juan had the water go up to his house.’]

sa OBL

bahay. house

Similarly, an intransitive verb like galit (‘be angry’) presents a stative prefix, incorporated in the morpheme that signals the argument role THEME (see [119]). It can, however, become bivalent if the ACTOR macro-role (which realizes the in [120]) is used as sentence trigger and is therefore realized on the verbal head: (119)

Na (ma-in)-galit ang TH-STAT.PST-be.angry TRIG ‘Mum was angry.’

nanay. mum

78

The structure of the verb phrase

(120)

Nakagalit ang mga kanya-ng ACT.PST.be angry TRIG PL PRO.3SG-ASSOC sa nanay. OBL mum ‘His statements made Mum angry.’

sabihin statement

As can be seen in (118) and (120), when the valency of a verb is enlarged (by the presence of a specific infix), the relevant argument must be realized as the trigger. The same is true in causative structures, where the needs to appear as the highest participant in the event (see sections 3.2.1., 3.4.1.): (121)

(122)

Nagdala si Juan ng ACT.PST.give TRIG Juan DIR ‘Juan brought me a letter.’ Nag-pa-dala

si

ACT.PST-CAUS-give TRIG

sa

liham sa letter OBL

Pedro ni Pedro DIR

Juan Juan

akin. PRO.1SG

ng DIR

liham letter

akin.

OBL PRO.1SG

‘Pedro had Juan bring me a letter.’ Verb valency in Tagalog can thus be enlarged through verbal morphology, provided that the added argument is realized as a trigger. In fact, there are further restrictions depending on the semantics of the verb. For example, the trigger for a verb like sit can be an ACTOR (123a) or a LOCATIVE (123b), provided that the locative denotes a referent related to the action of sitting ([123c] is consequently ruled out): (123)

a. Um-upo siya sa silya. ACT.PST-sit PRO.3SG.TRIG. OBL chair ‘S/he sat on the chair.’ b. In-upu-an niya ang silya. PST-sit-LOC PRO.3SG TRIG chair ‘S/he sat on the chair.’ c. *In-upu-an niya ang restaurante. ACT.PST-sit-LOC PRO.3SG TRIG restaurant ‘S/he sat in the restaurant.’

Arguments realized by clauses

79

We believe that the data presented are adequate evidence that the semantic structure of a verb substantially determines its valency and that the combination of these factors has important repercussions on the morpho-syntactic realization of sentences in typologically different languages.

2.4. Arguments realized by clauses In the last section of this Chapter, we will look at the formal realization of arguments in non-prototypical cases. In the sentences examined so far, arguments have always been NPs (or PPs). Now we are going to see what happens when an argument is realized by means of a whole clause, as in the following examples: (124)

a. b. c. d. e.

Kate hopes that Jim won’t come tomorrow. Kate said that Jim is coming tomorrow. Kate wishes that Jim wasn’t coming tomorrow. Kate knows that Jim is coming tomorrow. Kate has seen that Jim left.

In all these examples, the second argument – a THEME – consists of a clause which, in turn, is made up of a predication plus its arguments; this type of clause is sometimes called a . This clause is introduced by the complementizer that and has a finite verb form, marked for tense and agreeing in person and number with the argument functioning as the subject.46 Each of the verbs in (124) belongs to a semantic class defined as “verbs of saying, knowing and thinking, perception and volition” (cf. Frajzyngier 1996, Meinunger 2004). Some of them (e.g. hope) cannot have a simple NP as a second argument. On the other hand, the argument clause may assume a different surface form, as in (125): In this case, the argument clause has a non-finite verb and shows no agreement with the subject; its subject is interpreted as coreferent with the root clause subject: (125)

a. Kate hopes to come tomorrow. b. Kate wishes to come tomorrow.

In Generative Grammar, the subject of a non-finite clause is called PRO (in capital letters, as opposed to pro, the unpronounced subject of finite

80

The structure of the verb phrase

clauses), and its interpretation is the object of the Theory of Control. In particular, in (125a-b) there is coreference with the subject, but other arguments may also assume Control, as in (126a-b):47 (126)

a. Kate told Jim to come tomorrow. b. Kate ordered Jim to come tomorrow.

Here, the first argument of come is necessarily coreferent with Jim in both sentences and is assigned to the GOAL macro-role in the root clause. Ambiguous cases are also possible: (127)

Kate asked the coach to play a different position.

Here the ACTOR of play may be Kate, if ask is interpreted as ‘ask for permission’, or the coach, if ask means ‘require’. At any rate, once again the meaning of the root verb determines the number and type of the arguments required, as well as the type of Control applied, if one of the arguments is a clause. In all the sentences above, the clausal structure (with THEME macrorole) is projected with an object function. With mono-argument verbs, however, it assumes subject function, as in the following Italian examples: (128)

a. Sembra che i ragazzi seem.PRES.3SG that DET boy.PL cantino domani. tomorrow sing.SUB.PRES.3PL ‘It seems that the boys are singing tomorrow.’ b. È necessario che i ragazzi be.PRES.3SG necessary that DET boy.PL cantino domani. sing.SUB.PRES.3PL tomorrow ‘It is necessary that the boys sing tomorrow.’

(ITALIAN)

A sentence like (128a) may also appear with a different surface form, (129). The two sentences, however, are not perfectly synonymous:48 (129)

I

ragazzi sembrano cantare DET boy.PL seem.PRES.3PL sing.INF ‘The boys seem to be singing tomorrow.’

domani. tomorrow

Arguments realized by clauses

81

Here, the NP ragazzi is the subject of the root clause and as such triggers agreement with the verb sembrare. The subjectless subject clause, on the other hand, has a non-finite verb.49 This option is also made possible by the root verb, as shown by the fact that it is excluded in cases like (130): (130)

*The boys are necessary to sing tomorrow.

Let us now go back to sentence (124e), repeated below as (131). We can notice that there is again an alternative form, with a non-finite subordinate verb (132): (131)

Kate has seen that Jim left.

(132)

Kate has seen Jim leave.

Semantically, Jim in (132) is still an argument of leave, although it is not the syntactic subject of the subordinate clause, as it shows no agreement. From a structural point of view, it is the second argument (PATIENT) of the verb see: If the NP is replaced with a pronoun, this bears ACC Case: (133)

Kate has seen him leave.

This structure is typically found with verbs of perception (as in [134][135]) and is sometimes present in other languages (e.g. Italian, as in [136]) and with other verb classes (e.g. verbs of volition, as in [137]): (134)

a. I heard Kate sing. b. I heard her sing.

(135)

a. I watched the birds soar. b. I watched them soar.

(136)

L’

ho sentito have.PRES.1SG hear.PRT il pianoforte. DET piano ‘I heard him play the piano’. OCL.3SGM

suonare play.INF (ITALIAN)

82

The structure of the verb phrase

(137)

a. I wanted him to go. b. The President requires them to leave the country.

These structures are usually accounted for in the literature as cases of Exceptional (ACC) Case Marking (ECM; see Haegeman 1991): Since the constituent in the subordinate Spec,IP cannot be licensed as the subject of the non-finite verb, it exceptionally receives ACC Case from the root verb (as if it were its direct object). In other words, under this analysis, the whole clause is the argument (generated as a Compl) of the root verb, which assigns ACC Case to the constituent in Spec,IP: (138) [IP Kate [I’ has [VP tKate [V’ seen [IP Jim [I’ [VP tJim [V’ leave]]]]]]]]. [+ACC] Being exceptional, this mechanism clearly has no other independent motivation than to account for the ACC Case of the subordinate subject. Normally, however, this Case is assigned directly by the verb to its complement (regardless of its macro-role, as previously seen). One may ask why the same option is not available with raising verbs. An alternative solution to this problem could be to analyze such sentences as in (139), that is to say as Control structures in which the argument (PATIENT) of the perception verb is actually limited to the NP which, as such, receives ACC Case. In this picture, the embedded clause is a modifier of the object NP (i.e. a non-finite relative clause with a meaning equivalent to ‘Kate has seen John who was leaving’; see § 4.4.2. on relative clauses):50 (139)[IPKate [I’has [VPtKate[V’seen [NPJim[CP[IPPRO [I’[VPtPRO [V’leave]]]]]]]]]]. [+ACC] The subject of the infinitive subordinate clause is thus a PRO. Having no phonological realization, the PRO does not need to be licensed for Case (see Chomsky 1986a, Haegeman 1991). As a conclusion, we propose that the second argument of a verb of perception can be realized as a CP (when the verb is finite, as in [140a]), or an NP (as in [140b]):

Clauses as nouns (140)

83

a. Kate has seen [CP that Jim left]. b. Kate has seen [NP Jim [PRO leave]].

The choice of the phrase used to function syntactically as a direct object will again depend on foreground considerations (see note 17 to Chapter 2): In (140a), the departure of Jim is viewed as a global event, whereas in (140b) Jim is foregrounded as the object of seeing, and what is said about him is that he left.

2.5. Clauses as nouns So far, we have seen what happens when a clause occupies an argument position in English and Italian. Indeed, this option is universal, as recursivity is a property of natural languages. Furthermore, we expect that the semantic classes of verbs which admit a clausal argument are the same cross-linguistically, although their surface syntactic realization may vary depending on the morpho-syntactic and lexical characteristics of each language. This prediction is confirmed by facts. There are languages where a finite clause (i.e. whose verb has the same formal properties as in a root clause, including tense, mood and subject agreement features) may not occupy an argument position. This happens, for instance, in Afar and Turkish. In these languages, the verb of a subordinate clause must appear in a nominalized form. Let us consider the following data from Afar: (141)

Yo-k marahenna amateyo berra. (AFAR) O.PRO.1SG-to NEG.say.PERF.3SGF come tomorrow ‘She didn’t tell me she is coming tomorrow.’

(142)

Urrí diiné-m ubléh. kids sleep.PST.3SGM-NOMIN see.PST.1SG ‘I saw that the kids were sleeping.’

It can be noted that subordinate verbs bear no information regarding the tense (which is interpreted according to the root tense [142] or using adverbials [141]), but only regarding the person. The same happens in various languages for participial forms. In particular, the subordinate verb in (141) is considered a basic form by native speakers. On the other hand,

84

The structure of the verb phrase

the verb depending on a perception verb (142) bears a nominalizing suffix. Interestingly, the same suffix is found on the predicates of relative clauses: (143)

[[Mahammad Ali yuble-m] aaxéri-h] Mahammed Ali see.3SGM-NOMIN last.time-in ‘The last time Mahammed saw Ali was yesterday.’

kimáala. yesterday

Apparently, the perception verb in (142) does not select the whole clause as its argument, but only the NP urrì (‘kids’), which is in turn modified by a relative clause. This seems to confirm the hypothesis proposed for English in the previous section. The relative clause appears here to the right of the head noun, although Afar is a head-final language (see [143]). This is because the relative clause in question is an appositive one (as opposed to the restrictive relative clause in [143]), and such a clause follows its head noun in Afar (the same asymmetry is found in a number of languages; see § 4.4.2. on relative clauses). Afar data therefore show that the predicate of an argument clause has nominal (or at least not completely verbal) properties. Let us now look at some Turkish data from Kornfilt (1993). When used as arguments, Turkish subordinate clauses normally bear a nominalizing suffix which varies according to the event type. If the event is completed, the suffix is traditionally called factive and assumes the form dIk- (with a variable internal vowel, depending on vowel harmony); with all other events, the action suffix -mA- is used. In both cases, the subordinate clause bears an ACC Case marker, just like any direct object NP. Moreover, the subordinate verb is not marked for tense but is marked for person, although the subordinate subject is marked for GEN (as opposed to NOM) Case: (144)

[Ahmet-in öl-düg-ün]-ü Ahmet-GEN die-NOMIN-3SG-ACC ‘I heard that Ahmet is dead.’ [lit.: ‘I heard of Ahmet’s dying.’]

duy-du-m. (TURKISH) hear-PST-1SG

(145)

[Ahmet-in öl-me-sin]-i iste-me-z-di-m. Ahmet-GEN die-NOMIN-3SG-ACC want-NEG-AOR-PST-1SG ‘I would never have wanted Ahmet to die.’ [lit.: ‘I would never have wanted Ahmet’s dying.’]

Clauses as nouns

85

On the other hand, Turkish verbs of perception and opinion behave exactly like Italian and English verbs of perception, namely, the subordinate subject bears ACC Case, and the subordinate verb has no agreement but still shows tense markers: (146)

Ahmet [sen-i [yer-de yat-ar]] Ahmet PRO.2SG-ACC floor-LOC lie-AOR ‘Ahmet found you lying on the floor.’ [lit.: ‘Ahmet found you laid on the floor.’]

bul-du. find-PST.3SG

(147)

Herkes [sen-i [sinema-ya git-ti]] san-yor. PRO.2SG-ACC cinema-DAT go-PST believe-PROG.3PL all ‘Everybody believes that you went to the cinema.’ [lit.: ‘Everybody believes you gone to the cinema.’]

Finally, the verb of subordinate clauses depending on verbs of volition appears in the optative mood, with subject-related inflection but no tense markers: (148)

Herkes [yarin ben-im-le sinema-ya all tomorrow PRO.1SG-GEN-with cinema-DAT gel-esin] isti-yor. come-2SG.OPT want-PROG.3PL ‘Everybody wants you to come to the cinema with me tomorrow.’

In conclusion, Turkish embedded verbs do not display all the properties of a root verb. Furthermore, clausal subordination actually often turns out to be nominal subordination: The subordinate verb appears within a relative clause that modifies an object NP marked with ACC Case. There are, in fact, languages in which clausal arguments are invariably realized as relative clauses. Somali is one of these: All subordinate clauses may be described as relative clauses, although their noun head is not always realized. In most cases, however, all these clauses can easily be traced back to nominal constructions. Let us look at some examples of a relative clause occupying an argument position (the first two examples are proverbs from Saeed 1999: 211): (149)

Nin daad qaaday xumbo cuskay. (SOMALI) man wave take.PST.3SGM foam grasp.PST.3SGM ‘A man who is taken by a flood grasps at foam.’

86

The structure of the verb phrase

(150)

Timir aan laf lahayn horaa loo REFL.in date not seed have.NEG still.FM waayay. be missing.PST.3SGM ‘Dates which don’t have stones still haven’t been found.’

(151)

Naag-tii aad la hadashay woman-DET.AN SCL.2SG with talk.PST.3SGF walas-kaa. cousin-POSS.1.SG ‘The woman you talked to is my cousin.’

(152)

Waan aqaannaa mees-hii DECL.SCL.1PL know.1PL place-DET.AN soocdeen. travel.PRES.3PL ‘They know the place where they went.’

waa DECL

ay

u SCL.3PL at

In both (149) and (151), the relative clause is the argument functioning as the subject of the root verb. In (149), however, nin is the object of the relative clause daad qaaday, whereas in (151) naag is an indirect object governed by the preposition la. Similarly, in both (150) and (152), the relative clause is the argument functioning as the object of the root verb. In (150), however, timir is the subject of the relative clause, whereas in (152) meeshii is an indirect object governed by the preposition u. Let us now see sentences whose verbs require one of their arguments to be a clause, such as rejee ‘hope’, melee ‘think’ e xun ‘be sorry’: (153)

(154)

Waxaan rajeynayaa in-aad xashiyo ii (i-u) FM.SCL.1SG hope.PRES.1SG that-SCL.2SG papers OCL.1SG-to soo qori doonto. here write want.FUT.2SG.DEP ‘I hope you will write to me.’ Waxaan FM.SCL.1SG

u maleynayaa in-uu Xasan to think.PRES.1SG that-SCL.3SGM Xasan

doono. want.FUT.3SG.DEP ‘I think Xasan will come.’

iman come

Clauses as nouns (155)

87

In-uu Axmed dhoofo waan ka that-SCL.3SGM Axmed leave.3SG.DEP DECL.SCL.1SG from xumahay. be sorry.PRES.1SG ‘I’m sorry that Axmed is leaving.’

In all these examples, the subordinate clause functioning as an argument of the root verb is introduced by in, originally a noun meaning ‘part’ in Somali. This noun may still be used in its original meaning,51 but is also used to introduce argument subordinate clauses, whose structure perfectly mirrors that of relative clauses (see Antinucci, 1981: 243). In is no longer perceived as a noun by native speakers, but rather as a complementizer. Recent phonological data (Frascarelli & Puglielli 2009), however, show that this element is still pronounced with a primary stress. This means that phonologically in is a nominal head, and not a complementizer, as no complementizer bears primary stress in the languages that we have analyzed. Predictably, clauses introduced by in can also assume subject function when occurring with mono-argument predicates: (156)

In-uu yimaadaa waa that-SCL.3SGM come.PRES.PROG.3SG.NOM DECL ‘It’s true that he will come.’

(157)

In-aad timaaddaa waa that-SCL.2SG come.PRES.PROG.2SG.NOM DECL tahay. be.PRES.3SGF ‘It’s nice that you will come.’

run. truth

fiican nice

Relative clause properties are also found in Somali adverbial clauses which are adjuncts. Let us consider two examples for the moment (we will resume this topic in Chapter 5): (158)

Goor-ta qorrax-du dhacdo time-DET sun-DET.NOM set.DEP ‘Come when the sun sets.’

imaw. come.IMP.2SG

88

The structure of the verb phrase

(159)

In-t-aad shaqeenayso part-DET-SCL.2SG work.PRES.PROG.DEP akhrinayaa. read.PRES.PROG.1SG ‘While you are working, I’ll read a book.’

buug book

baan FM.SCL.1SG

To conclude this section, let us sum up the main points discussed in this chapter. Our analysis, as we have already stressed, is aimed at revealing universal aspects of language and, at the same time, language-specific characteristics and behavior. Both approaches are needed to understand how the various levels of language organization interact. We have seen that the semantic (argument-based) structure of the VP, determined by the meaning of the verb, interfaces with contraints imposed by syntax. A large number of argument roles corresponds to a limited number of macro-roles which assume a position within the VP and, consequently, a syntactic function in the sentence. Hierarchical relationships apply between macro-roles, which result in constraints regarding cooccurrence. In particular, the VP can contain three arguments at most, one of which (the highest in the hierarchy; see § 2.2.5.) needs to be projected as a syntactic subject. These, in our view, are the universal characteristics of the VP. The various ways in which macro-roles are overtly realized in typologically different languages, laid out in this Chapter, will be thoroughly discussed and illustrated in Chapter 3.

3. Syntactic functions

3.1. The relationship between deep roles and surface structure In Chapter 1 we proposed a syntactic approach to the traditional distinction among word classes, based on theoretical considerations and examples from different languages. In our assumption, lexical elements are assigned to individual categories within a syntagmatic environment, i.e. in the computational component of syntax. Lexical Insertion therefore contributes to the distinction between lexical heads (nouns, verbs, adjectives, adverbs) and functional heads (prepositions, determiners, complementizers and inflectional elements). Furthermore, it is responsible for the assignment of deep Cases and syntactic functions of phrases within sentences. For instance, if a noun head is merged into the VP, it will project a phrase with a specific argument role in the event described by the predicate. Conversely, if a phrase is merged outside the VP (in an adjunct position; see Chapter 1) or into an NP, it will assume a modifier function. Finally, if a noun head is merged in place of a verb head within a reduced sentential structure (as in copular sentences; see note 1 to Chapter 2), it will function as a predicate. Under our assumption, deep structure is universal. A logical consequence is that cross-linguistic differences depend on the surface realization of arguments. In other words, we assume that cross-linguistic variation basically consists of the formal realization of syntactic and pragmatic functions, and ultimately in the operations taking place at two interfaces: The morphology-syntax interface (IP, see § 1.2.3.2.) and the discourse grammar interface (CP; see § 1.2.3.3.). The former operations are analyzed in this chapter. Syntactic functions require the (inter)action of different grammatical means for their realization across languages. In particular, three main factors of variation may be identified: a) b)

Order of constituents; use of bound functional heads (i.e. affixes, bound morphemes and incorporating elements);

90

Syntactic functions

c)

use of free functional heads (e.g. prepositions, postpositions and particles).

There is a preference for distinct language types (see § 1.4.) to use one of these means (but seldom exclusively). Constituent order and free functional heads, for example, are typically found in isolating languages; inflectional morphemes and affixes characterize fusive and agglutinative languages, while incorporation processes are most often used in polysynthetic languages. Before we explore how these means interact in different languages, let us briefly consider the correlation between deep structure (extensively discussed in Chapter 2) and syntactic functions. Just as Case assignment (see below) is a surface phenomenon, some argument roles are closely related to specific syntactic functions; in other cases, the assignment of syntactic functions mainly depends on structural factors.

3.2. Deep and structural Cases Since Rouveret & Vergnaud (1980) and Chomsky (1981), the notion of Case has had a decisive role in the analysis of the distribution of nominal constituents and their morpho-syntactic characteristics. Within Generative Grammar, each nominal constituent needs to be marked for Case (i.e. has to be assigned a syntactic function in surface structure). This marking, however, may not be overtly realized. The idea of abstract Case was therefore introduced as a Universal Grammar principle to account for this requisite (initially formulated as Case Filter, see Chomsky 1981), in direct opposition to morphological Case, considered to be a parameter of languages. Research has been aimed at the analysis of the structural properties determining abstract Case assignment, which was defined in terms of government in GB and as feature checking in the Minimalist Program; see Chomsky (1981, 1995) and Haegeman (1991). While this perspective is strictly syntactic, a relationship between surface Cases and deep roles has always been acknowledged (see Fillmore 1968, 1970) and studied, leading to various solutions. Chomsky (1986a), for instance, proposed a direct link between Case and θ-roles through the Visibility Condition, where the logical order of the relationship was, as it were, reversed. He wrote that “Case makes an argument ‘visible’ for Theta role assignment” (p. 94) and is, as such, a pre-requisite for deep role assignment. This perspective presented obvious problems, e.g. in the case

Deep and structural Cases

91

of expletives (see below, § 3.2.2.) and was later abandoned. However, it opened a fruitful debate which led to a finer analysis of how syntactic functions are realized and relate to deep roles.1 This topic is crucial to our analysis; we will therefore summarize our discussion of deep structure from Chapter 2 here, and then move on to examine the realization of Case in surface structure. As we noted, the notion of ș-role (or argument role) is a strictly semantic one. A verb selects a certain number of arguments as they are part of the lexical information related to the semantic root included in the verb. A description of these arguments can be a very complex task as it involves extremely fine distinctions. Their impact on the structural level, however, shows that syntax is not sensitive to micro-semantic factors, but to broader distinctions, compatible with the number of positions admitted by the X' structure.2 Distinct argument roles are thus grouped into macro-roles (see section 2.2.1) as they share uniform or similar features and are merged into the same phrasal position. Based on structural conditions, these argument groups are assigned a particular deep Case, which defines how arguments are realized at a structural level, i.e. the final outcome of the syntaxsemantics interface in the VP (what Chomsky [1981] called Grammatical Function-ș). These deep Cases are then projected into surface structure as grammatical functions assigned on the basis of syntactic structure once the IP area is completed (§ 1.2.3.2). Case assignment therefore depends on where a phrase is located within sentential structure: Spec,IP (the subject function, SUBJ), complement of V (direct object, DO) or complement of P (indirect object, IO). These grammatical functions correspond to structural Cases proper and they can be associated with specific morphological markers in different languages: NOM Case for the subject function, ACC Case for the direct object function, DAT Case for the indirect object function and so on. We will examine the formal properties of Cases at a later stage. In the following section, we would like to show that there is no necessary correlation between deep Cases and structural Cases, although preferential associations do exist.

3.2.1. Correlations and tendencies It is well known that the grammatical function (GF in [1] below) of subject (NOM Case) is preferentially associated with the ACTOR macro-role (MR in the examples), whereas the grammatical function of direct object (ACC

92

Syntactic functions

Case) is related to the PATIENT macro-role. If the argument grid of a verb includes two arguments belonging to those macro-roles, their structural realization entails an (almost) automatic association with the abovementioned grammatical functions: (1) a. read: , Mark is reading {MR: ACT; GF: SUBJ}

Æ a book. {MR: PAT; GF: DO}

b. give: , , Æ Mark gives a book to Jim. {MR: ACT; GF: SUBJ}

{MR: PAT; GF: DO} {MR: GOAL; GF: IO}

c. put: , , Mark puts the book {MR: ACT; GF: SUBJ}

Æ on the table.

{MR: PAT; GF: DO} {MR: LOC; GF: OBL}

On the other hand, if a verb does not select the ș-role, the grammatical function of subject can be associated with different ș-roles. For example, in impersonal constructions this function is assumed by a semantically empty element (i.e. expletives), whose sole function is to realize the syntactic subject of the sentence3 (es in [3a]); in passive (2b) and middle (3b) constructions, this function is assigned to the argument that realizes the PATIENT/THEME macro-role. Let us consider the following examples from German (Bayer 2001): (2)

a. Oswald hat den Präsidenten Oswald.NOM have.PRES.3SG DET.ACC president.ACC ermordet. (GERMAN) kill.PRT ‘Oswald {ACT-SUBJ} has killed the President {PAT-DO}.’ b. Der Präsident wurde ermordet. DET.NOM president.NOM AUX.PASS.PST.3SG kill.PRT ‘The President {PAT-SUBJ} has been killed.’

(3)

a. Es EXPL.NOM

ist leicht, be.PRES.3SG easy

diesen DEM.ACC

fahren. drive ‘pro {SUBJ}is easy to drive this car {PAT-DO}.’

Wagen car.ACC

zu to

Deep and structural Cases b. Dieser Wagen fährt sich DEM.NOM car.NOM drive.PRES.3SG MIDD ‘This car {TH-SUBJ} drives easily.’

93

leicht. easy

What can be generalized from these data is that the grammatical function of subject is associated with an ACTOR whenever possible; in the absence of an ACTOR, another argument which occupies a lower position in the argument hierarchy can assume that function. Of course, the association of the ACTOR macro-role with the highest syntactic function in the structure is also sensitive to other semantic and pragmatic factors as can be clearly seen in languages like Tagalog. The internal structure of Tagalog verbs shows morphological variation depending on which argument has the pragmatic role of foreground, i.e. is treated by the speaker as the viewpoint for the narration of the event. Hence, a transitive verb like bumasa (‘read’) presents the alternatives in (4a-b) according to whether the foregrounded constituent (known as trigger and preceded by the marker ang; see Schachter 1990) has the ACTOR or 4 PATIENT macro-role: (4)

a. B-um-abasa ang titser ng diyaryo. (TAGALOG) ACT-read.PST TRIG teacher DIR newspaper ‘The teacher has read the/a newspaper.’ b. B-in-abasa ng titser ang diyaryo. PAT-read.PST DIR teacher TRIG newspaper ‘The/A teacher has read the newspaper.’

Considering this property, Tagalog causative constructions form another extremely interesting case. As it is known, two constituents with ACTOR macro-role may be present in causative constructions, one of them being selected by the semantically full verb (the ), the other (the ) by the verbal head (a light verb or morpheme) that determines the causative interpretation. Clearly, only one of the two arguments may occupy the highest syntactic function in the VP. It is therefore interesting to notice that in Tagalog the is the only constituent that may assume the syntactic role of trigger and consequently the pragmatic function of foreground. The is simply preceded by a particle (ng), indicating that its deep Case is that of a direct argument (DIR). Let us compare the following two sentences:5

94 (5)

Syntactic functions a. Mag-aalis ang tindero ng bigas sa sako ACT-extract.FUT TRIG seller DIR rice OBL sack para sa babae. OBL woman for ‘The seller will extract the rice from the sack for the woman.’ b. Ipag-aalis ng tindero ng bigas ACT.CAUS-extract.FUT DIR seller DIR rice sa sako ang babae. OBL sack TRIG woman ‘The woman had the seller extract the rice from the sack.’

Other languages, such as Quechua, present causative constructions in which the structural realization of the produces interesting semantic nuances. Apart from an ACC Case marker, this constituent can be assigned DIRECT Case, thus being associated with the ș-role (GOAL macro-role) and avoiding the coercitive reading implied by the causative verb. This can be observed in the following sentences (from van de Kerke 1996):6 (6)

a. Warmiy-ta aqha-ta llami-chi-saq. (QUECHUA) woman-ACC chicha-ACC taste-CAUS-1SG.FUT ‘I will make my wife taste the chicha.’ (coercitive) b. Warmiy-man aqha-ta llami-chi-saq. woman-DIREC chicha-ACC taste-CAUS-1SG.FUT ‘I will let my wife taste the chicha.’ (concessive)

Let us now consider how the syntactic function of DO is realized. Although this syntactic function is normally associated with a PATIENT, we saw earlier that the macro-role in question may also be realized as the sentential subject (see [2b]-[3b]). Moreover, other ș-roles can be associated with the grammatical function of DO, such as (part of the GOAL macro-role) in ditransitive constructions in languages like English or German: (7)

a. …when he taught me the rhyme. b. …als er mich den Reim when 3SGM.NOM PRO.1SG.ACC DET.ACC rhyme.ACC lehrte. (GERMAN) teach.PST.3SG ‘…when he taught me {GOAL-DO} the rhyme {PATIENT-DO}.’

Deep and structural Cases

95

The possibility for a sentence to include two constituents with distinct șroles but the same Case marker clearly shows that the realization of syntactic functions is a purely structural issue. In languages like English and German, the realization of the GOAL within the VP in ditransitive structures creates suitable structural conditions for that constituent to be assigned ACC Case.7 As Case assignment solely depends on structural factors, the same Case marker may be assigned to arguments which are realized within the VP, and to adjuncts. Let us consider the realization of DAT Case in the German verb stehlen (‘steal’) in the following sentences (from Bayer 2001): (8)

a. Er PRO.3SGM.NOM

einen

stiehlt steal.PRES.3SG

dem DET.DAT

Juwelier jeweler.DAT (GERMAN)

Ring. ring.ACC ‘He steals a ring from the jeweler.’ b. Er stiehlt seiner Freundin PRO.3SGM.NOM steal.PRES.3SG POSS.3.SG.DAT girlfriend.DAT einen Ring. DET.ACC ring.ACC ‘He steals a ring for his girlfriend.’ or ‘He steals a ring from his girlfriend.’ DET.ACC

In (8), DAT Case is assigned to both the constituent associated with the ș- role (8a)8 and the constituent serving as the of the action (8b). The latter sentence is, however, ambiguous. The first interpretation is preferred based on world knowledge; nevertheless, if the sentence is followed by a clause like um zu Geld zu kommen (‘to earn money from it’), the second reading becomes more immediate. These data might lead to the conclusion that both arguments are generated in the VP, where they receive a deep Case, morphologically expressed as a DAT. This conclusion, however, turns out to be wrong. As a matter of fact, if both constituents appear in the sentence, one of them must be realized as a PP. Let us consider sentences (9a) and (9b) along with (9c), which is ungrammatical although the DAT marker is retained:

96 (9)

Syntactic functions a. Er

stiehlt seiner Freundin steal.PRES.3SG POSS.3.SG.DAT girlfriend.DAT beim Juwelier einen Ring. jeweler.DAT DET.ACC ring.ACC at.DET.DAT ‘He steals a ring for his girlfriend from the jeweler.’ b. Er stiehlt dem Juwelier PRO.3SGM.NOM steal.PRES.3SG DET.DAT jeweler.DAT einen Ring für seine Freundin. DET.ACC ring.ACC for POSS.3.SG.ACC girlfriend.ACC ‘He steals a ring for his girlfriend from the jeweler.’ c. *Er stiehlt seiner Freundin dem Juwelier einen Ring. PRO.3SGM.NOM

As we can see, the simultaneous presence of the two constituents requires a preposition for one of them (bei or für). This means that these constituents cannot be generated at the same syntactic level and are consequently both not part of the VP. We can conclude that the verb steal corresponds to two distinct lexical entries (see § 2.2.3.) in which either the (as a GOAL) or the (as a LOCATIVE) is merged in Compl,VP and is consequently assigned DAT Case. The other is an adjunct. This is an important conclusion which leads us to a number of considerations. First, while it is possible that two constituents with different macro-roles are associated with the same Case marker (as in ditransitive structures, cf. [7]), it can be excluded that two constituents with different macro-roles merged in the same structural position in the VP (see [79] in Chapter 2) are both generated as arguments of the verb. This is expected as a direct consequence of the semantics-syntax interface: A specific position, and therefore a specific deep Case, may be assigned to a single constituent. The possibility for an adjunct with LOCATIVE macro-role to receive DAT Case in the absence of a GOAL (as in [8a]) must be considered a purely structural issue. Evidently, an economy principle applies whereby a constituent can be realized overtly through morphological Case alone if this option has not been applied earlier.9 We can assume that, if argument positions are left empty in deep structure, the system includes the possibility to apply to other constituents the formal properties of missing arguments. This is the case of the in causative constructions (see note 5 to Chapter 3), or the GOAL with verbs like kiss: As was extensively discussed in Chapter 2 (§ 2.2.3.), in a sentence like I kissed Jim, the (Jim) is realized as a DO (since the PATIENT is lexicalized in

Deep and structural Cases

97

the verb) and behaves accordingly from a syntactic point of view. For example, it can be the subject of a passive sentence. Finally, in some languages morphological Case and prepositions can be combined; we may therefore assume that the role of prepositions is to make finer distinctions between constituents with shared macro-roles. In other words, it could be assumed that arguments associated with the GOAL macro-role are assigned morphological DAT Case: If more constituents belonging to that group are present in a sentence, the simple morphological Case would then correspond to the deep Case, whereas prepositions would be used to make specific distinctions between the remaining semantic roles (realized as adjuncts). In this chapter, data from different languages will be examined which seem to confirm this assumption. Let us go back to the correlation between deep Cases and grammatical functions. We have seen that DAT Case appears to be strongly related to the GOAL macro-role in a language like German. This correlation resists syntactic transformations such as passivization. Let us consider examples (10) and (11), taken from Wunderlich & Lakämper (2001) and Bayer (2001) respectively: (10) a. … als Peter dem Touristen when Peter.NOM DET.DAT tourist.DAT den Dom zeigte. DET.ACC cathedral.DAT show.PST.3SG ‘… when Peter showed the cathedral {PAT-DO} to the tourist {GOAL-IO}.’ b. …als dem Touristen der Dom when DET.DAT tourist.DAT DET.NOM cathedral.NOM gezeigt wurde. show.PRT AUX.PASS.PST.3SG ‘…when the cathedral {PAT-SUBJ} was shown to the tourist {GOAL-IO}.’ (11) a. Oswald hat dem Präsidenten Oswald.NOM have.PRES.3SG DET.DAT president.DAT gehuldigt. salute.PRT ‘Oswald {ACT-SUBJ} has saluted the president (GOAL-IO}.’

98

Syntactic functions b. Dem Präsidenten /*der Präsident DET.DAT president.DAT DET.NOM president.NOM wurde gehuldigt. AUX.PASS.PST.3SG salute.PRT ‘The president {GOAL-IO} was saluted.’

The realization of (deep) DAT Case is in no way disrupted or modified by the different morpho-syntactic (and interpretive) conditions between active and passive sentences.10

3.2.2. When structural function is not associated with deep Case: Expletive subjects As was seen in example (3a), the grammatical function of subject can be realized by a constituent which is not part of the argument structure of the verb. In this case, the subject of a sentence is merged directly in its dedicated syntactic position (Spec,IP) and is not associated with any deep structure position. These semantically empty (or dummy) pronouns are called expletives. The need to realize subjects devoid of semantic roles is a direct consequence of a language universal, i.e. the requirement for a subject to be realized in every sentence (see note 19 to Chapter 1), formulated as the Extended Projection Principle (EPP) in Generative Grammar. Hence, when no argument is projected into Spec,IP in surface structure, an expletive fills that gap playing a purely syntactic function. This typically happens with monovalent verbs whose sole ș-role is a shadow argument (see § 2.2.2.), e.g. weather verbs (see[12]), monovalent verbs that select a THEME realized as a clausal structure (as in [13]), impersonal structures (14) and presentative structures (15): (12) a. It is raining. b. Il pleut. c. proEXPL piove.

(FRENCH) (ITALIAN)

(13) a. It seems that you are wrong. b. Il semble que vous avez tort. c. proEXPL sembra che abbiate torto.

(FRENCH) (ITALIAN)

Deep and structural Cases (14) a. It’s not true. b. Ce n’est pas vrai. c. proEXPL non è vero.

(FRENCH) (ITALIAN)

(15) a. There is a man in the garden. b. Il y a un homme dans le jardin. c. C’è un uomo nel giardino.

(FRENCH) (ITALIAN)

99

Expletive subjects can assume different forms cross-linguistically (it and there in English, il and ce in French) according to sentential structure,11 whereas they are silent in Null Subject languages such as Italian. In the latter case, as shown in examples (12c) to (15c), the expletive subject is indicated with the same symbol used for null subjects in finite sentences (‘pro’), with the notation EXPL (as a subscript). Apart from argument structure considerations (see above), the fact that expletive subjects are semantically null is also shown by structural phenomena: When a different constituent is realized in Spec,IP, the expletive is completely deleted with no consequences for sentence interpretation. Raising structures are a case in point. If the THEME of a verb like seem is realized as an infinitival clause, the embedded subject is raised to root position,12 thus avoiding the expletive: (16) a. It seems that Leo is a genius. b. *It seems Leo to be a genius. (17) Leo seems

to be a genius.

The presence of expletive elements is thus immediate evidence for the need to assume two distinct levels of analysis: A deep level, where syntax interfaces with lexical semantics (and deep Cases are realized), and a surface level, where syntactic functions are expressed on the basis of purely structural requirements and conditions. Consequently, not only are syntactic functions not necessarily related to any specific deep Case, but they may even lack one altogether. In some languages, however, the correlation between deep roles and surface Case markers is a two-way relationship. We will deal with these languages in the next section.

100 Syntactic functions 3.3. Deep roles in surface structure: Ergative languages In ergative languages (also ergative-absolutive, see Dixon 1980, 1994, Wierzbicka 1981, Manning 1996), the macro-role a constituent belongs to is crucial in determining its Case marking. Dixon (1994: 1) defines ergativity as follows: The term ergativity is, in its most generally accepted sense, used to describe a grammatical pattern in which the subject of an intransitive clause is treated in the same way as the object of a transitive clause, and differently from a transitive subject […] ‘ergative’ is the Case marking transitive subject, contrasting with […] ‘absolutive’ marking intransitive subject and transitive object.

The basic characteristic of an ergative language is therefore a formal distinction (mainly morphological, but with important consequences in syntax) within the subject function, depending on the deep role of the subject. If the ACTOR of a transitive structure assumes subject function, it takes ERG Case, whereas the subject of an intransitive verb is marked with absolutive Case (ABS), i.e. takes the same Case marker as the object of a transitive verb (belonging to the PATIENT macro-role). The following examples from Avar (Comrie 1981) illustrate this: (18) a. Vas-ass jas j-ecc-ula. SGF.ABS-praise-PRES boy-ERG girl.ABS ‘The boy is praising the girl.’ b. Vas v-eker-ula. SGM.ABS-run-PRES boy.ABS ‘The boy is running.’

(AVAR)

In (18a), the NP vas (‘boy’), interpreted as the sentence subject, is marked with ERG Case which signals its ACTOR macro-role. On the other hand in (18b), although vas is interpreted as the subject of run, it is marked with ABS Case, i.e. the same Case that marks jas (‘girl’), the PATIENT of the verb praise in (18a). In other words, while in nominative-accusative languages (such as English) the subject function is assigned on a purely structural basis, and consequently NOM Case marks every argument that satisfies the relevant licensing requirements, in ergative languages the ERG Case marker is

Deep roles in surface structure: Ergative languages 101 associated exclusively with the ACTOR macro-role. Similarly, ABS Case superficially marks the deep Case of PATIENT and THEME (both macro-roles being generated in the same position in the VP), regardless of whether they are projected as a subject or an object in sentential structure. It can be concluded that Case marking in these languages is determined by deep Cases (assigned in the VP) rather than grammatical functions.13 Although ergativity is not found in Indo-European languages (apart from a few languages of the Anatolian and Armenian subgroups), about one quarter of the world’s languages show ergative properties: Caucasian languages, Basque, many Austronesian languages, most Australian Aboriginal languages and Amerindian languages of the Eskimo-Aleutine family (see Crystal 1992). As noted earlier, the ergative character of a language has numerous syntactic consequences since the concept of grammatical subject (with its related phenomena) is no longer relevant. Let us consider, for instance, coordinate structures in Dyirbal (from Manning 1996):14 (19) Ba-yi burrbula ba-ngul gubi-ngu DET-ABS burrbula.ABS DET-ERG shaman-ERG (pro) baji-gu. fall-PST ‘The shamani hit Burrbulak and (pro*i/k) fell.’

bara-n hit-PST (DYIRBAL)

The second clause in the sentence (where coordination is expressed by means of juxtaposition) contains a null pronoun (pro), whose antecedent is necessarily burrbula, i.e. the PATIENT of the preceding clause. The opposite would be true in a nominative-accusative language. This is because the interpretation of the null subject in the second conjoint does not depend on the constituent with subject function in the first, but on the constituent whose deep Case is compatible with the argument selection of the coordinated verb.15 As a consequence, since fall selects a THEME as its first (and sole) participant, the ERG Case of shaman rules it out as an antecedent. Deep structure also determines the interpretation of null pronouns in object positions. In (22), for instance, the PATIENT of hit may only be coreferent with the PATIENT of the preceding clause, i.e. yara (‘the man’), marked with ABS Case:

102 Syntactic functions (20) Ba-yi yara ba-ngul gubi-ngu munda-n DET-ABS man.ABS DET-ERG shaman-ERG take-PST ba-ngun jugumbi-ru balga-n (pro). DET-ERG woman-ERG hit-PST ‘The shamani took the mank here and the woman hit (him*i/k).’ It is therefore clear that macro-role information is mostly preserved upon assignment of morphological Case and in coreference relationships in ergative languages.16 The relationship between deep roles and surface structure also emerges in antipassive constructions. This term is used to refer to a construction that is often related to the passive in nominative-accusative languages, although their properties are radically different. In the passive, the PATIENT assumes subject function. For this to be possible, upon lexical insertion the syntax-semantics interface needs to rule out the realization of an ACTOR within the VP. As we know, while there is no necessary correlation with deep structure, there are specific priorities in the assignment of grammatical functions (see § 3.2.1.): If a verb realizes both an ACTOR and a PATIENT in the VP, the former obligatorily receives subject function. Consequently, even though an is clearly implied in a passive event, it is not assigned a deep Case in the VP and is optionally resumed as an adjoined constituent. In a language where the grammatical function of subject is expressed differently (i.e. is not related to the Spec,IP position) and surface Cases are linked to deep Cases, it is impossible to assume the same type of operation: The of an intransitive verb cannot possibly be marked for ERG Case. Hence, something different occurs in the antipassive structure, as can be seen in the following sentences (again from Manning 1996): (21) Ba-la yugu ba-ngul DET-ABS tree.ABS DET-ERG ba-ngu barri-ngu. DET-INSTR axe-INSTR ‘The man cut the tree with the axe.’

yara-ngu man-ERG

(22) Ba-yi yara gunbal-na-nyu DET-ABS man.ABS cut-ANTIPASS-PST yugu-gu ba-ngu barri-ngu. tree-DAT DET-INSTR axe-INSTR ‘The man cut the tree with the axe.’

gunba-n cut-PST

ba-gu DET-DAT

Deep roles in surface structure: Ergative languages 103 Unlike in the active structure (21), in the antipassive (22) the of the verb cut presents an ABS Case marker, with the object of cut bearing DAT Case. The is marked in the same way in the two constructions. As we noted earlier, this structure bears little resemblance to a passive;17 under our assumption, it is instead comparable with a causative construction. If the hypothesis is maintained that Case assignment in these languages occurs at a deep level, we need to account for the fact that the of the corresponding transitive sentence is not removed, as in the passive, but demoted and realized as an object, just as it is in causatives (see note 99 and below). Similarly, the of the predication is not promoted to subject, but becomes the GOAL of the antipassive event. The sentence in (22) could plausibly be interpreted as a construction where a silent makes the man the object of an action of cutting with respect to a tree (by means of an axe). In other words, the antipassive morpheme does not appear to signal an intransitive structure, but rather a construction that could be called an impersonal causative, whose remains unexpressed (something like they made me work in English).18 Finally, notice that the NP marked with ABS Case is not only part of the argument structure of the antipassive verb, but is also the dominant NP (called pivot in Dixon 1994 and VanValin & LaPolla 1997), as shown by coreference relations. In (23), for example, dog, being the head of the relative clause, is interpreted as the subject of bite in the antipassive structure (Yidiny language, from Dixon 1994): (23) Waguja-ngu gudaaga wawal bajaajinyum man-ERG dog.ABS watch.PRES bite.ANTIPASS bunyaa-nda. woman-DAT ‘The man looks at the dog that is biting the woman.’

(YIDINY)

So far, we have examined the meaning and formal realization of deep Case in typologically different languages. Let us now go back to the structural Case in nominative-accusative languages and observe its properties. Let us first consider the Case corresponding to the syntactic function of subject.

104 Syntactic functions 3.4. The morpho-syntactic realization of structural Cases 3.4.1. Subject As observed in § 1.5.1., in a morphologically rich language (such as inflecting, agglutinative or incorporating languages) the subject has the specific property of determining agreement with the verb. In a formal framework, this property is expressed by the Spec-head configuration in IP. The syntactic subject is therefore the element located in Spec,IP and establishes a relationship with the verbal head raised to I° to check its inflectional features.19 Resuming the argumentation introduced in Chapter 1, in a sentence such as (24) each element is assumed to be merged into an X' structure projecting a phrase. Each phrase is then combined with the others, so as to form syntactic objects that make up the deep structure of the VP: (24) Jim begins his lesson on World War Two at ten. (25)

PP P’ VP (adjunct)

N’ N° Jim

ten

V’

ACTOR

NP

NP

P° at

VP

V° begins

PATIENT

NP N’ his

PP N° lesson

on WW2

The deep structure thus determined by the interface with (lexical) semantics then needs to be licensed from a morpho-syntactic point of view. This means that nominal constituents need to be assigned Case (i.e. to be licensed for a syntactic function) and the verbal head needs to check its

The morpho-syntactic realization of structural Cases 105 person and TAM features, through the means available in that language. These morpho-syntactic requirements, needed for interpretation, make it necessary for the verb to move to I° and for the subject to move to Spec,IP (for NOM Case). The Case of the direct object (ACC Case) is assumed to be made interpretable in situ: IP (26) I’

NP Jim

VP

I° begins [+TAM]

PP at ten

VP tNP

NP tv

his lesson

As we noted in § 1.5.1., the syntactic function of subject can be fulfilled by constituents with various argument roles with language-depending restrictions. In particular in a language like English, virtually any deep role may assume subject function and determine agreement with the verb. With a verb like begin, for example, it is possible to raise the THEME NP into Spec,IP, obtaining the sentence the class begins at ten. In this case the verb qualifies as an unaccusative monovalent verb (see note 22 to Chapter 2): IP

(27)

I’

NP The class [+NOM]

VP



begins [+TAM]

PP VP

at ten

Ø tv

tNP

106 Syntactic functions The fact that a Spec-head configuration with an inflected verb in I° is needed to realize NOM Case is confirmed by infinitive constructions where an NP, while being the first participant in an event, appears with Cases other than NOM. This occurs in sentences with exceptional (ACC) Case marking (ECM), previously examined in Chapter 2 (see § 2.4., examples (132) and ff.). Let us consider the following sentences: (28) a. I heard Jim play the piano. b. Ho sentito Jim suonare il piano.

(ITALIAN)

As we know, although Jim is the ACTOR of play, the verb – in its infinitive form – does not move to I°20 and therefore does not admit overt subjects (but only null subjects, indicated as PRO; see § 2.4.). The NP Jim thus receives ACC Case from the root perception verb as its object: (29) a. I heard him play the piano. b. L’ho sentito suonare il piano.

(ITALIAN)

Another case in which the deep subject does not emerge overtly as such is in causative sentences (considered in § 3.2.1.). Let us consider the following Italian examples: (30) a. Ho fatto leggere mio fratello {DO} have.PRES.1SG make.PRT read.INF POSS.1.SG brother per tutto il giorno. DET day for all ‘I had my brother read all day.’ b. L’ ho fatto leggere per OCL.3SGM have.PRES.1SG make.PRT read.INF for tutto il giorno. all DET day ‘I had him read all day.’ (31) a. Ho fatto leggere un libro have.PRES.1SG make.PRT read.INF INDEF.DET book tutto il giorno. a mio fratello {IO} per POSS.1.SG brother for all DET day to ‘I had my brother read a book all day.’

The morpho-syntactic realization of structural Cases 107 b. Gli

ho fatto leggere un have.PRES.1SG make.PRT read.INF INDEF.DET libro per tutto il giorno. DET day book for all ‘I had him read a book all day.’ IOCL.3SGM

My brother is the ACTOR of read in both (30) and (31), yet in no way can it be considered the subject of the corresponding sentences. Again, this is due to structural reasons: In causative constructions, the semantically full verb appears in its infinitive form. As we know, the realization of the grammatical subject requires the TAM features to be activated in I°. Since the verb does not raise to I°, however, the ACTOR moving to Spec,IP cannot assume the subject function. Yet, since each phonetically realized NP needs to be marked for Case (as it needs a syntactic function), an alternative Case assignment mechanism is required. For this reason, if the PATIENT is not realized, the ACTOR is marked for ACC Case by the full verb (as in [30]). If, on the other hand, the verb assigns ACC Case to its PATIENT (as in [31]), then the ACTOR cannot but receive structural Case from a preposition.21 We can thus conclude that my brother is not a subject in either (30) or (31) as in these sentences the only syntactic subject is the that determines agreement with the light verb fare. An analysis of causative constructions thus confirms that there are restrictions imposed by the surface structure on the formal realization of deep roles.22 If the subject is the constituent that determines agreement with the verb, then in sentences like the following, from French and Berber (from Ouhalla 1993) respectively, a subject function can be excluded for une bizarre petite femme in (32) and tamgharta in (33): (32) J’avais déjà commencé à manger lorsqu’il est entré une bizarre petite femme. (FRENCH) ‘I had already begun to eat when a weird little lady came in.’ [Camus, L’étranger] (33) Tamghart-a ay yzrin woman-DET INDEF.DEM see.PRT Mohand. Mohand ‘THIS LADY saw Mohand.’

/*t-zra /3SGF-see.PST (BERBER)

108 Syntactic functions In neither case does the relevant NP determine agreement with the verbal head. In the French example, entrer shows masculine agreement with the expletive subject il; in (33) agreement is lacking altogether with the verb showing participial agreement. Consequently, although the constituents une bizarre petite femme and tamgharta are interpreted as the individuals who perform the actions come in and see, they still do not fulfill the function of subjects. The presence of an expletive in (32) also suggests that une bizarre petite femme is not generated within the VP since it would then be the first participant in the event, and as such should raise to Spec,IP thus avoiding a dummy subject. Hence, not only is une bizarre petite femme not the subject of the sentence, but it is not even an argument of the verb entrer. Similarly, the indefinite demonstrative (ay) in (33) shows that the Berber sentence is a copular construction, which could be translated literally as ‘THIS LADY (is) the one that (has) seen Mohand’. In other words, the constituents intuitively interpreted as subjects in these sentences are actually predicates of sentences in which they are interpreted as new information (i.e. as Foci; see § 6.4.). Their syntactic function, therefore does not belong to the IP area, but to the CP domain (where the interface with discourse grammar takes place; see § 1.2.3.3.). As we have seen, subject-verb agreement is crucial in defining the syntactic function of subject and is expressed differently across languages. The analysis of agreement is one of the most interesting and active fields of typological research and authors from different theoretical frameworks have made a contribution to it (see, among others, Givón 1976, Barlow & Ferguson [eds.] 1988, Bresnan & Mchombo 1986, Aissen 1990, Doron 1986, Corbett 2006). This line of research has been mainly aimed at defining how agreement occurs and where it applies. Within our theoretical framework, the application domain and direction of agreement are defined by the Spec-head syntactic configuration, which has been assumed necessary for the realization of subject function. The remainder of this section will therefore focus on the modes of agreement from a crosslinguistic perspective. In some inflecting languages, the syntactic subject is signaled in two ways: By means of a specific NOM Case marker on the nominal constituent (morphological Case) and through some form of (phi and TAM-related) agreement on the verbal head. This is what happens in Modern Greek and Russian, as shown in (34) (from Holton et al. 1997) and (35):

The morpho-syntactic realization of structural Cases 109 (34) O

Nicos kerdise Nicos.NOM.M win.PST.3SG to loto. DET.ACC.SGN lottery.ACC.SG.N ‘Nicos won the lottery.’ DET.NOM.SGM

(35) Ivan proþital Ivan.NOM.M read.PERF.3SGM ‘Ivan read the (whole) book.’

(MODERN GREEK)

knigu. book.ACC.SGF

(RUSSIAN)

As we can see, in Russian, apart from person and number agreement, the past morpheme also bears information on subject gender. If the subject is feminine, the verb assumes the following form: (36) Marija proþitala Marija.NOM.F read.PERF.3SGF ‘Maria read the (whole) book.’

knigu. book.ACC.SGF

In other inflecting languages, such as Italian and French, morphological Case is not available; therefore, the subject function is signaled on the verbal head only: (37) Jean lisait Jean read.PST.3SG ‘Jean read the book.’

le DET

livre. book

(FRENCH)

Clearly, linear order is extremely important in these languages as it makes it possible to distinguish the subject and object functions in potentially ambiguous cases (for instance, sentences such as Luke saw Mary); conversely, the unmarked order does not need to be retained in languages with overt Case markers (such as Russian). In agglutinative languages, agreement features are indicated by means of dedicated morphemes, each one of which conveys a specific piece of grammatical information. Again, the amount and type of information codified by the morphological system is language-dependent. In Turkish (from Kornfilt 1997) the subject is signaled by both morphological Case on the noun and agreement morphemes on the verb (38); in Korean (from HoMin 1999), on the other hand, the NOM Case marker on the NP is the only indicator of the subject (see [39]) as the verb only presents information

110 Syntactic functions related to tense, clause type (declarative, interrogative, etc.) and, possibly, honorific morphemes: (38) Ben

(TURKISH)

(39) Hyeng-eykey-nun nay-ka sosik-ul brother-to-TOP S.PRO.1SG-NOM news-ACC cenhay-ss-ta. tell-PST-DECL ‘Your brother, I already told him the news.’

(KOREAN)

dün sinema-ya git-ti-m. PRO.1SG.NOM yesterday cinema-DAT go-PST-1SG ‘I went to the movies yesterday.’

In other agglutinative languages, such as Swahili, subject-predicate agreement is expressed through classifiers (see note 30 to Chapter 1), as in the following example (from Vitale 1981): (40) Wa-toto wa-nakula mkate. CL2-child CL2-eat fish ‘The kids are eating fish.’

(SWAHILI)

Here, the subject is marked with Classifier-2, specified for the semantic features [+animate] and [+plural]; the same classifier needs to be used as a verb prefix. In polysynthetic languages, information on the subject function is incorporated, like every other grammatical function, into words with an extremely complex template structure (see § 1.4.). Cross-linguistic analysis shows that the subject affix normally occupies a more external position than the object (regardless of the position of TAM morphemes; see Baker 1996). Let us look at some examples from Amerindian languages, respectively Cayuga (Evans & Sasse 2002) and Ngalakan (Baker 2002): (41) ĉ-s-kakhe-hona’tá-yĊthw-ahs. FUT-ITER-SCL.1SG/OCL.3PL-potato-plant-PERF ‘I will plant potatoes for them again.’ (42) Ngu-bu-may-woh-wo. SCL.1SG-OCL.3PL-(vegetable) food-give-PST ‘I gave them vegetables.’

(CAYUGA)

(NGALAKAN)

The morpho-syntactic realization of structural Cases 111 The hierarchical superiority of the subject over the object also applies to subjects of passive sentences, as shown in example (43) from Northern Tiwa (Aissen 1990). Interestingly, the (typically an adjunct in passive constructions) is the sole constituent with no corresponding morphemes incorporated on the verbal head (the object clitic -n- referring to child rather than man): (43) I-n-?u?u-wia-che-ban seuanide-ba.(SOUTH. TIWA) SCL.1SG-OCL.3SG-child-give-PASS-PST man-by ‘I was given the child by the man.’ Morpheme incorporation in polysynthetic languages confirms that subject and object occupy distinct hierarchical positions, and is consistent with the precedence principle (Comrie 1981): According to this, the subject precedes the object in the unmarked order in 95% of the world’s languages. As for isolating languages, the interpretation of the syntactic function of the subject chiefly relies on constituent order. Again, the crucial factor is the precedence principle, since the subject NP always precedes the object in linear order. Furthermore, these languages tend to assume an unmarked SVO order, as it immediately enables identification of syntactic functions.23 SVO, however, is not the only order found in isolating languages. Thus, along with (typically) SVO creole languages such as Saramaccan (see [44], from Byrne & Caskey 1993), there are languages with mixed SVO-SOV characteristics, such as Chinese (see [45a-b], from Li & Thompson 1981), and VSO languages such as Rapanui (see [46], from De Feu 1995). To the best of our knowledge, no isolating language presents a VOS or OVS order (OSV is not attested): (44) Dí

hógimbéti nján dí sindéki dí a bi téi DET jaguar eat DET snake DET 3SGM PST take páu kíi. (SARAMACCAN) stick kill ‘The jaguar ate the snake that it had killed with a stick.’

(45) a. Wǂ

măi shnj le. buy book PERF ‘I bought a book.’ b. Wǂ bă shnj măi le.24 PRO.1SG DEF book buy PERF ‘I bought the book.’ PRO.1SG

(CHINESE)

112 Syntactic functions (46) He

tikea e au te DYN see ACT 1SG DET ‘I can see the child.’

poki.25 child

(RAPANUI)

It can be concluded that the world’s languages employ various morphosyntactic strategies to signal the syntactic function of the subject unambiguously. While extremely diverse, the strategies examined share a number of constant factors; for example, the precedence of the subject over the object, indicating a hierarchically higher position for the subject, or, in non-isolating languages, the presence of verb agreement morphemes depending on a specific structural configuration.

3.4.2. Direct object As we saw in the preceding section, the syntactic function of object is also marked in languages, although object agreement morphemes on the verbal head are less frequent. In isolating languages, the syntactic function of object is typically fulfilled by the NP following the verb or the NP following the subject in linear order (see [44]-[46] above). Conversely, in languages with morphological Case, the object is usually assigned a specific marker called ACC Case, as observed in Modern Greek (34), Russian (35) and Korean (39). Finally, in polysynthetic languages the object morpheme always appears in a more internal position (i.e. closer to the verbal head) than the subject. Along with the examples from the previous section ([41]-[43]), let us consider the case of Somali (an SOV language) in the following sentence: (47) Jamacad-da shalay buu university-DET yesterday FM.SCL.3SGM igu (i-ku) arkay. OCL.1SG-to see.PST.3SGM ‘Cali saw me yesterday at the university.’

Cali Cali.NOM (SOMALI)

In Somali, a language with mixed inflectional and polysynthetic properties (see Svolacchia & Puglielli 1999), the subject clitic pronoun (uu) precedes the object clitic (i). Moreover, object clitics must appear (along with prepositions) within a clitic cluster, forming a verbal complex with the verbal head (see Puglielli 1981). On the other hand, the realization of the subject clitic is structurally less restricted, as it admits the presence of

The morpho-syntactic realization of structural Cases 113 nominal constituents between itself and the verbal complex. Again, a syntactic hierarchy is employed which enables interpretation and consequently makes it possible to distinguish between the two main syntactic functions. As was said at the beginning of this section, the syntactic function of the object is less frequently signaled in languages through overt markers than the subject function. This asymmetry may be ascribed to an economy principle operating in the system: Once the subject is marked, the object function can be subsequently interpreted as the morphologically unmarked Case. Interestingly, while an overt object marker is not as crucial for interpretation, there are languages where this marker is specialized, i.e. is associated with other types of information and takes on additional functional values. In particular, the ACC Case marker is sensitive to a number of semantic features of the object, such as definiteness and animacy. Let us see how. In languages like Russian, for instance, the ACC Case marker only appears on the object NP if the noun head is [+animate], as shown in the following example: (48) Ja videl mal’þika/ S.PRO.1SG see.PST.1SGM boy.ACC dub-Ø/ stol-Ø. oak table ‘I saw a boy/hippo/oak/table.’

begemota/ hippo.ACC (RUSSIAN)

The ACC marker -a is present on boy and hippo, but not on nouns indicating plants or inanimate things. This distinction is clearly connected to the need to clarify syntactic roles in potentially ambiguous contexts. For example, Russian admits a certain degree of variation in constituent order depending on discourse requirements. When a bivalent verb (such as see, meet, kill etc.) selects two [+human] arguments, both may qualify as ACTOR-subjects. In such cases, morphological Case makes it possible to unambiguously identify the different functions, independent of constituent order: (49) a. Ivan videl Ivan see.PST.1SGM ‘Ivan saw a boy.’ b. Ivan mal’þika videl. c. Mal’þika Ivan videl.

mal’þika. boy.ACC (= [49a]) (= [49a])

114 Syntactic functions d. Videl mal’þika, Ivan.

(= [49a])

This ambiguity is of course absent in a sentence where the is [-animate], and thus incompatible with an θ-role; here, an overt marker is totally unnecessary (as for dub and stol in [48] above). A specific Case marker for [+human] objects may also be realized by means of prepositions in languages without morphological Case. This is what happens in Spanish, where the strategy known as prepositional accusative is observed (see Berretta 1989 for an analysis of the same phenomenon in non-standard varieties of Italian): director busca al empleado/*perro. (SPANISH) manager seek.PRES.3SG to.DET employee/dog ‘The manager is looking for the employee / dog.’

(50) El

DET

The object empleado is preceded by a preposition that need not be present with a predicate like look for (which selects a PATIENT). The preposition is in fact needed to indicate a lower syntactic function than the subject for the relevant [+human] referent.26 This, of course, is no longer necessary if the direct object is dog. Thus, in Spanish, the preposition fulfills the same distinctive function as morphological Case in Russian. Let us now look at the following example: (51) El

director busca a un empleado / manager seek.PRES.3SG to INDEF.DET employee un empleado. INDEF.DET employee ‘The manager is looking for a (specific) employee / an (any) employee.’ DET

As shown in example (51), non-prepositional [+human] ACC is also possible in Spanish. The absence of the preposition is not coincidental, but associated with a different meaning which is clarified in the translation. When an indefinite article is used, the prepositional accusative expresses the feature [+specific] of the referent. In this case, the preposition means that the manager is looking for a specific employee; if no preposition is used, the referent receives a generic interpretation. The link between overt ACC Case marking and object definiteness is a frequently found feature in languages from different families and areas. It is observed in an Indo-European language like Persian (see [52], from Comrie

The morpho-syntactic realization of structural Cases 115 1981), as well as in a Semitic language like Tigrinya (see [53], from Tesfay 2002): (52) a. Hasan ketabra Hasan book.ACC ‘Hasan saw the book.’ b. Hasan yek Hasan INDEF.DET ‘Hasan saw a book.’

did. see.PST

(PERSIAN)

ketab did. book see.PST

(53) a. Njssu "anbesa q.til-u. (TIGRINYA) S.PRO.3SGM lion kill-SCL.3SG ‘He killed a lion.’ "anabis q. til-uww-om.27 b. Njssu nj-"it-om S.PRO.3SGM to-DET.PL-OCL.3PL lions kill-SCL.3SG-OCL.3PL ‘He killed the lions.’ In Tigrinya, a definite object determines not only the presence of an agreement morpheme on the verbal head, but also clitic resumption on the preposition nj (‘to’). In other words, a definite object in Tigrinya simultaneously causes verbal agreement with the subject and the phenomenon of prepositional accusative (53b). Conversely, an indefinite object does not require either (as in [53a]).28 Similarly, in a Ugro-Finnic language like Hungarian, a [+definite] direct object requires a specific agreement morpheme on the verbal head. Let us observe the following examples: (54) a. Olvas-ok. read-SCL.1SG ‘I am reading.’ b. Olvas-om a könyv-et. DET book-ACC read-SCL.1SG.OCL.3SG ‘I am reading the book.’

(HUNGARIAN)

Finally, let us consider the case of Swahili, a language where the object is realized by means of an affix on the verbal head (in a more internal position than the subject). The affix is repeated before the object NP in order to identify the semantic class of the nominal expression in question and express the information related to number (from Moravcsik 1978):

116 Syntactic functions (55) A-li-wa-ona SCL.3SG-PST-OCL.3PL-see ‘He saw the kids.’

(SWAHILI)

watoto. PL.kid

In this language too, object resumption in the verbal head appears to be sensitive to the definiteness feature, as shown in the following two examples (from Givón 1976): (56) a. U-me-leta kitabu? SCL.2SG-PERF-bring book ‘Have you brought a book?’ b. U-me-ki-leta SCL.2SG-PERF-OCL.3SG-bring ‘Have you brought the book?’

kitabu? book

Having discussed the formal expression of the subject and object functions, let us conclude the chapter with a detailed examination of the cross-linguistic strategies for the expression of morphological Case.

3.5. Morphological Case: Cross-linguistic strategies As we have seen in the previous sections, the presence of overt markers signaling the syntactic functions of nominal expressions is a parameter across languages. In languages with morphological Case, different strategies can be identified. First of all, in some languages Case is expressed through free morphemes (particles and prepositions), while in others it is realized by means of bound morphemes (affixes). A further distinction exists in the latter category, between languages where the Case marker appears on the nominal element (“dependence-marking”, according to Nichols 1986) and languages where it is found on the verbal head (“head-marking”). Finally, some languages mark both elements in the agreement relation. The following table presents an overview:29

Morphological Case: Cross-linguistic strategies 117 Table 1. Morpho-syntactic strategies for Case realization Case marker

Place and type of realization

Zero-marking (for subject/object only)

English, French, Italian prepositions

English, Italian, German, Russian, Tagalog

particles

Chinese, Somali, Tagalog

on the dependency (NP) on the head (V)

Modern Greek, Russian, Arabic, Somali, German Ewe, Albanian

on both (NP and V)

Warlpiri, Swahili, Avar

Free morphemes30

Bound morphemes

Languages (examples)

In this section, some examples from the languages mentioned in the Table will be used to show the various strategies at work. Zero-marking – a strategy found in many of the world’s languages – is used in the two main syntactic functions, i.e. subject and direct object. In a number of languages, these constituents thus lack any overt marker (since they are direct arguments of the verb); their syntactic function can be interpreted by constituent order alone, as shown in (57a-b) for Italian and French: (57) a. Leo (SUBJ) ha salutato Maria (DO) Leo have.PRES.3SG greet.PRT Maria ma Maria (SUBJ) non ha visto Leo (DO). but Maria not have.PRES.3SG see.PRT Leo ‘Leo said hello to Maria but Maria didn’t see Leo.’ salué Marie (DO) b. Leo (SUBJ) a Leo have.PRES.3SG greet.PRT Marie mais Marie (SUBJ) n’ a pas vu Leo (DO). but Marie not have.PRES.3SG NEG see.PRT Leo ‘Leo said hello to Marie but Marie didn’t see Leo.’ Subject and object, however, are not the only arguments to admit zeromarking. In some languages, indirect objects can also appear without overt markers; this is the case, for example, in English ditransitive structures:

118 Syntactic functions (58) a. b.

Leo (NOM) gave a book (DO) to/*Ø John (IO). Leo (NOM) gave *to/Ø John (IO) a book (DO).

It can be observed that the realization of the zero-morpheme (Ø) for the IO function in English requires adjacency to the verb as a structural condition. If that condition is not satisfied, a preposition is needed (see [58a]). As in the case of causative constructions (discussed in § 3.4.1.), this alternation shows once again that structural conditions must be satisfied for Case assignment. The argument selected by some motion verbs (like go) can also be realized as a direct object in some languages, as shown in the Somali example (59), where the NP guriga does not bear any Case marking. The syntactic function of the NP thus needs to be assigned directly by the verbal head yimid within deep structure: (59) Cali shalay buu guri-ga yimid. (SOMALI) Cali yesterday FM.SCL.3SG home-DET go.PST.3SGM ‘Cali went home yesterday.’ The syntactic function of subject or object can also be fulfilled by clausal constituents. In a language without morphological Case, such as Italian, this has no consequences from a formal point of view. As expected, the same type of clause may fulfill either function with no need for specific morphological markers (see [60]). Moreover, being (prosodically) heavy, these constituents will normally appear in post-verbal position, regadless of their syntactic function (see [61a-b]).31 The interpretation of the subject function, in this case, is not ensured by constituent order, but is determined by (third-person) agreement and considerations concerning the argument selection of the verb: (60) a. [Che Leo abbia trattato male that Leo have.SUB.PRES.3SG treat.PRT bad ha stupito tutti. have.PRES.3SG shock.PRT all ‘That Leo mistreated Maria shocked everyone.’ b. Non sapevo not know.PST.1SG [che Leo avesse trattato male that Leo have.SUB.PST.3SG treat.PRT bad ‘I did not know that Leo mistreated Maria.’

Maria]SUBJ Maria (ITALIAN)

Maria]DO. Maria

Morphological Case: Cross-linguistic strategies 119 (61) a. Dispiace [che Leo venga that Leo come.SUB.PRES.3SG displease.PRES.3SG proprio oggi] SUBJ. exactly today ‘It’s a shame that Leo is coming today of all days.’ b. Ci sembra [che Leo venga to.us seem.PRES.3SG that Leo come.SUB.PRES.3SG proprio oggi] SUBJ. exactly today ‘It appears that Leo is coming today of all days.’ Let us now move on to languages where explicit Case marking is realized through free or bound morphemes. As we said earlier, such morphemes may appear on one or both of the elements interacting in the Case assignment process. Various options exist. In some languages, such as Russian and Arabic, the Case marker is an affix on the NP, as the verbal head only bears information about agreement and TAM features. Let us consider the following examples: (62) Vþera utrom v universitete Ivan university.LOC Ivan.NOM yesterday morning.INSTR in dal Maši knigu. (RUSSIAN) give.PST.M Maša.DAT book.ACC ‘Ivan gave Maša a book at the university yesterday morning.’ (63) ‘Asbaha ’aknjh u become.PERF.3SGM brother.O.PRO.3SGM.NOM mumaththil-an. actor.ACC-INDEF ‘His brother became an actor.’

(ARABIC)

As is shown, in Russian (as in Arabic) oblique Case markers also appear as free functional morphemes (e.g. preposition v in [62]). Mixed Case assignment strategies are extremely frequent in languages: Zero-marking, prepositions and bound affixes can coexist within one system and so fine distinctions may be made between syntactic functions. Prepositional Case markers appear in the position defined by the head parameter. They are thus prenominal in VO languages such as English, and (usually) postnominal in OV languages such as Japanese (see § 1.2.2.).

120 Syntactic functions In some languages, free Case markers are not actual prepositions, but particles used to distinguish direct from indirect Cases. Elements such as ng and sa in Tagalog are a case in point (see [64]). These elements signal the deep relationship between verb and argument and are not directly comparable to prepositional Case markers. This is obvious as particles and prepositions can coexist to indicate a specific type of indirect complement (as in [65]-[66]):32 (64) Nag-bigay ang tatay ko ACT.PST-give TRIG father POSS.1.SG sa pulubi. OBL beggar ‘My father gave a beggar some money.’

ng DIR

pera money (TAGALOG)

(65) Nag-handa ang titser ng sorpresa ACT.PST-prepare TRIG teacher DIR surprise para sa mga bata. for OBL PL bambino ‘The teacher prepared a surprise for the kids.’ (66) Ikalima-ng gusali buhat sa kanto ang post office. fifth-ASSOC building from OBL corner TRIG post office ‘The post office is the fifth building from the corner.’ As is well known, an NP may be modified or coordinated. In these cases, the marker of its syntactic function can be realized on all the elements that make up the complex NP. This is observed in typologically distinct languages, such as Arabic (67) for adjectival modification, Korean (68) for nominal modification (from Ho-Min 1999), and Turkish (69) for coordinated structures (from Kornfilt 1997): (67) [Al-lugha al-arabiyya] DET-language.NOM DET-Arabic.NOM ‘The Arabic language is difficult.’

sa’abah. difficult

(ARABIC)

(68) Yumi-ka kay-eykey [phal-ul oynccok-ul kluth-ul] arm-ACC left side-ACC end-ACC Yumi-NOM dog-DAT mul-li-ess-ta. (KOREAN) bite-PASS-PST-IND ‘Yumi was bitten by the dog in the lower part of his left arm.’

Morphological Case: Cross-linguistic strategies 121 (69) Ahmet [uskumru-yu te istakoz-u] piúir-di. (TURKISH) Ahmet prawn-ACC and lobster-ACC cook-PST ‘Ahmet cooked shrimp and lobster.’ On the other hand, in languages like Somali, the Case marker only appears on the rightmost constituent in a complex NP (i.e. the most deeply embedded constituent, marking the right boundary of the phrase). The first element thus appears in an unmarked form: (70) [Qalin-ka iyo buggag-gu] miis-ka pencil-DET and book.PL-DET.NOM table-DET saran yihiin. be.PRES.3PL ‘The pencil and the books are on the table.’

way DECL.SCL.3PL (SOMALI)

The association of whole clauses with subject or object function is an interesting issue for analysis, as shown by the asymmetries that can be found in languages with morphological Case marking. In some of them (e.g. German, Russian and Modern Greek) the syntactic function of a subject or object clause is never marked. Consider the following German sentences (a translation of the Italian examples [60a-b]): (71) a. [Dass Leo Maria schlecht behandelt hat]SUBJ treat.PRT have.PRES.3SG that Leo.NOM Maria.ACC bad hat alle überrascht. (GERMAN) have.PRES.3SG all.PL.ACC shock.PRT ‘That Leo mistreated Maria shocked everyone.’ b. Ich wusste nicht [dass Leo PRO.1SG.NOM know.PST.1SG not that Leo.NOM Maria schlecht behandelt hat]DO Maria.ACC bad treat.PRT have.PRES.3SG ‘I didn’t know that Leo mistreated Maria.’ As is evident, no morphological marker is suffixed to the clause constituents to mark their respective syntactic functions. This would seem to lead to the generalization that Case marking is a phenomenon typically associated with NPs. However, data show that this is not the case. There are languages in which clausal constituents bear a Case marker depending on their function, just as nominal constituents do. This is observed in Afar, as shown below (from Bliese 1981):

122 Syntactic functions (72) [Han ‘ubée-m]-i milk drink.3SG-NOMIN-NOM ‘That s/he drinks milk is good.’

negay. good

(AFAR)

(73) a. [Ged‘de-m] ob’be. go.2SG-NOMIN.ACC hear.PST.1SG ‘I heard you left.’ b. (Na’nu) [gen‘na-m] fan’na. S.PRO.1PL go.1PL-NOMIN.ACC want.1PL.PST ‘We wanted to go.’ As can be seen, a clausal subject in Afar shows a NOM Case marker (-i) after the nominalizing suffix –m (cf. § 2.5.) (see [72]). In contrast, no Case marker is present in the case of object clauses [see (73)] since ACC Case is morphologically unmarked in this language. Turkish is another language in which the verb of an object clause bears a nominalizing affix (providing the verb a quasi-nominal status) and clausal complements are marked for Case. In the following examples (from Kornfilt 1993) the various Case markers show the different functions determined by the root verb: (74) a. (Ben) [Ahmed-in öl-düg-ün]-ü PRO.1SG.NOM Ahmed-GEN die-NOMIN-3SG-ACC duy-du-m. (TURKISH) hear-PST-1SG ‘I heard that Ahmed died.’ b. (Ben) [Ahmed-in öl-düg-ün]-den PRO.1SG.NOM Ahmed-GEN die-NOMIN-3SG-ABL kork-uyor-du-m. fear-PROG-PST-1SG ‘I was afraid that Ahmed had died.’ c. Zeynep [Ahmed-in sinema-ya git-me-sin]-e Zeynep Ahmed-GEN cinema-DAT go-NOMIN-3SG-DAT çok üz-ül-dü. very sad-PASS-PST ‘Zeynep was very sad about the fact that Ahmed had gone to the movies.’ The fact that complement clauses are directly comparable to nominalized structures is evident in other Western languages too (e.g.

Morphological Case: Cross-linguistic strategies 123 English, Italian, French and German) in the case of infinitival subordinates (also known as Control structures). These clauses are introduced by prepositions (like NPs can be). Although these prepositions are analyzed as C° heads, they do not admit the insertion of constituents before the verbal head, just as a preposition cannot be separated from its complement NP: (75) a. *I hope to, Jack, see him tomorrow. b. *I hope to, tomorrow, see Jack. c. *I hope to, JACK see tomorrow. In other words, infinitive clauses do not have a left periphery (i.e. a CP area; see § 1.2.3.3.) and their verbal forms obey a structural requirement of adjacency to the governing preposition. This requirement can be feasibly compared to a form of Case assignment. Therefore it appears that a verb with no TAM features shares a number of structural properties with a noun, although it cannot be considered a full-fledged noun, because it is able to assign the syntactic function of object directly (i.e. with no need for prepositions, unlike in the case of noun head complements). These data lead us to formulate the hypothesis that clausal subordination actually originates as nominal subordination. In other words, we can assume that complement clauses are originally introduced by a generic noun head (such as ‘fact’, ‘event’, etc.) which is then deleted (on silent heads, see Kayne 2005). A sentence like (74b) might therefore be paraphrased as ‘I was afraid of the fact that Ahmed had died’. This would account for the (pseudo)nominal properties shown by these constructions in many languages. We will come back to this topic in Chapter 5, when discussing adverbial clauses. For the moment, we will conclude by presenting some data which confirm our assumption in the case of finite subordinate clauses. We shall refer to the Spanish preposition de, which precedes clausal complements of verbs like decir (‘tell’), gustar (‘like’), pensar (‘think’) and ser (‘be’) (a phenomenon known as dequeismo; see De Mello 1995, Gómez Torrego 1998). Let us consider the following examples: (76) Sólo unas ganas desperadas de que me dijeran de que no era esto lo que me imaginaba. (SPANISH) ‘Only a desperate desire that they tell me that it was not what I imagined.’ [1990, ‘Guerra Privada’ Oddone, Pancho]

124 Syntactic functions (77) De lo que sì estoy siguro es de que me conoce más gente de la que compra libros. ‘What I am sure of is that more people know me than buy my books.’ [1990, the News, ABC: Lobo Carmen; Javaloyes Inigo] Since the complementizer is an invariable (non-nominal) head, the presence of a preposition before it could not be explained if a generic, silent NP between de and que (such as ‘the fact’) is not assumed. Based on this assumption, this NP was originally the head of the object clause and was later deleted (in other words, the sentence in [77] is originally equivalent to: what I am sure of is the fact that more people know me than buy my books). Let us now consider those languages in which Case marking does not appear on the noun head, but on the verbal head as an agreement morpheme (head-marking languages). Examples (78)-(79), from Tzutujil (Nichols 1992) and Lakhota (Van Valin 1992) are given below: ch’ooyaa" mouse

(78) X-ø-kee-tij tzyaq ASP-3SG-3PL-eat clothes ‘The mouse has eaten the clothes.’ (79) Lakhota ki thathaka ota Indian DET bison many ‘The Indian killed many bison.’

wicha-Ø-kte. 3SG-3PL-kill

(TZUTUJIL)

(LAKHOTA)

In these languages subjects and objects are not marked for Case, but are interpreted according to the markers on the verbal head. In passing, it should be noted that the subject marker always precedes the object marker, which proves once more the structural superiority of the subject in the incorporation process. Nichols (1992) observes that in some languages the markers of syntactic functions are not realized as affixes on the verbal head, but cliticized on TAM-related functional elements. This is the case of the infix -n- in the Uto-Aztecan example in (80): (80) Noo xu-n-po S.PRO.1SG MOD-1SG-PRES ‘I would kick John.’

xwaani John

"ari.

(UTO-AZTECAN)

kick

Although these data will not be analyzed in detail, we can observe that they support a structural analysis consistent with Generative Grammar, as

Morphological Case: Cross-linguistic strategies 125 illustrated in Chapter 1 (§ 1.2.3.2.). If the TAM categories are analyzed as functional categories licensed in I°, it may be concluded that Case marking takes place on the head in these languages too. The reason why the marker is overtly separated from the verb is that the verb has not been raised from its deep (VP-internal) position, consequently failing to incorporate functional morphemes.33 The possibility for the verb to remain in V° occurs fairly frequently in the world’s languages. What happens in Aztecan is not so different from what is observed in languages like Italian, where the object clitic appears to be incorporated on the auxiliary34 (and is thus separated from the verbal head): (81) Non l’ ho ancora not OCL.SGM have.PRES.1SG still ‘I haven’t seen him yet, Leo (I mean).’

visto, see.PRT

Leo (ITALIAN) Leo

There is, however, a substantial difference between Italian and the languages examined in Nichols (1992), i.e. the categorial status of the clitic. In Italian, the clitic is a pronoun, i.e. an element which receives an argument role within deep structure (in the VP) and is incorporated on the verbal head at a later stage as a consequence of its phonologically weak nature (see Rizzi 1986, Kayne 1991, Uriagereka 1998). The clitics presented by Nichols (and discussed in this chapter), on the other hand, are agreement markers, morphemes belonging to verbal inflection and, as such, comparable to suffixes such as -o in a word like cant-o in Italian. The different status of the clitic (pronoun vs. agreement morpheme) may be ascertained by verifying its mutual exclusiveness with a coreferential NP. In a sentence like (81), for example, it is possible to realize either the clitic lo or a full NP like Leo alone; this shows that each of the two elements may, on its own, fulfill the θ-role and saturate the verbal valency:35 (82) a. Non l’ ho ancora visto. have.PRES.1SG still see.PRT not OCL.SGM ‘I haven’t seen him yet.’ b. Non ho ancora visto Leo. not have.PRES.1SG still see.PRT Leo ‘I haven’t seen Leo yet.’

126 Syntactic functions On the contrary, when the clitic is an agreement marker, it needs to be realized regardless of the presence of a full NP (as in the Italian word canto, where subject agreement cannot be omitted). This is what happens in Spanish when the object appears as a pronoun (see [83])36 or in Romanian pe care interrogative sentences, where the wh-constituent must be doubled by a clitic within the sentence (see [84], from Dobrovie-Sorin 1995): (83) a. La

llamaron a ella. call.PST.3PL to O.PRO.3SGF ‘They called her.’ [lit: ‘They called her to her.’] b. *Llamaron a ella.

(SPANISH)

OCL.3SGF

(84) a. Pe care băiat l-ai văzut? (ROMANIAN) OCL.3SGM-AUX.2SG see.PRT to which boy ‘Which boy did you see?’ [lit.: ‘Which boy did you see him?’] b. *Pe care băiat ai văzut? The ungrammaticality of the corresponding sentences in a language like Italian proves that clitics in Italian have an argument role and therefore cannot be doubled by a full NP except in topicalized structures such as (81) (see Chapter 6). To conclude this section, we would like to consider those languages in which syntactic functions are signaled on both the verb and the nominal constituent, such as Warlpiri (from Falk 2005): (85) Karnta-ngku ka-ju woman-ERG PRES-SCL.3SG.OCL.1SG ngaju. PRO.1SG.ABS ‘The woman sees me.’

nya-nyi see-NONPST

(86) Pakirdi ka-rla karnta nyina enamored.ABS PRES-DAT woman.ABS sit.NONPST ‘The woman is in love with the man.’

(WARLPIRI)

wati-ki. man-DAT

In (85), ‘woman’ (karnta) is marked with ERG Case as it is the ACTORsubject of the action expressed by the verb, whereas the first person pronoun (ngaju) bears ABS Case as required by its PATIENT deep role. Moreover, the syntactic functions associated with these constituents also appear on the verbal head where they are synthetically conjoined in the

Morphological Case: Cross-linguistic strategies 127 morpheme -ju. Sentence (86) similarly shows that DAT Case is marked on both the NP ‘man’ (wati) and the functional head indicating present tense. The Case of ‘woman’, on the other hand, is not marked on the verbal head. We can assume that in Warlpiri, ABS Case may be realized on the verbal head only when associated with the object of a transitive clause (i.e. in its prototypical function), as in (85); the same is impossible when ABS Case is associated with a THEME-subject, as in (86), where a stative predicate is present. The marking on the verbal head is therefore sensitive to the argument structure as a whole, with a distinction between transitive and intransitive verbs in the realization of the deep ABS Case. There are, however, ergative languages where the ABS Case is marked on the verbal head independent of its transitivity, as in the Avar case in (18)-(19), repeated below as (87): (87) a. Vas-ass jas j-ecc-ula. boy-ERG girl.ABS SGF.ABS-praise-PRES ‘The boy praises the girl.’ b. Vas v-eker-ula. ragazzo.ABS SGM.ABS-correre-PRES ‘The boy is running.’

(AVAR)

It can be noted that ABS Case is present on the verbal head in both sentences (regardless of the argument structure of the verb). As for dependence-marking, only ERG Case is marked on nominal elements (see [87]), ABS being the morphologically unmarked Case. The morpho-syntactic realization of deep roles is, as we have seen, a fascinating topic which merits further research.

4. The structure of the noun phrase

4.1. The internal structure of the NP In Chapters 1 and 2, we saw that the X' structure is a universal and indefinitely recursive model, and showed how categories are formed as a consequence of lexical insertion in phrasal and clausal environments. From this perspective, the verb is the head of the phrase functioning as the predicate of a non-nominal sentence (i.e. in a predicative construction with a TAM system and, in some cases, agreement features). However, when a lexical root, after insertion, projects a phrase into an argument position (within the VP) or a non-argument position (outside the VP or within a different phrase, as a modifier), it assumes the category of noun. The corresponding phrase (an NP) has the following internal structure: NP

(1) Spec

N’ N°

Compl

The structure in (1) is thus assumed by an NP in any position, whether the NP is an argument, modifier, predicate or adjunct, and the lexical root in N° shows the distributional characteristics of a noun. The evidence for this is that it can cooccur with articles, demonstratives and quantifiers (see section 1.3.1. and below), it can be defined in terms of categories such as gender and number (in inflecting and agglutinative languages) and, finally, it can cooccur with what were previously defined as second-level modifiers (section 1.5.3.). In this chapter we will first concentrate on the internal structure of the NP and later examine argumental NPs in order to describe the syntactic relationship that exists between them and the verbal head of the VP. We will then focus on the functional area assumed for this category (the DP, Determiner Phrase, dominating the NP). Finally, we will present a brief

130 The structure of the noun phrase discussion of the most typical modifiers of the nominal head, i.e. adjectives and relative clauses. Let us begin by considering the types of elements that occupy the Specifier and Complement positions in the NP, as well as their semantic values and surface orders which vary across typologically different languages.

4.1.1. Elements in Spec,NP As we have seen, the X' structure is identical for each phrase, irrespective of the category its head belongs to; similarly, the semantic features which characterize the head upon lexical insertion do not change as a consequence of categorization. In other words, a lexical root presents a number of semantic properties which cannot be altered by its being categorized as a verb or a noun within sentential structure. This means that an English root like kiss selects an , a shadow (§ 2.2.3.) and a , and this selection remains unchanged whether the root is merged to be a verbal or a nominal category. As a matter of fact, the event described in (2a) as a predication can also be expressed by means of a nominal structure (2b): (2) a. b.

I kiss Mary/Leo kisses Mary. My kiss to Mary / Leo’s kiss to Mary.

It can be observed that the of the verbal construction appears as a genitive form (a possessive or GEN Case marked NP) in the corresponding nominal construction, or as a PP in languages with no morphological Case. The is also realized as a PP.1 In spite of these formal differences, it can be concluded that in both cases the (verbal or nominal) event has the same participants. The same can be observed in typologically different languages, such as Chinese and Turkish (from Kornfilt 1993). Let us look at the following examples: (3)

a. Wǂ

miaoshù [wǂ-de fángzi]. (CHINESE) describe PRO.1SG-ASSOC house ‘I am describing my house.’ b. [Wǂ-de fángzi] de miaoshù. PRO.1SG-ASSOC house ASSOC describe ‘My description of my house.’ PRO.1SG

The internal structure of the NP 131 (4) Ahmet-in [öl-düg-ün-ü] duy-du-m. (TURKISH) Ahmet-GEN die-NOMIN-3SG.POSS-ACC hear-PST-1SG ‘I have heard that Ahmet died.’ [lit.: ‘Of Ahmet, I heard his death.’] In (3a), the first person pronoun wǂ appears twice: in the first case, it is interpreted as the (and subject) of miaoshù; in the second, when followed by the associative particle de,2 it assumes the role within the NP wǂ-de fángzi. The latter is the of miaoshù in both (3a) and (3b). In (3a), it functions as a direct object, as it follows miaoshù in a predicative context; in (3b), on the other hand, it is a modifier of miaoshù as is shown by the presence of the associative particle de.3 It should be noted that the lexical root miaoshù does not change whether it appears as a verbal category (with predicate function, as in [3a]) or a nominal category (as in [3b]). In Turkish, a language with morphological Case, the of the subordinate clause (a nominalized clause, as seen in § 2.5.) is expressed by a possessive coindexed with an NP bearing GEN Case. The examples above show that nouns, like verbs, have an argument structure. These arguments are merged into the structure according to a hierarchical order which matches that of the VP.4 There is thus an important semantic and structural similarity between NP and VP (see Stowell 1981, Szabolcsi 1989 among others), which confirms our assumption that lexical heads have a shared origin in terms of semantic roots, and supports the idea that there is a universally valid hierarchical order for constituent insertion.5 We conclude that the argument selection of the root kiss is , regardless of its category. As we have seen in the VP structure, the constituent with ACTOR macro-role is generated in Spec,NP,6 whereas lower arguments occupy the Complement position: (5)

NP N’

John’s ACTOR

kiss

to Mary GOAL

132 The structure of the noun phrase Even nouns whose semantic root has no corresponding verb (i.e. nouns that denote objects, such as father, friend, professor, etc.) have an argument structure. Yet, while event nouns unambiguously determine their arguments, object-denoting nouns define non-intrinsic relationships. These relationships result in a or other types of complements, which (possibly) complete the meaning of the nouns (see § 4.1.2.). The following examples show that a definite noun may produce an incomplete result if it is not linked to an argument that completes its relational meaning: (6) a. b.

*The father is coming.7 My father / Jim’s father is coming.

(7) a. b.

*I met the friend. I met your friend / Jim’s friend.

(8) a. b.

*I need to talk to the teacher. I need to talk to my teacher / to the math teacher.

In (6)-(8), it can be noted that the role is expressed by a possessive NP or a genitive. This analogy confirms that the șrole is included in the ACTOR macro-role (as we proposed in § 2.2.1.). Also, in languages with morphological Case, the ș-role is typically associated with GEN Case, e.g. Turkish (4), Latin (9) and Somali (10). In Somali the GEN Case marker also has a tonal realization (10b): (9)

Domus Mariae. house.NOM Maria.GEN ‘Maria’s house.’

(10) a. Guri-ga nagood. house-DET woman.PL.GEN ‘The women’s house.’ b. Guri-ga Calí. house-DET Cali.GEN ‘Cali’s house.’

(LATIN)

(SOMALI)

It should also be noticed that possessive NPs can correspond to different θ-roles, namely or . This accounts for the ambiguity shown below:

The internal structure of the NP 133 (11) This is my present. interpretation (a): ‘the one that I gave’. interpretation (b): ‘the one that is for me’. One final consideration about the surface realization of an ACTOR as a PP in post-nominal position is in order. Let us consider the following sentences from German (12) and Tagalog (13): (12) a. Sein Geschenk. his.N.NOM present.N.NOM ‘His present.’ b. Das Geschenk von DET.N.NOM present.N.NOM of ‘Leo’s present.’ c. *Das von Leo DET.N.NOM of Leo.DAT (13) a. Ang TRIG

kanya-ng O.PRO.3SG-ASSOC

‘His book.’ b. Ang libro ni TRIG libro DIR ‘Juan’s book.’ c. *Ang ni Juan TRIG DIR Juan

(GERMAN)

Leo. Leo.DAT Geschenk. present.N.NOM

libro. libro

(TAGALOG)

Juan. Juan libro. libro

It is clear that, if the or appear as PPs, they cannot precede the noun. There seems to be a strong syntactic preference for the Spec,NP position to be occupied by NPs. We will not go into the structural reason for this restriction; let us simply observe that, even within the VP, an ACTOR-subject is never preceded by prepositions. This is therefore another structural analogy between VP and NP which confirms that our analysis is valid cross-linguistically.

4.1.2. Elements in Compl,NP and restrictive noun modifiers Based on the X' structure, we know that a noun head can have a complement with an argument role, as in (14)-(16):

134 The structure of the noun phrase (14) The description of your house was very accurate. (15) The phone call to your friend was really long. (16) The arrival in Berlin was punctual. In these three sentences the constituents occupying the Compl,NP position are the PATIENT (14), GOAL (15) and LOCATIVE (16) of the respective noun heads. In other words, they are arguments selected on the basis of the lexical properties of the head (which can be omitted if retrievable from the context; see note 5 to Chapter 4). The meaning of a noun head can also be completed by restrictive modifiers, as in the following examples: (17) Hot food is better for dinner. (18) The math textbook is very well written. (19) The dress you bought yesterday looks so good on you. The modifier is an AP in (17), a PP in (18), a relative clause in (19). Like arguments, these elements also provide information needed to define the reference of the noun head. If they were omitted in (17)-(19), it would be impossible to determine, out of context, which food, book or dress was being referred to. However, they are not needed as event participants, but as elements that help identify the referent of the noun. Given this property, restrictive modifiers were long analyzed as constituents generated in Compl,NP (like arguments): Their NP-internal position made it possible to distinguish them from accessory constituents, such as adjuncts. For this reason, we have chosen to deal with them in this section. Yet, as we will see later (see § 4.4.), recent analyses have identified a specific position for the lexical insertion of restrictive modifiers, distinguishing them structurally from arguments. The role of restrictive modifiers in identifying referents and their semantic difference from non-restrictive (or appositive) modifiers is clearly shown in the following example from Italian: (20) a. I

ragazzi furbi sono usciti. (ITALIAN) boy.PL clever.PLM be.PRES.3PL go out.PRT.PLM ‘The clever boys went out.’

DET

The internal structure of the NP 135 b. I

ragazzi, furbi, sono usciti. boy.PL clever.PLM be.PRES.3PL go out.PRT.PLM ‘The boys, being clever, went out.’

DET

In the former case, the AP identifies a subset within the set boys. In the latter, the appositive AP, signaled by specific parenthetical intonation, simply provides additional information about a single property of the set boys, all the members of which are clever (and went out). The AP in (20b), in other words, is not a complement of the noun head. With respect to the constituent order within the NP, Italian is a headinitial language (see § 1.2.2.), hence heads normally precede their modifiers. Conversely, in head-final languages (such as Turkish and Japanese), an adjective precedes its noun head. As is known, however, the headcomplement parameter admits variation both within a single language and across languages and, for this reason, the relative order of these constituents can change.8 Let us take, for example, English and Tagalog, both VO languages. In English, the modifier (other than a PP or a relative clause) precedes the head (see [21a-b]); in Tagalog the nominal head can both precede and follow adjectives (22) and relative clauses (23): (21) a. That green book is mine. b. This math book is well written. (22) a. Ang

maganda-ng beautiful-ASSOC ‘The beautiful voice.’ b. Ang boses na TRIG voice ASSOC ‘The beautiful voice.’ TRIG

boses. voice

(TAGALOG)

maganda. beautiful

(23) a. Nakita ko ang babae-[ng nagbasa PAT.PST.see SCL.1SG TRIG woman-ASSOC ACT.PST.read ng diyaryo].9 DIR newspaper ‘I saw the woman who was reading a newspaper.’ b. Nakita ko ang [nagbasa ng diyaryo PAT.PST.see SCL.1SG TRIG ACT.PST.read DIR newspaper na] babae. ASSOC woman ‘I saw the woman who was reading a newspaper.’

136 The structure of the noun phrase The alternative position of the modifier with respect to the head can be explained in derivational terms. It is thus assumed that the Complement has a specific position (which defines the unmarked order in a given language) and every variation is determined by movement operations of the nominal head. An analysis of these syntactic operations is, however, beyond the scope of this book.

4.1.3. When the head is a nominalization As we have seen above, the NP hosts elements needed to complete the meaning of the head, including its arguments. Let us go back to noun heads with an argument structure. These are traditionally described as deverbal nouns (from a morphological perspective) or nominalizations (from a syntactic perspective) and defined as “event nominals” in Grimshaw (1990). Typical examples of nominalizations are words like completion, destruction and defeat.10 Since our main interest is to show the interaction between the various levels of analysis, we will focus on how certain lexico-semantic mechanisms interact with the syntactic structure also in this case. As was shown in (14)-(16), noun heads determine selection restrictions on their complements. Let us observe the following contrast: (24) a. Your donation to the hospital. b. *Your book to the hospital. A noun like donation implies that the possession of an object x (PATIENT) is transferred from y (ACTOR) to z (GOAL), and selects a complement (the GOAL) introduced by the preposition to; on the other hand, a noun denoting an object (like book) cannot select a complement of this type (Grimshaw 1990: 94-95). Nouns denoting events thus have an argument structure;11 however, unlike verbs, these nouns cannot assign Case so prepositions must be used to fulfill that function (see note 1 to Chapter 1). Let us consider two Italian sentences such as (25a) and (26a) and the corresponding nominalizations in (25b) and (26b):

The internal structure of the NP 137 (25) a. Pochi concorrenti hanno partecipato few competitor.PL have.PRES.3PL participate.PRT alla gara. (ITALIAN) contest to.DET ‘Few competitors participated in the contest.’ b. La partecipazione di pochi concorrenti DET participation of few competitor.PL ha reso l’ evento poco have.PRES.3SG make.PRT DET event little interessante. interesting ‘The participation of few competitors made the event uninteresting.’ (26) a. I

nemici hanno distrutto enemy.PL have.PRES.3PL destroy.PRT la città. DET city ‘The enemies have destroyed the city.’ b. La distruzione della città DET destruction of.DET city era imprevedibile. be.IMPF.3SG unforeseen ‘The destruction of the city was unforeseen.’ DET

Both the ACTOR with subject function in (25a) and the PATIENT with object function in (26a) appear as PPs introduced by the preposition di in (25b) and (26b). The remaining argument selected by the verb (appearing in [25a] and [26b]) can also be included, as shown below: (25) c. La

partecipazione participation alla gara to.DET contest l’ evento poco DET event little ‘The participation of few event uninteresting.’ DET

di pochi concorrenti of few competitor.PL ha reso have.PRES.3SG make.PRT interessante. interesting competitors in the contest made the

138 The structure of the noun phrase (26) c. La

distruzione della città da parte destruction of.DET city from part dei nemici era imprevedibile. enemy.PL be.IMPF.3SG unforeseen of.DET ‘The enemies’ destruction of the city was unforeseen.’ DET

The LOCATIVE in (25c) does not change in the nominal construction, as it is a prepositional complement in both the VP and the NP. On the other hand, the ACTOR appears as a PP, da parte dei nemici (in [26c]). This is necessary because an ACTOR appearing as a PP cannot be realized in Spec,NP (see § 4.1.1.); however, since a di-phrase is already used to express the PATIENT, the grammar rules out this duplication. Indeed, a sentence like (27), where nemici is an ACTOR and città a PATIENT, is ungrammatical: (27) *La distruzione della città dei nemici...12 The only possible interpretation for (27) is one where dei nemici is a modifier of the NP città (i.e. a ). Sentences like (26c) thus require the complex preposition da parte di. It should be noticed that when a noun selects both an ACTOR and a PATIENT and just one of the two arguments is realized, the corresponding sentence is ambiguous. In one interpretation, the PP is a subject; in another, the same PP is the object of the nominalization: (28) a. La

distruzione dei nemici destruction of.DET enemy.PL era imprevedibile. be.IMPF.3SG unforeseen ‘The enemies’ destruction was unforeseen.’ b. La critica di Carlo è stata DET criticism of Carlo be.PRES.3SG be.PRT feroce. fierce ‘Carlo’s criticism was fierce.’ DET

Let us see two more examples which show the complexity of this area of research:

NPs as arguments 139 (29) a. La

lenta distruzione della slow.SGF destruction of.DET ‘The slow destruction of the city...’ b. *La distruzione lentamente della città… DET

(30) a. *Il distruggere lento una città... b. Il distruggere lentamente DET destroy.INF slowly ‘Slowly destroying a city...’

città… city

una INDEF.DET

città… city

Although distruggere and distruzione are both NP heads, as shown by their syntagmatic context (both are preceded by a DET), some semantic differences remain as the two noun heads require different types of modifiers: An adjective in (29) and an adverb in (30).13 To conclude, it should be observed in passing that different types of deverbal nouns show specific cooccurrence options with NP-internal elements such as determiners. In English, for instance, -ing nominals only admit definite determiners, cannot occur in the plural and cannot be used as predicates (from Grimshaw 1990): (31) a. b. c. d.

The shooting of rabbits is illegal. *A/one/that shooting of rabbits is illegal. *The shootings of rabbits are illegal. *That was the shooting of rabbits.

Similar phenomena can be found in all languages, although predictable differences exist in their surface expression. This, however, will not be discussed here. Let us now consider the case in which an NP is merged into the structure as an argument of a verbal head.

4.2. NPs as arguments As noted in § 4.1., the NP can fulfill different functions in the structure. In this section, its behaviour as an argument will be discussed in typologically different languages. In a number of languages, argument NPs may appear as nouns or pronouns (both free and clitic). This is the case in English:

140 The structure of the noun phrase (32) The child is eating lots of candy. (33) Paul talked to his friends. (34) He has just arrived. (35) I will give it to him tomorrow. In (32), the two arguments – ACTOR and PATIENT – are realized through two full NPs, the former being introduced by the article the and the latter by the quantifier lots of. In (33) Paul, a proper noun, is referential by definition and thus realizes the ACTOR NP independently,14 whereas the NP with a GOAL macro-role is preceded by a preposition and contains a . In (34), he (a THEME) is a free pronoun functioning as a subject, whose referent can be identified from the (linguistic or extralinguistic) context. Finally, in (35) ACTOR, PATIENT and THEME are all expressed by pronouns. Similar options are available in many other languages, such as Italian, French, German, Modern Greek, etc. The following examples are the Italian equivalents of the sentences given above (notice the use of null and clitic pronouns in the last one): (36) Il

bimbo sta mangiando child be.PRES.3SG eat.GER.PROG molte caramelle. many candy.PL ‘The child is eating lots of candy.’

DET

(37) Paolo ha parlato ai Paolo have.PRES.3SG talk.PRT to.DET ‘Paolo has talked to his friends.’ (38) Lui è appena he be.PRES.3SG just ‘He has just arrived.’

suoi POSS.3.PL

arrivato. arrive.PRT

(39) Glielo darò IOCL.3SGM-OCL.3SGM give.FUT.1SG ‘I will give it to him tomorrow.’

domani. tomorrow

(ITALIAN)

amici. friend.PL

NPs as arguments 141 Let us now consider another Italian sentence: (40) Maria, l’ hai incontrata? Maria OCL.3SGF have.PRES.2SG meet.PRT.SGF ‘Maria, did you meet her?’ Incontrare, a two-argument verb (ACTOR and PATIENT), occurs here with three NPs: Maria, the subject pro and the clitic pronoun la, coreferent with Maria. The reason for this clitic resumption is that the NP Maria appears in a marginal position, as it is the sentence Topic (see Chapter 6). The NP Maria therefore occupies an extrasentential position, while the PATIENT role and the direct object function are assigned to the clitic pronoun la. Indeed, no two NPs can have the same argument role: The NP Maria is a Topic whereas the argument position is filled by the clitic pronoun. This is shown by the fact that the clitic pronoun cannot be omitted (41a), while the extraposed lexical NP can (41b): (41) a. *Maria, hai incontrata? Maria have.PRES.2SG meet.PRT.SGF b. L’ hai incontrata? OCL.3SGF have.PRES.2SG meet.PRT.SGF ‘Did you meet her?’ The phenomenon observed in (41) for Italian is obligatory for all argument roles in some languages, where lexical NPs always occupy extrasentential positions. Argument positions are filled by clitic pronouns incorporated on the verbal head (V°) or on the inflectional head (I°), forming a verbal complex (see Chapter 1, § 1.4.). These pronouns are coreferent with the extraposed NPs, which make their referential identification possible. This is what happens in polysynthetic (or incorporating) languages such as Warlpiri (see [42], from Jelinek 1984), Mohawk (see [43], from Baker 1996) and Somali (see [44], from Puglielli 1981): (42) Ngujulu-rluk ka-rnak-ngkuj-rlaz 1SG-ERG PRES-SCL.1SG.NOM-OCL.3SG.ACC-OCL.2SG.BEN warri-rniz nyunktu-ku. (WARLPIRI) karli-kiij boomerang-DAT NONPST-2SG.DAT seek-NONPST ‘I am looking for a boomerang for you.’

142 The structure of the noun phrase wa-hik-yéna-" kìkak this FATT-SCL.1SG.OCL.3SGM-capture-PERF (MOHAWK) ne kwéskwesk.15 CLASS pig ‘I have captured this pig.’

(43) Ne

CLASS

wuu ik-la (44) a. Anigak O.PRO.1SG DECL.SCL.3SGM OCL.1SG-with hadlayaa. (SOMALI) talk.PRES.PROG.3SGM ‘He is talking to me.’ b. Axmedk waan kaaga (ka-u-Øk-ka) DECL.SCL.1SG OCL.2SG-to-OCL.3SG-from Axmed qaaday.16 take.PST.1SG ‘I took it from Axmed for you.’ Finally, in other languages morphemes agglutinated on the verb cooccur with lexically full argument NPs. These are morphemes (rather than clitics) because the presence of the lexical NP is obligatory: If the NP is deleted and the corresponding morphemes are left on the verb, the sentence becomes ungrammatical. This is what happens in Tzutujil (Nichols 1992) and Lakhota (Van Valin 1992), as seen in Chapter 3 (§ 3.5.). The relevant sentences are repeated here:17 tzyaqj ch’ooyaa"k. (TZUTUJIL) (45) a. X-øk-keej-tij ASP-3SG-3PL-eat clothes mouse.COLL ‘The mice have eaten the clothes.’ thathakaj ota wichak-Øj-kte. (LAKHOTA) b. Lakhotak ki Indian DET bison many 3SG-3PL-kill ‘The Indian has killed a lot of bison.’

4.3. The functional area of NP: The determiner phrase (DP) Based on the NP-VP analogy (in terms of argument selection and syntactic insertion of constituents into Spec and Compl positions), it is plausible to assume that the NP also has a functional area which completes its predicative-lexical core, just like the CP and the IP. This was proposed and argued by various authors (Abney 1987, Szabolcsi 1994, Giusti 1997,

The functional area of NP: The DP 143 Bernstein 2001; see Ramaglia 2004 for further discussion and references). It was shown, that the NP must be assumed as the complement of a functional head D°, whose phrase forms the left periphery of the NP: (46)

DP Spec

D’ D°

Compl

Subsequently, it was argued that the DP should be divided into a number of specific functional projections (as will be shown for CP; see Chapter 6). In this way, it is possible to account for the presence and distribution of the various elements which contribute to determining a noun, i.e. articles, demonstratives, cardinal and ordinal numerals, possessives and so on (see Giusti 1997, 2006 among others). Analyzing the full structure of the DP, however, is well beyond the scope of this book. The DP will be therefore treated as a single phrase and the behavior of its main components will be described from a cross-linguistic perspective. The syntactic function of elements placed in the DP head (chiefly articles, but also demonstratives) is that of determining (specifying, defining) the referential properties of the head noun of the complement NP. The article can have various functions and assumes different forms across languages depending on its agreement properties. In Italian, for example, there are two types of articles - definite and indefinite - which obligatorily agree in gender and number with the noun they occur with: (47) a. Il (SGM) the b. La (SGF) the c. I (PLM) the d. Le (PLF) the

cane (SGM). dog casa (SGF). house tavoli (PLM). tables sedie (PLF). chairs

(48) a. Un (SGM) cane (SGM). a dog

(ITALIAN)

144 The structure of the noun phrase b. Una (SGF) casa (SGF). a house In other languages, articles are invariable. This is true not only in languages in which grammatical gender is not expressed by means of morphological markers (such as English), but also in languages like Arabic, where gender determines agreement in other contexts, e.g. with verbs or adjectives. In these languages, articles bear no agreement markers: (49) a. The boy. Al-walad.

b.

The girl. Al-bint.

(50) a. The boys. Al-’awlƗd.

b.

The girls. Al-banƗt.

A number of languages lack a definite-indefinite opposition and have only one series of articles. Usually these are definite articles, while indefiniteness is not marked by any dedicated functional head. Therefore, in some languages, indefiniteness is the morphologically unmarked case. Somali is a case in point (see [51b]). In other languages, however, indefiniteness is marked by a specific form of the noun (and eventual modifiers). This is the case of Arabic tanwƯn, a nasal consonant added in final position, but left unpronounced (see [52b]): (51) a. Macallin-ku wuu yimi. (SOMALI) teacher-DET.M.NOM DECL.SCL.3SGM come.PST.3SGM ‘The teacher has come.’ b. Nin wuu yimi. man DECL.SCL.3SGM come.PST.3SGM ‘A man has come.’ (52) a. Ra?aytu al-imra?ata al-tawƯlata. (ARABIC) see.PERF.1SG DET-woman.ACC DET-tall.F.ACC ‘I saw the tall woman.’ b. Ra?aytu imra?ata-n tawƯlata-n. see.PERF.1SG woman.ACC-INDEF tall.F.ACC-INDEF ‘I saw a tall woman.’

The functional area of NP: The DP 145 In some languages, the article does not precede the noun, but is suffixed to it. This happens, for example, in Somali (see [51a] above), Bulgarian ([53], from Borriero 1976) and Albanian ([54], from Kallulli 1999): (53) Žená-ta kojáto vzeh woman-DET REL.PRO.SGF marry.PST.1SG leniva. lazy ‘The woman I married is lazy.’ (54) Diell-i i bukur e sun-DET DET beautiful and ‘The sun is beautiful and warm.’

e COP

ngrohtë.18 DET warm i

(BULGARIAN)

(ALBANIAN)

In dealing with articles, a few words should be spent on their semantic distribution in terms of definiteness. Let us observe the following examples: (55) Mario went out. (56) a. The boy went out. b. He was popular with the girls. (57) a. A boy went out. b. I bought a new car. In (55), the noun head Mario is a proper name and as such is referentially defined (see Lyons 1968). For this reason, an article is not normally used (but see note 14 to Chapter 1). In (56a), the article shows that the noun head refers to a specific individual which can be traced back in the linguistic or extralinguistic context. However, this interpretation is ruled out in (56b), where the definite article indicates a generic, completely non-specific referent. Similarly, in (57a), the article a, traditionally called indefinite, may be interpreted as any or a specific one, whereas in (57b) only a specific interpretation is available. Definiteness should not be confused with specificity. Consequently, no necessary relationship should be established between the [±definite] form of the article and a specific interpretation of the referent. Hawkins (1978) considers definiteness a semantic rather than pragmatic property of NP, thus ruling out the ‘definite = known/given’ equivalence. Although some

146 The structure of the noun phrase authors describe definiteness in terms of identifiability, considering the pragmatic implications often entailed by definiteness (Lambrecht 1994, Lombardi Vallauri 2002), a definite NP is not always a specific or known element in the discourse; conversely, a specific NP may be indefinite, as was shown in (56)-(57). This is why it is important to keep definiteness and specificity distinct. Definiteness has the semantic function of selecting an object from a class of possible objects (Ihsane & Puskás 2001) or restricting the range of entities it can refer to (Heim 1982). Specificity, on the other hand, has the semantic-pragmatic function of referring to an element which is present in the discourse or is part of the speaker’s knowledge. To conclude the present discussion on the formal and semantic properties of determiners, let us observe the following examples from Italian: (58) a. (Il)

dormire fa sleep.INF do.PRES.3SG ‘Sleeping is good for you.’ b. (Il) mangiare troppi DET eat.INF too many fa male. do.PRES.3SG bad ‘Eating too many sweets is bad for you.’ c. (Il) correre veloce non DET run.INF fast NEG ‘Running fast is not easy.’ DET

dopo si vedrà. later MIDD see.FUT.3SG ‘We’ll worry about it later.’ b. Quel domani sembra DEM tomorrow seem.PRES.3SG ‘That tomorrow seems so far away.’

bene. well

(ITALIAN)

dolci sweet

è facile. be.PRES.3SG easy

(59) a. Il

DET

molto lontano. very far

In all the sentences given in (58), the NP head functioning as a subject is an infinitive. In these cases, an article may marginally be present. Finally, in (59a-b), the NP head contains the lexical entities dopo and domani, which are more readily used in other contexts and with other functions (prepositions or adverbs, as in vengo dopo, ‘I am coming later’, arrivo dopo di te, ‘I am coming after you’, arrivo domani, ‘I am coming tomorrow’,

The functional area of NP: The DP 147 and so on). In this type of sentence, the article undoubtedly has a syntactic function, whereas its semantic value with respect to referentiality is less evident.19 Finally, some languages lack articles completely. In these cases, definiteness is conveyed by other means, such as demonstratives, restrictive noun modifiers and, above all, linear order (depending on informationstructural requirements; see Chapter 6). In Chinese, for instance, a preverbal subject is usually interpreted as given information and, consequently, specific: (60) Rén lài le. (CHINESE) PERF person come ‘The person (known to speaker and listener) has come.’ If, on the other hand, a presentative structure is used, the subject is in postverbal position and constitutes new information in the sentence. In this case, the noun is interpreted as definite: (61) Lài le rén. come PERF person ‘Someone has come.’ Similarly, a direct object in final position is interpreted as indefinite and new, while it has a given and topicalized interpretation in initial position: (62) a. W΅

m΁i le shnj. buy PERF book ‘I have bought a book.’ b. Shnj w΅ m΁i le. book PRO.1SG buy PERF ‘The book, I have bought it.’ PRO.1SG

The absence of articles should not be considered a specific property of isolating languages (like Chinese). On the contrary, the article is absent in a wide range of languages which are otherwise extremely rich in inflectional morphology. The category of article is missing in most Slavic languages (except Bulgarian and Macedonian); here, too, definiteness is expressed by means of the order of constituents. In Polish (63), for example, a subject noun before the verb is interpreted as given and definite, while it indicates

148 The structure of the noun phrase an indefinite referent when in final position (just as in Chinese; see examples [60]-[61]): (63) a. Kobieta weszła do pokoju. woman.NOM enter.PERF.F to room.GEN ‘The woman has walked into a/the room.’ b. Weszła do pokoju kobieta. enter.PERF.F to room.GEN woman.NOM ‘A woman has walked into the room.’

(POLISH)

Quantificational phrases (QP) also appear in the DP, specifically in Spec,DP position. QPs are syntactic operators20 and indicate that a given property (or predication) is valid for every element in a set (universal quantifiers; see [64]) or only for some (existential quantifiers; see [65]): (64) a. Every student is preparing for the exams. b. All the students are prepared. (65) a. Some students are preparing for the exams. b. A student is prepared. The domain of quantification is extremely complex, both on a syntactic and a semantic level. It is therefore impossible to discuss it thoroughly here. For further discussion and references, see Sportiche (1988), Reinhart (1995), Chierchia (1997).

4.4. Noun head modification We saw in § 4.1.2. that the NP can contain, along with noun head arguments, modifiers which restrict the reference of the noun head and, as a result, make its identification possible. While they are not arguments, these constituents may also be considered obligatory, since they are needed for interpretation. Following our examination of NP arguments and the NP functional structure, let us take a closer look at two of the most common modifiers in the world’s languages, i.e. adjectives and relative clauses. Our choice is solely determined by the space limitations of this book, as other elements can also function as modifiers. A noun head can be modified by a PP (see [17]) or another NP, as in English:

Noun head modification 149 (66) a. I’d like a chocolate cake. b. Public opinion is focused on the Afghanistan case. Furthermore, while adjectival modification is extremely widespread, there are languages which lack adjectives proper. In these cases, the properties of a noun head are expressed by other means (see Ramaglia 2008 for an overview of the different types of adjectival modifcation); most commonly, by relative clauses, in which NPs such as a good boy is expressed as a boy who is good (67a), or by predicative forms with a possessive meaning, as shown in (67b) for Somali: (67) a. Wiil wanaagsan. boy be good.PRES.RED ‘A good boy.’ b. Nin xoog leh. man strength possess.PRES.3SGM ‘A strong man.’ [lit.: ‘A man who has strength.’]

(SOMALI)

4.4.1. Adjectival modification The adjective can be defined as the constituent that modifies a noun head, indicating one of its properties (68a) (see Croft 1991: 67) or giving a “mere description” of it (68b) (see Wierzbicka 1988: 468): (68) a. I’d like a cool, refreshing, sugarfree drink. b. I’m not a fan of the baroque style. Syntactically, adjectival modifiers have an attributive value. APs are NP-internal modifiers and can appear in prenominal or postnominal position, as in Italian, il bel libro, vs. il libro bello, ‘the good book’.21 While some languages admit both options (with different meanings, as shown below), in others the position of the adjective with respect to the noun head is rigidly grammar-dependent. In English, for example, attributive adjectives, however many they may be, precede the noun they modify22 Any other constituent order produces ungrammatical sentences: (69) a. A good boy. (*A boy good.) b. That nice little girl. (*That nice girl little.)

150 The structure of the noun phrase In Chinese, too, adjectives always precede the nouns they modify, and the associative particle de is normally inserted between them (from Li & Thompson 1981): (70) a. Róngyì de wèntí. (*wèntí de róngyì.) (CHINESE) easy ASSOC problem ‘An easy problem.’ b. Pàng de rén. (*rén de pàng.) fat ASSOC person ‘Fat people.’ c. Fùzá de xiànxiàng. (*xiànxiàng de fùzá.) complex ASSOC phenomenon ‘A complex phenomenon.’ These examples from English and Chinese should not lead to the conclusion that the fixed position of the adjective depends on the (reduced) morphology of a particular language. There are languages with extremely rich morphology and an invariable position for APs. Arabic is one of them: (71) a. Walad hasan. (*Hasan walad.) (ARABIC) boy good ‘A good boy.’ b. Al-walad al-hasan. (*Al-hasan al-walad.) DET-boy DET-good ‘The good boy.’ c. Al-walad al-hasan wa al-jamƯl. DET-boy DET-good and DET-handsome (*Al-hasan al-walad wa al-jamƯl.) ‘The good and handsome boy.’ The examples show that adjectives in Arabic must follow the noun they modify and agree with it in gender, number and Case. Furthermore, if a noun is definite, its adjectives must also be preceded by the article (see [71b-c]).23 This makes it possible to distinguish between the attributive and predicative functions of adjectives (as the meaning of Al-walad hasan is ‘The boy is good.’). In other languages, adjective position is not fixed, as it can both precede and follow the noun. This is true in Italian and French, among others:

Noun head modification 151 (72) a. Una

bella giornata/ beautiful day / una giornata bella. INDEF.DET day beautiful ‘A beautiful day.’ b. Una difficile impresa / INDEF.DET difficult feat / un’ impresa difficile. INDEF.DET feat difficult ‘A difficult feat.’ INDEF.DET

(ITALIAN)

(73) a. C’

est un personnage be.PRES.3SG INDEF.DET person vulgaire. (FRENCH) vulgar ‘It’s a vulgar character.’ b. Ce vulgaire personnage sévit toujours. DEM vulgar person act cruelly.PRES.3SG always ‘This vulgar character is always cruel.’ DEM

Nevertheless, it would be wrong to believe that the position of adjectives in these languages is totally free and without consequences for interpretation. Constituent order variation, in fact, conveys specific meanings. Moreover, when a language admits variation in the linear order of constituents, there is still only one syntactically unmarked order. Therefore, if an adjective appears in different positions with respect to the head noun, this must be ascribed to either (a) a different point of lexical insertion (because a single adjective form has different meanings), or (b) a derivational process where the basic order is altered to convey specific semantic (see below) or pragmatic information. As usual, syntactic and distributional criteria should be used to identify the unmarked order of adjectives in a clear and unambiguous way. In particular, for languages in which adjectives can be pre- or postnominal, two specific diagnostics exist for identifying the unmarked order: (a) (b)

The position in which it is possible to modify the adjective itself; The position in which there can be more than one adjective.

On this basis, we can state that in languages like French and Italian the unmarked position of adjectives is definitely postnominal. As a matter of

152 The structure of the noun phrase fact, prenominal adjectives in these languages may not be modified or followed by complements (whereas postnominal adjectives show no such restrictions): (74) a. Una

giornata bella da day beautiful to ‘A wonderful day.’ *Una bella da morire giornata. b. Un’ impresa difficile INDEF.DET feat hard ‘A feat that is hard to describe. *Una difficile da raccontare impresa. c. Una giornata molto INDEF.DET day very ‘A really beautiful day.’ *Una molto bella giornata. INDEF.DET

morire. die.INF

da to

(ITALIAN)

raccontare. tell.INF

bella. beautiful

(75) a. Une

décision appropriée decision appropriate aux circonstances. (FRENCH) to.DET circumstance.PL ‘A suitable decision under the circumstances.’ *Une appropriée aux circonstances décision. b. Un garçon heureux de son sort. INDEF.DET boy happy of his destiny ‘A boy happy with his destiny.’ *Un heureux de son sort garçon. INDEF.DET

Similarly, it is impossible (or at least marginal or non-preferred)24 to insert more than one adjective to the left of the noun, either via juxtaposition or coordination:25 (76) a. È un ragazzo sveglio, simpatico e be.PRES.3SG INDEF.DET boy smart nice and intelligente. (ITALIAN) intelligent ‘He is a smart, nice and intelligent boy.’ *?È uno sveglio, simpatico e intelligente ragazzo.

Noun head modification 153 b. È un ragazzo simpatico INDEF.DET boy nice be.PRES.3SG ‘He is a nice and fun boy.’ *?È un simpatico e divertente ragazzo.

e divertente. and fun

Once the unmarked position has been defined and it has been verified that each variation in linear order determines a difference in meaning, it is important to see which specific semantic function is associated with the adjective depending on its position. In its unmarked position (i.e. postnominal in languages like Italian and French), the adjective usually has a restrictive function as it identifies the referent designated by the head noun by defining a subclass within the class of the noun: (77) Questo posto è pieno di funghi velenosi. DEM place be.PRES.3SG full of mushrooms poisonous ‘This place is full of poisonous mushrooms.’ (= the subclass of poisonous mushrooms is identified within the class of mushrooms) In its marked position, on the other hand, the adjective normally has an appositive (often connotative) function, as it does not restrict reference but contributes extra information about the head noun, possibly expressing the speaker’s opinion, as in examples (78b)-(79b):26 (78) a. Una INDEF.DET

donna woman

grande. big

‘A big woman.’ b. Una grande donna. INDEF.DET great/big woman ‘A great woman.’ (79) a. Un INDEF.DET

uomo man

‘A poor man.’ b. Un pover’ INDEF.DET poor ‘A miserable man.’

povero. poor uomo. man

154 The structure of the noun phrase When adjectives indicating physical properties (such as lungo, ‘long’, alto, ‘tall’, freddo, ‘cold’) occupy an unmarked position, they indicate an inherent characteristic of their referent: (80) a. I

piovosi inverni rainy winters ‘The rainy London winters.’ Le aride colline di questo DET barren hills of DEM ‘The barren hills of this landscape.’

DET

b.

londinesi. of-London paesaggio. landscape

Finally, adjectives in marked positions can modify one of the properties of the referent, rather than the referent itself. Let us consider the interpretation of the adjectives in examples (b): (81) a. Un INDEF.DET

amico friend

‘An old friend.’ (= an aged friend) b. Un vecchio INDEF.DET old ‘An old friend.’ (= a long-time friend) (82) a. Un

seminariste seminarist ‘A young seminarist.’ b. Un jeune seminariste. ‘A new seminarist.’ INDEF.DET

vecchio. old

amico. friend

jeune. young

(FRENCH)

(83) a. Un

cinéma ancien. cinema old ‘An old movie theater.’ b. Un ancien cinéma. ‘A former movie theater.’ INDEF.DET

Some adjectives, e.g. color and ability adjectives, cannot have an appositive function. Since they are only restrictive, the prenominal position is ruled out (see note 26 to Chapter 4):

Noun head modification 155 (84) a. Un

homme man ‘A blind man.’ b. Un livre INDEF.DET book ‘A green book.’ INDEF.DET

aveugle blind

/ *un aveugle homme

vert / *un vert livre green

It has been observed cross-linguistically that, if a noun is modified by several adjectives, their order within the NP is not free, and adjectives belonging to different semantic classes occur in fixed positions (Hetzron 1978, Dixon 1982, Sproat & Shih 1990, Cinque 1994, Scott 2002).27 In particular, adjective order in a language with postnominal adjectives – such as Arabic – mirrors the order that applies in a language with prenominal adjectives like English:28 (85) a. N > Acolor > Asize Al-kitƗbu al-’axd̞aru DET-book.NOM DET-green.NOM ‘The small green book.’ b. Asize > Acolor > N The small green book.

(ARABIC) al-s̞aƥƯru. DET-small.NOM (ENGLISH)

(86) a. N > Aprovenance > Acolor > Aquality (ARABIC) ŠƗyu-n s̞Ưniyyu-n ’axd̞aru29 Chinese.NOM-INDEF green.NOM.INDEF tea.NOM-INDEF jayyidu-n. excellent.NOM-INDEF ‘An excellent green Chinese tea.’ b. Aquality > Acolor > Aprovenance > N (ENGLISH) An excellent green Chinese tea. (87) a. N > Asize > Aquality (ARABIC) c al’abu bi-al-kurati al-kabƯrati play.IMPF.1SG with-DET-ball.F.GEN DET-large.F.GEN al-jamƯlati. DET-beautiful.F.GEN ‘I play with the beautiful big ball.’ (ENGLISH) b. Aquality > Asize > N I play with the beautiful big ball.

156 The structure of the noun phrase These restrictions prove that APs are inserted into the DP according to a hierarchy. In languages with unmarked prenominal adjectives (such as English), the following hierarchy applies (from Cinque 2010): (88) QUALITY > SIZE > SHAPE > COLOR > PROVENANCE Greenberg (1966: 87) first identified a hierarchy within the noun phrase, with a different distribution across languages. His Universal 20 is formulated as follows: (89) UNIVERSAL 20 When any or all of the items (demonstrative, numeral, and descriptive adjective) precede the noun, they are always found in that order. If they follow, the order is either the same or its exact opposite.

Cinque’s analysis (2005b) later showed that this universal is part of a broader generalization which includes the distribution of attributive APs in the DP (Cinque 2010) and adverbials in the sentence (Cinque 1999; see Chapter 5). In the case of unmarked postnominal adjectives, we can expect adjectives to follow the order in (88) or the exactly opposite order30 (as previously shown when contrasting English and Arabic). The unmarked position of adjectives is therefore essential in defining the direction of the hierarchy. We can expect that in a VSO language like Tagalog, the hierarchical order of prenominal adjectives (see [22a]) is the same as in English. This prediction is confirmed, as shown in the following examples: (90) Aquality > Acolor > N Ang akin-g TRIG O.PRO.1SG-ASSOC pula-ng libro. red-ASSOC book ‘My two heavy red books.’

dalawa-ng two-ASSOC

mabigat heavy

(TAGALOG) na ASSOC

(91) Aquality > Asize > N Ang akin-g magaganda-ng malalaki-ng larawan. TRIG O.PRO.1SG-ASSOC beautiful-ASSOC large-ASSOC photo ‘My beautiful large photos.’

Noun head modification 157 (92) Asize > Ashape > Acolor > N Ang maliit na bilog na TRIG small ASSOC round ASSOC ‘The small green round box.’

berde-ng green-ASSOC

kahon. box

It can therefore be concluded that the hierarchical order of adjectives shown in (88) is independent of the head-complement parameter. The unmarked position of adjectives is the only relevant factor, as adjectives are modifiers rather than primary functions (see Chapter 1, § 1.5.). This is extremely important from a syntactic point of view, as it rules out an analysis where APs are simply adjoined within the DP. As we know, adjuncts are merged with no restrictions on linear order (see Chapter 1), whereas data clearly show that this is not true of multiple adjectives. We must thus assume, following recent studies (Cinque 2010 in particular), that adjectives are merged into the Specifier of dedicated functional projections, each of which is specified for a semantic feature. These functional projections are located in a part of the structure which dominates the NP and ends with the DP. The heads of these functional projections, which will be simply referred to as FP (Functional Phrase), are associated with a distinctive semantic feature according to the hierarchy in (93): (93)

DP D’ FP1

D° AP1

FP2

[+quality] AP2

FP3

[+size] AP3

[+shape] AP4

FP4 [+color]

FP5

AP5 [+provenance]

NP

158 The structure of the noun phrase While we are not going to examine this hierarchy in detail, a few observations are needed to demonstrate its empirical validity. First of all, the functional heads where the various semantic features are encoded (which are associated with the different APs) may be considered as the syntactic locus where some elements of derivational morphology are merged. In particular, we are referring to diminutives, augmentatives, diminutives of endearment, etc. In a language like Italian, these express meanings corresponding to some adjectival modifiers.31 Let us consider the following sentences: (94) a. Ha davvero un INDEF.DET have.PRES.3SG really caratteraccio. bad-personality ‘S/he really has a bad temper.’ b. Per favore, prendi quello scatolone big-box please take.IMP.2SG DEM lo scaffale. DET shelf ‘Get that big box that is on the shelf, please.’

(ITALIAN)

sopra on

While English uses independent APs to convey certain meanings, in Italian the noun head incorporates the morphemes relating to a given semantic feature, after moving into the relevant functional head. This process resembles the V°-to-I° movement of the verb to incorporate inflection. Further evidence in favor of this analysis comes from the order of these affixes to the right of the noun: as is shown in (95), size morphemes are closer to the noun than quality morphemes. This shows that during its movement, the noun head incorporates quality morphemes last. Hence, the relevant functional projection must be higher in the DP functional area (see Baker’s [1996] Mirror Principle in § 5.3.):32 (95) a. tavolinetto, tavolinuccio /*tavolettino, *tavoluccino ‘tiny little table’ b. seggiolinetta, seggiolinuccia /*seggiolettina, *seggioluccina ‘tiny little chair’ c. bambolinetta, bambolinuccia /*bambolettina, *bamboluccina ‘tiny little doll’

Noun head modification 159 These functional heads could also be considered the place where classifiers are merged in languages like Chinese. As we saw in Chapter 1 (note 30), the insertion of these elements depends on specific semantic features of the noun head. They are not determiners (D°), but plausibly have a direct relationship with the NP functional area. So far, we have only considered adjectives modifying a noun head that denotes an object (book, box, ball, tea, etc.). However, adjectives are rigidly ordered even when they are associated with noun denoting events (e.g. invasion, attack, observation, etc). In particular, cross-linguistic studies have indicated the following hierarchy (from Cinque 1994):33 (96) SPEAKER-ORIENTED > SUBJECT-ORIENTED > MANNER > ARGUMENTAL This hierarchy is illustrated in the following examples: (97) a. A probable immediate awkward US attack. b. Ses autres probables gauches réactions immédiates à ta lettre. (FRENCH) ‘His other probable immediate awkward reactions to your letter.’ The hierarchy in (96), just like that in (88), refers to the unmarked order of prenominal adjectives. The conditions laid out in Universal 20 (see [89]) hold in languages with unmarked postnominal adjectives. For example, Arabic shows a perfect mirror-image order (from Ramaglia 2006): (98) ord > N > Athematic > Aspeaker-or c TƗlitu hujnjmi-n amirƯkiyyi-n third.NOM attack.GEN-INDEF American.GEN-INDEF muh̞tamali-n ‘alƗ al-snjdƗn probable.GEN-INDEF on DET-Sudan.GEN ‘A third probable US attack on Sudan.’

(ARABIC)

Again, a hierarchy of corresponding functional projections must be assumed, into which these adjectives are merged. Finally, let us spend a few words on argumental (or thematic) adjectives. This term refers to adjectives relating to the argument selection of the noun. An argumental adjective realizes one of the arguments of the noun head, as shown in example (99), where German represents the ACTOR of invasion, replacing a corresponding prepositional structure (100):

160 The structure of the noun phrase (99) The German invasion of Albania. (100) The invasion of Albania by the Germans. As in the case of restrictive adjectives, these modifiers are exclusively postnominal in a language like Italian (see Giorgi 1988: 304-307). This is clearly shown in the following ungrammatical example: (101) *La

tedesca invasione dell’ Albania. German invasion of.DET Albania (intended: ‘The German invasion of Albania.’)

(ITALIAN)

DET

Argumental adjectives preferably express the ACTOR macro-role. Indeed, if a noun selects two arguments, it is impossible to express the PATIENT with an adjective and the ACTOR with a PP in a language like Italian: (102) *L’

invasione invasion

albanese (PAT) da Albanian from della Germania (ACT). of.DET Germany (intended: ‘Albania’s invasion of Germany.’)

DET

parte part

Nevertheless, when no ACTOR is present, other arguments may appear as APs. Let us consider the following examples, where the adjectives expresses a PATIENT or THEME (with nouns relating to unaccusative verbs): (103) a. La

produzione automobilistica (PAT). production automotive ‘Car production.’ b. L’ estrazione petrolifera (PAT). DET extraction petroliferous ‘Oil extraction.’ c. Il genocidio curdo (PAT). DET genocide Kurdish ‘The Kurdish genocide.’ d. La trasferta londinese (LOC). DET transfer of-London ‘The away game in London.’ DET

Noun head modification 161 partenza italiana (TH). departure Italian ‘The Italians’ departure.’

e. La

DET

4.4.2. The relative clause The relative clause is a very interesting field of research. First, the relevant data are extremely complex; second, this construction is very important from a theoretical point of view, as it is employed in various syntactic structures, including adverbial subordination (see Chapter 5, § 5.5.1.) and Focus structure (see Chapter 6, § 6.4.). The relative clause is a sentential modifer within the NP which is used to (a) identify the referent expressed by the noun head (restrictive relative clauses) or (b) provide additional information on that referent when it is already semantically identified (appositive relative clauses): (104) a. The boy who is coming soon is Ann’s brother. b. John, who is coming soon, is Ann’s brother.

(RESTRICTIVE) (APPOSITIVE)

In (104a), the relative clause is needed to identify the boy that is being talked about. In (104b), on the other hand, the proper noun John is referential by definition, and the relative clause simply provides accessory information regarding the relevant head noun. Apart from this semantic difference (depending on the nature of the NP head), restrictive and appositive relative clauses have different syntactic properties across languages. We will come back to this topic in § 6.4.2.3. and concentrate first on the properties of restrictive relative clauses.

4.4.2.1. The internal structure of restrictive relative clauses Several theories have been proposed since the Sixties to account for the syntax and interpretation of relative clauses within the generative framework (for an overview see Alexiadou et al. [eds.] 2000, Bianchi 1999, 2002a,b). These theories provide adequate answers to two different crucial questions, namely (a) how can the head of the relative clause play two roles in the structure at the same time (in the relative clause and in the main clause where the relative DP is merged)? (b) how is the relative clause syntactically related to the noun head it modifies?

162 The structure of the noun phrase Until the late Eighties, the most widely accepted theory was that the relative clause was adjoined to the noun head (Ross 1967, Jackendoff 1977). This made it possible to distinguish relative clauses (see [105a]) from object clauses (105b), the latter occupying the Compl,VP position): (105) a. John talked [to the boy [that was tired]]. b. John told the boy [that he was tired]. The analysis of relative clauses as adjuncts was, however, completely abandoned in the Nineties, as a number of studies provided morphosyntactic evidence of relative clauses being external to the head noun. Nevertheless, the head noun originates within the relative clause and this explains its dual role (see below). In this respect, Kayne’s (1994) proposal is one of the most influential in the generative framework. Following studies by Vergnaud (1974) and Chomsky (1977), Kayne argues for a raising analysis, in which the D° head selects as its sole complement the relative clause (a CP) containing the NP head (see [107]). In other words, a restrictive relative clause such as (106a) would be generated as shown in (106b). This structure accounts for the macro-role of the NP head (book) in the relative clause (i.e. PATIENT), and shows how the surface structure is derived by movement of the NP head into an operator position (see note 20 to Chapter 4), namely Spec,CP. This movement is represented below: (106) a. The book that I bought. b. [DP the [CP that [IP I bought [NP book]]] (107)

DP D’ D° the NP book

CP that

IP VP

Ik tk

bought

tNP

Noun head modification 163 According to this hypothesis, the relativization process implies movement of the NP head. The relevant NP is merged within the relative clause and, once reached scope position (Spec,CP), it can establish a relationship with (i.e. obtain a syntactic function from) the predicate of the root clause. The dual role of the NP is thus explained. Consider the following sentences: (108) a. b. c. d.

The boy [that is talking to Jim] studies math. The boy [(that) I invited] studies math. I met the boy [that studies math]. I met the boy [(that) you introduced me to at the party].

In the previous sentences, the NP head boy has different syntactic functions with respect to the predicates of the relative and root clauses. In (108a), it is the subject of both; in (108b), it is the subject of the main predicate (study) and the object of invite in the relative clause; in (108c), it is the object of the root verb (meet) and the subject of the relative clause; finally, in (108d), it is the object of both predicates. The dual role of the NP is immediately explained, if we consider the movement in (107). Here, the NP head is generated within the relative clause with a specific ș-role and syntactic function; after movement to Spec,CP it assumes the syntactic function that the DP containing the relative clause has in the root clause. As an example, let us consider the structural derivation of (108a) and (108b): (108a’)

IP I’ DPk

VP

studies D’

D° the NP boy

V’ CP that t’NP

tDPk

tv DP

IP is

VP

tNP

talking

math

PP to Jim

164 The structure of the noun phrase (108b’)

IP I’ DPk

VP

studies D’

V’

D° the

CP

NP

(that)

boy

tDPk

tv DP

IP VP

Ik tk

invited

math

tNP

As we can see, the function of the NP head in the relative clause depends on the place in the relative clause where it is merged, i.e. subjectACTOR in (108a’) and object-PATIENT in (108b’). The function of the NP head in the root clause, on the other hand, depends on the argument role and syntactic function of the DP containing the relative CP.34 This analysis accounts for the different morpho-syntactic realization of NP heads in typologically different languages. In a language with morphological Case such as Russian, for example, the NP head bears a Case marker that shows its syntactic function in the root clause; at the same time, within the relative clause, a relative pronoun indicates its role in the subordinate CP (from Comrie 1981): (109) a. Devuška [kotor-aja prišla] maja girl.NOM REL.PRO-3SGF.NOM arrive.PST.SGF POSS.1.SGF sestra. (RUSSIAN) sister ‘The girl who has arrived is my sister.’ b. Devuška [kotor-uju videla] girl.NOM REL.PRO-3SGF.ACC see.PST.SGF maja sestra. POSS.1.SGF sister ‘The girl she has seen is my sister.’

Noun head modification 165 c. Ja videl devušku S.PRO.1SG see.PST.SGM girl.ACC prišla]. arrive.PST.SGF ‘I have seen the girl who came.’ d. Ja videl devušku S.PRO.1SG see.PST.SGM girl.ACC ty tože videl]. S.PRO.2SG too see.PST.SGM ‘I saw the girl that you also saw.’

[kotor-aja REL.PRO-3SGF.NOM

[kotor-uju REL.PRO-3SGF.ACC

In Somali, on the other hand, the different roles of the head are signaled by verbal morphology, which presents specific variations. Let us look at the following sentences (from Frascarelli & Puglielli 2005a, Antinucci 1981): (110) a. Wiil-ka [Maryan la hadlayaa] waa with talk.PRES.PROG.RED.NOM DECL boy-DET Maryan walaal-kay. (SOMALI) brother-POSS.1.SG ‘The boy that is talking to Maryan is my brother.’ b. Moos-ka [aad cunaysaa] waa cerin. banana-DET SCL.2SG eat.PRES.PROG.2SG.NOM DECL unripe ‘The banana you are eating is unripe.’ c. Wiil-ka [af talyaani-ga hadlayá] boy-DET language Italian-DET speak.PRES.PROG.RED baan jeclahay. FM.1SG love.PRES.1SG ‘I love the boy that is speaking Italian.’ d. Boosta-da gee warqad-da [aan post office-DET take.IMP.2SG letter-DET SCL.1SG qorayó]. write.PST.SUBOR ‘Take the letter I wrote to the post office.’ When the NP head is a subject in the relative clause (110a,c), the corresponding verb appears in a reduced form35 (as in all cases in which the subject is realized in the CP area). The reduced paradigm is predictably absent in (110b), where the NP head (‘banana’) is the object of ‘eat’. Finally, when the NP head is the object of both sentences, the relative verb shows the subordinate paradigm (110d). This led Antinucci (1981) to

166 The structure of the noun phrase assume that subordination in Somali originated from this type of construction, as the same pattern is found in object and adverbial clauses. As we will see in Chapter 5, this hypothesis can be extended to a number of typologically different languages, where clausal subordination seems to result from various forms of nominal modification.

4.4.2.2. Relative clauses from a typological perspective Typologically, there are interesting restrictions on the syntactic function of the NP head of relative clauses. As stated in Keenan & Comrie (1977), an important correlation exists in languages between the syntactic function of an NP within the relative clause and the possibility that it functions as an NP head. This correlation led the authors to formulate the following hierarchy: (111) Accessibility Hierarchy (Keenan & Comrie 1977: 66) SUBJ > DO > IO > OBL > GEN > COMPAR According to this hierarchy, subject NPs can always be relative heads, followed by direct objects, indirect objects and so on. The syntactic function most difficult to relativize is the object of comparison. Therefore, no language admits relativized indirect objects unless it admits relativized subjects and direct objects; in addition, no language can admit relativized direct objects only, and so on. In English, NPs can always head relative clauses, regardless of their syntactic function, as shown in the following examples: (112) a. b. c. c'. d. d'. e. f. f'. g. h. i.

The woman [that is reading a book] is my sister. (SUBJ) The book [(that) the woman is reading] is interesting. (DO) The boy [I will give this book to] is a good friend of mine. (IO) The boy [to whom I will give this book] is a good friend of mine. The knife [I cut the meat with] has broken. (OBL) The knife [with which I cut the meat] has broken. The town [where we went] was really beautiful. (OBL) The person [I’m going out with] is an old friend. (OBL) The person [with whom I’m going out] is an old friend. The day [I met you] it was raining. (OBL) The reason [why I called you] is important. (OBL) The writer [whose book I read] is very famous. (GEN)

Noun head modification 167 j. The author [I read the book of] is little known. j'. The author [of whom I read the book] is little known. (GEN) k. The man [Maria is taller than] is her brother. (COMPAR) As illustrated, not only subjects and direct objects, but every type of indirect and oblique complement (corresponding to ș-roles , , , etc), as well as genitive NPs and objects of comparisons may head relative clauses.36 The same is also true for Italian. Since the object of comparison can be relativized in this language, so can all the other functions, given the hierarchy in (111): (113) L’

uomo [di cui Maria è più alta] man of REL.PRO Maria be.PRES.3SG more tall è suo fratello. (ITALIAN) be.PRES.3SG POSS.3.SG brother ‘The man Mary is taller than is her brother.’ (cf. [112k]) DET

Not all languages, though, are so permissive. French, for instance, rules out the final function of the hierarchy in (111), the relativization process stopping at genitive NPs (114a) and excluding objects of comparisons (114b): (114) a. L’auteur [dont j’ai lu le livre] est très fameux. (cf. [112 i/j]) b. *L’homme [que que Marie est plus grande] est son frère. (cf. [112k]) In other languages, such as Welsh, relativization possibilities stop at direct objects. The sentence is therefore ungrammatical if an oblique NP (e.g. a LOC) is relativized (McCloskey 2002): (115) a. Y

bachgen a oedd yn darllen. (WELSH) boy that be.PST.3SG in read.PROG ‘The boy that was reading.’ b. *Dyma ‘r llyfr y darllenais y storia. here-is DET book that read.PST.1SG DET story (intended: ‘Here is the book where I read the story.’) DET

Clearly, the most restrictive languages are those that only admit relativization of the subject function. This is the case of Toba-Batak, shown in the following examples (from Comrie 1981):

168 The structure of the noun phrase (116) a. Manussi abit boru-boru i. (TOBA-BATAK) wash clothes woman DET ‘The woman is washing the clothes.’ b. Boru-boru na manussi abit i. woman that wash clothes DET ‘The woman that is washing the clothes.’ c. *Abit na manussi boru-boru i. clothes that wash woman DET (intended: ‘The clothes that the woman is washing.’) Relativization of the direct object in (116c) yields ungrammaticality. Nevertheless, the grammar offers a means to obviate these restrictions. The hierarchy in (111) refers to the syntactic functions of the NP head, not to its deep role: Therefore, if a direct object needs to be relativized, the sentence can be made passive so that the PATIENT assumes subject function (see [117a]), becoming available for relativization (117b): (117) a. Disussi ni boru-boru abit i. wash.PASS by woman clothes DET ‘The clothes were washed by the woman.’ b. Abit na disussi ni boru-boru i. clothes that wash.PASS by woman DET ‘The clothes that were washed by the woman.’ Another interesting typological factor relates to the type of relative clauses found in the world’s languages. In particular, linguistic typology considers the structural position of the NP head and the way it is resumed (if it is) in the relative clause (see Comrie 1981): Four types of relative clause are thus identified. In the nonreduction type the NP head occupies the position within the relative clause determined by the argument structure of the relative verb. Therefore, if the NP head is a PATIENT and, as such, assumes direct object function in the relative clause, it appears in the position defined for that role in the specific language. For example, in an SOV language like Bambara (from Comrie 1981), the direct object occupies the preverbal position (see [118a]), and consequently the NP head in the relative clause appears in the same position (followed by a relative particle; see [118b]):

Noun head modification 169 (118) a. N ye so ye. (BAMBARA) S.PRO.1SG PST house see ‘I saw the house.’ b. Tyİ be [n ye so min ye] dy2. man PRES S.PRO.1SG PST house REL see build ‘The man is building the house I saw.’ In other words, in these languages the NP head does not raise to Spec,CP (see [107]). In most languages, however, the NP head is extracted from the relative clause and appears in an external position (as shown in [108a’-b’] for English). In this case, languages use different strategies to resume the head in the relative clause. One of these is the gap strategy, according to which no pronominal element is present in the relative clause to indicate the syntactic function of the NP head. This is the case of Somali (see [110]) and Afar:37 (119) a. [Yemee’te] ab’ba tuble. 3SGM.PERF.come father.ACC 2SG.PERF.see ‘You saw the father that came.’ b. [A‘nu ub’le] aw’ki ‘daa S.PRO.1SG see.PST.1SG boy.NOM stone.ACC ‘The boy I saw threw the stone.’ c. ‘is [‘yo-h yex’e] S.PRO.3SGF O.PRO.1SG-to give.PERF.3SGM bey’te. take.PERF.3SGF ‘She took the shoes he had given to me.’

(AFAR)

cammi’se. throw.PST.3SGM kabel’la shoes.ACC

Whether the NP head has subject function in the relative clause (119a) or is the subject of the root clause (119b) or the object of both predicates (119c), its resumptive pronoun within the relative clause is null. The NP head and relative clause are simply juxtaposed and the role of the head in the relative clause is inferred from the argument structure of the verb. In other languages, the NP head is resumed in the relative clause by means of pronouns with variations depending on its syntactic function: a resumptive pronoun is ruled out if the NP has a subject function in the relative clause, it is optional if it functions as a direct object and obligatory when it has any other function. This is the pronoun retention strategy, which is clearly illustrated in the following Persian examples (from Comrie

170 The structure of the noun phrase 1981). Notice that in this case relative clauses usually contain a C° head (which is lacking in the languages that employ the previous strategy): (120) a. Mard-i [ke (*u) bolandqadd bud] PRO.3SG tall be.PST.3SG man-DET that jijera košt. chicken.ACC kill.PST.3SG (PERSIAN) ‘The man that was tall killed the chicken.’ b. Hasan mard-i-ra [ke zan (ura) Hasan man-DET-ACC that woman PRO.3SG.ACC zad] mišenasad. know.PRES.3SG hit.PST.3SG ‘Hasan knows the man that the woman hit.’ [lit.: ‘Hasan knows the man that the woman hit (him).’] c. Man zan-i-ra [ke Hasan *(be u) S.PRO.1SG woman-DET-ACC that Hasan to PRO.3SG jijera dad] mišenasad. chicken.ACC give.PST.3SG know.PRES.1SG ‘I know the woman Hasan gave the chicken to.’ [lit.: ‘I know the woman that Hasan gave the chicken to her.’] Finally, in some languages the relative clause is introduced by a relative pronoun which resumes the NP head indicating its syntactic function in the relative clause. This strategy has been illustrated in the Russian examples (109a-d) for subject and direct object functions. For the sake of completeness, another example from Russian is given here, where the NP head has the syntactic role of indirect object: (121) Devuška [kotor-oj ja girl.NOM REL.PRO-3SGF.DAT S.PRO.1SG knigu] maja sestra. book.ACC POSS.1.SGF sister ‘The girl I gave the book to is my sister.’

dal give.PST.SGM (RUSSIAN)

Interestingly, these strategies not only define cross-linguistic variation in the formation of relative clauses as variation may also appear within languages: A single language can employ different strategies, depending on the syntactic function of the NP head. Italian is a case in point. The first four sentences from (112), repeated in (122), show that Italian seems very much like Persian in the case of subjects and direct objects (although

Noun head modification 171 pronoun resumption is ruled out in both cases in Italian). For every other function, however, a relative pronoun is needed (DET+quale, or cui), preceded by the appropriate preposition. Note that the complementizer cannot be used in this case, according to a restriction applying in many languages across the world:38 (122) a. La

[che legge un libro] that read.PRES.3SG INDEF.DET book è mia sorella. (SUBJ) (ITALIAN) be.PRES.3SG POSS.1.SG sister ‘The woman that is reading a book is my sister.’ b. Il libro [che la donna legge] DET book that DET woman read.PRES.3SG è interessante. (DO) interesting be.PRES.3SG ‘The book the woman is reading is interesting.’ c. Il ragazzo [al quale darò DET boy to.DET which give.FUT.1SG questo libro] è un mio DEM book be.PRES.3SG INDEF.DET POSS.1.SG caro amico. (IO) dear friend ‘The boy I will give this book to is a good friend of mine.’ d. Il coltello [con cui ho tagliato DET knife with which have.PRES.1SG cut.PRT la carne] si è rotto. (OBL) DET meat MIDD be.PRES.3SG break.PRT ‘The knife I cut the meat with has broken.’ etc. DET

donna woman

Before concluding this section, a final clarification is in order regarding the complementizer preceding relative clauses. We have glossed this element (English that, French que, Italian che, Persian ke, etc) as COMP, and it is especially important to avoid interpreting it as a relative pronoun: While this is the way it has been traditionally analyzed, morpho-syntactic and distributional evidence proves that this interpretation is incorrect. In fact, the formal properties of a C° head like that (and corresponding elements in other languages) are evidently different from those of a pronoun. A pronoun, as such, bears information on phi features (gender, person, number) and is marked for Case. For example, in Russian – a

172 The structure of the noun phrase language with morphological Case – the relative pronoun kotor – has these types of inflectional characteristics (see [109]), as opposed to that-type elements (Russian þto), which are typically invariable. In addition, heading an NP, a pronoun may be preceded by a preposition whenever required for Case assignment, while this is impossible for that, che and que.39 As we saw in (112), when the NP head has a function other than subject or direct object in English, that cannot be used. A relative pronoun is needed in this case, since it can be preceded by the preposition required to signal its indirect (or oblique) complement function: (123) a. Il

ragazzo [a cui darò questo boy to which give.FUT.1SG DEM libro] è un mio caro amico. POSS.1.SG dear friend book be.PRES.3SG INDEF.DET ‘The boy [to whom I will give this book] is a good friend of mine.’ b. *Il ragazzo [a che darò questo libro] è un mio caro amico. DET

In conclusion, an element like che is simply a functional head (in C°), and its function is to introduce the subordinate clause, just like in finite object clauses.

4.4.2.3. Types of relative clauses and their meaning As stated at the beginning of § 4.4.2., relative clauses can be classified as restrictive and appositive depending on their function, which is either to identify a referent or to provide additional information on the NP head. This semantic distinction (see de Vries 2001, Del Gobbo 2003, Rebuschi 2005 for further discussion) is reflected in interesting structural differences across languages. For example, appositive relative clauses appear almost universally postnominally, even in OV languages, where restrictive relative clauses are usually prenominal. This is the case in Afar (from Bliese 1981): (124) a. [A‘nu ub’le]RESTR aw’ki S.PRO.1SG see.PST.1SG boy.NOM cammi’se. throw.PST.3SGM ‘The boy that I saw threw a stone.’

’daa stone.ACC (AFAR)

Noun head modification 173 wee’ce. b. ‘awki [‘abba ge’da-hiyya]APPOS boy.NOM father go.PRES.3SGM-REL.PRO cry.PST.3SGM ‘The boy, who is going to his father, has cried.’ Second, in some languages appositive relative clauses are signaled by special introducing elements (here simply glossed as associative particles; see Frascarelli & Puglielli 2005a for details), which are lacking in restrictive clauses. This is what happens in Somali: (125) a. Wiil-ka [Maryan la hadlayaa] RESTR waa with talk.PRES.PROG.RED.NOM DECL boy-DET Maryan walaal-kay. (SOMALI) brother-POSS.1.SG ‘The boy that is talking to Maryan is my brother.’ waa b. Cali [oo Maryan la hadlayá]APPOS Cali ASSOC Maryan with talk.PRES.PROG.RED DECL walaal-kay. brother-POSS.1.SG ‘Cali, who is talking to Maryan, is my brother.’ Furthermore, restrictive relative clauses cannot be separated from the NP head because the NP is generated within the relative clause and moves to its Spec,CP. Appositives, on the other hand, can be separated. This contrast is shown in the English examples in (126)-(127) from Huck & Na (1990): (126) a. The man [that I want to introduce you]RESTR is here. b. *The man is here [that I want to introduce to you]. (127) a. A man [who had blond hair]APPOS entered the room [that I had just finished painting] RESTR. b. A man entered the room [that I had just finished painting]RESTR [who had blond hair]APPOS. A result of this property is that if an NP head is modified by both a restrictive and an appositive relative clause, the latter necessarily occupies a more external position. This restriction is shown in (128) for Somali:

174 The structure of the noun phrase [oo aan ku (128) a. Wiil-ka [hadlayaa]RESTR boy-DET talk.PRES.PROG.RED.NOM ASSOC SCL.1SG OCL.2SG baa Landan ka yimíd. (SOMALI) baray]APPOS introduce.PST.1SG FM London from come.PST.RED ‘The boy that is talking, who I introduced you to before, comes from London.’ b. *Wiilka [oo aan ku baray]APPOS [hadlayá]RESTR baa Landan ka yimíd. Based on these and other formal differences, a number of researchers have concluded that the structure of appositive relative clauses cannot be the same as that proposed for restrictive relative clauses, as in (107). The syntactic relationship between the NP head and the appositive relative clause is definitely not as close as that between the NP and a restrictive clause. This fully reflects the semantic differences previously described. Finally, from a prosodic point of view, the appositive clause appears as a parenthetical element, forming an independent prosodic domain with respect to the NP head and the root clause where it is merged (unlike restrictive relative clauses). All these differences led Frascarelli & Puglielli (2005b) to propose that the appositive relative clause is a declarative clause with predicative function – which accounts for its tendentially postnominal position – included in a Small Clause structure (see Chapter 2).40 To conclude the present description of relative clauses, at least two other extremely common types must be mentioned, namely, free relatives and correlative clauses. Free relatives are restrictive relative clauses whose NP head is not overtly realized. This type of relative clause is used if the head is an indefinite or generic NP, as in the following example: (129) Ø [chi arriva per primo] vincerà Ø who come.PRES.3SG for first win.FUT.3SG il premio. DET prize ‘(The one) who comes first will win the prize.’

(ITALIAN)

In (129) the NP head is phonologically null (as indicated by the Ø symbol), but still present for interpretation. The same construction is found in typologically different languages, for example Tagalog:

Noun head modification 175 [binili ng nanay (130) Nakita ko [DP ang Ø PAT.PST.see SCL.1SG TRIG buy.PST.PAT DIR mother para sa akin-g kaarawan]]. (TAGALOG) PRO.1SG-ASSOC birthday for OBL ‘I have seen what mother bought for my birthday.’ As we know, ang is the D° element that precedes the NP functioning as a trigger (see note 17 to Chapter 2). This clearly shows that ang is followed by a (phonologically null) NP in (130), which is the head of the relative clause. Finally, correlative clauses are relative clauses in which the noun head appears in its full form both in extraposed position and within the relative clause. This construction is typically used in Hindi (see [131], from Srivastav 1991), but also Italian (132) and Latin (see [133], from Pompei 2007) among others: (131) [jo

laRkii khaRii hai] vo laRkii girl standing be.3SG DEM girl lambii hai. tall be.3SG ‘The girl who is standing is tall.’ [lit.: ‘The girl who is standing, that girl is tall.’]

REL.PRO

(HINDI)

(132) Per for un

questo intervento ho inventato DEM speech have.PRES.1SG invent.PRT titolo [il quale titolo voleva INDEF.DET title DET which title want.IMPF.3SG riassumere una tematica summarize INDEF.DET topic molto complessa]. (ITALIAN) very complex ‘For this speech, I have come up with a title that was meant to sum up a very complex topic.’

(133) In castris Helvetiorum tabulae repertae sunt litteris Graecis confectae et ad Caesarem relatae [quibus in tabulis nominatim ratio confecta erat, qui numerus domo exisset eorum… (Caesar, The Gallic Wars, 1.29.1) (LATIN) ‘In the camp of the Helvetii, lists were found drawn up in Greek characters, and were brought to Caesar, in which an estimate had

176 The structure of the noun phrase been drawn up, name by name, of the number which had gone forth from their country...’ The type of relative structure in the Italian sentence (132) is used in high registers and depends on textual cohesion considerations with specific semantic-pragmatic functions.

5. Adverbial modification

5.1. Adverbs and adverbials: Preliminary considerations From a formal point of view, adverbs are usually described as invariable lexemes whose function is to modify predicates, other modifiers (adjectives or adverbs) or larger syntactic units, including clauses. Adverbs are thus undoubtly modifiers (a second-level function; see § 1.5.) but, unlike adjectives, they do not operate on nominal constituents. This is an important difference, as it determines an immediate categorial distinction on a distributional basis for those languages in which adverbs and adjectives are formally identical, e.g. German: (1)

a. Diese Arbeit war sorgfältig. (GERMAN) DEM.NOM work.NOM be.PST.3SG accurate ‘This work was accurate.’ b. Der Mann arbeitet sorgfältig. DET.NOM man.NOM work.PRES.3SG accurate ‘The man works accurately.’

As is shown, while in some languages adverbs are often identified through a specific derivational suffix (e.g., -ly in English, -ment in French, -mente in Italian, etc.), in German (predicative) adjectives and adverbs are distinguished on a purely distributional basis. However, the function of adverbial modifiers is not limited to these invariable lexemes. Let us consider the following examples: (2) a. b. c. d.

I have just arrived. I arrived one minute ago. I arrived in time. I arrived right after you came.

Noun phrases (2b), prepositional phrases (2c) and whole sentences (2d) may modify a predicate as well as an adverb proper (2a), conveying very similar meanings.

178 Adverbial modification Furthermore, in other languages such as German, this function can also be fulfilled by particles,1 i.e. functional heads like ja and doch, which activate specific meanings within the sentence:2 (3)

Sie

heiȕt be named.PRES.3SG ‘Of course her name is Petra!’ PRO.3SGF

(4)

A: Du

bist be.PRES.2SG ‘You’re not ready yet!’ B: Ich bin PRO.1SG be.PRES.1SG ‘Of course I am!’ PRO.2SG

ja yes

Petra! Petra

noch still

nicht not

doch sure

fertig! ready

(GERMAN)

fertig! ready

In this work, the term adverbial modifier (or simply adverbial) will be used to refer globally to the various elements that can function as adverbs. This set will include the heads of adverbial phrases (AdvP) (adverbs proper), as well as every other constituent (DP, PP, CP) used with this function.3

5.2. Properties and distribution of adverbial modifiers A distinctive feature of adverbial modifiers seems to be their accessory character. What they do is add information to the predication and their absence does not make the sentence ungrammatical or incomplete.4 Adverbials might be thus considered circumstantial constituents par excellence and their structural position could be easily ascribed to the overgenerating mechanism of adjunction. As we saw in Chapter 1 (§ 1.2.1.), this mechanism involves the replication of a maximal category, to which a non-argument constituent (i.e. a constituent that is not selected by the verb) is adjoined. This constituent supplies additional information regarding the head of the relevant phrase. Based on their position in the sentence, adverbials might be thus analyzed as adjoined to the VP, to the sentential node (IP) or to the CP, with no real restriction on their number and mutual order. Hence, in the following sentences, the adverbial tomorrow may be (left- or right-) adjoined (as adjunct nodes have no predetermined direction) to the CP (5a), the IP (5b), the VP (5c) or to any of these constituents (5d). Similar

Properties and distribution of adverbial modifiers 179 considerations could apply to probably, which would be left- and rightadjoined to the CP in (5b) and (5d) respectively, adjoined to the IP in (5a) and to the VP in (5c): (5)

a. b. c. d.

Tomorrow, probably Jim will come. Probably, Jim will come tomorrow. Jim will probably come tomorrow. Jim will come tomorrow, probably.

Adjunction appears to be extremely easy to apply, as it only relies on the linear order of non-obligatory constituents. However, admitting that some constituents can be merged in several positions – and various directions – on the basis of linear order alone is a serious problem for a theory that aims at identifying the universal principles governing language acquisition (i.e. Universal Grammar). Apart from theoretical considerations, we also noticed above that even some (nominal) modifiers may be necessary to identify referents (see § 4.1.2.) and that APs are ordered according to a rigid hierarchy in languages. Over time, this has led scholars to get rid of adjunction for APs and assume a hierarchy of functional projections (between NP and DP) whose heads are related to specific semantic features expressed by the different adjectives (see § 4.4.1.). In this chapter, we will argue that something similar holds for adverbial modifiers. This will help further reduce the role of adjunction and restrain its over-permissive character. In our discussion on the nature and position of modifiers, we will consider typologically different languages, starting with Italian. It is often maintained that adverbial modifiers can be inserted at any point of the sentence. In fact, even a cursory analysis will show that some positions are ruled out for some types of adverbials, and that different interpretations can be associated with different positions. In other words, the position of adverbials is anything but random, as will become clear from the following Italian examples: (6)

a. Leo parla bene l’ inglese. Leo speak.PRES.3SG well DET English ‘Leo speaks English well.’ b. *Leo bene parla l’ inglese. Leo well speak.PRES.3SG DET English L’ INGLESE. c. Leo, bene, parla Leo well speak.PRES.3SG DET English ‘Leo speaks ENGLISH well.’

(ITALIAN)

180 Adverbial modification (7)

a. Leo parla sempre di sé. always of REFL Leo talk.PRES.3SG ‘Leo always talks about himself.’ b. *Leo sempre parla di sé. of REFL Leo always talk.PRES.3SG c. Leo, SEMPRE parla di sé. Leo always talk.PRES.3SG of REFL ‘Leo ALWAYS talks about himself.’

(8)

a. Leo parlerà chiaramente con tutti. Leo speak.FUT.3SG clearly with everyone ‘Leo will speak clearly (= in a clear way) to everyone.’ b. Leo, chiaramente, parlerà con tutti. Leo clearly speak.FUT.3SG with everyone ‘Clearly (= as is clear), Leo will speak to everyone.’

(9)

a. Visto che/poiché sei un amico, friend since be.PRES.2SG INDEF.DET ti parlerò sinceramente. IOCL.2SG speak.FUT.1SG honestly ‘Since you are my friend, I will be honest with you.’ b. Ti parlerò sinceramente IOCL.2SG speak.FUT.1SG honestly visto che/perché sei un amico. friend because be.PRES.2SG INDEF.DET ‘I will be honest with you because you are my friend.’

An adverb like bene cannot be preverbal (6b), unless it is topicalized (6c) and thus forces a contrastive interpretation of the nominal constituent conveying new information (English in [6c]; on Topic and Contrast see Chapter 6). The same is true for a time adverb like sempre (7b), unless it is contrastively focused (as in [7c], which is interpreted as: Leo ALWAYS talks about himself, not just every now and then). An adverb like chiaramente acts as a modal modifier of the predicate when it is postverbal (8a), whereas it assumes evidential modality over the sentence when preverbal (8b). Finally, a causal adverbial clause (9a-b) is interpreted as given information if it precedes the root sentence (being the reason for what is expressed therein) and as new information if it follows it. Interestingly, different subordinators may be necessary for the two cases (poiché ‘since’ vs. perché ‘because’).

Properties and distribution of adverbial modifiers 181 These few examples lead to the conclusion that adverbials do not modify the whole predication, but parts of it, contributing information on specific aspects (manner, time, modality, speaker’s attitude, etc.). It is therefore clear that the syntax and interpretation of adverbial modifiers cannot be explained using adjunction.5 Various theories have been formulated to account for these complex phenomena (cf. among others, Quirk et al. 1985, Nølke 1990, Lonzi 1991, Ernst 2002). Within different theoretical frameworks, these theories have aimed at defining formal properties and semantic relationships of adverbial elements. Within Functional Grammar (Hengeveld 1987, 1989, Dik 1997), for instance, the sentence is considered a layered structure comprising a set of hierarchically ordered levels. Under this theory, adverbial expressions are considered as satellites, which operate at semantically-based modification levels. A semantic layered structure is also proposed in Ramat & Ricca (1998), using the traditional bipartition in high and low adverbs6 (thus described with reference to their position in the sentence). The higher level includes (a) connective adverbs (e.g. anyway), (b) speech act adverbs (e.g. frankly), (c) domain adverbs (e.g. linguistically), (d) modal adverbs (such as probably, evidently, hopefully) and sentential adverbs (eventevaluating, such as fortunately or subject-evaluating, such as kindly). The lower level includes event-oriented adverbs (time, aspect, locative adverbs) and predicative adverbs (e.g. completely). This is clearly a semantics-based classification, which shows the need for an ordering hierarchy for adverbs, although no precedence relationship within each of the two groups is defined. In the following sections, we will present examples from different languages to show that a hierarchy based on semantic properties is required, as well as finer structural distinctions within the individual domains. In doing so, we will begin by describing and analyzing adverbial phrases.

5.3. Structural insertion of adverb phrases Apart from considerations relating to language families, basic constituent orders (SVO, SOV, VSO) and morphological types, the world’s languages clearly show that the order of AdvPs is definitely not free within the sentence. On the contrary, their position is fixed and their mutual order is constrained.7 Let us consider data, starting with two SVO languages, an isolating language like Chinese (from Li & Thompson 1981) and an inflecting one like Modern Greek (from Alexiadou 1997):

182 Adverbial modification (10) AdvSpeechAct> AdvEval a. Laoshi-shuo wǂ buxing dui tƗmen PRO.1SG unfortunately toward PRO.3PL honestly you pian-jian. (CHINESE) have prejudice ‘Honestly, unfortunately I am biased against them.’ a'. *Buxing wǂ laoshi-shuo dui tƗmen you pian-jian. AdvPerfAsp > AdvRetrospAsp b. Mei ci wǂ pengjian tƗ, tƗ zongshi PRO.3SG PRO.3SG always each time PRO.1SG meet ganggang cong guowai huilai. just from abroad return ‘Every time I see her/him, s/he has just been abroad.’ b'. *Mei ci wǂ pengjian tƗ, tƗ ganggang zongshi cong guowai huilai. (11) AdvEval > AdvEpistMod a. Efithos o Yanis pithanos tha figi. (MODERN GREEK) luckily DET Yanis probably FUT go.3SG ‘Fortunately, Yanis will probably leave.’ a'. *Pithanos o Yanis efithos tha figi. AdvHabitAsp> AdvComplAsp b. Katalavene sinithos kala completely understand.IMPF.3SG usually ta themata. DET.N.PL.ACC problem.N.PL.ACC ‘Usually, he fully understood the problems.’ b'. *Katalavene kala sinithos ta themata. These sentences show that in both Chinese and Modern Greek, adverbs belonging to the same semantic domains require obligatory positions and mutual orders. In Chinese, for example, a speech act adverb must precede an evaluative adverb, even though both are high adverbs; similarly, the equivalent of always must precede just, although both are low aspectual adverbs. In Greek, evaluative adverbs must precede those expressing epistemic modality (both being high adverbs), habitual adverbs must precede completion ones, as far as aspectual adverbs are concerned.

Structural insertion of adverb phrases 183 Five pairs of sentences from two languages with SOV properties will be compared in the following examples. The first of each pair is from Finnish, an agglutinative language, the second from German, an inflecting language.8 Data show that in these two languages too, a hierarchy must be assumed: (12) AdvEval> AdvEviden a. Onne-ksi Per on ilmeise-sti lähte-nyt. (FINNISH) lucky-ADV Per be.PRES.3SG evident-ADV leave-PRT.PST.SG ‘Fortunately, Per has clearly left.’ a'. *Ilmeisesti Per on onneksi lähtenyt. b. Peter ist glücklicherweise offensichtlich luckily evident Peter be.PRES.3SG gegangen. (GERMAN) go.PRT ‘Fortunately, Peter has clearly left.’ b'. *Peter ist offensichtlich glücklicherweise gegangen. (13) AdvEviden > AdvEpistMod a. Per ilmeise-sti ymmär-tää mahdollise-sti hyvin Per evident-ADV understand-3SG probable-ADV well onglema-n. (FINNISH) problem-ACC ‘Clearly, Per probably understands the problem well.’ a'. *Per mahdollisesti ymmärtää ilmeisesti hyvin ongleman. a''. *Per hyvin ilmeisesti ymmärtää mahdollisesti ongleman. b. Peter versteht offensichtlich das Problem DET.N problem Peter understand.PRES.3SG evident wahrscheinlich gut. (GERMAN) probable well ‘Clearly, Peter probably understands the problem well.’ b'. *Peter versteht wahrscheinlich gut offensichtlich das Problem. (14) AdvIrrealis > AdvHabilMod > AdvAnterT a. Per on ehkä viisaa-sti jo already Per be.3SG maybe wise-ADV lähette-tty kotti-in. home-LOC send-PRT.PASS ‘Maybe Per has wisely already been sent home.’ a'. *Per on viisaasti ehkä jo lähettetty kottiin.

(FINNISH)

184 Adverbial modification a''. *Per on jo ehkä viisaasti lähettetty kottiin. b. Peter wurde möglicherweise klugerweise bereits already Peter AUX.PASS.PST.3SG possible.way wise.way nach Hause geschickt. (GERMAN) to home.DAT send.PRT ‘Maybe Peter has wisely already been sent home.’ b'. *Peter wurde klugerweise möglicherweise bereits nach Hause geschickt. b''. *Peter wurde bereits möglicherweise klugerweise nach Hause geschickt. (15) AdvHabitAsp > AdvAnterT a. Kahd-e-lta Per on yleensä two-SG-ABL Per be.3SG usually jo syö-nyt. (FINNISH) already eat-PRT ‘At two o’clock, Peter has usually already eaten.’ a'. *Kahdelta Per on jo yleensä syönyt. b. Um zwei Uhr hat Peter normalerweise at two hour have.PRES.3SG Peter normal-way bereits gegessen. (GERMAN) already eat.PRT ‘At two o’clock, Peter has usually already eaten.’ b'. *Um zwei Uhr hat Peter bereits normalerweise gegessen. (16) AdvPerfAsp > AdvComplAsp a. He ymmär-tav-at aina täysin S.PRO.3PL understand-PRES-3PL always totally mi-stä puh-un.9 (FINNISH) REL.PRO.INDEF-ELA talk-1SG ‘They always totally understand what I am talking about.’ a'. *He täysin aina ymmärrä mistä puhun b. Sie verstehen immer vollständig S.PRO.3PL understand.PRES.3PL always complete das, was ich sage. (GERMAN) DEM.ACC.SGN REL.PRO.ACC.SGN S.PRO.1SG say.PRES.1SG ‘They always totally understand what I am saying.’ b'. *Sie verstehen vollständig immer das, was ich sage.

Structural insertion of adverb phrases 185 These examples show that a hierarchical order has to be assumed for adverbs in all semantic domains. In particular, it emerges that evaluative adverbs precede evidential modality adverbs (12), and these, in turn, precede epistemic modality adverbs (13). In addition, irrealis modality precedes habilitative modality (14), and both modal and aspectual adverbs precede anteriority temporal adverbs (14)-(15). Finally, as noticed for Greek, completive aspect occupies the lowest position in the hierarchy (16). Let us now examine how adverbs behave in Tagalog, a partly isolating VSO language (where inflection only appears on the verbal head) and Arabic, an inflecting language.10 For the sake of simplicity, we present equivalent sentences of the Chinese and Modern Greek examples (10)-(11). The hierarchy remains the same: (17) AdvSpeechAct > AdvEval a. Sa totoo lang, hindi ako nagkaroon OBL honestly unluckily NEG PRO.1SG.TRIG possess nang maganda-ng saloobin sa kanila. (TAGALOG) REINF good-ASSOC attitude OBL PRO.OBL.3PL ‘Honestly, I am unfortunately biased against them.’ a'. *Lang sa totoo, hindi ako nagkaroon nang magandang saloobin sa kanila. b. Bi-sarƗcat-i-n ’ind-Ư ma’a al-asaf-i with-honesty.F-OBL-INDEF at-O.PRO.1SG with DET-sorrow-OBL tahƗmulƗt-u-n sayyi’at-u-n prejudice.PLF-NOM-INDEF bad.SGF-NOM-INDEF ’an-hum. (ARABIC) on-O.PRO.3PL ‘Honestly, I am unfortunately biased against them.’ b.’ *Ma’a al-asafi’indii bi-sarƗcatin tahƗmulƗtun sayyi’atun ’anhum. (18) AdvEval> AdvEpistMod a. Sa kabutihang-palad, si Juan ang malamang OBL goodness-luck TRIG Juan TRIG probably a-alis. (TAGALOG) FUT-leave ‘Fortunately, Juan will probably leave.’ a'. *Ang malamang si Juan sa kabutihang-palad aalis.

186 Adverbial modification al-cazz-i Ahmadu min b. Li-cusn-i for-beauty-OBL DET-luck-OBL Ahmad.NOM from sayamdƯ. (ARABIC) al-muctamili DET-possible.OBL FUT.go.IMPF.3SGM ‘Fortunately, Ahmad will probably leave.’ b'. *Min al-muctamili Ahmadu li-cusni al-cazzi sayamdƯ. (19) AdvHabitAsp > AdvComplAsp a. Karaniwan niya lubusa-ng usually PRO.3SG totally-ASSOC nauunawaan ang mga usapin. (TAGALOG) understand.PAT.PST TRIG PL problematic case ‘Usually, he totally understood problems.’ a'. *Lubusang niya karaniwan nauunawaan ang mga usapin. b. Huwa ’Ɨdiyy a-n kƗna S.PRO.3SGM habitual.ACC-INDEF be.PERF.3SGM yafhamu al-mašƗkila. understand.PST.3SGM DET-problem.PL.ACC bi-kamƗlihƗ. (ARABIC) with-totality.OBL-O.PRO.3SGM ‘Usually, he totally understood problems.’ b'. *Huwa bi-kamƗlihƗ kƗna yafhamu al-mašƗkila ’Ɨdiyyan. While linguists like Ramat & Ricca (1998) argue that the unmarked (or preferential) order of adverbs is related to their semantic properties and distinguish between various macro-levels of modification, data analysis shows that adverb order is much more rigid than can be predicted on the basis of semantic domains. For this reason, within Generative Grammar, authors like Alexiadou (1997) and Cinque (1999) opposed adjunctionbased analyses and proposed a universally valid syntactic ordering. In particular, Cinque (1999) presented a hierarchy of functional projections, in which adverbs are rigidly ordered. Under this analysis, AdvPs are not merged as adjuncts, but as Specifiers of the functional projections comprising complex sentential structure. Let us discuss this last point briefly before we proceed with our analysis. Following Pollock’s (1989) influential work, in the late Eighties the IP sentential node was divided into a number of functional projections, each associated with a specific piece of information regarding the inflection of the verbal head (e.g. Tense, Mood, Aspect, etc.). The hierarchical order of these projections was defined in the following years thanks to cross-

Structural insertion of adverb phrases 187 linguistic studies (especially on agglutinative and polysynthetic languages). A fixed linear order of verbal functional morphemes was defined, whereby subject agreement is consistently more external (i.e. higher) within the verbal head than object agreement (wherever present), tense morphemes are normally more external than aspect morphemes and so on. Instead of a single IP node, representations such as that in (20) below began to be used in the structural analysis of the inflection area, where AgrSP is the functional projection where subject agreement takes place and NOM Case is assigned (Spec,AgrSP being equivalent to Spec,IP), TP is the tense phrase, AspP the aspect phrase and AgrOP the phrase devoted to object agreement: (20)

AgrSP AgrS’ Agr° [SUBJ Agr] [+NOM]

TP T’ T° [+tense]

AspP Asp’ Asp° [+aspect]

AgrOP AgrO’ AgrO° [DO Agr]

VP

Given this structure and the Locality principle,11 the realization of verbal functional categories does not entail a single instance of movement (as for I° earlier), but a cyclic movement of the verbal head along the various functional heads, resulting in the incorporation of the features contained in each of them. In this process, the first morpheme to be incorporated is necessarily the one closest to the VP; conversely, the last to be incorporated appears in a more external position. This is known as the ‘Mirror Principle’ (Baker 1996), and accounts for the fact that the subject agreement morpheme is normally the most external morpheme on the verbal head. As an illustration, consider the derivation of the verbal head in a polysynthetic language like Yareba (from Weimer 1972):

188 Adverbial modification (YAREBA)

(21) Yau-r-eb-a-su. sit.CLASS-HAB-PST-3SGM ‘He used to sit down.’ (22)

AgrSP yaur-eb-a-su

AgrS’ TP T’

yaur-eb-a

AspP Asp’ yaur-eb

VP V’ yaur

As the diagram in (22) shows, during the derivation the verb passes through all the functional heads making up the complex sentential structure, and incorporates any morpheme found in its cyclic movement. In this process the morpheme incorporated first is the closest to the lexical root, whereas the highest morpheme in the structure is necessarily the outermost. This digression was aimed at illustrating two crucial points in language analysis. First, it shows that sentential structure is considerably richer than was hitherto assumed for ease of exposition, as it comprises a hierarchy of functional projections, each responsible for the realization and interpretation of an individual verbal category. Secondly, we illustrated the process of incorporation, an extremely useful tool for understanding the morphological structure of words in polysynthetic and agglutinative languages. Let us now return to adverbs. As we said, AdvPs are merged as Specifiers of functional projections. This means that each adverb corresponds to a functional head that codifies a semantic feature relating to the meaning of the adverb in question. For example, often is merged in an aspect-related functional projection, probably is associated with epistemic modality, etc. Based on cross-linguistic analysis, Cinque (1999) identified as many as 24 syntactic positions, each related to a semantic specification. These are listed below, with an example for each functional position (a bracketed notation is used for ease of exposition):12

Structural insertion of adverb phrases 189 (23) [MoodP frankly [Mood°[+SpeechAct] [MoodP unfortunately [Mood°[+evaluative] [MoodP evidently [Mood°[+evidential] [ModP probably [Mod°[+epistemic] [TP then [T°[+pst/fut] [MoodP perhaps [Mood°[+irrealis] [ModP necessarily [Mod°[+necessity] [ModP possibly [Mod°[+possibilty] [ModP gladly [Mod°[+volitional] [ModP obligatorily [Mod°[+obligation] [ModP wisely [Mod°[+ability] [AspP usually [Asp°[+habitual] [AspP again [Asp°[+repetitive] [AspP often [Asp°[+frequentative] [TP already [T°[+anteriority] [AspP no longer [Asp°[+terminative] [AspP still [Asp°[+continuative] [AspP always [Asp°[+perfect] [AspP just [Asp°[+retrospective] [AspP long [Asp°[+durative] [AspP almost [Asp°[+prospective] [AspP soon [Asp°[+progressive] [AspP completely [Asp°[+completive] [VoiceP well [Voice° [+voice] As we can see, the verbal categories Mood (MoodP), Tense (TP), Modality (ModP), Aspect (AspP) and Voice (VoiceP) are divided into functional projections, whose head hosts a semantic feature expressed by the adverb (AdvP) merged into Specifier position. So, a sentence like (24) corresponds to the tree diagram in (24’): (24) Leo unfortunately maybe again totally messed up his test. (24’)

AgrSP

AgrS’ MoodP Leo Mood’ unfortunately MoodP Mood° Mood’ [+evaluative] AspP Mood° maybe Asp’ [+irrealis] again

Asp° [+repetitive] totally

AspP Asp’

Asp° VP [+completive] tLeo messed up his test

190 Adverbial modification According to Cinque’s (1999) proposal, the hierarchy in (23) is universal. It accounts for data from various languages, and possible alternative (or contrasting) orders may be explained on the basis of pragmatic considerations.13 Besides avoiding adjunction and accounting for obligatory orders in different languages, this analysis also has another advantage: The order of TAM particles in isolating languages reflects adverb order. In other words, in languages where TAM information is not incorporated on the verbal head as inflectional or agglutinative morphemes but appears as free morphemes, these morphemes present the same order as the functional projections in (23). It follows as a logical conclusion that these particles occupy the functional heads corresponding to the Spec positions where adverbs are merged. Such particles are especially frequent in creole languages (see [25] from Gibson 1986 and [26] from Jones 1971) and in many Niger-Congo languages (see [27], from Omamor 1982): (25) Shi

aalweez de a sing. (GUYANA CREOLE) always DUR PROG sing ‘She usually always keeps singing.’

PRO.3SGF

(26) A PRO.1SG

a

HAB

bin

m2 s

d2 n

de

PST

MOD

PERF

PROG





rait. write

(KRIO)

‘I should have been writing.’ (27) O

rè rèn. go PERF ‘He will have started going habitually.’ PRO.3SGM

waá



FUT

HAB

(ISEKIRI)

As can be verified by comparing these sentences with the hierarchy in (23), the particles follow the expected order. Furthermore, (25) shows that a single semantic feature can be realized by an adverb rather than a functional head, as in the case of aalweez (‘always’). The diagram for (25) is the following:

Structural insertion of adverb phrases 191 (28) shi

AgrSP AgrS’ AspP Asp’ a [+habitual] aalweez

AspP Asp’ Asp° [+perfect]

AspP Asp’ de [+durative]

AspP Asp’

a [+progressive] VP tshi sing One last observation is in order regarding agglutinative and polysynthetic languages. Here, the linear order of the relevant suffixes is the mirror-image of the order in (23). This is expected according to the Mirror Principle, previously described and illustrated in (22), and confirms the validity of the proposed hierarchy. Moving upwards along the functional projections, the verb cyclically incorporates the different morphemes, thus obtaining a mirror ordering. Let us look at the following examples from Korean, an agglutinative language (Sohn 1994) and Macushi, a polysynthetic language (Abbot 1991):14 (29) Minca-nun ttena-ss-te-kwun-yo. Minca-TOP leave-PST-EVID-EVAL-POLITE ‘Evidently, Minca has fortunately left.’ (30) I’-po’-pîtî-tu’ka-‘pî-i-ya. SCL.3SG-whip-ITER-PST-finally-OCL.3SG-ERG ‘S/he finally whipped him/her.’

(KOREAN)

(MACUSHI)

In (30), the subject pronoun (I’-) is phonologically cliticized onto the verbal head (po’ ‘whip’), whereas the agent-subject agreement morpheme (ERG Case) is the outermost, as expected on the basis of the split-IP hierarchy (see [20]). To conclude, the derivation of the verbal head (ttenass-te-kwun-yo) in (29) is shown below:

192 Adverbial modification (31)

MoodP Mood

ttena-ss-te-kwun-yo [+polite]

MoodP Mood’ MoodP Mood’ TP ttena-ss-te T’ [+evidential]

ttena-ss-te-kwun [+evaluative]

ttena-ss [+pst]

VP V’ ttena

Unexpected linear orders, according to Cinque, can be ascribed to four possible reasons: (i) the adverb modifies another adverb; (ii) the adverb has moved to the CP area as a questioned element; (iii) the adverb is used as a Focus; (iv) the adverb is merged in different positions as it corresponds to different meanings (as for some adjectives; see the Italian example un amico vecchio vs. un vecchio amico, cf. Chapter 4, [81a-b]). This means that, except cases (i) and (iv), deviant orders are normally caused by focalization processes (Cinque 1999: 27). Cinque, however, does not discuss these phenomena any further. The possibility for adverbs to move to a marked, sentence-initial position is also argued by Rizzi (2004), who proposes a functional category dedicated to this type of movement in the left periphery of the sentence, i.e. MoodP. Adverb movement is thus associated with a rather generic prominence requirement: “We assume that the left periphery can contain dedicated Mood heads which can host adverbs as their Specs; the functional motivation for such heads is that they make the moved adverb […] ‘prominent’, ‘salient’ (as opposed to ‘ground’)” (Rizzi, 2004: 241)

Recent studies on spoken corpora, however, have shown that adverbs may occupy unexpected positions as a consequence of informationstructural requirements (Bonvino, Frascarelli & Pietrandrea 2007). In fact, the analysis of intonational contours proves that there is a relationship between pragmatic (or semantic) functions and tonal events which show the

Structural insertion of adverb phrases 193 properties recently associated with specific discourse categories (i.e. Topic, Focus, Comment; see Chapter 6).15 Adverbs can thus be left- or righttopicalized (see [32]-[33]), focused (see [34] (examples taken from De Mauro et al. [eds.] 1993) and may also convey new information (the Comment) in the sentence (see [35], from Bonvino 2006): (32) Spesso la borsa di studio è often DET grant of study be.PRES.3SG roba simile al contratto. (LIP, RC6) stuff similar to.DET contract ‘Often, a grant is something like a contract.’

una INDEF.DET (ITALIAN)

(33) Esatto, lei è un po’ exactly she be.PRES.3SG INDEF.DET little timida, probabilmente. (LIP, RA1) shy probably ‘Exactly, she is a little shy, probably.’ (34) Ma lui sempre studia, but he always study.PRES.3SG sempre fa i compiti. (LIP, RA9) always do.PRES.3SG DET homework.PL ‘But he ALWAYS studies, he ALWAYS does his homework.’ (35) A: Ma il fatto che ci sono delle that there be.PRES.3PL of.DET but DET fact immagini che ci sono delle cose image.PL that there be.PRES.3PL of.DET thing.PL non ti aiuta a capire la parola not OCL.2SG help.PRES.3SG to understand.INF DET word maggiormente che in un testo? more than in INDEF.DET text ‘The fact that there are pictures, that there are things, doesn’t it help you understand the word better than in a text?’ B: Sì, quello sicuramente. yes that certainly ‘Yes, definitely.’ Based on the data presented so far, we can conclude that adverbs (or rather the AdvPs they head) are merged as Specifiers of dedicated

194 Adverbial modification functional projections within the area of the original IP node, depending on their semantic properties. Adverbs do not modify predicates in general, but a specific category of the verbal head, i.e. the relevant functional head. This head is the place where inflectional morphemes and modal particles with similar meanings are merged. Adverbs thus follow a specific order with variations required by discourse grammar.

5.4. Noun and prepositional phrases with adverbial function As already said in § 5.1., different phrases, i.e. NPs, PPs and CPs, can have an adverbial function in a number of languages. In this section, we will focus on the first two.16 Noun and prepositional phrases with adverbial functions modify the predicate with regard to various types of information. In particular, the verbal categories of Time, Place and Manner can be modified: (36) a. Incontrerò Leo domani. meet.FUT.1SG Leo tomorrow ‘I am going to meet Leo tomorrow.’ b. Incontrerò Leo all’ università. meet.FUT.1SG Leo at.DET university ‘I am going to meet Leo at the university.’ c. Incontrerò Leo con molto piacere. meet.FUT.1SG Leo with much pleasure ‘I am going to meet Leo with great pleasure.’ Unlike adverbs proper, these constituents are usually found at the beginning or at the end of the sentence: While an intermediate position is also possible, it is perceived as marginal: (37) a. b. c.

??

Incontrerò domani Leo. Incontrerò all’università Leo. ?* Incontrerò con molto piacere Leo. ??

The intermediate position of adverbial phrases does not make the relevant sentences ungrammatical. However, they are not appropriate as allnew sentences although they would be perfectly acceptable in a context where the adverbial element is new information and the object is produced after a short pause (as a marginalized constituent; cf. Cardinaletti 2002,

NPs and PPs with adverbial function 195 Frascarelli 2008), that is to say, as marked sentences. This means that apparently adverbial NPs and PPs cannot occupy the Spec positions of the functional projections that make up the sentence, although they fulfill a function similar to adverbs. This difference can be explained as a consequence of the prosodic heaviness of these constituents. Being normally longer than one word, they are preferentially placed sentence-finally (see Nespor 1991). In initial position, on the other hand, they need a pause to separate them from the rest of the sentence and this makes them dislocated elements. Another possible explanation for this difference was indicated by Aoun (1985): As they are nominal (rather than adverbial) elements, these constituents need to be marked for a syntactic function. Indeed, in languages with morphological Case, they are Case-marked (see [38]). Hence, when in a sentence-median position, they may interfere with Case assignment to other constituents (e.g. ACC Case in sentences [39]-[40]): (38) Montags /am Montag Monday.GEN at.DET.DAT Monday.DAT ich Maria. S.PRO.1SG Maria ‘I am meeting Maria on Monday.’

treffe meet.PRES.1SG (GERMAN)

(39) a. John will read the letter tomorrow. b. *John will read tomorrow the letter. (40) a. For John/him to read the letter tomorrow is impossible. b. *For tomorrow John/him to read the letter is impossible. Whatever the explanation may be, it is clear that even though these elements convey verb-related information (such as Time and Manner), they cannot appear in IP-internal positions (unlike adverbs). Does this mean that adverbial NPs/PPs are not merged in these functional projections? Or, rather, that they are merged there (where they are licensed depending on the functional feature they instantiate) but need to move to the periphery? Some data seem to point to the latter option. The first observation to be made is that these adverbial elements can see the temporal feature encoded in the verbal head in the same way as the inflected verbal head (in I°) can see the phi features (person, gender and number) of the constituent in Spec,IP (i.e. the subject). Consider the following Italian sentences:

196 Adverbial modification (41) a. Gianni incontra Maria domenica. Gianni meet.PRES.3SG Maria Sunday ‘Gianni will meet Maria on Sunday.’ b. Gianni ha incontrato Maria domenica. Sunday Gianni have.PRES.3SG meet.PRT Maria ‘Gianni met Maria last Sunday.’ Since such feature sharing typically results from a Spec-head relation, data like those in (41) lead us to assume that the adverbial NP domenica is generated in Spec,TP and enters a Spec-head relation with the [+pst/fut] functional feature (see [23]), thus obtaining the correct interpretation regarding the time at which the action expressed by the predicate takes place (i.e. in the past or in the future). From this preverbal position (which is always possible, as in Giovanni domenica incontra Maria), the NP subsequently moves to a peripheral position. If preceded by an article, an adverbial NP can also be associated with an aspectual feature. The article activates the NP functional area, which hosts a number of elements including quantifiers (in Spec,DP position; see section 4.3.). When an adverbial NP like domenica follows an article, it is associated with a [+habitual] quantified piece of information, as shown in the following sentence: (42) Gianni incontra Maria la domenica (= ogni domenica). each Sunday Gianni meet.PRES.3SG Maria DET Sunday ‘Gianni meets Maria on Sundays.’ Based on what has been said so far, there is solid evidence in favour of adverbial NPs not just being extrasentential adjuncts, but elements merged in the Spec of functional projections. The fact that their interpretation conflicts with the interpretation of pure adverbs when the latter activate contrasting functional features, confirms this. Therefore, a temporal NP (expressing a punctual action in the past or in the future) cannot be combined with a frequency aspectual adverb like spesso, ‘often’ (43a). On the other hand, an aspectual (+DET) NP such as la domenica, ‘Sundays’, in (43b) can: (43) a. *Gianni domenica incontra spesso Maria. b. Gianni la domenica incontra spesso Maria.

NPs and PPs with adverbial function 197 Similarly, a habitual aspectual DP (see [42]) cannot be employed when the verb expresses a punctual (rather than habitual or reiterated) action in the past or in the future (44): (44) *Gianni incontrerà/ha incontrato Maria la domenica. Languages like German, where right-hand extraposition is only permitted for PPs (and marginally so), provide further evidence that temporal NPs are generated in Spec,TP. In fact, linear order clearly shows that these NPs must occupy a higher position than Manner adverbs, according to the hierarchy in (23). Let us consider the following sentence: (45) a. …weil Hans gestern den Brief DET.ACC letter.ACC because Hans yesterday sorgfältig geschrieben hat. (GERMAN) have.PRES.3SG carefully write.PRT ‘[...] because Hans wrote the letter carefully yesterday.’ b. *…weil Hans sorgfältig den Brief gestern geschrieben hat. What has been argued so far for adverbial NPs applies to adverbial PPs: These are also generated in the IP area and are then extraposed to a peripheral position. Some cross-linguistic data can support this assumption. A type of information that is usually expressed by means of adverbial PPs concerns the location of events. Yet, this is not included as one of the verbal functional categories proposed by Cinque (1999) (see [23]). This category, however, must appear in the IP area, as proved by polysynthetic languages like Mohawk. In this language locative information is expressed by a morpheme on the verbal head which occupies a lower position than the polyfunctional morpheme ‘Λ’ whenever the latter conveys temporal information (future in the past; see [46]); conversely, it is higher than that morpheme when the latter expresses aspectual (habitual) information (see [47]). Consider the following examples (from Baker 1996):17 (46) Wak-ateryΛtare-" tsi akwéku Λ-t-hun-e-" OCL.1SG-know-PERF that all FUT-LOC-SCL.1PL-come-PERF ne sha’-khe-hukarΛ-". (MOHAWK) ASSOC FACT-SCL.1SG/OCL.3PL-invite-PERF ‘I learned they would all come (here) when I invited them.’

198 Adverbial modification

Λ-yùk-kΛ nònΛ (47) Tyótku uhkàk HAB-SCL.1SG.OCL.1SG-see when always someone kanàta-ku y-Λ-k-e-". town-to LOC-HAB-SCL.1SG-go-PERF ‘Someone always sees me when I am going to town.’ Based on these data, we can assume that the functional projection expressing locative information occupies an intermediate position in the IP area, between TP [+pst/fut] and AspP [+habitual]. In any case, this projection must be higher than Manner adverbs, as shown by the fixed order (Time>Place>Manner) of German (cf. Hinterhölzl 2000): (48) a. …weil Hans gestern in der Universität because Hans yesterday in DET.DAT university.DAT das Buch sorgfältig gelesen hat. (GERMAN) DET.ACC book.ACC carefully read.PRT have.PRES.3SG ‘[…] because Hans read the book carefully at the university yesterday.’ b. *… weil Hans gestern sorgfältig das Buch in der Universität gelesen hat. c. *… weil Hans in der Universität sorgfältig das Buch gestern gelesen hat. We can conclude that adverbial NPs and PPs are also generated in a hierarchical order, in particular in the following Specifier positions: (49) a) ‘TEMPORAL’ NP/PP b) ‘LOCATIVE’ NP/PP c) ‘MODAL’ NP/PP

= SpecTP or SpecAsp [+habitual] = an intermediate functional projection between Time and Manner adverbs = SpecVoiceP

Although in Italian, unlike in German, right-hand extraposition of adverbial NPs/PPs is always possible (or even preferred), the assumption that adverbial NPs/PPs are generated according to a specific hierarchy is confirmed by coreference relations. As is known, for a nominal referent to bind a possessive pronoun, the former needs to assume scope over the latter (on the notion of scope, cf. Chomsky 1976, May 1977, 1985).18 This scope condition remains unchanged after movement of argument NPs, as shown by the following sentences:

NPs and PPs with adverbial function 199 il suok/j libro (DO) (50) a. Mariak darà Maria give.FUT.3SG DET POSS.3.SG book a Giannij (IO). to Gianni ‘Maria is going to give her/his book to Gianni.’ b. Mariak darà a Giannij (IO) Maria give.FUT.3SG to Gianni il suok/j libro tj (DO). DET POSS.3.SG book ‘Maria is going to give Gianni her/his book.’

(ITALIAN)

As is shown by indices, the antecedent of the possessive suo in (50a) may either be the DP Maria or the DP Gianni, as both the subject DP and the PP in V’, containing the DP Gianni, occupy a structurally higher position in the diagram than the DP il suo libro, as illustrated below: (51)

IP I’

DP Maria

VP

darà

V’

tMaria

V’ tv

PP a Gianni DP

il suo libro The possessive suo is thus in the scope of both referents, and this condition does not change (see [50b]) even if the DP is moved to a higher position (which does not need to be specified here) in the IP functional area for reasons related to information structure.19 This means that what is relevant for binding, at least for this type of movement,20 is the deep (i.e. argument) position of bound DPs.

200 Adverbial modification The situation of adverbial PPs/NPs is rather different. Let us consider the following sentences:21 volentieri [nel (52) a. Mariak domani forse incontrerà Maria tomorrow maybe meet.FUT.3SG eagerly in.DET suok/*j nuovo appartamento] il vecchio amico Leoj. POSS.3.SG new apartment DET old friend Leo ‘Perhaps Maria will eagerly meet her old friend Leo in her new apartment tomorrow.’ b. *?Maria volentieri domani forse incontrerà nel suo nuovo appartamento il vecchio amico Leo. c. *?Maria forse domani incontrerà volentieri nel suo nuovo appartamento il vecchio amico Leo.22 d. *?Maria domani volentieri nel suo nuovo appartamento forse incontrerà il vecchio amico Leo. Let us first note that while Italian admits variation in its linear order, sentences (52b-d) are extremely marginal, which indicates that the constituent order in sentence (52a) is the unmarked case. This confirms that there is a specific hierarchical order for adverbs in the IP area: (53) [TP [pst/fut] tomorrow > [MoodP [+irrealis] perhaps > [ModP [+volitional] eagerly > [XP LOCATIVE Secondly, we can observe that the only possible antecedent of the possessive included in the locative PP is the NP Maria (thus excluding Leo). This shows that, unlike in (51), the locative PP is not in the scope of the DO in deep structure: It is merged in a higher position and its placement in the IP area fully accounts for this structural condition. Finally, note that when this PP is moved into final position, the extraposition process creates the necessary structural conditions for the DO to qualify as a possible antecedent for the possessive: (54) Mariak domani forse incontrerà volentieri il vecchio Maria tomorrow maybe meet.FUT.3SG eagerly DET old suok/j nuovo appartamento]. amico Leoj [nel apartment friend Leo in.DET POSS.3.SG new ‘Maybe tomorrow Maria will eagerly meet her old friend Leo in his/her new apartment.’

Clausal structures with adverbial function 201 The same is true of temporal NPs: In (55a) the possessive can only refer to Maria, whereas in (55b) it can also refer to each student: incontrato [il giorno del (55) a. Mariak ha Maria have.PRES.3SG meet.PRT DET day of.DET suo k/*j compleanno]z ognij studente. POSS.3.SG birthday each student ‘Mariak met each studentj on herk/*hisj birthday.’ b. Mariak ha incontrato ognij studente Maria have.PRES.3SG meet.PRT each student compleanno]z. [il giorno del suo k/j DET day of.DET POSS.3.SG birthday ‘Mariak met each studentj on herk/hisj birthday.’ To conclude, we have examined the syntax of adverbial NPs and PPs and shown that, despite appearances, a hierarchical structure can also be assumed for these elements which further limits the use of adjunctions. We can now move on to the syntax of adverbial clauses.

5.5. Clausal structures with adverbial function Like all other adverbial elements, Generative Grammar has traditionally considered adverbial clauses as circumstantial constituents (recursively) adjoined to the sentential node. Following a more rigorous approach which began in the Nineties (Kayne 1994, Cinque 1999) and a more recent interest in information structure (Haegeman 2003, Rizzi 2004), this analysis has been questioned by many authors, but no agreement has so far been reached. In this section, we will propose as clear and concise a description as possible of the vast range of phenomena involving adverbial clauses, and offer some points for analysis. In particular, in analyzing adverbial clauses we will first focus on their internal syntax (i.e. the structure of the adverbial subordinator) and subsequently on their external syntax (i.e. their merge point in the root clause).

5.5.1. The internal syntax of adverbial clauses In the literature, the subordinating element of an adverbial clause is usually ascribed to the category of complementizers (i.e. C° heads; see Grimshaw

202 Adverbial modification 1991, Lasnik & Saito 1992, Vikner 1994). However, cross-linguistic data show that this type of introductory element should not be considered to be equivalent to an invariable functional head as a complementizer (e.g. English that, Italian che, German dass, etc.) but as an element having nominal properties. Let us now present some data on time and manner adverbial clauses from typologically different languages, starting with Somali, an inflecting and partly polysynthetic language (from Frascarelli & Puglielli 2005a): (56) a. Gaba-rtii oo [mar-kaas gur-iga gashay] girl-DET.AN ASSOC time-DEM house-DET enter.PST.RED baan arkay. (SOMALI) FM.SCL.1SG see.PST.1SG ‘I saw the girl after she had walked into her house.’ b. [In-t-aad shaqeemaysó] buug baan part-DET-PRO.2SG work.PROG.DEP book FM.SCL.1SG akhrinayaa. read.PROG.1SG ‘While you are working, I am going to read a book.’ c. [Goorta qorra-xdu dhacdó] imaw. momentDET sun-DET.NOM set.DEP come.IMP.2SG ‘Come when the sun sets.’ d. [Si-da Ahmed uu doonayó] erey-ga u way-DET Ahmed SCL.3SGM want.PROG.DEP word-DET to qor. write.IMP.2SG ‘Write the word the way Ahmed wants.’ First of all, it should be noted that the introductory element of subordinate clauses (corresponding to English conjunctions after, when, while and the way [that]) is a noun, as it can be modified by a DET or a DEM and can be used in sentences with a subject or object function. Such nouns are usually generic (‘time’, ‘moment’, ‘manner’, etc.); however, a certain degree of subcategorization can be observed, relating to a specific type of (adverbial) temporal modification: Mar for posteriority, in and goor for simultaneity (the former with a durative aspect, the latter with a punctual one). There is thus a precise semantic correlation between the generic noun introducing the subordinate clause and the adverbial meaning of that clause.

Clausal structures with adverbial function 203 Let us now have a look at the following data from Turkish (from Kornfilt 1993), a head-final agglutinative language: (57) a. Müdür [tatil-e çik-ti÷-i zaman] manager vacation-DAT go-NOMIN-POSS.3.SG time ofis kapa-n-ir. (TURKISH) office shut-REFL-AOR ‘When the manager goes on vacation, the office closes.’ b. [Müdür-ün tatil-e çik-ti÷-i gece] ofis manager-GEN vacation-DAT go-NOMIN-POSS.3.SG night office yan-di. burn-PST ‘On the night when the manager went on vacation, the office burned down.’ c. Hasan [kitab-i san-a ver-di÷-im Hasan book-ACC PRO.2SG-DAT give-NOMIN-POSS.1.SG takdir-de] çok kiz-acak. case-LOC much anger-PST ‘Hasan got really mad because I gave you the book.’ d. Müdür [tatil-e çik-tik-tan sonra] ofis manager vacation-DAT go-NOMIN-ABL after office yan-di. burn-PST ‘After the manager went on vacation, the office burned down.’ Turkish also shows that the element introducing the adverbial clause is a noun head which can be either generic (57a) or referential (57b).23 The subordinate verb shows the nominalizing suffix Diö24, typical of subordination in this language (see § 2.5.); as a consequence, the temporal interpretation of the adverbial clause depends on the root clause and its subject is realized by means of a possessive suffix (consistent with the nominal status of the subordinate clause), correlated with a GEN or NOM Case NP (a Topic in the latter case). The nominal nature of the introductory element of adverbial clauses is even more evident in reason clauses, where the noun head (‘case’) is marked for locative Case (57c). Finally, it should be noted that the adverbial subordinate clause may be introduced by a prepositional element (57d): In this case, the adverbial clause can be considered a noun phrase proper. As the glosses in (57d) show, no subject agreement is present on the nominalized verbal head, which, on the contrary, bears ablative Case (for details, see Kornfilt 1993).

204 Adverbial modification To conclude the description of the different language types, let us finally observe some data from MƗori (Bauer 1993), a head-initial isolating language in which adverbial subordination shows similar properties to the two previously analyzed languages: (58) a. [Kia tae mai koe] ka kai taatou. (MƖORI) time arrive here S.PRO.2SG T/ASP eat S.PRO.1PL.INCL ‘We eat when you come.’ b. [Inaa tae mai koe] ka kai moment arrive here S.PRO.2SG T/ASP eat ai taatou. FUT S.PRO.1PL.INCL ‘We will eat when you come.’ c. I hoki maatou ki te kaainga [i T/ASP return S.PRO.1PL.EXCL to DET house from te mea e ua ana]. DET thing T/ASP rain PST ‘We went back home because it was raining.’ d. E haere ana ahau ki te taaone [ki te T/ASP move PST S.PRO.1SG to DET town to DET hoko mai teetahi koha]. buy here INDEF.DET gift ‘I went to town to buy a present.’ Again, in this language adverbial clauses are formed as subordinate clauses modifying a nominal element which is a generic head (‘time’, ‘moment’, ‘thing’, etc.). As noted above for Turkish, these sentences can also be introduced by a preposition selecting a DP (see [58c-d]). This pattern is found in a number of languages of different families and types. Consider the following examples from Polish (Testa 2007), Afar (Bliese 1981), Hungarian (Lipták 2006), Rembarunga (Nordlinger & Saulwick 2002) and Tagalog (Schachter & Otanes 1972): (59) Został ukarany [za AUX.PERF.3SGM punish.PRT.SGM.NOM for siĊ obchodził]. REFL.PRO behave.PERF.3SGM ‘He was punished for the way he behaved.’

to DEM

jak how (POLISH)

Clausal structures with adverbial function 205 (60) [’usuk ’yoo yoogo’re-emi-k duma] S.PRO.3SGM O.PRO.1SG hit.PST.3SG-NOMIN-from earlier time inniki’se. (AFAR) fall.PST.1SG ‘I fell before he hit me.’ (61) a. [Ami-kor Péter nincs ott-hon] REL.PRO.NONHUM-TEMP Peter COP.3SG.NEG here-land Anna egyedül van. (HUNGARIAN) Anna alone COP.PRES.3SG ‘When Péter is not at home, Anna is alone.’ b. [Az-elĘtt hogy/ mi-elĘtt el-ment], Péter DEM-before that REL.PRO-before PV-go.PST.3SG Péter meg-nézte a postá-já-t. PERF-check.OBJ.PST.3SG DET mail-ACC-POSS.3.SG ‘Before leaving, Péter checked his mail.’ (62) Jurla Ø-nga-ngara-ma [guwa water SCL.1SG-OCL.3SG-get-FUT aim ga-Ø-rdom-ma-gan]. NONPST-SCL.3SG-drink-FUT-BEN ‘I will get some water so he drinks.’

(REMBARUNGA)

(63) [Bago natulog] ang mga bata uminom ACT.drink.PST previous ACT.sleep TRIG PL kids ng gatas. (TAGALOG) DIR milk ‘Before sleeping, the kids drank some milk.’ This pattern is clearly too frequent to be coincidental. It can thus be argued that adverbial clauses have the syntactic properties of subordinate clauses depending on a noun head and sometimes accompanied by a preposition. It is now important to establish what type of subordination applies to adverbial clauses. As we know, clausal nominal modification is divided into two main types, i.e. relative and complement clauses, illustrated in (64a-b) respectively: (64) a. I do not believe the story [that you have told]. b. I do not believe (the story) [that Jack has left].

206 Adverbial modification In the case of the relative clause in (64a) the noun head is part of the argument selection of the subordinate verb and, according to the raising analysis, it is generated within the relative clause and subsequently moves as a syntactic operator (see § 4.4.2.1.). In other words, in (64a) the NP story is the PATIENT and DO of tell and the relevant structure is the following: (65)

DP D’ CP

the

C’ IP that

I’ you

VP

have

V’ tyou

NP

told

story The insertion of the NP-head within the relative clause is confirmed by a number of diagnostics (see § 4.4.2.1.), e.g. the fact that no resuming pronoun is admitted within the relative clause (see [66]); this shows that the two elements compete for the same macro-role and syntactic function: (66) *I do not believe the storyk [that you have told itk to me]. On the other hand, a complement clause is an argument of the head noun. It is therefore selected by the noun and merged in Compl,NP: (67)

DP D’ NP N’

the

CP story

C’ IP that

I’ Jack

VP V’

has tJack

told

Clausal structures with adverbial function 207 A careful examination of the internal structure of adverbial clauses reveals that they are mostly relative clauses. In Somali, for instance, the verb of the adverbial clause shows a reduced paradigm when it (appositively) modifies an NP with subject function, as in relative clauses. For the sake of convenience, sentence (56a) is repeated below as (68), and compared to an appositive relative clause (69) with the same characteristics (i.e. where the head is the subject of the subordinate clause): (68) Gabartii oo [markaas guriga gashay] girl.DET.AN ASSOC time.DEM house.DET enter.PST.RED baan arkay. (SOMALI) FM.SCL.1SG see.PST.1SG ‘I saw the girl after she had walked into her house.’ [lit.: ‘The girl, that time she walked into her house, I saw her.’] (69) Cali oo [toosayá /*toosayaa] baa in Cali ASSOC wake.PROG.RED /wake.PROG.3SGM FM amount yar seexday. small sleep.PST.3SGM ‘Cali, who is waking up right now, has not slept much.’ In both cases the reduced paradigm must be used, confirming the assumption that the two sentences present the same type of subordination (i.e. relative clauses). A fact confirming this hypothesis is that whenever the adverbial does not refer to a subject constituent, but is only used as a restrictive modifier of a generic noun head (‘time’, ‘manner’, etc.), the dependent paradigm appears, i.e. the same found in relative clauses when their head is an object of any kind. Let us again consider sentence (56b), repeated here as (70), and compare it with the relative clause in (71):25 (70) [In-t-aad shaqeemaysó] buug baan akhrinayaa. part-DET-2SG work.PROG.DEP book FM.SCL.1SG read.PROG.1SG ‘While you are working, I am/will be reading a book.’ (71) Boosta-da gee warqad-da [ay Maryan post office-DET take.IMP.2SG letter-DET SCL.3SGF Maryan qoraysó]. write.PROG.DEP ‘Take to the post office the letter that Maryan is writing.’

208 Adverbial modification The same analogy is found in the other languages illustrated previously. In Turkish, for instance, the verb of relative clauses typically bears the NOMIN affix which was found in adverbial clauses (see [57a-d]). Let us now compare (57a), repeated as (72), and the relative clause (73): (72) [Müdür-ün tatil-e çik-ti÷-i zaman] manager-GEN vacation-DAT go-NOMIN-POSS.3.SG time ofis kapa-n-ir. (TURKISH) office shut-REFL-AOR ‘When the manager goes on holiday, the office closes.’ (73) [Hasan-in iç-ti÷-i] úarap. Hasan-GEN drink-NOMIN-POSS.3.SG wine ‘The wine Hasan drinks.’ In this context, it is interesting to note that in Turkish relative clauses, along with the form Diö, we can find the form (y)An, as in the following case: (74) [Bali÷-i yi-yen] fish-ACC eat-NOMIN ‘The man who ate fish.’

adam. man

The choice of the nominalizer depends on the syntactic role of the noun head: When it is linked to the subject (or possessor) function, the nominalizer is (y)An (as in [74]); in all other cases it is Diö (see Kornfilt 1993 for details). Consequently, in adverbial clauses, where the NP-head of the relative clause is not connected with the subject function (as it expresses information on time, manner, goal, cause etc.), the nominalizer will invariably be Diö. The order of constituents also supports the present analysis. As is evident, the generic noun precedes the adverbial clause in head-initial languages (such as MƗori, Polish, Tagalog) and follows it in head-final languages (Turkish, Afar). The assumption that adverbial clauses originate from relative clauses is thus supported by a number of facts. In fact, even in languages like English or Italian where adverbial clauses are typically introduced by wh-phrases (e.g. English when, how, why, Italian quando, come, perché), the same sentence can be reformulated using a generic NP (possibly preceded by a preposition), as the head of a restrictive relative clause:

Clausal structures with adverbial function 209 (75) a. [When you come], everyone will be happy. b. [At the time at which you come], everyone will be happy. (76) a. [Da come mi hai salutato], OCL.1SG have.PRES.2SG greet.PRT from how ho capito che c’ era have.PRES.1SG understand.PRT that there be.IMPF.3SG un problema. (ITALIAN) INDEF.DET problem ‘From the way you said hi to me, I sensed there was a problem.’ b. [Dal modo in cui mi hai from.DET way in which OCL.1SG have.PRES.2SG salutato], ho capito che greet.PRT have.PRES.1SG understand.PRT that c’ era un problema. there be.IMPF.3SG INDEF.DET problem ‘(=[76a])’ Nonetheless, the relative clause is not the only realization for adverbial clauses, although it is certainly the most common. In some cases the adverbial clause is preceded by a noun head which cannot be considered a part of the argument structure of the subordinate verb (e.g. the MƗori causal clause in [77a] below).26 In other cases, the adverbial clause is immediately introduced by a preposition (as in the MƗori final clause [58d] and in the Turkish anteriority temporal clause [57d]). The two sentences are repeated here for convenience as (77b) and (78): (77) a. I

hoki

T/ASP return

te

maatou

ki te kaainga [i to DET house from ana]. (MƖORI)

S.PRO.1PL.EXCL

mea e ua DET thing T/ASP rain PST ‘We went back home because it was raining.’ b. E haere ana ahau ki te taaone [ki T/ASP move PST S.PRO.1SG to DET town to hoko mai teetahi koha]. buy here INDEF.DET gift ‘I went to town to buy a present.’

te DET

Ø

210 Adverbial modification (78) Müdür [tatil-e çik-tik-tan Ø sonra] go-NOMIN-ABL after manager vacation-DAT ofis yan-di. (TURKISH) office burn-PST ‘After the manager went on vacation, the office burned down.’ In these cases, we surmise that the adverbial clause is not a relative clause but the complement of a noun head which can be phonologically null27 (cf. Testa 2007). A noun head may be omitted as a consequence of its semantic properties; a generic head like ‘event’, ‘thing’, ‘purpose’ etc. can be considered unnecessary for interpretation and therefore omitted.28 The presence of a null NP is evident in MƗori as there is a DET (te) after the preposition (ki) in (77b); it can thus be assumed that there is a noun head interpreted as ‘purpose’ and the relevant sentence can be paraphrased as: (79) ‘I went to town for the purpose of buying a present.’ Similarly, in Polish, the presence of a null NP in causal clauses is evident, as in the following example (from Testa 2007): (80) Został ukarany [dla-tego Ø Īe AUX.PERF.3SG punish.PRT.SGM.NOM for-DEM.GEN that Ĩle siĊ obchodził]. (POLISH) REFL.PRO behave.PERF.3SGM badly ‘He was punished because he behaved badly.’ The adverbial clause is introduced by an element (dlatego) that represents the lexicalization of a preposition and a demonstrative with GEN Case, which is not followed by a noun head but by the complementizer Īe. Plausibly, a null NP like ‘event’ is present, modified by the following complement clause, thus obtaining the following inter-pretation: (81) ‘He was punished for the event that he behaved badly.’ Accordingly, it can be assumed that the same type of implicit noun head (‘event’) precedes the postposition sonra (‘after’) in the Turkish sentence (78), interpreted as follows: (82) ‘After the event that the manager went on vacation, the office burned down.’

Clausal structures with adverbial function 211

Based on cross-linguistic data, our conclusion is that adverbial clauses can be thought to originate from nominal subordination; consequently, these clauses are inserted as relative or complement clauses. What this means in structural terms is that adverbial clauses are not simply CPs, but DP-embedded CPs. This subordination relationship is evident in some languages, whereas in others lexicalization and null generic NPs make it more obscure. Once the internal nature of these clauses has been defined, we can move on to our second question, i.e. the relationship between adverbial and root clauses. Where are adverbial clauses merged? Are they generated as Specifiers of functional projections (like pure adverbs and adverbial NP/PPs) or are they extrasentential elements?

5.5.2. The external syntax of adverbial clauses As Bianchi (1997) notes, adverbial clauses do not seem able to overtly occupy an unmarked position within the IP. In other words, it appears that adverbial clauses cannot be specifiers of functional projections, as is evident for Italian (83), Dutch (84), Hungarian (85) and Polish (86), from Testa (2007):29 (83) *Avevo quando hai telefonato have.PRES.2SG phone.PRT have.IMPF.1SG when finito di mangiare. end.PRT of eat.INF [lit.: ‘I had when you called just finished eating.’]

appena just (ITALIAN)

(84) *Ik heb toen ik een kind S.PRO.1SG AUX.PRES.1SG when S.PRO.1SG INDEF.DET kid was hier gewoond. (DUTCH) AUX.PST.1SG here live.PRT [lit.: ‘I when I was a kid lived here.’]

212 Adverbial modification (85) *Huszonnégymillió volt ami-kor COP.PST.3SG REL.PRO.NONHUM-TEMP 24 million kit-ört a forradalom a francia. (HUNGARIAN) DET French break out-PST.3SG DET revolution [lit.: ‘The French population was when the Revolution broke out 24 million.’] (86) *Mieszkałem kiedy wybuchła wojna live.IMPF.PST.1SGM when break out.PERF.PST.3SGF war w Moskwie. (POLISH) in Moscow ‘I lived in Moscow when the war broke out.’ One reason why adverbial clauses are necessarily peripheral could be their prosodic heaviness, as was previously suggested for adverbial NPs/ PPs. However, cases in which no phonological factors can explain obligatory extraposition are frequently found, as is shown in the following example from Testa (2007): (87) a. Hai chiamato Anna? Se sì, have.PRES.2SG call.PRT Anna if yes avresti potuto sentire anche Andrea. (ITALIAN) have.COND.2SG can.PRT hear.INF also Andrea ‘Did you call Anna? If you did, you could have rung Andrea too.’ b. Hai chiamato Anna? *Avresti se sì potuto sentire anche Andrea. If the adverbial clause se sì is placed in an intermediate position within the IP, the sentence is ungrammatical in spite of the prosodic lightness of this constituent. On the other hand, if a prosodically heavier adverb like allora (‘then’) is used, both positions are acceptable: (88) a. Hai chiamato Anna? Allora, have.PRES.2SG call.PRT Anna then avresti potuto sentire anche Andrea. hear.INF also Andrea have.COND.2SG can.PRT ‘Did you call Anna? Then you could have rung Andrea too.’

Clausal structures with adverbial function 213 b. Hai chiamato Anna? Avresti allora Anna have.COND.2SG then have.PRES.2SG call.PRT potuto sentire anche Andrea. hear.INF also Andrea can.PRT ‘Did you call Anna? Then you could have rung Andrea too.’ This contrast shows that what really counts is not prosodic heaviness, but the syntactic complexity of a constituent (see Austin, Engelberg & Rauh [eds.] 2004): Although formed by just two syllables, se sì is a sentence, even though a reduced one, and as such it contains a larger amount of structure than a simple adverb. The relevant restriction seems therefore to apply to embedded CPs. In other words, it should be assumed that the syntactic structure of a CP is somehow too rich, and consequently too heavy, to appear in an embedded position even when its terminal nodes are not all overtly realized. Evidence for this is provided by the observation that adverbial PPs (whose internal structure is certainly not as complex as that of a CP) behave like AdvPs in this context: (89) a. Hai chiamato Anna? In tal caso, have.PRES.2SG call.PRT Anna in such case avresti potuto sentire anche Andrea. hear.INF also Andrea have.COND.2SG can.PRT ‘Did you call Anna? In that case, you could have rung Andrea too.’ b. Hai chiamato Anna? Avresti Anna have.COND.2SG have.PRES.2SG call.PRT in tal caso potuto sentire anche Andrea. hear.INF also Andrea in such case can.PRT (= [89a]) What causes the restriction on adverbial CPs? Following Testa (2007), we argue that adverbial clauses are always extrasentential because they are necessarily focused (as new information) or topicalized (as given information). As seen in Chapter 1 (§ 1.2.3.3.), this type of information is realized in the CP area, where syntax interfaces with discourse grammar. We therefore propose that these clauses cannot be part of an all-new sentence but form distinct syntactic and prosodic domains. Furthermore, the extrasentential position of adverbial clauses is consistent with the pragmatic principle of Progression in new information (see note 19 to

214 Adverbial modification Chapter 5): Focused adverbial clauses normally follow the root clause, whereas topicalized ones precede it. It can be easily verified that while a time clause is not appropriate (as indicated by the symbol #) in an all-new sentence as (90), it is totally acceptable when the preceding question focuses on the temporal information (91): (90) A: What’s up? B. #Jim left when they called him. (91) A: When did Jim leave? B. (Jim left) when they called him. Similarly, when an adverbial clause is new information, it cannot precede the root clause: (92) A: Why did you leave? B. I left because I was tired. B'. *Because I was tired I left. (93) A: When are you leaving? B. I am leaving when you come. B'. *When you come I am leaving. The same is true for other languages, such as Polish (from Testa 2007): (94) A: Dlaczego odszedłeĞ? (POLISH) for-what.GEN leave.PST.2SG ‘Why did you leave?’ B. (Odszedłem) dla-tego Īe byłem leave.PST.1SGM for-DEM.GEN.SG that COP.PST.1SGM głodny. hungry.NOM.SGM ‘(I left) because I was hungry.’ B'. *Dlatego Īe byłem głodny odszedłem. In Focus-oriented languages (see § 6.4.3.), such as Somali and Hungarian, NPs conveying new information need to appear preverbally (in a marked position with regard to the new information progression). It is therefore interesting to notice that this is exactly the position that an

Clausal structures with adverbial function 215 adverbial clause conveying new information must occupy in these languages. This is shown in the following examples (from Puglielli 1981 and Testa 2007, respectively): (95) A: Goormaad (goor+ma+baa+aad) buug-gan when.Q.FM.SCL.2SG book-DEM akhrinaysaa? (SOMALI) read.PROG.2SG ‘When are you going to read this book?’ B. Intaad shaqeenaysó baan buug-gan while.SCL.2SG work.DEP.2SG FM.SCL.1SG book-DEM akhrinayaa.30 read.PROG.2SG ‘I’m going to read this book while you are working.’ B'. *Buuggan baan akhrinayaa intaad shaqeenaysó. (96) A: Mi-kor ment el? (HUNGARIAN) REL.PRO.NONHUM-TEMP go.PST.3SG PV ‘When did she leave?’ B. AKKOR ment el, ami-kor DEM.TEMP go.PST.3SG PV REL.PRO.NONHUM-TEMP megjött Mari. PERF.come.PST.3SG Mari ‘(She left) when Mari came.’ B. *Ment el, AKKOR amikor megjött Mari. A telling characteristic can be observed in the Hungarian example: Since it is impossible to place the whole adverbial clause initially (probably owing to its heaviness), a resumptive element (akkor) is used in its place, coreferent with the right-dislocated adverbial clause. A literal translation of the Hungarian sentence would therefore be ‘she left then, when Mari came’. The interpretation of adverbial clauses as Topics was already noted by Bianchi (1997). The author showed that if these sentences carry given information they are pronounced with an upward contour whenever they appear in the left periphery of the sentence (97a) (thus resembling left-hand Topics; see § 6.5.); on the other hand, they are prosodically flat when they appear to the right after a pause (97b) (like right-hand Topics):

216 Adverbial modification (97) a. Dopo che Gianni è uscito, after that Gianni be.PRES.3SG exit.PRT ho cominciato a lavorare. to work.INF have.PRES.1SG begin.PRT ‘After Gianni went out, I began to work.’ b. Ho cominciato a lavorare, have.PRES.1SG begin.PRT to work.INF dopo che Gianni è uscito. after that Gianni be.PRES.3SG exit.PRT ‘I began to work, after Gianni went out.’

(ITALIAN)

In these cases the time clause is not part of the assertion but is presented as presupposed information. In the former, it is interpreted as a temporal frame (see Chafe 1976, 1987); in the latter, it is a sort of afterthought, which retrieves previously activated information (see Frascarelli 2000a, 2004a and § 6.5.1.). Evidence of the topicalized status of these adverbial clauses comes from languages like Hungarian and Polish. In particular, pro-forms are available in Hungarian for both conditional (see [98], from Testa 2007) and time clauses (albeit optional; see [99], from Bhatt & Lipták 2009), whereas in Polish only for conditional sentences (see [100], from Testa 2007): (98) a. Ha meg-jav-ult, akkor veg-e if PERF-improve-PST.3SG DEM.TEMP end-POSS.3.SG a móká-nak. (HUNGARIAN) DET joke-DAT ‘If he has improved, the fun is over.’ b. *HA MEGJAVULT, akkor vege a mókának (99) a. Ami-kor János meg-jött REL.PRO.NONHUM-TEMP János PERF-come.PST.3SG (akkor) Mari fĘzĘtt. then Mari cook.PST.3SG ‘When János came, Mari was cooking.’ b. Ame-ddig János alsz-ik REL.PRO.NONHUM-till Janos sleep-PRES.3SG (addig) Mari fĘz. DEM.TERM Mari cook.PRES.3SG ‘While János is sleeping, Mari is cooking.’

Clausal structures with adverbial function 217 c. Ami-óta János dolgoz-ik REL.PRO.NONHUM-TERM Janos work-PRES.3SG (az-óta) Mari rendszeresen fĘz. DEM-from Mari regularly cook.PRES.3SG ‘Since János has been working, Mari has cooked regularly.’ (100) a. Jakbym wiedział Īe lubi kwiaty, if.COND.1SG know.PRT COMP love.PRES.3SG flowers.ACC to kupiłbym je dla niej. (POLISH) then buy.COND.1SG OCL.3PLN for O.PRO.3SGF.GEN ‘If I had known she liked flowers, I would have bought some for her.’ b. *TO KUPIŁBYM JE DLA NIEJ, jakbym wiedział Īe lubi kwiaty. Topicalized complement clauses can be resumed in Italian as well, unlike focused clauses,31 as is shown in the following sentences: (101) A. Tutti si sono sorpresi per all MIDD be.PRES.3PL surprise.PRT for l’ assenza di Leo… DET absence of Leo ‘Everyone was surprised at Leo’s absence...’ B. Io no: che Leo non sarebbe venuto, I no that Leo not be.COND.3SG come.PRT lo sapevo da tempo. OCL.3SGM know.IMPF.1SG from time ‘I wasn’t: I had known for a while that Leo wouldn’t come.’ B'. Io no: *che Leo non sarebbe venuto, sapevo da tempo. (102) A. Che cosa non sapevi? what not know.IMPF.2SG ‘What didn’t you know?’ CHE LEO SAREBBE VENUTO. B. Non sapevo not know.IMPF.1SG that Leo be.COND.3SG come.PRT ‘I didn’t know that Leo would come.’ B'. *Non lo sapevo CHE LEO SAREBBE VENUTO. On the other hand, no resumptive forms are generally available for adverbial clauses, with the exception of conditional clauses in which allora

218 Adverbial modification can be considered (some type of) an (adverbial) pro-form, as shown in the following example (from Testa 2007): (103) Se hai abbastanza soldi, money if have.PRES.2SG enough allora puoi partire. then can.PRES.2SG leave.INF ‘If you have enough money, then you can leave.’ Also in this case, this pro-form is unacceptable when the conditional clause is new information: (104) *Allora then

puoi can.PRES.2SG

partire leave.INF

HAI ABBASTANZA SOLDI. money if have.PRES.2SG enough (intended: ‘You can leave IF YOU HAVE ENOUGH MONEY.’)

SE

To conclude, the extraposed position of adverbial clauses is not generically due to their being merged as adjuncts to the sentential node. Adverbial clauses are merged in the same way as other adverbials, i.e. as Specifiers of functional heads associated with specific semantic roles and hierarchically ordered in the root IP area. However, they have to be extraposed as Topics or Foci. This means that what we have described as a strong preference for adverbial NPs/PPs is a requirement for adverbial clauses as they cannot occupy in surface structure the unmarked position where they are merged. In other words, they cannot appear within the root clause. This obligatory extraposition may be ascribed to information-structural requirements which make it possible for the different discourse roles to be interpreted at the interfaces.

6. The sentence as utterance

6.1. Discourse grammar and information structure In the previous chapters we discussed the lexical and functional categories needed to describe languages in their universal features and cross-linguistic variation. Furthermore, we showed some aspects of the interaction between lexical semantics and syntactic structure (VP area) and between morphology and syntax (IP area). As we know, however, there is another domain of analysis in the representation of sentential structure in X' terms (Chapter 1, § 1.2.3.), i.e. the CP, where the interface between syntax and pragmatics takes place. This is the topic of this chapter. This interaction will be discussed both in terms of the communicative function of each sentence within the discourse context and, more broadly, from the point of view of information structure, i.e. the way in which constituents are located in the sentence to meet specific discourse requirements. We will therefore study the mechanisms and strategies used by languages for pragmatic and communicative purposes. As usual, we will focus on those aspects which are significant at the interface levels, i.e. phonology (prosody and intonation), morphology and syntax. Our objective will again be to show how universal mechanisms take different forms in typologically different languages, while at the same time the number of possible options remains limited.

6.2. Speech acts Let us begin our discussion on the communicative function of utterances by considering the following sentences: (1)

Jim is leaving tomorrow.

(2)

Is Jim leaving tomorrow?

(3)

Jim, leave tomorrow!

220 The sentence as utterance Though similar, these utterances are clearly used “to do different things” in discourse (see Austin 1962, Searle 1969) and their semantic-pragmatic value reflects their distinct discourse functions. Sentence (1) is a statement, because the speaker utters it to inform the hearer of a given event; sentence (2) is used to ask for information; finally, sentence (3) is a way for the speaker to get the hearer to perform an action. The three examples correspond to three formally and pragmatically distinct types of sentences, usually called declarative, interrogative and imperative. These terms refer to the communicative function conventionally associated with each sentence type. In uttering any of these sentences, the speaker performs what is known as a locutionary act, i.e. produces a meaningful linguistic expression. The speaker inevitably also performs an illocutionary act, as s/he uses language in a meaningful way by interacting with another person to inform, give orders, caution, concede and so on. Furthermore, the illocutive act may be associated with a perlocutionary act: The speaker may then “bring about or achieve [something] by saying something, such as convincing, persuading, deterring” (Austin 1962: 109). These acts are actually performed only if the speaker succeeds in achieving the desired purpose. Speech acts are therefore associated with a specific communicative force. In particular, locutionary force corresponds to the surface morphosyntactic form assumed by the different clause types (declaratives, interrogatives, imperatives), whereas illocutionary force is the communicative purpose of the speaker (which may be direct or indirect; see below and, more specifically, Chapter 7). As an illustration of the acts performed when someone speaks, consider our analysis of the following example from Chierchia & McConnell-Ginet (1990): (4) The bull is in the field. The sentence in (4) is clearly a statement. To state, however, is much less concrete than to argue, to bet, to caution or to alarm (i.e. the types of speech acts which, according to many authors, form the literal force associated with utterances); a sentence like (4) can be asserted with any of these illocutionary forces. In other words, the locutionary act performed by asserting sentence (4) is stating that the bull is in the field. But what is the purpose that induces the speaker to utter such a sentence? The illocutionary force of this assertion, though not made explicit by the speaker, can vary according to the context: It may be to create a common ground between speaker and hearer, or to caution (by pointing to a danger), or to indirectly

Speech acts 221 ask for an intervention (such as shutting the gate), and so on. A possible conclusion is that sentence (4) is an indirect illocutionary act. Note that, while direct illocutionary purposes may be made explicit, this is not the case for indirect ones:1 (5) The bull is in the field. And this is a warning. (6) #The bull is in the field. And this is an order. This means that a declarative sentence cannot be directly associated with an order, as the illocutionary force of an order needs to be expressed by means of an imperative sentence. With a declarative sentence, a speaker can perform a number of functions, among which are asserting, informing, suggesting, cautioning and so on. All of these locutionary acts can be subsumed under a single illocutionary macro-function (i.e. speaking to inform). Similarly, other macro-functions such as asking for information, requiring an action or exclaiming can be conveyed through different expressions. These illocutionary functions are realized by formal means with specific morphological, syntactic and prosodic features. As an example, see the following sentences from English, Italian, Somali and Tagalog: (7)

a. John leaves tomorrow. b. Does John leave tomorrow? c. Eat!

(8)

a. Mario parte domani. Mario leave.PRES.3SG tomorrow ‘Mario leaves tomorrow.’ b. Mario parte domani? Mario leave.PRES.3SG tomorrow ‘Does Mario leave tomorrow?’ c. Mangia! eat.IMP.2SG ‘Eat!’

(9)

a. Calí berri Cali.NOM tomorrow ‘Cali leaves tomorrow.’

waa DECL

(ITALIAN)

dhoofayaa. (SOMALI) leave.PRES.PROG.3SGM

222 The sentence as utterance b. Calí berri ma Q Cali.NOM tomorrow ‘Does Cali leave tomorrow?’ c. Cun! eat.IMP.2SG ‘Eat!’

dhoofayaa? leave.PRES.PROG.3SGM

(10) a. Si

Juan ay aalis bukas. (TAGALOG) Juan TM leave.FUT tomorrow ‘Juan leaves tomorrow.’ b. Si Juan ba ay aalis bukas? TRIG Juan Q TM leave.FUT tomorrow ‘Does Juan leave tomorrow?’ c. Kain! eat ‘Eat!’ TRIG

In the four languages examined, the sentences in (a) have declarative force, those in (b) interrogative force and those in (c) are orders. There are, however, important differences in their formal realization. In English, Italian and Tagalog, declarative force is only signaled with prosodic and intonational means; the declarative sentence has a downgrading contour2 and represents the unmarked construction (§ 6.3.1.) in these languages. In Somali, on the other hand, along with a downgrading contour (Frascarelli & Puglielli 2007), a marker (waa) is needed to signal the declarative force of the sentence. Let us now consider yes-no questions.3 In English, interrogative force requires a specific syntactic construction where the auxiliary precedes the subject, and intonation may be downward, as in declaratives, or change (according to varieties; see Brown 1983). In Italian, the same constituent order is used as in declaratives; as a consequence, the difference in illocutionary force is only expressed by the intonational contour (upward at the end). Finally, in Somali and Tagalog, specific morphological markers are needed: Ma in Somali (in the same position that waa occupies in the corresponding declarative sentence; see [9b]), ba in Tagalog (following the topicalized constituent; see [10b]). As regards the sentences in (c), orders lack overt subject in the languages examined (regardless of whether they are null subject languages), and the verb appears in a particular form, often identical to its base form.4

Information structure in declarative clauses 223 A more extensive and detailed description of speech acts will be presented in Chapter 7, where it will also be discussed how types of sentences corresponding to different speech acts relate to each other, and how information is distributed among them. For the moment, we simply wish to emphasize that illocutionary force is crucial in discussing information structure, and to show its main characteristics. Next, we will examine the mechanisms used in languages to codify the distribution of information, starting with declarative clauses.

6.3. Information structure in declarative clauses Each sentence has an information content. Speaker-hearer interaction is developed by means of sentence construction through a process of information packaging (using Vallduví’s 1992 terms) by which given and new information are differentiated.5 In simple terms (though see note 5 to Chapter 6), the informative part of the sentence may be defined as new, whereas shared, active and presupposed information is given. This common ground also includes the speaker’s assumptions about the interlocutor’s knowledge and all that comprises the linguistic or extra-linguistic context. We will now consider how information is distributed within a declarative sentence and describe the way in which languages signal this distribution. We will limit the discussion to predicative sentences6 thus excluding thetic utterances (such as there is a dog in the garden), where the whole sentence is considered new information. Predicative sentences are made up of two parts: A new part, the predication (i.e. the object of assertion), and an element presented and assumed as the starting point of the predication (see the idea of “thema” in Calabrese 1986). These two parts are normally described in the literature as Comment and Topic respectively. Consider the following question-answer pair: (11) A: What are the children doing? B: The children are playing hide-and-seek. In (11B), are playing hide-and-seek is the new part, i.e. the Comment, whereas the constituent the children is given, i.e. the Topic. The Comment (the verb with its complements and any adjuncts) is indispensable, since it is new information; the Topic, on the other hand, may be omitted, provided it can be retrieved from the context. Indeed, another possible answer to (11A) is:

224 The sentence as utterance (12) They are playing hide-and-seek. The opposite is also possible: The [verb+complement] portion may be given information and the nominal element new information, as in the following case: (13) A: Who is playing hide-and-seek? B: The kids (are playing hide-and-seek). Here, the nominal constituent the kids, being new information, is defined as the Focus and the rest of the sentence, which can be omitted, is the Presupposition. As a matter of fact, the sentence in (13B) is not a predication about the kids, but implies the presupposition that ‘someone is playing hide-and-seek’ (expressed in the question), and specifies the kids as that ‘someone’ (on Focus interpretation, see § 6.4.). It can be concluded that information in sentences can be distributed according to two patterns (see Chomsky 1971, Neeleman & Reinhart 1997, Krifka 2007, Frascarelli 2010): (14) INFORMATIONAL PATTERNS a) Topic-Comment b) Focus-Presupposition We now intend to establish a clear link between discourse grammar, phrasal categories (basically NP and VP) and syntactic functions. For this purpose, we will start our analysis with discourse strategies and the distribution of informational content in unmarked and marked sentences.

6.3.1. The notion of markedness We have repeatedly mentioned the notion of unmarked (or basic) order of constituents. What is the exact meaning of this term (and its antonym) in linguistics? Three distinct levels of analysis must be taken into consideration for a comprehensive definition, that is to say, phonology, syntax and pragmatics. A sentence is phonologically unmarked when the intonational contour associated with it can be represented as a continuous curve; in a prosodically marked sentence, on the other hand, there are interruptions, pauses and intonational peaks.

Information structure in declarative clauses 225 A sentence is syntactically unmarked when its constituents follow the order determined by the argument selection of the verb (see Chapter 2). This is true for all-new sentences, i.e. answers to questions like What happened?, What’s new? As a consequence, a sentence is syntactically marked when its constituents do not occupy their canonical positions, but are dislocated to express special meanings. It should not be forgotten that a marked order is never arbitrary, but fulfills specific pragmatic and discourse requirements. A syntactically marked sentence usually has a special intonation, hence syntactic and phonological markedness are closely related. Finally, pragmatic markedness corresponds to the idea of appropriateness: A sentence is pragmatically unmarked when it is suitable for a very large number of contexts and language situations. In particular, a pragmatically unmarked sentence is typically one in which given information precedes new information, based on the principle of Progression in new information (see Antinucci & Cinque 1977, Renzi, Salvi & Cardinaletti [eds.] 1988, and note 19 to Chapter 5). Of course, the extent of new information can vary depending on the context. So, in a sentence like (15): (15) Ann gave Jim a book. either the whole VP (gave Jim a book) or one of the two objects may be new information, according to whether the sentence is an answer to one of the following questions: (16) a. What did Ann do? Ann [gave Jim a book]. b. What did Ann give Jim? She gave him [a book]. c. Who did Ann give a book to? She gave it [to Jim]. It is thus extremely important to take the preceding context into consideration whenever the distribution of information is analyzed.

226 The sentence as utterance 6.3.2. Given and new in marked and unmarked structures As argued above, given information precedes new information in a pragmatically unmarked sentence; in other words, it appears to be cognitively reasonable to utter first the element to which the information provided in the predication refers. If this element is inserted into the structure as an NP, it assumes Topic function in discourse grammar. In the same sentence, new information will necessarily be inserted in the structure as a VP,7 and will function as a Comment. The Topic does not have a syntactic function within the IP because, as we know, discourse categories are hosted in the CP area, and a Topic-NP cannot occupy an IP-internal position as a subject or object. The Topic is therefore connected with a pronominal/null element within the sentence, which transfers to the Topic-antecedent the interpretation concerning its syntactic function (by means of an anaphoric chain). So, for example, in (17) and (18), the Topic is not a subject, but is connected with the syntactic function of subject, realized by a pronoun.8 Notice that in a null subject language like Italian, the relevant pronoun is phonologically null (pro): (17) Leok malheureusement ilk Leo unfortunately PRO.3SGM arrivé tard. arrive.PRT late ‘Unfortunately, Leo arrived late.’ sfortunatamente prok (18) Leok Leo unfortunately PRO.3SGM arrivato tardi. arrive.PRT late ‘Unfortunately, Leo arrived late.’

est be.PRES.3SG (FRENCH)

è be.PRES.3SG (ITALIAN)

What has been said so far about unmarked sentences can be summed up as follows:

Information structure in declarative clauses 227 Table 2. Correlation between levels of analysis in unmarked sentences pragmatics

GIVEN

NEW

discourse grammar

TOPIC

COMMENT

(correlated to)

PREDICATION

semantic value

ARGUMENT

phrasal category

syntactic function

NP (correlated to)

VP

PREDICATE

SUBJECT

Discourse functions are thus not unrelated to syntax; on the contrary, they strongly correlate with syntactic structure, both at a deep and surface level (as seen in Chapters 2 and 3). In particular, Table 2 shows that the ‘natural’ coincidence between subject and Topic in unmarked sentences directly results from this correlation. Based on syntactic requirements, the argument promoted to the subject function takes sentence-initial position (Spec,IP); this element is, by default, assumed as the starting point of the predication and consequently interpreted as a Topic. This correspondence also plausibly explains why SVO and SOV languages are the most common in the world: The Topic is a (generally nominal) discourse category assumed as “what the sentence is about” (Reinhart 1981), and the subject is the syntactic function that exerts the highest control within a sentence (see the concept of superiority in Chapter 3, § 3.4.1.). It is clear that the correlation between Topic and subject also satisfies cognitive requirements of naturalness in the distribution of information. This prominent role of the subject-Topic in the sentence is confirmed by prosodic and intonational properties: As shown in recent studies, this type of NP is usually associated with a rising tone and forms a prosodic domain, which is separate from the rest of the sentence (Frascarelli & Trecci 2007, Frascarelli 2007, 2008). Let us consider the underscored sentence in the following text (all Italian sentences examined for intonation in this chapter are taken from Bonvino 2006):

228 The sentence as utterance (19) Praticamente i giochi avvenivano sempre durante la giornata cioè non è non era come com’è per noi adesso per cui magari che ne so uno va al cinema la sera o a teatro no - in realtà la vita dei romani avveniva sempre durante il giorno e quindi anche i giochi avvenivano durante il giorno. ‘Basically, the games always took place during the day, I mean, it’s not... it wasn’t like... like it is for us now, maybe, you know, you go to the movies in the evening, or to the theater, you know, really the life of the Romans always took place during the day and so the games took place during the day too.’

Figure 1. Subject-Topic intonational contour in a subject-initial language

After the rising tone marking the NP the life of the Rómans, the intonational curve departs again from a minimal value of the speaker’s fundamental frequency (the ‘baseline’ is indicated in the Figure with an arrow). This prosodic pattern indicates a prosodic boundary between the Topic-NP and the rest of the sentence, which shows an intonational peak on the adverb always followed by a downgrading contour (typical of Comments; see note 2 to Chapter 6). In predicate-initial languages, on the other hand, the verb occupies initial position in an unmarked sentence. Clearly, no correlation between subject and Topic can take place in this case; on the contrary, it must be assumed that in languages such as Tagalog (VSO) or Malagasy (VOS) the starting point of a predication is the verb itself. However, since the Topic is a nominal constituent, we expect that the verb cannot assume Topic function in the discourse. This prediction is borne out by prosodic data: Unmarked sentences in these languages present an upward-downward

Information structure in declarative clauses 229 contour, and the verb is not separated from the rest of the sentence. This is shown in Figure 2, corresponding to example (20): (20) B-um-abasa ang titser ACT.PST-read TRIG teacher ‘The teacher read a newspaper.’

ng DIR

diyaryo. (TAGALOG) newspaper

Figure 2. Intonational contour in a predicate-initial language

In this case, after the intonational peak on the verb (indicated with an arrow), the rest of the sentence proceeds downward with no restarts (i.e. within a single prosodic domain).9 This shows that the verb is definitely the starting point of the predication, but is not a Topic. In other words, in predicate-initial languages an unmarked sentence is made of a Comment alone (a Topic necessarily requires a marked construction which needs to be signaled in specific ways; see § 6.5.3.). Let us now examine the correlation between discourse categories, phrasal categories and syntactic functions in marked sentences. Consider the following question-answer pair: (21) Chi ha who have.PRES.3SG ‘Who called?’

telefonato? phone.PRT

(22) (Ha telefonato) have.PRES.3SG phone.PRT ‘MARIA (did).’

MARIA. Maria

(ITALIAN)

230 The sentence as utterance In formulating the question in (21), the speaker knows that there was a phone call, and asks his/her interlocutor to identify its author (unknown to the speaker); thus, in (22), the existence of ‘somebody who has called’ is presupposed information (and can be omitted), whereas the NP Maria is new information. Maria is therefore the content of the semantic predication, even though it is formally a noun. In fact, this option is not surprising, as NPs are normally predicates in copular sentences (see § 1.5.2.). This means that, in a sentence with a marked relationship between phrasal categories and discourse categories, the given part is realized by means of a VP structure, while the new part by means of a nominal constituent (an NP). In such cases, we can say that a Presupposition and a Focus assume, respectively, subject and predicate functions. In more explicit terms, we can conclude that a sentence like (22) is equivalent to a copular structure like: (23) [Colei che ha she that have.PRES.3SG è [MARIA]PRED. be.PRES.3SG Maria ‘The one who called is Maria.’

telefonato]SUBJ phone.PRT

Pragmatic markedness in (22) thus corresponds to syntactic markedness and is also reflected in prosody. If the Presupposition is overt, the constituent order necessarily becomes VS, and the intonational contour shows a pitch on the Focus in final position. As an example, see the following conversational exchange and, in particular, the underlined marked sentence which is represented in Figure 3 (and examined in Frascarelli 2004b): (24) A: Tu l’ hai chiesto S.PRO.2SG OCL.SGM have.PRES.2SG ask.PRT o te l’ ha detto lui? S.PRO.3SGM or IOCL.2SG OCL.SGM have.PRES.3SG say.PRT ‘Did you ask him or did he tell you?’

Information structure in declarative clauses 231 B: Io non ho mai parlato di S.PRO.1SG not have.PRES.1SG never speak.PRT of soldi, l’ ha sempre tirato OCL.3SGM have.PRES.3SG always draw.PRT money fuori LUI, il discorso. out S.PRO.3SGM DET topic ‘I’ve never talked about money, he’s always been the one to bring up the topic.’

Figure 3. Intonational contour of a Focus-final (VS) sentence

Note that the NP lui is marked by a pitch, whereas the verb is completely included in the downgrading contour of the sentence. Intonational data thus show that when the verb is given it is not a Topic, but simply presupposed information. The relationship between different levels of analysis in marked sentences is illustrated in Table 3:

232 The sentence as utterance Table 3. Correlation between levels of analysis in marked sentences pragmatics

discourse grammar

GIVEN

NEW

PRESUPPOSITION

FOCUS

ARGUMENT

PREDICATION

semantic value

phrasal category

VP

NP

syntactic function

SUBJECT

PREDICATE

We can now include in a single schema the analysis we have proposed for information structure in declarative sentences and for the TopicComment and Focus-Presupposition pairs: pragmatics

GIVEN

NEW

discourse grammar

TOPIC

PRESUPPOSITION

FOCUS

phrase

NP

VP

NP

COMMENT

VP

information structure syntactic function semantic value

(correlated with) SUBJ/OBJ

SUBJ

(correlated with)

PREDICATE PREDICATION

ARGUMENT

Figure 4. Information structure and levels of analysis

The schema in Figure 4 clarifies what is intended by the term information structure, that is to say, a strong correlation between the distribution of information, discourse categories and syntactic functions

Information structure in declarative clauses 233 (with links to semantic value and phrasal categories). Based on this correlation, what is given (i.e. what is talked about or is part of the common ground) corresponds to (or is correlated with the function of) subject (or object, as we will see); on the other hand, the new part is the predicate. This means that in marked structures (Focus-Presupposition), the predicate is a nominal element and the subject is the nominalization of a predication. Data supporting this hypothesis will be presented in the following section, from languages using different strategies in realizing Focus structures. To conclude this part of the chapter, note that Topic-Comment and Focus-Presupposition structures may combine. In particular, a Focus and a Topic may be present in the same sentence. This means that a nominal element, which would otherwise be part of the Presupposition, is made salient and proposed as a Topic. Consider the Italian question in (25): (25) Chi ha mangiato who have.PRES.3sg eat.PRT ‘Who ate the cake?’

la DET

torta? cake

(ITALIAN)

The Presupposition in (25) is that ‘x ate the cake’ and the interlocutor is asked to specify a value for the variable (x) contained in the question. The NP cake is thus part of the Presupposition and, consequently, can be realized as a clitic in the answer (cf. [26a]). However, the speaker might choose to realize the NP la torta as a Topic and link it to a clitic. This is because s/he might want to say more about it (in this case the Topic will appear on the left; see [26b]), or because s/he wants to repeat that element sentence-finally, as in (26c): (26) a. L’ OCL.3SGF

ha mangiata have.PRES.3SG eat.PRT.SGM

PAOLO. Paolo

PRESUPPOSITION

FOCUS

‘PAOLO ate it.’ b. La DET

torta, l’ cake OCL.3SGF

TOPIC

ha mangiata have.PRES.3SG eat.PRT.SGF

PAOLO. Paolo

PRESUPPOSITION

FOCUS

‘The cake, PAOLO ate it.’

234 The sentence as utterance

c. L’

ha

OCL.3SGF have.PRES.3SG PRESUPPOSITION

mangiata eat.PRT.SGF

PAOLO, la torta. Paolo DET cake FOCUS

TOPIC

‘PAOLO ate the cake.’ The NP la torta in (26b-c) is an extrasentential constituent (i.e. appearing outside the IP). This is shown syntactically by the fact that the VP includes a resumptive clitic pronoun which cannot be omitted. Based on the Thematic Criterion (i.e. a θ-role can be assigned to only one constituent), the Topic and the clitic cannot be merged in the same argument position. The Topic is therefore not generated in the VP and has no (direct) grammatical function in the sentence; rather, it is correlated with an argument clitic that determines its syntactic connection with the rest of the sentence (as in the case of subject Topics, which were discussed earlier).10 In conclusion, Tables 2 and 3 and the final schema in Figure 4 show the relationships between different levels of analysis in the communicative process determining the distribution of information. The interpretation of concepts such as new and given information does not take place within the domain of pragmatics alone, but requires the fundamental contribution of structure at all levels. Certainly, the categories of Focus and Topic are much more complex functionally than we have illustrated so far. Some points will be discussed in more depth in the following sections. Here, our preliminary intention has been to present and define them as a multifaceted and complex system (namely, information structure) relating to the interpretation of the illocutionary force of sentences.

6.4. Focus The term Focus denotes the constituent that conveys new information in a structure where the rest of the sentence is the Presupposition. As said earlier (see note 5 to Chapter 6), a constituent may be new not just when it answers a specific request for information, but also when it corrects or integrates previously stated information. It is therefore necessary to distinguish between two main types of Focus: a) Information Focus, which specifies the value of a variable within a set of alternatives; b) Contrastive Focus, where the alternative set must include the meaning of the utterance corrected (for details and discussion, see Rooth 1992, Kenesei 2006, Krifka 2007, Frascarelli 2010).11 Information Focus (henceforth, simply Focus) is

Focus 235 a constituent that typically answers a wh-question (see note 3 to Chapter 6 and [27] below), whereas a contrastive Focus is used when previous information needs to be corrected or excluded, as shown in (28)-(29):12 (27) A: Chi ha mangiato who have.PRES.3SG eat.PRT.SGM ‘Who ate the cake?’ B: L’ ha mangiata OCL.SGF have.PRES.3SG eat.PRT.SGF ‘Paolo did.’

la DET

torta? cake

PAOLO. Paolo

(28) A: Ho saputo che Leo ha have.PRES.3SG have.PRES.1SG know.PRT that Leo mangiato la torta. DET cake eat.PRT.SGM ‘I heard that Leo ate the cake.’ B: No: l’ ha mangiata PAOLO. No OCL.SGF have.PRES.3SG eat.PRT.SGF Paolo ‘No, Paolo did.’ (29) A: Ho saputo che Leo e Paolo have.PRES.1SG know.PRT that Leo and Paolo verranno alla festa. to.DET party come.FUT.3PL ‘I heard that Leo and Paolo are coming to the party.’ B: No: verrà solo PAOLO. No come.FUT.3SG only Paolo ‘No, only Paolo is.’ As clearly shown in (29), a contrastive Focus may be a referent mentioned in the previous context but still be new information. In structural terms, Focus since Chomsky (1977) has been analyzed as a syntactic operator (cf., among many others, Horvath 1986, Vallduví 1992, Kiss [ed.] 1995, Rebuschi & Tuller [eds.] 1999). Because of this property, Focus is a quantificational element and, as such, assumes scope over the rest of the sentence after movement (either overt or in LF) into the CP area (see § 6.4.4.) in order to specify a variable included in the Presupposition. In other words, the two sentences in the question-answer pair in (27) share a Presupposition. This Presupposition contains a variable (‘x’), which is specified by the Focus thanks to its scope position (∃ = ‘exists’):

236 The sentence as utterance (30) Chi ha mangiato la torta? DET cake who have.PRES.3SG eat.PRT ‘Who ate the cake?’ [PRESUPPOSITION= ∃ x, x ate the cake; who = operator requiring specification of x] ha mangiata PAOLO. have.PRES.3SG eat.PRT Paolo [PRESUPPOSITION = ∃ x, x ate the cake; PAOLO = Focus-operator specifying a value for x]

(31) L’

OCL.SGF

It follows necessarily from these semantic-syntactic considerations that Focus is unique per sentence. Constructions with multiple informative Foci are ungrammatical (for interface evidence and discussion, cf. Bocci 2009). Given these properties, which are assumed to be universal, we will now see how Focus is realized in typologically different languages.

6.4.1. Focus strategies in typologically different languages The world’s languages resort to different strategies for expressing new information. Several mechanisms have been proposed for their description, some of which are based on prosodic means alone, others based on morphology and many indicating syntax as a major means for interpretation.13 In fact, despite cross-linguistic variation, languages can be classified according to three main strategies for Focus realization. These are: a) the in situ strategy, b) the extra situm strategy,14 and c) the copular strategy. The in situ strategy is the syntactically simplest case, as the focused constituent appears in its unmarked position. This is what happens in languages such as English and Italian. Given a question like What did you give Leo?, the Focus is post-verbal in the answer: (32) I gave him A BOOK. (33) Gli

ho dato have.PRES.1SG give.PRT LIBRO. book ‘I gave him A BOOK.’

IOCL.3SGM UN INDEF.DET

(ITALIAN)

Focus 237 In this case, following the hypothesis that the Focus is a syntactic operator, it is assumed that the scope movement to the CP area does not take place overtly, but in Logical Form.15 In other languages, the Focus must be realized in the CP area: Any other position for the Focus yields ungrammaticality.16 This means that Focus movement to scope position must be overt and, for this reason, this construction is described as extra situm. Languages using this strategy include Hungarian and Basque, as illustrated in the following examples (respectively from Horvath 1995 and Ortiz de Urbina 1999): (HUNGARIAN)

(34) MARI jött el tegnap. Mari come.PST PV yesterday ‘MARI came yesterday.’ irakurri (35) JONEK read.PST Jon.ERG ‘JON read the letter.’

du AUX

eskutitza. letter

(BASQUE)

Finally, a great number of languages employ a strategy where Focus realization is closely related to copular constructions and to the presence of relative clauses (see Schachter 1973, Appleyard 1989, Ouhalla 1999). Also in these languages Focus appears in initial position (as in the extra situm strategy), but with crucial differences. Let us observe the following data from Tigrinya, Chaha (both from Appleyard 1989) and Wolof (Kihm 1999): ‘Ψyyu waga (36) NΨSSU PRO.3SGM COP.3SGM price zäxΨbbΨr. lift.REL ‘HE raises prices in town.’

‘ab in

kätäma town (TIGRINYA)

MÄDÄR-u yäcänänä. (37) BÄHΨ from.DEM place-COP come.PST.REL ‘We come FROM THAT PLACE.’

(CHAHA)

WI la jaaykat bi jënd. (38) FAS horse DET COP.3SG merchant DET buy.PST.REL ‘The merchant bought THE HORSE.’

(WOLOF)

238 The sentence as utterance As we can see, in these sentences the Focus is followed by a copula, which may have a resumptive pronoun incorporated, as in (36) and the verb of the following clause (the Presupposition) shows a specific paradigm, i.e. the one used in relative clauses. Therefore, what follows the copula is not simply the rest of the sentence (from which the Focus has been extracted), but an independent (relative) clause that formalizes the Presupposition. This means that, unlike extra situm Focus, the present strategy requires a bi-clausal construction. This structure is reminiscent of another construction used for Focus in Indo-European languages, namely the cleft sentence (as in It is JIM that is going to the movies; cf., among others, Hedberg 1990, Collins 1991, Declerck 1988, Den Dikken 2001, Frascarelli & Ramaglia in press). In these constructions, too, the Focus appears sentence-initially and is followed by a relative clause (see Frascarelli 2000b). However, there is a basic difference from the strategy discussed here. In cleft sentences Focus is often associated with contrast, whereas the Focus in (36)-(38) is informative.17 This led Frascarelli (2010) to define it as the ‘cleft-like’ strategy. This strategy is frequently used in creole languages, although in these cases no overt marker on the verb signals its relative paradigm. This is a consequence of simplification in these languages, which usually lack inflectional features. Let us consider sentences (39) and (40), from Bickerton (1993) and Veenstra & den Besten (1994) respectively: (39) A

JAAN COP Jaan ‘JAAN saw me.’

sii see

me. O.PRO.1SG

(40) Na

AKI wi COP Aki S.PRO.1PL ‘We saw AKI.’

bin PST

(GUYANA CREOLE)

si. see

(KRIO)

This strategy also includes those languages where Focus markers are used: According to recent studies, these markers are derived from original copular forms (see Frascarelli & Puglielli 2005a, Frascarelli 2010). Sentences like (41) from Somali can therefore be equated to copular constructions: (41) CALI baa soomaali FM Somali Cali ‘CALI is Somali.’

ah.18 be.RED

(SOMALI)

Focus 239 Finally, in some cases this strategy is found in the absence of a copula (a possible option for copular sentences in many languages; see § 2.1.). However, other morpho-syntactic elements show that this is the same construction we have examined here. For instance, the presence of a relative pronoun between the Focus and the Presupposition (as in [42], from Ouhalla 1999),19 or an element such as the trigger in Tagalog (43): al-ladi (42) HUWA PRO.3SGM DET-REL.PRO ‘I saw HIM.’

ra’aytu. see.PERF.1SG

ibinigay ng titser (43) LIBRO ang book TRIG PAT.give.PST DIR teacher sa estudiante. OBL pupil ‘The teacher gave the pupil A BOOK.’

(ARABIC)

(TAGALOG)

Since, as we know, the trigger always precedes a nominal constituent, the fact that it is followed by a verb in (43) leads us to assume that there must be a null (generic) NP head between ang and ibingay functioning as the head of the following relative clause. In other words, (43) can be paraphrased as ‘it is A BOOK, the thing that the teacher gave the pupil’ (see also Chapter 4, § 4.4.2.3.). In conclusion, Focus interpretation according to this strategy requires a copular structure where the Presupposition qualifies as a nominalized sentential structure (as it is included in a relative clause). What does this mean structurally? How are these elements inserted into the sentence? And, most importantly, how is this strategy related to the previous ones and to the hypothesis of Focus as a syntactic operator? An answer to these questions will be provided in the following section. 6.4.2. Focus as a specificational predicate As is well-known, copular sentences are Small Clauses merged as the complement of a superordinate copula (see note 1 to Chapter 2):

240 The sentence as utterance

(44)

CP C’ IP I’ SC

COP SUBJECT

PREDICATE

Once we have ascertained that in the cleft-like strategy the Focus moves to a pre-copular position (in the CP area) while the presupposed relative clause remains in situ, we need to define their merge positions in the Small Clause. In other words, we want to establish whether the Focus or the Presupposition is generated as the predicate in the relevant structure. A thorough analysis here is beyond the scope of this book (see Frascarelli & Puglielli 2005a, Frascarelli 2010); we will therefore only consider data which can provide an immediate (albeit simplified) account. First, note that the Focus in these languages does not show the syntactic properties of a nominal constituent merged into an argument position. For instance, a focused subject does not bear NOM Case, but the morphologically unmarked Case of the language which is typically used for predicative NPs. Furthermore, the verb in the relative clause does not fully agree with the focused subject, which results in the phenomenon known as Antiagreement (see Ouhalla 1993, Frascarelli 1999). Consider the following sentence (from Frascarelli & Puglielli 2005a): (45) Hilib-ka NIMAN-KA baa cunayá. (SOMALI) meat-DET.ABS men-DET.ABS FM eat.PRES.PROG.RED(REL) ‘THE MEN are eating the meat.’ The NP interpreted as the subject of ‘eat’ is marked for ABS Case and the verb of the relative clause shows reduced agreement. NOM Case or a full verbal agreement would make the sentence in (45) ungrammatical: (46) a. *Hilib-ka meat-DET.ABS b. *Hilib-ka meat-DET.ABS

NIMAN-KU baa men-DET.NOM FM NIMAN-KA baa men-DET.ABS FM

cunayá. eat.PRES.PROG.RED(REL) cunayaan. eat.PRES.PROG.3PL

Focus 241 Similar phenomena are found in many other languages using this strategy. In Kikuyu (from Schwarz 2004), for example, subject focusing requires a participal form of the verb: (47) Ne

KÁRIOKI ó-tem-ire/ COP Karioki PRT-cut-PST ‘KARIOKI cut a tree.’

*á-tem-ire 3SG-cut-PST

mote. tree

(KIKUYU)

Another fact is particularly significant. Not only does a focused subject not bear NOM Case, but in these languages a NOM Case marker appears on the relative clause, i.e. on the nominalized clause which expresses the Presupposition. This is more clearly seen in dependence-marking languages (see § 3.5.) such as Somali and Afar, as the Case marker appears on the rightmost constituent within the relative clause: (48) Wiilkaas baa [aan af boy.DEM.ABS FM NEG language hadlini]. speak.NEG.NOM ‘THAT BOY doesn’t speak Italian.’

Talyaani-ga Italian-DET.ABS

ku in (SOMALI)

(49) ÀWKA [huurí-h adda-l kullumta habte.m]. (AFAR) boy.ACC boat-of inside-in fish leave.PST.3SGF.NOMIN.NOM ‘THE BOY left the fish in the boat.’ It can be concluded that the cleft-like strategy requires a complex structure where the Focus is merged as the predicate of a Small Clause, while the Presupposition is a free relative clause20 inserted as a subject whose generic NP head is the variable to be specified. The Focus moves to the CP area in order to reach the functional projection (the Focus Phrase, see [52] below) where it is interpreted, and assumes scope over the Presupposition. A sentence like ‘CALI is a Somali’ (see [41] above) may be therefore paraphrased as ‘CALI is the person who is Somali’ and analyzed as follows (for details, see Frascarelli 2010):

242 The sentence as utterance

(50)

CP IP NP

I’ SC

COP

CALI[+foc]

tCali

DP D’ CP C’ NP value k specification

that

IP tk is a Somali

The idea that new information is the predicative part of a sentence has been discussed in many studies from different perspectives and with different results (cf. Higgins 1973, Szabolcsi 1981, Ajello 1995, Huber 2000). Chomsky (1971: 72) also states that “the focused constituent is the predicate of a dominant sentence” and ascribes the semantics of an assertion of identity to the Focus-Predication structure. In other words, what Chomsky argues is that the interpretation of the sentence (51a), where the PP in his study is focused, corresponds to that of the copular sentence (51b): (51) a. John writes poetry IN HIS STUDY. b. The place where John writes poetry is HIS STUDY. This hypothesis has thus been considered in the literature. However, no syntactic formalization has yet been proposed. Hence, the crucial contribution of the present analysis21 is that the Focus is proposed as a predicate not only in semantic-functional terms but also syntactically. Since an NP can assume a predicative role only within reduced clauses, the Small Clause is the only appropriate structure for this construction. Once the Focus is inserted as a predicate, its derivation in the CP area is due to discourse grammar requirements.

Focus 243 Now that the structure and derivation of the copular strategy have been clarified, let us consider the final question asked at the end of § 6.4.1., i.e. how this strategy is correlated with the previous two. As we have said, semantic-pragmatic considerations clearly attest that new information is the predicative portion of a sentence. Using a copular structure, we then showed that this correlation (Focus = predicate) is encoded in syntax where its interpretation is made possible. It is therefore plausible to suggest that syntax is also the interface between semantics and discourse grammar for the other two strategies. Our hypothesis is therefore that Focus is a predicate in the in situ and extra situm strategies as well. In other words, the sentence in (51b) is not simply a paraphrase of (51a), but rather the way in which a Focus is generated, derived and interpreted at a syntactic level in the world’s languages (i.e. as the predicate of a copular structure). The basic difference between the languages examined in (32)-(35) and in (36)-(41) is that the copular structure has been preserved more overtly in the latter, whereas it has been grammaticalized (and, consequently, simplified) in the former where the original structure has been obscured. This is why a relative clause can no longer be identified in the Presupposition in a number of languages such as English and Italian. However, the fact that in many languages Focus bears non-NOM Case even when it is interpreted as a subject (see [45] and [49]) cannot be accidental: This difference depends on the Focus being generated as a predicate, which means that it does not move to Spec,IP to take NOM Case but assumes the default Case for the language in question.

6.4.3. Focus-prominent languages From a discourse-grammar point of view, the world’s languages are often classified into three types: Subject-prominent, Topic-prominent and Focusprominent (see Comrie 1981, Kiss [ed.] 1995 among others). This distinction refers to the fact that languages belonging to each of these groups show phenomena which reveal that either subject, Topic or Focus have a higher level of control in the syntactic relations within the sentence (e.g. null subject interpretation [cf. Frascarelli 2007], superiority effects [cf. Büring & Hartmann 1994, Haider 2000], and so on). In particular, Focusprominent languages (cf. Kiss [ed.] 1995)22 are characterized by a Focus in a fixed scope position, necessarily adjacent to the verb or a verb-related functional head.

244 The sentence as utterance To account for this rigid structural requisite, authors such as Horvath (1986) and Brody (1990) proposed that Focus-related semantic-pragmatic information is codified within a feature, which is generated in the head of a specific functional projection in the CP area (where the syntax-discourse grammar interface takes place). In other words, along with an extended and complex functional area within the DP (Chapter 4, § 4.3.) and IP (Chapter 5, § 5.3.), there must be specific functional projections in the CP area, each dedicated to a discourse grammar category. These authors assume that the [+foc] feature is generated in the head of the Focus phrase (FocP).23 Hence, in Focus-prominent languages, the Focus needs to move to the FocP specifier to activate the interpretation of the [+foc] feature; non-Focusprominent languages, on the other hand, do not have to satisfy this requirement and can maintain the focused constituent in its unmarked position (i.e. in situ), as in English or Italian. However, cross-linguistic analysis shows that in situ Focus has the same interpretive properties as extra situm Focus (see Rizzi 1997, Frascarelli 2000a among others). Brody (1990) argues that the [+foc] feature must be present in every language (which is plausible from a semantic-pragmatic point of view) and is activated for in in situ Focus languages as well, though in Logical Form. That is why its effects are not overt. In conclusion, a classification of languages in terms of inclusion in the Focus-prominent group does not appear to have a bearing on interpretation. Nevertheless, it can help single out those languages where specific derivational processes involving the Focus in the CP area are overtly realized.

6.4.4. Syntactic interface: The complex structure of CP Based on what has been said so far, we can conclude that Focus is an element that moves from the position where it was merged (as a predicate) to a functional projection in the CP where the [+foc] feature is encoded, and that this operation can take place in overt syntax (extra situm and cleftlike strategy) or in covert syntax (in situ Focus languages). To account for the relationship between syntax and discourse categories, a division of the CP node into a number of functional projections has been proposed in recent years. Each of these projections is dedicated to the encoding and interpretation of a specific semantic-pragmatic feature. This approach, which was described as cartographic (see Cinque [ed.] 2002, Rizzi [ed.] 2004, Belletti [ed.] 2004) led to the definition of the following hierarchy within the CP area, also known as the left periphery of the

Focus 245 sentence (Rizzi 1997), where the FocP is located in between two Topic positions:24 (52)

ForceP Force°

Force’ TopP

CP area

Top’ Top°

FocP Foc’ Foc° [+foc]

TopP Top’ Top°

FinP Fin’ Fin°

IP/SC

The functional projections that make up the original CP node are extremely important for information structure as they represent the relationship between the sentential content (IP or SC) and discourse grammar. Along with Focus and Topic projections, Rizzi (1997) includes a Force and a Finiteness projection in the left periphery. The idea is to allow a structural position for the interpretation of speech acts (see § 6.2.), encoded in Force°, as well as explaining why the choice of a specific speech act has important consequences at a morpho-syntactic level (the degree of finiteness of the verb, subject agreement, etc.) within the sentential structure (encoded in Fin°). According to Rizzi: Complementizers express the fact that a sentence is a question, a declarative, an exclamative, a relative, a comparative, an adverbial of a certain kind, etc., and can be selected as such by a higher selector. This information is called… the specification of Force… The choice of the Complementizer reflects certain properties of the verbal system of the clause… I will assume that the CP-system expresses a specification of finiteness, which in turns selects an IP system with the characteristics of finiteness: mood distinctions, subject agreement licensing nominative case, overt tense distinctions. (Rizzi 1997: 283-284)

246 The sentence as utterance The relationship between speech acts, illocutionary force and syntactic structure will be dealt with in Chapter 7. We will conclude our discussion of Focus here, with some considerations on its phonological properties.

6.4.5. Prosodic interface The idea that Focus is associated with a specific tonal event is shared by various authors. In particular, within Generative Grammar, the association between Focus and sentential pitch was already identified in early studies on this topic. Chomsky (1971: 201) asserts that “the Focus is the phrase containing the intonation center” and, according to Jackendoff (1972: 234), “the Focus Marker contains a feature marking the pitch contour”. These intuitions have been confirmed in recent years by empirical analyses performed using specific tools and softwares for the measuring of fundamental frequency. In particular, research in this field has shed light on an important phonological distinction between Focus and Comment: They are both new information, but have a different prosodic realization: the Comment is included in a continuous, downgrading contour (see note 2 to Chapter 6), whereas the Focus forms a single prosodic group with the verb and requires extraposition of non-focused constituents in distinct intonational phrases (cf. Frascarelli 2000a, 2004b). In other words, prosodic and intonational analyses show that a non-focused NP/PP in a Focus construction is necessarily a Topic, consistent with the argumentation discussed in § 6.3.2.25 This can be verified by analyzing sentences such as that in (24), illustrated in Figure 3 (repeated here as 3bis):

Figure 3bis. Intonational contour of a post-focal NP

Topicalization 247 As is shown, after the pitch on the Focus lui a short break is produced and the NP il discorso is uttered in a low tone, as is typical of right-hand Topics (see below).

6.5. Topicalization As we saw in (52), the CP area has been extended to include a functional projection dedicated to new information (FocP) and two positions for given information, both generically called Topic Phrases (TopP). Based on Rizzi’s (1997: 295-297) hypothesis, these projections are freely recursive, meaning that they may be freely iterated. This would account for the fact that, in languages like Italian, a potentially unlimited number of Topics can appear in a sentence, as opposed to Focus uniqueness. This is shown by the following contrast: (53) Domani, a Leo, questa storia, con ogni tomorrow to Leo this story with each probabilità, non gliela racconterò. (ITALIAN) probability NEG IOCL.SGM-OCL.SGF tell.FUT.1SG ‘Most probably, I will not tell Leo this story tomorrow.’ (54) a. *A chi che cosa domani racconterai come? to who what thing tomorrow tell.FUT.2SG how ‘*To whom what are you going to tell how?’ A LEO b. *QUESTA STORIA domani this story tomorrow to Leo racconterò CON GIOIA. tell.FUT.1SG with joy [lit.: ‘*THIS STORY I am going to tell TO LEO GLADLY.’] However, we will see that the supposed recursion of Topics is not at all free. Indeed, different types of Topics must be distinguished, based on their formal and semantic-pragmatic properties. We will conclude that there are in fact specific restrictions on the number and linear order of Topics in a sentence, and we will identify their hierarchy in an attempt to improve the structure in (52) (as argued in Frascarelli & Hinterhölzl 2007).

248 The sentence as utterance 6.5.1. The functions of Topic in discourse The realization of Topics is often associated with given information as a means of emphasizing salient discourse referents. This generalization, however, is only partially correct and can lead to erroneous conclusions about the actual nature of topicalization. First of all, as previously said (see note 19 to Chapter 5), the notion of ‘given’ should not be automatically associated with previous mention of the relevant referent. A Topic may be part of the shared knowledge (and thus present in the speakers’ minds), but not necessarily active (or semi-active). This is the case of new Topics, as described in Frascarelli (2003). We will come back to this, when discussing the Aboutness-shift Topic (Frascarelli & Hinterhölzl 2007). An example of this type of Topic is given in (55) (from De Mauro et al. [eds.] 1993): (55) A: ho sentito con piacere che hai divulgato molto jazz ‘I was glad to hear that you have been broadcasting a lot of jazz’. B: ah sì sì ma ora eh se porteremo a fondo il nostro progetto il sabato_ sera ci sarà un avvenimento sportivo intervallato da brani esclusivamente jazzistici ‘Oh yes, right but now if we manage to go on with our plan, there will be a Saturday-night sports event with breaks for jazz pieces only.’ A: bene ‘Good.’ B: quindi questo lo dovremo eh dire non appena potremo dirlo però eh sarà questa la nostr il nostro progetto ‘So we will have to say this as soon as we can, anyway this will be our plan.’ A: e Caniggia lo pigli? ‘And Caniggia, are you going to get him?’ B: ah è probabile_ di sì è probabile di sì ‘It’s likely, yes, it’s likely, yes.’ A: mi piace a me quel giocatore [...] ‘I like that player.’ In this radio conversation (between a host and a football team manager), the referent Caniggia was not previously mentioned. It is nonetheless introduced as a Topic in the host’s question as a way of changing subject and surprising the other speaker.

Topicalization 249 Secondly, ‘emphasizing’ is far too vague a notion and cannot account for the various functions that Topics fulfill in a sentence. If one is to identify a common semantic-pragmatic function of Topics, then the best description is Reinhart’s (1981), later developed by Lambrecht (1994). In this description, Topics are connected with whatever the speaker assumes as the subject matter of his discourse. Quoting Reinhart, the Topic is “what the sentence is about”. Topicalized constituents are therefore all connected with a property which may be defined as aboutness, in pragmatic terms. However, as we said, this pragmatic characterization assumes very different values and formal properties when Topics are actually used in sentences. Let us consider the following sentence from De Mauro et al. (eds.) (1993), discussed in Frascarelli & Hinterhölzl (2007): (56) La situazione è questa: l'insegnante come ho detto ai ragazzi è in maternità ha una gravidanza difficile e sta usufruendo di quella legge particolare della maternità anticipata per ora ha avuto un mese io penso che non tornerà però lei m'ha detto ah di non dirlo ancora ai ragazzi perché per motivi suoi- comunque io signora penso di chiudere l'anno […] questo comunque io ai ragazzi non l'ho detto direttamente ‘This is the situation: As I told the class, the teacher is on maternity leave, she is having a difficult pregnancy and is benefiting from that particular law on early maternity leave, she has had a month so far, I don’t think she is coming back, but she told me not to tell the class yet, for personal reasons, anyway, madam, I think I’m going to stay until the end of the year... This, however, I haven’t told the children explicitly.’ The underscored sentence contains three Topics: The NP questo (correlated with a DO clitic), the pronoun io (correlated with a null subject in the sentence) and the PP ai ragazzi (correlated with the IO function). The speaker is a teacher talking to the mother of one of his pupils about his position as substitute teacher. I ragazzi is clearly a Familiar Topic in this conversation, as it is mentioned at the beginning and repeated several times. The pronoun io, on the other hand, is contrastive, since the teacher’s aim is to argue that s/he is not going to tell the class about his/her true situation, although s/he cannot exclude that others are. Finally, the DO Topic questo marks a turning point in the conversation: Once the teacher has finished discussing the general situation, his/her purpose is to divert the interlocutor’s attention to a specific topic.

250 The sentence as utterance This example indicates that (at least) three types of Topics must be specifically identified from a semantic-pragmatic point of view. Based on the analysis in Frascarelli & Hinterhölzl, we will show that these Topics correspond to specific formal (intonational and structural) properties. When it is used to propose, i.e. introduce or re-introduce a subject matter, a Topic occupies the highest position in the left periphery of the sentence, with a rising intonational contour and a prosodic domain separated from the rest of the sentence. This Topic may be newly introduced, and therefore new in the discourse (as in [55]), but its being new has different implications for information structure than in the case of a Focus or a Comment. Here, what is new in the conversational dynamics (cf. Krifka 2007) is the instruction to change to a new topic, not its informational content. This type of Topic thus combines the aboutness property with a shifting function. For this reason, it is described as Aboutness-shift Topic by Frascarelli & Hinterhölzl (2007) and illustrated by sentences like (57) below (Topics are given in bold here and in the following examples): (57) Il materiale era tantissimo quindi all'inizio l'ho fatto tutto di corsa cercando di impiegarci magari anche il tempo che dicevate voi magari facendolo un po' superficialmente pur di prendere tuttol'ultima unit la sto facendo […] ‘The material was quite a lot, so at the beginning I did it all in a rush, trying to do it in the time that you had fixed, perhaps a little superficially, so as to do everything- the last unit, I’m doing it now […]’

Figure 5. Intonational contour of the Aboutness-shift Topic

Topicalization 251 In this short text, a student is expressing her opinion about a self-study language course. She first speaks in general terms, then wants to shift to a specific point, namely the final unit of the course. She therefore proposes this element as a left-hand Topic (resumed by the clitic la) and produces it with a rising contour, whose rise begins on the diphthong [ju] and peaks on the post-tonic vowel. The rest of the sentence (the Comment) does not present significant variations in its curve. When the aboutness property is associated with familiar/given information (i.e. mentioned or inferrable by the context/situation), the function of the Topic is to maintain topic continuity (see Givón 1983). When given information is not associated with aboutness, the relevant Topic only serves to retrieve backgounded information (especially in the case of a distant reference, for clarification). This type of Topic (called Familiar) can appear either in a low position in the left periphery, or at the end of the sentence, as a right-hand Topic.26 In both cases the Familiar Topic is characterized by a low tone. Let us take examples (58) and (59), as discussed in Frascarelli & Hinterhölzl (2007). In the former, the Familiar Topic (the NP inglese) follows an Aboutness-shift Topic and precedes the Comment, which signals the hierarchical order between these elements. In the latter, the Familiar Topic (la parola) follows the Comment in sentencefinal position. (58) Era tutto molto nuovo nel senso che comunque la lingua inglese attraverso i programmi sul computer diciamo non l’avevo […] comunque l’inglese risultava anche facendolo da solo più interessante […] io inglese non- premetto non l’avevo mai fatto] ‘Everything was totally new to me in the sense that I had never studied English through computer programs […] but doing English on my own was more interesting […] I have to add that I’d never studied English before.’

252 The sentence as utterance

Figure 6. Intonational contour of an Aboutness-shift and a Familiar Topic

(59) A:

B:

A:

B:

Come impari le parole? le studi sui libri? e come le impari quando le studi davanti al computer? C’è qualcosa di diverso? ‘How do you learn words? do you study them in books? and how do you learn them when you study with a computer? is there anything different?’ in genere nei libri aspetto cioè leggo tutto e aspetto di capirla magari nel computer se so che c'è il vocabolario sono portata a cercarla subito ‘well, in general, when I study in books I wait, that is to say, I read the whole text and try to understand it – perhaps with the computer, if I know there’s a dictionary, I usually look it up straight away il fatto che ci sono delle immagini non ti aiuta a capire la parola maggiormente che in un testo? ‘doesn’t the fact that there are pictures help you to understand the meaning of a word better than reading a text?’ sì quello sicuramente cioè basta che mi trattengo un po’ e probabilmente nemmeno ho bisogno di cercarla la parola ‘yes, of course – that is, if I just hold back and, probably, I do not even need to search the word.’

Topicalization 253

Figure 7. Intonational contour of a Right-hand Familiar Topic

Finally, Topics can create oppositional pairs with the effect of contrasting the information contained in the relevant Comments. In this case we have a Contrastive Topic (see Kuno 1976, Büring 1999, Molnár 2002). We will come back to the notion of Contrast in § 6.6. For the moment, we will simply illustrate it through the two underscored sentences in the following text: (60) A: Come mai hai fatto due lingue, cioè, inglese e francese? ‘Why did you take two languages, I mean, English and French?’ B: francese l’ho fatto alle medie per tre anni con una professoressa con cui mi sono trovata benissimo […]- con l’inglese mi son trovata sempre a disagio. ‘French I studied at school for three years with a professor that I liked a lot […]with English I never felt at ease.’ In this case, the rising tone that characterizes the Topic no longer peaks on the post-tonic vowel (as in the case of Aboutness-shift Topics), but on the tonic syllable. This formal distinction in extremely important, as it makes it possible to distinguish between these Topics also prosodically.

254 The sentence as utterance

Figures 8a-b. Intonational contours of Contrastive Topics

Based on what has been said so far, the hierarchy in (52) must be reconsidered and substituted with the one given in (61), where ShiftP is the functional projection hosting the Aboutness-shift Topic, FamP is the position of Familiar Topics and ContrP is used for Contrastive elements. A specific semantic-functional feature is encoded in each of these projections, determining the intepretation of the Topic inserted in the corresponding Spec position. The asterisk following FamP means that this is a recursive projection. Indeed, while a sentence may contain just one Aboutness-shift and one Contrastive Topic, multiple Familiar Topics are allowed, since different backgrounded discourse referents may be retrieved within the same sentence.27

Topicalization 255 (61)

ForceP Force’ ShiftP Force° Shift’ ContrP Shift° Contr’ FocP Contr° Foc’ Foc°

FamP* Fam’ Fam°

FinP Fin’ Fin°

IP/SC

6.5.2. Phrasal restrictions on the topicalized constituent As we have seen, the Topic is an extra-sentential element inserted on the left of FinP (which can also appear right-dislocated after the Comment). In our overview of discourse grammar categories (see § 6.3.2.), we repeatedly noted that the Topic function can only be fulfilled by an NP (possibly embedded in a PP). We will now show how this syntactic restriction is highly significant since any NP may be a Topic regardless of its correlation with an argument role, the function of its coreferential clitic and its internal structure. Consider the following Italian sentences: (62) a. Anna, la incontro molto spesso. (ITALIAN) Anna OCL.3SGF meet.PRES.1SG very often ‘Anna, I meet her very often.’ b. In giardino, ci lavoro volentieri. in garden OBLCL work.PRES.1SG readily ‘In the garden, I readily work.’ c. Di queste cose, preferisco non parlarne. things prefer.PRES.1SG NEG speak.INF-OBLCL of DET ‘About these things, I’d rather not talk.’

256 The sentence as utterance d. Con le matite, non ci scrivo mai. with DET pensils NEG OBLCL write.PRES.1SG never ‘With pencils, I never write.’ e. Con il mio amico Fabio, ci friend Fabio OBLCL with DET POSS.1.SG esco nel fine settimana. go out.PRES.1SG in.DET end week ‘With my friend Fabio, I go out at the weekend.’ f. (Quanto a) Maria, esce sempre volentieri. as for Maria go out.PRES.3SG always eagerly ‘(As for) Maria, she is always keen on going out.’ As we can see, only the Topics in (62a) and (62f) are correlated with an argument position in the sentence; all the others convey additional information of various nature. Notice that even when the topicalized constituent seems to be a sentence, it is actually a nominalized clause, as was discussed in Chapter 3, § 3.5. So, the Topics in (63a-b) are themselves NPs, just like the deverbal nouns in (64a-b): (63) a. Partecipare alla riunione, Maria non lo participate.INF to.DET meeting Maria NEG OCL.SGM fa volentieri. eagerly do.PRES.3SG ‘Maria is not happy about attending the meeting.’ b. Che Leo arrivi sempre in ritardo always in delay that Leo arrive.SUB.PRES.3SG lo sappiamo, e non possiamo OCL.SGM know.PRES.1PL and NEG can.PRES.1PL farci niente. do.INF-OBLCL nothing ‘That Leo always arrives late we know, and we can do nothing about it.’ (64) a. Una

tale partecipazione alla riunione non such participation to.DET meeting NEG me la aspettavo. IOCL.1SG OCL.3SGF expect.PST.1SG ‘Such a large turn out at the meeting I didn’t expect.’ INDEF.DET

Topicalization 257 b. L’

arrivo di Leo in ritardo, francamente, arrival of Leo in delay frankly non ha stupito nessuno. NEG have.PRES.3SG surprise.PRT nobody ‘Leo’s late arrival, frankly didn’t surprise anyone.’ DET

As shown in (63a) and (65a) below, the topicalization of a nominalized clausal constituent may require a light verb (such as fare, ‘do’), functioning as a pro-form for the infinitival nominalized clause in Topic position. The semantic root of the nominalized verb may also be repeated within the sentence, this time in an inflected form (as in [65b]). In such cases, the topicalized NP must not contain complements (see [65c], ungrammatical): (65) a. Mangiare, Leo non lo fa volentieri. Leo NEG OCL.SGM do.PRES.3SG willingly eat.INF ‘As for eating, Leo does not do it willingly.’ b. Mangiare, Leo non mangia volentieri. eat.INF Leo NEG eat.PRES.3SG willingly ‘As for eating, Leo does not eat willingly.’ c. *Mangiare la pasta, Leo non mangia eat.INF DET pasta Leo NEG eat.PRES.3SG la pasta volentieri. DET pasta willingly One final observation is in order regarding the semantic properties of the topicalized NP and in particular, its being necessarily characterized by a [+specific] feature. As argued in Chapter 4 (§ 4.3.), the concept of specificity corresponds to the semantic-pragmatic function of referring to an element in the discourse or to the speaker’s knowledge. It should therefore not be confused with definiteness. It is thus possible to topicalize both definite NPs (66) and indefinite NPs (67), as well as quantified NPs (68), provided a specific referent is intended (for further discussion, see Cinque 1990b, Ihsane & Puskás 2001, Frascarelli 2007): (66) a. Il

di Leo non lo of Leo NEG OCL.3SGM vedo da anni. see.PRES.1SG from years ‘Leo’s brother, I haven’t seen him in years.’

DET

fratello brother

258 The sentence as utterance b. Non gli

voglio parlare, want.PRES.1SG talk.INF al tuo amico. to.DET POSS.2.SG friend ‘I don’t want to talk to him, your friend.’

NEG IOCL.3SGM

(67) a. Una

persona così gentile non la person so kind NEG OCL.3SGF trovi più al giorno d’ oggi! find.PRES.2SG anymore to.DET day of today ‘People as friendly as him, you don’t find these days!’ b. Oggi gli studenti ce l’ hanno today DET students OBLCL OCL.3SGF have.PRES.3PL tutti, una macchina. INDEF.DET car all ‘These days every student has a car.’ INDEF.DET

(68) a. Qualcuno l’ ho incontrato al bar, someone OCL.3SGM have.PRES.1SG meet.PRT at.DET bar ma nessuno di interessante. but nobody of interesting ‘Some people I did meet in the bar, but nobody interesting.’ b. La dice anche lui, qualche bugia, OCL.3SGF say.PRES.3SG also he some lie ogni tanto. every while ‘He, too, tells the occasional lie.’

6.5.3. Topicalization strategies in typologically different languages After examining the semantic-pragmatic properties of topicalization with respect to information structure, we will now show the various topicalization strategies in typologically different languages. This will reveal a correlation between the specific characteristics of Topics and the general properties of the grammatical systems of the languages in question. A preliminary distinction should be made between languages with and without clitic/weak pronouns. In the former, pronouns are used to anaphorically resume topicalized elements, whereas in the latter the topicalized constituent is not overtly correlated with its syntactic function

Topicalization 259 within the sentence. Clitic resumption was illustrated in the previous sections for Italian, an inflecting language. Other inflecting languages with similar properties include French and Arabic: (69) Ce

livre, à Jean, je le lui book to Jean S.PRO.1SG OCL.3SGM IOCL.3SGM donnerai demain. (FRENCH) give.FUT.1SG tomorrow [lit.: ‘This book, to Jean, I will give it to him tomorrow.’] DEM

(70) Al-kitab, li Ahmad, ’actayatahu Mariam. (ARABIC) DET-book to Ahmad give.PST.F-OCL.3SGM Mariam [lit.: ‘The book, to Ahmad, Mariam gave it to him.’] This mechanism is also highly productive in polysynthetic languages, since, as we know, argument structure in these languages is realized by means of clitics incorporated on the verb, whereas full NPs (or PPs) all appear in dislocated positions to the left or right of the verbal complex. Let us observe the following sentences from Warlpiri (Jelinek 1984) and Somali: (71) Ngujulu-rluk ka-rnak-ngkuj-rlaz. 1SG-ERG PRES-SCL.1SG.NOM-OCL.3SG.ACC-OCL.2SG.BEN warri-rniz nyunktu-ku. (WARLPIRI) karli-kiij boomerang-DAT NONPST-2SG.DAT seek-NONPST ‘I am looking for a boomerang for you.’ (72) a. Axmedk annagaj buu (baa-uuk) naj-la. Axmed O.PRO.1PL FM.SCL.3SGM OCL.1PL-with ordayey. run.PRES.PROG.3SGM ‘Axmed is running with us.’ waan (waa-aan) kuk-Øj-u b. Adigak O.PRO.2SG DECL.SCL.1SG OCL.2SG-OCL.3SG-to geeyey Axmedj. take.PST.1SG Axmed ‘I took you to Axmed.’

(SOMALI)

The availability of clitic pronouns in a language appears to be connected with the realization of multiple Topics. At a merely intuitive level this fact,

260 The sentence as utterance still the cause of much debate among scholars, may be interpreted in terms of an information retrievability requirement: In the absence of clitic pronouns and morphological Case markers, the presence of multiple Topics makes it difficult to relate each extraposed NP to its syntactic function within the sentence.28 Another strategy used to signal a Topic is to mark it with a special functional word (known as a Topic marker). This strategy, however, does not seem to be very common in the world’s languages, and is definitely less common than Focus markers. Interestingly, however, it is mainly found in languages where just one Topic per sentence is admitted, which is typically of the shifting type. Let us look at the following examples from Japanese29 (Van Valin & LaPolla 1997), Korean (Choe 1995), Gungbe (Aboh 2004) and Tagalog: (73) Sono okasi wa huntora-nai. DEM sweets TM get fat-NEG ‘(With) those sweets, you don’t get fat.’ (74) Sonamu-nun hangsang phulu-ta.30 pine-TM always green-DECL ‘As regards pines, (they) are always green.’ (75) [Olay sal-ki-nun]k kepuki-ka Øk long live-NOMIN-TM dove-NOM sal-n-ta. live-PRES-DECL ‘As regards living long, doves live longer.’

(JAPANESE)

(KOREAN)

kacang olay more long

(76) VƱ lò yà è yì wéxò mȑ.(GUNGBE) kid DET TM PRO.3SG.NOM go.PERF school in ‘(As regards) the kid, (he) went to school.’ (77) Ang libro ay ibinigay ko kay TRIG book TM PAT.give.PST SCL.1SG to ‘The book, I gave it to Pedro.’

Pedro. (TAGALOG) Pedro

As in the case of Focus markers, the presence of Topic markers does not depend on the language family or morphological type. Furthermore, the nature and origin of Topic markers does not seem to be connected to specific structural properties (as we proposed for Focus markers, derived

Topicalization 261 from original copular forms; see § 6.4.1.). No generalizations have been proposed for this issue in the literature; we will therefore leave the topic for future research.

6.5.4. Topic-prominent languages A classification of languages as Topic-prominent (as opposed to Subjectprominent) was proposed by Li & Thompson (1976: 459): In subject-prominent languages the structure of languages favours a description in which the grammatical relation subject-predicate plays a major role; in topic-prominent languages, the basic structure of sentences favours a description in which the grammatical relation topic-comment plays a major role […]. All the languages we have investigated have the topic-comment construction, and although not all languages have the subject predicate construction there appear to be ways of identifying subjects in most topic-prominent languages.

Although many of the considerations and linguistic facts presented by the two authors are still valid today, our observed data lead us to believe that notions such as Topic and Comment cannot possibly be set against the notions of subject and predicate as they belong to separate levels of analysis. The former have to do with discourse grammar and information structure, while the latter refer to grammatical functions resulting from the surface structure projection of VP-generated semantic structures. Both pairs must therefore be present in every language, as all sentences have a subject and all (non-thetic) sentences predicate something about a Topic. The differences in languages merely lie in the formal means used to signal those elements. So, a Topic may be identified through its intonation, its correlation with a clitic or a specific marker. Similarly, a subject may be signaled by verbal agreement (in inflecting or agglutinative languages) or a fixed position within the sentence (typically preverbal; see § 1.5.1.). Li & Thompson’s distinction, however, appears to be confirmed by some properties of Topic-prominent languages such as Chinese, as opposed to subject-prominent languages such as Italian. One of these properties is the interpretation of null subjects in coordinate clauses. While in Italian null subject coreference seems to depend on the overt subject in the first member of the coordination (78), in Chinese its interpretation definitely relies on the Topic (79):

262 The sentence as utterance risoj] cresce bene (78) [In quel campok], [il in DEM field DET rice grow.PRES.3SG well è prezioso. pro*k/j precious.SGM be.PRES.3SG ‘In that fieldk ricej grows well, and it*k/j is valuable.’

e and (ITALIAN)

(79) [Nà kuài dàotián], [m΃] zhăng de hƟn wàng DEM piece ricefield rice grow ASSOC very prosperous hƟn zhéngùi de. (CHINESE) very valuable ASSOC ‘(In) that fieldk ricej grows well, and it k/*j is valuable.’ As can be noticed, in the Italian sentence (78), the adjectival predicate ‘valuable’ can only refer to the NP ‘rice’, while in Chinese its antecedent is ‘field’. This difference seems to be significant from a structural and interpretive point of view, as it shows that a higher degree of control is associated with Topics in languages like Chinese. However, if some structural and morpho-syntactic conditions are altered in (78), the contrast is no longer so clear. Let us consider the following example: (80) [In quei campik], [il risoj] cresce bene, e in DEM field.PL DET rice grow.PRES.3SG well and preziosi. (ITALIAN) pro?k/*j sono precious.PLM be.PRES.3PL ‘In those fieldsk ricej grows well, and they?k/*j are valuable.’ When the NP ‘field’ is in the plural, verbal agreement is modified and a short pause is inserted before the coordinate clause, so the null subject assumes the topicalized NP as its antecedent. If in Italian the coreference of pro depended necessarily on an overt subject, the sentence in (80) would be ungrammatical (as no interpretation would be possible). Yet, the possibility for (80) to be interpreted (albeit marginally) as stating that ‘the fields are valuable’ means that this type of coreference is possible in languages like Italian too. A possible objection is that the interpretation of (80) is forced by agreement. We believe, however, that the marginality effect is simply due to the fact that the NP campi is embedded in a PP, which makes it a bad antecedent for the pro in subject position. On the other hand, in Chinese the Topic is an NP, and its coreference with the subject of the coordinate clause

Topicalization 263 is immediate. Notice that the intended reading in (80) can be made much more natural just by adding a connecting word such as quindi, ‘therefore’: risoj] cresce bene e quindi (81) [In quei campik], [il in DEM fields DET rice grow.PRES.3SG well and so prok/*j sono preziosi. be.PRES.3PL precious.PLM ‘In those fieldsk ricej grows well, and so theyk/*j are valuable.’ It can be concluded that null subject coreference works in a rather similar way in Italian and Chinese. How can the apparent contrast in (78)(79) and, in general, the way in which a null subject selects its antecedent be accounted for? Following Frascarelli’s (2007) interface analysis, we believe that null subject interpretation always depends on its coreference with a Topic and, specifically, with the Aboutness-shift Topic. The difference between languages like Italian and Chinese is that in the latter the Topic must always appear explicitly sentence-initially, whereas in Italian it can be ‘silent’, provided it has been introduced earlier in the discourse. In other words, the interpretation in (78) depends on the fact that the structure of the sentence, out of context, leads to the assumption that ‘rice’ is the Aboutness-shift Topic, since in an SVO language like Italian the Topic is usually coreferent with the subject in an unmarked sentence (see Table 2 in § 6.3.2.) while the other element is a PP. If, however, (78) is found in a context where ‘the field’ is clearly the Aboutness-shift Topic, then its role as an antecedent becomes immediately evident with no need for a plural verb agreement: (82) [Quel campok] non si tocca, perché [in quel DEM field NEG REFL touch.PRES.3SG because in DEM campok], [il risoj] cresce bene e dunque field DET rice grow.PRES.3SG well and therefore prok/*j è prezioso. be.PRES.3SG valuable.PLM ‘That fieldk is off-limits, because ricej grows well in that field, and so itk/*j is valuable.’ It can be easily shown that it is the Aboutness-shift Topic, rather than the grammatical subject, which is the antecedent for null subject interpretation. In a sentence like (83), both Gianni and Leo might be antecedents of pro; this means that both NPs are Topics and null subject coreference

264 The sentence as utterance depends on which of the two is the current Aboutness-shift Topic.31 Since Gianni is located in a higher position in the C-domain, it is the best candidate for a shifting function and, indeed, if we assume a rising intonation for it, then it qualifies as the only possible antecedent for the subject pro (though connected with a DO clitic in the first conjoined clause): (83) [Giannik], [Leoj] lok ha salutato Gianni Leo OCL.3SGM have.PRES.3SG greet.PRT con affetto ed prok/*j è felice per questo. with affection and be.PRES.3SG happy for DEM ‘Giannik, Leoj greeted him warmly, and hek/*j is happy about that.’ Without further discussion on this point, we can conclude that the difference between Topic-prominent and subject-prominent languages does not refer to two distinct types of languages, but, as in the case of Focusprominent languages, to languages in which the same phenomena happen more or less overtly. As a final observation in this section, we will briefly examine an undoubtedly important distinction, namely the distinction between argument-initial (SVO and SOV) and predicate-initial (VSO and VOS) languages. As noted in § 6.3.2., what seems to differentiate languages is the starting point of the utterance. In other words, when projecting a sentence in surface form, some languages (e.g. English, German, Italian) usually begin with one of the participants, while others (such as Tagalog and Malagasy) place the predicate first. We observed that a ‘given’ predicate is not (and cannot be) a Topic, but a foregrounded constituent (on this concept, see Chafe 1976, Givón 1976, Talmy 1985). The world’s languages could thus be classified into argument-prominent, with an unmarked TopicComment structure, and predicate-prominent, whose unmarked constructions and information structure present special phenomena (for discussion and references, see Carnie & Guilfoyle [eds.] 2000, Pearson 2000, Carnie, Harley & Dooley [eds.] 2005, Frascarelli in press).

6.6. Contrast Contrast is yet another phenomenon to be included in discourse grammar. It was long considered a special instance of Focus, with a different function from information Focus (see § 6.4.). More recently, however, it has become

Contrast 265 clear that Contrast is used to convey important distinctions with a specific role in discourse and is not necessarily related to new information. Let us start by looking at the following conversation in Italian, where contrastive Focus is presented: (84) A: Franco ha riscosso molta simpatia. Franco have.PRES.3SG gain.PRT much sympathy ‘Franco gained much sympathy.’ B: No, è MARIA che ha riscosso no be.PRES.3SG Maria that have.PRES.3SG gain.PRT molta simpatia. much sympathy ‘No, it’s MARIA that gained much sympathy.’ As noted in § 6.4., the constituent Maria in (84B) is not only focused, but also opposed to the parallel constituent in (84A). In particular, sentence (84B) is a cleft sentence whose marked structure (of the type: ‘It is x that y’) is used to identify x as the contrastive Focus in the sentence (in languages like Italian and English). The same type of information can be conveyed by simply using a special intonational contour: (85) No, MARIA ha riscosso no Maria have.PRES.3SG gain.PRT ‘No, MARIA gained much sympathy.’

molta simpatia. much sympathy

In particular, the contrastive Focus is marked by a high pitch on the tonic vowel, which also has a special duration and intensity. Consider as an example the following sentence (from Frascarelli 2004b): (86) Non è che quello che fai TE NEG be.PRES.3SG that DEM that do.PRES.2SG S.PRO.2SG lo devo fare IO e viceversa. OCL.3SGM must.PRES.1SG do.INF S.PRO.1SG and viceversa ‘It’s not like I have to do what you do and the other way around.’

266 The sentence as utterance

Figure 9. Intonational contour of two Contrastive Foci

It can be noted that the two contrastive Foci (te and io) are part of a single intonational phrase, entirely included within the scope of the negative operator non, which is placed sentence-initially. This is shown by the downgrading contour of the whole sentence, which reaches its lowest point on viceversa. However, an intonational peak can be noted on each Focus, along with vowel lengthening (although both are monosyllabic pronouns). But Contrast, as said earlier, may also be associated with a Topic. In this case, too, a high pitch on the tonic vowel is observed (see Figures 8a-b in section 6.5.1.). The interpretation for this constituent, however, is completely different from a contrastive Focus. Let us take example (60), repeated here as (87): (87) A: Come mai hai fatto due lingue, cioè, inglese e francese? ‘Why did you take two languages, I mean, English and French?’ B: francese, l’ho fatto alle medie per tre anni con una professoressa con cui mi sono trovata benissimo […]- con l’inglese, mi son trovata sempre a disagio ‘French I studied at school for three years with a professor that I liked a lot […] with English I never felt at ease.’ Clearly, a contrastive Topic does not deny (part of) a preceding statement replacing it with a new piece of information. It does not have an impact on the truth value of a sentence, but creates oppositional pairs with respect to other Topics. Contrastive Topics thus refer to a salient set of

Contrast 267 referents, whose oppositional value is provided by the information that follows (the Comment). Subcategories thus need to be identified within the domain of Contrast, so that the meaning of the term is less vague. We can describe the contrastive Focus in (84) as corrective, whereas that in (86) is oppositive. In the same way, contrastive Topics can be classified as a) (purely) contrastive (87), b) replacing and c) partial (cf. Büring 1999). Let us consider two examples that illustrates the (b) and (c) types respectively (from Büring 1999): (88) A: Do you think that Fritz would buy this suit? B: Well, I certainly wouldn’t! (89) A: What did the German musicians wear? B: The Bavarian musicians wore ‘Lederhosen’. In (88B), the speaker does not answer the question asked in (88A) but states something different (albeit related) about himself in order to move the conversation away from Fritz by means of a contrast. In (89), on the other hand, the Contrastive Topic is used to ‘narrow down’ the Topic given in the previous sentence: Speaker B has nothing to say about all the musicians, or perhaps s/he is only interested in saying something specific about the Bavarian ones. The Contrast feature can thus assume various semantic-pragmatic nuances, depending on the context and, in particular, its correlation with either a Focus or Topic. That is why many authors, such as Frascarelli & Hinterhölzl (2007), have argued that the meaning associated with Contrast must be encoded in an independent functional projection (ContrP; see [61]), whose Spec can host either a Topic or a Focus. This is confirmed by a number of facts. First of all, it was noted that contrastive Focus and contrastive Topic are in complementary distribution (no sentence can contain both); this is consistent with the fact that the Spec,ContrP position can host just one constituent. Secondly, both the contrastive Focus and the Contrastive Topic occupy the position indicated in the tree diagram in (61), i.e. following the Aboutness-shift Topic and preceding a Familiar Topic. Finally, both present the same tonal event (a high pitch on the tonic syllable), which points to a systematic correlation between functional projections in the CP area, their semantic-pragmatic interpretations and their phonological realizations. This is an extremely important result, as our research aims at understanding how information structure works and is

268 The sentence as utterance interpreted. Clearly, a number of discourse grammar issues still remain open and require further research. We believe, however, that a systematic and comparative study of phenomena at different levels of analysis (the interfaces) is the best way to advance our knowledge on these subjects.

7. Illocutionary force

7.1. Types of illocutionary acts In Chapter 6 (§ 6.2.) the pragmatic functions of utterances were introduced with a brief overview on speech acts, classified into locutionary, illocutionary and perlocutionary acts. We will now give a more detailed description of the illocutionary force of utterances, in order to see how the various pragmatic functions emerge at the syntactic, morphological and phonological levels and how they are related to information structure and its expression in typologically different languages.1 First of all, it needs to be stressed that different terms will be used to refer to the pragmatic level and the interface level of analysis. Using traditional terminology for the classification of sentences with respect to the syntactic level, sentences can be classified into declarative (positive and negative), interrogative, imperative and exclamative. Each of these types corresponds, at a pragmatic level, to a different illocutionary act, i.e. the realization of a different communicative goal enacted by the speaker through a locutionary act: Assertion, request for information, request for action, exclamation. As we will see later on, there is no one-to-one relationship between these two levels; for example, it is possible for a question to be produced not to obtain information but to solicit an action (see below). Declarative sentences have traditionally been studied at various levels of analysis: Semantic (truth conditions, etc.), syntactic (in terms of structure, constituent order, etc.) and prosodic (particularly regarding questions, since intonation is the only formal distinction between a declarative sentence and a question in a number of languages). For those languages where interrogative sentences also have structural markers, many different syntactic descriptions have been proposed according to the theoretical framework. Imperative and exclamative sentences have been much less studied at all levels of analysis.2 We will attempt to show the correlations between pragmatic analysis, semantic interpretation and formal properties for the various types of sentences with respect to illocutionary acts.

270 Illocutionary force 7.1.1. Overt performatives The functions of assertion and request for action may be overtly realized in a sentence. Let us look at some examples from different languages: (1)

I announce/ declare/ confirm that John was present at the meeting.

(2)

Affermo/ dichiaro/ dico che Maria affirm.PRES.1SG/ declare.PRES.1SG/ say.PRES.1SG that Maria ha partecipato a quella to DEM have.PRES.3SG participate.PRT manifestazione. (ITALIAN) demonstration ‘I state/ declare/ say that Maria attended that demonstration.’

(3)

U’linu anna Ahmada kƗna that Ahmad.ACC be.PERF.3SGM declare.PRES.1SG maw÷njd-a-n fi al-÷alsati. (ARABIC) present-ACC-INDEF in DET-meeting.F.OBL ‘I declare that Ahmad was present at the meeting.’

(4)

Waxaan FM.SCL.1SG

caddaynayaa declare.PRES.PROG.1SG

in-aad that-SCL.2SG

tahay. be.PRES.2SG ‘I declare that you are Somali.’

Soomaali Somali (SOMALI)

(5)

I order/ beg/ solicit you to behave yourself.

(6)

a. Chiedo/ ordino che tu ask.PRES.1SG/ order.PRES.1SG that you partecipi alla riunione. (ITALIAN) to.DET meeting participate.SUB.PRES.2SG ‘I ask/order that you attend the meeting.’ b. Proibisco a Maria di partecipare alla forbid.PRES.1SG to Maria of participate.INF to.DET riunione. meeting ‘I forbid Maria to attend the meeting.’

Types of illocutionary acts 271 (7)

(8)

Ɩmuruka an order.PRES.1SG.O.PRO.2SG that sulnjkaka. behaviour.ACC.O.PRO.2SG ‘I order you to behave well.’ Waxaan

ku

FM.SCL.1SG

OCL.2SG

tucsina do.good.SUB.PRES.2SG (ARABIC)

amrayaa order.PRES.PROG.1SG

cunto. eat.PRES.DEP ‘I order you to eat.’

in-aad that-SCL.2SG (SOMALI)

Sentences in (1)-(4) are all assertions, while those in (5)-(8) are orders whose illocutionary force is realized by the verb of the main clause. These verbs are used performatively here,3 meaning that, being used in the first person (which identifies the speaker) and in the present tense, they convey the action of asserting and commanding. We could say that sentences (1)(4) all express statements, declarations, etc. and, as such, are all instances of assertions. Similarly, the verbs in (5)-(8), used performatively, are themselves orders rather than descriptions of orders. Explicit performatives are thus formal acts with specific morpho-syntactic characteristics. The illocutionary force of these utterances is expressed in a way that could be described as formally fixed or frozen (i.e. constrained from a structural point of view), given the aforementioned restrictions (first person subject and a tense which is that of the time of locution). In their surface projection, the sentence representing the propositional content necessarily assumes a syntactically subordinate form, its formal properties depending on the specific language system. These sentences will thus present different forms in typologically different languages. In English, for instance, the subordinate clause is introduced by a C° head, which varies according to illocutionary force, as illustrated below: (9)

I declare / confirm that Anna was present at the meeting.

(10) I ask / wonder whether Anna was present at the meeting. (11) I order/ solicit you to be present at the meeting. In (9) and (10), the clauses are introduced by that and whether respectively, depending on the type of performative, whereas in (11) the

272 Illocutionary force verb is non-finite and there is a prepositional complementizer (to), which determines a Control structure (see § 2.4.). The presence of a complementizer is the grammaticalization of the dependency relationship established between the performative verb and the sentence that expresses the propositional content. In some contexts this can be avoided, for example in reported speech: (12) I declare: “Anna was present at the meeting”. (13) I ask: “Was Anna present at the meeting?” (14) I order: “Be present at the meeting!”4 In languages whose syntax rules out subordinate clauses like those in (9)-(11), it is expected that this relationship is expressed by juxtaposing two (syntactically) independent clauses, as in a quotation. This prediction is borne out in languages like Somali, where no indirect speech or embedded questions are admitted (see [4], [8] and below). Thus, the pragmatic functions through which the speaker’s communicative intentions are realized can be considered macro-functions. Their verbalization shows that their semantic-lexical realization may be extremely rich with a broad range of meanings, all included in the macrofunction identified. The breadth of the semantic realization is determined by the lexicon of a language (with a common base core).5 On the other hand, the interface with the remaining levels of analysis is diversified to the extent that different languages use morpho-syntactic (word order, specific markers, agreement, etc.) and phonological means to convey that part of the meaning. Let us go back to examples (9)-(11); as we have observed, the overt performative determines a complex sentence, including a finite root clause and a subordinate clause that may be finite or non-finite.6 This variation is found both within single languages and cross-linguistically (see § 3.5.). In this context, it is worth spending a few words about the concept of finiteness, which we defined in terms of phi and TAM features, encoded in functional heads in the IP area. According to some authors, however, the categories defining finiteness, along with tense and person, should include illocutionary force and politeness (see Bisang 2007). In Kistane (or Soddo), for example, the verb of a main clause presents a person morpheme and a declarative suffix (see [15]), the latter being absent in subordinate clauses

Types of illocutionary acts 273 (16). Consider the following sentences from Bisang (2007) and Leslau (1992), respectively (MOD = modification marker): (15) Abi yä-bayy-äw father.DEF MOD-son-POSS.3.SGM ‘Father loves his son.’

y-wädd-u. (KISTANE) IMPF.3SGM-love-DECL

(16) Tä-kätäma-yy yä-mä-ho-m ä-u. to-town-toward MOD-PV-3SGM.come-COMP 1SG.PERF.know-DECL ‘I know he came to town.’ In Japanese, the polite form -mas- (-masu- for present tense and -masi-ta for past tense) is, according to Bisang (2007), the clearest sign of finiteness since it can occur almost exclusively in finite and main clauses (17b), while it is ruled out in dependent clauses such as relative clauses (see [18]): (17) a. Ringo-o tabe-ta. apple-ACC eat-PST ‘S/he/they ate an apple.’ b. Ringo-o tabe-masi-ta. apple-ACC eat-POL-PST ‘S/he/they ate an apple.’

(JAPANESE)

(18) *[Tabe-masi-ta] ringo apple eat-POL-PST ‘The apple s/he/they ate.’ Similarly, whenever converbs are used7 the politeness marker does not appear on them, but on an independent verbal head (from Hinds 1986): (19) Koobe-e it-te, tomodati-ni at-te, issyo-ni Koobe-LOC go-CONV friends-DAT meet-CONV together-DAT gohan-o tabe-masu. meal-ACC eat-POL.PRES ‘(I am) going to Kobe, meeting my friend (and we are) eating together.’ This shows that pragmatic aspects such as illocutionary force and politeness markers have a morphemic realization in these languages; their

274 Illocutionary force positions must therefore be explained in structural terms. We will also show that these markers only occur in root clauses. Before we conclude this section on overt performatives, it is interesting to note that a sentence containing a verb used performatively may occur in a different position than expected on the basis of the unmarked order of constituents. Thus, in languages such as English and Italian, the root clause can appear justapoxed on the right, as in the following examples: (20) a. I will defend you, I swear. b. I’ll kill you! I promise. (21) Ti

proteggerò, giuro. protect.FUT.1SG swear.PRES.1SG ‘I will protect you, I swear.’

(ITALIAN)

OCL.2SG

(22) Chi ha invitato Maria, ti chiedo. who have.PRES.3SG invite.PRT Maria IOCL.2SG ask.PRES.1SG ‘Who invited Maria, I ask you.’ While not very frequent, these usages are possible. When the verb used performatively follows its complement, it must be considered, in terms of information, as a ‘clarification’ of the illocutionary act, and forms a separate sentence (see below). Consider a dialogue such as the following: (23) A: I will help you. B: Is this a promise or a threat? B’s question shows that the interpretation of the (covert) performative verb may sometimes be ambiguous and not retrievable from the context. Another interesting case is given by sentences like the Somali example in (4), repeated here as (24), with an overt performative verb: (24) Waxaan caddaynayaa IN-AAD FM.SCL.1SG declare.PRES.PROG.1SG that-SCL.2SG TAHAY. be.PRES.2SG ‘I declare THAT YOU ARE SOMALI.’

SOOMAALI Somali (SOMALI)

The sentence in (24) conveys declarative illocutionary force. As we know, this pragmatic function is signaled by a specific marker in Somali

Types of illocutionary acts 275 (i.e. waa; see note 3 to Chapter 2). This marker, however, is absent in (24), where a Focus marker (waxaa, see note 18 to Chapter 6) is used to focus the subordinate clause. This shows that the clause containing the performative verb cannot include declarative markers, whereas the focusing of the subordinate clause is a sign of its being new information and, therefore, the sole propositional content of the complex sentence. Based on these observations, we can conclude that a verb used performatively only expresses a predication on the communicative intention of the sentence and necessarily selects three arguments: A speaker (ACTOR), a hearer (GOAL) and the clause forming the propositional content of the illocutionary act (PATIENT). On the other hand, it cannot include modality markers, nor can it admit other (phi or TAM) functional features apart from those described above; finally, it is not sensitive to the distribution of information. It is thus a frozen structure (as was said in section 7.1.1.), hence the necessity to analyze sentences such as those in (20)-(22) as two independent (i.e. not forming a subordination relation) sentences.

7.1.2. Covert performatives We said previously that every utterance is actually a speech act which takes the form of a complex structure in its overt realization. We can therefore assume that every independent sentence is embedded in a covert performative structure. This entails, at an underlying level, that each utterance comprises a complex structure whose (covert) root clause bears illocutionary force. This clause determines the realization of different markers and formal properties within the overt clause (which is its complement), making its interpretation possible in terms of communicative goal. There are two points to be discussed with respect to this assumption, namely: a) How is the illocutionary force of the covert performative encoded in the clause that appears as an independent sentence? b) What data support the hypothesis that an underlying performative is present in every sentence? These questions will be dealt with in the sections that follow.

276 Illocutionary force 7.1.2.1. Formal markers of the covert performative The statement is the unmarked speech act in a language. It is therefore predictable that the functional head of Force is less frequently lexicalized in statements. Declarative sentences lack complementizers in many languages, as shown in the following examples ([26] taken from Van Valin & LaPolla 1997): (25) Nick will come tomorrow. (26) Hanako-ga yakuza-o Hanako-NOM gangster-ACC ‘Hanako killed the gangster.’ (27) Juan está Juan be.PRES.3SG ‘Juan is tired.’

korosi-ta. kill-PST

cansado. tired

(28) Maryan waa timid. Maryan DECL come.PST.3SGF ‘Maryan has come.’

(JAPANESE)

(SPANISH)

(SOMALI)

In all these sentences, declarativeness is mainly conveyed by intonation, whereas in Somali the marker waa is additionally needed (cf. Chapter 6, examples [7]-[10]). Other types of illocutionary force (interrogative, imperative, etc.) are expressed through different formal means in languages. Most often intonation, but also constituent order, TAM features and specific morphological markers are used. It can be concluded that languages can encode pragmatic functions (i.e. the various types of illocutionary acts) in different ways, so that it is not necessary to express the corresponding performative verb overtly.

7.1.2.2 Formal markers in the presence of a performative In the sentences examined in the previous section, the functional head of Force is not lexicalized. Now consider the following Italian sentences:

Types of illocutionary acts 277 (29) a. Che parti? that leave.PRES.2SG ‘Are you leaving?’ b. Che parta! that leave.SUB.PRES.3SG ‘May s/he leave!’ c. Che bello partire! that beautiful leave.INF ‘How nice it is to leave!’ It can be noted that che, in conjunction with different intonations and verbal moods, introduces yes-no questions (29a), imperative-optative sentences (29b)8 and exclamatory sentences (29c). These data seem to lead to the conclusion that the complementizer is associated with distinct illocutionary functions. We believe, however, that the clause introduced by che receives an indication of its illocutionary force from the covert performative that underlies every utterance. The presence of a null performative is supported by data from languages in which sentences that are superficially independent actually have subordinate-like morpho-syntactic properties (see particularly Evans 2007 on this topic). In Latin, for example, sentences like (30) are considered to be paraphrases of (31) (from Lakoff 1968): (30) Venias. come.SUB.PRES.2SG ‘Come!’/ ‘May you come!’ (31) Impero/ volo order.PRES.1SG/ want.PRES.1SG ‘I order/want you to come.’

(LATIN)

ut COMP

venias. come.SUB.PRES.2SG

The choice of the negative item clearly shows a syntactic, as well as semantic, relationship between (30) and (31). Root clauses that express a command select ne (see [32a]), whereas those that express possibility select non (see [32b]) (from Evans 2007): (32) a. Ne /*non

venias. NEG come.SUB.PRES.2SG ‘Don’t come!’

278 Illocutionary force b. Non /*ne venias. NEG come.SUB.PRES.2SG ‘May you not come!’ As it is impossible to realize negation by means of non in (32a), it follows that the clause is (covertly) selected by a performative verb of command (such as impero in [31b]); on the other hand, (32b) is selected by a root verb expressing possibility, such as potest fieri in a structure like the following: (33) Potest fieri ut non NEG can.PRES.3SG happen COMP ‘It may be that you are not coming.’

venias. come.SUB.PRES.2SG

Many other languages admit similar cases, e.g. English, where root infinitive clauses are possible (34a-b), as well as exclamative sentences with finite verbs introduced by the complementizer that (35):9 (34) a. Leo, go home? Never! b. Me, waste time for you? Never! (35) That he should be so ungrateful! It is interesting to notice that the possibility of verb deletion, usually reserved for performatives, is extended in some languages to the root verb of a complex sentence whenever that verb is a verbum dicendi. This is the case in Arabic. In this language, a sentence like (36) also appears as in (37), without the verb of saying; as a result, the complementizer is the first element of the relevant declarative sentence (from Ross 1970): (36) ’Aquulu ’inna al-walada say.PRES.1SG that DET-boy.ACC al-baita. DET-home.ACC ‘I say that the boy left home.’ (37) ’Inna al-walada DET-boy.ACC that (= [36])

qad PST

qad taraka PST leave.PERF.3SGM

taraka leave.PERF.3SGM (ARABIC)

al-baita. DET-home.ACC

Types of illocutionary acts 279 The same is observed in a number of other languages. For example, in Wari’ subordinate structures are extremely rare and dependency relationships are normally expressed by juxtaposition. It is therefore completely standard for verba dicendi to be covert (thus avoiding subordination), obtaining sentences such as (38) (from Everett & Kern 1997): (38) ’Om ca

hwap ra con (ca-on) pije’ be.quick PRO.2SG COMP.SCL.3SGM-OBJ.3SGM kid taxi? (WARI’) husband.POSS.1.SG ‘Didn’t my husband tell the kid to hurry up?’ [lit.: ‘Not ‘be quick’ that to the kid, my husband?’]

NEG IRR

As we can see, in this head-final and incorporating language, full NPs appear in extrasentential positions (as right-hand Topics) and clitics related to syntactic functions are incorporated on the C° head. This introduces the subordinate clause that is realized in the absence of the root verb to say. The omission of the verb to say in a complex structure is also found in Spanish, with an interesting semantic nuance deriving from the presence of the complementizer. Consider the following sentences (from Etxepare 2008): (39) a. Oye, el Barça ha ganado Barça have.PRES.3SG win.PRT hear.IMP.2SG DET la Champions. (SPANISH) DET Champions ‘Listen, Barcelona has won the Champions’ League.’ b. Oye, que el Barça ha ganado hear.IMP.2SG that DET Barça have.PRES.3SG win.PRT la Champions. DET Champions ‘Listen, (they say that) Barcelona has won the Champions’ League.’ As Etxepare (2008) puts it: “The apparent optionality of the complementizer masks an important 10 semantic difference between the (a) and the (b) cases. Consider [39] : as a typical declarative sentence, [39a] constitutes an assertion, whose propositional content is that a given soccer team (Barcelona) has won the

280 Illocutionary force Champions League. When compared with [39a], [39b] contributes the additional meaning that someone else (who is not the speaker) said [39a], such that the (speaker’s utterance of [39b]) constitutes a report of what has been said.” (Etxepare, 2008: 79).

The presence of a complementizer is thus relevant for evidentiality: By uttering (39b), the speaker assumes no responsibility as to what is asserted in the subordinate clause, but simply reports what was said by someone else. The deletion of verbs like decir, ‘say’, is also shown by the acceptability of sentences like (40), where the predication includes all the arguments semantically determined by the phonologically null verb of saying (from Etxepare 2008):11 (40) Si viene mi madre, tu a ella if come.PRES.3SG POSS.1.SG mother S.PRO.2SG to O.PRO.3SGF que el tabaco es tuyo. that DET tobacco be.PRES.3SG POSS.2.SGM ‘If my mum comes, you (tell) her that the tobacco is yours.’ In a number of Chadic languages too, the most common verb of saying may be omitted in the root clause. In this case, either a complementizer or an auxiliary verb associated with the main verb (or both) must be present. Let us consider Western Mupun (from Frajzyngier 1996):12 (41) a. Wu

sat n say COMP ‘He said yes.’ / ‘He agrees.’ i. b. Wu n PRO.3SGM COMP yes (= [41a]) PRO.3SGM

i. yes

(WESTERN MUPUN)

n-nas mo. sat n say COMP 1SG-hit PRO.3PL ‘He said I hit them.’ b. Wu n n-nas mo. PRO.3SGM COMP 1SG-hit PRO.3PL (= [42a])

(42) a. Wu

PRO.3SGM

A great amount of data confirms the hypothesis in § 1.2; we therefore expect that in a language like Somali, where subordinate clauses may only

Illocutionary force and performative structure 281 be DPs (see § 4.4.2.), the deletion of the performative yields nominalized clauses. This prediction is confirmed by data: (43) a. Dhib badanaa wiil-ku! (SOMALI) disturbance much boy-DET.NOM ‘The boy is so annoying!’ b. Muxuu (ma-wax-baa-uu) gabyaa yahay?! Q-thing-FM-3SGM poet be.PRES.3SGM ‘What a poet he is! / ‘He is such a poet!’ [lit.: ‘(I say) how much he is a poet.’] c. Muxuu (ma-wax-baa-uu) cunaa yahay wiil-ku?! Q-thing-FM-SCL.3SGM eater be.PRES.3SGM boy-DET.NOM ‘The boy is such a glutton!’ It is worth noting that these phenomena are manifested in typologically distinct languages belonging to different language families. A more extensive analysis of these phenomena might reveal interesting structural correlations with other morpho-syntactic properties of the languages in question. This topic is left for future research.

7.2. Illocutionary force and performative structure The breadth and variety of meanings that each speaker can express in the various speech acts, based on his/her knowledge and the vocabulary of his/her language, must be conveyed through the formal structure of the language itself (at various levels of organization). This structure imposes a rather limited number of options. We saw in previous chapters that in our formal framework different parts of the syntactic structure relate and interface with the other levels of analysis needed to describe (and understand) languages. In particular, the VP area hosts the interface with predication and its arguments (see Chapter 2); in the IP area, the predication is completed with phi and TAM features (see Chapter 3); finally, the CP area comprises a number of functional projections which interface with discourse grammar (see Chapter 6). According to the sentential structure proposed in Chapter 6 (§ 6.5.1.), the highest node in the representation is ForceP, whose head is assumed to host the complementizer (following Rizzi 1997). However, as a consequence of what has been said in this chapter, we propose that the structure must include an area dedicated to illocutionary force (i.e. to the expression of the

282 Illocutionary force speaker’s communicative intention), and that this area is not included in the CP but dominates the whole sentential structure. The realization of ForceP is thus the expression of a piece of information that is located even higher in the syntactic structure and encoded in a node which we will call IFP (Illocutionary Force Phrase).13 This phrase qualifies as a clausal structure which is reduced (or at least constrained) in terms of functional areas (IP and CP). Consider the following diagram: (44)

IFP IF’

[+assertion] IFVP [+question] IFV’ [+order] speaker [+exclamation] IFV’

VPERFOM

hearer ForceP Force’ ……… FinP Fin’ IP

As we can see, the relevant performative structure selects ForceP and is made up of a single functional projection (IFP), whose head hosts the illocutionary act, and an Illocutionary Force Verb Phrase (IFVP), where the performative verb and its arguments are merged.14 This operation of lexical insertion may happen overtly or covertly. In the former case, the performative verb raises to IF°, lexicalizing the relevant features and appearing in its fixed form, as described above. As such, it cannot be modified (see § 7.4.1.) and its arguments cannot be defined as either new or given. In the (more frequent) case that the performative verb is phonologically null, movement to IF° is also covert and the communicative intention is expressed through formal means (both intonational and morpho-syntactic) in the subordinate clause. Note that whenever the portion of structure expressing an illocutionary act (IFP) is overt, what normally appears as an independent sentence (as in

Illocutionary force and performative structure 283 [45a]) becomes subordinated to a root clause containing the performative verb (45b-c): (45) a. Everybody get out! b. I order that everybody gets out. c. I order everybody to get out. All three sentences are used to express a command. In (45a), the performative verb is covert and the illocutionary value is indicated by the imperative mood and the intonation of the overt subordinate clause. In both (45b) and (45c), on the other hand, the verb order is used performatively, and the clause that expresses an order is a that-clause in (b) and an infinitive clause in (c). The difference between these three sentences is basically syntactic in nature, though the varying constructions and lexical choices inevitably determine semantic nuances (which will not be discussed here). Let us now consider the following sentences: (46) a. I ordered everybody to get out. b. Rick says that the police were wrong. c. Mary asked if everybody was leaving. The verbs in (46a-c) are not used performatively. In (46a) a past tense is used; in (46b) the statement is made by Rick rather than the speaker; in (46c) both of these conditions apply. As a result, there are three complex declarative sentences, which include a root clause (with performative-like verbs used non-performatively) and a subordinate clause - introduced by a complementizer - itself part of the assertion. We will come back to this type of complex sentence later on, in order to discuss their illocutionary force and the distribution of information. In the previous chapter we examined phenomena and data regarding information structure in declarative sentences; we will do the same for sentences with an illocutionary force other than that of an assertion.

7.2.1. The distribution of information in different types of sentences In declarative sentences resulting from an illocutionary act of assertion, given and new information are distributed creating pairs described as Topic-Comment and Focus-Presupposition. In the light of what was argued in Chapter 6, it is clear that the illocutionary force of a sentence acts upon

284 Illocutionary force the new (informative) portion of it, i.e. Comment or Focus, whereas Topic and Presupposition are given and, as such, are not part of any assertion. Let us now see how information is distributed in sentences with different types of illocutionary force (interrogative, imperative and exclamative).

7.2.1.1. Requests for information Traditionally, interrogative sentences used by the speaker to obtain a piece of information from his interlocutor have been classified into two subtypes: Yes-no and wh-questions (see note 3 to Chapter 6). Examples are presented in (47) and (48a-c) respectively: (47) Are you happy to leave? (48) a. Who is leaving tomorrow? b. What are you going to eat for lunch? c. When are you leaving? The question in (47), pronounced with an unmarked interrogative intonation, is about the whole propositional content and is related to its truth conditions15; the expected answer is therefore yes or no. In (48a-c), on the other hand, a single constituent is asked about, i.e. the wh-element occurring in preverbal position in English and in many other languages. Therefore, the presupposition in (48a) is that ‘there is an x leaving tomorrow’ and the speaker is asking the interlocutor to specify a value for x. The same can be argued for the two other sentences, where the element x is a direct object (48b) and a time adjunct (48c) respectively. As we saw for declaratives, yes-no questions too may be asked about a portion of the sentence, the rest being presupposed: (49) a. Is Jack coming WEDNESDAY MORNING? b. Is Jack COMING Wednesday morning? c. Is JACK coming Wednesday morning? In the sentences in (49), the questioned constituent is the one on which the main pitch falls (signaled here in small capitals). Note that a cleft sentence may be used as an alternative way of formulating these questions:16

Illocutionary force and performative structure 285 (50) a. Is it WEDNESDAY MORNING that Jack is coming? b. Is it JACK that is coming Wednesday morning? Regardless of their surface form, the questions in (49) and (50) are of the yes-no type; thus, based on what we said about informational distribution, the questioned element is the focused one. We can conclude that the Focus is the portion of the sentence the illocutionary force has scope on, while the rest of it is presupposed. According to our universalist hypothesis, these observations lead us to assume that interrogative force can have scope on just a portion of the sentence in any language. It is therefore interesting to see what happens at the interfaces in typologically different languages, i.e. how the syntactic, morphological and phonological structures concur to convey this piece of information. Consider the following English sentences: (51) a. b. c. d.

Did Fred sell the house to Mary? Did Fred sell THE HOUSE to Mary? (or just the land?) Did Fred sell the house TO MARY? (or to Jane?) Did FRED sell the house to Mary? (or John?)

In (51a), the request for information has scope on the whole sentence and has the unmarked yes-no interrogative intonation; in (51b), the intonational peak is on the house, making it the questioned constituent; in (51c), the questioned part is to Mary; finally, in (51d) it is Fred. Thus, in English, both syntactic and intonational means are used: Syntax requires a question to present an [auxiliary-subject-verb] structure; intonation signals the possibility for a single constituent to be questioned (the rest of the sentence being presupposed). As we saw in Chapter 6 (§ 6.2.), in Somali distinct morphological markers are used in declarative and interrogative clauses. In a sentence such as (52), the interrogative marker ma occurs in preverbal position and this construction is used to question the whole sentence (i.e. the predicate and its clitic arguments): (52) Cali jaamacad-da buugag ma geeyey? take.PST.3SGM Cali university-DET book.PL Q ‘Did Cali take the books to the university?’

(SOMALI)

When the interrogative force has scope on just one constituent, we have sentences like the following:

286 Illocutionary force CALI baa jaamacad-da buugag geeyèy? Cali FM university-DET book.PL take.PST.RED ‘Did CALI take the books to the university?’ JAAMACAD-DA buu buugag b. Cali ma Cali Q university-DET FM.3SGM book.PL geeyey? take.PST.3SGM ‘Did Cali take the books TO THE UNIVERSITY?’ c. Cali jaamacad-da ma BUUGAG buu Cali university-DET Q book.PL FM.3SGM geeyey? take.PST.3SGM ‘Did Cali take THE BOOKS to the university?’

(53) a. Ma Q

It can be observed that in (53a-c) ma always precedes the questioned NP (Cali in [53a], jaamacadda in [53b] and buugag in [53c]) and is additionally followed by baa, the marker used in this language to indicate the focused NP. This clearly shows that the questioned constituent is also necessarily the sentence Focus, i.e. the element on which the illocutionary force has scope. Furthermore, since the interrogative marker ma precedes the verb in (52) and the Focus in (53), these constructions further support the analysis proposed in § 6.4.2. We can therefore conclude that the interrogative marker always precedes the sentential predicate (i.e. the propositional content of illocutionary force). Sentences like the following are thus all ungrammatical: (54) a. *CALI baa jamacadda buugag ma geeyey. b. *Cali JAMACADDA buu buugag ma geeyey. c. *Cali jamacadda BUUGAG buu ma geeyey. This shows that the FM and Q must necessarily operate on one and the same constituent as it is impossible to question one constituent and focus another (see specifically Frascarelli & Puglielli 2007 on this topic). In the other type of interrogative sentences, the wh-element is necessarily the one the speaker asks about. Let us look at a few examples from Italian:

Illocutionary force and performative structure 287 (55) a. Chi è arrivato? (ITALIAN) who be.PRES.3SG arrive.PRT.SGM ‘Who has come?’ (È arrivato) LUIGI. be.PRES.3SG arrive.PRT.SGM Luigi ‘Luigi (has come).’ b. A chi ha regalato il libro Maria? to who have.PRES.3SG give.PRT DET book Maria ‘Who did Maria give the book to?’ (L’ ha regalato) A GIOVANNI. OCL.3SGM have.PRES.3SG give.PRT.SGM to Giovanni ‘(She gave it) to Giovanni.’ c. Quando incontri Leo? when meet.PRES.2SG Leo ‘When are you seeing Leo?’ DOMANI. (Lo incontro) OCL.3SGM meet.PRES.1SG tomorrow ‘(I am seeing him) tomorrow.’ As is clear from the question/answer pairs in (55a-c), chi, a chi and quando are the constituents on which illocutionary force has scope (the objects of the questions), whereas the rest is presupposed. An informal interpretation of (55a) could be the following: ‘It is known that someone has come, and the speaker asks for this individual to be specified’. The new information in the answer is Luigi, the rest of the sentence being presupposed (and usually omitted). Note that in Italian, like in English, these sentences have a special constituent order as the wh-constituents occur sentence-initially regardless of their argument role and syntactic function (subject, object, adverbial, etc.). This means that the questioned constituent, irrespective of its deep position, must necessarily move to the CP area to receive the Focus feature and be in the exclusive scope of the illocutionary force. In this case, too, we expect that wh-questions behave in a similar way from an illocutionary-semantic point of view in all languages. As usual, differences will be found in the interface mechanisms at the syntactic, morphological and phonological levels. Let us observe what happens in an incorporating language such as Somali:

288 Illocutionary force (56) a. Cali muxuu (ma-wax-baa-uu) cunay? (SOMALI) Cali Q-thing-FM-SCL.3SGM eat.PST.3SGM ‘What did Cali eat?’ b. Goormuu (goor-ma-baa-uu) yimid? come.PST.3SGM time-Q-FM-SCL.3SGM ‘When did he come?’ c. Af soomaali-ga maxaad (ma-wax-baa-aad) u language Somali-DET Q-thing-FM-SCL.2SG for baraynaysaa? study.PRES.PROG.2SG ‘Why do you study Somali?’ The three sentences all include a generic noun (‘thing’, ‘time’) which is focused (as it precedes the FM) and accompanied by the interrogative marker ma. This shows that it is the sole element included in the scope of the interrogative force. In Somali, a wh-question is thus overtly realized through a complex word formed by the incorporation of a lexical head with the Focus and the interrogative marker.17 This element signals the illocutionary force of the deleted performative. To conclude our description of interrogative sentences in Somali, we will consider the two following sentences: (57) a. Ma

baa yimid? FM come.PST.3SGM(RED) ‘Has A MAN come?’ b. NINmaa (nin-ma-baa) yimid? man-Q-FM come.PST.3SGM(RED) ‘WHICH MAN has come?’ Q

NIN

man

The former is a yes-no question and can be answered as in (58a), whereas the latter is a wh-question and an appropriate answer must be formulated as in (58b): (58) a. Haa (NIN baa yimid). yes man FM come.PST.3SGM(RED) ‘Yes (A MAN has come).’ b. CALI baa yimid. Cali FM come.PST.3SGM(RED) ‘CALI has come.’

Illocutionary force and performative structure 289 The two different interpretations are associated with the two distinct surface orders of the functional elements combining with the NP (i.e. [ma NP baa] in [57a] vs. [NP ma baa] in [57b]). The realization of interrogative sentences in an isolating language such as Chinese shows interesting similarities with the inflecting and incorporating languages examined so far. In Chinese, like in Somali, yes-no questions are normally signaled by an interrogative marker (ma also in this case), placed sentence-finally (from Li & Thompson 1981): (59) a. NƱ

néng xiƟ Zhǀngguó zì ma? can write Chinese character Q ‘Can you write Chinese?’ b. TƗ zài nàr sànbù ma? PRO.3SG at there stroll Q ‘Is s/he taking a stroll there?’

(CHINESE)

PRO.2SG

In this case, the constituent order is the same as in a declarative sentence, the only difference being the presence of the final marker ma. In Chinese, however, yes-no questions may also be formulated using a construction known as disjunctive (or A-not-A question). Here, no interrogative marker is used, and the question entails choosing one of two alternatives: (60) a. NƱ

qù bu qù? go NEG go ‘Are you going?’ [lit.: ‘Are you going or not going?’] b. NƱ xƱhuƗn tƗ-de chènshƗn bu xƱhuƗn PRO.2SG like PRO.3SG-ASSOC shirt NEG like tƗ-de chènshƗn? PRO.3SG-ASSOC shirt ‘Do you like (or don’t you like) his/her shirt? PRO.2SG

It can be noted that this type of question requires a positive or negative value to be defined for the same predicative content and, therefore, the presence of a negative operator having scope over the predication. The illocutionary force of the disjunction (and consequently of the question) again applies to the predicate.18 In Chinese, wh-constituents need to appear within the sentence in the very position determined by the argument selection of the verb (that is to say, the position where the Focus appears in the corresponding answer). In

290 Illocutionary force (61), the wh-constituent takes initial position because it is correlated with the subject role; in (62), conversely, it requires a post-verbal position, being interpreted as a direct object: ZhƗngsƗn chƯ-fàn? (61) SHÉI qƱng who invite Zhangsan eat-food ‘Who invited Zhangsan for lunch?’ (62) NƱ

qƱng SHÉI chƯ-fàn? invite who eat-food ‘Who did you invite for lunch?’ PRO.2SG

Any other position for the wh-element is ruled out, as shown by the ungrammaticality of the following sentences: (63) a. *SHÉI nƱ qƱng chƯ-fàn? b. *NƱ qƱng chƯ-fàn SHÉI? c. *NƱ SHÉI qƱng chƯ-fàn? The position of wh-constituents related to non-argument information is also rigidly determined and always reflects the position of the corresponding phrase in a declarative sentence. So, an adverbial wh-phrase must be placed between subject and verb (i.e. in the IP area; see § 5.3.), just like the corresponding adverbial NP/PP in a declarative sentence: (64) NƱ

ZċNME xiČ xiăoshuǀ? how write novel ‘How did you write the novel?’ PRO.2SG

These data demonstrate that wh-constituents are inserted within the sentential structure (IP). In languages like English, Italian and Somali, their sentence-initial position is thus the result of movement due to interface requirements. In languages like Chinese this movement does not occur in overt syntax, but in Logical Form, in order to obtain the right interpretation.

7.2.1.2. Requests for action We have already considered sentences expressing requests for action. These are orders lacking an overt performative, expressed with a specific

Illocutionary force and performative structure 291 verbal form (imperative) or the base form of the verb in various languages. Furthermore, specific intonation is used.19 We will now see how pairs such as Topic-Comment and Focus-Presupposition can be used in a sentence whose illocutionary force is a request for action. In sentences such as (65ab), the propositional content on which the illocutionary force has scope is the whole sentence: (65) a. Stirami la camicia. iron.IMP.2SG-IOCL.1SG DET shirt ‘Iron my shirt.’ [lit.: ‘Iron the shirt for me.’] b. Porta il regalo a Tullia. to Tullia take.IMP.2SG DET present ‘Take the present to Tullia.’

(ITALIAN)

Conversely, in sentences like (66a-c) the pre-verbal constituent is a Topic. This is shown by the presence of a clitic pronoun coreferent with that NP and by intonational evidence, since the Topic forms an independent prosodic domain with respect to the Comment on which the request for action has scope: (66) a. La

camicia, stiramela! DET shirt iron.IMP.2SG-IOCL.1SG-OCL.3SGF ‘Iron the shirt for me!’ [lit.: ‘The shirt, iron it for me!’] b. Il regalo, portalo a Tullia! DET present take.IMP.2SG-OCL.3SGM to Tullia ‘Take the present to Tullia!’ [lit.: ‘The present, take it to Tullia!’] c. A Tullia, portale il regalo! to Tullia take.IMP.2SG-IOCL.3SGF DET present ‘Take the present to Tullia!’ [lit.: ‘To Tullia, take her the present!’]

An imperative sentence may thus contain Topics. But can it also include a Focus? As we know, Focus is (both syntactically and semantically) a predicative NP as well as the element on which illocutionary force has scope. Since the propositional content of a request for action may only apply to an action, the answer to our question is necessarily negative. In an imperative sentence, new information can only be realized by an eventive structure as opposed to a nominal predicate. In other words, there appears to be a semantic-syntactic incompatibility between narrow focusing and sentences expressing a request for action.20

292 Illocutionary force Nevertheless, sentences like (65a) and (65b) may also be uttered with a marked intonation, with an intonational peak on one of the arguments (in small capitals here): CAMICIA (non la gonna). (67) a. Stirami la iron.IMP.2SG-IOCL.1SG DET shirt NEG DET skirt ‘Iron my shirt (not my skirt).’ b. Porta il regalo a TULLIA (non alla madre). take.IMP.2SG DET present to Tullia NEG to.DET mother ‘Take the present to TULLIA (not her mother).’ IL REGALO a Tullia (non il libro c. Porta to Tullia NEG DET book take.IMP.2SG DET present che ti ha prestato). IOCL.2SG have.PRES.3SG lend.PRT that ‘Take THE PRESENT to Tullia (not the book she lent you).’

What is observed here, however, is a contrastive rather than information Focus (see § 6.4.), as the part in brackets indicates. These facts further confirm our hypothesis that Contrast does not coincide with Focus and is to be considered an independent functional category in the CP area. Its presence, therefore, does not determine comparable semantic-pragmatic implications for the scope of illocutionary force. The possibility for several Topics or contrastive elements to be found in imperative sentences appears to be possible cross-linguistically. To show this, let us consider data from Somali. Morphological markers are used in this language to signal sentence type and Focus, which clarifies our argument. In sentences like the following, the illocutionary force of an order is provided by the verb form and intonation, while specific morphological markers are missing, in contrast to declaratives (waa) and interrogatives (ma): (68) a. Cun! eat.IMP.2SG ‘Eat!’ b. Hal-kan igu (i-ku) place-DEM OCL.1SG-at ‘Wait for me here.’

(SOMALI)

suga. wait.IMP.2PL

However, in a negative imperative sentence, illocutionary force is signaled by a morphological marker:

Illocutionary force and subordinate clauses 293 (69) a. Wax-ba ha keenin. thing-REINF IMP bring.NEG.2SG ‘Don’t bring anything!’ b. Ha i labina. IMP OCL.1SG bother.NEG.2PL ‘Don’t bother me.’ One consideration can account for this apparent asymmetry: While in (68) the imperative interpretation is determined univocally by the verbal form, in (69) interpretive difficulties might arise as the verbal ending -in is also used for the negative verb paradigm in declarative sentences (compare [69a] with anigu waxba ma keenin, ‘I haven’t brought anything’);21 thus, ha is needed to induce an imperative interpretation. In addition, note that the marker ha is also used in imperative-optative sentences, combined with a verbal form known in the dependent mood: (70) Ha

sugo. aspettare.DEP.3SG ‘Let him wait.’

OPT

As this verbal mood cannot occur in the root clause of a complex structure, this once again proves that a (deleted) main clause containing a performative verb must be assumed in these structures.

7.3. Illocutionary force and subordinate clauses As for the relationship between illocutionary force and subordination, we will concentrate on subordinate clauses with an argument role and consider the following questions: Should it be assumed that a complex sentence involves more than one illocutionary act? Can the subordinate clause have a different pragmatic value than the root clause? Should it be assumed that the CP area of a subordinate clause hosts the same number and type of functional projections as those required, as we have seen, in a root clause? While we will not go into detail here, we would like to use crosslinguistic data to show how the generalizations in our descriptions so far may be extended to this issue, and how the relationship between universal and language-specific characteristics remains constant.

294 Illocutionary force 7.3.1. Declarative sentences As usual, we will begin by examining sentences whose pragmatic value is that of an assertion, i.e. declarative sentences. The following examples all include a root clause and a subordinate argument clause: (71) a. Jim declared that Ann missed her train. b. Rick asked if Jack told the truth. In none of these sentences is the root verb used performatively; therefore, the performative structure is not projected to the surface, remaining covert (‘I declare that’). In other words, those appearing as root clauses in (71) are actually themselves selected by a performative structure determining their illocutionary force. This is also true for (71c): Although the verb ask and the subordinate complementizer if are present in the sentence, from a pragmatic point of view this is a declarative rather than interrogative sentence. We conclude that, even when a complex sentence is uttered, there is a single communicative goal and, consequently, a single illocutionary act. Naturally, there are still differences between the three sentences in (71), depending on the factuality properties of the root verb and signaled by the different complementizers and verbal moods of the complement clause.22 If, as we assume, the complex sentences in (71a-b) contain just one performative, then they are wholly included within the scope of the assertion, provided they are pronounced with an unmarked intonation; in this case, they are the propositional content of that assertion. Of course, by using different intonational contours a different distribution of information can be obtained, as in the case of simple sentences (see § 7.2.1.1.) resulting in a different extension for the scope of the illocutionary force. The examples in (72) show some of these possibilities. In each example, the preceding context makes it clear which part of the sentence is new information (and, as such, in the scope of the illocutionary force): (72) a. What did Jim say? (Jim said that) ANN MISSED HER TRAIN. b. Who said that Ann missed her train? JIM (said that Ann missed her train). c. Who did Jim say missed her train? (Jim said that) ANN (missed her train).

Illocutionary force and subordinate clauses 295 d. What did Jim do? (Jim/He) SAID THAT ANN MISSED HER TRAIN. In (72a), the declarative force only has scope on the subordinate clause, in (72b) on Jim and in (72c) on Ann, all the rest being presupposed. Finally, in (72d) Jim is the only piece of given information, while all the rest is new and defines the propositional content of the illocutionary act. Still in the domain of declarative sentences, a language like Somali clearly shows that the illocutive force appears just once in complex sentences. The marker waa, signaling declarative force, may occur only once in a complex structure, and it must be in the root clause: (73) a. In-uu Axmed (*waa) dKRRIy waan that-SCL.3SGM Axmed DECL leave.DEP.3SGM DECL.SCL.1SG ka xumahay. (SOMALI) da be.sorry.PRES.PROG.1SG ‘I am sorry that Axmed is leaving.’ b. In-ay (*waa) timid waan that-SCL.3SGF DECL come.DEP.3SGF DECL.SCL.1SG rajeynayaa. hope.PRES.PROG.1SG ‘I hope she will come.’ The object clause introduced by in precedes the verb (Somali being an SOV language) and contains no marker: Waa signals the beginning of the verbal complex in the root clause. The fact that declarative markers are ruled out in subordinate clauses is largely confirmed cross-linguistically. We noticed earlier how the declarative suffix -u in Kistane can only mark root clause verbs (see [16]). The same is true for the declarative suffix -ta in Korean. Consider the following sentence (from Ho-Min 1999): (74) Minca-nun caki [apeci-ka ku nal ku kos-ey Minca-TM self father-NOM DEM day DEM place-to o-si-ess-n-unya-ko(*-ta)] mwul-ess-ta (KOREAN) come-HON-PST-IND-Q-QUOT(-DECL) ask-PST-DECL ‘Minca wondered if his father had gone there on that day.’ As we can see, the DECL marker -ta cannot appear within the object clause. Interestingly, the relevant marker cannot be used in both members

296 Illocutionary force of what is generally considered a coordinate structure, but only in one.23 This shows again that there is just one illocutionary act: (75) Minca-ka hakkyo-ey [chak-ul kaci-ko an(*-ta)] Minca-NOM school-to book-ACC take-and NEG(-DECL) ka-ss-ta. go-PST-DECL ‘Minca went to school and did not take his books.’ As we saw previously, it is possible in English to focus a constituent of a subordinate clause in situ, and concentrate the illocutionary force on it (see (72)). In many of the languages that use the extra situm or the cleft-like strategy (see § 6.4.1.), however, this option is not available. So, in Hungarian (Lipták, p.c.), a question like (76A) cannot be answered as in (76B). The same is true for Basque (see [77] from Ortiz de Urbina 1999), where a Focus in the embedded CP necessarily assumes contrastive value, inappropriate as an answer to a wh-questions: (76) A: ‘Who does János want us to invite for dinner?’ B: #János azt akarja, hogy JULISKÁT hívjuk János DEM.ACC want.3SG that Juliska.ACC invite.1PL meg vacsorára (HUNGARIAN) PV dinner.SUBL ‘#It is JULISKA that János wants us to invite.’ (77) A: ‘Who do you think wrote that letter?’ B: Nik uste dut JONEK idatzi PRO.1SG think.1SG AUX Jon.ERG write eskutitza letter ‘#It is JON that I think wrote that letter.’

du-ela AUX-DEM (BASQUE)

It seems therefore that information Focus cannot be realized in the CP area of subordinate clauses, as if the left periphery of embedded clauses were somehow defective and lacking a FocP functional projection. Consequently, the embedded constituent has to be raised to the root CP in order to be visible for interface interpretation (for the syntactic and theoretical implications of this analysis, see Frascarelli 2010). The same phenomenon is found in Somali. In this language it is possible to focus – and concentrate the declarative illocutionary force – on an entire

Illocutionary force and subordinate clauses 297 subordinate clause (using the FM waxaa, as illustrated in [78]), while it is impossible to focus an individual constituent within the subordinate clause using baa, as shown by the ungrammaticality of (79): (78) Wuxuu ii sheegay FM.SCL.3SGM OCL.1SG.to say.PST.3SGM XAMAR TEGEY. Mogadishu go.3SGM(RED) ‘He told me HE WENT TO MOGADISHU.’

IN-UU that-SCL.3SGM

(SOMALI)

DHAR (79) *Iyada suuq-a tegtay in-ay S.PRO.3SGF market-DET go.PST.3SGF that-SCL.3SGF clothes bay soo iibsato. FM.SCL.3SGF LOC buy.DEP (intended: ‘She went to the market to buy CLOTHES.’)

Again, if the constituent in question is to become the only element in the scope of declarative force, the sentence structure needs to be changed, and a verb ‘buy’ has to be used which lexically includes the idea of motion (expressed in the root clause in [79]): (80) DHAR bay suuqa u soo iibsanaysay. clothes FM.SCL.3SGF market.DET for LOC go.buy.PRES.PROG.3SGF ‘She went to the market to buy CLOTHES.’ Clearly, different languages express meanings in different ways depending on the restrictions imposed by their grammars; however, under no circumstances can information Focus appear in the left periphery of a subordinate clause. The possibility of concentrating the declarative force on a single embedded constituent seems to be restricted to those languages in which in-situ Focus is available (such as English).

7.3.2. Interrogative sentences Let us now consider the following sentences: (81) a. Did Jim say that Mary is coming? b. Did Jim SAY that Mary is coming? c. Did Jim say that MARY IS COMING?

298 Illocutionary force The question in (81a) has an unmarked intonation and can be considered the object of a request for information, being thus included in the scope of a single interrogative-type performative. Hence, it can be answered with ‘yes’ or ‘no’. The sentence in (81b) is also a yes-no question; this time, however, it is associated with a marked intonation (i.e. a pitch on the verb) and the question can be used in a context where it is presupposed that ‘Mary is coming’, and the speaker asks the hearer whether Jim gave that piece of information or not. Finally, in (81c) the root clause is presupposed and the subordinate clause is questioned (thus being in the scope of the illocutionary act). Possible answers to this question include: ‘Yes, he said that’ and ‘No, (he said that) she is not coming.’ If the subordinate clause is introduced by whether instead of the complementizer that, we have the following: (82) Did Jim tell you whether Mary is coming? In this case, a simple answer such as ‘yes’ has an ambiguous meaning as it can be interpreted either as ‘Yes, she is coming’ or as ‘Yes, he said she is coming.’ This means that the illocutionary force can have scope either on the whole complex structure or on the subordinate clause alone. If the root verb is ask, and the question in (83) is answered with a simple ‘yes’, this statement excludes the subordinate clause, as shown in (84b-c): (83) Did Jim ask you whether Mary is coming? (84) a. Yes (= He asked me that). b. *Yes, she is coming. c. *Yes, he asked me that: She is coming. We can conclude that a verb like ask is necessarily the propositional content of an interrogative act; conversely, the subordinate clause can only be backgrounded, although it is non-factual;24 in any case, it cannot be the questioned element. These data lead us to the following conclusions. In a yes-no question, comprising a root clause and a subordinate clause, there is always a single illocutionary act whose scope can vary according to the interaction between presupposition and questioned elements. It should again be stressed that the semantics of the root verb has a decisive role for interpretation.

Illocutionary force and subordinate clauses 299 To complete the description of interrogative sentences, we now intend to examine complex sentences including multiple wh-constituents. Let us consider the following: (85) a. Who said when John is coming? b. Who discussed what happened to John? Since there are two wh-constituents (one in the root and the other in the subordinate clause), one might surmise that two distinct questions are asked by means of a single complex sentence. This is easily disproved, as shown in the possible answers to (85a) and (85b), in (86) and (87) respectively: (86) a. Mark. b. *Tomorrow. (87) a. Laura. b. *He was run over by a motorbike. (86a) is an appropriate answer to (85a), whereas (86b) is not; similarly, (87a) is an appropriate answer to (85b), as opposed to (87b). Thus, who is the sole element in the scope of interrogative force in (85a-b), while when and what are indefinite expressions with no interrogative value. Though they look like wh-constituents, they are simply relative pronouns (see section 4.4.2.2.) and the object clauses in (85) are nominalized structures (i.e. relative clauses). Indeed, when in (85a) may be paraphrased as ‘the time at which’ and what in (85b) is equivalent to ‘the thing/the event that’. Consequently, (85a) expresses the presupposition that ‘someone stated the moment when John is coming’, and the speaker asks the hearer to specify a value for ‘someone’. The wh-constituent in the root CP is thus the sole Focus of the complex sentence and the only content of interrogation. We shall now consider cases in which the wh-constituent is part of the argument structure of the subordinate clause. Let us consider the following sentences: (88) a. Who did you say was coming? b. What did you suppose Mary had done? In (88a), who is interpreted as the subject of the subordinate clause, whereas what in (88b) is a direct object of the subordinate clause. Again, these elements, like who in (85a-b), are the only ones included in the scope

300 Illocutionary force of the interrogative performative, the rest of the sentence being presupposed. Note that these wh- elements must necessarily appear in the CP of the root clause:25 (89) a. *You said who was coming? b. *You supposed what Mary had done? Furthermore, if the wh-constituent is an adjunct (its insertion not being determined by the argument selection of the verb), ambiguous cases may arise, as in: (90) When did you say John was coming? It is evident that (90) admits two different interpretations: In the former, when refers to the predicate of the root clause (presupposition: ‘You said that John was coming’; question: When did you say that?); in the latter, when refers to the subordinate predicate (presupposition: ‘You said that John was coming’; question: When is John coming?). In other words, the adjoined constituent could have been raised from either the subordinate or the main clause. The requirement for an embedded wh-constituent to be raised to the root CP is made evident in languages with extra situm and cleft-like Focus strategies, such as Hungarian and Somali. Let us first consider the following sentences from Somali (from Xaange 1988 and Saeed 2007 respectively): (91) a. War adna, maxaad (ma-wax-baa-aad) jeclaan lahayd VOC PRO.2SG Q-thing-FM-SCL.2SG like have.PST.2SG in-aad heshid? (SOMALI) that-SCL.2SG find.DEP.2SG ‘And you, what would you like to find?’ [lit.: ‘And you, what would you like that you find?’] b. Kumaad (kuma-baa-aad) rumeysantahay Kidal-ka who-FM-SCL.2SG believe.PRES.PROG.2SG rumor-DET ih in-uu ,PjDQNti la kulmay? Imam.DET.AN with meet.PST.3SGM be.RED that-SCL.3SGM ‘Which of them do you believe [to be the source of] the rumor that he met the Imam?’

Illocutionary force and subordinate clauses 301 Maxaad in (91a), while correlated with the verb heshid, appears in the root CP area; similarly, the interrogative pronoun kuma in (91b) must rise to that position, although it modifies a subordinate-internal nominal element. While it is totally impossible to realize a wh-constituent in a subordinate CP in Somali, this seems to be admitted in Hungarian. However, the interpretation of such sentences confirms our assumptions. Consider the following questions (from Lipták p.c.): nem tudsz hogy érkezik? (HUNGARIAN) (92) a. KIT who.ACC NEG know.2SG that arrive.PRES.3SG ‘Who do you not know is coming?’ b. Nem tudod hogy ki érkezik? NEG know.2SG.OBJ that who.NOM arrive.PRES.3SG ‘Do you not know who is coming?’ The interrogative sentence in (92a), where the wh-constituent has risen to the root CP, is a genuine wh-question. On the contrary, the sentence in (92b), with ki in the subordinate CP, is not a wh-question: As shown in the translation, it is a yes-no question and its propositional content is the whole utterance. Possible answers include: ‘Yes, of course I do’ and ‘No, I don’t know who is coming.’ Ki, in (92b), is therefore not a wh-constituent, but a relative pronoun, as argued for examples (85a-b).26 These data show that a wh-constituent is interpreted as an element with interrogative force only when it is located in the left periphery of the root clause, just like Focus in declarative sentences (see § 7.3.1.).27 This corroborates the conclusion that the CP area of subordinate clauses is defective in that it lacks a FocP projection. These data also show that a complex sentence includes a single performative structure which expresses a specific illocutionary force and pragmatic function.28 As a final consideration on complex questions, we can say that regardless of their types (yes-no or wh-questions) they always include a single interrogative act, which may have scope on the whole utterance or just part of it (in this case the rest is presupposed). In particular, in the case of a wh-question, just one constituent is in the scope of the interrogative act, i.e. that corresponding to a focused NP in declarative sentences. The fact that a Focus and a wh-constituent cannot be present in the same sentence is not simply a consequence of their “competing for the same position” (contra Rizzi 1997), but results from the fact that they both constitute the propositional content of the performative in utterances with

302 Illocutionary force different illocutionary force: Declarative force for a Focus, interrogative force for a wh-constituent. Therefore, the type of illocutionary act determines their mutual exclusion. Let us now briefly examine apparent counterexamples like the following: (93) a. You said that Anna wants you to buy FRUIT (not sweets). b. Did you ask Leo to call on you TOMORROW? (rather than today?) c. Frank wants to make Ann stop going to THAT disco (not discos in general). In these sentences, intonation shows that there is an element marked by an intonational peak in the subordinate clause. This, however, is not an information Focus, but a contrastive one. As argued in Chapter 6 (§ 6.6.) and again in this Chapter (§ 7.2.1.2.), Contrast is not the same as Focus, and its presence does not entail the same semantic-pragmatic implications regarding the scope of illocutionary force. Further evidence of this is that the contrastive element in the subordinate clause may cooccur with a whconstituent in Italian: CINEMA non (94) Chi hai detto che AL who have.PRES.2SG say.PRT that to.DET cinema NEG vuole più andare (e non A TEATRO)? (ITALIAN) and NEG to theater want.PRES.3SG more go.INF ‘Who did you say did not want to go to the MOVIES any more (and not to the THEATER)?’

Here, the wh-constituent chi, part of the subordinate clause, is realized in the CP of the root clause and does not interfere with the PP al cinema, a contrastive element. This shows that the interpretation of focused/whelements and Contrast takes place in different functional projections.

7.3.3. Imperative sentences As we have seen, the sentences whose pragmatic value is that of an order assume a specific verb form in all languages if they are projected overtly with no performative structure. Based on what we have argued so far about the distribution of information in declarative and interrogative sentences, it can be easily foreseen that the propositional part of imperative sentences –

Illocutionary force and subordinate clauses 303 i.e. the constituent on which the illocutive act has scope – must be the whole predicate selected by the (covert) performative structure, which overtly appears as an independent sentence.29 In the Italian example (95), produced with unmarked intonation and the pragmatic value of an order, the whole sentence is in the scope of the covert performative: (95) Proibisci a Mario forbid.IMP.2SG to Mario ‘Forbid Mario to tell lies.’

di of

raccontare tell.INF

bugie. lie.PL

Again, as for declaratives, Topics may be inserted in this type of utterance: (96) a. A Mario, proibiscigli di raccontare to Mario forbid.IMP.2SG-IOCL3SGM of tell.INF ‘As for Mario, forbid him to tell lies.’ b. Di raccontare bugie, proibiscilo a of tell.INF lie.PL forbid.IMP.2SG-OCL3SGM to ‘Forbid MARIO to tell lies. c. A Mario, di raccontare bugie, to Mario of tell.INF lie.PL proibisciglielo. forbid.IMP.2SG-IOCL3SG-OCL3SGM ‘FORBID Mario to tell lies.’

bugie. lie.PL Mario. Mario

It is thus possible that an imperative sentence is accompanied by elements representing given information as the presence of resumptive clitics demonstrates. Similarly, Contrast may also apply in imperative sentences: (97) Proibisci A MARIO di raccontare bugie forbid.IMP.2SG to Mario of tell.INF lie.PL (non a Luisa, che non ne dice). NEG to Luisa that NEG OBL.CL say.PRES.3SG ‘Forbid MARIO to tell lies (not Luisa, as she never lies).’

304 Illocutionary force (98) Proibisci a Mario DI RACCONTARE forbid.IMP.2SG to Mario of tell.INF (non di frequentare quella ragazza). NEG of frequent.INF DEM girl ‘Forbid Mario TO TELL LIES (not to see that girl).’

BUGIE

lie.PL

In both cases, Contrast is marked by intonation, while the constituent order is unchanged with respect to the basic structure. Again, it is possible for an overt performative verb to be used; in this case the whole complex structure is in its scope:30 (99) a. I order you to tell John to come back soon. b. I order you to prohibit everyone from attending the meeting. c. I forbid anyone to tell John to leave. In conclusion, the idea that there is a (covert/overt) performative structure accounts for similarities and differences between illocutionary acts comprising simple and complex sentences in typologically different languages. Specifically, consider the following labeled structures: (100) a. Simple sentence [IFP VPERFORM [IFVP [ForceP1 (COMPL) [ShiftP [ContrP [FocP [FamP [FinP [IP1…]]]]]]]]] b. Complex sentence [IFP VPERFORM [IFVP [ForceP1 (COMPL) [ShiftP [ContrP [FocP [FamP [FinP [IP1… [ForceP2 COMPL [TopP [ContrP [FinP [IP2…]]]]]]]]]]]]]] As is shown, we assume the presence of a sentential structure above the ForceP node which hosts the performative and expresses the illocutionary force of a sentence. The performative root clause is usually covert, as is the complementizer in ForceP1 (shown in brackets); on the other hand, the complementizer introducing ForceP2 must be present, and its form depends on the illocutionary force and the finite vs. non-finite form of the predicate in IP2.31 It must also be admitted that the hierarchical extension of functional projections following ForceP is not the same in all sentences. In particular, FocP seems only to be available in the left periphery immediately selected by the performative (IFP), i.e. the clause that appears as a root clause after deletion of the performative (ForceP1 in [100a-b]).32 Topic and Contrast, on the other hand, are present in subordinate clauses too. The pragmatic

More on illocutionary acts 305 functions of subordinate Topics, however, are arguably not the same as those of Topics in a root clause (for this reason, we have described the Topic phrase as simply TopP in [100b]).33

7.4. More on illocutionary acts We would like to introduce two final issues, in order to provide a basis for future cross-linguistic research. First, we will deal with the possibility of modifying the proposed performative structure; secondly, with the relationship between the morpho-syntactic properties of different sentence types and their illocutionary functions (i.e. communicative intention).

7.4.1. Performative structure modification In our analysis, the performative portion of an utterance is the highest element in the sentence and consists of a structure whose tense-moodaspect possibilities are limited as it requires a first person ACTOR (the speaker), a GOAL (the hearer) and a PATIENT (the subordinate clause); see [44] above). One may ask whether this structure can be modified by adverbial elements. In his hierarchy, Cinque (1999) proposes a Speech Act projection where adverbs such as sincerely, frankly and honestly are merged (see § 5.3.). Can it be argued that these adverbs actually modifiy the verb in its performative use? Note that these adverbials cannot appear in overt speech acts: (101) a. #I sincerely pronounce you to be man and wife. b. #I honestly swear to tell the truth. c. #I frankly order everyone to get out. This fact, however, seems to be denied by the acceptability of sentences like the following: (102) a. I forcefully pronounce my innocence. b. I honestly swear I will make an effort. We believe, however, that the verba dicendi in (102a-b) are not used performatively, and are therefore not the surface realization of a speech act.

306 Illocutionary force This is proven by the possibility for the sentences in question to be continued as follows: (103) a. I forcefully pronounce my innocence, but I know I am guilty. b. I honestly swear I will make an effort, but I cannot guarantee it. On the other hand, an overt speech act cannot be denied: (104) a. #I swear to tell the truth, but I cannot guarantee it. b. #I pronounce you to be man and wife, but I am not sure. Let us now consider the interpretation of an adverbial clause, which might lead us to assume that a performative verb can actually be modified. (105) Mario went out because the windows are closed. This is clearly an ambiguous sentence. Under its first reading, the fact that the windows are closed is the reason for Mario’s going out. In other words, this interpretation may be paraphrased as follows: (106) INTERPRETATION (a) of (105) (I declare that) Mario went out (and the reason for his going out is) that the windows are closed. In this case, the adverbial clause in (105) does not modify the performative, but the clause included in the scope of the performative (with a causal function). Under the second interpretation, the fact that the windows are closed is the reason for my assertion, and therefore the adverbial clause seems to modify the (covert) performative. In fact, the adverbial clause expresses the reason for my thinking that Mario went out. This means that the adverbial clause actually modifies a (covert) verb of opinion which is the object of the assertion. This interpretation can be expressed as follows: (107) INTERPRETATION (b) of (105) (I declare that I think that) Mario went out, (and I think so) because the windows are closed. We are therefore assuming that, along with a covert performative, the sentence includes a covert attitude verb whose complement is the sentence Mario went out. This analysis is plausible and very similar to Etxepare’s

More on illocutionary acts 307 (2008) proposal regarding some Spanish sentences, which appear independent but actually need to be introduced by the complementizer que: The author argues that a covert verb to say must be assumed (see examples [39]-[40]). In Spanish, too, adverbial clauses can be used to modify the covert verb to say (see note 11 to Chapter 7). The difference between English and Spanish is that the English complementizer that cannot be realized in a context without presuppositions (i.e. as the first utterance in an exchange). Imagining a question-answer context, however, we can admit a sentence like (108B): (108) A: What do you think? B: That John went out, because the windows are closed. We conclude that performative verbs cannot be modified and therefore the sentential structure which forms them is rigidly limited (which is consistent with their being fixed structures; see §§ 7.1.1. and 7.2.).

7.4.2. Indirect illocutionary acts What has been said in this chapter about illocutionary acts may be summarized as follows. On producing an utterance, the speaker may have two types of communicative goal, namely: (109) ILLOCUTIONARY GOAL (I). Speaking to exchange information, i.e. speak to know something, which means that the speaker: (α) provides information to the interlocutor; (β) asks the interlocutor for information. ILLOCUTIONARY GOAL (II). Speaking in order to get an action performed, i.e. speak to cause an action, which means that the speaker: (γ) offers his/her action to the interlocutor; (δ) requires an action from the interlocutor. For these two types of speech acts to reach their goals, in the case of goal (I) the speaker needs to receive a verbal response, in the case of goal (II) an action needs to be performed, while a verbal response may or may not be present. This can be represented in the following table:

308 Illocutionary force Table 4. Speech Acts and illocutionary goals ILLOCUTIONARY FORCE speak to know something speak to cause an action speaker hearer speaker hearer α

β

giving information

obtaining information

γ

obtaining information

giving information

δ

offering answer speaker’s action requesting

(answer)+ hearer’s action

The elements in this table must then be associated with the information on the related linguistic form, i.e. the types of sentences which correspond syntactically, morphologically and phonologically to each of these options. In the speak to know something section, the uses in Į correspond to the structure of declarative sentences, whereas those in ȕ are expressed through interrogative sentences (the difference between yes-no and wh-questions being that the former are requests for information on a specific propositional content and the latter requests for specification of a portion of the sentence). In the speak to cause an action section, the uses in Ȗ correspond to questions, whereas those in į typically assume the form of imperative sentences, although they may appear as interrogatives too. In both cases, the questions are not used to obtain information. As shown in Table 4, the communicative goal of the questions in uses in Ȗ consists in the speaker’s offer to perform some action, such as in the following: (110) a. Shall I go buy some food? b. Shall I put some salt in the water? c. Would you like some more light? Clearly, these sentences are not uttered to obtain information, but a ‘yes’ or ‘no’ in order to perform the consequent action. The speech act reaches its goal if, once an answer is received, the speaker acts accordingly. As regards the cases in į, as we said, an imperative or interrogative sentence may be used:

More on illocutionary acts 309 (111) a. Open the window. b. Will you open the window? In both sentences in (111) the speaker directs a request for action to the hearer; in (111a) what is expected is not a verbal answer, but an action (i.e. the hearer opening the window), while in (111b) there may be an answer, but it would not be enough. For the speech act to achieve its goal, an action by the interlocutor must take place.34 Then what is the difference between (111a) and (111b)? They share the same communicative goal and consequently the same illocutionary value. What changes is the syntacticintonational form and the semantic value of the constraint that the speaker wants to (or can, depending on context and social conventions) place on the hearer. In fact, for the uses in į other forms can be employed which differ in modals and verb moods. They occupy different positions that correlate to different degrees on a formality and politeness scale, depending on the type of constraint. Here are some examples: (112) a. b. c. d. e. f.

Open the window! Please open the window! Will you open the window? Can you open the window? Would you open the window? Could you open the window, please?

CONSTRAINT

FORMALITY

+





+

Note that the constraint and formality values are inversely related. The linguistic form expressing a strong and clear constraint corresponds to a low level of formality, and vice versa. A more specific description of these concepts, taking all levels of interaction into account, will not be attempted here. We would like to underline once more that, if it is assumed that communicative goals are universally valid, a number of constants in surface realization and crosslinguistic differences will be found, as each language will use only some of the options available from a limited set admitted by the language system. Therefore, in the various languages we have analyzed, declarative sentences always express a type-Į illocutionary act which corresponds to the unmarked structure from a morpho-syntactic and intonational point of view. The question is clearly the unmarked form in requests for information (type-ȕ), but it can also be used (the so-called non-institutional uses) for

310 Illocutionary force other types of illocutionary acts (Ȗ and į). This is hardly surprising, since the common factor in all these cases is a pragmatic one, i.e. the presence of a request. Therefore, languages also use questions in requests for actions, whenever allowed by the system.

Conclusions

In this final section we will summarize some of the main points of our linguistic analysis which has centered on two main issues: The interaction between the different levels of organization in languages and crosslinguistic comparison. As far as the former is concerned, we have tried to single out both universal aspects and parametric differences between typologically different languages. At the same time, through the analysis of data we have shown what means are used in languages to convey and make accessible all the information included in a sentence (the basic descriptive unit in our theoretical framework). To sum up our descriptive hypotheses and provide some final generalizations, let us return to our theoretical framework and see what our crosslinguistic analysis has told us about its components. According to Generative Grammar, the two universal components of human language, i.e. the lexicon and the computational system, are linked by the operation known as lexical insertion (see Chapter 1, § 1.2.). This operation is the basis of a derivational process, divided into four representation levels, whose schematization is repeated here:

(1)

lexical insertion

D-STRUCTURE

Syntax S-STRUCTURE PF

LF

In our interpretation, the lexicon comprises a number of units defined as semantic roots. Each unit is formed by a set of semantic features, which are assigned a specific category only upon lexical insertion. In other words, their category is determined by the syntagmatic context in which they are

312 Conclusions placed. The syntagmatic context is provided by the computational system, which imposes a recursive hierarchical structure, formulated in terms of X' theory. Upon lexical insertion, the roots are assigned a morphological form which is appropriate to their semantic-lexical contents, and the predicate and argument roles are defined. The number of arguments (three at most) is determined by the meaning of the predication; each receives a θ-role, itself predication-dependent. This semantic structure interfaces with the predicate and its arguments at the (syntactic) level of deep structure; here, the arguments are grouped together as macro-roles (or deep cases). A propositional core is thus formed, composed of the VP and various NPs (with the verb in the V° position, and the NPs in the Specifier and Complement positions). This deep structure level and the assignment of a syntactic function to each constituent determine the surface structure. This is where languages exhibit differences, in terms of both constituent order and surface markers of syntactic functions (morphemes, particles and prepositions). In other words, the rich and varied nuances found at a semantic level must be re-organized (i.e. structured) in groupings in order to enter the (formal) computational system. Each grouping shows a specific syntactic behavior. Restrictions are thus imposed by the formal system due to its very nature and the means available; violations of these restrictions may lead to incomprehensibility.1 Furthermore, pragmatic mechanisms are at work from the very moment of lexical insertion: For instance, the speaker’s choice as to which piece of information should be foregrounded and backgrounded. Examples are sentences such as The workers loaded bricks on the truck vs. The workers loaded the truck with bricks, or Jack gave Amy a kiss vs. Jack kissed Amy. Cross-linguistic variation, of course, is expected also in this case. Let us go back to the structural assignment of θ-roles within macro-roles which takes place in the VP. This is based on structural conditions obtaining between sister nodes (see Chapter 2, § 2.2.5.). The structure thus obtained reveals the hierarchy which exists between macro-roles (which we derived from data), as well as any restrictions on their combinations: (2)

MACRO-ROLE HIERARCHY ACTOR > PATIENT/THEME > GOAL/LOCATIVE/QUANTITY/MEASURE/ COMITATIVE

Conclusions 313 According to the Thematic Criterion, a syntactic position may contain just one constituent with a macro-role; it is therefore evident that only one of the alternative macro-roles in (2) may be present in a sentence. The syntax-semantics interface thus makes strong predictions possible regarding the combinatorial possibilities between macro-roles and the complementary distribution of some of them. The semantic contents interfacing with categories and the syntactic structure in terms of VP (predicate) and NPs (arguments) assume a syntactic function which is determined by the structural position of the phrase that hosts them. Syntactic functions (first-level only) also form a hierarchy: (3)

SYNTACTIC FUNCTION HIERARCHY SUBJ > DO > IO

The two hierarchies interact, hence determining a surface structure which must include a subject (be it lexicalized, null or appearing as an expletive), plus possible additional complements (according to the hierarchy in [3]). There is not a one-to-one relationship between syntactic functions and macro-roles. If there is no ACTOR (or if the ACTOR is realized as an adjunct), the subject function can be assumed by a PATIENT; in turn, if PATIENT is absent or expressed as an adjunct, the function of direct object can be assumed by the GOAL. In other words, a generalization can be made that core syntactic structure is formed by the predicate and three argument positions at most and that there is a tendency in many languages for these positions to be filled according to the above-mentioned hierarchical order. The various possibilities offered by the system to express all this, used differently in various languages, have been extensively described throughout this book. A functional area is also assumed for each of the two constituents, VP and NP, discussed so far. In the case of the NP, this is the DP and includes semantic aspects such as definiteness, referentiality, deixis, quantification, and the various categories (gender, number, etc.) that apply to the head noun. Some of these topics have been discussed in Chapter 4. As regards the VP, its functional area is contained in the IP structure. A simplified structure of the sentence first presented in Chapter 1 is repeated below:

314 Conclusions

(4) CP

Syntax-Discourse grammar interface C’



IP

Syntax-Morphology interface I’



VP

Syntax-Semantics interface V’

V° The IP area hosts both TAM and adverbial modification categories (see Chapter 5). While we will not discuss the latter, we would like to mention briefly a few aspects of the former which merit further attention. In surface structure tense-related functional projections are used to realize the temporal location of the propositional content with respect to the time of locution (whether expressed as verb inflection or as free elements in non-inflecting languages). In other words, the event that the speaker informs the hearer about is temporally located as taking place during, before or after the time at which the speaker utters the sentence. This semantic-pragmatic function interfaces with morpho-syntactic aspects of sentence structure. We will not look into the valencies of aspect and its interrelationship with tense, except to note that is a complex topic which has been little described cross-linguistically. It is worth mentioning, however, that these characteristics are closely related to the meaning of the verb, and their location within the IP is determined by their structural relationship with the relevant verbal head. The semantically complex and typologically differentiated category of modality is also hosted in this part of the structure. The IP area also contains all the information relating to agreement. In most languages, this consists of the codification of the syntactic function of subject; in some languages, the function of object is also hosted here. Again, differences in realization have been observed in languages belonging to different morphological types. Let us now consider the most significant conclusions pertaining to the part of the syntactic structure that interfaces with pragmatics and discourse grammar, namely the CP area. We introduced the concept early on, and completed the relevant structure by assuming a higher structure (IFP) in our discussion of illocutionary acts (see Chapter 7).

Conclusions 315 The CP comprises a vast number of functional projections, each dedicated to host a constituent with a specific information value. This information value bears a strong link to syntactic structure, as shown by our definitions of the pairs Topic-Comment and Focus-Presupposition (see Chapter 6). At the same time, a sentence is a meaningful unit insofar as it is used by a speaker with a specific communicative intention (see Chapter 7). This intention is, again, part and parcel of syntactic description, as it is related to the presence of a complementizer and the requirement for the performative part to assume scope over the propositional content. A description of the sentence as utterance therefore needs to include a dominant performative structure which we have assumed to be frozen. This structure has a verb used performatively as its predicate. Once again, crosslinguistic differences may arise. A varying number of lexical units can be used, with different nuances in meaning, to refer to what we identify as performative verbs expressing basic illocutionary acts in conversational exchanges (speak to know something and speak to cause an action). Lexical and formal differences aside, it must be possible to express these communicative intentions in every language. An utterance must therefore include one and only one performative structure regardless of its overt realization and the complexity of the sentence. Even when several clauses are embedded in a root clause, e.g. a complement clause and an adverbial clause (Ann informed us that Jim was sad because nobody agreed with him), there is still only one performative. On the other hand, any [VP-NP] sequence determines a syntactic-semantic structure of a sentential type, i.e. a sequence of clauses, one of which the root clause and the rest embedded as complements or adjuncts (based on the recursivity principle). From an information point of view, contextual choices are possible regarding what is to be included in the scope of the assertion (or the request for information) within a sentence: A comment, if a predication is overtly realized as a VP, or a Focus, if it is overtly realized as an NP. What is left outside the scope of the illocutionary force is the Presupposition when it includes a predicate, and the Topic when it is an NP. Foci and wh-elements have similar functions in utterances whose syntactic forms (declarative and interrogative, respectively) are related to two distinct illocutionary acts. The syntactic and semantic-pragmatic differences between functional categories such as Topics and Foci determine their different distribution within the CP area. While several types of Topics (and therefore several dedicated functional projections) must be assumed, just one Focus can be realized in a sentence. Finally, our approach has clarified a number of

316 Conclusions issues regarding the interfaces between discourse grammar, semantics, syntax and, whenever possible, phonology (prosody and intonation). This has led us to revise concepts such as Focus-prominent and Topic-prominent languages (see Chapter 6). The topics included in our analysis, based on data from typologically different languages, have been discussed using a syntax-centered approach. However, we have also tried to show how all the other levels of analysis should be part of that description. The formal nature of the system and the constraints it imposes have helped us understand how the whole system works by examining interactions between levels in detail. This type of approach can be conducive to the identification of universal and parametric aspects of languages. We have attempted to account for this by showing that differences mainly arise from the various strategies by which semanticpragmatic structures and deep syntax aspects are brought to the surface in typologically different languages. A complete description within our theoretical framework would certainly require additional elements: A detailed analysis of the theoreticalformal structure, a consideration of the mechanisms needed for interrelating deep and surface structures, and a formalization of the proposed interface interactions with the other components of grammar. These can be the subjects of future research. One possible outcome might well be a new definition of the idea of language type: The various parameters used so far in morphology (inflecting, agglutinative, etc.), constituent order (SVO, SOV, etc.) and so on, all of which pertain to the surface form, might show strong correlations with the other means used as surface signals of semantic-pragmatic and structural-functional (i.e. syntactic) aspects which are necessary for interpretation.

List of languages

Afar Albanian Arabic Avar Bambara Basque Bemba Berber Breton Bulgarian Cayuga Chaha Chichewa Chinese Chukchi Dutch Dyirbal English Ewe Finnish French

Afro-Asiatic family; Cushitic group; Eastern Cushitic subgroup Indo-European family; Albanian group Afro-Asiatic family; Semitic group; Western Semitic subgroup Northern Caucasian family (debated); Northwestern Caucasian group Niger-Congo family; Mande group Language isolate Niger-Congo family; Atlantic-Congo group; Bantu subgroup Afro-Asiatic family; Berber group Indo-European family; Celtic group Indo-European family; Balto-Slavic group; Southern Slavic subgroup Amerindian family; Iroquoian group; Northern Iroquoian subgroup Afro-Asiatic family; Semitic group; Southern Semitic subgroup Niger-Congo family; Atlantic-Congo group; Bantu subgroup Sino-Tibetan family; Chinese family Chukchi-Kamchatkan family; Chukchi group Indo-European family; Germanic group; Western Germanic subgroup Australian family; Pama-Nyungan group Indo-European family; Germanic group; Western Germanic subgroup Niger-Congo family; Atlantic-Congo group; Kwa subgroup Uralic family; Ugro-Finnic group Indo-European family; Romance group; GalloRomance subgroup

318 List of languages German Greenlandic Gungbe Gunwinjgu Guyana creole Haiti creole Hebrew Hindi Hungarian Isekiri Italian Japanese Kashmiri Kikuyu Kistane Korean Krio Lakhota Latin Macushi Malagasy Maǀri Mayali Modern Greek Mohawk Mojave Ngalakan Oneida Persian

Indo-European family; Germanic group; Western Germanic subgroup Eskimo-Aleutian family; Inuit group Niger-Congo family; Benue-Congo group; Kwa subgroup Australian family; Pama-Nyungan group Creoles; English-based creoles Creoles; French-based creoles Afro-Asiatic family; Semitic group; Western Semitic subgroup Indo-European family; Indo-Iranic group Uralic family; Ugro-Finnic group Niger-Congo family; Atlantic-Congo group; Yoruboid subgroup Indo-European family; Romance group; Eastern Italic subgroup Language isolate Indo-European family; Indo-Iranic subgroup Niger-Congo family; Atlantic-Congo group; Bantu subgroup Afro-Asiatic family; Southern Semitic group; Ethiopic subgroup Altaic family or language isolate Creoles; English-based creoles Amerindian family; Siouan group; Sioux group Indo-European family; Italic group Amerindian family; Cariban group Austronesian family; Malayo-Polynesian group; Borneo-Philippines subgroup Austronesian family; Malayo-Polynesian group; Polynesian subgroup Australian family; Pama-Nyungan group Indo-European family; Greek group Amerindian family; Iroquoian group; Northern Iroquoian subgroup Amerindian family; Hokan group; Yuman subgroup Australian family; Pama-Nyungan group Amerindian family; Iroquoian group; Northern Iroquoian subgroup Indo-European family; Indo-Iranic group

List of languages 319 Polish Quechua Rapanui Rembarunga Romanian Russian Saramaccan Somali Southern Tiwa Spanish Swahili Tagalog Tigrinya Toba Batak Turkish Tzutujil Uto-Aztecan Warì Warlpiri Welsh Western Mupun Wolof Yareba Yidiny

Indo-European family; Balto-Slavic group; Western Slavic subgroup Amerindian family; Quechua group Austronesian family; Malayo-Polynesian group; Polynesian subgroup Australian family; Pama-Nyungan group Indo-European family; Romance group; Eastern Italic subgroup Indo-European family; Balto-Slavic group; Eastern Slavic subgroup Creoles; English-based creoles Afro-Asiatic family; Cushitic group; Eastern Cushitic subgroup Amerindian family; Kiowa-Tanoan group Indo-European family; Romance group; Western Romance subgroup Niger-Congo family; Atlantic-Congo group; Bantu subgroup Austronesian family; Malayo-Polynesian group; Borneo-Philippines subgroup Afro-Asiatic family; Semitic group; Southern Semitic subgroup Austronesian family; Malayo-Polynesian group; Batak subgroup Altaic family; Turkic group Amerindian family; Mayan group; Quiché subgroup Amerindian family; Uto-Aztecan group Amerindian family; Chapacuran group Australian family; Pama-Nyungan group Indo-European family; Celtic group Afro-Asiatic family; Chadic group; Angas group Niger-Congo family; Atlantic-Congo group; Senegambian subgroup Papuan family; Trans-New Guinea group Australian family; Pama-Nyungan group

List of abbreviations

– . ABL ABS ACC ACT ADV AN ANTIPASS AOR ASSOC AUG AUX BEN CAUS CHAN CL CLASS COLL COMP COMPAR CONV COP DAT DECL DEF DEM DEP DES DET DIR DIRECT DO DUR

morpheme boundary fusive morpheme boundary ablative (Case) absolutive (Case) accusative (Case) ACTOR (macro-role) adverb anaphoric (determiner) antipassive construction aorist associative particle augmentative auxiliary beneficiary causative change of state clitic (pronoun) classifier collective complementizer comparative converb copula dative (Case) declarative force marker definite demonstrative dependent paradigm desiderative determinant marker of direct case directional direct object durative aspect

DYN ELA EP ERG EXCL EXPL F FACT FM FORCE FREQ FUT GEN GER HAB HON IMP IMPERS IMPF INCL IND INDEF INSTR INTENS IO IOCL IP IRR ITER LOC M MIDD MOD N

dynamic action elative (Case) epententic (vowel) ergative (Case) exclusive expletive feminine factual Focus Marker illocutive force frequentative future genitive (Case) gerund habitual honorific imperative impersonal imperfective inclusive indicative indefinite instrumental (Case) intensifier indirect object indirect object clitic inflectional phrase irrealis iterative locative masculine middle voice modal neuter

322 List of abbreviations NEG NOM NOMIN NONHUM NONPST NP O.PRO OBJ OBL OBLCL OCL OPT PASS PAT PERF PL PLF PLM POL POSS PREP PRES PRO PROG PRT PST

negation nominative (Case) nominalizer non human non past noun phrase object pronoun objective conjugation oblique (Case) oblique clitic object clitic optative passive PATIENT (macro-role) perfective plural plural feminine plural masculine politeness possessive preposition present tense pronoun progressive past participle past tense

PV Q QUOT RED REFL REINF REL S.PRO SC SCL SG SGF SGM STAT SUB SUBJ SUBL SUBOR TEMP TERM TH TM TOP TRIG VOC

preverbal particle question marker quotation reduced paradigm reflexive reinforcer relative subject pronoun small clause subject clitic singular singular feminine singular masculine stative subjunctive subject sublative (Case) subordinate paradigm temporal (Case) terminative (Case) THEME (macro-role) Topic Marker Topic trigger vocative (Case)

Note: a) Words in the examples are not always fully specified regarding their inflectional features in the gloss; specifications are functional to the analysis in question. b) Tagalog infixes may determine a discontinuity in lexical roots. Whenever one of the examples includes this type of infix, the first part of the root is not taken into consideration in the interlinear gloss.

Notes Introduction 1 Data collection was partly carried out within the PRIN Project Grammatical, lexical and ethno-cognitive aspects of minority languages with primary orality, which made it possible to work with informants (title of the Research unit: Discourse grammar and information structure in typologically different languages).

1.

Categories and functions

1 In the Minimalist program (2004) Chomsky’s objective is to establish not only what the principles of Universal Grammar are, but also why the properties of language are thus defined. The notion of interface therefore takes a leading role in linguistic analysis, as any syntactic operation must be explained according to output conditions, i.e. the interpretation at Logical Form and Phonetic Form. 2 According to innatism, it is believed that the human mind is genetically predisposed to language, and that therefore language is not learned during life, but activated. Exposure to data makes it possible for the child to develop a mental faculty possessed since birth (i.e. the Universal Grammar principles) and, at the same time, to fix the parameters specific to his or her own language. See also Lightfoot (1982), Cook (1988) for an extensive discussion of Universal Grammar and specific properties of language; regarding the broad issue as to the linguistic input being inadequate for grammar learning, see the literature on the “poverty of the stimulus” (Chomsky 1988, Pullum 1996, Fodor 2000, Lasnik & Uriagereka 2002, Pullum & Scholz 2002). 3 Within the Minimalist Program, the levels of representation and their mutual relationships were modified. In particular, the D-structure and S-structure levels were done away with, whereas the two interpretive levels, Phonetic Form and Logical Form, were retained. This specific development, however, is irrelevant for the purpose of this work. The previous and more familiar model will thus be used here for the sake of convenience. For a description of the Minimalist model, see Moro (1995), Marantz (1995), among others. 4 Chomsky (1981) in fact proposes an additional level before surface structure, called “Syntactic structure”. For the sake of simplicity, this subdivision will be avoided here, and the term S-Structure will be used to refer to the surface structure as the endpoint of syntactic derivation. 5 The level where a linguistic expression is realized by means of a sequence of articulated sounds (Graffi 1994: 134).

324 Notes 6 The level where the meaning of a sentence is represented as determined solely by linguistic units (Graffi 1994: 135). 7 The term refers to a rule that can be applied to its output an infinite number of times. 8 The notation XP (X-Phrase) is equivalent to the notation X'' (read as ‘X-doublebar’). 9 The presence of lexical entries with complex structures, such as phrasal words, also needs to be taken into account (see Jackendoff 1995, 2002 for a discussion of this topic). 10 These concepts will be taken up, explained and discussed in detail in Chapter 2. 11 In this book, the phrase labels usually found in the literature will be used: NP (Noun Phrase), VP (Verb Phrase), PP (Prepositional Phrase), IP (Inflectional Phrase), CP (Complementizer Phrase), etc. Other phrase labels will be introduced in the course of the present work. 12 The independence of the subject from this parameter has a specific structural reason, as will become clear in Chapter 3, where the syntactic realization of argument structure will be discussed. 13 Because a great number of examples from various languages are included in this work, some of the conventions used should be noted: a) the name of the language is mentioned in brackets after each example; b) if consecutive examples are drawn from the same language, its name is only mentioned after the first; c) the source of each example is provided in the preceding text; d) where no source is given, the example is original (i.e. drawn from interviews with native informants). For the abbreviations used in glosses, see the relevant List. 14 Similar observations are also valid in English (an SVO language), as can easily be verified in the translated examples. 15 OVS languages are extremely rare, and OSV languages are not at all attested. This point will be further discussed in § 1.4. 16 The classification of German as an OV language might appear surprising and questionable, since in fact German presents different orders in different clause types: VO in main clauses and OV in subordinate clauses. Such variation, though, merely results from a central property in the grammar of German (and other Germanic languages), known as the Verb-second (V2) requirement. Thus, in languages like German, the verb must always appear after the first constituent in the clause. This means that in unmarked main clauses, where the subject usually takes first position, the verb follows it, thus creating an SVO structure. On the other hand, in embedded clauses, where V2 phenomena are highly restricted (cf. Emonds 1970, Hooper and Thompson 1973, Heycock 2006, among others), the lexical verb occupies the final position, preceded by its object(s) (yielding the SOV order). This goes to prove that the SVO order in main clauses is not the basic order in German: if it were so, it should apply in all clausal types, contrary to fact. For a discussion of this point, see the contributions included in Svenonius (ed.) (2000).

Notes 325 17 As Chomsky notes (1986a: 3), “generative means nothing more than explicit”. The syntactic component thus generates structural descriptions, which then need to be interpreted at the Phonetic Form and Logical Form levels. 18 The list of ș-roles in (19) is not exhaustive; see Chapter 2 for a detailed description. 19 The Extended Projection Principle (EPP) was originally proposed by Chomsky (1981) to account for expletives (see § 3.2.2.) in languages with obligatory subjects (non-pro-drop languages). According to this principle, the Spec,IP position needs to be filled in every sentence by an NP functioning as a subject. Since Chomsky (1995), however, the EPP term has no longer been considered as applying to the sentential subject alone, but as an extended feature requiring the presence of a scope-assuming constituent (i.e. a constituent located in the highest Spec) in each of the areas in (18): the VP level, the clause level, and the discourse grammar level. This extended formulation of the EPP recently led Rizzi (2006) to propose what is known as the “Subject Criterion” to account for the need of a sentential subject. 20 On the opposition between free forms and bound forms see Bloomfield (1933), Lyons (1968) and Scalise (1983). 21 Each one of these areas contains more structure than is visible in (29), as will be discussed in the following chapters. See in particular Chapters 6 and 7 for the content of the CP area. 22 One example is the (logical) distinction between nouns and verbs (respectively as subjects of predications and expressions of the predicated actions or qualities) made by Plato in the Sophist. The distinction was used by Aristotle in the Poetics, where conjunctions were brought in (and possibly articles, although some sources ascribe these to the Stoics). On the other hand, the classification into eight parts of speech, later adopted by the Western tradition via Latin, is attributed to Dionysius Thrax (see Lyons 1968, Lepschy 1990, Robins 1997). 23 See Schachter (1985), Hopper & Thompson (1984, 1985), Lazard (1999), Anward (2001), Sasse (2001), Givón (2001), among others. 24 Dixon (1982) suggests that the words of a language are divided into a number of semantic types, viewed as linguistic universals – an idea that closely resembles that of prototype proposed by Rosch (1978). Dixon shows how some semantic types are linked to individual lexical classes as well as to specific syntactic and morphological properties. For example, the ideas related to motion, affecting (things and/or people) and giving, among others, tend to be classified in the verb class; on the other hand, objects, kin and other types are grouped together as nouns. 25 In the Minimalist Program, Chomsky repeatedly states that he considers lexical elements as indissoluble sets of features. In particular, Chomsky (1995, 2004) assumes that lexical elements are formed by a set of phonological (PHON), formal (FOR) and semantic (SEM) features, merged as a whole at the beginning of the derivation. These need to be distinguished from “grammatical-

326 Notes

26

27 28 29 30

functional” items, i.e. irregular inflectional elements and suppletive forms, which are merged in the course of derivation (late insertion). Lexical elements are thus syntactic atoms functioning as phrasal heads, whose internal structure cannot undergo any changes in syntax. Chomsky (2004) clearly states that there are no “substantive lexical categories”, but only naked roots, which assume categorial status after combining with a functional (F) element in an F-XP configuration. This approach is called the Lexical Integrity Hypothesis and has been the object of numerous debates in recent literature; see Harley & Noyer (2000), Lieber & Scalise (2005) on this issue; for a “dressed root” analysis, see Ramchand (2008). Our choice of assuming a syntactic criterion in word-class categorization might cause problems if derivational morphology is taken into consideration. In the case of derived words such as laughter or believable, it is plausible that their category is defined before they are merged into the phrasal structure. These words are blocked in their morpho-syntactic functions and the unique base hypothesis (Scalise 1995) rules out the possibility that a derivational suffix may be applied to a semantic root. This point raises questions which would require a more extensive and in-depth analysis than is possible in this book. We can, however, sketch two possible solutions and some suggestions for analysis. Let us suppose that the presence of a derivational component within the Lexicon is to be avoided, since it would necessarily entail accepting that categories are defined in the Lexicon itself; we can then assume that, after Lexical Insertion and the consequent formation of the phrase structure, this structure does not immediately proceed to syntactic operations, but is led through a morphological sub-component where the derivation takes place. Under this hypothesis, derivational morphology maintains an independent status and is presented as a sub-routine within the syntactic computational system (whose output moves on to the various derivations). This solution has a lot in common with the principles of Distributed Morphology (Halle & Marantz 1993). According to this theory, word formation is purely syntactic or post-syntactic (contrary to Minimalist principles; see note 25 to Chapter 1). Alternatively, it can be assumed that derivational morphology, like inflectional morphology, fully takes place within syntax; after it is merged into the phrase structure, a lexical head undergoes any number of movements to functional projections dedicated to derivation. This more radical syntacticist position is supported by Starke (2001), amongst others. An asterisk indicates the ungrammaticality of the relevant example. The example from Breton is taken from Borsley & Stephen (1989). See note 17 to Chapter 2 for a discussion on triggers. The concept of classifier is a rather complex one. It is related to the interpretation of the noun head and, in particular, to the possibility of referring to an individual entity and modifying it by describing its properties, quantity and so on. For a speaker of Italian or English, noun modification by means of a

Notes 327 numeral, a demonstrative or a quantifier is quite clear, as is shown in the following examples: (i) a. b. c. d. e.

Tre persone / three people. Questa lampada / this lamp. Queste poche mucche / these few cows. Quei sei libri / those six books. Tutte le case / all the houses.

However, translating these expressions into a language without nominal inflection, such as Chinese, presents problems. As can be seen, the corresponding phrases are ungrammatical: (ii) a. *SƗn three b. *Zhèi this c. *Zhèi this

rén. person dƝng. lamp jƱ niú. few cow

d. e.

*Nèi that *ZhƟng all

liù shnj. six books fángzi. house

The NPs in (ii), however, become perfectly grammatical when a ‘particle’ is added between the modifier and the noun. The choice of this particle – the classifier – depends on the semantic features of the noun (human, animate, vertical/horizontal object, and so on, including extremely fine distinctions): (iii)a. SƗn ge rén. d. Nèi liù bƟn shnj. three CLASS person that six CLASS book b. Zhèi zhăn dƝng. e. ZhƟng ge fángzi. this CLASS lamp all CLASS house c. Zhèi jƱ tiáo niú. this few CLASS cow The function of the classifier is therefore to ensure that a linguistic item has an individual reference and relates it to other elements in the sentence. Its presence also depends on morphological typology issues. In languages with no (or poor) morphology, where dependency relations between constituents are expressed almost exclusively by syntax, classifiers make it possible to interpret a word immediately as belonging to a specific class, modify it within its phrase and realize long-distance dependencies, as in the Mayali case in (39b). See Aikhenvald (2003) for a discussion of this issue. 31 It is important to note that our formal approach does not deny the crucial role of semantics in defining categories. It rather aims at integrating the various levels of analysis, while specifying their properties and functions. 32 Interestingly, there are no functional categories in the VP.

328 Notes 33 In this context, Baker’s (1988, 1996) work is particularly crucial. Baker justified incorporation from a theoretical point of view (according to the Morphological Visibility Condition) and proved the existence of specific sequential orders between functional categories, based on an extensive crosslinguistic analysis. The morpho-syntactic interpretation of these sequences later led to the identification of a hierarchical order among the functional projections that make up the syntactic structure of the IP area (Cinque 1999). 34 In this language the lengthening of the final vowel of verbs (represented by the IPA symbol “:”) is a functional morpheme meaning perfectiveness. 35 These considerations provide an opportunity to reflect on the meaning of word. The previous examples clearly show that, semantically, the concept of word is an extremely elusive one, as it does not refer to a single entity. In our approach, the word has to do with the phonological output of syntactic derivations (see the model in [1]), i.e. incorporation operations involving lexical and functional heads. For a discussion see Lyons (1968), Simone (1997), De Mauro (2005). 36 The syntactic function of predicate corresponds to what is known as predication in semantics (see Chapter 2, § 2.1.). 37 The analysis of argument roles and their correlation with event structures will be discussed more specifically in Chapter 2. 38 The same phenomenon is common to a number of Austronesian languages and will be studied in more depth in the following chapters with respect to Tagalog. 39 We will come back to the system and characteristics of ergative languages in Chapter 3. 40 Non-agent subjects may have morpho-syntactic consequences even in nominative-accusative languages. For example, in Italian the clitic si causes verbs such as sciogliere (‘melt’) and muovere (‘move’) to be interpreted as middles in the second sentence of each of the following pairs: (i) a.

b.

(ii) a.

Io sciolgo il ghiaccio. S.PRO.1SG melt.PRES.1SG DET ice ‘I melt the ice.’ Il ghiaccio si scioglie. DET ice MIDD melt.PRES.3SG ‘(The) ice melts.’

(ITALIAN)

Io

muovo questa macchina. move.PRES.1SG DEM car ‘I move this car.’ b. Questa macchina si muove lentamente. DEM car MIDD move.PRES.3SG slowly ‘This car moves slowly.’ S.PRO.1SG

41 Identifying and specifying copular sentences are distinguished in the literature. The former have a definite subject, identified by the predicate (as in [69]); in

Notes 329 the latter, the subject denotes a set and the predicate specifies what defines that set (as in [71]). See also Huber (2000) and Den Dikken (2001) on this issue. 42 As shown in (76), Russian uses a copula in these constructions in the past tense. This is an extremely frequent case in the world’s languages: zero copula is therefore found in the unmarked case (declarative, affirmative sentence in the present tense), whereas explicit copula appears in negative sentences and in the past and future tenses.

2.

The structure of the verb phrase

1 A Small Clause (SC) is by definition a verbless sentence. This structure has in recent years come to be considered the best way to account for the syntax of copular (or nominal) sentences, as well as predicative structures with raising verbs. As shown cross-linguistically in some of the papers in Cardinaletti & Guasti (1995), a Small Clause is a sentential structure with no IP functional area, in which a subject NP is directly linked (as a sister node) to an XP functioning as the predicate. The SC is typically a subordinate clause. In the case of a copular sentence, the SC is subordinate to a copula (i), whereas in the case of a raising verb the root clause verb is believe or seem (ii). The subject of the subordinate Small Clause cannot receive its syntactic function from a non-verbal predicate (as in that case the required Spec-head structural configuration could not be verified; see section 1.5.1.); therefore, the subject needs to be raised to the Spec,IP position of the root clause, where an inflected verb (or copula) can fulfill this function (also see Moro 1997): (i) [IP[I is [SC[NP Jim] [AP intelligent]]]] Æ [IP [NPJim] [I is [SC tNP[AP intelligent]]]]. (ii) [IP [I seems [SC [NP Jim] [AP intelligent]]]] Æ [IP [NP Jim] [I seems [SC tNP [AP intelligent]]]]. In languages with no overt copula (such as those in examples [7] to [11] below), we assume that a copula may have a phonologically null realization, at least in affirmative declarative sentences. 2 It is important to note, however, that if a tense other than the present or negative polarity is to be expressed in these languages, an overt copula is normally used to realize TAM features or negation. This confirms that the absence of a copula in affirmative declarative sentences must actually be interpreted as a null realization of the copula (see note 1 to Chapter 2). In some languages, e.g. Tagalog, adverbial NPs (such as yesterday, tomorrow, etc.) must be used to give a sentence a temporal location.

330 Notes 3 Waa is not a copula, but rather a sentential illocutive (declarative) force marker (cf. Frascarelli & Puglielli 2007). 4 Tagalog is a predicate-initial language. Therefore, a nominal/adjectival predicate (like a verb in a predicative sentence) obligatorily precedes its subject NP. 5 The insertion of derivational and inflectional morphology has been a much debated topic in the generative theoretical model since the Seventies. This book will mainly cover that portion of morphology which interfaces with syntax (VP structure, see § 1.2.3.1.). As far as derivational morphology is concerned, our discussion will be limited to understanding the relevant phenomena and data (see also note 26 to Chapter 1). This is because derivational morphology is more closely related to lexical organization than to syntax, regardless of which section of our formal description deals with its data and behavior. 6 The concept of valency, borrowed from chemistry, was introduced into linguistics by Tesnière (1959). 7 The number and type of argument roles (or deep Cases) has been a much debated topic in the literature, with various approaches from Fillmore (1968, 1970) and Anderson (1968, 1971) to more recent theoretical proposals (Dowty 1991, Reinhart 2000, 2002, Alexiadou & Anagnostopoulou 2006). In this book, a framework is adopted which satisfies both theoretical and typological approaches. 8 This is the proposal made by Van Valin & LaPolla (1997: 139), whose definition of macro-role is the following: “macroroles are generalizations across the argument-types found with particular verbs which have significant grammatical consequences; it is them, rather than specific arguments in logical structure, that grammatical rules refer to primarily.” 9 From this point on, argument roles (or θ-roles), as selected by the lexical semantics of the predicate, will be graphically distinguished from macro-roles, which represent the syntactic realization of argument roles: The former will be indicated in angle brackets (, , , etc.), the latter in small capitals (ACTOR, PATIENT, etc.). In the glosses, macro-roles will be indicated with acronyms (see the List of Abbreviations). 10 Of course, our assumption does not imply a perfect coincidence between the sets combined in a lexical item in different languages. It can be assumed that different lexicalizations result in the same argument structure. Let us compare examples from English (i) and Italian (ii): (i)

The plane flies over the Alps.

(ii) L’

aereo sorvola plane fly.over.PRES.3SG ‘The plane flies over the Alps.’ DET

le DET

Alpi. Alps

(ITALIAN)

11 This reciprocal relationship has been formulated in Generative Grammar as the Thematic Criterion (or θ-Criterion). According to this, “each argument bears

Notes 331

12

13

14

15

16 17

18 19

20 21

one and only one θ-role, and each θ-role is assigned to one and only one argument” (Chomsky 1981: 36) (also see § 2.2.). The same is not true for modal particles (i.e. functional heads) in the IP area: Being heads, they are incorporated after movement of the verb along the various functional projections. This topic will be discussed in detail in Chapter 5, § 5.3. VanValin & LaPolla (1997: 140) also affirm that while different argument roles need to be distinguished semantically, from a formal perspective the grammars of languages treat some argument roles in the same manner as part of a single set. As affirmed by VanValin & LaPolla (1997: 141): “Actor does not necessarily mean agent, but rather something more general. Even a category such as ‘doer of the action’ is too restrictive, because experiencers and perceivers do not really do anything. The term actor is perfectly compatible with a non-volitional thing, such as in ‘the key opened the door’ where ‘key’ is the actor, and certainly is inanimate and non-volitional”. Interestingly, in some languages such as Tagalog, verbal morphology marks the and the in the same way, as both are first participants in an event. Quoting VanValin & LaPolla (1997: 142) once again, we may say that these participants “are things which undergo something in a general sense”. In Tagalog, verb morphology signals the argument role of the constituent functioning as a trigger (also see § 3.2.1.), i.e. the constituent playing the main role in the event expressed by the predicate. Its function is similar, while not identical, to that of a subject, and is best described with the idea of foreground (i.e. as the most salient constituent in a predication; see Schachter 1990). The trigger is usually preceded by the marker ang (see [36]) or si, when the trigger is a proper name. In (30) no trigger is overtly realized; nevertheless the morpheme um- indicates that a trigger must be present (though unpronounced), which is either an ACTOR or a THEME (see note 15 to Chapter 2). See the idea of shadow argument in Tenny & Pustejovsky (eds.) (2000). As we know, according to EPP (see note 19 to Chapter 1), each sentence must have a subject. Formally, this means that the Spec,IP position needs to be filled by an NP with that function. In English and French a restriction applies whereby this position must obligatorily be filled by an overt constituent. On the other hand, in the grammar of languages such as Italian, Spanish and Russian (called pro-drop languages) this restriction does not apply: In these languages, the subject position may be realized by a phonologically null constituent (called pro in Generative Grammar). This constituent may have an argumental role (as in: pro sta cantando, ‘pro is singing’) or be an expletive (as in: pro piove, ‘pro is raining’). These have traditionally been considered intransitive uses of transitive verbs. In both cases, however, a quantifier such as a lot is acceptable in the answer (Yes, a lot). This is because this type of quantifier can modify the verb

332 Notes (functioning as an adverbial) rather than its argument. Its presence is thus independent of the argument structure of the verb. 22 As is well-known, motion verbs belong to the class of unaccusative verbs, i.e. intransitive verbs whose subject is not an ACTOR but a THEME. Ample syntactic evidence shows that the subject of these verbs behaves syntactically as an argument generated in Compl,VP. For example, in Italian the subject of verbs such as andare, venire, arrivare (‘go’, ‘come’, ‘arrive’) etc. may be replaced by the clitic ne (ia,b), just like the object of a transitive verb (iia,b). This is ruled out for ACTOR-subjects (iiia,b): (i) a.

b.

(ii) a.

b.

Sono arrivati arrive.PRT.PLM be.PRES.3PL ‘Many students have arrived.’ Ne sono OBL.CL be.PRES.3PL ‘Many (of them) have arrived.’

molti many

arrivati arrive.PRT.PLM

Leo ha mangiato Leo have.PRES.3SG eat.PRT ‘Leo has eaten many sandwiches.’ Leo ne ha Leo OBL.CL have.PRES.3SG ‘Leo has eaten many (of them).’

(iii) a. Hanno telefonato have.PRES.3PL phone.PRT ‘Many students have called.’ b. *Ne hanno OBL.CL have.PRES.3PL ‘Many (of them) have called.’

studenti. student.PL

(ITALIAN)

molti. many

molti many

panini. sandwiches

mangiati eat.PRT.PLM

molti many

studenti. student.PL

telefonato phone.PRT

molti. many

molti. many

These verbs consistently select the auxiliary essere (‘be’) in Italian, just like verbs with passive morphology (whose subject is a PATIENT, i.e. another macrorole generated in Compl,VP). For details and an in-depth discussion, see Burzio (1986). 23 A possible objection is that in sentences such as (53) and (55) it is actually possible to omit the PPs to Jim and from Frankfurt, respectively. It is known, however, that the possibility of omitting an argument does not change the argument structure of a verb, but implies that the omitted argument is considered irrelevant to the utterance (and therefore uninformative) in that particular context. The GOAL in the predicate be clear and the LOCATIVE in the predicate leave are part of the semantic selection of the two verbs. Even when these argument are absent, they should thus be considered underlying (as in the cases in [42] and [43]).

Notes 333 24 The verbs used to realize the PATIENT overtly are those known as light verbs (see Larson 1988) such as give or do. 25 Also note that the instrumental complement may be omitted in (65) and (66); the two remaining constituents, on the other hand, are obligatory, as is expected for arguments. 26 This sentence is acceptable in specific contexts, such as the following: Mary works at a textile factory. She cuts for a living. 27 It should be noted that the semantics of the verb interacts with verb tenses. A sentence like (69) in the past (The knife cut well) again has a dynamic-verb interpretation, as the predication of a property requires an atemporal use of the present tense; see Bertinetto (1994). In this case, the macro-role of knife is ACTOR, whereas the PATIENT is a non-lexicalized (generic) shadow argument, as in the case of eat in (42)-(43). 28 On the notion of trigger, see note 17 to Chapter 2. As for the alignment of gumaling with glosses, cf. the notice in the List of Abbreviations. 29 In the late Eighties, Larson (1988) developed a more elegant syntactic hypothesis to account for this VP-internal doubling (known as VP-Shell). The same proposal was later adopted and modified by various authors (Hale & Keyser 1993, Chomsky 1993, Kayne 1994). Since the details of these proposals are beyond the scope of this work, we will assume a doubled V' for the sake of simplicity. 30 Within Generative Grammar, the possibility (or requirement) for an NP to refer to an antecedent NP is traditionally explained on the basis of Binding Theory. We will not discuss the topic here: For details, see Graffi (1994), Haegeman (1991). 31 On these constructions (found in typologically different languages such as Korean and Modern Greek), see Oehrle (1976), Bresnan (1978), Barss & Lasnik (1986), Aoun & Li (1989), Goldberg (2006). 32 We have seen that GOAL may become the subject of a passive construction, when it is realized as a direct object in the active sentence. This directly mirrors the case of the shadow PATIENT, seen with verbs like kiss, or baciare in Italian (see § 2.2.3.). When the PATIENT is included in the verb root, the GOAL is realized as a direct object, which justifies its being projected as a subject in the passive sentence. Clearly, these mechanisms are based on purely structural criteria. They show that there is a hierarchy in macro-roles and that no vacant position is normally left when an argument is not realized in the VP. 33 Notice that participles are also marked for gender. 34 The priority of the ACTOR over other macro-roles as a candidate for subject function also reflects more strictly semanticist and functionalist theories, according to which the ACTOR role is assigned the highest degree of control in the predicate event. 35 The animateness feature appears to make a difference. 36 The verb preoccupare (‘worry’) can also have an unaccusative realization:

334 Notes (i) Tutti si sono preoccupati all MIDD be.PRES.3PL worry.PRT.PLM *dall’ esame di stato. by.DET exam of state ‘Everybody has worried about the final exam.’

dell’/ of.DET

per for

il/ DET

37 Note that while Swahili is morphologically agglutinative, it behaves syntactically like a polysynthetic (incorporating) language: The subject and object clitics appearing on the verbal head are not simply agreement markers, but argument pronouns, as proved by the fact that the corresponding full NP may be omitted. 38 Notice that the clitic referring to the PATIENT macro-role is sensitive to the feature [±animate]. For this reason it is realized as -wa- in (101), where it refers to kids, and as -m- in (102), where it refers to letter. 39 The surface form of the verb depends on morpho-phonological readjustment rules. 40 NOM Case can be segmental (when it follows a DET, as in [105]) or tonal, as in this case (see Puglielli 1981). NPs with functions other than subject appear in their default form (traditionally called non-nominative): For this reason, no Case is indicated in the glosses in such cases. 41 As we remarked in Chapter 1 (§ 1.3.2.), in Somali, an incorporating language, prepositions are realized in clitic clusters immediately preceding the verbal head, and as such are separated from their respective noun complements. 42 See Puglielli & Bruno (1988), Saeed (1987) for a more thorough analysis. 43 Another strategy for avoiding the ACTOR in Somali entails using the impersonal pronoun la, whereby the verb retains a transitive-active form: (i) Libaaxyo waa la DECL IMPERS lion.PL ‘Lions were killed.’

dilay. kill.PST.3SGM

(SOMALI)

44 While VSO is the usual notation, it is important to stress that Tagalog is actually a predicate-initial language as the initial position may also be occupied by non-verbal predicates (e.g. NPs or APs in nominal constructions; see [11]). 45 See Pearson (2000, 2001) and Travis (2003) on nag as a derivate from the infixation of -um- into the causative morpheme pag. 46 In a language like Italian, these clauses differ with respect to mood: the subjunctive would be used in clauses equivalent to (124a) and (124c), the indicative in others. This difference depends on various semantic characteristics of the subordinate clause and is always determined by the semantics of the root verb. 47 PRO may also have an arbitrary interpretation, as in (i). See Chomsky (1986b), Kayne (1991), Haegeman (1991) for Control structures:

Notes 335 (i) It’s nice PRO to leave. 48 The difference in meaning between the two sentences probably depends on different viewpoints, whereby the speaker emphasizes distinct pieces of information: The predicate sembrare in (128a) and the argument of the THEMEclause i ragazzi in (129). 49 Verbs like sembrare (‘seem’) are called raising verbs in the literature because they allow raising of the embedded subject to root position. This movement, though, is subject to a structural requirement as the embedded clause must be non-finite. In particular, if the clause is non-finite, the movement is necessary (see [i]); if it is finite, the movement is totally blocked (see [ii]): (i) *It seems the boys to be singing tomorrow. (ii) *The boys seems that are singing tomorrow. Subject raising in (ii) is required for the assignment of the syntactic function of subject: In order for the constituent moving to Spec,IP to receive that function, there has to be an inflected verb in I° whose phi-features agree with that constituent (see § 1.2.3.2.). Since a non-finite verb does not move to I°, the constituent raising to Spec,IP cannot be licensed as a subject and needs to undergo raising to enter a Spec-head relation with the inflected verb of the root clause. Conversely, when the verb of the argument clause is inflected, its subject is fully licensed and its raising is blocked, according to the general principle of economy of derivation (see Chomsky 1981 and subsequent works on raising structures). 50 Since the NP head of the relative clause is a fully referential proper name, this can only be an appositive clause: Consequently, a structure in which the head noun is not extracted from a clause-internal position (see Frascarelli & Puglielli 2005a). We will, however, not focus on these specific formal issues which fall outside the scope of this work. 51 Here is an example: (i) In lacag ah part money be.PRES.3PL ‘Give me some money.’

i OCL.1SG

sii. give.IMP.2SG

(SOMALI)

For a detailed description of Somali subordinate clauses, see Antinucci (1981).

336 Notes 3.

Syntactic functions

1 These correlations led to the well-known distinction between inherent and structural Case in the literature. In simple terms, structural Case identifies a syntactic function which is not necessarily related to a specific deep role. As a consequence, its assignment solely depends on a number of structural requirements: For example, NOM Case is assigned to any constituent (be it an argument or a non-argument) occupying the Spec,IP position. Conversely, inherent Case is directly connected to a specific deep role and as such depends on semantic information codified in the verbal head. Examples of inherent Case in the literature are Genitive (GEN) and Dative (DAT); diverging analyses exist, however, for different languages. For a thorough critical discussion of the subject, see the references cited in this Chapter and the papers in Lingua, Vol. 111 (2001), an issue entirely devoted to this topic. Here, we have chosen to avoid this distinction as we consider it redundant and not sufficiently explanatory. Under our hypothesis, the relationship between lexical semantics (expressed as argument roles) and syntax takes place in the VP, as arguments are merged into specific structural positions (see § 2.2.1.). The realization of syntactic functions (i.e. Case) is thus invariably a structural phenomenon, more or less closely related to specific macro-roles, with varying forms in different languages. 2 As we saw in Chapter 2 (§ 2.2.5.), syntax imposes restrictions on the number and type of arguments that may appear within the VP. These restrictions are immediately explained in acquisition terms, as they avoid system redudancies. 3 In a language like Italian, an expletive subject is realized as a null pronoun. 4 Authors such as Zwicky (1977), Dixon (1994), Jelinek (1998) and Nichols (2001, 2003) discuss the possibility of also considering other hierarchies, e.g. a referential hierarchy, a person hierarchy, an animateness hierarchy and a topicality hierarchy, respectively. For example, in Kashmiri, “object person must not outrank subject person” (Nichols 2003: 219), as the hierarchy is 1PERS>2PERS>3PERS. Consequently, with a 1PERS subject and a 2PERS object, the former bears NOM Case and the latter ACC Case (ia); on the other hand, if there is a 2PERS subject and a 1PERS object, the latter bears DAT Case, making the 1PERS object invisible to the hierarchy and avoiding its violation: (i) a.

b.

Chu-s-ath parina:v-a:n. be-1SG.NOM-2SG.ACC teach-PRES ‘I {ACT-SUBJ} am teaching you {GOAL-DO}.’ Chu-h-am parina:v-a:n. be-2SG.NOM-1SG.DAT teach-PRES ‘You {ACT-SUBJ} are teaching me {GOAL-IO}.’

(KASHMIRI)

Notice that while the second argument in (ia) belongs to the GOAL macro-role, it is realized as a DO, following our hierarchy in (81), Chapter 2. This is

Notes 337 because no PATIENT is present and the GOAL assumes the highest syntactic function available. 5 The is the deep role with maximum control in other languages such as English, French and Italian (see Shibatani 1975, 1976, Guasti 1993, Simone 2006) and is consequently realized as the subject in surface structure. As the of the full verb is affected by the ’s action, it is associated with a lower syntactic function, i.e. DO (also IO in a language like Italian, if DO is already assigned to another constituent with PATIENT macro-role). Let us consider the following example in Italian: (i) a.

maestra {-SUBJ} lo {-DO} ha teacher him have.PRES.3SG fatto leggere. (ITALIAN) make.PRT read.INF ‘The teacher had him read.’ La maestra {-SUBJ} gli {-IO} ha DET teacher him have.PRES.3SG fatto leggere un libro {-DO}. make.PRT read.INF INDEF.DET book ‘The teacher had him read a book.’ La

DET

b.

We will come back to causative constructions in §§ 3.2.1 and 3.4.1. 6 In other languages, the same semantic nuance is expressed by lexical means relating to the choice of the light verb: For example, the pair make/let in English or the corresponding pair fare/lasciare in Italian. 7 In English, this depends on the linear order of arguments (compare he taught the rhyme to me). In German, a language with morphological Case, linear order has no role. 8 This argument role was not dealt with in Chapter 2. It belongs to the LOCATIVE macro-role as it expresses the symbolic place where the event begins. 9 It should be noticed, in fact, that a can be realized with a PP even in the absence of a , as shown in (i): (i) Er stiehlt beim he.NOM steal.PRES.3SG at.DAT ‘He steals a ring from the jeweler.’

Juwelier jeweler.DAT

einen DET.ACC

Ring. ring.ACC

Clearly, sentences (8a) and (i) have slightly different meanings. By using a simple DAT a possession relationship is expressed, so that the sentence in (8a) means that the jeweler owned the ring. On the other hand, tsentence (i) means that the ring was in the jeweler’s shop, but might have been owned by someone else (although this is quite unlikely). Irrespective of these nuances, the fact that a simple DAT (and its relevant interpretation) cannot be realized if a is present (as in [9b]) and the possibility of uttering (i) are

338 Notes evidence in favor of our analysis. Here is an adjoined constituent and its realization depends on structural factors. 10 The realization of DAT remains unchanged even in the presence of an ACTOR, which clearly appears as an adjoined prepositional constituent since passive Voice is activated: (i) Dem Präsidenten wurde vom Volk gehuldigt. ‘The president was saluted by the people’. 11 There is an extremely rich literature on expletives, the discussion of which is well beyond the scope of this volume. For a cross-linguistic perspective, see Travis (1984), Safir (1985), amomg others. 12 This raising operation is made necessary by structural requirements as the subordinate verb in its infinitive form cannot license a subject position. This determines the raising of the embedded subject into the Spec,IP of the superordinate finite clause where the assignment of subject function is possible. 13 In some ergative languages, however, the correlation between Case and deep roles is made more complex by semantic aspects of the verb, especially when relating to the category of action. In Mandan (a North-American language), for example, intransitive verbs of motion (such as run) assign ERG Case to their argument as opposed to stative verbs (such as stand), which assign ABS Case. For a detailed discussion of ergativity, see Dixon (1994). 14 As usual, indexes are used to indicate coreference between pronouns and full NPs. 15 Furthermore, this interpretation is independent of constituent order; thus, the relative position of the constituents with ERG and ABS Cases is completely irrelevant. 16 This picture is in fact more complex, since in a number of ergative languages the pronominal system is nominative-accusative instead. In most Australian languages, for example, first and second person pronouns apply a nominativeaccusative system, whereas nouns and adjectives use an absolutive-ergative one (see Dixon 1980). 17 For a description of shared properties and interpretive differences between passive and antipassive constructions in a semantic-functional framework, see Dixon (1994). In particular, the author stresses that “a passive typically focusses on the state which the referent of the underlying O(bject) NP is in […] An antipassive, in contrast, focusses on the fact that the referent of the underlying A(gent) is taking part in some activity that involves an object, while backgrounding the object” (Dixon 1994: 149). 18 Of course, important structural differences with causatives exist, since in this case the ACTOR bears ABS Case and the PATIENT bears DAT Case, unlike in actual causative constructions. The semantic similarities we noted earlier, however, justify our formal hypothesis, although these implications cannot be dealt with here.

Notes 339 19 It should be remembered that tree diagrams are a graphic way of representing the syntactic derivation and understanding how interpretation is construed. 20 Based on the Economy Principle (Chomsky 1993, 1995), syntactic movement only takes place if needed for the interpretation of specific morpho-syntactic requirements. If the verb head does not contain grammatical features to be checked in I°, it has no reason for moving and remains in the VP. 21 The ACTOR of the full verb in a causative construction can also be realized as an adjunct (i.e. as an agent by-phrase). In this case, the coercitive meaning implied by DAT Case is avoided (or at least mitigated). Let us compare the interpretation of the two following sentences: (i) L’

ho fatto have.PRES.1SG make.PRT ‘I had Mario do it.’

OCL.3SG

fare do.INF

(ii) L’

ho fatto fare have.PRES.1SG make.PRT do.INF ‘I had Mario do it.’ OCL.3SG

a to

Mario. (coercitive) Mario

da Mario. (non-coercitive) by Mario

22 These restrictions assume particular importance from an innatist and acquisitional perspective. Universal Grammar must provide a limited number of options so that a child can acquire the language in a short time. 23 This also applies in languages with extremely reduced morphology, such as English. 24 As we can see, when the object is in postverbal position (as in [45a]), it is an indefinite individual introduced for the first time (new information). On the other hand, a direct object preceded by bă in preverbal position (cf. [45b]) is a definite element previously mentioned in the discourse context (given information). This shows that the relationship between syntactic roles and discourse grammar is crucial within a language system. We will discuss this interaction further in the final chapters of this volume. 25 The sentence-initial morpheme he is a functional head signaling dynamic actions. It is not found with stative predicates or nominal clauses. Furthermore, the morpheme indicating the grammatical person of the subject is preceded by a morpheme signaling the ACTOR macro-role of the subject itself. No morpheme is used to express any other possible argument role. 26 Notice that in passive constructions the PATIENT-subject must be realized without a preposition, even if it is [+human] and [+specific]: (i) El /*al empleado es buscado por el director. Passive thus requires a non-prepositional subject, making the distinction which applies in active sentences no longer relevant.

340 Notes 27 The third-person clitic appears as -uww in (53b) as opposed to -u in (53a). This difference is the result of a simple phonological adjustment: The long semivowel -ww- has an epenthetic function, as required by the mid back vowel (o) in the subsequent morpheme. 28 The different morpho-syntactic realizations are, as is to be expected, closely connected to information-structural options. Specifically, lion in (53a) is new information, whereas the lions in (53b) is a given element. The final chapters of this volume will be specifically devoted to information structure. 29 As shown in Table 1, some languages use several strategies for Case realization. 30 Here, the generic term preposition refers to the head of a PP governing an NP as its complement, which may also appear as a postposition (in parametric variation). The preposition is thus, by definition, an element close to the NP which it governs and to which it assigns Case. Particles, on the other hand, are discontinuous elements with respect to the NP they refer to, and do not assign Case. 31 Information distribution also has a crucial role in determining the position of clausal complements. We will not dwell on this topic here. 32 The agentive prefix nag- on the verb give in (64) signals that my father, the of the event, is the foregrounded element and functions as the point of view of the predication. For a throrough discussion, see Schachter & Otanes (1972), Schachter (1990), Travis (1998). 33 Nichols (1992), on the other hand, concludes that in these languages Case is not marked either on the head or on the dependency. 34 It can be assumed that the auxiliary is simply a functional head that realizes TAM- and Voice-related categories in analytic language systems such as Italian. This is because the auxiliary lacks an argument structure, and its presence does not influence verbal selection or Case assignment in any way. 35 The presence of the NP Giovanni in (81) should therefore be syntactically unacceptable. As we will see in Chapter 6 (§ 6.5.), however, this is not the case, since this type of construction meets specific needs related to the syntaxdiscourse grammar interface. 36 When, however, the direct object is a full NP, the clitic is optional: (i) (Les) ofrecieron queso y leche a OCL.3PL offer.PST.3PL cheese and milk to de pocos medios. of few means.PL ‘Cheese and milk were offered to poor families.’

familias family.PL (SPANISH)

Notes 341 4.

The structure of the noun phrase

1 The need for both the and the to appear as PPs in languages like Italian depends on purely structural requirements. In nominal constructions the noun head cannot assign Case to its direct arguments (unlike verbs, as seen in Chapter 3); thus, a preposition (itself a Case assigner) must be inserted to satisfy what is known as the Case Filter (see § 3.2.). 2 This particle is conventionally associated (as a suffix) with the modifier. 3 In (3b), the associative particle preceding miaoshù (‘description’) does not require the pronoun wǂ, although its presence is understood by speakers, as is clear in the translation (‘my description’). Plausibly, the presence of the pronoun in the modifying NP (‘my house’) makes the doubling of the pronoun on the NP head unnecesssary. 4 Of course, the two ways of expressing the same event imply differences in terms of Aktionsart: Whereas the verbal category expresses an action, i.e. a present, future or simply possible event, a noun presents the event as a fact (or a state). 5 It is, however, important to observe that there is a fundamental difference between nouns and verbs in terms of overt realization of argument roles. While the norm is that verbs require their valency to be overtly saturated so that the sentence has a meaning, the arguments of nouns need not be realized whenever the context makes it possible to retrieve event participants. Hence, whereas omitting the PATIENT in (i) makes the sentence incomplete, sentence (ii) is grammatical although neither argument is overtly expressed: (i) *Jack met with pleasure. (ii) The meeting was pleasant. 6 We have so far assumed, for the sake of simplicity, that the Spec,NP position is occupied by determiners (articles, demonstratives etc.). As we will see in § 3, however, these elements are functional heads that project an independent phrase (DP) dominating the NP. 7 The sentence is acceptable if the referent is given or retrievable from the utterance context. 8 In a number of languages (including French, Italian, Spanish, Russian and Tagalog), adjectives can occupy both a prenominal and a postnominal position. This variation reflects interpretive differences, as will be seen in § 4.4.1. 9 The associative particle in Tagalog, indicating the subordination relationship between the modifier and the noun head (similar to the Chinese de; see [3]), has two allomorphs, depending on whether the head precedes or follows the modifier. 10 For a discussion on the semantic properties of nominalization, see Chomsky (1970), Giorgi & Longobardi (1991).

342 Notes 11 Grimshaw (1990) proposes a distinction between nominals requiring an argument structure vs. nominals in which “only participants are present”. See Grimshaw (1990) and Siloni (1994) for further discussion. 12 This sentence is ungrammatical regardless of constituent order: (i) *La distruzione dei nemici della città… 13 Infinitival nominalization appears to be closer to a verb than nominalization by means of suffixes (such as –zione, ‘-tion’). This verbal character is made evident by the fact that the infinitival nominalization is followed by a direct object, no preposition being necessary (unlike in the other nominalized forms). 14 Some languages (e.g. some Italian dialects), however, admit articles on proper names. In this case, it can be assumed that proper nouns are not associated by default to a definiteness value, and an article is therefore needed (in D°) to activate that feature (see Longobardi 1991). Sentences like (i) are completely different: Here, the article is not used to define a whole individual, but to specify some of his properties. The truth-value of the sentence only refers to that specific subpart of the referent: (i) The Jim I know would have never done that. 15 The extraposed NP ‘pig’ appears as a discontinuous constituent (see Chapter 1, § 1.3.2) whose semantic unity is ensured by the particle ne. Baker (1996), however, does not present any specific information on this particle. We have chosen to gloss it as a classifier, considering the way it is used. 16 In Somali, the verbal head does not form a lexical unit with clitic pronouns (which is usually the case in polysynthetic languages), but appears – at least in writing – isolated at the end of the sentence. Object clitics, on the other hand, form a clitic cluster with prepositions and the subject clitic precedes this cluster and is incorporated on the element marking the Focus or the illocutive force of the sentence (cf. waa in [44a-b]). Despite this formal property, verb and object clitics form a morpho-syntactic unit, defined as a verbal complex by Puglielli (1981), insofar as they cannot be separated by other elements. Full NPs thus need to be placed in a peripheral position with respect to this predicative core. Finally, notice that the third-person object clitic is null; therefore, the clitic coreferent with Axmed in (44b), depending on the preposition ka, is indicated as Ø. 17 For further discussion of the NP structure, see Giusti (1994), Sportiche (1988), Szabolcsi (1989), Shlonsky (1991), Giorgi & Longobardi (1991). 18 In Albanian, articles can also appear before adjectives. Their presence seems to depend on an elliptical operation involving the noun head (before each DET), which takes place with some types of modifier (see Kallulli 1999 for discussion).

Notes 343 19 The use of articles, of course, has to do with the semantics of the noun head and, therefore, with the subclassificaton of nouns. For example, proper names should be kept distinct from common nouns. In the former, D° is normally empty, whereas the presence of an article in the latter depends on their [±countable] property. If a noun is [+countable] it can occur in the plural, along with various types of quantifiers (all the boys, the four boys); if the noun is [-countable] it has no plural form and cannot be modified by numeral quantifiers: (i) a. b. c.

Sugar makes you fat. *Sugars make you fat (except with the meaning ‘different types of sugar’). *The three sugars.

20 The notion of ‘operator’ comes from logic (see Carnap 1934, Reichenbach 1947). A syntactic operator is a constituent that assumes scope over a phrasal domain (usually a sentence) after a movement operation (be it overt or LF movement), in order to identify a variable, i.e. any element whose value is not defined. Quantifiers are operators insofar as they quantify the NP they modify. 21 Predicative adjectives, on the other hand, are APs functioning as the predicate of a Small Clause (see note 1 to Chapter 2) and are therefore outside the DP (e.g. il libro è bello, ‘the book is good’). 22 In fact, some English adjectives do appear postnominally with a specific meaning (defined as stage-level in the literature; see Cinque 2010). This is shown in the contrast between (ia) and (ib): (i) a. b.

There are few navigable rivers. There are few rivers navigable.

The prenominal adjective in (ia) indicates an inherent property of rivers (individual-level interpretation): What is said in this sentence is that there are only a few rivers that are always navigable. The postnominal position in (ib), on the other hand, means that there are only a few rivers that are navigable now (which does not rule out the possibility for them to be navigable at other times). The meaning and interpretive variation between adjectives depending on their positions is extremely fine and complex. See Ramaglia (2008) for references and further discussion. 23 In distributional terms, the presence of a DET shows that the relevant modifier is in fact a DP. However, since we are focusing on the order of nominal modifiers, this issue will not be considered here and a traditional AP definition will be maintained for these constituents in Arabic. 24 Judgments are extremely variable among speakers, both in terms of length (as argued by Nespor 1988) and types of adjectives. For example, phrases such as un vecchio e caro amico (‘a good old friend’), una pura e semplice casualità

344 Notes

25

26

27

28

29

30

31

(‘a pure and simple coincidence’) and so on, are completely natural. We will not study the issue in detail here, but simply note that adjectives undergo no restrictions in postnominal position. All these properties lead to the assumption that an AP in its marked (i.e. prenominal) position is not always a real modifier (at least in Italian), but forms a sort of complex word ([A+N]) with the following noun head. This issue merits further research using appropriate diagnostics, possibly from a crosslinguistic perspective. However, it must be said that the unmarked position admits both interpretations (restrictive and appositive). This means that a postnominal adjective in Italian can be used with a metaphorical meaning, as extra information or as the speaker’s opinion, the intended meaning being clarified by the linguistic or extra-linguistic context. Under no circumstances, on the other hand, can an adjective in a marked position assume a restrictive function. This order is fairly fixed in some languages (e.g. Germanic languages), whereas it shows some variation in others (e.g. Italian). In Italian, adjectives may occur in different positions depending on their specific semantic and pragmatic functions (such as emphasis, contrast, etc.) Here, we will only consider the order of adjectives in unmarked constructions. For a discussion of adjectival modification in Arabic marked constructions, see Ramaglia (2006) from which examples (85)-(87) are taken. Notice that in the Italian equivalents, the head noun is placed in an intermediate position between the modifiers. In particular, size and quality adjectives (usually) precede the head noun, while color and provenance adjectives follow it. From a semantic viewpoint, the position of the former group of adjectives presumably depends on their connotative value, while the latter obligatorily appear to the right of the noun because they are invariably restrictive. According to syntactic theory, the intermediate position of the head noun (both in Italian and in other languages) is interpreted in derivational terms. Adjectives are inserted according to a basic hierarchy (see [93]), and the NP is assumed to move along the functional projections within the NP functional area (see Cinque 2010). The suffix -n is absent in the indefinite form of the adjective ’axd̞ar ‘green’, which is a diptote adjective and, as such, lacks a specific indefiniteness marker (the diptote declension presents just two endings: nominative –u and genitiveaccusative –a, both in the definite and in the indefinite form). The mirror-image order of adjectives is explained in derivational terms in Generative Grammar (see note 28 to Chapter 4). A cyclic movement of NP is assumed into each of the functional projections, resulting in each AP being incorporated to the right of the NP. This gives the opposite order to the one in (88). For further discussion and references, see Cinque (1994, 2010) and Ramaglia (2008). In fact, these elements are not considered fully derivational, as they do not alter the category of the noun they are attached to and they are more complex than

Notes 345 derivational suffixes in terms of productivity. However, since they cannot be classified as inflectional suffixes, our hypothesis remains plausible. 32 This seems to confirm our idea that derivational morphology is part of syntax (see note 26 to Chapter 1). However, further research is needed on this topic. 33 Manner adjectives (such as violent) and modal adjectives (such as probable) modify an event and cannot apply to noun heads denoting objects (see [i]). In other words, these adjectives function in the DP in the same way as adverbs in the sentence (see [ii]-[iii]). Their location within a functional area with (semantics-based) dedicated projections is thus consistent with the current analysis of adverbs in the IP (to be dealt with in Chapter 5): (i) a. *The probable green book. b. *The violent brown bag. (ii) a. Germany’s violent invasion of Albania. b. Germany invaded Albania violently. (iii)a. The probable intervention of the armed forces against the enemy. b. The armed forces will probably intervene against the enemy. 34 Ample evidence is given in the literature supporting this analysis, including anaphor binding or the narrow scope of quantifiers (for further discussion, see Bianchi 1999, 2002a,b, Cinque 2005a, Frascarelli & Puglielli 2005a). 35 The reduced paradigm of Somali verbs has just three forms: One for the third person feminine, one for the first person plural and one for all the remaining persons. In this paradigm, agreement is strongly participial (see Puglielli 1981). 36 However, different syntactic functions of the NP head require different introducing elements in the relative clause. A simple complementizer (that) is only used for subjects and objects (where it is optional); in all other cases either a relative pronoun with the appropriate preposition or a stranded preposition is required. The relativization of the various syntactic roles thus entails using different strategies, which are dealt with in this section. 37 When the noun head is extracted, its position with respect to the relative clause depends on the head-complement parameter. Thus, in Afar, the head always appears at the end of the relative clause, whereas in VO languages it is found before it. 38 In non-standard Italian varieties, resumptive pronouns are also available for indirect objects, which makes relative pronouns unnecessary, as in example (i) (see Fiorentino 1999). In this case, the rules of the pronoun-retaining strategy apply, as shown in Persian (120c): (i) Il DET

ragazzo boy

che that

gli IOCL.3SGM

ho parlato have.PRES.1SG talk.PRT

346 Notes ieri. yesterday ‘The boy I talked to yesterday.’

(ITALIAN)

39 In Spanish and Portuguese, complementizers can be preceded by a preposition (see the phenomenon of dequeísmo, Chapter 3, § 3.5.). As previously observed, however, this is due to the presence of a null (i.e. deleted) noun head between the preposition and subordinate C° head. Consequently, our generalization can still be considered valid. 40 According to this analysis, the appositive clause is neither a complement nor an adjunct of the NP head. The latter is merged as an independent and fully referential DP, whereas the predicative CP provides additional information about it. The relationship between the NP head and the appositive clause is guaranteed by a relative pronoun coindexed with the former, located in the Spec,CP of the appositive clause.

5.

Adverbial modification

1 The main function of Modalpartikeln (Engel 1988, Werner 1998, Blasco Ferrer 1999, Coniglio 2009) is to express the attitude and participation of the speaker towards the communicative act; for this reason, these particles are often called Kontaktsignale, ‘contact signals’ in German grammars (see Weinrich 1993). They are used very frequently in spoken language. As their basic function is to give an utterance specific nuances, usually their semantic value is not easily definable, but varies according to different conversational contexts. 2 Their status as heads, rather than phrases, is evident as these elements cannot move around in the sentence, unlike adverbial constituents. The correlation between functional heads and adverbial modification will become clear later in this chapter. 3 The term adverbials, introduced by Jespersen (1924), is widely used in recent linguistic studies. In Jespersen’s analysis words are linked to each other in a syntactic hierarchy: The noun, i.e. the first-rank category, is determined by the verb, the second-rank category (which also includes adjectives), in turn determined by the adverb, the third-rank category. The category of adverb is thus presented as an extremely broad class, which can include all third-rank elements. 4 Some verbs, however, select adverbials obligatorily (e.g. behave, last). Given their limited occurrences, these verbs are considered exceptions and, as such, are not included in our analysis. 5 From the Nineties, adjunction was totally refuted by a number of authors, following Kayne’s (1994) proposal on the antisymmetry of syntax. Based on cross-linguistic data, Kayne shows that “linear precedence entails c-command”:

Notes 347

6 7

8

9 10 11

12

13

As adjunction makes it possible for two phrases to be linked to the same maximal node, it must be ruled out from the theory. Indeed, no dominance relationship can be defined in an adjunction configuration, although adjoined phrases (apparently) have a linear order. In this book the theoretical option of adjunction has been retained, in line with Chomsky (1995, 2001). However, its limitations and alternatives to it will be considered when necessary to describe and explain cross-linguistic data (as in the case of adverbials). These terms were first proposed by Jackendoff (1972) for high adverbs and Vikner (1997) for low adverbs. It should be noted that our examples, as well as those we take from other sources, are based on syntactically and intonationally unmarked sentences. Thus, it cannot be ruled out that other constituent orders are possible in marked structures as a means to convey specific meanings (as in examples [6]-[9]). We will deal with the marked position of adverbs at the end of this section. However, both Finnish and German have mixed characteristics in their basic constituent order. In particular, German is VO in root(-like) clauses and OV in subordinates. Nevertheless, these languages can be taken as representative of the OV type because they have both prepositions and postpositions (absent in head-initial languages) and above all because the position of adverbs (and adjectives) is consistently head-final. The elative Case on the relative pronoun introducing the subordinate clause usually indicates origin (for details, see Vilkuna 2003). Polysynthetic languages have not yet been taken into consideration; this exclusion will be clarified later. According to the Locality principle, each movement operation must target the closest position compatible with the moving category. Consequently, a head undergoing movement (as in this case) must pass through all the available X° positions. This restriction, which is part of the Economy principle of derivation (Chomsky 1993), is also known as the Minimal Link Condition. It should be noticed that the hierarchy in (23) is the maximum extension of functional projections, based on Cinque’s (1999) data. Of course, the whole sequence is never found, above all because it would be semantically impossible. A continuative adverb, for example, can never cooccur with a terminative one. To establish this hierarchy, Cinque often uses a transitivity criterion: Given a number of attested combinations, others are inferred as a logicalmathematical consequence (“if A precedes B in language x and B precedes C in language y, then A precedes C in the assumed linear order”). The existence of restrictions on cooccurrence should not be forgotten: This would form an interesting subject for research with a view to identifying groupings and possible internal hierarchies. It is important to notice, however, that given the great variety of cross-linguistic data it might be argued that some distinctions are overly subtle, and that semantic features are unlikely to all have a direct link to syntax in every language. Plausibly, languages present macro-semantic groupings within

348 Notes

14 15

16

17

18 19

20

21 22

individual functional heads. These relate to syntax, exactly like the interface between ș-roles (cf. Chapter 2) and syntax takes place via macro-roles. It is also expected that single nuances within each macro-function are realized in different ways across languages. Alternative orders are thus possible within a single semantic domain, with no need to resort to special pragmatic requirements. Notice that in polysynthetic languages adverbial-like information is also realized as morphemes incorporated into the verbal head. In particular, when an adverb is a (left-hand) Topic, it is associated with a rising tone; when it is right-dislocated, the curve is generally flat, or has a more limited range of tonal variation than what precedes it (as proposed by Frascarelli & Hinterhölzl 2007; see § 6.5.1.). If an adverb forms the sentence Comment on its own or is part of it, it is associated with a downgrading contour with a terminal pattern on the final tonic vowel. Finally, when an adverb is a Focus operator, its tonic syllable is marked by a pitch, forming a typical hat contour (as shown in Frascarelli 2004b). For details and discussion, see Bonvino, Frascarelli & Pietrandrea (2007). Noun and prepositional phrases with adverbial functions have not been the object of extensive study, as authors have preferred to focus on adverbs or, more recently, adverbial clauses. Our considerations are based on analyses regarding different topics, such as Vergnaud (1985) and Hinterhölzl (2000). For a discussion of adverbial PPs with a similar approach to the analysis proposed for AdvPs in § 3, see Schweikert (2005). It can be noted that in Mohawk a formal distinction is made between a locative morpheme indicating cislocative movement (i.e. towards the speaker, as in [46]) and a LOC morpheme indicating translocative movement (as in [47]). As is known, scope requires a c-commanding relation (see note 5 to Chapter 5). In fact, in (50b), the DO in final position is interpreted as new information, unlike in (50a). This is due to a general pragmatic principle of ‘Progression in new information’ (cf. § 6.3.1.), according to which the distribution of information in languages follows a pragmatically unmarked order whereby given information normally precedes new information. Within Generative Grammar two types of movement must be distinguished: Amovement (where A stands for ‘argument’), aimed at Case assignment and A'movement, targeting non Case-related positions (generally connected with discourse requirements). In the former case, binding conditions are determined by the final position of the constituent undergoing movement; in the latter, the initial position is decisive, as in the case under consideration. The grammaticality judgment on the following sentences is based, as usual, on an unmarked intonational contour with no pauses or intonation peaks. With this particular linear order, the adverb forse is interpreted as a modifier of domani only, and not of the whole predicate.

Notes 349 23 Note that in the absence of an article, the noun gece (‘night’) is interpreted as definite and specific. To obtain an indefinite interpretation, the indefinite article eer (‘a’) needs to be used. 24 The initial and final consonants (indicated as the archiphonemes D and G) may be realized as voiced or voiceless (t/d and k/g) depending on the adjacent segment (see [57a-d]). 25 The dependent paradigm is actually the most common in Somali adverbial clauses, since these are normally formed as restrictive relative clauses on a generic head which does not have a subject function within the relative clause (for details, see Antinucci [1981]). 26 Indeed, the NP mea (‘thing’) in (77a) cannot be selected by the verb ‘rain’, a monovalent verb with a shadow argument (see § 2.2.2.). 27 Notice that the nominalizer Di÷ is also found in complement clauses in Turkish (cf. Kornfilt 1993). 28 We have already noted that a (generic or retrievable) NP can be omitted whenever it is the second argument of verbs like ‘eat’ (I have eaten [something]), ‘go’ (Alex went [somewhere]) and so on (see § 2.2.2.). 29 It should be remembered that informants’ judgments are always based on an unmarked interpretation (and intonation) of the relevant sentences. Thus, sentences like those in (83)-(84) are acceptable if the adverbial CP is pronounced as a parenthetical element (preceded and followed by a pause). This option, however, is irrelevant for our analysis. 30 Note that in Somali the focused DP, besides being preverbal, must be immediately followed by a focus marker (FM), i.e. baa (see Puglielli 1981). 31 The presence of a resumptive pronoun for topicalized, as opposed to focused, constituents is one of the most common diagnostics used to distinguish the two types of extrasentential constituents in the absence a disambiguating context or intonation.

6.

The sentence as utterance

1 The number/hash sign (#) is used in this work to indicate pragmatically inappropriate (and therefore unacceptable) sentences. 2 In particular, the intonational contour shows a gradual lowering in its fundamental frequency, with a final fall on the last tonic syllable (see Brown 1983, Hayes & Lahiri 1991 for English, D’Imperio 2002, Frascarelli 2004b for Italian, Frascarelli in press for Tagalog). 3 As we know, questions are classified on the basis of the type and extension of the constituent asked about. In yes-no (or ‘polar’) questions a confirmation is asked regarding some piece of information. When the question includes an indefinite interrogative expression (such as who, what, when, etc.), it means

350 Notes

4

5

6 7

8

9

that only that part of the sentence is totally unknown and therefore asked about: This type of question is known as wh-question. Italian follows this general rule for verbs ending in -are while other imperative forms coincide with those for present indicative. An order may also be expressed by means of an infinitive form. This happens in negative imperatives (non mangiare, ‘do not eat’) and whenever orders are impersonal, as in recipes or user manuals (e.g. sbattere le uova, ‘beat the eggs’, togliere con cura l’imballaggio, ‘carefully remove the packaging’, acquistare il biglietto prima di salire sull’autobus, ‘purchase a ticket before boarding the bus’, etc.). What is presented here as a binary (given/new) opposition represents in fact a much more complex situation. These notions express the so-called activation state of information (cf. Chafe 1987), which depends on different, mutually related parameters. An element may be ‘given’ because the speaker assumes it as known by the hearer as part of shared knowledge, it might have been mentioned in the previous context, or it can be present in (or entailed by) the extra-linguistic context/situation (see among others Venneman 1975, Givón 1976, [ed.] 1983, Prince 1981, Heim 1982, Vallduví 1992, Roberts 1996, 2004). The semantic-pragmatic notion of ‘new’, on the other hand, is usually described as information which is not shared between interlocutors at a certain point in the conversational interaction. Again, however, this should not be taken as an absolute definition: What is offered as new information may be also part of the speakers’ shared knowledge or may have been previously mentioned, the purpose then being to underscore or contrast that specific element with respect to other pieces of information (see Dik 1980, Calabrese 1992, Zubizarreta 1994, Frascarelli 2000a, Krifka 2007). We will come back to these notions in the sections devoted to Topic and Focus. A predicative sentence is an assertion whereby a certain property is assigned to the individual denoted as the subject of the predication (see Kuroda 1972). In fact, the Comment (just like the Presupposition; see below) includes an inflected verb which may be followed by adjuncts, as discussed earlier. The phrase that forms the Comment is therefore larger than the VP: It is a sentence and, as such, should be indicated as IP in our schema. However, we have chosen to use lexical, as opposed to functional, categories to indicate the predication, so that the various levels of analysis might be compared. For the same reason, we indicate nominal categories as NPs, rather than DPs. Of course, a Topic may also be connected with an object (direct, indirect or oblique), using a clitic pronoun (see § 6.5. and below). However, as we are dealing here with unmarked constructions, we will focus on the correlation between Topic and subject. Here and elsewhere, the interruptions in the intonational curve are not due to prosodic pauses, but to the presence of unvoiced consonants in the sentence. The absence of voicing determines a sort of phonetic silence, which is reflected in the waveform and in the pattern of the fundamental frequency (i.e. what we refer to as the intonational curve).

Notes 351 10 Finally, note that the presence of a clitic determines agreement with the past participle, even though the auxiliary is avere. As is known, in the presence of a full NP this type of agreement never takes place with non-unaccusative verbs. (i) Paolo ha Paolo have.PRES.3SG ‘Paolo ate the cake.’ (ii) *Paolo Paolo

11

12

13 14

15

16

17

ha have.PRES.3SG

mangiato eat.PRT.SGM

mangiata eat.PRT.SGF

la DET

la DET

torta. cake

torta. cake

Agreement in (26) is due to the incorporation of the clitic on the verbal head. We will not further discuss the mechanism here. For details, see Frascarelli (2000a: Chapter 3). Indeed, additional Focus categories can be identified, such as ‘verum Focus’ (i.e. Focus on the truth value of a sentence), ‘closed Focus’ (referring to a closed set of alternatives) and ‘exhaustive Focus’ (whose denotation can only lead to a true proposition). For an overview and discussion, see Krifka (2007). For the moment, this type of constituent will be referred to as contrastive Focus, as is usual in the literature. As we will see in the final section of this chapter, however, this is not a primitive concept of grammar, since Contrast is an independent discourse category which can combine with Focus and Topic alike. There are many studies on this topic which cannot all be referred to here. For references, see Frascarelli (2000a), (2010). In the literature, the term ex situ is often used to refer to this type of strategy. We believe, however, that extra situm (‘outside the site’) more accurately expresses the distance of the focused constituent from its original location. Indeed, the term ex situ (‘from the site’) expresses adjacency to the original location (as in ex cathedra, ‘from the chair’), which is not the case of fronted Focus. Ample syntactic evidence shows that movement of the Focus into the CP area can be assumed in Logical Form, although the Focus occupies its unmarked position. This evidence includes weak crossover effects, clitic pronominal resumption, parasitic gap licensing etc. A discussion of these diagnostics is well beyond the scope of this book. For details, see Frascarelli (2000a). As Basque is an SOV language, the order in (35) not only indicates that the Focus has moved into the CP area, but also that the verb must have undergone some variation with respect to its basic position. This is because this strategy requires the Focus to be in pre-verbal position, rather than simply sentenceinitial. In other words, the Focus construction we are examining is an appropriate answer to a wh-question. Notice, however, that some Indo-European languages also use cleft sentences to this end, which confirms the extent to which this

352 Notes strategy is used in the world’s languages. A sentence like (i) in French, for instance, is a well-formed answer to the question Who fell down the stairs? (i) C’

est JEAN qui est DEM COP.3SG Jean S.REL.PRO AUX.3SG ‘JEAN fell down the stairs.’

tombé fall.PRT

dans l’ escalier. in DET stairs

18 In Somali, there are two focusing structures. One uses baa, which structurally follows the focused NP. An alternative structure uses waxaa (previously discussed in Chapter 2, § 2.5.): This element occurs pre-verbally and determines the realization of the focused constituent to a post-verbal position as in a pseudo-cleft construction (see Antinucci 1981). 19 The use of pronouns as opposed to copular forms in Focus constructions is frequently attested in the literature. Consider, for example, the following sentences from Hebrew. Here, the mere presence of a separating (strong) pronoun helps distinguish a simple copular sentence from a Focus construction: (i)

Dani more. Dani teacher ‘Dani is a teacher.’

(ii) DANI hu Dani PRO.3SGM ‘DANI is a teacher.’

(HEBREW)

more. teacher

20 A free relative clause is a relative clause whose noun head is generic (e.g., ‘person’, ‘thing’, ‘place’, etc.) and null (see § 4.4.2.3.). 21 Originally proposed in Frascarelli & Puglielli (2005a) for Cushitic languages, this analysis was extended cross-linguistically in Frascarelli (2010). 22 Topic-prominent languages will be discussed in § 6.5.4., also in relation to subject-prominent languages. 23 We will return to the structural position of FocP in the next section. 24 In the cartographic approach, the FocP position hosts all syntactic operators (see note 20 to Chapter 4) associated with new information, therefore including wh-constituents, whose target position had so far been described generically as Spec,CP. The syntactic properties of Focus and wh-phrases coincide (see Rizzi 1997 and references cited therein, among others). Their similarity accounts for their complementarity, which will be analyzed in terms of illocutionary acts in Chapter 7. 25 Interestingly, other authors came to the same conclusion for typologically different languages, such as Hungarian (see Kenesei & Vogel 1990) and Chichewa (see Kanerva 1990).

Notes 353 26 The derivation of right-hand Topics requires movement of the Comment to the left periphery. This operation will not be discussed here (cf. Frascarelli 2004b, 2008). 27 For a comparative analysis of this topic in typologically different languages, see Frascarelli & Hinterhölzl (2007), Frascarelli (2008), Frascarelli & Puglielli (2009). It is also important to notice that the distinction between different types of Topic, with specific formal properties and a hierarchical order, is confirmed by non-verbal languages (see Puglielli & Frascarelli 2007). 28 The reasons for this difference are, of course, much more complex and should be investigated in the semantic and discourse properties of Topics as well as in the mechanism of anaphoric reference (e.g., the difference between strong and weak pronouns, cf. Cardinaletti and Starke 1999). 29 In fact, Japanese seems to admit two Topics in the left periphery of the sentence (both followed by the Topic marker). However, this is limited to spoken language and possibly connected with different types of Topics. 30 As argued in Choi (1999), the TM nun may also be used to mark Contrastive Topics (see § 6.5.1. and below). However, as shown in recent analyses (Hetland 2007), the interpretive difference is expressed through a different intonational contour (as expected, based on our analysis in § 6.5.1.). 31 For additional data supporting the central role of Topics in defining antecedence relations with null subjects and objects, see Huang (2000), Sigurdsson & Maling (2008), Frascarelli (2007, 2008).

7.

Illocutionary force

1 For a semantic analysis and detailed discussion of the notion of illocutionary force cf. Chierchia & McConnell-Ginet (1990). 2 The scarce number of studies on imperatives is also noted in Chierchia & McConnell-Ginet (1990); Postal (1967) examined them from a syntactic point of view, arguing for the presence of an underlying second person subject. Exclamative sentences have not been the object of much research either (for a recent study, see Gonzáles Rodriguez 2007). We will not discuss them in depth, due to limitations of space. 3 On performative verbs, see Austin (1962), Searle (1969) and Fraser (1975), Levinson (1983). 4 For differences between direct and indirect speech with respect to markers like complementizers, see Frajzyngier (1996). 5 For example, we assume that all languages have verbs like to say, to state, to order and to exclaim. 6 It is usually assumed that root clauses are prototypically finite and subordinate clauses prototypically non-finite. We will see, however, that this is not always the case (see below and, in particular, Evans 2007).

354 Notes 7 The syntactic analysis of converbs (lexicalized serial verbal forms, coordinate and subordinate structures, etc.) is an object of debate in the literature. We will not discuss it in detail here. Readers should refer to Nedjalkov (1998) and the papers included in Haspelmath & König (eds.) (1995). 8 Note that in (29b) the lexicalization of Force is incompatible with a verb in the imperative; the subjunctive mood, typical of dependent clauses, must be used. Under our hypothesis (see above), such sentences are in fact always dependent, even when they occur with no overt complementizer, as in: (i) Entrino in campo enter.SUB.PRES..3PL in field ‘Let the players enter the field.’

i DET

giocatori. player.PL

9 Root infinitives are also found in Italian, German, Russian and Hebrew (for details and discussion, see Evans 2007). 10 The numbering of the examples is ours. 11 Additional evidence is also provided by the fact that in the presence of que a sentence like (i) can be continued with an adverb modifying the (deleted) verb of saying, while in the absence of que this is impossible (cf. (ii)): (i) Si viene mi madre, que el tabaco es tuyo, y rápidamente/educatamente. If my mum comes, (you tell her) that the tobacco is yours, and fast/politely.’ (ii) Si viene mi madre, el tabaco es tuyo, # y rápidamente/educatamente. If my mum comes, the tobacco is yours, # and fast/politely.’ 12 No temporal markers are present in the author’s glosses. However, the past tense is used in the English translations. 13 The underlying presence of a performative verb in utterances has been a major topic for generative semantics in the 70’s and a syntactic implementation of this idea was originally put forth in Ross (1970) through the so-called Performative Hypothesis. This proposal, however, was abandoned in subsequent works. 14 It is in fact plausible to hypothesize that there is more structure than indicated in (44) and, in particular, that a frozen functional area should be assumed, devoted to the realization of (present) tense, (first) person, the speaker’s degree of politeness (see § 7.4.2.) and including projections used to convey the spatial (here) and temporal (now) deictics that characterize a performative by definition. This is consistent with the formal framework assumed so far to describe the morphology-syntax interface and rules out the possibility for this information to be directly encoded in the Lexicon. While these problems are crucial from a structural point of view, they are not central to the present proposal. We will therefore leave these issues open for future research. The category of number, including its relationship with the speaker, is also a topic deserving further consideration.

Notes 355 15 The question is equivalent to: Is it true or false that you are happy to leave? 16 The cleft sentence option is not available for (49b) because in the context it is x that y the element x needs to be a phrase (DP/PP/CP) as opposed to a head (a verbal head in this case). 17 For details on the syntactic formalization of these structures and, in particular, how complex functional heads (such as muxuu, goormuu and maxaad) are formed, see Frascarelli & Puglielli (2007). 18 Notice that, although both are yes-no questions, the two types can be distinguished from a pragmatic-semantic point of view due to the type of presupposition involved: (a) A-not-A questions are used in the absence of presuppositions, i.e. in a neutral context, where the asker does not expect a positive or a negative answer; (b) ma questions, on the other hand, may be used in both neutral and nonneutral contexts, being appropriate in situations where the question expresses surprise or a request to deny a previous presupposition. 19 From a semantic point of view, imperatives can also express warnings, permissions and wishes; in any case, “whatever imperative meaning might amount to, it seems that it must be sufficiently abstract to accommodate a wide range of actions” (Chierchia & McConnell-Ginet 1990: 224). 20 This semantic incompatibility between imperative sentences and focusing is also caused by other factors. From a strictly semantic viewpoint, sentences can be divided into two groups depending on their realis vs. irrealis modality (see Saeed 2007). The former group includes declarative and interrogative sentences (both factual and non-factual); the latter comprises imperative, optative and potential sentences where factuality and time distinctions do not have a role. These are the characteristics that rule out the presence of focused NPs in these sentences (see also Lyons 1977 for further discussion on this topic). 21 For a description of Somali negative sentences, see Antinucci (1981). 22 As is known, complement-embedding verbs can either be factual (i.e. have facts as their objects, not a hypothesis or a guess) or non-factual. The notion of factuality is typically applicable to attitude verbs such as say and know, while think and believe are non-factual verbs. Hence, in (71a) the verb say is followed by that plus a finite mood, which results in a factual subordinate clause, while in (71b) the verb hope is followed by a non-factual subordinate clause (counterfactual if a past tense is used: Mark hoped Jack would tell the truth). Finally, in (71c) the verb ask, whose semantics typically implies a non-factual subordinate clause, is followed by a complementizer that expresses its nonfactuality (compare with *Jack asked that Ann told the truth). 23 In fact, a sentence like (75) shows that the second member of coordination is syntactically embedded in Korean (an SOV language) and its verb is non-finite. Without going into the details of syntactic analysis, it is important to notice that the DECL marker can only appear once, on the finite verb. 24 This is also shown by the fact that interrogative sentences such as *Did Jim ask you that Mary is coming? are ungrammatical.

356 Notes 25 Wh-constituents in fact also admit an in situ realization (as in You saw what?, You went where?); these sentences, however, are not genuine wh-questions, but are interpreted as echo-questions (aimed at expressing emphasis, irony, etc.). 26 We will not discuss here the reasons for the different Case marking of this element, i.e. accusative in (92a) vs. nominative in (92b). For a discussion, cf. Frascarelli (2010). 27 As we said earlier (§ 7.2.1.1.), this position may be reached by a Focus or a whconstituent, either in overt syntax (English, Italian, Somali, Hungarian, Basque, etc.) or in Logical Form (Chinese, Turkish, Japanese). This movement and the predicative role of these constituents (as they are new information) ensure their visibility at the PF interface for main prominence assignment. 28 As a consequence, sentences such as (i) are all in the scope of a single declarative-type performative (i.e. they are assertions) and the constituents who in (ia), what in (ib) and who in (ic) are relative rather than interrogative elements: (i) a. b. c.

29

30 31

32

33

34

Jack told Ann who (= the person that) he wants to give his pictures to. Jack asked Ann what (=the thing that) she intends to do. You were wrong to say who (= the person that) you are voting for.

The traditional classification of these subordinate clauses as indirect interrogatives is therefore based on weak arguments of an interpretive-semantic nature which are not justified at a pragmatic or syntactic level. Note that imperative sentences impose semantic restrictions on predicate type; for example, no imperative sentences may have stative predicates (*Be tall!), as a request for action may only operate on a predication which is semantically characterized as such. The non-finite form of the subordinate verb is syntactically motivated; the same is true for every other non-finite subordinate clause. In some languages, the complementizer in ForceP2 may also be omitted - for example, that in English. These cases are, however, marked and only possible in specific contexts. According to the structure in (100a-b) the ShiftP projection is also only available in the matrix C-domain. However, based on semantic and syntactic considerations, Bianchi & Frascarelli (2010) have reached the conclusion that shifting Topics constitute independent Speech Acts and, as such, the relevant projection should not be included in the scope of the sentential illocutionary force. Since the syntactic details of this proposal are still work in progress, we maintain here the structure proposed in Chapter 6 without further discussion. Similar hypotheses were recently proposed for subordinate CPs in Haegeman (2002). For a semantic characterization of topics in root/non-root contexts, cf. Bianchi & Frascarelli (2010). These topics were discussed thoroughly in the 1970s. See Searle (1969, 1975), Grice (1968, 1975) and the papers in Cole & Morgan (1975).

Notes 357 Conclusions 1 This is not to say that restrictions do not exist at a semantic-pragmatic level. However, interpretation is generally feasible, provided the formal structure is preserved. Consider the well-known Chomskyan example, Colorless green ideas sleep furiously: While semantically peculiar, the sentence is definitely interpretable. In fact, during an experiment, American students deemed it a line from a poem.

References

Abbott, Miriam 1991

Macushi. In Handbook of Amazonian Languages. Vol. 3. Desmond C. Derbyshire and Geoffrey K. Pullum (eds.), 23– 160. Berlin/New York: Mouton de Gruyter.

Abney, Steven P. 1987 The English noun phrase in its sentential aspect. Ph.D. diss., MIT. Aboh, Enoch O., Katharina Hartmann, and Malte Zimmermann (eds.) 2007 Focus Strategies in African Languages. The Interaction of Focus and Grammar in Niger-Congo and Afro-Asiatic. Berlin/New York: Mouton de Gruyter. Aikhenvald, Alexandra Y. 2003 Classifiers: A Typology of Noun Classification Devices. Oxford/New York: Oxford University Press. Aissen, Judith L. 1990 Towards a theory of agreement controllers. In Studies in Relational Grammar. Vol. 3. Paul M. Postal and Brian D. Joseph (eds.), 279–320. Chicago/London: The University of Chicago Press. Ajello, Roberto 1995 La focalizzazione in somalo. In Scritti Linguistici e Filologici in Onore di Tristano Bolelli, Roberto Ajello and Saverio Sani (eds.), 1–28. Pisa: Pacini Editore. Alexiadou, Artemis 1997 Adverb Placement: A Case Study in Antisymmetric Syntax. Amsterdam/Philadelphia: John Benjamins. Alexiadou, Artemis, and Elena Anagnostopoulou 2006 From hierarchies to features: Person splits and direct inverse alternations. In Agreement Systems, Cedric Boeckx (ed.), 41– 62. Amsterdam/Philadelphia: John Benjamins. Alexiadou, Artemis, Paul Law, André Meinunger, and Chris Wilder (eds.) 2000 The Syntax of Relative Clauses. Amsterdam/Philadelphia: John Benjamins. Anderson, John M. 1968 On the status of ‘lexical formatives’. Foundations of Language 4: 308–318. 1971 The Grammar of Case: Towards a Localistic Theory. Cambridge: Cambridge University Press.

360 References Andrzejewski, Bogumil W. 1968 Inflectional characteristics of the so-called weak verbs in Somali. African Language Studies 9: 1–51. Andrzejewski, Bogumil W. 1969 Some observations on hybrid verbs in Somali. African Language Studies 10: 47–89. Antinucci, Francesco 1981 Tipi di frase. In Sintassi della Lingua Somala, Annarita Puglielli (ed.), 219–300. (Studi Somali 2.) Roma: Ministero degli Affari Esteri. Antinucci, Francesco, and Guglielmo Cinque 1977 Sull’ordine delle parole in italiano: l’emarginazione. Studi di Grammatica Italiana 6: 121–146. Anward, Jan 2001 Parts of speech. In Language Typology and Language Universals, Martin Haspelmath, Ekkehard König, Wulf Österreicher, and Wolfgang Raible (eds.), 727–735. Berlin/ New York: Mouton de Gruyter. Aoun, Joseph 1985 A Grammar of Anaphora. Cambridge, MA: MIT Press. Aoun, Joseph, and Yen-hui Audrey Li 1989 Scope and constituency. Linguistic Inquiry 20 (2): 141–172. Appleyard, David L. 1989 The relative verb in Focus constructions: an Ethiopian areal feature. Journal of Semitic Studies 34 (2): 291–305. Austin, Jennifer R., Stefan Engelberg, and Gisa Rauh (eds.) 2004 Adverbials. Amsterdam/Philadelphia: John Benjamins. Austin, John L. 1962 How to Do Things with Words. Oxford: Clarendon Press. Baker, Brett J. 2002 How referential is agreement? The interpretation of polysynthetic dis-agreement morphology in Ngalakan. In Problems of Polysynthesis, Nicholas Evans and Hans-Jürgen Sasse (eds.), 51–85. Berlin: Akademie Verlag. Baker, Mark C. 1988 Incorporation. A Theory of Grammatical Function Changing. Chicago/London: The University of Chicago Press. Baker, Mark C. 1996 The Polysynthesis Parameter. Oxford/New York: Oxford University Press. Baker, Mark C. 2003 Lexical Categories: Verbs, Nouns and Adjectives. Cambridge: Cambridge University Press.

References 361 Barlow, Michael, and Charles A. Ferguson (eds.) 1988 Agreement in Natural Language. Approaches, Theories, Descriptions. Stanford, CA: CSLI Publications. Barss, Andrew, and Howard Lasnik 1986 A note on anaphora and double objects. Linguistic Inquiry 17 (2): 347–354. Bauer, Winifred A. 1993 Maori. London: Routledge. Bayer, Josef 2001 Asymmetry in emphatic topicalization. In Audiatur Vox Sapientiae, Caroline Féry and Wolfgang Sternefeld (eds.), 15– 47. Berlin: Akademie Verlag. Belletti, Adriana (ed.) 2004 Structures and Beyond. The Cartography of Syntactic Structures. Vol. 3. Oxford/New York: Oxford University Press. Bernstein, Judy B. 2001 The DP hypothesis: Identifying clausal properties in the nominal domain. In The Handbook of Contemporary Syntactic Theory, Mark R. Baltin and Chris Collins (eds.), 536–561. Oxford: Blackwell Publishing. Berretta, Monica 1989 Sulla presenza dell’accusativo preposizionale in italiano settentrionale: note tipologiche. Vox Romanica 48: 13–37. Bertinetto, Pier Marco 1994 Temporal reference, aspect and actionality: Their neutralization and interactions, mostly exemplified in Italian. In Tense, Aspect and Action: Empirical and Theoretical Contributions to Language Typology. Proceedings of Seminars on Verbal Semantics at Odense University in 1986 and 1987, Carl Bache, Hans Basbøll, and Carl-Erik Lindberg (eds.), 113–137. Berlin/New York: Mouton de Gruyter. Bhatt, Rajesh, and Anikó Lipták 2009 Matching effects in the temporal and locative domains. In Correlatives Cross-Linguistically, Anikó Lipták (ed.), 343–373. Amsterdam/Philadelphia: John Benjamins. Bianchi, Valentina 1997 On the structural position of time clauses. Quaderni del Laboratorio di Linguistica 11: 66–90. Bianchi, Valentina 1999 Consequences of Antisymmetry: Headed Relative Clauses. Berlin/New York: Mouton de Gruyter. Bianchi, Valentina 2002a Headed relative clauses in generative syntax – Part I. Glot International 6 (7): 197–204.

362 References Bianchi, Valentina 2002b Headed relative clauses in generative syntax – Part II. Glot International 6 (8): 235–247. Bianchi, Valentina, and Mara Frascarelli 2010 Is topic a root phenomenon?. Iberia: An International Journal of Theoretical Linguistics 2 (1): 18–65. Bickerton, Derek 1993 Subject Focus and pronouns. In Focus and Grammatical Relations in Creole Languages, Francis Byrne and Donald Winford (eds.), 189–212. Amsterdam/Philadelphia: John Benjamins. Bisang, Walter 2007 Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In Finiteness: Theoretical and Empirical Foundations, Irina Nikolaeva (ed.), 115–137. Oxford/New York: Oxford University Press. Blasco Ferrer, Eduardo 1999 Italiano e Tedesco. Un Confronto Linguistico. Torino: Paravia. Bliese, Loren F. 1981 A Generative Grammar of Afar. Dallas: The Summer Institute of Linguistics and the University of Texas at Arlington. Bloomfield, Leonard 1933 Language. New York: Allen & Unwin. Bocci, Giuliano 2009 On syntax and prosody in Italian. Ph.D. diss., University of Siena. Bonvino, Elisabetta 2006 Le Sujet Postverbal en Italien Parlé: Syntaxe, Zones et Intonation. Paris: Ophrys. Bonvino, Elisabetta, Mara Frascarelli, and Paola Pietrandrea 2007 Semantica, sintassi e prosodia di alcuni avverbi nel parlato spontaneo. In La Comunicazione Parlata. Vol. 1. Federico Albano Leoni, Francesco Cutugno, Massimo Pettorino, and Renata Savy (eds.), 547–586. Napoli: Liguori Editore. Booij, Geert 2005 The Grammar of Words: An Introduction to Linguistic Morphology. Oxford/New York: Oxford University Press. Borriero, Lavinia 1976 Grammatica Bulgara. Firenze: Licosa. Borsley, Robert D., and Janig Stephens 1989 Agreement and the position of subjects in Breton. Natural Language and Linguistic Theory 7 (3): 407–427.

References 363 Bresnan, Joan 1978

A realistic transformational grammar. In Linguistic Theory and Psychological Reality, Morris Halle, Joan Bresnan, and George A. Miller (eds.), 1–59. Cambridge, MA: MIT Press. Bresnan, Joan, and Sam Mchombo 1987 Topic, pronoun, and agreement in Chichewa. Language 63 (4): 741–782. Brody, Michael 1990 Remarks on the order of elements in the Hungarian Focus field. In Structures and Arguments. István Kenesei (ed.), 95–121. (Approaches to Hungarian 3.) Szeged: Jate. Brown, Gillian 1983 Prosodic structure and the given/new distinction. In Prosody: Models and Measurements, Anne Cutler and D. Robert Ladd (eds.), 67–78. Berlin/New York: Springer-Verlag. Büring, Daniel 1999 Topic. In Focus – Linguistic, Cognitive and Computational Perspectives, Peter Bosch and Rob van der Sandt (eds.), 142– 165. Cambridge: Cambridge University Press. Büring, Daniel, and Katharina Hartmann 1994 The dark side of wh-movement. Linguistische Berichte 149: 56–74. Burzio, Luigi 1986 Italian Syntax. A Government and Binding Approach. Dordrecht: Reidel. Byrne, Francis, and Alexander F. Caskey 1993 Focus, emphasis and pronominals in Saramaccan. In Focus and Grammatical Relations in Creole Languages, Francis Byrne and Donald Winford (eds.), 213–232. Amsterdam/Philadelphia: John Benjamins. Calabrese, Andrea 1986 Pronomina. Some properties of the Italian pronominal system. In Papers in Theoretical Linguistics, Naoki Fukui, Tova R. Rapoport, and Elizabeth Sagey (eds.), 1–46. (MIT Working Papers in Linguistics 8.) Cambridge, MA: MIT. Calabrese, Andrea 1992 Some remarks on Focus and Logical Structure in Italian. Harvard Working Papers in Linguistics 1: 91–127. Cardinaletti, Anna 2002 Against optional and zero clitics. Right dislocation vs. marginalization. Studia Linguistica 56: 29–57. Cardinaletti, Anna, and Maria Teresa Guasti (eds.) 1995 Small Clauses. (Syntax and Semantics 28.) San Diego, CA: Academic Press.

364 References Cardinaletti, Anna, and Michael Starke 1999 The typology of structural deficiency. A case study of the three classes of pronouns. In Clitics in the Languages of Europe, Henk van Riemsdijk (ed.), 145–233. Berlin/New York: Mouton de Gruyter. Carnap, Rudolf 1934 Logische Syntax der Sprache. Wien: Julius Springer. Carnie, Andrew, and Eithne Guilfoyle (eds.) 2000 The Syntax of Verb Initial Languages. Oxford/New York: Oxford University Press. Carnie, Andrew, Heidi Harley, and Sheila Ann Dooley (eds.) 2005 Verb First. Amsterdam/Philadelphia: John Benjamins. Chafe, Wallace 1976 Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Subject and Topic, Charles N. Li (ed.), 25–55. New York: Academic Press. Chafe, Wallace 1987 Cognitive constraints on information flow. In Coherence and Grounding in Discourse, Russel S. Tomlin (ed.), 21–51. Amsterdam/Philadelphia: John Benjamins. Chierchia, Gennaro 1997 Semantica. Bologna: Il Mulino. Chierchia, Gennaro, and Sally McConnell-Ginet 1990 Meaning and Grammar. Cambridge, MA: MIT Press. Choe, Hyon Sook 1995 Focus and Topic movement in Korean and licensing. In Discourse Configurational Languages, Katalin É. Kiss (ed.), 269–334. Oxford/New York: Oxford University Press. Choi, Hye-Won 1999 Optimizing Structure in Context. Scrambling and Information Structure. Stanford, CA: CSLI Publications. Chomsky, Noam 1957 Syntactic Structures. The Hague/Paris: Mouton. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam 1970 Remarks on nominalization. In Readings in English Transformational Grammar, Roderick Jacobs and Peter Rosenbaum (eds.), 184–221. London/Waltham: Ginn. Chomsky, Noam 1971 Deep structure, surface structure and semantic interpretation. In Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, Danny D. Steinberg and Leon A.

References 365 Jacobovitz (eds.), 183–216. Cambridge: Cambridge University Press. Chomsky, Noam 1976 Chomsky, Noam 1977

Chomsky, Noam 1980 Chomsky, Noam 1981 Chomsky, Noam 1982 Chomsky, Noam 1986a Chomsky, Noam 1986b Chomsky, Noam 1988 Chomsky, Noam 1993

Chomsky, Noam 1995 Chomsky, Noam 2001 Chomsky, Noam 2002

Chomsky, Noam 2004

Conditions on rules of grammar. Linguistic Analysis 2: 303– 351. On wh-movement. In Formal Syntax, Peter W. Culicover, Thomas Wasow, and Adrian Akmajian (eds.), 71–132. New York: Academic Press. Rules and Representations. New York: Columbia University Press. Lectures on Government and Binding. Dordrecht: Foris. Some Concepts and Consequences of the Theory Government and Binding. Cambridge, MA: MIT Press.

of

Knowledge of Language. Its Nature, Origin and Use. New York: Praeger Publishers. Barriers. Cambridge, MA: MIT Press. Language and Problems of Knowledge. Cambridge, MA: MIT Press. A minimalist program for linguistic theory. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel J. Keyser (eds.), 1–52. Cambridge, MA: MIT Press. The Minimalist Program. Cambridge, MA: MIT Press. Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge, MA: MIT Press. An interview on minimalism. In On Nature and Language, Adriana Belletti and Luigi Rizzi (eds.), 92–161. Cambridge: Cambridge University Press. Beyond explanatory adequacy. In Structures and Beyond. The Cartography of Syntactic Structures. Vol. 3. Adriana Belletti (ed.), 104–131. Oxford/New York: Oxford University Press.

366 References Chomsky, Noam 2005 On phases. Ms., MIT. Cinque, Guglielmo 1990a Ergative adjectives and the lexicalist hypothesis. Natural Language and Linguistic Theory 8 (1): 1–39. Cinque, Guglielmo 1990b Types of A’-Dependencies. Cambridge, MA: MIT Press. Cinque, Guglielmo 1994 On the evidence for partial N-movement in the Romance DP. In Paths Towards Universal Grammar. Studies in Honor of Richard S. Kayne, Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini (eds.), 85–110. Washington, DC: Georgetown University Press. Cinque, Guglielmo 1999 Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford/New York: Oxford University Press. Cinque, Guglielmo 2005a The prenominal origin of relative clauses. Paper presented at Linguaggio a Roma Tre, University of Roma Tre. Cinque, Guglielmo 2005b Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36 (3): 315–332. Cinque, Guglielmo 2010 The Syntax of Adjectives. A Comparative Study. Cambridge, MA: MIT Press. Cinque, Guglielmo (ed.) 2002 Functional Structure in DP and IP. The Cartography of Syntactic Structures. Vol. 1. Oxford/New York: Oxford University Press. Cole, Peter, and Jerry L. Morgan (eds.) 1975 Speech Acts. (Syntax and Semantics 3.) New York: Academic Press. Collins, Peter C. 1991 Cleft and Pseudo-Cleft Constructions in English. London: Routledge. Comrie, Bernard 1981 Language Universals and Linguistic Typology: Syntax and Morphology. Chicago/London: The University of Chicago Press. Coniglio, Marco 2009 Die Syntax der deutschen Modalpartikeln: Ihrer Distribution und Lizenzierung in Haupt- und Nebensätzen. Ph.D. diss., University of Venice and Humboldt-Universität zu Berlin.

References 367 Cook, Vivian J. 1988

Chomsky’s Universal Grammar. An Introduction. Oxford: Blackwell Publishing. Corbett, Greville G. 2006 Agreement. Cambridge: Cambridge University Press. Croft, William 1991 Syntactic Categories and Grammatical Relations. The Cognitive Organization of Information. Chicago/London: The University of Chicago Press. Crystal, David 1992 The Encyclopedic Dictionary of Language and Languages. Oxford: Blackwell Publishing. Declerck, Renaat 1988 Studies on Copular Sentences, Clefts and Pseudo-Clefts. Dordrecht: Foris. De Feu, Veronica 1995 Rapanui: A Descriptive Grammar. London: Routledge. Del Gobbo, Francesca 2003 Appositives at the interface. Ph.D. diss., University of California at Irvine. De Mauro, Tullio 2005 La Fabbrica delle Parole. Il Lessico e Problemi di Lessicologia. Torino: UTET. De Mauro, Tullio, Federico Mancini, Massimo Vedovelli, and Miriam Voghera (eds.) 1993 Lessico di Frequenza dell’Italiano Parlato. Milano: Etaslibri. De Mello, George 1995 El dequeísmo en el español hablado contemporáneo: ¿Un caso de independencia semántica?. Hispanic Linguistics 6 (7): 117– 152. Derbyshire, Desmond C., and Geoffrey K. Pullum (eds.) 1991 Handbook of Amazonian Languages. Vol. 3. Berlin/New York: Mouton de Gruyter. Dik, Simon C. 1980 Studies in Functional Grammar. London/New York: Academic Press. Dik, Simon C. 1997 The Theory of Functional Grammar. Berlin/New York: Mouton de Gruyter. Dikken, Marcel den 2001 Specificational copular sentences and pseudoclefts. Ms., CUNY Graduate Center. [Published in The Blackwell Companion to Syntax. Vol. 4. Martin Everaert and Henk van Riemsdijk (eds.) (2006), 292–409. Oxford: Blackwell Publishing.]

368 References D’Imperio, Mariapaola 2002 Italian intonation: An overview and some questions. Probus 14 (1): 37–69. Dixon, Robert M. W. 1980 The Languages of Australia. Cambridge: Cambridge University Press. Dixon, Robert M. W. 1982 Where Have All the Adjectives Gone? and Other Essays in Semantics and Syntax. Berlin/New York: Mouton de Gruyter. Dixon, Robert M. W. 1994 Ergativity. Cambridge: Cambridge University Press. Dixon, Robert M. W. 2004 Adjective classes in typological perspective. In Adjective Classes. A Cross-Linguistic Typology, Robert M. W. Dixon and Alexandra Y. Aikhenvald (eds.), 1–49. Oxford/New York: Oxford University Press. Dobrovie-Sorin, Carmen 1995 Clitic clusters in Romanian: Deriving linear order from hierarchical structure. In Advances in Romanian Linguistics, Guglielmo Cinque and Giuliana Giusti (eds.), 55–82. Amsterdam/Philadelphia: John Benjamins. Doron, Edit 1986 The pronominal copula as agreement clitic. In The Syntax of Pronominal Clitics, Hagit Borer (ed.), 313–332. (Syntax and Semantics 19.) New York: Academic Press. Dowty, David 1991 Thematic proto-roles and argument selection. Language 67 (3): 547–619. Emonds, Joseph 1970 Root and structure-preserving transformations. Ph.D. diss., MIT. Engel, Ulrich 1988 Deutsche Grammatik. Heidelberg: Groos. Ernst, Thomas 2002 The Syntax of Adjuncts. Cambridge: Cambridge University Press. Etxepare, Ricardo 2008 Aspects of quotative constructions in Iberian Spanish. In Crosslinguistic Studies of Clause Combining: The Multifunctionality of Conjunctions, Laury Ritva (ed.), 35–78. Amsterdam/Philadelphia: John Benjamins. Evans, Nicholas 1991 A draft grammar of Mayali. Ms., University of Melbourne.

References 369 Evans, Nicholas 2002

Evans, Nicholas 2007

The true status of grammatical object affixes. In Problems of Polysynthesis, Nicholas Evans and Hans-Jürgen Sasse (eds.), 15–50. Berlin: Akademie Verlag.

Insubordination and its uses. In Finiteness: Theoretical and Empirical Foundations, Irina Nikolaeva (ed.), 366–431. Oxford/New York: Oxford University Press. Evans, Nicholas, and Hans-Jürgen Sasse (eds.) 2002 Problems of Polysynthesis. Berlin: Akademie Verlag. Everett, Daniel L., and Barbara Kern 1997 Wari’: the Pacaas Novos Language of Western Brazil. London: Routledge. Falk, Yehuda N. 2005 Open argument functions. In Proceedings of the LFG ’05 Conference, Miriam Butt and Tracy Holloway King (eds.), 136–153. Stanford, CA: CSLI Publications. Fillmore, Charles J. 1968 The case for case. In Universals in Linguistic Theory, Emmon Bach and Robert T. Harms (eds.), 1–88. New York: Holt, Rinehart and Winston. Fillmore, Charles J. 1970 Lexical entries for verbs. Synthese 21: 251–274. Fiorentino, Giuliana 1999 Relativa Debole. Sintassi, Uso, Storia in Italiano. Milano: Franco Angeli. Fodor, Jerry A. 2000 The Mind Doesn’t Work That Way. Cambridge, MA: MIT Press. Foley, William A., and Robert D. van Valin 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Fox, Barbara A., and Paul J. Hopper (eds.) 1994 Voice: Form and Function. Amsterdam/Philadelphia: John Benjamins. Frajzyngier, Zygmunt 1996 Grammaticalization of the Complex Sentence: A Case Study in Chadic. Amsterdam/Philadelphia: John Benjamins. Frascarelli, Mara 1999 Subject, nominative case, agreement and Focus. In Boundaries of Morphology and Syntax, Lunella Mereu (ed.), 195–215. Amsterdam/Philadelphia: John Benjamins.

370 References Frascarelli, Mara 2000a The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. Dordrecht: Kluwer Academic Publishers. Frascarelli, Mara 2000b Frasi scisse e ‘Small Clauses’: Un’analisi dell’inglese. Lingua e Stile 35 (3): 417–446. Frascarelli, Mara 2003 Topicalizzazione e ripresa clitica. Analisi sincronica, confronto diacronico e considerazioni tipologiche. In Italia Linguistica Anno Mille. Italia Linguistica Anno Duemila, Nicoletta Maraschio and Teresa Poggi Salani (eds.), 547–562. Roma: Bulzoni. Frascarelli, Mara 2004a Dislocation, clitic resumption and minimality: A comparative analysis of left and right Topic constructions in Italian. In Romance Languages and Linguistic Theory 2002, Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe, and Petra Sleeman (eds.), 99–118. Amsterdam/Philadelphia: John Benjamins. Frascarelli, Mara 2004b L’interpretazione del Focus e la portata degli operatori sintattici. In Il Parlato Italiano, Federico Albano Leoni, Francesco Cutugno, Massimo Pettorino, and Renata Savy (eds.), B06. Napoli: D’Auria. Frascarelli, Mara 2007 Subjects, topics and the interpretation of referential pro. An interface approach to the linking of (null) pronouns. Natural Language and Linguistic Theory 25 (4): 691–734. Frascarelli, Mara 2008 The fine structure of the topic field. In The Bantu-Romance Connection. A Comparative Investigation of Verbal Agreement, DPs, and Information Structure, Cécile De Cat and Katherine Demuth (eds.), 261–292. Amsterdam/Philadelphia: John Benjamins. Frascarelli, Mara 2010 Narrow Focus, clefting and predicate inversion. Lingua doi:10.1016/j.lingua.2010.03.021. Frascarelli, Mara in press Intonation, information structure and the derivation of inverse VO languages: An interface analysis of Tagalog and Malagasy. In Austronesian Contributions to Linguistic Theory: Selected Proceedings of AFLA, Raphael Mercado, Eric Potsdam, and Lisa Travis (eds.). Amsterdam/Philadelphia: John Benjamins.

References 371 Frascarelli, Mara, and Roland Hinterhölzl 2007 Types of topics in German and Italian. In On Information Structure, Meaning and Form, Kerstin Schwabe and Susanne Winkler (eds.), 87–116. Amsterdam/Philadelphia: John Benjamins. Frascarelli, Mara, and Annarita Puglielli 2005a The Focus system in Cushitic languages. In Proceedings of the 10th Meeting of Hamito-Semitic (Afroasiatic) Linguistics, Pelio Fronzaroli and Paolo Marrassini (eds.), 333–358. Firenze: University of Florence. Frascarelli, Mara, and Annarita Puglielli 2005b A comparative analysis of restrictive and appositive relative clauses in Cushitic languages. In Contributions to the Thirtieth “Incontro di Grammatica Generativa”, Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano (eds.), 307–332. Venezia: Cafoscarina. Frascarelli, Mara, and Annarita Puglielli 2007 Focus in the Force-Fin system. Information structure in Cushitic languages. In Focus Strategies in African Languages. The Interaction of Focus and Grammar in Niger-Congo and Afro-Asiatic, Enoch O. Aboh, Katharina Hartmann, and Malte Zimmermann (eds.), 161–184. Berlin/New York: Mouton de Gruyter. Frascarelli, Mara, and Annarita Puglielli 2009 Position, function and interpretation of topics in Somali. In Information Structure and Its Interfaces, Lunella Mereu (ed.), 325–348. Berlin/New York: Mouton de Gruyter. Frascarelli, Mara, and Francesca Ramaglia in press Pseudo-clefts at the Syntax-Prosody-Discourse interface. In Cleft Constructions, Tonjes Veeenstra, Katharina Hartmann, and Malte Zimmermann (eds.). Amsterdam/Philadelphia: John Benjamins. Frascarelli, Mara, and Alessandra Trecci 2007 Subjects in a pro-drop language. Syntactic roles, discourse categories and the interpretation of pro. In La Comunicazione Parlata. Vol. 1. Massimo Pettorino, Antonella Giannini, Marianna Vallone, and Renata Savy (eds.), 587–625. Napoli: Liguori Editore. Fraser, Bruce 1975 Hedged performatives. In Speech Acts, Peter Cole and Jerry L. Morgan (eds.), 187–210. (Syntax and Semantics 3.) New York: Academic Press.

372 References Geeraerts, Dirk 2002

Gibson, Kean 1986

Conceptual approaches III. Prototype theory. In Lexikologie/Lexicology. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen/An International Handbook on the Nature and Structure of Words and Vocabularies. Vol. 1. D. Alan Cruse, Franz Hundsnurscher, Michael Job, and Peter Rolf Lutzeier (eds.), 304–318. Berlin/New York: Mouton de Gruyter. The ordering of auxiliary notions in Guyanese Creole. Language 62 (3): 571–586.

Giorgi, Alessandra 1988 La struttura interna dei sintagmi nominali. In Grande Grammatica Italiana di Consultazione. Vol. 1. Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds.), 273–314. Bologna: Il Mulino. Giorgi, Alessandra, and Giuseppe Longobardi 1991 The Syntax of NPs: Configuration, Parameters and Empty Categories. Cambridge: Cambridge University Press. Giusti, Giuliana 1994 Enclitic articles and double definiteness: A comparative analysis of nominal structure in Romance and Germanic. The Linguistic Review 11: 241–255. Giusti, Giuliana 1997 The categorial status of determiners. In The New Comparative Syntax, Liliane Haegeman (ed.), 95–123. London: Longman. Giusti, Giuliana 2006 Parallels in clausal and nominal periphery. In Phases of Interpretation, Mara Frascarelli (ed.), 163–184. Berlin/New York: Mouton de Gruyter. Givón, Talmy 1976 Topic, pronoun and grammatical agreement. In Subject and Topic, Charles N. Li (ed.), 149–188. New York, Academic Press. Givón, Talmy 1983 Topic continuity in discourse: An introduction. In Topic Continuity in Discourse: A Quantitative Cross-Language Study, Talmy Givón (ed.), 5–41. Amsterdam/Philadelphia: Benjamins. Givón, Talmy 1984 Syntax. A Functional-Typological Introduction. Vol. 1. Amsterdam/Philadelphia: Benjamins. Givón, Talmy 2001 Syntax. An introduction. Amsterdam/Philadelphia: John Benjamins.

References 373 Givón, Talmy (ed.) 1983 Topic Continuity in Discourse: A Quantitative Cross-Language Study. Amsterdam/Philadelphia: John Benjamins. Goldberg, Adele 2006 Constructions at Work. Oxford/New York: Oxford University Press. Gómez Torrego, Leonardo 1998 Gramática Didáctica del Español. Madrid: Ediciones SM. Gonzáles Rodríguez, Raquel 2007 Reconstruction and scope in exclamative sentence. In Coreference, Modality and Focus, Luis Eguren and Olga Fernández-Soriano (eds.), 89–112. Amsterdam/Philadelphia: John Benjamins. Graffi, Giorgio 1994 Sintassi. Bologna: Il Mulino. Greenberg, Joseph 1966 Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, Joseph Greenberg (ed.), 73–113. Cambridge, MA: MIT Press. Grice, H. Paul 1968 Utterer’s meaning, sentence meaning and word meaning. Foundations of Language 4: 225–242. Grice, H. Paul 1975 Logic and conversation. In Speech Acts, Peter Cole and Jerry L. Morgan (eds.), 41–58. (Syntax and Semantics 3.) New York: Academic Press. Grimshaw, Jane B. 1990 Argument Structure. Cambridge, MA: MIT Press. Grimshaw, Jane B. 1991 Extended projections. Ms., Brandeis University. Haegeman, Liliane 1991 Introduction to Government and Binding Theory. Oxford: Blackwell Publishing. Haegeman, Liliane 2002 Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2: 117–180. Haegeman, Liliane 2003 Conditional clauses: External and internal syntax. Mind and Language 18 (4): 317–339. Haider, Hubert 2000 Towards a superior account of superiority. In Wh-Scope Marking, Uli Lutz, Gereon Müller, and Arnim von Stechow (eds.), 231–248. Amsterdam/Philadelphia: John Benjamins.

374 References Hale, Kenneth, and Samuel J. Keyser 1993 On argument structure and the lexical expression of syntactic relations. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel J. Keyser (eds.), 53–109. Cambridge, MA: MIT Press. Halle, Morris, and Alec Marantz 1993 Distributed morphology and the pieces of inflection. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel J. Keyser (eds.), 111–176. Cambridge, MA: MIT Press. Harley, Heidi, and Rolf Noyer 2000 Formal versus encyclopedic properties of vocabulary: evidence from nominalizations. In The Lexicon-Encyclopaedia Interface, Bert Peeters (ed.), 349–374. Amsterdam: Elsevier Press. Hasselgård, Hilde, Stig Johansson, Bergljot Behrens, and Cathrine FabriciusHansen (eds.) 2002 Information Structure in a Cross-Linguistic Perspective. Amsterdam/New York: Rodopi. Hawkins, John A. 1978 Definiteness and Indefiniteness. London: Croom Helm. Hayes, Bruce, and Aditi Lahiri 1991 Bengali intonational phonology. Natural Language and Linguistic Theory 9 (1): 47–96. Hedberg, Nancy 2000 The referential status of clefts. Language 76 (4): 891–920. Heim, Irene 1982 The semantics of definite and indefinite noun phrases. Ph.D. diss., University of Massachusetts at Amherst. Hengeveld, Kees 1987 Clause structure and modality in Functional Grammar. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 53–66. Dordrecht: Foris. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25 (1): 127–157. Hetland, Jorunn 2007 The Korean particle nun, the English fall-rise accent, and thetic/categorical judgements. In On Information Structure, Meaning and Form, Kerstin Schwabe and Susanne Winkler (eds.), 117–127. Amsterdam/Philadelphia: John Benjamins. Hetzron, Robert 1978 On the relative order of adjectives. In Language Universals, Hansjakob Seiler (ed.), 165–184. Tübingen: Gunter Narr Verlag.

References 375 Heycock, Caroline 2006 Embedded root phenomena. In The Blackwell Companion to Syntax. Vol. 2. Martin Everaert and Henk van Riemsdijk (eds.), 174–209. Oxford: Blackwell Publishing. Higgins, Francis Roger 1973 The pseudo-cleft construction in English. Ph.D. diss., MIT. Hinds, John 1986 Japanese. London: Routledge. Hinterhölzl, Roland 2000 Licensing movement and stranding in the West Germanic OV languages. In The Derivation of OV and VO, Peter Svenonius (ed.), 293–326. Amsterdam/Philadelphia: John Benjamins. Holton, David, Peter Mackridge, and Irene Philippaki-Warburton 1997 Greek: A Comprehensive Grammar of the Modern Language. London: Routledge. Ho-Min, Sohn 1999 The Korean Language. Cambridge: Cambridge University Press. Hopper, Paul, and Sandra A. Thompson 1973 On the application of root transformations. Linguistic Inquiry 4: 465–497. Hopper, Paul, and Sandra A. Thompson 1984 The discourse basis for lexical categories in Universal Grammar. Language 60 (4): 703–752. Hopper, Paul, and Sandra A. Thompson 1985 The iconicity of the universal categories ‘Noun’ and ‘Verb’. In Iconicity in Syntax, John Haiman (ed.), 151–183. Amsterdam/Philadelphia: John Benjamins. Horvath, Julia 1986 Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Horvath, Julia 1995 Structural focus, structural case and the notion of feature assignment. In Discourse-Configurational Languages, Katalin É. Kiss (ed.), 176–206. Oxford/New York: Oxford University Press. Huang, Yan 2000 Anaphora. A Cross-Linguistic Approach. Oxford/New York: Oxford University Press. Huber, Christian 2000 Es-Cleft und det-Cleft. Zur Syntax, Semantik und Informationsstruktur von Spaltsätzen in Deutschen und Schwendischen. Stockholm: Almquist and Wiksell International.

376 References Huck, Geoffrey J., and Younghee Na 1990 Extraposition and Focus. Language 66 (1): 51–77. Ihsane, Tabea, and Genoveva Puskás 2001 Specific is not definite. Generative Grammar in Geneva 2: 39– 54. Jackendoff, Ray S. 1972 Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jackendoff, Ray S. 1977 X-bar Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press. Jackendoff, Ray S. 1995 The boundaries of the lexicon. In Idioms: Structural and Psychological Perspectives, Martin Everaert, Erik-Jan van der Linden, André Schenk, and Rob Schreuder (eds.), 133–165. Hillsdale: Lawrence Erlbaum Associates Publishers. Jackendoff, Ray S. 2002 Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford/New York: Oxford University Press. Jelinek, Eloise 1984 The ergative split in Warlpiri. Paper presented at Linguistics Colloquium, University of Arizona. Jelinek, Eloise 1998 Voice and transitivity as functional projections in Yaqui. In Projections from the Lexicon, Miriam Butt and Willi Geuder (eds.), 195–224. Stanford, CA: CSLI Publications. Jespersen, Otto 1924 The Philosophy of Grammar. London: Allen & Unwin. Jones, Elred 1971 Krio: An English-based language of Sierra Leone. In The English Language in West Africa, John Spencer (ed.), 66–94. London: Longman. Kallulli, Dalina 1999 The comparative syntax of Albanian. On the contribution of syntactic types to propositional interpretation. Ph.D. diss., University of Durham. Kanerva, Jonni M. 1990 Focusing on phonological phrases in Chichewa. In The Phonology-Syntax Connection, Sharon Inkelas and Draga Zec (eds.), 145–161. Chicago/London: The University of Chicago Press. Kayne, Richard S. 1984 Connectedness and Binary Branching. Dordrecht: Foris.

References 377 Kayne, Richard S. 1991 Romance clitics, verb movement and PRO. Linguistic Inquiry 22 (4): 647–686. Kayne, Richard S. 1994 The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kayne, Richard S. 2005 On parameters and on principles of pronunciation. Ms., NYU. Keenan, Edward L., and Bernard Comrie 1977 Noun phrase accessibility and Universal Grammar. Linguistic Inquiry 8 (1): 63–99. Kenesei, István 2006 Focus is identification. In The Architecture of Focus, Valéria Molnár and Susanne Winkler (eds.), 137–168. Berlin/New York: Mouton de Gruyter. Kenesei, István, and Irene Vogel 1990 Focus and phonological structure. Ms., University of Szeged and University of Delaware. Kerke, Simon C. van de 1996 Affix order and interpretation in Bolivian Quechua. Ph.D. diss., UVA. Keyser, Samuel J., and Tom Roeper 1984 On the middle and ergative constructions in English. Linguistic Inquiry 15 (2): 381–416. Kihm, Alain 1999 Focus in Wolof: A study of what morphology may do to syntax. In The Grammar of Focus, Georges Rebuschi and Laurice Tuller (eds.), 245–273. Amsterdam/Philadelphia: John Benjamins. Kiss, Katalin É. 1998 Identificational Focus versus information Focus. Language 74 (2): 245–273. Kiss, Katalin É. (ed.) 1995 Discourse Configurational Languages. Oxford/New York: Oxford University Press. Kornfilt, Jaklin 1993 Infinitival wh-constructions and complementation in Turkish. In Subordination and Complementation, Kersti Börjars and Nigel Vincent (eds.), 66–83. (Eurotyp Working Papers III/4.) Strasbourg: European Science Foundation. Kornfilt, Jaklin 1997 Turkish. London: Routledge. Krifka, Manfred 2007 Basic notions of information structure. In The Notions of Information Structure, Caroline Féry, Gisbert Fanselow, and

378 References Manfred Krifka (eds.), 13–55. (Interdisciplinary Studies on Information Structure 6 – Working Papers of the SFB 632.) Potsdam: University of Potsdam. Kuno, Susumu 1976

Subject, theme, and the speaker’s empathy. A reexamination of relativization phenomena. In Subject and Topic, Charles N. Li (ed.), 417–444. New York: Academic Press.

Kuroda, Sige-Yuki 1972 The categorical and thetic judgment. Foundations of Language 9: 153–185. Lakoff, Robin T. 1968 Abstract Syntax and Latin Complementation. Cambridge, MA: MIT Press. Lambrecht, Knud 1994 Information Structure and Sentence Form. Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Larson, Richard K. 1988 On the double object construction. Linguistic Inquiry 19 (3): 335–391. Lasnik, Howard, and Mamoru Saito 1992 Move Į. Cambridge, MA: MIT Press. Lasnik, Howard, and Juan Uriagereka 2002 On the poverty of the challenge. The Linguistic Review 19: 147–150. Lazard, Gilbert 1999 La question de la distinction entre nom et verbe en perspective typologique. Folia Linguistica 33 (3/4): 389–418. Lefebvre, Claire 1986 Relexification in creole Genesis revisited: The case of Haitian Creole. In Substrata versus Universals in Creole Genesis, Pieter Muysken and Norval Smith (eds.), 279–300. Amsterdam/Philadelphia: John Benjamins. Lehmann, Christian 2004 Interlinear morphemic glossing. In Morphologie/Morphology. Ein internationales Handbuch zur Flexion und Wortbildung/An International Handbook on Inflection and Word-Formation. Vol. 2. Geert Booij, Christian Lehmann, Joachim Mugdan, and Stavros Skopeteas (eds.), 1834–1857. Berlin/New York: Mouton de Gruyter. Lepschy, Giulio C. 1990 Storia della Linguistica. Vol. 2. Bologna: Il Mulino.

References 379 Leslau, Wolf 1992

Outline of Soddo. In Gurage Studies, Wolf Leslau (ed.), 153– 180. Wiesbaden: Harrassowitz. Levinson, C. Stephen 1983 Pragmatics. Cambridge: Cambridge University Press. Li, Charles N., and Sandra A. Thompson 1976 Subject and topic: A new typology of language. In Subject and Topic, Charles N. Li (ed.), 457–489. New York: Academic Press. Li, Charles N., and Sandra A. Thompson 1981 Mandarin Chinese: A Functional Reference Grammar. Berkeley/Los Angeles: University of California Press. Lieber, Rochelle, and Sergio Scalise 2005 The lexical integrity hypothesis in a new theoretical universe. Lingue e Linguaggio 5 (1): 7–32. Lightfoot, David 1982 The Language Lottery. Towards a Biology of Grammars. Cambridge, MA: MIT Press. Lipták, Anikó 2004 On the correlative nature of Hungarian left-peripheral relatives. In Proceedings of the Dislocated Elements Workshop, Benjamin Shaer, Werner Frey, and Claudia Maienborn (eds.), 287–313. (ZAS Working Papers in Linguistics 35.) Berlin: ZAS. Lipták, Anikó 2006 Relativization strategies in temporal adjunct clauses. In Linguistic Variation Yearbook 2005, Pierre Pica, Johan Rooryck, and Jeroen van Craenenbroeck (eds.), 65–117. Amsterdam/Philadelphia: John Benjamins. Lobeck, Ann 1995 Ellipsis: Functional Heads, Licensing and Identification. Oxford/New York: Oxford University Press. Lombardi Vallauri, Edoardo 2002 La Struttura Informativa dell’Enunciato. Firenze: La Nuova Italia. Longobardi, Giuseppe 1991 Proper names and the theory of N-movement in syntax and Logical Form. University of Venice Working Papers in Linguistics 1: 1–60. Lonzi, Lidia 1991 Frasi subordinate al gerundio. In Grande Grammatica Italiana di Consultazione. Vol. 1. Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds.), 571–592. Bologna: Il Mulino.

380 References Lyons, John 1968

Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press.

Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Manning, Christopher D. 1996 Ergativity: Argument Structure and Grammatical Relations. Stanford, CA: CSLI Publications. Marantz, Alec 1995 The minimalist program. In Government and Binding Theory and the Minimalist Program, Gert Webelhuth (ed.), 351–382. Oxford: Blackwell Publishing. May, Robert 1977 The grammar of quantification. Ph.D. diss., MIT. [1991 facsimile edition. New York: Garland Publishing.] May, Robert 1985 Logical Form: Its Structure and Derivation. Cambridge, MA: MIT Press. McCloskey, James 2002 Resumption, successive cyclicity, and the locality of operations. In Derivation and Explanation in the Minimalist Program, Samuel D. Epstein and T. Daniel Seeley (eds.), 184– 226. Oxford: Blackwell Publishing. Meinunger, A. 2004 Verb position, verbal mood and the anchoring (potential) of sentences. In The Syntax and Semantics of the Left Periphery, Horst Lohnstein and Susanne Trissler (eds.), 313–341. Berlin/New York: Mouton de Gruyter. Merchant, Jason 2001 The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford/New York: Oxford University Press. Molnár, Valéria 2002 Contrast – from a contrastive perspective. In Information Structure in a Cross-Linguistic Perspective, Hilde Hasselgård, Stig Johansson, Bergljot Behrens, and Cathrine FabriciusHansen (eds.), 147–161. Amsterdam/New York: Rodopi. Moravcsik, Edith A. 1978 Agreement. In Syntax, Joseph Greenberg (ed.), 331–374. (Universals of Human Language 4.) Stanford, CA: Stanford University Press. Moro, Andrea 1995 Virtual conceptual necessity: La semplificazione della grammatica generativa nei primi anni novanta. Lingua e Stile 30 (4): 637–674.

References 381 Moro, Andrea 1997

The Raising of Predicates. Predicative Noun Phrases and the Theory of Clause Structure. Cambridge: Cambridge University Press. Nedjalkov, Igor’ V. 1998 Converbs in the languages of Europe. In Adverbial Constructions in the Languages of Europe, Johan van der Auwera and Dónall P. Ó. Baoill (eds.), 421–455. Berlin/New York: Mouton de Gruyter. Neeleman, Ad, and Tanya Reinhart 1997 Scrambling at the PF interface. In The Projections of Arguments: Lexical and Compositional Factors, Wilhelm Geuder and Miriam Butt (eds.), 309–353. Stanford, CA: CSLI Publications. Nespor, Marina 1988 Il sintagma aggettivale. In Grande Grammatica Italiana di Consultazione. Vol. 3. Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds.), 425–441. Bologna: Il Mulino. Nespor, Marina 1993 Fonologia. Bologna: Il Mulino. Nichols, Johanna 1986 Head-marking and dependent-marking grammar. Language 62 (1): 56–119. Nichols, Johanna 1992 Language Diversity in Time and Space. Chicago: Chicago University Press. Nichols, Johanna 2001 Why “me” and “thee”?”. In Historical Linguistics 1999, Laurel J. Brinton (ed.), 253–276. Amsterdam/Philadelphia: John Benjamins. Nichols, Johanna 2003 The Nakh-Daghestanian consonant correspondences. In Current Trends in Caucasian, East European and Inner Asian Linguistics, Dee Ann Holisky and Kevin Tuite (eds.), 207–264. Amsterdam/Philadelphia: John Benjamins. Nølke, Henning 1990 Les adverbiaux contextuels: Problèmes de classification. Langue Française 88 (1): 12–27. Nordlinger, Rachel, and Adam Saulwick 2002 Infinitives in polysynthesis: The case of Rembarrnga. In Problems of Polysynthesis, Nicholas Evans and Hans-Jürgen Sasse (eds.), 185–201. Berlin: Akademie Verlag.

382 References Oehrle, Richard T. 1976 The grammatical status of the English dative alternation. Ph.D. diss., MIT. Omamor, Augusta P. 1982 On native speakers and linguistic theory. Kiabara Journal of the Humanities 3: 36–54. Ortiz de Urbina, Jon 1999 Focus in Basque. In The Grammar of Focus, Georges Rebuschi and Laurice Tuller (eds.), 311–333. Amsterdam/Philadelphia: John Benjamins. Ouhalla, Jamal 1993 Subject-extraction, negation and the antiagreement effect. Natural Language and Linguistic Theory 11 (3): 477–518. Ouhalla, Jamal 1999 Focus in Arabic clefts. In The Grammar of Focus, Georges Rebuschi and Laurice Tuller (eds.), 335–359. Amsterdam/ Philadelphia: John Benjamins. Payne, Doris L. 1997 The Maasai external possessor construction. In Essays on Language Function and Language Type: Dedicated to T. Givón, Joan Bybee, John Haiman, and Sandra A. Thompson (eds.), 395–422. Amsterdam/Philadelphia: John Benjamins. Pearson, J. Matthew 2000 Two types of VO languages. In The Derivation of OV and VO, Peter Svenonius (ed.), 327–363. Amsterdam/Philadelphia: John Benjamins. Pearson, J. Matthew 2001 The clause structure of Malagasy: A minimalist approach. Ph.D. diss., UCLA. Pollock, Jean-Yves 1989 Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry 20 (3): 362–424. Pompei, Anna 2007 Sulla tipologia e l’evoluzione delle frasi relative in latino. Paper presented at Linguaggio a Roma Tre, University of Roma Tre. Postal, Paul M. 1967 Performatives and person. Ms., IBM – Thomas J. Watson Research Center. Potsdam, Eric 2006 The cleft structure of Malagasy wh-questions. In Clause Structure and Adjuncts in Austronesian Languages, HansMartin Gärtner, Paul Law, and Joachim Sabel (eds.), 195–232. Berlin/New York: Mouton de Gruyter.

References 383 Prince, Ellen F. 1981

Towards a taxonomy of given-new information. In Radical Pragmatics, Peter Cole (ed.), 223–255. New York: Academic Press.

Puglielli, Annarita 1981 Frase dichiarativa. In Sintassi della Lingua Somala, Annarita Puglielli (ed.), 3–45. (Studi Somali 2.) Roma: Ministero degli Affari Esteri. Puglielli, Annarita 2004 More on categories and interfaces. The Linguistic Review 21: 277–292. Puglielli, Annarita (ed.) 1985 Aspetti Morfologici, Lessicali e della Focalizzazione. (Studi Somali 5.) Roma: Ministero degli Affari Esteri. Puglielli, Annarita, and Biancamaria Bruno 1988 Middle voice in Somali. In Cushitic-Omotic. Papers from the International Symposium on Cushitic and Omotic Languages, Marianne Bechhaus-Gerst and Fritz Serzisko (eds.), 515–533. Hamburg: Helmut Buske Verlag. Puglielli, Annarita, and Mara Frascarelli 2007 Interfaces: The relation between structure and output. In Verbal and Signed Languages. Comparing Structures, Constructs and Methodologies, Elena Pizzuto, Paola Pietrandrea, and Raffaele Simone (eds.), 133–169. Berlin/New York: Mouton de Gruyter. Pullum, Geoffrey K. 1996 Learnability, hyperlearning, and the poverty of the stimulus. In Proceedings of the 22nd Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on the Role of Learnability in Grammatical Theory, Jan Johnson, Matthew L. Juge, and Jeri L. Moxley (eds.), 498–513. Berkeley: Berkeley Linguistics Society. Pullum, Geoffrey K., and Barbara C. Scholz 2002 Empirical assessment of stimulus poverty arguments. The Linguistic Review 19: 9–50. Quirk, Randolph, Sydney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rackowski, Andrea, and Lisa Travis 2000 V-initial languages: X or XP movement and adverbial placement. In The Syntax of Verb Initial Languages, Andrew Carnie and Eithne Guilfoyle (eds.), 117–141. Oxford/New York: Oxford University Press.

384 References Ramaglia, Francesca 2004 La struttura dell'informazione nel sintagma nominale: CP = DP?. MA diss., University of Roma Tre. Ramaglia, Francesca 2006 La sintassi degli aggettivi attributivi in arabo. Ms., University of Roma Tre. Ramaglia, Francesca 2008 La sintassi degli aggettivi e la proiezione estesa del NP. Ph.D. diss., University of Roma Tre. Ramat, Paolo, and Davide Ricca 1998 Sentence adverbs. In Adverbial Constructions in the Languages of Europe, Johan van der Auwera and Dónall P. Ó. Baoill (eds.), 187–275. Berlin/New York: Mouton de Gruyter. Ramchand, Gilliam 2008 Lexical items in complex predications: Selection as underassociation. Tromsoe Working Papers on Language & Linguistics: Nordlyd 35 (1): 115–141. Rebuschi, Georges 2005 Generalizing the antisymmetric analysis of coordination to nominal modification. Lingua 115 (4): 445–459. Rebuschi, Georges, and Laurice Tuller (eds.) 1999 The Grammar of Focus. Amsterdam/Philadelphia: John Benjamins. Reichenbach, Hans 1947 Elements of Symbolic Logic. London/New York: MacMillan. Reinhart, Tanya 1981 Pragmatics and linguistics: An analysis of sentence Topics. Philosophica 27 (1): 53–94. Reinhart, Tanya 1995 Interface strategies. OTS Working Papers in Linguistics TL-95002. Reinhart, Tanya 2000 The theta system: Syntactic realization of verbal concepts. OTS Working Papers in Linguistics TL-00-01. Reinhart, Tanya 2002 The theta system – An overview. Theoretical Linguistics 28 (3): 229–290. Renzi, Lorenzo, Giampaolo Salvi, and Anna Cardinaletti (eds.) 1988 Grande Grammatica Italiana di Consultazione. Vol. 1. Bologna: Il Mulino. Rizzi, Luigi 1986 Null objects in Italian and the theory of pro. Linguistic Inquiry 17 (3): 501–557.

References 385 Rizzi, Luigi 1997

Rizzi, Luigi 2004

Rizzi, Luigi 2006

The fine structure of the left periphery. In Elements of Grammar. Handbook in Generative Syntax, Liliane Haegeman (ed.), 281–337. Dordrecht: Kluwer Academic Publishers. Locality and left periphery. In Structures and Beyond. The Cartography of Syntactic Structures. Vol. 3. Adriana Belletti (ed.), 223–251. Oxford/New York: Oxford University Press. On the form of chains: Criterial positions and ECP effects. In Wh-Movement: Moving On, Lisa Lai-Shen Cheng and Norbert Corver (eds.), 97–133. Cambridge, MA: MIT Press.

Rizzi, Luigi (ed.) 2004 The Structure of CP and IP. The Cartography of Syntactic Structures. Vol. 2. Oxford/New York: Oxford University Press. Roberts, Craige 1996 Information structure in discourse: Towards an integrated formal theory of pragmatics. In Papers in Semantics, Jae-Hak Yoon and Andreas Kathol (eds.), 91–136. (The Ohio State University Working Papers in Linguistics 49.) Columbus, OH: The Ohio State University, Department of Linguistics. Roberts, Craige 2004 Context in dynamic interpretation. In Handbook of Pragmatics, Lauerence R. Horn and Gregory L. Ward (eds.), 197–220. Oxford: Blackwell Publishing. Robins, Robert H. 1997 Storia della Linguistica. Bologna: Il Mulino. Rooth, Mats 1992 A theory of focus interpretation. Natural Language Semantics 1: 75–116. Rosch, Eleanor 1978 Principles of categorization. In Cognition and Categorization, Eleanor Rosch and Barbara B. Lloyd (eds.), 27–48. Hillsdale: Lawrence Erlbaum Associates Publishers. Ross, John Robert 1967 Constraints on variables in syntax. Ph.D. diss., MIT. Ross, John Robert 1970 On declarative sentences. In Readings in English Transformational Grammar, Roderick A. Jacobs and Peter S. Rosenbaum (eds.), 222–272. London/Waltham: Ginn. Rouveret, Alain, and Jean-Roger Vergnaud 1980 Specifying reference to the subject: French causatives and conditions on representations. Linguistic Inquiry 11 (1): 97– 202.

386 References Saeed, John I. 1987 Saeed, John I. 1999 Saeed, John I. 2007 Safir, Kenneth J. 1985 Sapir, Edward 1921

Somali Reference Grammar. Wheaton: Dunwood Press. Somali. Amsterdam/Philadelphia: John Benjamins. The role of Focus across sentence types. Paper presented at Linguaggio a Roma Tre, University of Roma Tre. Syntactic Chains. Cambridge: Cambridge University Press. Language: An Introduction to the Study of Speech. New York: Harcourt, Brace & Co.

Sasse, Hans-Jürgen 2001 Scales between nouniness and verbiness. In Language Typology and Language Universals, Martin Haspelmath, Ekkehard König, Wulf Österreicher, and Wolfgang Raible (eds.), 495–509. Berlin/New York: Mouton de Gruyter. Saussure, Ferdinand de 1916 Cours de Linguistique Générale. Paris: Payot. Scalise, Sergio 1983 Morfologia Lessicale. Padova: Clesp. Scalise, Sergio 1995 La formazione delle parole. In Grande Grammatica Italiana di Consultazione. Vol. 3. Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds.), 471–514. Bologna: Il Mulino. Schachter, Paul 1973 Focus and relativization. Language 49 (1): 19–46. Schachter, Paul 1985 Parts-of-speech systems. In Language Typology and Syntactic Description. Vol. 1. Timothy Shopen (ed.), 3–61. Cambridge: Cambridge University Press. Schachter, Paul 1990 Tagalog. In The World’s Major Languages, Bernard Comrie (ed.), 936–958. Oxford/New York: Oxford University Press. Schachter, Paul, and Fe T. Otanes 1972 Tagalog Reference Grammar. Berkeley/Los Angeles: University of California Press. Schwarz, Florian 2004 Kikuyu Focus constructions. Syntactic and semantic issues. Paper presented at Topic and Focus: Information Structure and Grammar in African Languages, University of Amsterdam. Schweikert, Walter 2005 The order of prepositional phrases. In Contributions to the Thirtieth “Incontro di Grammatica Generativa”, Laura Brugè,

References 387 Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano (eds.), 29–147. Venezia: Cafoscarina. Scott, Gary-John 2002

Searle, John R. 1969 Searle, John R. 1975

Stacked adjectival modification and the structure of nominal phrases. In Functional Structure in DP and IP. The Cartography of Syntactic Structures. Vol. 1. Guglielmo Cinque (ed.), 91–120. Oxford/New York: Oxford University Press. Speech Acts. Cambridge: Cambridge University Press.

Indirect speech acts. In Speech Acts, Peter Cole and Jerry L. Morgan (eds.), 59–82. (Syntax and Semantics 3.) New York: Academic Press. Shibatani, Masayoshi 2004 Voice. In Morphologie/Morphology. Ein internationales Handbuch zur Flexion und Wortbildung/An International Handbook on Inflection and Word-Formation. Vol. 2. Geert Booij, Christian Lehmann, Joachim Mugdan, and Stavros Skopeteas (eds.), 1145–1165. Berlin/New York: Mouton de Gruyter. Shlonsky, Ur 1991 Quantifiers as functional heads: A study of quantifier float in Hebrew. Lingua 84 (2/3): 159–180. Sigurðsson, Halldór Á., and Joan Maling 2008 Argument drop and the empty left edge condition (ELEC). Working Papers in Scandinavian Syntax 81: 1–27. Siloni, Tal 1994 Noun phrases and nominalizations. Ph.D. diss., University of Geneva. Simone, Raffaele 1997 Esistono verbi sintagmatici in italiano?. In Lessico e Grammatica. Teorie Linguistiche e Applicazioni Lessicografiche, Tullio De Mauro and Vincenzo Lo Cascio (eds.), 155–170. Roma: Bulzoni. Simone, Raffaele 2006 Classi di costruzioni. In Zhì. Scritti in Onore di Emanuele Banfi in Occasione del suo 60° Compleanno, Nicola Grandi and Gabriele Iannaccaro (eds.), 383–409. Cesena/Roma: Caissa Italia Editore. Simpson, Jane 1991 Warlpiri Morpho-Syntax: A Lexicalist Approach. Dordrecht: Kluwer Academic Publishers. Sohn, Ho-Min 1994 Korean. London: Routledge.

388 References Speas, Margaret 1991

Functional heads and the mirror principle. Lingua 84 (2/3): 181–214. Sportiche, Dominique 1988 A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19 (3): 425–449. Sproat, Richard, and Chilin Shih 1990 The cross-linguistic distribution of adjective ordering restricttions. In Interdisciplinary Approaches to Language. Essays in Honor of S. Y. Kuroda, Carol Georgopoulos and Roberta Ishihara (eds.), 565–593. Dordrecht: Kluwer Academic Publishers. Srivastav, Veneeta 1991 The syntax and semantics of correlatives. Natural Language and Linguistic Theory 9 (4): 637–686. Starke, Michal 2001 On the inexistence of specifiers and the nature of heads. Ms., NYU. [Available online at http://ling.auf.net/lingBuzz/000004.] Stowell, Timothy A. 1981 Origins of phrase structure. Ph.D. diss., MIT. Svenonius, Peter (ed.) 2000 The Derivation of OV and VO. Amsterdam/Philadelphia: John Benjamins. Svolacchia, Marco, and Annarita Puglielli 1999 Somali as a polysynthetic language. In Boundaries of Morphology and Syntax, Lunella Mereu (ed.), 97–120. Amsterdam/Philadelphia: John Benjamins. Szabolcsi, Anna 1981 The semantics of Topic-Focus articulation. In Formal Methods in the Study of Language, Vol. 2. Jeroen Groenendijk, Theo Janssen, and Martin Stokhof (eds.), 513–541. Amsterdam: Mathematisch Centrum. Szabolcsi, Anna 1989 Noun phrases and clauses: Is DP analogous to IP or CP?. Ms., Hungarian Academy of Sciences. [Published in The Structure of Noun Phrases, John R. Payne (ed.) (1996), The Hague/Paris: Mouton.] Szabolcsi, Anna 1994 The noun phrase. In The Syntactic Structure of Hungarian, Ferenc Kiefer and Katalin É. Kiss (eds.), 179–274. (Syntax and Semantics 27.) San Diego, CA: Academic Press.

References 389 Talmy, Leonard 1985

Lexicalization patterns: Semantic structure in lexical forms. In Language Typology and Syntactic Description. Vol. 3. Timothy Shopen (ed.), 57–149. Cambridge: Cambridge University Press. Tenny, Carol L., and James Pustejovsky (eds.) 2000 Events as Grammatical Objects. Stanford, CA: CSLI Publications. Tesfay Tewolde, Yohannes 2002 A modern grammar of Tigrinya. Africa 57 (4): 646–647. Tesnière, Lucien 1959 Éléments de Syntaxe Structurale. Paris: Klincksieck. Testa, Giorgio 2007 La sintassi interna ed esterna di alcune frasi avverbiali. Uno studio comparativo. Ph.D. diss., University of Roma Tre. Toman, Jindrich 1987 Wortsyntax. Tübingen: Niemeyer. Travis, Lisa 1984 Parameters and effects of word order variation. Ph.D. diss., MIT. Travis, Lisa 1998 Theta-positions and binding in Balinese and Malagasy. Canadian Journal of Linguistics 43 (3/4): 435–467. Travis, Lisa 2003 Agents and causes in Malagasy and Tagalog. In Proceedings of the 8th Meeting of the Austronesian Formal Linguistics Association (AFLA 8), Andrea Rackowski and Norvin Richards (eds.), 355–370. (MIT Working Papers in Linguistics 44.) Cambridge, MA: MIT. Tsimpli, Ianthi Maria 1995 Focusing in Modern Greek. In Discourse Configurational Languages, Katalin É. Kiss (ed.), 176–206. Oxford/New York: Oxford University Press. Uriagereka, Juan 1998 Rhyme and Reason: An Introduction to Minimalist Syntax. Cambridge, MA: MIT Press. Valin, Robert D. Jr. van 1992 Incorporation in Universal Grammar: A case study in theoretical reductionism. Journal of Linguistics 28 (1): 199– 220. Valin, Robert D. Jr. van, and Randy J. LaPolla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Vallduví, Enric 1992 The Informational Component. New York: Garland.

390 References Vallduví, Enric 1995

Structural properties of information packaging in Catalan. In Discourse Configurational Languages, Katalin É. Kiss (ed.), 122–152. Oxford/New York: Oxford University Press. Veenstra, Tonjes, and Hans den Besten 1994 Fronting. In Pidgins and Creoles: An Introduction, Jacques Arends, Pieter Muysken, and Norval Smith (eds.), 303–315. Amsterdam/Philadelphia: John Benjamins. Vendler, Zeno 1967 Verbs and times. In Linguistics in Philosophy, Zeno Vendler (ed.), 97–121. Ithaca, NY: Cornell University Press. Venneman, Theo 1975 Topic, sentence accent and ellipsis: A proposal for their formal treatment. In Formal Semantics of Natural Language, Edward L. Keenan (ed.), 313–328. Cambridge: Cambridge University Press. Vergnaud, Jean-Roger 1974 French relative clauses. Ph.D. diss., MIT. Vergnaud, Jean-Roger 1985 Dependances et Niveaux de Representation en Syntaxe. Amsterdam/Philadelphia: John Benjamins. Vikner, Sten 1994 Finite verb movement in Scandinavian embedded clauses. In Verb Movement, David Lightfoot and Norbert Hornstein (eds.), 117–147. Cambridge: Cambridge University Press. Vikner, Sten 1997 V°-to-I° movement and inflection for person in all tenses. In The New Comparative Syntax, Liliane Haegeman (ed.), 189– 213. London: Longman. Vilkuna, Maria 2003 On expletive subjects and topics in Finnish. Paper presented at Syntactic Functions – Focus on the Periphery, University of Helsinki. Vitale, Anthony J. 1981 Swahili Syntax. Dordrecht: Foris. Vries, Mark de 2001 Patterns of relative clauses. In Linguistics in the Netherlands 2001, Ton van der Wouden and Hans Broekhuis (eds.), 231– 243. Amsterdam/Philadelphia: John Benjamins. Weimer, Harry 1972 Yareba verb morphology. Te Reo 15: 58–70. Weinrich, Harald 1993 Textgrammatik der Deutschen Sprache. Mannheim: Dudenverlag.

References 391 Werner, Angelika 1998 Deutsche Modalpartikeln im Kontrast zum Japanischen. Ph.D. diss., Siegen University. Wierzbicka, Anna 1981 Case marking and human nature. Australian Journal of Linguistics 1 (1): 43–80. Wierzbicka, Anna 1988 The Semantics of Grammar. Amsterdam/Philadelphia: John Benjamins. Williams, Edwin 1994 Thematic Structure in Syntax. Cambridge, MA: MIT Press. Wunderlich, Dieter and Renate Lakämper 2001 On the interaction of structural and semantic case. In Lingua 111 (4): 377–418. Xaange, Axmed C. 1988 Sheekoxariirooyin Soomaaliyeed Folkstales from Somalia. Uppsala: Scandinavian Institute of African Studies. Zubizarreta, Maria Luisa 1994 The grammatical representation of Topic and Focus: Implications for the structure of the clause. University of Venice Working Papers in Linguistics 4 (1): 97–126. Zwicky, Arnold M. 1977 On Clitics. Bloomington, IN: Indiana University Linguistics Club.

Subject index

adjunct, 8-10, 12, 37, 41, 47-49, 51, 59, 61, 66, 68-74, 87, 89, 9597, 104, 111, 129, 134, 157, 162, 178, 186, 196, 218, 223, 284, 300, 313, 315, 339n, 346n, 350n adjunction, 9, 178-179, 181, 186, 190, 201, 346n-347n adverb, 23-24, 32, 44, 139, 177, 180182, 186, 188-190, 192, 196, 212-213, 228, 346n-348n, 354n adverbial clause, 87, 123, 166, 180, 201-205, 207-218, 306-307, 315, 348n-349n adverbial modification, 2, 177, 314, 346n adverbial NP, 48, 195-198, 201, 211-212, 218, 290, 329n adverbial PP, 197, 200, 213, 348n agreement, 16-18, 34-37, 43, 65, 73, 79, 81, 83, 85, 104-110, 112, 115-119, 124-126, 129, 143144, 187, 191, 201, 203, 240, 245, 261-263, 272, 334n, 345n, 351n antiagreement, 240 antipassive, 102-103 argument (role), 16, 27, 35-37, 48, 53-54, 59, 61, 65, 67, 74, 77, 89, 91, 125-126, 133, 141, 164, 255, 287, 293, 331n, 337n, 339n , 17-18, 20, 36-38, 46, 50, 54, 56, 59, 66, 68-71, 76, 92, 102-103, 111, 114, 130-133, 330n-331n, 337n, 340n-341n , 17-18, 20, 36-37, 58, 92, 94-96, 330n, 337n

, 93-94, 96 , 46, 78, 93, 103, 107, 337n , 51 , 51 , 46, 50, 66, 7071, 77 , 17-18, 33, 36, 46, 50, 54, 56, 70-71 , 50, 56, 59-61, 66, 74, 103, 333n , 17-18, 20, 36, 47, 78, 92, 118, 167, 197-198, 348n , 51 , 17-18, 20, 33, 36, 38, 50, 58-60, 92, 103, 114, 125, 130-131, 330n, 337n, 341n , 51, 56, 131-133, 138, 140, 208 , 50 , 51 , 51, 56 , 51, 167 , 17-18, 50-51, 54, 102, 131, 331n argument structure, 2, 5, 17-19, 33, 45-46, 51, 53, 58-59, 61, 63, 73, 77, 98-99, 103, 127, 131132, 136, 168-169, 209, 259, 299, 324n, 330n, 332n, 340n, 342n background, 56, 254, 298, 312, 338n Case (marking), 16, 27, 82, 90-91, 95-96, 99-101, 103, 107, 109, 114, 116-125, 164, 171-172, 195, 241, 340n, 348n, 356n absolutive (ABS), 38, 100-101, 103

394 Subject index 126-127, 240, 338n abstract, 90 accusative (ACC), 27, 58, 81-85, 91, 94-95, 105-107, 112114, 122, 195, 336n, 344n dative (DAT), 91, 95-98, 103, 127, 336n, 338n-339n deep, 2, 46, 54, 63, 66, 89, 91, 93, 95-99, 101-103, 312n ergative (ERG), 36, 38, 59, 100102, 126-127, 191, 338n genitive (GEN), 130-132, 203, 210, 336n, 344n morphological, 90, 96-97, 102, 108-109, 112-114, 116, 118, 121, 130-132, 164, 172, 195, 260, 337n nominative (NOM), 17-18, 27, 84, 91, 100, 105-106, 108-109, 122, 187, 203, 240-241, 243, 334n, 336n, 344n, 356n structural, 2, 90-91, 103-104, 107, 336n category, 2, 20, 22-24, 28, 30-32, 38, 41, 44, 50, 116, 129-131, 147, 178, 188, 194, 197, 201, 311, 314, 326n, 331n, 338n, 341n, 344n, 346n-347n, 354n discourse, 227, 244, 351n functional, 17-18, 67-68, 192, 292 lexical, 29-30 phrasal, 39, 227, 232 causative, 68, 71, 74, 77-78, 93, 103, 106, 334n, 338n construction, 93-96, 103, 107, 118, 337n-339n verb, 75-76, 94 clause, 17, 19, 34, 79, 81, 83, 95, 99101, 110, 118, 127, 209, 214, 216, 218, 220, 238, 245, 262, 264, 275-278, 283, 306, 324n325n, 335n, 338n adverbial, 123, 166, 180, 201-218, 306-307, 315, 348n-349n

argument, 79-80, 82-84, 86, 294, 335n complement, 121-124, 205-206, 210-211, 217, 294-295, 315, 349n nominalized, 131, 241, 256-257, 281 relative, 2, 13-14, 39, 41, 82, 8487, 103, 130, 134-135, 148-149, 161-175, 206210, 237-243, 273, 299, 335n, 345n, 349n, 352n root, 19, 79-83, 161-164, 169, 174, 201, 203, 211, 214, 218, 271-277, 280, 283, 293, 324n, 329n, 334n-335n, 347n, 353n, 355n-356n subordinate, 13-14, 19, 81-85, 87, 123, 131, 172, 202-205, 207, 271-272, 275, 279283, 293-299, 301-305 cleft, 238, 240-241, 244, 265, 284, 296, 300, 351n-352n, 355n Comment, 2, 193, 223-227, 229, 232233, 246, 250-251, 255, 261, 264, 267, 283-284, 291, 315, 348n, 350n, 353n complementizer, 14, 16, 19, 79, 87, 124, 171, 202, 210, 245, 272, 277-281, 283, 294, 298, 304305, 315, 324n, 345n, 354n356n Contrast, 3, 180, 253, 264-267, 292, 302-304, 338n, 343n-344n, 350n-351n definiteness, 56, 113-116, 132, 139, 143, 145-147, 150, 257, 313, 328n, 339n, 344n, 349n discourse grammar, 2, 16, 19, 20, 89, 108, 194, 213, 219, 224, 226227, 232, 242-245, 255, 261, 264, 268, 281, 314, 316, 323n, 325n, 339n

Language index 395 expletive, 54, 98-99, 108, 313, 331n, 336n event, 2, 18, 32, 37, 46-47, 51-52, 6061, 68-69, 71, 73, 76, 78, 8384, 89, 93, 102-103, 106, 108, 123, 130, 132, 134, 136, 181, 220, 314, 331n, 333n, 337n, 340n-341n, 345n dynamic, 2, 50-51, 54, 56, 60, 66, 68-69, 73-76, 333n, 339n stative, 2, 50-51, 54-56, 60, 66, 7276, 127, 338n-339n, 356n structure, 2, 51, 54, 60-61, 66, 69, 72-73, 75-77, 328n Focus, 2-3, 161, 192-193, 224, 230247, 250, 264, 266-267, 283292, 296, 299-302, 315, 342n, 348n, 350n-352n, 356n cleft-like, 238, 240-241, 296, 300 contrastive, 180, 234-235, 265267, 292, 296, 302, 351n extra situm, 236-238, 243-244, 296, 300, 351n information, 234, 264, 292, 296297, 302 in situ, 236, 240, 243-244, 296297, 356n marker, 29, 238, 246, 260, 275, 349n narrow, 267, 291 (as a) predicate, 242-244 force (projection), 20, 245, 255, 281282, 304, 354n, 356n declarative, 222, 295, 297, 302, 330n illocutionary, 2-3, 5, 16, 19, 220223, 234, 246, 269-277, 281-287, 289, 291-296, 298, 304, 308, 315, 342n, 353n, 356n illocutive (see illocutionary) interrogative, 222, 284-285, 287, 298-299, 301-302

foreground, 56, 83, 93, 264, 312, 331n, 340n function, 2, 5, 8, 16, 23, 62, 64-65, 68, 91-94, 98, 102, 110, 114, 118, 121, 127, 136, 146, 153154, 164, 167, 169-172, 194195, 201, 226, 228, 234, 249251, 255, 264, 306, 327n, 329n, 331n, 340n, 344n-346n, 348n adverbial, 194-195, 201 communicative, 219-220 illocutionary, 221, 277, 305 modifier, 41, 89, 148, 177-178 object, 80, 91, 109, 113, 116, 121, 137, 141, 168, 170, 202, 313-314 pragmatic, 89, 93, 146, 176, 249, 257, 269, 272, 274, 276, 301 predicate, 24, 32, 38-39, 41, 89, 131, 174, 230, 314 syntactic, 23, 27, 32-33, 37, 4345, 64-67, 69, 73, 88-95, 98-99, 103-108, 111-114, 116-124, 126, 143, 147, 163-164, 166, 168-170, 195, 206, 224, 226-227, 229, 232, 258, 260, 279, 287, 312-314, 328n-329n, 335n-337n subject, 33-34, 37, 56, 65-66, 7176, 80, 87, 91, 100-102, 105, 107-110, 113, 118, 137, 167-169, 207-208, 227, 233, 313, 333n, 338n, 349n Topic, 226, 228, 255 hierarchy, 9-10, 17, 46, 59, 63, 88, 93, 113, 156-159, 166-168, 179, 181, 183-186, 188, 190191, 197-198, 244, 247, 254, 305, 312, 336n, 344n, 346n347n

396 Subject index (of) macro-roles, 63-65, 68-69, 312, 333n (of) syntactic functions, 64-65, 68-69, 313, 336n incorporation, 29, 31, 90, 111, 124, 187-188, 288, 328n, 351n information structure, 199, 201, 219, 223, 232, 234, 245, 250, 258, 261, 264, 267, 269, 283, 323n, 340n innatism, 6, 15, 323n interface, 2, 6-7, 15-20, 43, 46, 4950, 54, 59, 63-64, 66, 88-89, 91, 96, 99, 102, 104, 108, 213, 218-219, 236, 243-244, 246, 263, 268-269, 272, 281, 285, 287, 290, 296, 312-31, 323n, 330n, 340n, 348n, 354n, 356n intonation, 2, 135, 219, 222, 225, 227, 246, 253, 261, 264, 269, 276, 282-285, 291-292, 294, 298, 302-304, 309, 316, 347n, 349n intonational contour, 192, 222, 224, 228-231, 234, 246, 250, 252-254, 265-266, 294, 348n-350n, 353n intonational peak, 224, 228-229, 266, 285, 292, 302, 348n intonational phrase, 246, 266 language(s) agglutinative, 28-30, 32, 34, 46, 49, 65-66, 72-73, 90, 104, 109-110, 129, 183, 187191, 203, 261, 316, 334n ergative, 36-38, 59, 100-103, 127, 191, 328n, 338n Focus-prominent, 234, 243, 316 head-final, 11-12, 14, 46, 84, 135, 203, 208, 279, 347n head-initial, 11-12, 46, 204, 208, 347n

incorporating, 31, 34, 73, 104, 141, 279, 287, 289, 334n inflecting, 28, 30, 32, 34-35, 4546, 66, 73, 104, 108-109, 129, 181, 183, 185, 202, 259, 261, 289, 314, 316 isolating, 21, 23, 28, 32, 45-46, 90, 111-112, 147, 181, 185, 190, 204, 289 polysynthetic, 28-32, 48, 59, 65, 90, 110-112, 141, 187-188, 191, 197, 202, 259, 334n, 342n, 347n-348n predicate-initial, 77, 228-229, 264, 330n, 334n SOV, 13, 77, 111-112, 168, 181, 183, 227, 295, 351n, 355n Subject-prominent, 243, 261, 264, 324n SVO, 12, 32, 66, 72, 77, 111, 181, 227, 263-264, 324n Topic-prominent, 243, 261, 264, 316, 352n VSO, 12, 77, 111, 156, 181, 185, 264, 334n lexical insertion, 7-8, 15-16, 44, 89, 102, 129-130, 134, 151, 282, 311-312, 326n lexicon, 5, 7-8, 20-22, 32, 38, 41, 272, 311, 326n, 354n macro-role, 2, 46, 49, 51, 55-57, 59, 61-69, 72-73, 77, 82, 88, 91, 94, 96-97, 100, 102, 162, 206, 312-313, 330n, 333n, 336n, 348n ACTOR, 46, 50-51, 53-78, 80, 91, 93, 100-102, 104-107, 113, 126, 131-133, 136-141, 159-160, 164, 275, 305, 312-313, 330n-334n, 338n339n, 355n COMITATIVE, 62-64, 312 DURATION, 51, 57, 62

Language index 397 GOAL,

51, 57-58, 61-67, 80, 92, 94-98, 103, 131, 134, 136, 140, 275, 305, 307, 312313, 332n-333n, 336n337n LOCATIVE, 51, 56-57, 61-64, 78, 96, 134, 138, 198, 200, 312, 332n, 337n MEASURE, 63-64, 312 PATIENT, 50-51, 55-58, 60-65, 6770, 72-76, 81-82, 92-94, 96, 100-102, 104, 107, 114, 126, 134, 136-141, 160, 162, 164, 168, 206, 275, 305, 312-313, 330n, 332n334n, 337n-339n, 341n QUANTITY, 51, 57, 62-65, 312 THEME, 50-51, 53-57, 60, 62-66, 68-69, 73-80, 92, 98-99, 101, 105, 127, 140, 160, 312, 331n-332n, 335n modification, 181, 186, 273, 305 adjectival, 120, 149, 344n adverbial, 2, 177, 202, 314, 346n noun head (nominal), 120, 148, 166, 205, 326n morphology, 5, 19, 24, 27, 34, 37, 129, 150, 219, 236, 314, 316, 326n-327n, 330n, 339n, 354n derivational, 76, 158, 326n, 330n, 345n inflectional, 16-18, 147, 326n, 330n verbal, 44, 46, 61, 68, 75, 78, 165, 331n-332n nominalization, 2, 136, 138, 233, 341n-342n null subject, 99, 101, 106, 226, 243, 249, 261-263, 353n order, 8, 11-12, 14, 23, 26, 32, 35, 65, 72, 109, 111, 130-131, 151, 155-159, 178, 181-182, 185-

186, 190-191, 194, 198, 224225, 251, 272, 289-290, 313, 324n, 328n, 343n-344n, 351n, 353n constituent, 2, 32, 35, 43, 89-90, 111, 113, 117-118, 135, 147, 149, 151, 181, 200, 208, 222, 225, 230, 269, 276, 287, 289, 304, 312, 316, 338n, 342n, 347n linear, 7, 11-14, 25-26, 109, 112, 147, 151, 153, 157, 179, 187, 192, 197, 200, 247, 337n, 347n-348n parameter, 10-15, 28, 90, 116, 119, 135, 157, 316, 323n-324n, 345n, 350n part(s) of speech, 20-22, 325n passive, 37, 51, 57-58, 66-70, 72-76, 92, 97-98, 102-103, 111, 168, 332n-333n, 338n-339n predication, 2, 38, 43-46, 51, 54, 60, 63, 76, 79, 103, 130, 148, 178, 181, 223-224, 226-233, 242, 275, 280-281, 289, 312, 315, 325n, 328n, 331n, 333n, 340n, 350n, 356n presupposition, 2, 224, 230, 232-236, 238-241, 243, 283-284, 291, 298-300, 307, 315, 350n, 355n pronoun, 29, 54, 71, 81, 126, 131, 140, 169, 171-172, 191, 198, 206, 226, 238, 249, 301, 334n, 338n, 341n, 345n, 349n, 352n353n clitic, 26, 48-49, 59, 72-73, 111113, 115, 125-126, 139142, 234, 249, 251, 255, 258-261, 264, 279, 285, 291, 303, 328n, 332n, 334n, 340n, 342n, 350n-351n null, 101, 140, 169, 226, 336n, 342n

398 Subject index relative, 164, 170-172, 239, 299, 301, 345n-347n question, 225, 229-230, 233, 235236, 245, 267, 269, 272, 275, 282, 285-287, 296, 298-301, 307-310, 349n, 352n, 355n356n wh-, 235, 284, 287-288, 296, 301, 308, 350n-351n, 356n yes-no, 19, 222, 277, 284-285, 288-289, 298, 301, 308, 349n, 355n recursivity, 10, 17, 83, 315 relative, 161, 165-166, 168, 176, 238, 356n clause (see clause) pronoun (see pronoun) relativization, 163, 167-168, 345n semantic role, 38, 46, 62, 67, 97-98, 218 semantic root, 22, 24, 32, 61, 91, 131-132, 257, 311, 326n sentence (marked/unmarked), 19, 35, 195, 224-232, 263, 347n complex, 272, 275, 278, 283, 293295, 299, 301, 304 copular, 89, 230, 239, 242, 328n329n, 352n declarative, 221-223, 232, 269, 276, 278-279, 283, 289290, 293-295, 301, 308309, 329n interrogative, 126, 269, 284, 286, 288-289, 294, 297, 299, 301-302, 308, 355n simple, 294, 304 Small Clause, 44-45, 56, 174, 239242, 329n, 343n specificity, 145-146, 257 speech act, 181-182, 219-220, 223, 245-246, 269, 275-276, 281, 305-309, 356n

illocutionary act, 2, 220-221, 269, 274-276, 278, 283, 293296, 298, 302, 304-307, 309-310, 314-315, 352n assertion, 216, 220, 223, 242, 269-271, 279, 282-284, 294, 306, 315, 350n, 356n direct, 220-221, 353n exclamation, 269, 282 imperative, 220-221, 269, 276277, 283-284, 291-293, 302-303, 308, 350n, 353n-356n indirect, 220-221, 272, 307, 353n, 356n request for action, 269-270, 291, 309, 356n request for information, 234, 269, 285, 298, 315 locutionary act, 220, 269 perlocutionary act, 220 structure, 7-8, 11, 15, 43, 201, 226, 233, 261, 265, 304, 309, 312, 347n deep (D-structure), 7, 16, 32, 35, 53, 89-91, 96, 98, 101102, 104, 118, 125, 200, 311-312, 329n event, 2, 49, 51, 54, 60-61, 66, 69, 72-73, 75-77, 328n performative, 275, 281-282, 294, 301-305, 315 phrase, 2, 10, 17-18, 24, 28, 32, 43, 53, 62, 64, 88, 104, 129, 131, 230, 326n, 330n, 342n semantic, 45, 55, 76-77, 79, 261, 312, 315 sentence, 12, 25, 28, 32, 43, 80, 89, 91, 98-101, 130, 186, 188, 219, 239, 245, 281282, 290, 297, 304, 307,

Language index 399 314, 329n surface (S-structure), 7, 33, 8991, 98, 100, 102, 107, 162, 218, 261, 311-314, 316, 329n, 337n syntactic, 15, 36-37, 44, 57-59, 65-66, 76, 91, 136, 161, 213, 219, 227, 246, 281282, 313-315, 323n, 328n X', 9-10, 12-13, 15-17, 91, 104, 129-130, 133 ș-role (see also argument role), 27, 36-38, 46, 49-52, 54, 56, 6364, 66, 69-70, 76, 90-95, 98, 114, 125, 132, 163, 167, 234, 312, 325n, 330n-331n, 348n Topic, 2-3, 30, 141, 180, 193, 203, 215, 217-218, 223-224, 226229, 231-234, 243, 245-264, 266-267, 279, 283-284, 291292, 303-305, 315-316, 348n, 350n-353n, 356n aboutness-shift, 248, 250-254, 263-264, 267 contrastive, 253-254, 266-267, 292, 353n familiar, 249, 251-254, 267 given, 147, 213, 215, 223, 227, 232-233, 247-248, 250251, 350n marker, 29-30, 260, 353n

verb, 12-13, 17-18, 24-25, 32, 35, 37, 48-49, 55, 59, 63-65, 72-74, 77, 93, 104, 108-112, 115119, 123-127, 129, 139, 141, 185-187, 190-191, 194-195, 197, 203, 273, 314, 334n, 336n, 342n, 348n, 351n, 355n ditransitive, 67, 94-96, 117 intransitive, 37, 57, 74-77, 100, 102, 108, 127, 331n-332n, 338n mono-argument (also monovalent), 52-55, 60, 63-64, 73, 77, 80, 87, 98, 105, 349n performative, 3, 272, 274-276, 278, 282-283, 293, 304, 306-307, 315, 353n-356n two-argument (also bivalent), 50, 55, 57, 59-60, 62-64, 74, 77, 113, 141 three-argument (also trivalent), 9, 47, 51-52, 61-63, 67, 72 transitive, 37-38, 50, 55, 57, 68, 74-77, 93, 100, 103, 127, 331n-332n, 334n wh-constituent, 126, 208, 284, 286287, 289-290, 299-302, 315, 352n, 356n wh-question (see question) X'-theory, 8, 10, 312

valency, 52, 57, 78-79, 125, 330n, 341n

Language index

Afar, 11, 13-14, 83-84, 121-122, 169, 172, 204-205, 208, 241, 317, 345n Albanian, 26, 117, 145, 160, 317, 342n Arabic, 11-12, 19, 26, 30-31, 39, 44, 52-53, 56, 117, 119-120, 144, 150, 155-156, 159, 185-186, 239, 259, 270-271, 278, 317, 343n-344n Avarm 37, 100, 117, 127, 317 Bambara, 168-169, 317 Basque, 37, 101, 237, 296, 317, 351n, 356n Bemba, 40-41, 317 Berber, 107-108, 317 Breton, 26, 317, 326b Bulgarian, 26, 145, 147, 317 Cayuga, 32, 110, 317 Chaha, 237, 317 Chichewa, 48-49, 317, 352 Chinese, 23, 28, 34, 44, 52, 111, 117, 130, 147-148, 150, 155, 159, 181-182, 185, 261-263, 289290, 317, 327n, 341n, 356n Chukchi, 40-41, 317

243-244, 264-265, 271, 274, 278, 284-285, 287, 290, 296297, 307, 317, 324n, 326n, 330n-331n, 337n, 339n, 343n, 349n, 354n, 356n Ewe, 117, 317 Finnish, 183-184, 317, 347n French, 14, 39, 53-54, 98-99, 107109, 117, 123, 140, 150-154, 159, 167, 171, 177, 212, 226, 259, 318, 331n, 337n, 341n, 352n German, 14-15, 30-31, 39-40, 92, 9495, 97, 117, 121, 123, 133, 140, 159, 177-178, 183-184, 195, 197-198, 202, 264, 317, 324n, 337n, 344n, 346n-347n, 354n Greenlandic, 31-32, 318 Gungbe, 29, 260, 318 Gunwinjgu, 48, 318 Guyana creole, 29, 39, 190, 238, 318

Dutch, 211, 317 Dyirbal, 37, 101, 317

Haiti creole, 40, 318 Hebrew, 318, 352n, 354n Hindi, 175, 185, 318 Hungarian, 19-20, 52-53, 115, 204205, 211-212, 214-216, 237, 296, 300-301, 318, 352n, 356n

English, 11, 13-14, 19, 22-25, 33-34, 37, 44, 48, 53, 55, 58-60, 6667, 73, 83-85, 94-95, 99-100, 103, 105, 117-119, 123, 130, 135, 139, 144, 148-150, 155156, 158, 166, 169, 171-173, 177, 202, 208, 221-222, 236,

Isekiri, 190, 318 Italian, 11, 14, 24, 33, 35, 39, 52, 53, 55, 58, 67-70, 73, 75-76, 8081, 83, 85, 98-99, 106, 109, 114, 117-118, 123, 125-126, 134-135, 137, 140-141, 143, 146, 149-153, 158, 160, 167,

402 Language index 170-171, 174-177, 179, 192193, 195, 198-200, 202, 208209, 211-212, 216-217, 221222, 226, 229, 233, 236, 243244, 247, 255, 259, 261-265, 270, 274, 276, 286-287, 290291, 302-303, 318, 326n, 328n, 330n-334n, 336n-337n, 340n-342n, 344n-346n, 349n350n, 354n, 356n Japanese, 11, 119, 135, 160, 173, 176, 318, 353n, 356n Kashmiri, 318, 336 Kikuyu, 241, 318 Kistane, 272-273, 295, 318 Korean, 13, 30, 109-110, 112, 120, 191, 260, 295, 318, 333n, 355n Krio, 190, 238, 318 Lakhota, 124, 142, 318 Latin, 27, 60, 132, 175, 277, 318 Macushi, 191, 318 Malagasy, 12-13, 36-37, 228, 264, 318 Maǀri, 204, 208-210, 318 Mayali, 27, 318, 327n Modern Greek, 19-20, 108-109, 112, 117, 121, 140, 181-182, 185, 318, 333n Mohawk, 141-142, 197, 318, 348n Mojave, 40-41, 318 Ngalakan, 110, 318 Oneida, 31, 318 Persian, 115, 169-171, 318, 345n Polish, 147-148, 204, 208, 210-212, 214, 216-217, 319

Quechua, 94, 319 Rapanui, 111-112, 319 Rembarunga, 48, 204-205, 319 Romanian, 126, 319 Russian, 12, 30-31, 34-35, 39, 44, 5657, 108-109, 112-114, 117, 119, 121, 164, 170-172, 319, 329n, 331n, 341n, 354n Saramaccan, 29, 111, 319 Somali, 26, 44, 52, 55, 73-77, 85, 87, 112, 117-118, 121, 132, 141142, 144-145, 149, 165-166, 169, 173-174, 202, 207, 214215, 221-222, 238, 240-241, 259, 270-272, 274, 276, 280281, 285, 287-290, 292, 295297, 300-301, 391, 334n-335n, 342n, 345n, 349n, 352n, 355n, 356n Southern Tiwa, 111, 319 Spanish, 114, 123, 126, 276, 279, 307, 319, 331n, 340n, 341n, 346n Swahili, 30, 49, 72-73, 77, 110, 115117, 319, 334n Tagalog, 26, 44, 52-53, 60-61, 77-78, 93, 117, 120, 133, 135, 156, 174-175, 185-186, 204-205, 208, 221-222, 228-229, 239, 260, 264, 319, 328n-331n, 334n, 341n, 349n Tigrinya, 115, 237, 319 Toba Batak, 167-168, 280, 319 Turkish, 11, 13, 30, 83-85, 109-110, 120-122, 130-132, 135, 203204, 208-210, 319, 349n, 356n Tzutujil, 124, 142, 319 Uto-Aztecan, 124, 319

Language index 403 Warì, 29, 279, 319 Warlpiri, 36, 59, 117, 126-127, 141, 259, 319 Welsh, 167, 319 Western Mupun, 280, 319

Wolof, 237, 319 Yareba, 187-188, 319 Yidiny, 103, 319