Categorial Grammar and Word-Formation: The De-adjectival Abstract Noun in English 9783110943207, 9783484421226

Table of contents :
Contents
Preface
Abbreviations
I . Introduction
2. Methods of Analysis
3. Categorial Grammars and Formal Semantics
4. Word-Formation within a Categorial Framework
5- The De-adjectival Abstract
6. Conclusion
Bibliography

Citation preview

BUCHREIHE DER ANGLIA ZEITSCHRIFT FÜR E N G L I S C H E P H I L O L O G I E

Herausgegeben von Helmut Gneuss, Hans Käsmann, Erwin Wolff und Theodor Wolpers

22. Band

KARL REICHL

Categorial Grammar and Word-Formation : The De-adjectival Abstract Noun in English

MAX N I E M E Y E R V E R L A G T Ü B I N G E N 1982

Als Habilitationsschrift auf Empfehlung des Fachbereichs Sprach- und Literaturwissenschaft I der Universität München gedruckt mit Unterstützung der Deutschen Forschungsgemeinschaft.

CIP-Kurztitelaufnahme der Deutschen Bibliothek Reichl, Karl: Categorial grammar and word-formation: the de-adjectival abstract noun in English / Karl Reichl. - Tübingen : Niemeyer, 1982. (Buchreihe der Anglia, Zeitschrift für englische Philologie ; Bd. 22) NE: GT I S B N 3-484-42122-3

I S S N 0340-5435

© Max Niemeyer Verlag Tübingen 1982 Alle Rechte vorbehalten. Ohne ausdrückliche Genehmigung des Verlags ist es auch nicht gestattet, dieses Buch oder Teile daraus auf photomechanischem Wege zu vervielfältigen. Printed in Germany. Satz und Druck: Allgäuer Zeitungsverlag G m b H , Kempten. Bindearbeiten: Heinrich Koch, Tübingen.

Contents

Preface

VII

Abbreviations

IX

1 . Introduction

i

2. Methods of Analysis

6

2.1 Derivation as a Morphological Process 2.1.1 Structuralist Analyses 2. i .2 Formalization

8 8 14

2.2 Derivation and Syntax 2.2.1 Non-transformational Analyses 2.2.2 The Transformationalist Position

21 21 23

2.3 Derivation and the Lexicon 2.3.1 Pre-lexical'Semantico-Syntax' 2.3.2 The Lexicalist Position

27 29 33

3. Categorial Grammars and Formal Semantics

41

3.1 Categorial Grammars 3.1.1 The'Standard Model' 3.1.2 Applicative Grammar

41 41 54

3.2 Formal Semantics 3.2.1 Montague Grammar 3.2.2 Variants and Extensions

60 61 76

4. W o r d - F o r m a t i o n within a Categorial F r a m e w o r k

89

4.1 Functional Structure 4.1.1 The Functional Principle 4.1.2 Syntactical, Lexical and Grammatical Categories 4.1.3 Deep Structure Categorial Grammar 4.1.4 The Categorization of the Adjective

89 89 92 95 103

4.2 Syntax, Semantics and the Lexicon 4.2.1 Syntax 4.2.2 Word-Formation 4.2.3 Semantics and the Lexicon

113 113 130 147

V

5- T h e De-adjectival Abstract 5.1 Derivational Morphology 5.1.1 Segmentation and Derivational Types 5.1.2 Suffixation 5.2 Syntax and Semantics 5.2.1 Problems of Quantification 5.2.2 Abstraction and Nominalization 5.2.3 Abstract Types and Lexicalization

157 157 158 174 189 190 199 221

6. Conclusion

236

Bibliography

240

VI

Preface

This study of the de-adjectival abstract noun in English is an attempt to apply categorial grammar in its various forms to a well-defined body of empirical data. This application of a theoretical framework, first formulated and predominantly explored in the field of syntax and propositional semantics, to a range of phenomena in the field of derivational morphology has led to a revision of both surface structure oriented and deep structure oriented categorial grammars. Given the restrictions of the topics to be discussed in this book, these revisions must be of a tentative nature. Further modifications might be called for when other areas of linguistic structure are investigated more thoroughly. Nevertheless, it is to be hoped that this study will contribute to a better understanding of the possibilities of a categorial grammar. As a familiarity with the various versions of categorial grammar cannot be assumed in all readers interested in derivational phenomena in English, a rather large part of this book is devoted both to an exposition of the principles of categorial grammar and a discussion of the theoretical model adopted here. It is hoped that this exposition and discussion will prove helpful for the evaluation of the analyses and interpretations proposed in Chapter Five. This book is the revised English version of my German Habilitationsschrift, submitted to the Fachbereich Sprach- und Literaturwissenschaft I of the University of Munich. I am grateful to the Deutsche Forscbungsgemeinscbaft for a generous research grant. I am furthermore grateful to Professor Dr. Helmut Gneuss, who has been most encouraging during my years of studying and teaching at the English Department of the University of Munich, and to Professors Dr. Hans Kasmann, Dr. Leonhard Lipka and Dr. Theo Vennemann for their helpful criticism of this study, as well as to Dr. Raymond Hickey for correcting the English version. Any shortcomings in this book, however, remain my responsibility. I would also like to thank Professor Dr. Winfried Lenders and Dr. Gerd Willee of the Department for Linguistic Computer Research in the Institutfiir Kommunikationsforscbung und Phonetik at the University of Bonn for their help with the Brown University Corpus. Finally, my VII

thanks are also due to the editors of the Buchreihe der Anglia and to the publisher for kindly accepting my study. It is customary to end the preface by thanking one's wife for having typed the manuscript. Having typed (and re-typed) the manuscript myself, I would prefer to thank my wife and children for having prevented me from becoming too engrossed in my scholarly work and thus missing out on all the other joys of life. Bonn

VIII

January, 1981

Abbreviations

AL: ALD: BSL: Chambers:

CLS: CSL: EWF: EP h: FL: GCE: HEW: HWPh: IBAL: IF: IULC: JL: JL: L: MED: MEG: MSSE: NHLS: OED:

REW:

Acta Linguistica Hafniensia Hornby, A.S., A.P.Cowie, eds. (1980), Oxford Advanced Learner's Dictionary of Current English, third rev. ed., Oxford. Bulletin de la Société de Linguistique de Paris Macdonald, A.M., ed. (1973), Chambers Twentieth Century Dictionary, New Edition, rev. ed., Edinburgh [edition used for the compilation of the corpus]. Papers from the... th Regional Meeting Chicago Linguistic Society Cambridge Studies in Linguistics Marchand, H. (1969), The Categories and Types of Present-Day English Word-Formation, second ed., München. Edwards, P., ed. (1967), Encyclopaedia of Philosophy, 8 vols., New York. Foundations of Language Quirk,R., S.Greenbaum, G.Leech, J.Svartvik (1973), A Grammar of Contemporary English, second impr. with corr., London. Koziol, H. (1972), Handbuch der englischen Wortbildungslehre, second ed., Heidelberg. Ritter, J., K.Gründer, eds. (1971- ), Historisches Wörterbuch der Philosophie, Basel, Darmstadt. Internationale Bibliothek für allgemeine Linguistik Indogermanische Forschungen Indiana University Linguistics Club Janua Linguarum Journal of Linguistics Language Kurath, H., S.M. Kuhn, J . Reidy, eds. (1952- ), Middle English Dictionary, Ann Arbor, Mich. Jespersen, O. (1901-49), A Modern English Grammar on Historical Principles, 7 vols., Heidelberg, Copenhagen. Stockwell, R.P., P.Schachter, B.H.Partee (1973), The Major Syntactic Structures of English, New York. North-Holland Linguistic Series Murray, J . A . H . , H.Bradley, W.A. Craigie, C.T. Onions, eds. (1933), The Oxford English Dictionary, 13 vols., Oxford. Burchfield, R.W. (1972- ), A Supplement to the Oxford English Dictionary, Oxford. Meyer-Lübke, W., ed. (1935), Romanisches etymologisches Wörterbuch, third ed., Heidelberg. IX

RWEG:

Lehnert, M., ed. (1971), Rückläufiges

Wörterbuch der englischen

Gegen-

wartssprache, Leipzig. SOED:

Little, W., C . T . Onions, eds. (1973), The Shorter Oxford English Dictionary on Historical Principles, third rev. ed. with etymologies rev. by G . W . S . Friedrichsen and with rev. addenda, Oxford.

TBL:

X

Tübinger Beiträge zur Linguistik

I.

Introduction

This book deals with a topic which at first sight might seem rather specialized. It sets out to analyse a fairly restricted class of linguistic phenomena, the de-adjectival abstract nouns in English. Even a quite superficial characterization of de-adjectival abstracts, however, reveals that the questions to be raised and discussed in the course of this study are both of a ramified and rather complex nature, intimately connected to some of the most fundamental problems of linguistic analysis. If we look at a de-adjectival abstract like goodness and contrast it with other 'nominal forms' of the adjective good as for instance in (i) to (5), some of the most salient features of the adjective-derived abstract noun become apparent. (1) (2) (3) (4) (5)

His being so good surprised me. He denied the bad results and admitted only the good ones. The good must not be treated badly. He moved his goods somewhere else. He certainly is a goody.

Without going into details at this stage, we might say that the main difference between goodness, goods in (4) and goody in (5) on the one hand and the nominalizations of good in (1) and (2) is that in the derivation of the former words a change of the lexical class of good is involved, while in (1) and (2) good is still an adjective. The so-called partial conversion of good in (3) is a border-line case, one of many. It is classified as a conversion, because there is no overt mark for the change of lexical class the adjective good has undergone in (3), as distinct from e.g. suffixation like in goodness or goody; but this conversion has been effected only in part: in (3) good retains some of its adjectival qualities such as modification by adverbs (cf. the very good etc.).1 As to the distinction 1

For the view that partially converted adjectives are adjectives functioning as head of a noun phrase (and that thus no change of lexical class has occurred) see GCE: 251-3, I

between the nominalization of good in goodness on the one hand and in goods or goody on the other, it is obviously of a semantic kind; goodness denotes an abstract object, while goods and goody do not. There are more syntactic and semantic differences (as well as morphological ones) between the various nominalizations of good; furthermore those mentioned are not yet stated in sufficient detail or with enough precision. Terms like 'lexical class', 'denote', 'abstract object' and, of course, 'nominalization' present a great number of problems and suggest the wider context in which this study is placed: the theory of the parts of speech, the lexicalist-transformationalist controversy, the relationship between propositional and lexical semantics and, from a philosophical point of view, the problem of universals, to name only the most important. Although concentrating on de-adjectival abstract nouns, a number of points which are of a more general interest will thus have to be discussed in the following chapters. The derivation of an abstract noun from an adjective was seen as a predominantly philosophical problem in traditional linguistic theory. If nouns denote substances, more specifically 'substances with quality' as Priscian and Donat had taught, what kind of substances do abstracts like albedo, "whiteness", denote, the medieval grammarians asked. D o they only signify the 'form' of substances or merely 'qualities' inherent in substances? Petrus Helias, a twelfth century grammarian, brings a lengthy and tortuous discussion of this problem in his Summa on Priscian, and the definition of abstracts continued to intrigue grammarians and logicians alike throughout the Middle Ages. 2 The analysis of abstracts as words standing for 'abstract ideas' is intricately bound up with the theory of knowledge in rationalist and empiricist philosophy, and this aspect is also reflected in philosophically oriented grammatical treatises, as for instance in James Harris' Hermes: Agen, by a more refin'd operation of our Mind alone, we abstract any Attribute from its necessary subject, and consider it apart devoid of its dependence. For example, from Body we abstract to Fly; from Surface, the being White; from Soul, the being Temperate. A N D thus 'tis we convert even Attributes into Substances, denoting them on this occasion by proper Substantives, such as Flight, Whiteness, Temper-

2

I O I O - I ; for the interpretation of partially converted adjectives as class abstractions cf. below 5.2.2. The full conversion of adjectives is treated in Bergener (1928); see also Biese (1941) and EWF: 3 59 ff. Cf. De Rijk (1962-7: II, 221 ff.), Reichl (1976: i6 7 ff.). 2

ance; or else by others more general, such as Motion, Colour, Virtue. These we call A B S T R A C T S U B S T A N C E S ; the second sort we call A R T I F I C I A L . (Harris, 1 7 5 1 : 3 7 - 8 )

Henry Sweet, at the end of the 19th century, still used, in his New English Grammar, the distinction between a substance as something concrete and an attribute as something abstract to define abstract nouns (cf. Sweet, 1892-8: 1,12,61-2). In post-Saussurean modern linguistics, however, the tendency towards a formal approach to language has led to a dissociation of the various questions the analysis of abstracts poses. With the definition of word classes by non-semantic criteria, as exemplified in the works of C . C . Fries and Z.S. Harris for instance, the problem of what sort of entities abstract nouns denote ceased to puzzle linguists. Abstract nouns are now merely a sub-class of nouns, with whose morphological and syntactical behaviour alone a grammar has to cope. It is only with the increasing influence of logic and analytic philosophy on linguistics that the semantic aspect of abstracts has again come into the range of grammatical theory, where it can now be formulated with more precision than in traditional treatments of language. It is in this vein that the following study attempts to consider the morphological, syntactic and semantic characteristics of the abstract noun in English, more precisely of the de-adjectival abstract. The limitation to de-adjectival nouns was necessitated by reasons of length; many of the observations made apply also to deverbal abstracts, however. It was with a view to analysing abstracts on the various linguistic levels that a categorial framework was adopted. Categorial grammars were first formalized in the fifties and have received new attention through the development of formal semantics, notably in 'Montague grammar'. While the earlier versions of categorial grammar-especially in the form of an 'applicative grammar'-are particularly suitable for the analysis of the morphological structure of derived lexemes, the model-theoretic categorial grammars advocated by R. Montague, D. Lewis, M. Cresswell and others provide a precise instrument for the description of their semantic structure. It will be one of the concerns of this book to show how some of the insights of a philosophically oriented categorial grammar can be fruitfully integrated into a linguistic theory with a categorially based syntax. Before, however, explaining the basic notions of categorial grammar and formal semantics in Chapter Three, the various possibilities of 3

analysing de-adjectival abstracts-and word-formation processes in general-according to the most widely accepted grammatical models are discussed in Chapter Two. In Chapter Four the theoretical framework adopted in this book is sketched, while in Chapter Five this framework is applied to the empirical data collected. This study is based on a corpus which was compiled from Chambers Twentieth Century Dictionary, supplemented by the Shorter Oxford English Dictionary and M. Lehnert's Rückläufiges Wörterbuch der englischen Gegenwartssprache. On approximately 3200 index cards the deadjectival abstracts have been arranged in 'word-families' (more properly 'derivational nests', cf. 3.1.2), that is to say, that together with the abstract its base adjective is listed; if more than one abstract is derived from the base adjective, these are also recorded; if the base adjective is itself derived its base and all adjectival derivations from this base with their de-adjectival abstracts are also listed; this process is repeated for all derived bases (cf. J.I for the application of this information). In addition to this the Brown University Corpus has been used for frequency counts and the nouns ending in -ness and -ity contained in that corpus have been analysed as to their syntactic environments.3 A last point to be made in this introductory chapter concerns some technical matters of definition and typographical convention. The terms 'grammar' and 'derivation' are both used in their wider sense, unless otherwise indicated; this means that by 'grammar' not only the morphology and syntax of a language are understood, but the systematic description of the language structure as a whole. Likewise the term 'derivation' (and analogously 'derivational') covers all kinds of wordformation processes, including composition. For the purposes of this study the terms 'word', 'lexeme' and 'lexical item' are treated as being interchangeable (for details see 2.1.2); furthermore, no distinction is drawn between a 'lexical class' and a 'word class'. When there are two lexical items A and B, and B is derived from A, A is called the 'base' or 'base-form' and B the 'derived form'. The terms 'free form', 'bound form', 'unique form' and 'morpheme' are understood as defined by L. Bloomfield (cf. 2.1.1), although his principles of morphological analysis will have to be modified in the course of this study (cf. 5.1). The definition of 'stem', however, is not taken over from Bloomfield; this term 3

The Standard Corpus of Present-Day Edited American English contains 500 texts of approximately 2000 words each from 1 j areas of writing, ranging from various journalistic styles to different forms of fiction. This corpus was compiled at Brown University during 1963 and 1964 under the direction of W. Nelson Francis.

4

will be used only occasionally, designating the form of a word which is not overtly marked for some morphological category (cf. 2.1.2). As general terms for linguistic units of various kinds-excluding, however, semantic units-'expression' and 'form' or 'linguistic form' are used. Phonetic transcriptions are put between square brackets, graphemes between angular brackets-< > - a n d phonemic transcriptions between slanting lines; it is usually the latter that are used in the present context. Morphemes and allomorphs are here not normally marked by braces. In rules parentheses symbolize facultative linguistic forms, braces alternative forms; an arrow means that the expression to the left of the arrow is to be rewritten as the expression to the right of the arrow, a simple arrow standing for a phrase-structure expansion, a double arrow for a transformation. A simple arrow is also used to mark the derivational relationship between lexical items, the lexeme to the left of the arrow being the base for the form to the right of the arrow. Graphs for the symbolizing of derivational 'nests' are introduced and explained in 3.1.2. For diachronic developments a special arrow is used; A > B means " A developed into B " , A < B means " A developed from B " . Unacceptable linguistic forms are preceded by an asterisk. The meaning of a linguistic expression is put within double quotation-marks; features-both semantic features and others-are enclosed by square brackets; predicates are spelled with capitals (cf. 2.3, 4.2.3). There are a number of special symbols which will be introduced and explained in the course of this study; the most important are the symbols of logic, the brackets used in set-theory (pointed brackets for ordered n-tuples, braces for sets; cf. 3.1.x, 3.2.1), and the various symbolizations of the categorial formalism (cf. 3).

5

2. Methods of Analysis

Before taking a closer look at the structure of the de-adjectival abstract noun in English, it will be helpful for later discussions to consider the issues involved in somewhat broader terms. The analysis of nouns like goodness or stupidity is traditionally seen as falling within the field of word-formation. To study word-formation is to study the linguistic devices a language possesses to build words (as opposed to borrowing words from other languages or dialects). These devices might be of various kinds, the most common being composition, affixation and 'modification'; an example of the former would be darkroom from dark and room, of the latter two length from long, where the addition of the suffix- th (affixation) is accompanied by a qualitative vowel-alternation (modification).1 The rules of word-formation form, according to this view, a part of their own in the description of language, complementing the lexicon, which indicates the semantic structure of words, by specifying their morphological structure. This classification of word-formation has not remained unchallenged-witness F. de Saussure's modification of the threefold distinction morphology-syntax-lexicology according to his differentiation between associative or paradigmatic and syntagmatic relations among linguistic units (cf. Saussure, 1967-71: II, f. 302 v -308 v )-, but by and large it is typical of what might be called traditional grammatical theory. The place of word-formation within a grammatical model rests on at least three preconditions: the definition of what is to count as a derivational process, the structure of the language under investigation, and the linguistic theory one subscribes to. Crucial for the definition of a derivational process is the point of view adopted. Linguist A might be studying

1

For a more detailed distinction of morphological processes-differentiating between affixation, reduplication and modification (vowel change, suppletion, subtraction)-see Matthews (1974: n 6 f f . ) ; cf. also below 5 . 1 . 1 .

6

the morphological structure of a given lexicon, taking the lexicon as a finite list of recorded lexical items and basing his analysis on the segmentability of the items; linguist B, while using the same lexicon and following the same technical procedure, might interpret this lexicon as an open class, which can be supplemented by new, unrecorded items. For B, but not for A, derivational rules will have a predictive force and B will have to take phenomena like productivity or frequency of occurrence into account when formulating his rules. Thus A and B might differ in their analysis of Modern English length, though their analysis of Old English lengdu, where -du can be regarded as a productive suffix, would presumably be identical.2 It is often asserted that grammar, meaning in the narrow sense of the word syntax and morphology, deals with the regular formations of language and the lexicon with the irregular ones. In this case derivational rules could be either part of grammar or of the lexicon, depending on the regularity of the derivational structure of the language under investigation. Languages with a 'mixed lexicon' like Modern English (cf. J . I . I ) normally exhibit idiosyncrasies and irregularities which might call for a 'lexical' treatment of word-formation, while it is typical of agglutinating languages, for instance, that they are derivationally transparent and regular, a fact which might speak in favour of an incorporation of derivational rules into the morphological component of grammar. By the same token, however, it is arguable that morphological processes marking grammatical categories like tense, mood etc. should, when irregular, be treated in the lexicon, an argument that assigns to the criterion of regularity for the classification of word-formation within a linguistic model a rather dubious value.3 The most decisive factor for the position that word-formation rules are given in a grammar is of course the theoretical framework assumed. Derivational rules will have a different status in a grammar that postulates a word level of analysis than in a linguistic model where an independent word level is not recognized; similarly derivational processes will be viewed differently in grammars where the emphasis is on syntax than in grammars where the emphasis is on semantics, differently also in linguistic descriptions where the orientation is toward 'surface 2

3

Compare die distinction between analysis and synthesis in L y o n s (1968: 1 j 8 ff.) and in Lipka ( 1 9 7 1 b : 222ff.). This is also realized by Sweet, w h o uses this distinction between regular and irregular processes to define the domain of grammar and lexicon, respectively; cf. Sweet (1892-8: I, 7 - 1 0 ) . C f . also Levkovskaja (1952).

7

structure' than in those where one operates with more 'abstract' underlying structures. There are roughly three main approaches to derivational processes in modern linguistic theory, (i) One might argue from a predominantly morphological point of view that goodness for instance is a noun derived from an adjective good by means of the suffix -ness just as friends is derived from friend by means of the suffix -s. (2) There are also arguments in favour of a syntactic view of derivation, in this case stressing the relationship between goodness in John's goodness with good in John is good. (3) This relationship can be interpreted not only-and not even primarily-as a syntactic one, but rather as a semantic relationship between lexical items, which has to be analysed on the 'sub-lexemic' level. We will consider these positions in turn before giving a fourth alternative, the treatment of derivation within a categorial framework. 4

2.1 Derivation as a Morphological Process 2.1.1

Structuralist Analyses

A 'classical' formulation of the first approach is found in Leonard Bloomfield's Language. Bloomfield presents the grammatical structure of a language in terms of a hierarchy of linguistic levels: at the bottom lies the phonological level, then come the morphème level, the word level, the level of syntactic constructions, and finally the sentence level. The immediate constituents of the sentence-which is defined as "an independent linguistic form, not included by virtue of any grammatical construction in any larger form"-are syntactic constructions, which, depending on the complexity of the sentence, might have other syntactic constructions as immediate constituents. 5 When the immediate constituents of a construction are what Bloomfield calls 'minimum free forms' the word level is attained. Words can be further analysed if they show

4

5

For a general discussion of the approaches to derivational morphology see Pennanen (1972), Serebrennikov (1970-3: II, 344-393), (1973-6: II, 284-320), Lloyd (1964) [Romance languages], Stepanova (1968), (1973) [German], Erben (1975) [German]; see also DeArmond (1969) for the morphological approach to word-formation. An introduction to English word-formation is Adams (1973); standard reference books are MEG: VI, HEW and especially EWF. For bibliographical references cf. Seymour (1968) [Germanic languages], Stein (1973) [English], Bloomfield (1935: 170); cf. also Lyons (1968: 172-180). For a discussion of immediate constituent analysis see Wells (1947).

8

partial phonetic-semantic resemblances to other words, as e.g. birds to girls, both having the same element -5 meaning roughly "several". A morpheme is then defined as "a linguistic form which bears no partial phonetic-semantic resemblance to any other form", as is the case with bird when compared to girl.6 With the morpheme level the ultimate constituents of a sentence are reached. Outside the bounds of sentence analysis lies the phonological level, the level of linguistic units that differentiate meanings but are in themselves meaningless. The phoneme is the domain of phonology, words and morphemes fall within morphology, sentence types and syntactic constructions are dealt with in syntax. Word-formation is in this model part of morphology: B y the

morphology

of a language we mean the constructions in which

bound forms appear among the constituents. B y definition, the resultant forms are either bound forms or words, but never phrases. Accordingly, w e may say that morphology includes the constructions of words and parts of words, while syntax includes the constructions of phrases. (Bloomfield, 1 9 3 5 : 207)

Bloomfield distinguishes on the morpheme level between free forms {room in rooms), bound forms (-s in rooms) and unique forms (cran- in cranberry), on the word level between primary words and secondary words.7 Primary words are free forms which have no free forms as constituents. There are two kinds of primary words, derived primary words, which contain more than one bound form, and morphemewords, which consist of only one free morpheme. An example of the first kind is receive {re- plus -ceive), of the second man. Secondary words are free forms which have free forms as constituents. Here too there are two categories: compound words, which are made up of free forms {door and knob in door-knob), and derived secondary words with only one free form among their constituents {boy in boyish). This classification entails three types of derivational processes: composition (yielding compounds), secondary derivation (yielding derived secondary words) and primary derivation (yielding derived primary words). It is important to realize that for Bloomfield the class of derived words comprises words like boyish and words like glasses, i.e. that by

6

7

Bloomfield (1935: 161); compare also Lyons (1968: 180-94), Matthews (1974: 77-95), Bierwisch (1962). C f . Bloomfield (1935: 160, 209ff., 237ff.). Cf. also the classification into simple stems, secondary derived stems and primary derived stems in Hockett (1958: 240).

9

the term derivation both word-formation and inflexion are covered. The distinction between the two is one of degree: while inflectional affixes constitute the outer morpheme ranks of a word, derivational affixes constitute its inner ranks.8 There are further distinctions which Bloomfield drew in his analysis. If a morpheme has different forms depending on whether it occurs as a free form or as the base of a derived form, he differentiates between a kernel (as German) and a stem (as Germano- in Germanophobia). Primary words like hammer can be split into two elements by analogy to secondary words like leader; Bloomfield calls hamm- in hammer a root, -er a primary affix. Words that do not admit of this analysis, as for instance boy, are termed primary root-words. From this follows that there are free and bound roots (boy vs. hamm-). Secondary words derived by means of a zero-affix (as the verb man from the noun man), whose underlying free form is a root, are classified as secondary rootwords. Secondary words can also be formed from phrases; Bloomfield's example is old-maidish from old maid. If this derivation is not formally marked, as in jack-in-the-pulpit, the derived forms are phrase-words, cases which lie on the border between morphology and syntax: A s a border region w e have phrase-words ( j a c k - i n - t h e - p u l p i t ) and some compound words ( b l a c k b i r d ) , which contain no bound forms among their immediate constituents, and yet in some ways exhibit morphologic rather than syntactic types of construction. (Bloomfield, 1 9 3 5 : 207)

It is at this point that morphology and syntax meet. Bloomfield makes this clear in his discussion of compounds (cf. Bloomfield, 1935: 227-37). In principle, however, he restricts derivational processes to the morphological level of grammar and stays within the bounds of a 'morphological model' of word-formation. This model is characterized on the one hand by the assumption of a hierarchy of linguistic levels, among them a word level, which, although part of a higher level, can be studied by itself. It is immaterial in this context whether the word is defined, with Bloomfield, syntactically as a minimum free form, stressing the independence and isolability of the word, or whether semantic or formal criteria are employed in its defini-

8

C f . Bloomfield (1935: 222). Compare also Jespersen's treatment of derivational and inflectional morphology in MEG: VI. The interdependence of the two is stressed in Stankiewicz (1962).

10

tion. 9 Secondly, it is typical of this model that words are treated as linguistic units which can be broken down into constituents by purely formal procedures, analytic procedures that were formulated in great detail by E. Nida in his book on morphology. 1 0 In contemporary linguistic theory Bloomfield's conception of language as a hierarchy of linguistic levels with a clearly circumscribed word level is best represented by tagmemic and by stratificational grammar. K . Pike's tagmemic theory systematically extends the notion of the function of a linguistic unit as Bloomfield defined it to all linguistic levels." A linguistic form cannot be analysed correctly, so tagmemic grammarians argue, if it is segmented only into formal or only into functional elements. T o take an example: The sentence John ran is neither a concatenation of word classes-noun + verb-nor just a sequence of functions-subject + predicate-, but both (for the notion of a function cf. below 3.1.1, 4.1.1). It consists of two pairs, tagmemes, each of which is the correlation of a functional element or slot to a formal element or filler; the sentence can thus be represented by the formula S : N + P:V, i.e. a subject position filled by a noun plus a predicate position filled by a verb (cf. C o o k , 1969; Elson, Pickett, 1965: 57-8; Pike, Pike, 1977: 3*ff-). O n the word level, where derivation is treated on a par with inflexion, words are also segmented into tagmemes. Thus the abstract noun goodness is in tagmemic terms a noun stem (ns), consisting of the slot 'core', which is filled by an adjective stem (ajs), and the slot 'nominalizer' (nom), which is filled by one of a set of nominalizing morphemes (-ness): ns = core:ajs + nom: {-ness} (cf. C o o k , 1969: 116-39; Elson, Pickett, 1965: 75—81). Similar to this is the position word-formation has in the stratificational model developed by S. Lamb. There language is viewed as a hierarchy of stratal systems, where syntactic processes are specified by the 'lexotactics' and morphological phenomena-of an inflectional and derivational kind-by the 'morphotactics':

' For the various criteria of definition see Kramsky (1969: 17-40) and the critical bibliography by Juilland, Roceric (1972); cf. also Lyons (1968: 202-4). 10 Cf. Nida (1949: 6-61); cf. also Harris (1951: 5-6, 1 j-16) for the distributional criteria of morphemic analysis. 11 The different use of the term 'tagmeme' by Bloomfield and the tagmemic grammarians is discussed in Pike (1958). For tagmemic theory see Pike (1967), Elson, Pickett (1965), C o o k (1969), Pike, Pike (1977). A 'formans model' of word-formation influenced by Pike's tagmemics is advocated in Pilch (1968).

II

T h e morphotactics

deals with the specifically morphological tactic features of

the language, accounting for inflectional and productive derivational patterns. It has the grammatical w o r d as its basic domain. 1 2

An example will illustrate this. If we take two linguistic units A and B, say the words stone and wall, these two elements can be in what Lamb calls an And-relationship or in an Or-relationship. Stone and wall stand in an And-relationship (or syntagmatic relationship) in stone wall, i.e. they are linearly ordered, they stand in an Or-relationship (or paradigmatic relationship) in The ( ) is big, where either stone or wall can be inserted for ( ); These relationships can be ordered or unordered; from roof and garden both roof-garden and garden-roof is formable (unordered And-relation), while from good and -ness only goodness, but not *nessgood is derivable (ordered And-relation). Finally, a relationship can be upward or downward; it is downward when an element is realized by another element, it is upward when an element realizes another element. The relationship between stone, stone wall and stony is downward when looking at the way stone wall and stony are segmentable into stone plus some other element, it is upward when looking at the way stone can become part of different words, like stone wall and stony. The symbols for these relationships are as follows (Lamb, 1966: 9):

Ordered

Unordered

(0

Or

And

And

Or

Downward

Upward

12

A vV :V

e

f

n/

\o

J\

L o c k w o o d (1972: 1 1 4 ) ; cf. Lamb (1966: 21). Tagmemic theory and stratificational grammar are compared in Lockwood (1972: 254-7); for improvements on the stratificational model see L o c k w o o d (1972: 120-6).

12

If we now take the nouns clearness and evenness, their structure could be reflected in the following (simplified) diagram:'3

That is to say, the class of adjectives consists of two ordered elements (middle triangle), a prefix, which is either un- or zero, but not both together (left downward brace), and some base adjective, as clear or even (middle downward brace). These adjectives-so far clear, even, unclear and uneven-can either stand by themselves as adjectives or be part of a noun (middle upward brace). The class of nouns consists in the present example of two ordered elements, an adjective and a suffix -ness (right triangle), and comprises according to (2) clearness, evenness, unclearness and unevenness. In spite of the interconnections between the various stratal systems, every stratum can be analysed by itself. This means that the derivational structure of a language, though related to syntactical processes via the stratal network, is seen as a predominantly morphological process.'4 The same can be said of M. A. K. Halliday's systemic grammar, where there is a similar system of interrelations between various linguistic levels, but where nevertheless the analysis of grammatical and derivational morphology is mainly confined to the level of grammatical form. 15 As a last example for this kind of analysis one could mention Z. S. Harris' procedure in his Methods in Structural Linguistics. Harris, who works from the phoneme upward to the sentence, represents the latter as I}

14

15

In the following diagram the linguistic units are left unspecified; for the terms 'morphons', 'morphemes', 'lexons', 'lexemes' as used by stratificational grammarians see Lamb (1966: 1 8 - 2 1 ) , Lockwood (1972: 14-25). See also the comparison of stratificational and transformational grammars in Lockwood (1972: 263-270). C f . Berry (1975-7: I, 68-9). For the linguistic levels postulated by systemic grammar see Halliday (1961).

13

a sequence of morpheme classes (and single morphemes, if the class consists of only one element), listing among his morpheme classes also classes of derivational morphemes. Hence a word like goodness is seen by Harris as an instance of the morpheme class sequence consisting of A (adjective) and An (class of suffixes which, when combined with adjectives, yield nouns) (cf. Harris, 1 9 5 1 : 262-80). The sentence (3)

I admire goodness.

would then have the following structural formula (with I = class of pronouns): (4)

I V A An

This formula can be simplified, insofar as the sequence A An stands in a paradigmatic relationship to N , as is seen when substituting Mary for goodness: (5)

a. I admire Mary. b. I V N

There is thus a functional equivalence between A An and N in certain syntactic structures (cf. Harris, 1 9 5 1 : 275ff.). Before, however, pursuing this syntactic aspect of word-formation, a formalization of the morphological model of word-formation might help to clarify some of the problems connected with this view of derivation. 2.1.2 Formalization A formalization along the lines of algebraic linguistics conceives the subpart of grammar in which derivational processes are treated as a set of rules which specify the class of well-formed lexical items of the language described. These rules can be brought into the form of a context-free grammar, with L (lexical item) as the 'axiom' of the system, analogous to the category S of a phrase-structure grammar. 1 * What is meant by L in this context needs some clarification. If we take the morphemes to be the ultimate constituents of the sentence as Bloomfield did, including with him among bound morphemes morphological processes like vowel-alternation etc., and if we take the word to be a free form consisting either of one morpheme or an internally coherent se16

For an introduction to algebraic linguistics see Wall (1972). For other formalizations see Mötsch (1962) [includes transformations], Kiefer (1973a), Brockhaus (1975).

14

quence of morphemes, we can define the lexeme or lexical item as an abstract entity standing for a class of words which differ from each other only in respect of their grammatical morphemes. There is thus a difference between the lexeme or lexical item and the word (or more precisely the various word-forms). In addition to this the lexeme has a citationform, i.e. a form that customarily appears in the lexicon. In a language like English it is this citation-form which enters into derivational constructs; compare manful vs. *menful or '-man'sful. It so happens that the citation-form of lexemes in English is the word-form not overtly marked by grammatical morphemes, i.e. morphemes marking categories such as number, person, tense etc.; of course, this does not mean that these word-forms do not express some grammatical category (cf. man encoding [singular] for instance). More correctly these particular word-forms are from the point of view of derivation 'stems'; as these stems are, however, formally identical with lexemes in their citation-form, the practice of loosely equating lexemes with words and stems has been adopted for this study. 17 If L then stands for the class of objects to be defined by the grammar, we might postulate as the auxiliary alphabet the categories 'simple' or 'underived lexical item' (SL), 'complex' or 'derived lexical item' (CL) and 'derivational element' (D). The basic alphabet would then consist of simple lexical items like man, power etc. and derivational elements like -/«/, -ness etc. Adding to this a context-free rewrite rule we can formulate the following 'grammar': (6)

(i)

L

-»

L + L SL J CL

(ii)

CL

( SL + D ) ( CL + D*

(iii)

SL —> {man, power}

(iv)

D —> {-/«/, -we«}

This grammar generates among other lexical items man, power, manpower, manful, powerful, manfulness, powerfulness. Unfortunately it also generates a great many (to be precise, infinitely many) incorrect lexical items, such as *powernessful, *powerfulpowerful, *manpower17

Cf. Lyons (1968: 197-8), Matthews (1974: 20-35), Lyons (1977: I, 18-25). r

5

nessfulman etc. Certain refinements will have to be introduced in order to make (6) into an acceptable generative device. Some of the incorrect forms, like ''powerness or *powerfulful, can be 'filtered out' if we subcategorize the lexical items into lexical classes; -/«/ is characteristically added to nouns, while -ness can be suffixed only to adjectives. Restricting ourselves for simplicity's sake to the word classes 'noun', 'verb' and 'adjective', a more realistic 'word grammar' of English would be the following (for lexical classes see 4.1.2): (7)

(0 L

,L+ L SL„ 1 SLV

SLa CL n CL V

CLa (ii) CLn

i (SLn > * CLn)

n

IclJ+Dv !clJ+d" (iii) CLV

-»

/iSLn) »CL n '

Un

IclJ+d" (iv) CLa

-

, |SLn , {man, centre, power ... } —* {grow, read...}

(vii) SLa

-»

(viii)

16

Dnn

-»

{free, good ... } {-hood...}

—

{-th...}

(ix)

Dv„

(X)

D„

{-dom, -ity, -ness ... }

(xi)

Dnv

{-6...} {de-, over- ... }

(xii) (xiii)

Dav

{-¿, -ize ...}

(xiv)

D„a

{-al, -/«/...}

(XV)

Dva

—»

{-ing, -able ...}

(xvi)

Daa

—»

{un- ...}

This grammar generates besides compounds and simple lexical items the following complex lexical items (among others): manhood (SLn + Dnn). growth (SLV + D vn ), freedom (SLa + Dan), (to) man (SLn + Dnv), overgrow (SLV + D m ), (to) free (SLa + D av ), powerful (SL„ + D na ), growing (SLV + D va ), unfree (SLa + Daa), but also more complex lexical items like decentralizability:

The rules as they stand are however still insufficient, quite apart from the incompleteness of the basic vocabulary and the restriction to only three lexical classes. A number of morphological and morphophonemic rules have to be added. Thus adjustments will have to be made for the correct placement of prefixes. One way of doing this is to introduce a transformational rule (for transformational rules see 2.2.2) which puts 17

derivational elements marked as prefixes (recognizable by the position of the hyphen) to the front of the lexical category they are combined with. This rule would have the following form (where SD gives the structural description of the sequence to be transformed and SC its structural change): (9)

SD:

X

-

£

L

a}

+

[Z-]^

a SC:

X

-

-

Y

=>

aP

[Z-]D

+

ap

{ C s L L a} a

Y

where X and/or Y can be zero, where Z is some prefix, and where a and P = n or v or a, with a = P or a =t= p.

Applied to (8) this rule will yield de-+centre+-al+-ize+-able+-ity. In order to convert this string of morphemes into decentralizability-or rather /disen tr3laiz3'biliti/-s tress assignment rules and morphophonemic rules to the effect of changing /'senta/ into /'sentr/ in the context /al/ and /abl/ into /a'bil/ in the context /iti/ will have to be introduced (cf. Chomsky, Halle, 1968: 83ff.). Another shortcoming of grammar (7) is that it does not allow for the co-occurrence of more than one derivational process, as for instance in warm-hearted. This compound is neither derived from * warm-heart by suffixation nor from warm and *hearted by composition, but from warm and heart by suffixation and composition simultaneously (see 5.1). In fact, the compounding rule as it stands (L —> L + L) has to be restricted, as not all word classes can build compounds; L v + L v for instance yields unacceptable forms (though not in all varieties of English, cf. Pidgin talktalk etc.). While the problems mentioned so far can be solved by various modifications of the rules of (z)-subcategorization, introduction of morphological and morphophonemic rules, provisions for 'simultaneous derivation'-, there are a number of points that are less easily amended. The rules formulated are recursive. Now one might argue that compounds can be of indefinite length in English, at least in principle, though compounds of more than three elements are far less likely to be met with in English than for instance in German. But is this also true of affixal derivation in English? Is it possible, at least in principle, to derive a word of indefinite length by means of affixes only? If we take the sentence (10a), we can paraphrase it by (10b): 18

(10)

a. John has (shows) pity. b. John is pitiful.

From pitiful in (iob) we can derive pitifulness as in (na), with ( n b ) as a paraphrase: (n)

a. John has (shows) pitifulness. b. John has the quality of being pitiful.

If we ascribe a quality to John as in (i ib), we can also ascribe to him the quality of having this quality, and again the quality of having the quality of having this quality, ad infinitum. Although it might be difficult to understand, there is nothing wrong with a sentence like: (12)

John has the quality of having the quality of having the quality of being pitiful.

The possibility of a syntactical structure like (12) seems to entail the possibility of a corresponding lexical structure. If John has the quality of being pitiful, we might say that he is full of pitifulness or ''pitifulnessful, and if he has the quality of being the latter, then he might be described as *pitifulnessfulnessful, and so on. Why is it then that these forms have to be marked by an asterisk? One might contend that these forms need not be marked by an asterisk at all; although presumably never recorded, the forms are basically buildable. Compare in this context the arguments for and against an indefinite incorporation process in Mohawk and the evidence for 'regular' affixation in agglutinating languages.'8 In Turkish, for instance, it is morphologically possible to form causatives of the third and fourth (and nth) degree, as e.g. oldurttiirtmek, "to get someone to get someone to get someone to make someone die", although such forms "are rarely if ever found outside the pages of grammar-books" (Lewis, 1975: 146). Does the same principle apply to the affixation of -ful and -ness in English? Going back to (12) we might say that this sentence is both grammatical and that it makes sense, although the state of affairs it describes is of a very peculiar kind, presumably not very often made the subject of discourse. There might be philosophical reasons for denying that a state of affairs as described in (12) exists, but this is a different matter. From this it follows that it cannot be for purely semantic reasons that pitifulnessfulnessful (with or without asterisk) is 'odd', to say the least, in 18

For the Mohawk example see Postal (1964) and Reich (1969).

!9

English. Yet there is a difference between this word and sentence (12). Lexical items are far more subject to linguistic conventions than sentences. I can utter a sentence like (12) irrespective of whether the speakers of English think that there is any point in talking about qualities of qualities of qualities etc., but I cannot build a lexical item without paying attention to the types of lexical items already in existence (for the notion of 'derivational type' cf. J . I . I ) . This does not exclude the formation of new items and it does not impinge on the innovative use of language both in ordinary discourse and in poetic language, but it does exclude the derivation of adjectives meaning "having a " from abstract nouns, where a stands for the meaning of an abstract noun. I can certainly utter a string of morphemes like pitifulnessfulnessful and a native speaker of English might, after reflection, even understand what I mean by it. He will, however, take it as an idiosyncratic expression, which really needs a gloss in order to be fully intelligible, and he will wonder why I did not say having the quality of having the quality of being pitiful in the first place. This is a difficult and ramified problem, and we will have to return to it in due course (see 5.1.2, 5.2.3); for the moment we can observe that a recursive word grammar of English would generate a great many forms that are unacceptable to speakers of English by the standard of the existing lexicon. There ist yet another side to the problem of acceptability. According to (7X) an adjective can be converted into a noun by suffixing either -dom, -ity or -ness. This rule does not indicate whether each element of SL a (or CL a ) can have all three suffixes. If we take the adjectives free, good, rapid, happy and elastic, we get the following abstracts (the asterisk marking again unacceptable forms): (13)

freeness

goodness

rapidness

happiness

freedom

"'gooddom

:

'rapiddom

*happidom

elasticness '•'elasticdom

*freeity

''goodity

rapidity

*happity

elasticity

While -ness can be affixed to all adjectives, -dom and -ity are more restricted. Our grammar must somehow account for these restrictions. Here the problem of acceptibility is intimately connected with the question of productivity: when there is affixal variation as in the present example, not all affixes are normally equally productive. This factor needs to be considered when constructing a 'projective' grammar of the type linguist B would favour (see above).-Having stated the most important problems encountered when formulating a word grammar of 20

English, w e must n o w turn to a further defect of (7), its lack of information on the syntactic aspect of derivation.

2.2 D e r i v a t i o n a n d S y n t a x 2.2.1 Non-transformational Analyses T h e syntactic aspect of word-formation has been stressed especially by linguists concerned with derivation from a diachronic point of view. T h e neo-grammarian H . Paul saw the development of derivation as a sequence starting w i t h a syntactic construction and ending up, via c o m position, with affixation. T h u s a phrase like the condition O l d English frizes

of the free,

döm, might be transformed into a c o m p o u n d - n o w

taking the nominative case of free ( O l d English freo-döm)-and into a suffixed form, as in M o d e r n English freedom.19

finally

K . Brugmann was

even more emphatic in his study of compounds: Für die Entwicklung der menschlichen Sprache heißt es bekanntlich nicht: im Anfang war das Wort, sondern: im Anfang war der Satz. (Brugmann, 1900: 388) In synchronic analysis the syntactic aspect of word-formation is already touched on w h e n recognizing the need for the introduction of lexical classes into derivational rules (cf. also 2.1.1 for Harris' procedure). Lexical classes differ f r o m each other not only in form but also in function (more of this in 4.1.2) and there must therefore be a syntactic difference between the various derivations according to the lexical class of the base-form and of the resulting form. This is underlined b y J. K u r y l o w i c z in an article on the parts of speech, which appeared in 1936. K u r y l o w i c z distinguishes the fonction tion syntaxique.

lexicale of a w o r d from its fonc-

B y the former he means the lexical class or w o r d class to

w h i c h a w o r d belongs, of w h i c h there are four according to K u r y l o w i c z , noun, adjective, verb and adverb; these are defined basically b y semantic criteria (nouns denote things, adjectives qualities; cf. 1 above). F r o m the fonction

lexicale of a w o r d (or its sens lexical) results its fonction

syntaxi-

que primaire. This is the 'normal' function of a w o r d or w o r d class in the sentence, in the case of a noun that of the subject, in the case of an adjective that of determinant-epithete 19

of the subject, in the case of a verb

C f . Paul (1920: 343ff.); -döm has already in O l d English a 'semi-suffixal' character; compare Meid (1967:

F o r a criticism of this theory, the 'agglutination theory',

see Jespersen (1922: 375 ff.).

21

that of the predicate, and in the case of an adverb that of déterminant du verbe. The derivation of a lexical item from another word without change of lexical class is called by Kurylowicz dérivation lexicale. In this case semantic changes within a given word class are brought about by derivational processes, changes that cannot be captured by syntactic rules. If, however, the derivational process engenders a change of lexical class, syntactical phenomena are also involved. Kurylowicz terms this type of derivation dérivation syntaxique.10 The fonction syntaxique primaire of a word is now increased by a fonction syntaxique secondaire; the latter is a syntactic function which is typical of a word class other than the one the base word belongs to. An example would be the use of a noun A as déterminant of another noun B. It is important to see that this secondary syntactic function of A can be brought about both by syntactical and by derivational means; compare the function of duke's and ducal in the following pair of sentences (with duke = A, park = B): 21 (14)

a. The duke's park is not open to the public. b. The ducal park is not open to the public.

It should be noticed in passing that duke's and ducal are not fully equivalent. While the expression the duke's park presupposes that there is a duke and that it is indeed a duke and not a duchess, the phrase the ducal park lacks such a presupposition; we will return to this question in a later chapter (cf. 4.2.2, 5.2.3). The syntactic aspect of derivational phenomena is also considered in H. Marchand's standard reference work on English word-formation. 22 Marchand classifies derived words into morphological units like archenemy and syntactic groups like stone wall. Following Saussure words are defined as signs (signifiant) encoding a meaning (signifié; for a discussion of 'meaning' cf. 4.2.3). It is only those signs which are formally and 20

21

22

For a similar distinction between 'governing derivation' (word class change) and 'restrictive derivation' (word class maintenance) cf. Hockett (1958: 243). Compare also Marchand's terms 'transposition' (~dérivation syntaxique) and 'expansion' (~dérivation lexicale); cf. Marchand (1967); Marchand's terms overlap with the ones used by Dokulil ('transposition', 'modification', 'mutation'); cf. Dokulil (1968), Giinther (1974), Kubrjakova (1974). The syntactic equivalence between a construction with a noun in the genitive and one with a denominal adjective was already noticed by Priscian; cf. Keil (1855-1923: II, 68-82). Especially in the second edition of EWF. For the overlapping of syntactic and derivational processes see also Dokulil (1964). A general discussion of the delimitation of grammar (in the narrow sense) and lexicon is found in Schopf (1969: 4-51). 22

semantically segmentable (cf. J.I.I) into constituent signs that lie within the pale of word-formation. Two cases have to be distinguished: (1) The derived word is a grammatical syntagma, i.e. a combination of full linguistic signs; to this category belong composition, affixation, zero-derivation and backderivation. (2) The derived word is not a grammatical syntagma, i.e. it is not formed from full linguistic signs; here belong cases of phonetic symbolism, rhyme combination, clipping and blending. Marchand extends the notion of the determinative compound-a compound where one element modifies the other as in bloodtest, blood specifying the kind of test denoted by the compound-to all derivatives by dividing them into two parts, a determinant and a determinatum. Both parts are realized as morphemes; where a word cannot be formally and semantically analysed (Marchand is here in disagreement both with Bloomfield and with Nida), it has to be treated as a moneme, an unanalysable whole. In the case of compounds the syntagma is in opposition to both morphemic elements, in the case of prefixal and suffixal derivations the syntagma is in opposition to the base morpheme and to all the lexical items having the same affix. Marchand includes diachronic information in his survey of English word-formation, but stresses that only backderivation is an exclusively diachronic derivational process. 2.2.2 The Transformationalist Position A systematic analysis of governing derivation (cf. note 20 above) as a syntactic process was first undertaken within the framework of generative transformational grammar. It is in particular the formation of nouns which has attracted transformational linguists. The derivation of nouns like silliness or arrival from silly and arrive, respectively, is seen as a special case of a nominalizing process which turns both word classes and syntactic constructions (including sentences) into 'syntactic nouns' (noun phrases or nomináis), syntactic units that stand in a paradigmatic relationship to simple nouns. The nouns silliness or arrival are thus comparable to other nominalizations of their respective bases: (15)

(16)

a. b. c. a. b. c.

That John arrived upset me a lot. John's having arrived upset me a lot. John's arrival upset me a lot. That Bill is silly hasn't gone unnoticed. Bill's being silly hasn't gone unnoticed. Bill's silliness hasn't gone unnoticed.

23

The parallelism between these various nominalizations can be captured, if they are related to a common source, in the case of (15) to (17a), in the case of (16) to (17b): (17)

a. John arrived. b. Bill is silly.

It is for this reason that nominalizations, including the derived lexical noun, are not generated by the phrase-structure grammar, but derived from their underlying structure by transformational rules. A first step in this direction was made by 2 . S. Harris, when he correlated two or more syntactical structures containing the same number n of word classes (whatever else they contain) and occurring with the same n-tuples of elements of these classes in the same environments. 23 According to this criterion (18a), (18b) and (18c) are correlated or transforms of one another (compare 2.1.1; T = definite or indefinite article): (18)

a. That (T) N is A b. (T) N'J be-ing A c. (T) N'J A An

This is borne out by the fact that with each insertion of an element of N and of an element of A into formula (18a) the same element can also be inserted into the formulas (18b) and (18c), while in addition to this the phrases thus formed occur in the same environments (cf. (16) with Bill for N and silly for A). In order not to complicate matters we will leave out of account here that the construction (18c) is ambiguous and is found in contexts impossible for (18a) and (18b) (cf. 5.1.1, 5.1.2). While according to Harris (1957) transformations correlate structures generated by a phrase-structure grammar, in N . Chomsky's model transformations are functions which map strings of constituent structure elements onto a new string of such elements: A grammatical transformation T operates on a given string (...) with a given constituent structure and converts it into a new string with a new derived constituent structure. (Chomsky, 1957: 44)

In this case the phrase-structure part of grammar need only generate untransformed structures like the ones underlying (17a) and (17b). According to the model of transformational grammar as presented in Chomsky's Syntactic Structures de-adjectival abstracts are transforma23

C f . H a r r i s ( 1 9 5 7 : sec. 3). F o r his later definition of transformations as operators see Harris (1968).

24

tionally derived from sentences where the adjective functions as the predicate. This can be illustrated by one of Chomsky's examples (cf. Chomsky, 1962): (19)

The country's safety is in danger.

In order to generate (19) Chomsky postulates two underlying structures on which the nominalizing transformation operates: (20)

a. 1 -it + 0 - VP b. NP - C - be - Adj

(20a) corresponds to the sentence (21a) and is generated by phrasestructure rules as represented by the tree (21b), where T symbolizes the definite or indefinite article, 0 the singular morpheme of the noun, N a b a subcategory of nouns comprising abstracts and their pronominalized form it: (21)

a. It is in danger.

b.

in danger

(20b) on the other hand corresponds to the sentence (22a) and is generated by phrase-structure rules as represented by (22b) (here C symbolizes the class of tense morphemes): (22)

a. The country is safe.

b.

The

country

25

The nominalizing transformation does two things: it embeds (20b) into (20a) and it changes (20b) so that it can replace T - it + 0 in (20a). It can be formalized as follows: (23)

SD-.T-it

+ 0-VP, NP-C-£e-Adj

SC: N P +

S-

=>

nom + A d j - V P

S is the genitive suffix (as 5 in country's) and nom the nominalizing affix (as -ty in safety). Chomsky leaves structures in which the nominalizations function as something other than the subject out of account; these can be accommodated, however, without much difficulty into the model illustrated. The fullest transformationalist account of nominalization in English has been given by R . B . Lees (cf. Lees, i960). His analysis is along the same lines as that of Chomsky. Lees points out that there are a number of morphological problems with derived abstract nouns-problems of productivity and suffixal variation-, but he does not go into these questions. The transformational treatment of nominalization is modified in the Aspects model of generative grammar in a number of cases.24 The nominalizing transformation has no longer two strings as its input, since the constituent sentence is already embedded into its matrix sentence by the rules of phrase-structure grammar. This is made possible by the introduction of recursive rules which allow S to appear to the right of a rewrite phrase-structure rule. A further difference to the earlier model relevant to derivation concerns the expansion of terminal symbols. The different lexical categories are now rewritten as complex symbols, i.e. as sets of syntactic features, inherent features in the case of the noun, selection features in the case of the verb. This means that de-adjectival nouns like goodness or sincerity have the feature [-(-abstract] (cf. Chomsky, 1965: 79 ff.). Parts of the last chapter of Aspects is devoted to questions of wordformation. While the derivation of a noun from a verb or an adjective is comparatively regular, many derivational processes show irregularities and idiosyncrasies: Derivational processes create much more of a problem for any sort of generative (that is, explicit) grammar than do inflectional systems. This results from the fact that they are typically sporadic and only quasi-productive. ( C h o m sky, 1965: 184) 24

Cf. Chomsky (i965); for a comparison of the two models see Maclay ( 1 9 7 1 : 169-176).

26

It may be sufficient to list only good or sincere in the lexicon and generate goodness and sincerity by the grammar, but the same cannot be done as easily with words in -or, -id or -ify for instance (cf. horror, horrid, horrify; terror, *terrid, terrify; candor, candid, *candify; see also example (13) in 2.1.2). And yet, Chomsky contends, it is unsatisfactory to treat these words as unanalysable lexemes: . . . since it is clear that from the point of view of both the semantic and the phonological interpretation it is important to have internal structure represented in these w o r d s . Their meaning is clearly to some extent predictable (or at least limited) b y the inherent semantic properties of the morphemes that they contain, and it is easy to s h o w that internal structure must be assigned to these items if the phonological rules are to apply properly in forming their phonetic representations... ( C h o m s k y , 1 9 6 5 : 186)

It is at this point, where the issue of the semantic and deep structure syntactic analysis of derived lexical items is raised, that generative grammarians have split into different 'schools'. The remainder of this chapter is devoted to a discussion of some of the most important features of these diverging grammatical models relevant to derivation.

2.3 Derivation and the Lexicon If, following Chomsky's suggestion in the foregoing quotation, inherent semantic properties of morphemes like -ify are taken into account, it is not enough to say that -ify is a transpositional suffix; -ify must also be characterized as the morphological expression of a semantic component conventionally labelled "causative". While in words like horrify or beautify the correlation between a meaning element "causative" and a morphological feature can be stated by purely synchronic analytical procedures (we will refer to these verbs as 'class one verbs'), it is only with the help of diachronic information that the morphological structure of verbs like set, "make sit", or lay, "make lie", reveals a causative semantic feature ('class two'). There is also a third class of causative verbs in English-verbs such as kill, "make die", or show, "make see"-, where neither a synchronic nor a diachronic morphological analysis shows the reflection of "causative" in the form of the word. In the 'standard model' of transformational grammar there are several ways of accounting for semantic regularities in the lexicon such as the presence of a meaning component "causative" in a class of lexical items. 2

7

As to verbs of the first class (verbs overtly marked by some derivational element, including a zero element as in walk in the sense of "make walk") a transformational derivation could be proposed. In this case the 'causative transformation' would have roughly the following form: (24)

SD: X - Vp - [[NP; - Aux - have - Nq]s]Np - Y => SC: X - Nq + -ify - NP; - Y

In (24) SD specifies structures underlying sentences like John caused Bill to have horror, i.e. V p is a subclass of verbs comprising cause and make and N q is a subclass of nouns restricted to words like beauty, horror, zinc and a few others to which -ify can be affixed. The rule in its present form is of course oversimplified. It does not take into account adjectives modifying N q ) a fact which complicates matters considerably, nor does it allow for the derivation of verbs like fishify or solidify, where instead of have - N q the structures V r - N s and V r - Adj t , respectively, have to be postulated (with V r —> become, N s —> fish, alkali etc., Adj t —> solid, liquid etc.). Furthermore, there are quite a number of morphological problems involved in a transformational solution; besides the necessity of formulating morphophonemic rules like /'hora/ —» /'hor/ in the context /ifai/ there is the difficulty of deriving verbs like mollify where no base exists in English.25 Some of the difficulties connected with a transformational analysis of ify-verbs are avoided if derivational phenomena are dealt with in the lexicon. According to M. Halle's proposal the rules of word-formation are to operate on the morphemes of the language under investigation generating the set of lexical items, which in turn are then inserted into syntactic structures; unacceptable formations are eliminated by a 'filtering device' attached to the word-formation component.26 While Halle's model is designed to deal primarily with morphologically analysable derivatives like the verbs of class one, arguments have been advanced in favour of derivational rules within the lexicon which also generate synchronically 'opaque' words like the verbs of class two (for diachronic vs. synchronic analysis see 5.1.2). Once diachronic information is used, however, in the analysis of lexical items, the boundaries between wordformation and etymology become blurred. This is illustrated by T. M. Lightner's approach where words like generic, pregnant, kind and kin25

26

For an analysis of ify-verbs in English see Mahn (1971). In the present context the difference between causative and factitive verbs has been neglected. We will come back to the morphological problems only hinted at here in 5.1. C f . Halle (1973) and for pertinent criticism Boas (1974) and Lipka (1975).

28

dergarten (among many others) are derived from one and the same base by the rules of English (!) word-formation (cf. Lightner, 1975). But even in such an extreme version kill could not be derived from die. For the verbs of class three "causative" can be registered only in the semantic description of the lexical item, for instance in the way J . J . Katz and J . A . Fodor suggested within the framework of the standard theory. 27 2.3.1 Pre-lexical 'Semantico-Syntax' There are, however, other alternatives, one of them an alternative advocated by a group of linguists commonly grouped together as generative semanticists. These linguists share with the proponents of componential analysis in semantics the belief that the semantic content of the lexical items of a language can be stated by the members of a presumably finite and universal set of 'basic' meaning elements.28 These elements are, however, represented as semantico-syntactic units in generative semantics and not as features, more precisely as predicates in the logical sense. This implies that syntactic structures, too, are modelled on the predicate calculus of formal logic. 2 ' Taking sentence (25a) as an example, its underlying structure will have to be segmented into a two-place predicate K I L L and two ordered arguments, John and Bill (see also 4.2.1):

(25)

a. John kills Bill.

John

Bill

KILL

Generative semantics, however, does not stop at this point. K I L L is seen as a complex predicate which can be further decomposed into elementary or atomic predicates. The structure underlying (25a) is thus

27 28

25

C f . Katz, Fodor (1963) and Katz (1972). The two respective sets need not be identical. For componential analysis see Lyons (1977: I, 3 1 7 - 5 5 ) . C f . Seuren (1974: 1-27), Maclay ( 1 9 7 1 : 176ff.), Lyons (1977: II, 409-22), Lakoff (1970a), Lakoff (1970-1). For a discussion of atomic predicates cf. Wierzbicka (1972). For a criticism of generative semantics cf. Brame (1976).

more correctly represented by (26), an analysis first proposed by J . D. McCawley and subsequently modified by McCawley and others: 30

Arg

Pred

Bill

ALIVK

An operation termed 'predicate raising' combines predicates of embedded sentences with higher level predicates, thus generating lexical items that correspond to various configurations of predicates, such as dead = N O T + A L I V E , die = N O T + A L I V E + B E C O M E , kill = N O T + A L I V E + B E C O M E + C A U S E . The possibility of such a pre-lexical syntax was discussed already by J . Gruber, who postulated a polycategorial lexical attachment rule, displacing for instance go + into in a structure like [[go]v + [into + NP]pp]yp by the lexical item enter (cf. Gruber, 1976). It is to be noted in this context that structures like (26) are far from being uncontroversial; although the predicate calculus has been taken for a model, the treatment of a logical operator like ~ ("not") as predicate is, to say the least, very unorthodox from a logical point of view. Derivational processes are also analysed on the basis of the predicate calculus by H. E. Brekle in his study on nominal compounds in English (cf. Brekle, 1976). Brekle derives nominal compounds from sentences, 30

Cf. McCawley (1968a, 1968b, 1971a, 1971b); for a critical discussion see Fodor (1970), Chomsky (1970a), Lipka (1972: 61-72), Wierzbicka (1975), Lipka (1976).

3°

thus for instance fieldmouse and mousehole from (27a) and (27b) respectively: (27)

a. The mouse lives in the field. b. The mouse lives in the hole.

Although the structures underlying (27a) and (27b) are represented as a combination of n-place predicates with n arguments, Brekle differs from the classical predicate calculus in a number of points. Instead of classifying live in as a two-place predicate with mouse and field and mouse and hole, respectively, as arguments, Brekle sets up a hierarchy of predicates to capture the logical structure of the sentences in question. He terms mouse {—w), field (=y) and hole (=y) first-order predicates, live ( = R ) a second-order predicate and introduces generalized predicates like L O C ("locative") and A F F ("affective") etc. Brekle holds that between lives and field in (27a) and lives and hole in (27b) there is a locative relationship, morphologically realized by the preposition in. Between this expression, symbolized as L O C (R, y) and paraphrasable as "There is a relationship of 'being in local proximity' between R and y", and mouse there is, in Brekle's view, the further relationship of 'being affected by one another'. In other words, the formula L O C (R, y) and the symbol w are arguments of the two-place predicate AFF. This is symbolized by (28a) and, in tree form, by (28b): (28)

a. AFF ((LOC (R, y)), w)

y

R

LOC

In order to derive nominal compounds from (28), an element is topicalized. If w is topicalized, one gets fieldmouse and holemouse, if y is topicalized one gets mousefield and mousehole.JI For a further analysis 31

Although unusual both holemouse and mousefield are perfectly acceptable words. Brekle defends his analysis against the transformationalist position as in Lees (1960: 1 1 3 ff.) in Brekle (1975).

31

of the structures underlying compounds like fieldmouse-\.e. structures paraphrasable as the mouse such that it lives in the field—see the discussion of definite descriptions in j.2.1. Pre-lexical semantico-syntactic structures as the source of derived lexical items are also found in the linguistic models proposed by C. C. Fillmore and by W. L. Chafe. In Fillmore's case grammar a sentence is divided into a modality component, expressing tense, mood etc., and a propositional component (cf. 4.1.3). In the latter a 'predicator', characteristically a verb, is linked to n linguistic units, characteristically though not necessarily noun-phrases, by 'deep structure cases' (cf. Fillmore, 1971a: 37). It is easy to see that Fillmore's analysis of the propositional content of a sentence can be re-formulated in the terms of the predicate calculus, with the 'predicator' as n-place predicate and the various case-marked linguistic entities as arguments. The predicates, however, are here not only subcategorized according to the number of arguments they take, but also according to the cases these arguments are in. Thus a verb like kill is characterized as a two-place predicate, taking as arguments a unit marked as agentive and one marked as objective; if the instrument is specified, as in John killed Bill with a knife, kill would have to be classified as a three-place predicate with a third argument in the instrumental case (cf. Fillmore, 1968). The verb kill is, however, decomposable also within the framework of case grammar. Fillmore argues for a treatment of causatives like kill as instances of event causation, that is by taking into account that John can kill Bill only by doing something. John's doing something (with John as agentive) functions as the instrument, Bill's dying as the result or goal; between instrument and goal there is a causal relation. John killed Bill is then more properly represented by (29) (cf. Fillmore, 1971a; Kastovsky, 1973):

S

(29)

Modality

Proposition

V

CAUSE

32

S,

[Instr.]

[Goal]

John does someth.

Bill dies

In Chafe's model sentences are analysed as semantically characterized structural relations between sets of features. The basic structure underlying John killed Bill could be represented by (30), where the choice of a verb marked by the features [process] and [action] engenders the addition of two nouns classified as [patient] and [agent] respectively: (30)

I V [process] faction I

I

I

[pat] N

[agent] N

Bill

John

I

kill

Although Chafe's examples of causatives are taken from class one verbs, it is conceivable that kill could be derived from die according to his model. The derivational rule in question states that a verb-root with the feature [process] can be optionally marked also by the feature [action]; in this case a causative morpheme has to be added (an extreme case being substitution) and an agent noun has to be present besides the patient noun engendered by [process] (cf. Chafe, 1970: 128-131). (30) could then be re-analysed as (31): (3')

V [process] [action]

[pat] N

[agent] N

die + [causative]

Bill

John

2.3.2 The Lexicalist Position In his article "Remarks on Nominalization" Chomsky distinguishes between two possibilities of describing some trait of the language under investigation in its grammar: either the phenomenon is analysed by the rules of the base (phrase-structure and lexicon) or by the rules of the transformational component. In the first case an extension of the base 33

simplifies the transformational part; this is the lexicalist solution. In the second case the base is simplified, while the transformational component becomes more complicated; this is the transformationalist position (cf. C h o m s k y , 1970b). We have seen already that morphologically analysable causatives can be derived transformationally in the standard model, though it is more usual to adopt a lexicalist position (cf. Halle's proposal, 2.3.1). Much more controversial is the case of nominalization. While Lees and C h o m s k y in his earlier writings opt for a transformationalist solution (cf. 2.2.2), C h o m s k y now argues that de-adjectival and deverbal abstract nouns have to be derived within the lexicon. In his opinion a transformational analysis is unable to account satisfactorily for (1) the nominal character and (2) the frequent morphological idiosyncrasies of abstract nouns. Both characteristics of the 'derived nominal' become particularly noticeable when comparing it with the gerund. Whereas the gerund for instance cannot have an article or an attributive adjective (very often not even the corresponding adverb), this is possible with derived nouns analogously to primary nouns. Compare the following examples (taken mostly from C h o m s k y , 1970b): (32)

(33)

(34)

a. John's intelligence b. The intelligence displayed by John c. John's being intelligent d. *The being intelligent displayed by John a. John's sudden refusal of the offer b. John's sudden refusing the offer c. John's suddenly refusing the offer a. John's amazing resemblance to Bill b. :i John's amazing resembling Bill c. :i John's amazingly resembling Bill

Similarly the regularity with which gerunds can be derived from sentences contrasts with the irregularity and idiosyncrasy of derived nouns: (35)

a. John is eager to please. b. John's being eager to please c. John's eagerness to please

But: (36)

34

a. John is easy to please. b. John's being easy to please c. *John's easiness to please

We will return to these examples later and therefore postpone a discussion of their structure till then (cf. j.2.2). 32 An extension of the base in the lexicalist sense implies that the syntactic structures underlying expressions like John's intelligence, John's refusal of the offer etc. must be generated directly by the rules of the phrase-structure grammar. In order to be able to do this, Chomsky introduces what he calls the 'X-bar convention'. It is embodied in the following rule (Chomsky, 1970b: 210): (37)

X —> [Spec, X ] X

This rule allows for the generation of analogous verb-phrase, nounphrase and adjective-phrase structures. It states that any of the categories N , V or Adj can be expanded into a category of the same type; the latter category is combined with some other category which functions as its specifier or modifier. In the case of a verb the specifier is characteristically an auxiliary, in the case of a noun a determiner, in the case of an adjective an adverb. This rule can be illustrated by the following examples (compare Chomsky, 1970b: 2 1 1 ; MSSE: 6): (for (39) see p. 36) (38)

a. The enemy destroyed the city.

Det

N

the

city

For both destroyed in (38a) and destruction in (39a) a common lexical item destroy appears in (38b) and (39b). This analysis results from a revision in the lexicon. In order to be able to explain the partial similarity 3J

For a discussion of the lexicalist position see inter alia MSSE: i f f . , Wik (1973: 48-63), Kastovsky (1977a), Jackendoff (1977). 35

(39)

The enemy's

destroy

[Spec, N ]

N

1 Det I 1

N I 1

the

city

between base and derived w o r d - f o r instance in respect of selectional restrictions-lexical units are listed ... with fixed selectional and strict subcategorization features, but with a choice as to the features associated with the lexical categories noun, verb, adjective. The lexical entry may specify that semantic features are in part dependent on the choice of one or another of these categorial features. (Chomsky, 1970b: 190)

This means that in the case of a de-adjectival abstract both the base adjective and the derived noun could be analysed by the same lexical entry, indicating that the choice of the nominal form has the meaning "the quality of being a " (with a = the meaning of the base adjective) etc. The relatedness of the base adjective and the derived abstract could be further stressed, when deriving the latter from the former in a wordformation component along the lines of Halle's "Prolegomena". In this case it would presumably be more natural, however, to list the two forms separately in the dictionary; semantic information need not be duplicated; for shared semantic features, selection restrictions etc. the lexical entry of the derived form could refer back to the entry of the base form. 3 3 There are still other variants of the lexicalist position. R. S. Jackendoff, while firmly lexicalist, constructs his lexical entries nevertheless 33

C f . Jackendoff (1975: 642ff.); Jackendoff, however, favours what he calls the 'fullentry theory', i.e. the reduplication of semantic information in two separate lexical entries. For a critique of Jackendoff's position as presented in Jackendoff (1972) see Rosenberg (1974-5), Hust, Brame (1976).

36

very much in the spirit of generative semantics by formulating at least part of the meaning as semantico-syntactic structures (cf. Jackendoff, 1972: 36ff.). A modification of the lexicalist position is also advanced by R. Beard, whose phrase-structure grammar is based on the model of generative semantics; while prohibiting lexical transformations of the predicate raising type, he advocates the formulation of rules that allow for the insertion of derived lexemes into appropriate deep structure contexts, not unlike the lexeme insertion rules with selectional restrictions of the standard model of transformational grammar (cf. Beard, 1976a). The lexicalist position is coupled with Fillmore's case grammar in The Major Syntactic Structures of English by R. P. Stockwell, P. Schachter and B. Hall Partee. Taking (38) and (39) as an example, the 'X-bar analysis' is criticized in a number of points. According to the authors of MSSE the graphs (38b) and (39b) do not show that enemy stands in the same relation to destroy and to destruction and that a noun when combined with both destroy and destruction must have the feature [concrete]. Although there are indications to this effect in the lexicon, a syntactic theory where these relationships and selectional restrictions can be immediately derived from the deep structure analysis is to be preferred to Chomsky's lexicalist 'extended standard theory'. By modifying the base component of grammar in the sense of case grammar, diagrams (38b) and (39b) can be re-written as (40a) and (40b) respectively (cf. MSSE: 7, with destruction instead of destroy in (40b)). In (40a) and (40b) the character of the base verb as a two-place predicate with a noun phrase in the agent case and a noun phrase in the neutral case as arguments is reflected in the syntactic structure of the derived noun. Nevertheless, the derivation of this noun is a sub-lexemic process. While Chomsky derives gerunds only transformationally and is dubious about

37

N

destruction

Neutral

Agent

NP

NP

the city

the enemy

the character of 'mixed nouns'-deverbal action nouns in -ing (cf. 5.2.3) - , the authors of MSSE derive both the gerund and the deverbal action noun transformationally; P. Schachter proposes, however, in a later article (1976) that even the gerund has to be analysed by taking a lexicalist position; we will come back to these analyses and further arguments for and against 'lexicalism' (cf. 4.2.2, 5.2).

Before discussing a further alternative to analyse derivational processes, the formulation of word-formation phenomena within a categorial framework, in the following chapters, it might be useful to summarize some of the points made in this chapter by considering the kind of questions a theory of word-formation should account for. If A and B are two lexical items of which B is derived from A (more precisely, formally and semantically derived from A; cf. 5.1), the following questions can be asked: (1) What is the formal, morphological relationship between A and B? B might be a modified form of A-achieved by various morphological means-or it might be formally identical with A (for the possibility of A being a modified form of B see 5.1). (2) What is the relationship between A and B in terms of lexical classes (for the term 'lexical class' and related terms see below 4.1.2)? They might belong to different classes (as e.g. silliness, N , from silly, Adj), they might belong to different subclasses of the same 'major category' (as e.g. walk, V tr , in John walks the dog every morning, from walk, Vitr), or they might belong to the same class (as e.g. kitchenette, N, from kitchen, N). (3) Are there syntactic constructions 1), in which B can occur, that are transforms (in Harris' sense, cf. 2.2.2) of syntactic constructions i|)I...i|Jm (m> 1), in which A can occur? We have seen that in

38

the case of silly and silliness such constructions do exist, with zimuska ("little winter") = R J R 2 0 (denominal noun). These derivations can be represented graphically. A derivational graph consists of nodes representing words and branches representing relators, the different directions of the branches differentiating the various relators in the following way:

We can thus plot the derivations from zima by the following graph (cf. Saumjan, 1965a: 220): (see (19), p. $7). In the phrase-generator phrases are formed from words, again by the operation of application. Here in addition to words and relators connectors and adnectors are also used. Connectors are syntactical categories that map two identical categories onto an identical category, such as for instance the category and belongs to (it makes noun phrases out of noun-phrases-e.g. John and his father-, sentences out of sent-

56

(i9) , o zime (R 4 R 2 0)

zima ( R 2 0 )

zimnij (R,RjO) zimttska (R 2 R 2 0)

zimuet (R,R 2 0)

perezimuet / (R,R,R 2 0)«

^ zimovka

(R 2 R,R 2 0)

tzimovocnyj

(R,R 2 R,R 2 0)

ences - e.g. John squints and Bill lisps-etc.; cf. 4.2.1). Adnectors are relators whose value-category is identical to their argument-category, such as denominal noun affixes, deverbal verb affixes, verb-adverbs etc. Although the phrase-generator generates all possible phrases, it does not indicate which phrases are transforms of each other (or as Saumjan puts it, are in an invariance relationship). In order to accomplish this, a generator of phrasal transform-fields is introduced. The workings of this generator is somewhat complicated and is best illustrated by a simplified example. If we have the phrases (20)

a. The boy runs. b. The running boy c. The boyish runner

we can assign to these phrases the following categories (disregarding the article for simplicity's sake): (21)

a. R 2 0 R.O b. R , R , 0 R 2 0 c. R , R 2 0 R 2 R , 0

( = a a0) ( = a p . a a a) ( = a . a a a^.a)

The formulas ( 2 i a - c ) can be directly generated by the phrase-generator. But they can also be generated in the phrasal transform-field-generator. In this case the phrase (21a) R 2 0 R , 0 is taken as operand and the transforms of the first degree in the joint generator are calculated by the following method (for details see Saumjan, 1965a: 225ff.): The parts of the operand phrase are each connected with a relator and then connected with each other in a certain order.

57

R2O • Rs R3

R2 R, R4

RSR2O R,R2O R2R2O < R,R2O < R4R2O

R,O • • R,RJO

~

— R4R,0

This yields eight phrases, pairs of derived 'words' that are connected by a line. These phrases with an approximative interpretation are as follows: (23)

i. R,R2O .R3R,O ii. R5R10 R3R2O iii. R3R2o R2R,O iv. RJR,0 R2R2O V.

R2R2O

vi. R2R,O vii. R,R2O viii. R,R,O

R,R,O R,R2O R4R10

R4R2O

boyishly running runningly boyish the boyish runner the running boy the boy runs the runner is a boy is a boy runningly runs boyishly

Here (2oa-c) are shown to be transforms of each other (where R ^ O = R , 0 , R 2 R 2 0 = R j O ) ; we will have to come back to this effect of categorial grammars below in 4.2.1. In recent years Saumjan has modified his model in a number of points, of which some of the most important shall be briefly touched upon (cf. Saumjan, 1973, 1974, 1977)- On the level of episemions, a (=n) is now taken as a variable for six constants, which, following case grammar, stand for nouns in various situational roles: The constant a, known as an agentive, is interpreted as an animate object which produces some action. The constant i, called an instrumentive, is interpreted as an inanimate object in the broadest sense which serves as the cause of the action or state expressed by the predicate. The constant/, called an affective, is interpreted as an animate object which is subjugated to the action or is in the state expressed by the predicate. The constant /, called a locative, is interpreted as an object localized in space or time. The constant c, called a completive, designates a complement to an empty predicate. The constant o, called an objective, is interpreted as an object with a neutral role in a situation; the meaning of this role depends entirely on the meaning of the predicate. (Saumjan, 1973: 270)

58

Consequently the category of the predicate-Saumjan's category Aa(3 on the level of episemions and P Aa p on the level of semions-has to be subcategorised according to the various classes of nouns it takes as arguments. Putting s for (3 and using one of the notations of the standard model, we can represent Saumjan's examples for congruent sentences in the following slightly simplified form (cf. Saumjan, 1973: 271-277):

(24)

a.

Mal'cik begaet

( T h e b o y is running),

b.

s.

mal'cik (25)

a.

Voda kipit

( T h e w a t e r is boiling),

b.

s.

voda (26)

a.

Otec spit

( F a t h e r is s l e e p i n g )

b.

s.

otec (27)

a. Afrike

iarko

(It is cold in Africa),

b.

^

Afrike (28)

a.

Otec prinimaet vannu

b.

( F a t h e r t a k e s a bath), ^ [s;f,c]

otec

vannu

. .I prinimaet

59

(29)

a- Molnija ubila korovu (Lightning killed the cow), b.

korovu

molnija

The so-called 'de-predicators', i.e. those functions that have predicates as arguments, have to be subclassified in a similar way. Hence the agentive-suffix -er as in runner from run is not R 2 R , 0 (cf. (21c)) any longer, but more precisely a category that makes agentives out of verbs, in one form of the standard notation an element of category [a; (s; n)] in the case of an intransitive verb as base (cf. Saumjan, 1973: 277-280). Saumjan also introduces other inflexional and derivational elements into his model, categories denoting aspect, tense, mood, causatives, diminutives etc. (cf. Saumjan, 1973: 300-3). Most importantly, the invariance relationships between sentences are now captured by semantic rules, rules that formulate the derivability of some sentence form A from another sentence form B in the sense that the meaning of A follows from the meaning of B. Thus, for instance, one rule says that from P 0 T 0 the structure P(0) is derivable. By P 0 T 0 a sentence with an objective as noun (T0) and a verb of the category [s; o] (P0) is symbolized; Saumjan's example is Veter duet ("The wind is blowing"). From this structure the sentence Duet ("It is blowing") is semantically derivable, where the subject-noun has been deleted (cf. Saumjan, 1973: 283-333; Saumjan, 1974; Saumjan, 1977; for the treatment of verbs like duet in the framework of dependency grammar cf. Hickey, 1980: 66ff.). More about the semantic aspect of categorial grammars will have to be said in the following sections.

3.2 Formal Semantics Husserl's attempt to construct a 'pure grammar of meanings' has its modern counterpart in the work of those logicians and philosophers who concern themselves with the development of a formalized semantics, notably for the calculi of logic, but also, especially in recent years, for natural languages. While the group of 'formal semanticists' as a whole (by no means unified; cf. Lewis, 1970-1 ¡19) has little connection to categorial grammar as outlined so far, some of its most influential 60

members-especially R. Montague, D. Lewis and M. J. Cresswell-have employed a 'categorial notation' for their logical formulae, which thus mirror a categorial syntax for natural languages. This parallelism between syntactic and logical structures is in no way fortuitous; it is rather the expression of an endeavour to extend the scope of formal semantics to natural languages. This is accomplished all the more easily when both the syntactic and the logico-semantic 'level' of a language are described within the same theoretical framework. There is thus a powerful argument in favour of a categorial grammar that can be advanced from the point of view of a general theory of language, a 'universal grammar' as Montague called it. Before elaborating this argument, however, an exposition of the main characteristics of a 'Montague grammar' (and its variants) must be given. In view of the complexity of the theories concerned this exposition can be only of an informal kind; more details, in as far as they are relevant in this context, will have to be deferred till later.11 3.2.1 Montague Grammar By way of an introduction let us first consider the syntactic part of a 'Montague grammar' as presented in Montague's "The Proper Treatment of Quantification in Ordinary English" (henceforth abbreviated as PTQ).11 This part consists of the following sub-parts: (1) The definition of syntactic categories; (2) the categorization of the lexicon; (3) a kind of phrase-structure grammar, recursively defining the class of meaningful expressions (well-formed or grammatical expressions) of the language under consideration, i.e. in this case a fragment of English. As in the standard version of a categorial grammar the set of syntactic categories is specified with the help of two basic categories-in PTQ e (= entity expression) and t (= truth-value expression), corresponding, at least superficially, to n and s, respectively-and the notion of a function, allowing the formation of functor-categories. Montague writes these functors as A/B, where A represents the value-category and B the argument-category, a notational convention-corresponding to [A; B-] in 11

For a general discussion of the aims and methods of formal semantics see for instance Rogers (1963); Moravcsik (1975); Kutschera (1976). I2 ' The most important works by Montague on the philosophy of language are collected in Montague (1974), from which quotations will be taken. For introductory remarks see especially Thomason in Montague (1974: 1-69); Allwood, Andersson, Dahl (1977: 127-147); Partee (1975); Link (1976: 110-161). As PTQ is the paper by Montague most often discussed by linguists it is taken as a basis for the following exposition.

61

a linearly ordered categorial grammar or to simply [A; B] in a categorial grammar, where the concatenating operation is not linearly ordered -which will be employed also in the remainder of this book. A double dash serves to subcategorize functor-categories with identical valuecategories and argument-categories, but symbolizing different sets of lexical expressions (cf. also 4.1.2); besides A/B there is thus also a category A//B. In addition to e and t the following categories are used in PTQ (under the heading 'alternative notation' the categories are given according to the standard theory, employing, however, capitals instead of small letters and the dash instead of the notational conventions discussed in 3.1):

Syntactic Category

Abbreviation Explanation

Example from Alternative the Lexicon Notation

entity expression truth-value expression tIt t/IV

IV T

V itr term

IV/T IV/IV

TV IAV

V tr adverb to

t//e t/t

CN

common noun sentential adverb preposition V taking a sentential complement V taking an infinitival complement

IAV/T IV/t

IV//IV

N S run John, he0. he,... find rapidly

S/N S/(S/N)

man necessarily

S/N S/S

in believe that

((S/N)/(S/N))/(S/(S/N)) (S/N)/S

try to

(S/N)/(S/N)

(S/N)/(S/(S/N)) (S/N)/(S/N)

The set of syntactic categories is an index set; this means that the categories serve as indices for the set B of basic expressions (the lexicon). Thus Biv, for example, the set of basic expressions of category IV, comprises in PTQ the verbs run, walk, talk, rise and change (for the 62

other lexical items and expressions in addition to the ones given above see Montague, 1974: 250). It will be noted that both B e and B t are empty sets. In the case of B t this is due to the fact that no linguistic expression of category t, i.e. no sentence, is a basic expression; all sentences in the fragment of English considered by Montague consist of more than one basic expression. The case of B e is more complicated. If we look at the noun phrases in (ia-c): (1)

a. John squints. b. The man squints. c. Every man squints.

then both John, the man and every man should belong to the same category from a purely syntactic point of view, as they stand in a paradigmatic relationship to each other. But only John is a proper name. To bring out this syntactic parallelism between proper names and other noun phrases, Montague categorizes all these expressions as terms. While, however, only John and other proper names are basic expressions of category BT, other elements of the category T are compound expressions, generated by the syntactical rules. As to the semantic difference between proper names and other expressions of category T, this will be dealt with in the semantic part of PTQ. Thus the category e (or N) is eliminated from syntax and the category T (or S/(S/N)) introduced instead, a functor that makes sentences out of intransitive verbs or verb phrases. Before proceeding any further it might be helpful to parse a few expressions. Confining ourselves to proper names as noun-phrases and disregarding morphology (cf. 4.1.2), let us consider the following sentences: (2)

a. John sings. b. John finds Bill. c. John believes that Bill sings.

(2a) is straightforward and can be analysed as in (3) (bearing in mind that instead of e or N for John we have T or S/(S/N)): (3)

a. Johnj h.

sings \m

S/(S7N)

I

John

S/N

I

sings

63

The verb finds in (2b) is a two-place predicate, i.e. a functor with two terms as arguments: it makes a sentence when combined with two terms or, alternatively, it makes a one-place predicate when combined with its object-term (cf. 4.2.1 for a further discussion of these two categorizations of transitive verbs; alternative two is given below): (4)

a. Johnj b.

findsjy

Billy

.S/N

S/(S/N)

(S/N) / (S/ (S/N)

I

Bill

finds

John

S/(S/N)

The verb believes in (2c) finally is categorized like a transitive verb with the only exception that its second argument is an S instead of a term:

(5)

a. John-x b.

believes that\\,x S

Johny

sings\\>

S/ (S/N)

John

believes that

Montague's version of a phrase-structure grammar should be more properly characterized as a 'phrase-structure cum transformational grammar', incorporating like the standard version of categorial grammar transformational rules into the syntactic component. The syntactic rules of PTQ comprise five parts, (1) basic rules, (2) rules of functional application, (3) rules of conjunction and disjunction, (4) rules of quantification, and (5) rules of tense and sign. Basic rules. Basic rule 1 (S 1) states that every basic expression of some category X is also a phrase of the same category, i.e. a 'noun' is also a 'noun phrase', an 'intransitive verb' also a 'verb phrase' etc. Rule S2 states that combinations of every, the definite and the indefinite article 64

with a common noun yield phrases of the term-category; in Montague's formulation:' 3 (6)

S2. If t e P C N , t h e n F 0 (¡¡), F ,

F 2 (£) e P T ,

every £ F, (S) = the 5

w h e r e F0(£) =

F 2 (£) =

d £ o r ¿in £ a c c o r d i n g as t h e f i r s t w o r d i n £ t a k e s a o r

an. This rule enables us to parse expressions like (7)

The man sings.

While man is of category C N or S/N and will thus when combined with sings of category IV = S/N not yield a sentence, the expression the man as a whole will. The article and quantifiers generally convert a term (syntactically a noun phrase in subject and object position) out of a common noun, which, it will be noted, is syntactically equivalent to an intransitive verb or verb phrase; this is Montague's 'proper treatment of quantification'. We then have for (7):

The third basic rule introduces appositional sentences to common nouns; it states that when a common noun (more exactly a Pcn) is combined with a sentence (PT), then by the insertion of such that at the beginning of the sentence and appropriate pronominal changes in the sentence the whole expression is a common noun (Pcn)- Thus we get from e.g. the man (e Pcn) a n d he0 squints (e PT) the expression the man 13

M o n t a g u e (1974: 251). (6) can be paraphrased as follows: If some expression for instance man, is a c o m m o n noun (a phrase of category C N ) , then every man, the man and a man are terms (phrases of category T). The expressions every man etc. are written as functions, numbered f r o m o to 2, with an expression of category C N , in our example man, as a r g u m e n t - F 0 {man) etc.-, because this is precisely what quantifiers are: functions f r o m c o m m o n nouns to terms. The form of the indefinite article is regulated according to the initial sound of the w o r d that follows. Instead of a simple c o m m o n noun, X, could also be a phrase, e.g. musical man, man whom I met yesterday etc.; cf. b e l o w and 4.1.4.

65

such that he squints (E PCN)- THIS RULE IS A KIND of embedding transformation of relative sentences into NPs, with some modifications due to Montague's treatment of co-referentiality (cf. below the rules of quantification). Rules of functional application. Rules S4 to Sio correspond to rewrite rules in a categorial framework. Some of these rules also indicate morphological changes (amounting roughly to the agreement of the verbform with the subject-noun in S4 and the change from the nominative case to the accusative case when the personal pronoun stands in object position in S5). S4 specifies that the combination of a term (more exactly a PT) with a verb phrase (Piv) is a sentence (Pt); S j specifies that the combination of a term (PT) with a Vtr (PTV) is a verb phrase (Piv). S6 says that a preposition (PIAVAT) combined with a term (PT) gives a verb-adverbial (PIAV) and S7 that a verb taking a sentential object (Piv/t) combined with a sentence (Pt) yields a verb phrase (Piv)- In the same way, the concatenation of a verb taking an infinitival complement (Piv//iv) with a verb phrase (Piv) results in a verb phrase (Piv) according to S8. There are finally S9 and Sio. S9 states that a sentential adverbial (P^,) combines with a sentence (PT) to make a sentence (PT), while Sio says that a verb-adverbial (PIAV) concatenated with a verb phrase (PIV) yields a verb phrase (Piv)- It is interesting to note that this last rule incorporates a permutation transformation, which means that in PTQ 'functional expansions' are conceived of as expansions of linearly ordered elements: (9)

Sio. If 5 e Piv/iv and P 6 Pry, then F 7 (5, 0) e P I V , where F 7 ( 5 , P) = (3 6. (Montague, 1974: 251)

In other words: Given two expressions 8 and |3, 6 of category IV/IV, e.g. the verb-adverb beautifully, (3 of category IV, e.g. the intransitive verb sings; the combination of these two expressions-more exactly a combining function, application, with these two expressions as arguments, symbolized as F (6, p)-yields as value an expression of category IV, i.e. a verb phrase (sings beautifully). In this verb phrase the order of elements is P 5:

(.0)

sings

66

Rules of conjunction and disjunction. These rules state that the combination of two sentences by and or or is a sentence (Si i), the combination of two Piv's by and or or is a PiV (Si2), and the combination of two P T 's by and or or is a term phrase (Si3). They amount to a categorization of and and or as either t/t,t (=S/S,S), 14 , IV/IV,IV or T/T,T. Compare:

T

T/T,T

John

and

T

IV

Mary dance

IV/IV.IV and

IV sing and Bill

watches

Rules of quantification. Rules S14 to S16 have to do with problems concerning the scope of quantifiers and hence also of reference and coreference. Much discussed examples are sentences like: (12)

a. Every man loves a woman. b. John seeks a unicorn.

In (12a) a woman can either denote a specific woman loved by all men or else for every man some member (not necessarily the same) of the female species. The first reading can be symbolized by (13a) and paraphrased by (13b), the second can be symbolized by (13c) and paraphrased by (13d). (13)

a. 3 x (WOMAN(x) A Vy(MAN(y)

LOVE (y, x))

b. There is a woman such that all men love her.

c. Vy (MAN(y) —* 3 x (WOMAN(x) A LOVE (y, x))) d. For all men there is some woman which they love.

A similar ambiguity is found in (12b). This sentence has, as Montague points out (Montague, 1974: 255), both a non-referential or de dicto reading and a referential or de re reading. In the first case John is looking for any specimen of the species unicorn, something which might or might not exist; he is guided in his search presumably by some idea of what a unicorn ought to look like. Seek is here an intensional verb (see below). The phrase a unicorn is here used obliquely (Frege's ungerade 14

In more-place functors the arguments will be separated by commas. I will write S instead of t in order to make Montague's categorizations more immediately comparable to the ones proposed in the following chapters.

67

Bedeutung), it is, as Quine puts it, referentially opaque (cf. Quine, i960: 141 ff.). It might be helpful to quote Church in this context (1956: 8): A c c o r d i n g to the Fregean theory of meaning which w e are advocating, "Schliemann sought the site of T r o y " asserts a certain relation as holding, not between Schliemann and the site of T r o y (for Schliemann might have sought the site of T r o y though T r o y had been a purely fabulous city and its site had not existed), but between Schliemann and a certain concept, namely that of the site of T r o y .

But John might also be looking for a particular unicorn, in which case a unicorn refers to a specific entity and is used referentially. In order to capture this ambiguity, sentence (12b) can be analysed in two ways. For the non-referential reading the following structural tree can be graphed: (•4)

CN

John

seeks

a

unicorn

For the referential reading, on the other hand, Montague postulates a function S14 which converts terms and sentences into sentences, with the specification that-by some kind of a generalized transformation-the term (in our example a unicorn) be inserted into the place occupied by a pronoun in the 'embedded sentence' (in our example John seeks himo).1' This gives the following structural tree (cf. Montague, 1974: 255):

15

Personal pronouns are conceived of as proper names, which is the reason why they have subscript numbers in Montague's grammatical model.

68

To bring out the function of S/T, S, one could perhaps paraphrase (i 5) as (16)

There is a unicorn and John seeks it.

This example shows incidentally that the syntax of PTQ admits of structurally ambiguous sentences (as well as semantically ambiguous ones). In addition to S14 there are two further rules which combine a term phrase with a PCN containing personal pronouns (Si 5) and with a PIV containing personal pronouns (Si6), yielding a PCN and a P I V , respectively.' 6 Rules of tense and sign. Rule Si 7 specifies that the combination of a PT and a Pjv is a sentence also when the verb is put into the negative form, into the future form, into the negative future form, into the present perfect form and into the negative present perfect form (Montague, 1974: 252-3; cf. also below 4.1.3). Intensional logic. It is when we come to the semantic part of this grammatical model, that Montague goes markedly beyond the standard version of a categorial grammar. In PTQ the semantic interpretation is introduced indirectly. First an artificial language, the calculus of intensional logic, is developed; then this language is interpreted; in a third step the syntactic structures of English are translated into intensional logic and thus supplied with a semantic interpretation. Crucial for an understanding of intensional logic-and of formal or model-theoretic semantics in general-is an understanding of the notions 'intension', 'possible world' and several related concepts. The difference between the intension and the extension of a descriptive expression has been briefly described already (cf. 3.1.1). It was said then that the intension of a predicate is a property (a universal), of a name an individual concept and of a declarative sentence a proposition. It is customary to relate the term 'extension' to Frege's Bedeutung (denotation or reference) and the term 'intension' to Frege's Sinn (sense or meaning).'7 16

17

These rules, assigning referents or rather other names as co-referents to personal pronouns, also include morphological specifications as to the correct form of the pronoun. For a more detailed explanation and justification of S 1 5 and S 1 6 see Partee (1975: 234-6). Cf. Frege (1892); for a discussion of these terms see Church (1956: § 01), Lyons (1977: chapter 7).-Well-known is Quine's critique of the term 'proposition' when defined as the meaning of a sentence; cf. Quine (i960: § 42), (1970: iff.). It is based on the following reasoning: It is traditionally argued that synonymous sentences express the same proposition (i.e. John squints and John schielt for instance express the same proposition). According to Quine's behaviouristic position this synonymy of sentences (in distinction to the synonymy of smaller linguistic expressions) cannot be tested, as there are no clues available to the observer to decide whether two sentences spoken in

69

Disregarding the tangle of problems connected to this terminology and equating for the moment 'intension' with 'meaning', let us first consider the meaning of (declarative) sentences. Knowing the truth-value, i.e. the extension, of a sentence is obviously not equivalent to knowing its meaning. I might not know whether a sentence like (ia) John squints is true or false and yet I know what it means. This knowledge of the sentence's meaning depends largely on my knowing what squint(s) means. Here, too, a purely extensional point of view is not sufficient for determining the meaning of the predicate-expression. It might be the case that the predicate SQUINT is co-extensive with the predicate B A L D . In this case, both predicates denote the same class of individuals, while the corresponding predicate-expressions squint and bald are by no means synonymous or in any other way semantically related. Without going into epistemological questions and leaving some modifications till later, we might say that our knowledge of the meaning of squint consists in being able to correlate certain phenomena of non-linguistic reality with this linguistic expression. This knowledge has to be acquired; a speaker of English can be said to have attained this knowledge, when he uses the linguistic expression squint-in its various forms, including, of course, derivatives like the verbal noun-correctly. It is thus languagerelative knowledge. As Wittgenstein puts it in his Blue Book: But if w e had to name anything which is the life of the sign, w e should have to say that it w a s its use. T h e sign (the sentence) gets its significance from the system of signs, from the language to which it belongs. R o u g h l y : understanding a sentence means understanding a language.

(Wittgenstein, 1 9 6 9 : 4 , 5 )

We can articulate this knowledge by describing what squinting is, i.e. by using other linguistic expressions correlatable to squinting or parts or aspects of it. Those other expressions, however, are not the meaning of squint. This is so, even if the paraphrasing linguistic expressions belong to some artificial, constructed language like the calculi of logic or the pseudo-lexemes (predicates) postulated in componential analysis or generative semantics (compare Lewis' invective against 'markerese' in Lewis, 1970: 18-9). the same situation have the same meaning. Model-theoretic semantics, however, adheres to the distinction between sentences and propositions, defining the latter as non-mentalistic, objective concepts; cf. Church (1956: § 04).

7°

The ability to use squint etc. and analogously also John correctly (we will come back to proper names) is a prerequisite to the understanding of the meaning of John squints. If we know the meaning of John squints, we also know under what conditions this sentence will be true-which is not the same as knowing whether it is true or not. It is not true, for instance, if John's eyes are completely normal or if he rolls his eyes in a funny way etc. etc. Model-theoretic semantics equates knowing the conditions under which a sentence will be true with being able to assign a truthvalue to this sentence in all possible circumstances, or rather in all possible worlds. In the words of D. Kalish: To understand a sentence is to know under what conditions it would be true and under what conditions it would be false; that is to say, to know the sense (meaning) of a sentence is to know not its actual truth-value but its truthvalue in each possible world. (EPh s.v. "Semantics": VII, 354) This brings us to the notion of a possible world. This term, originating with Leibniz (cf. Mates, 1968) was revived by Carnap and defined as a state-description in Meaning and Necessity. Carnap's state-description is relative to a particular language, in Carnap (1956) relative to an interpreted logical calculus S,. This calculus consists of variables for individuals, the usual logical connectives (negation, conjunction etc.), the existential and the universal quantifier, two operators to be discussed later (the iota- and the lambda-operator; cf. below), descriptive constants for individuals and for predicates. An atomic sentence is defined as a sentence consisting of an n-place predicate plus n individual constants (cf. Carnap, 1956: 5). If there are, as Carnap stipulates, finitely many predicate signs and possibly infinitely many individual signs, S, may contain infinitely many possible atomic sentences. Out of these sentences classes can be formed. A state-description in S, is then defined as "a class of sentences in S „ which contains for every atomic sentence either this sentence or its negation, but not both, and no other sentences"; such a class of sentences "gives a complete description of a possible state of the universe of individuals with respect to all properties and relations expressed by predicates of the system. Thus the state-descriptions represent Leibniz' possible worlds or Wittgenstein's possible states of affairs." (Carnap, 1956: 9) A more ontologically oriented definition of a possible world postulates, in the same atomistic vein, a set B of 'basic particular situations', i.e. the set of all space-time points. A possible world is then defined as any subset w of B (cf. Cresswell, 1973: 37ff., with further elaborations). D. Lewis (1970-1: 24) states simply that "a 71

possible world corresponds to a possible totality of facts, determinate in all respects."18 Although this can be no more than an informal introduction to highly complex and controversial issues, we are now in a better position to pursue Montague's argumentation. Montague's intensional logic has four subparts, (i) the definition of types, (2) a lexicon, (3) a recursive definition of the meaningful (well-formed) expressions of intensional logic (its syntax), and (4) an interpretation or intensional model. Types. There are first the two types e and t (now called types not categories); furthermore 'functor-types' written as ordered pairs ( a , |3), where a is the argument-type and P the value-type. This notation corresponds to the notation |3/a, a notation which will be generally used here for simplicity's sake. There is finally a specific functor-type (s, a ) , where a is any type and s a third primitive type besides e and t (to be explained presently). Lexicon. The 'lexicon-rule' specifies that "denumerably many variables and infinitely many constants of each type" are employed (Montague, 1974: 256). Syntax. The syntax of intensional logic is somewhat simpler than the corresponding syntax of the fragment of English. The first rule states that every variable and every constant of some type a is a meaningful expression of type a, abbreviated as ME a . Rule (2) introduces the lambda-operator. This operator is customarily used to convert a sentence with a free variable into a predicate.1' If we have the sentence 'x squints', then '1.x (x squints)' can be paraphrased as "(belongs to) the class of x such that x squints" (extensional interpretation) or as "(has) the property of squinting" (intensional interpretation). As the whole sentence is of category S and the expression 'Xx (x squints)' is a predicate, i.e. of category S/N, the lambda-operator can be categorized as a functor with a sentence as argument and a predicate as value:20

/.x 18

19

20

x

¡quints

For a detailed discussion of the notion 'possible world' see Link (1976: 75-109). For an elementary exposition cf. Bradley, Swartz (1979: 4ff.). A variable is said to occur freely when it is not bound by a quantifier. Thus in 'x squints' x occurs freely, while in ' V x (x squints)', "for all x: x squints" or "every x squints", x falls within the scope of the universal quantifier. For simplicity's sake I will talk of categories when strictly speaking the term 'type' should be employed.

72

Montague, like Carnap, generalizes the lambda-operator, applying it to any expression with a free variable irrespective of whether it is a sentence or not (cf. Carnap, 1 9 5 8 : i 2 9 f f . ; Montague, 1 9 7 4 : 256f.). Rule (3) corresponds to the rules of functional application S4 to Sio. It says that the combination of a M E of type b/a-ME< a > b ) - w i t h a M E of type a - M E a - y i e l d s a M E of type b - M E b : (18) b/a

a

According to (4) the concatenation of two M E s of the same type by means of an equation sign results in a M E of type t (a sentence); these sentences can be paraphrased as " a is identical to |3", where a and P belong to the same type. Rule (5) introduces the logical connectives, quantifiers and modal operators. It states that if 9 and i|> are sentences, then the following expressions are also sentences (more properly, are of type t): ~cp ("it is not the case that cp"), [cp A ij>] ("cp and op"), [cp v

("cp or tJj"), [cp —> i|>]

("cp implies ijj"), [cp ip] ("cp implies ip and ip implies cp"), • cp ("it is necessary that cp"), Wcp ("it will come about that cp") and Hep ("it came about that cp"). If cp is a sentence and u a variable, then the following expressions will be sentences: 3ucp ("there is a u such that cp") and Vucp ("for all u : cp") (cf. Montague, 1 9 7 4 : 256). In rules (6) and (7) the notations for intensions and extensions are given. If a is a M E of type a, then the intension of a is notated as f a ] . This intension is defined as a meaningful expression of type a/s (ME( S ; a )), i.e. as a function from s to a, where s stands for 'sense'. If on the other hand a is a M E of type a/s, then f a ] is the notation for the corresponding extension, i.e. a M E of type a. Converting an intension into an extension consists, so to speak, in eliminating s. This will be further explained in the semantic part. Semantics.

The semantics correlates the meaningful expressions to

their denotata. First, three sets are postulated: A, the set of individuals; I, the set of possible worlds; J, the set of time-points. The set of possible worlds in the wider sense (i.e. including the time-factor) is all possible worlds at all possible times; more exactly, this set is the Cartesian product I X J , i.e. the set of ordered pairs (i, j ) , where i is an element of I and j an element of J (for all elements of I and J). It is interesting to note that while the elements of these three sets are placed, so to speak, on the 73

level of reality, there is no explicit ontological theory connected to these objects. The sets are introduced as undefined primitive terms. What the interpretation then does is, strictly speaking, not to correlate elements of language with elements of reality, but rather elements of language with a model of reality. 21 In intensional logic there are four kinds of expression: basic expressions of type e ('proper names') and t ('sentences'); compound expressions of type b/a, where b and a are either simple types (e.g. when b/a = e/e, t/e etc.) or compound types (e.g. when b/a = (e/e)/t, (e/e)/(t/t) etc.); finally expressions of type a/s, where a is some simple or compound type and s a third simple type besides e and t. In order to give denotations for these four kinds of expression four rules are formulated (cf. Montague, 1974: 258). The first rule says that the denotation of an entity-expression is an individual. More formally, the set of possible denotations of type e with respect to A , I and J is identical with A . In Montague's notation: (19)

D ei A, 1. ] = A.

The second rule says that the denotation of a truth-bearing expression is in the Fregean tradition a truth-value. More formally, the set of possible denotations of type t with respect to A , I, J is the set {o, 1 } , where o symbolizes falsehood and 1 truth. In Montague's notation: (20)

Dt>

A>

J = {o, 1}.

The third rule states that the denotation of an expression of type b/a ( = (a, b)) is a function with the denotation of a as argument and the denotation of b as value. In Montague's notation: (21)

D(a>

b), A, I, J = Db, A, I, j D °. A, I, J

For an example let us take the conjunction and as a sentence-connector (although strictly speaking an expression like and does, of course, not belong to intensional logic; the logical connector A stands here in its place). This conjunction is of category t/t, t (or S/S, S or in Ajdukiewicz' notation cf. above 3.1.1): S„

and 21

Montague refrains from ontological remarks; compare his footnote 8 in PTQ; Montague (1974: 257). For more explicit comments cf. Cresswell (1973: chapters 3 and 7).

74

In S o / S , , S the value-category S corresponds to b, the argumentcategories S, and S2 correspond to a. The denotation of and is hence according to (21) a function from the denotations of S, and S2 to the denotation of S0. The denotation of S, is according to (20) either o or 1; similarly the denotation of S2 is either 0 or 1. The denotation of and as an expression of category S0/S,,S2 is now defined by stating that if the denotation of S, is 1 and the denotation of S2 is 1, the denotation of S c is also 1; in all other cases the denotation of S0 is o. This is to say that the extension of the sentence-connector and in a truth-functional model is a truth only if both its argument-sentences are true. Finally, the denotation of an expression of type a/s (= (s, a)), i.e. the denotation of the intension of a meaningful expression of type a, is the denotation of the expression of type a under all possible circumstances. More formally, the set of possible denotations of type (s, a) is a function with I x J , the possible worlds in the wider sense, as argument and the set of denotations of a as value. In Montague's notation: 2

(v 2 3•>')

D ,( s , a ), , ,A ,, I,. J =

0

Da,

A , I, J

If a is an entity expression, then its intension is an individual concept; that is by (23) the individual is related to all possible worlds, all possible circumstances ('Napoleon in 1980', 'I in a world without decay' etc.). If a is a truth-bearing expression, then its intension is a proposition; that is by (23) the truth or falsehood of a (declarative) sentence is related to all possible worlds, all possible circumstances. Montague then defines the interpretations of the various meaningful expressions generated by the syntactic rules of intensional logic. These are based in a straightforward fashion on the denotation assignments above. To give just two examples: If qp is an expression of type t, then the extension of ~(p is 1, if and only if the extension of cp is o; the extension of Dcp ("it is necessary that (s: (s/sys)

where '(S/S)/S' —> {because, although, when...} The phrase-structure grammar generating the deep structures of our fragment of English would then have the following rules: (43)

(i) 2 ->• S. US'

(iii) S/S

OR: (ii) S —»i S.S/S \ \ s.'sn/s' ( N.S/N.('S/nS') . 'S/tS'. CS/mS')l

— i S/S J'S/S' f S.'(S/S)/S'

(iv) N

i'N' \ jN.N/N [ (C.'N/C'Î

(v) C

CC Ic.c/cl

(vi) N/N

S.'(N/N)/S'

(vii) C/C

-» t'C/C' [='AA']. ('AA/AA'))

iS.S. S/S,S'{

where S/S,S

—»

{and, or ...}

We thus get for (49a)-where the comma in writing, corresponding to a (potential) pause in speaking, is an alternative surface realization of and-the structural tree (49b): (49)

a. John won the race {and} Bill got the prize and Jim didn't get anything. >

S

John won the race

120

•S/S.S"

rd)

S

'S/S.S'

s

BUI got the prize

and

Jim didn't get anything

I

Subject- and object-clauses are conceived of as nominalized sentences possible with certain predicates (cf. 4.2.3). As has been mentioned already, while the sentence complements of verbs like think or realize are propositions, the sentence complements of performative verbs are illocutionary acts rather than simply propositions (cf. 4.1.3). Rule (iv) will therefore have to be elaborated in roughly the following way: (50)

(iv') N where j'N/Sj ' N/2''

jS.'N/S' < f z . 'N/2'l [Nom]

Compare: (51)

a. That John is stupid is obvious.

b.

[Pres]

John is stupid

[Nom]

obvious

c. Mary doesn't think that John is stupid.

d.

S

'S/„S'

•N/S' [Neg]

[Pres] Mary

think 'John is stupid

[Nom] 121

e. John wonders whether Mary likes Bill. f.

'N/T

[Pres]

John

wonder

[Nom]

Mary likes Bill

When looking at rule (viii), it will be noted that there is a discrepancy between (36) and (43) as regards the categorization of transitive verbs. In (36) transitive verbs are defined as two-place functors, in (43) as oneplace functors with a complex value-category. The two categorizations are, however, equivalent; for any functor Y / X , X we can write (Y/X)/X, for any functor Y / X , X , X we can write ((Y/X)/X)/X and so forth for many-place functors (cf. Curry, 1977: 32):

S/N,N

(S/N)/N Despite this equivalence, there seems to be a difference between (52a) and (52b), at least at first glance: (52b) reflects the dichotomy of subject and predicate-something is predicated of N , - , while in (52a) something seems to be predicated of N , and of N 2 . Here, however, we have to distinguish between various kinds of transitive verbs. Verbs like resemble or equal express a symmetric relation-if John resembles Bill, then Bill resembles John as well-, but most transitive verbs are not symmetric-if John loves Mary, Mary does not necessarily love John-, many even asymmetric-if John's height exceeds Bill's, then Bill's cannot exceed John's (for a formal definition of these relations cf. Wall, 1972: 122

noff.). Furthermore we have to distinguish between the use of a transitive verb in an active and in a passive sentence. This use does not affect the symmetry or non-symmetry of the relation; read for instance is still a non-symmetric relation between the reader and the thing read, even if the thing read is the subject of a passive sentence, i.e. the first argumentN in the linear order of the sentence. There is, then, a logical and a linear order of arguments. Any twoplace predicate has a first and a second argument, the first being so-tospeak the logical subject of the sentence. In an active sentence this first argument is also the first argument in the sentence and hence the grammatical subject, while in a passive sentence the (logically) second argument takes the first place in the sentence. The active sentence is topically neutral as it reflects the logical order of arguments, while the passive sentence is topically marked (although often more natural, especially in the case of an indefinite first argument).34 From this it follows that (52a) is indeed equivalent to (52b) also in the sense that something is predicated of N „ insofar as N, is the first argument of the predicate. It is only when writing out a phrase-structure grammar as in (43) that it is more convenient to have an overall category S/N. As (43) generates deep structures without regard to pragmatic aspects, it might be preferable to categorize two-place predicates as twoplace functors. This becomes even more desirable in the case of threeplace verbs (e.g. give), where it seems artificial to expand the category of the transitive verb (in one categorization) into an N and a three-place verb; clearly (53a) is to be preferred to (53b): (53) S/N,N,N

((S/N) / N ) / N 34

For the passive in English cf. Svartvik (1966); for the typology of passive constructions cf. Khrakovsky (1973) and for the analysis of the passive in a variety of languages cf. Xolodovic (1974). The passive in English is treated within the framework of Montague grammar in Partee (1973/76: 66), Bennett (1976: 141-4).

123

The logically first argument of a transitive verb is normally called an agent, the second an object. As Saumjan has shown, the semantic roles the arguments of a predicate perform in a sentence can be incorporated into a categorial grammar (cf. 3.1.2). Instead of writing N, and N 2 we could therefore adopt Fillmore's analysis of the sentence in terms of deep cases and write N a and N 0 . 35 This notation guarantees that the various other categories which are formulated irrespective of the role an N performs in the sentence need not be changed (e.g. the is a C/N regardless of whether N is agent, object etc.). Alternatively, one could indicate the case-frames in the lexical entry of the individual verb (cf. Jackendoff, 1972: 38). The respective merits of these two alternatives cannot be explored in this context; in order to simplify the syntactic component needed for an analysis of the de-adjectival abstract, alternative two will be favoured in this book (cf. below 4.2.3). Instead of (ii) we have therefore: (54)

(ii') S —» /. . . ) L N.S/N ) jN.-N^S/N.N | . CS/ n S').'S/ t S'. CS/n.SV ( ( N,.Nj.N,.S/N,N,N / J

Analogously (viii) will have to be emended: (55)

(viii') S/N

t I ('IV' . ('IV/IV'))

(viii") S/N.N ->• 'TV' . ('TV/TV') (viii"') S/N.N.N -> 'DV' . ('DV/DV) This amendment does still not cover all the facts. Besides di-transitive verbs like give (abbreviated as DV), there are complex-transitive verbs (cf. GCE: 40) like elect as in The committee elected John president. Here, a case-notation might be useful; while di-transitive verbs could be categorized as S/Na,N0,N' cf. also Nolan (1970), Leeman (1973), Ungeheuer (1969). See also in this connection the procedure in Leech (1974: 263ff.).

141

(89)

If 6 e Ptv, then FWI (6) € PADj (where ADJ = tlilt), and FW1 (8) = 8 + able. Translation: X x 0 3 y [ 6 ' ( y , P[P{x}])]

The first part of (89) formulates what a rule like (86) formulates; an expression 5 of category TV is given, the function F WI (a word-formation function) with 6 as argument has as value an expression of category ADJ (defined as a predicative adjective; for the three dashes as symbolizing subcategorization cf. above 4.1.2), namely the expression b+able. The meaning of b+able is specified by translating it into intensional logic. The formula in (89) can be best understood by proceeding in steps. Taking 6 as a variable which stands for instance for decipher, 3y[S(y, x)] reads "there is a y such that y deciphers x". With the modality operator O this gives "it is possible, there is a y such that y deciphers x". 4S The lambda-operator was explained above as making predicates out of sentences (cf. 3.2.1, 3.2.2); we then have for the formula Xx03y[8(y, x)] the paraphrase "the class of x such that it is possible, there is a y such that y deciphers x". Predicates like decipherable define in other words the class of objects of which it is possible that someone (y) deciphers them; put intensionally, they define the property of objects to be such that they can be deciphered by someone. In the formula in (89) 5' symbolizes the translation of 6 into intensional logic and P [ P { x } ] is the intensional formulation of x, reading as "the class of properties characteristic of x", i.e. denoting the individual concept (intensional) rather then the individual (extensional) (for the reason behind this cf. 3.2.1 and below 4.2.3). Alternatively, a meaning can be assigned to -able by the following rule (to be modified and made more precise in 4.2.3 and 5.2.3): (90)

The denotation of -ablep\/rv is that function from the denotation of TV to the denotation of PA such that S, S2, where

S, = [N1.['TVj,.[-a^/e].pA/TV']'PA,]s S2 = [Poss]. S / m s'-[[Someotte} N ..N i .TV | '] s

(90) states first that -able by its categorization denotes a function from TV to PA, and secondly that this denotation is of such a kind that S, S,. This means that once a denotatum is assigned to -able qua the syntactic category it belongs to, this denotatum can be further specified by replacing the structure -able occurs in by a structure conforming to S2 and semantically interpreting this structure. Taking any adjective ending in -able, we can say that qua predicate it defines (in extensional terms) 45

The operator O , denoting logical possibility, is incorrect here; cf. above.

142

a class of objects, the class of breakable, decipherable ... things. In fact, PA by itself defines a rather comprehensive class K of objects, the class of objects that can instance a quality expressed by an adjective. As adjectives in -able can be segmented into T V and P A / T V , the class K characterized by PA can be limited to that subclass of K which is formed from mapping the class characterized by T V into K. Transitive verbs denote ordered pairs of objects, objects that stand to each other in the relation expressed by the individual transitive verb (the class of all pairs x and y, where x loves y, x hits y etc.). This subclass of K is, however, still larger than the class adjectives in -able denote; it contains also the objects characterized by all other de-verbal adjectives whose base verb is transitive, the class of objects denoted by adjectives like defiant, observant etc. The interpretation of the category alone does therefore not suffice to specify the meaning of -able- adjectives. It is for this reason that the additional information provided by the paraphrase of S, by S 2 is needed. By (90) we then get the meaning of an individual adjective regularly employing the suffix -able when interpreting semantically S 2 and looking up the lexical entry for TVj. Before pursuing the problems raised in the preceding paragraph (the interpretation of syntactical categories, the make-up of lexical entries) in the next section, the question of lexical idiosyncrasy and morphological opacity must be briefly touched upon. As in the deep structures generated by (43) lexemes are left unanalysed, the lexicon will have to give a lexical entry for every lexeme. Making use of (90), we get for breakable for instance the following information:46 (91)

breakable: [TV]PA, with TV = break, PA/TV = -able.

If we consider breakable as completely regular, then the meaning of the lexeme can be ascertained by checking the lexical entry for break and for -able as in (90). If on the other hand there are any irregularities, these can be noted in the lexical entry. For breakable a more adequate analysis might have to specify that in S 2 'TVj' should be replaced by [

iji

•)hn is Bill's father

Bill is John's son

The logical formula for (102) is (103a), which can be transformed into (103b) when semantically analysing the relations of fatherhood and sonship in p and q (103c): (103)

a. •

(p

q)

b. D ( p - > p ) c. x is the father of y = y is the son of x

In (103b) the modal operator is strictly speaking superfluous, because (p —> p) is by itself logically true or necessary. This observation raises a host of questions, which belong to the philosophy of logic and cannot be gone into here; 54 suffice it to say that in a more comprehensive categorial grammar of English the modality component as outlined here might have to be considerably revised. Even more problems are raised by the interpretation of modal operators that are postulated on the basis of the system of modal auxiliaries in English. In (104) there is some modal operator of moral obligation: (104) John had to invite his parents to his wedding. For this operator-let it be symbolized by MN-the following interpretation could be suggested: (105)

The denotation of M N in relation to M and U is the function from the denotation of S1 relative to M and U to the denotation of S 2 relative to M and U such that S 2 = T , if there is some behavioural standard (moral code etc.) according to which what is stated in S, takes place.

Of the various other functors represented by non-lexemes, some are of a purely transpositional kind, such as [Attrib], [Rel] or [Nom]. These 54

C f . Montague's interpretation of • in Montague (1974: 259); for an interpretation of modal operators cf. also Bradley, Swartz (1979: 323ff.); for a résumé of the discussion about modal logic cf. Haack (1978: 170ff.).

152

can be fully characterized by interpreting their respective syntactic category (cf. the first set of rules; for [ N o m ] cf. also below 5.2). Others have to be supplied by paraphrases. A m o n g these, the 'Genitive-functor' might serve as an illustration. In 4.2.2 (68) the genitive-more properly: the possessive genitive-has been categorized as representing the category ' ( C / C ) / N \ i.e. as converting terms into modifiers of common n o u n s . " If w e take the linguistic expressions in (106) as paraphrases of one another, then the functional structures (107a) and (107b) can also be taken as being paraphrastic: (106)

a. b. c. d. e.

The The The The The

man's house house house house

house of the man (which) the man has (which) the man owns (which) the man possesses

(107) 'N/C

(C/C) /N'

The b.

house

[Gen. Poss.]

the man N

'N/C'

'(C/C) /S'

has th£ house

[Rel]

owns \possesse& 5S

C f . Weinreich (1972: of the Latin genitive ch.i 1 ; for a typology objective genitive cf.

1 1 4 ) for the polysemy of the genitive and cf. the 'classic' analysis by Groot (1956-7). For the genitive in English cf. also MSSE: of the possessive genitive cf. Ultan (1978). For the subjective and 5.2 below.

153

On the basis of (107) the following interpretation of the possessive genitive can be given: (108) The denotation of [Gen. Poss.] in relation to M and U is the function from the denotation of N; relative to M and U to the denotation of C/Cj relative to M and U such that S, *-* S2, where S, = [...[... [N;.[Poss. Gen.]c/cj]N • • -]s S,= [ . . . [ . . . [ [N|

( has i Cj]s . [Rel]]C/c,]N • •-]s < owns ?

In (108) has, owns, possesses (or more correctly have, own, possess) are synonymous and themselves to be interpreted. This interpretation, which also shows their synonymity, is provided by the third part of the semantic component. Here lexical entries for the lexemes are given. These contain the following items of information: (1) the phonological/phonetic structure of the lexical item; (2) its syntactic category, including its sub-lexemic categorization (cf. 4.2.2); (3) the semantic interpretation proper (which might not have to be given in the case of derived lexemes). In (3) the 'lexicographer's meaning' (cf. 3.2.2, note 27) is specified, based on paraphrases and analytic procedures like componential analysis. In the light of what has been said at the end of 3.2.2, these meaning-specifications are part of a vast synonymical and paraphrastic system (in the objectlanguage and possibly also a meta-language postulating pseudo-lexemes like C A U S E etc.), which is ultimately related to its denotata only by the interpretation of the syntactic categories these various lexemes and pseudo-lexemes represent. Part (3) of the lexical entry has therefore basically the same form as the interpretation-rules of derivational and grammatical categories, with S, containing the lexeme and S 2 a paraphrase with—in the spirit of componential analysis-more basic lexemes or pseudo-lexemes. Thus the lexical entry for woman might be for instance: (109)

1. woman ['wuman] 2. C 3. The denotation of woman in relation to M and U is the function from D $ ' u to DSM'U such that S, «-> S2, where S, = [ . . . [ . . . [ ® o m u n } c ] s / N ] s

S2 = [...[... [human and female and di/«/t]pA]s/N]s What (109) states is in other words that woman defines a class of individuals and that the predicate (is a) woman is synonymous with the predi-

154

cate (is) human and female and adult, i.e. that the class defined by woman is the intersection of the classes defined by human, female and adult. How these predicates can be further defined a look-up of their lexical entry will reveal, which will in turn lead to a further look-up of lexical entries, ad infinitum, unless semantic primitives as undefined (or only deictically definable) (pseudo-) lexemes are postulated (cf. 3.2.2). As semantic roles were not included in functional structure, they have to be given in the lexical entry of predicates. The subcategorization of predicates according to characteristics of their arguments such as 'doer', 'receiver' etc. presupposes a semantic analysis of the predicate in question. In order to postulate the cases 'agentive' and 'objective' (or 'neutral') for the arguments of for instance open, the sentence (no)

John

opened the door.

has to be analysed into two components: (1) "somebody does something", (2) "something is/becomes something (i.e. open)". This, as Fillmore points out himself (1970), is, of course, the reformulation of a case-analysis in terms underlying the analysis proposed by generative semanticists. Case grammar and generative semantics depend both on sub-lexemic, semantic analyses (see in this connection also Chomsky's discussion (1970a)). It might therefore be possible to dispense with deep cases altogether in favour of fuller semantic analyses.'6 There are, finally, lexemes like fact, problem, true, order, ask etc. which have to be interpreted. While these words might be used as 'meaningful lexemes', they are 'semantically vacuous' as metalinguistic expressions. If we have (in)

a. It is true that John b. Mary is John's

loves

Mary.

true love.

then true in ( m a ) is a meta-linguistic word signalling that the truthvalue of the embedded sentence is T, while true in (11 ib) is a word of the object-language. Similarly, in (112)

a. I order you to leave this room. b. I ordered you to leave this room.

order in (112a) is a performative verb, while in (112b) it is a 'descriptive verb'. The denotation of true and order can then be given in the following two rules (for the interpretation of fact and problem cf. 5.2): 56

For a specification of lexical information cf. also Fillmore (1971b). Compare Jackendoff J s position in Jackendoff (1972).

155

(113)

The denotation of true in relation to M and U is the function from D n 1 u (more properly [S] N as true is predicated of sentences) to Ds 1 ' u such that S2, where

S, = [[S;]N.tr«e]s S 2 = S, (with DSM' U = T) (114) The denotation of order in relation to M and U is the function from Dn,' U and D # ; u (more properly [2] N z) to DSM'U such that S, deaf-muteness ~ muteness, colour-blind—> colour-blindness ~ blindness (for a discussion of compound adjectives in English and German cf. Lipka, 1966). They are analysable into the categories of the bases of the compound-«* and a in the case of deafmute, n and a in the case of colour-blind-plus a two-place 'composition functor' having a as value and the base-categories as arguments (cf. 4.2.2, note 43). Compounds cannot be incorporated into derivational graphs, but must be seen as the result of amalgamating two graphs: (7) 9

colour

blind

colour-blind

Disregarding compounds then we have the following types (compare also Stein, 1971: i8off.): (8)

Type 1: a + n/a

Type 2: n + a/n + n/a

160

Type 3: v + a/v + n/a

vegetate

vegetative •

vegetativeness

Type 4: a + a/a + n/a

bluish

# bluishness Type 5: n + n/n + a/n + n/a 0 1 > felon •l,felony cleverish pseudocleveris h —» pseudo-cleverishness). More important is the fact that the graphs plotted so far are imperfect in several respects. They are on the whole limited fragments of larger derivational graphs; the example used to illustrate type 9 could be expanded in the following manner: (9)

satis-

satisfactory

dissatisfy

satisfactoriness dissatisfaction dissatisfactoriness

Furthermore, derivational graphs often show multiple branching. When abstracts are derived from adjectives ending in -able, as a rule both the suffixes -ity and -ness are possible (though not necessarily of equal frequency):

167

Q

(io)

cipher

decipher

decipherableness •

•

decipherability

Connected to the problem of suffixal variation is the problem of homonymity. Here too multiple branching is typical. P. A . Soboleva points out that the verb to black has to be analysed into two homonyms on the basis of its two meanings, "make black" and "polish (boots etc.) with blacking" (cf. ALD

s.v. black v.t.) respectively, where the former is

a de-adjectival verb (blacky = "make black") and the latter a denominai verb (black 2 = "polish with blacking"). This is captured in graph ( n ) (cf. Soboleva, 1976: 106):

(>0

A s has been remarked above (cf. 3.1.2), derivational graphs like the ones presented here have been developed by linguists working with the model of applicative grammar. Here the basic concern is with Russian, a language marked by its derivational transparency, witness the 'derivational nests' listed in Ginzburg (1973). (Cf. also Soboleva, 1 9 7 3 ; Ginzburg, 1979.) The advantage of such a treatment of derivation is that individual derivational processes can be seen in the context of the derivational mechanisms of a language as a whole. A given derived w o r d - s a y 168

decentralizable-czn thus be related to some type, allowing not only the formulation of its derivational history-e.g. centre decentralizable—, but predicting also further buildable forms-such as decentralizability and decentralizableness. But English, as distinct from Russian, as a 'lexically mixed' language-i.e. as a language with a great number of loans from French, Latin and other languages-is only partially transparent; large parts of the lexicon are derivationally opaque, a fact which accounts for the many complications met with in the setting up of derivational graphs.2 Words are not always easily segmentable into derivational elements, and even if they can be segmented the morphological elements analysed are not always productive. This means that a derivational graph is often only a picture of the existing lexicon and not a truly generative device (for the difference between an analytic and a synthetic point of view in the study of word-formation see above 2, footnote z). This does not necessarily invalidate derivational graphs as a means of studying the structure of a lexicon; as long as the relationships between the various elements of a graph are definable, derivational graphs can throw some light on the derivational mechanisms of languages like English as well. These definitions rest on the segmentability of lexical items, a problem we will turn to now. At the beginning of this section four cases were differentiated, deadjectival abstracts identifiable as such only on semantic grounds (bliss), abstracts identifiable on morphological grounds with no base adjective (velocity), a primary adjective (badness) and a secondary adjective-comprising tertiary, quartary etc. adjectives-as basis (decentralizability). For cases one and two no revealing derivational graphs can be plotted, nor can restrictions on the suffixes-as far as they are recognizable-be formulated in terms of the morphological form of the base adjective. Case three has base adjectives, primary in English. This group of adjectives comprises both adjectives like bad, wise or noble as well as adjectives like agile or albescent. The primary character of the base adjective is, however, relative to the period of the language under investigation and, in the case of loans, to the giving or receiving language. While the adjectives wise and noble are primary in Modern English as well as in Old and Middle English and while noble is primary in Modern as well as in Old French, neither adjective is primary when going still one step further

1

For a general discussion of derivational patterns in 'lexically mixed' languages and a formal definition of derivational relationships see Reichl (forthcoming).

169

back in history: wise is a deverbal adjective in Germanic (cf. Meid, 1967: 55 [§ 65], 142 [§ 118.1]), Latin nobilis is derived from nosco (cf. Buck, 1933: 330-1 [§ 474.2]). There are two reasons against segmenting noble in Modern English into no- and -ble, the one being that the Latin suffix -bilis has normally the form -ablel-ible (and less frequently -uble) in English, the other that while Latin nobilis can be semantically segmented into "can be" (corresponding roughly to the suffix) and "known" (linking the word to nosco) this cannot be done in English. 3 More problematic are agile and albescent. Both adjectives could be formally analysed into a bound (root) morpheme and a suffix, agile into ag- + -He on the basis of contractile . . . ] . [p

...].. .] s , [ N o m ] . N / s . ] N

=> a) [(t£)

[...]S|]N/if[i]

b) [ . . . a + ( s ) . . . p + | i n g } . . . ] N / i f ( 2 ] c) [ . . . a . . . p + { i n g ) . . . ] n / if [3]

The structural description of (65) enumerates only those elements of functional structure which are relevant for the transformations, i.e. the nominalizing functor, the noun serving as subject of the sentence and the predicate-verb. All three derived structures are limited as to their availability by certain conditions symbolized by [1] to [3]. Informally, these conditions are for a): it is not possible if the whole structure has subject position; for b): the gerund is only possible for certain nominal structures of the embedded sentence, i.e. it is for instance not possible if the subject-N is modified by a relative clause; for c): this structure is only possible if the matrix-N has object position. Apart from the incompleteness of the derived structures-there are also nominalizations as in I don't like (it for) John to smoke thick cigars-, there are further restrictions and variants.27 The structures (65a-c) are not yet surface structures; the sentential operator regulating tense, for instance, will also have to be considered, the morphophonemic rules for {s} will have to be given etc. As the representation of the tense-operator is a transformation independent of nominalization and of a more general nature, it is presumably prior to the affixation of -ing to the verb. This means that there are intermediary structures between functional and surface structure and that the transformations are ordered.28 Rule (65) in its generality will also cover the cases where P is of category TV, DV, PA or T A ('transitive adjective', cf. below), although the morphological rules will of course differ in the case of predicative adjectives (copula in surface structure etc.). 27 28

Cf. esp. Lees (1960), Rosenbaum (1967), MSSE-. ch. 8, Dubois-Charlier (1971). For a detailed discussion of transformational rules within the extended standard theory cf. Chomsky (1977: 81 ff.); for a succinct discussion of the ordering of transformations cf. Grinder, Elgin (1973: 135—148). On the raising process in (65c) see the extensive study by Postal (1974).

203

There are two further cases of sentential nominalization which ought to be mentioned briefly: (66) a. I don't like to be stupid / being stupid. b. To be stupid/Being stupid can be an advantage. In (66a) the subject of the embedded sentence is identical to the subject of the matrix sentence; it is therefore deleted by the transformation of Equi-NP deletion well studied in the framework of generative grammar. 2 ' This process would have to be added to (65). In (66b) an indefinite subject-noun of the embedded sentence has been deleted, an analysis suggested by paraphrases like: (67) a. For anyone to be stupid can be an advantage. b. Anybody's being stupid can be an advantage. Here, too, this further optional transformation has to be added to (65). Abstracts, both deverbal and de-adjectival (as in (5 2d, h)), although left unanalysed on the level of functional structure, are yet related to the corresponding nominalized sentences and gerunds. It can be argued that in the context of a factive predicate (such as is regrettable in (j2f-h)) an abstract admits of an interpretation which makes the N it occurs in virtually synonymous with a nominalized sentence: (68)

a. I regret that Mary {departed.) Us ill. i b. I regret Mary's i departure, i t illness. )

Factive abstracts (cf. below) differ, however, from nominalizations in that the category of tense is not expressed (Mary's departure could also correspond to that Mary is going to depart; cf. also the relationship between unfortunately and unfortunate in 4.2.2 (71), (72)). Although fully analysed only in the lexicon, the derivational structure of abstracts can be signalled in syntax by employing the categorization [PA] N or [S/ a N] N and [TA] N or [S/ a N,N] N for de-adjectival abstracts, abbreviated as N a (cf. 5.1.2 (22i)). A similar categorization can be applied to deverbal 29

C f . esp. the analyses of Rosenbaum (1967), Lakoff (1968), Bonney (1976), Menzel (1975), Bresnan (1979) in the context of complement structures. For an analysis within the framework of Montague grammar cf. Partee (1973/76: 6/ff.). Arguments against the E q u i - N P deletion transformations are advanced in Brame (1976).

204

abstracts ([IV] N , [TV] N etc.), but these will have to be left out of account here. The expressions that Mary is ill and Mary's illness will then be generated by the syntax of the given model in the following form: (69)

^ "N/S"

"N"

'PA'

Mary

ill

I

[Nom]

[Pres]

I

b. ^ N / N ^ ^ ^ ^

I

[PAf

I

"N"

'(N/N) /N'

Mary

[Gen. sub.]

illness

It is when the semantic interpretation of (69a) and (69b) is given that their relatedness can be made explicit; the derivational structure of illness allows an identification of the base with ill in (69a), while the semantic interpretation of [Gen. sub.] reveals the sameness of the relationship between Mary and being ill in (69a) and Mary and illness in (69b). It is consequently in the lexicon/semantic component that (69a) and (69b) are shown to be systematically related. This procedure applies also to morphologically opaque or problematic abstracts. Here, as in (70) (for (70b) see p. 106),

N/N

[PA]N

Bill's

speed

A

I

the semantic synonymy relationship between speed, velocity, swiftness and fastness, and the derivationally encoded semantic relationship be205

N/N

[PAf

The train's

velocity

tween speed and speedy ("having speed") help to interpret (70a) as being related to (71a) and (70b) as being related to (71b): (70 'N/S'

[Nom]

Bill is speedy

b. 'N/S'

[Nom]

More details on semantic interpretation will have to be deferred to 5.2.3; in the remainder of this section the syntactic structures de-adjectival abstracts occur in will have to be discussed. Parallel to the treatment of sentential nominalizations with deleted subjects of the embedded sentence as in (66)/(6y), generalized abstracts will be treated as structures where there is an indefinite genitive complement (subjective genitive; cf. below) in deep structure which can be optionally deleted. We have thus in (72a) the surface realization of a functional structure as proposed in (73b), in (72b) the surface realization of the same kind of functional structure according to (73c): (72)

a. Anybody's stupidity can be a nuisance. b. Stupidity can be nuisance.

(73)

[[•••«iv-On- [[Gen. sub.] . [...Pw•-]n]n/n]n =>

a) [[the...a...]

206

+of+

[,..p...]]N

b) [ [ . . . ß + { s } . . . ] + [ . . . « . . . ] ] „ / i f [I] c) [,..a...]N /if [2]

[1] specifies the constraints on the prepositive genitive (for some of these see e.g. GCE: 198 ff.), [2] states that (73c) is optional if |$ = anybody (anyone) and P is not further modified. Class abstractions seem to vacillate between partial conversions such as the poor ("poor people as a class" according to the OED; cf. footnote 1 in chapter 1) and substantivizations like the beautiful ("the name given to the general notion which the mind forms of the assemblage of qualities which constitute beauty" according to the OED). But only in the first case is a paraphrase with "the class of (adj.) objects/persons" possible: (74)

a. The poor present a great problem to the Indian government. b. The class of poor people presents a great problem to the Indian government. c. The Romantics idealized the beautiful. d. The Romantics idealized the class of beautiful objects. e. The Romantics idealized beauty.

While (74a) is synonymous with (74b), (74c) is not synonymous with (74d), but rather with (74c), which shows that expressions like the beautiful are in fact synonymous with generalized abstract nouns. As to class abstractions proper, the definite article can here be viewed as a kind of generic article used to denote the whole class of poor people etc., i.e. as being equivalent to the universal quantifier (cf. 5.2.1). So far the discussion was in the main limited to 'one-place adjectives'. There are however adjectives which can be regarded as a kind of 'transitive adjective', such as fond of or aware of.}0 In order to arrive at the correct categorization of these adjectives and the corresponding abstracts a comparison with transitive verbs might be useful. If we have (75)

a. John loves Mary. b. John's love of Mary

we can analyse (75a) as (76a) and consequently (75b) as (76b):

30

Cf. König (1971); for an analysis of adjectival structures in German in a transformationalist framework see Mötsch (1964), in a logically oriented framework see Eisenberg (1976).

207

( 7 6)

a.

•N,'

John

'(N/N) /N"

[Gen. sub.]

[TV]N

love

N/N '(N/N) /N" I [Gen. ob.]

•n; I Mary

It is to be noted that the passivization of sentences with transitive verbs is not matched with passivizations of 'abstract structures' in a straightforward manner. While passivization is optional for most sentences containing a transitive verb (for constraints cf. GCE: 803 ff.), the interchange of N , and N 2 in the case of structures like (76b) is limited to certain lexical items; thus we have both (77a) and (77b), but not (77d), we have both (77c) and (77f), but not (77b) (this, too, is an argument for a lexicalist position): (77)

a. The enemy's destruction of the city b. The city's destruction by the enemy c. John's love of Mary d. ''The love of Mary by John e. The hunters' killing of the elephants f. The killing of the elephants by the hunters g. My wife's passion for grapes h. ''The passion for grapes by my wife

We therefore have to add a rule, allowing for the permutation of N , and N 2 in structures like (76b) in certain cases. This brings us to another point. There are two kinds of genitive in (76b), identically categorized but differentiated according to their argu208

ment-N; if it is N „ then we have a genitivus subiectivus, if it is N 2 , then we have a genitivus obiectivus. The former can be morphologically represented by both the synthetic and the analytic genitive, the latter only by the analytic genitive, exceptions being some passivized structures such as (77a). The preposition used with the object need not always be of; other prepositions are also possible (for, about etc.). This analysis of the genitivus obiectivus implies that the prepositions found with abstracts are syntactically different from the prepositions found with the respective adjectives. When combined with abstracts they have the function of making an N / N out of the object. When, however, found with the adjective they are functionally empty; it is a morphological idiosyncrasy of English that two-place adjectives cannot be combined with a noun as object without a preposition. That this preposition is normally the same as the one used with the abstract is a coincidence; it need not be so: compare for instance keen on, keenness for. The structures (79a) and (79b) for (78a) and (78b) respectively can therefore be proposed: (78)

a. John is fond of Mary. b. John's fondness for Mary

(79)

b.

/

'(N/N) /N) John

[Gen. sub.]

fondness

[Gen. ob.]

Transitive adjectives differ, however, from transitive verbs in that they cannot be passivized. One could argue that this is so because the objects 209

of two-place adjectives are syntactically not nouns but adverbials, as is suggested by the prepositional phrase of Mary in (78a) and the like. That this is not so can be easily demonstrated by looking at the seemingly two-place adjective brilliant: (80)

a. John is brilliant at maths. b. John's brilliance at maths

In (80a) the phrase at maths modifies the kind of brilliance John has got, i.e. it is a genuine adverbial, on the same level with adverbs like very, extraordinarily etc.; the same can not be said of of Mary, which does not narrow down in the same way the kind of fondness John has got. Modifiers like extraordinarily or at maths are adverbs when combined with the adjectival predicates, but attributes when combined with the abstracts. For the phrases brilliant at maths and brilliance at maths the following structural trees can be graphed: (81) 'PA'

I

brilliant

[PA]N . !

brilliance

PA/PA at maths

N/N at maths

The question arises here whether nominal modifiers of category N/N are syntactically derived and can be related to the adverbial modifiers of category PA/PA. If we look at the following expressions: (82)

a. John's extraordinary intelligence b. John's intelligence is extraordinary. c. John is extraordinarily intelligent.

we find that attributive adjectives modifying abstracts are both derived from predicative adjectives (82b) and from adverbs {82c); they modify in Bolinger's terms both the referent and the reference of their head (cf. 4.2.4; the term 'reference' here is, of course, different from the term 'reference' as understood in 4.2.1). This double function of modifiers stems from the meaning of abstracts, which is characterized as denoting 210

the 'reference' of the lexeme as 'referent' (cf. 5.2.3). It can be expressed by the syntactic derivation of the attributive adjective from the predicative one, analogous to referent-modifying adjectives (extraordinary intelligence ~ red flower), on the one hand, and by showing, on the other hand, that the category N / N is functionally equivalent to PA/PA, i.e. that the structures underlying (82a) and (82c) must be systematically related. This equivalence arises from the categorization of the de-adjectival abstract as [PA] N . If N / N modifies N and if N is derived from PA, then N / N corresponds to a modifier of PA, which, according to the categorial formalism, must be of category PA/PA. For John's extraordinary intelligence the following structural tree can therefore be graphed:

'(N/N) /S"

John's

intelligence

John's intelligence is extraordinary

[Attrib]

There are, however, adjectives which cannot be used predicatively (cf. (21) in 4.1.4): (84)

a. John's utter stupidity b. '•'John's stupidity is utter. c. John is utterly stupid.

It was argued in 4.1.4 that these adjectives cannot be syntactically derived from predicative adjectives and should be categorized as expressions of category 'N/N' or 'C/C'. This position will be maintained here, although a semantic analysis might show that adjectives like utter are predicates when taking their synonyms into account (complete, absolute), their restriction to an attributive position being possibly a surface phenomenon only. Two-place adjectives like aware (of) have a nominalized sentence or an abstract-structure as second term, i.e. they are of category S / N „ N 2 with

211

c. John's awareness of the fact that Mary is present d. John's awareness of Mary's presence (85c, d) has the following structure: (86)

John's

awareness

[Gen. ob.]

t the fact that . . .1 \ Mary's presence1

A further group of two-place adjectives and their abstract derivations are construed with an infinitive as object: 3 ' (87)

a. John is

/ ready \ eager anxious willing hesitant

b. John's

1 readiness \ eagerness | anxiousness/anxiety ( I willingness hesitancy

to leave,

to leave

T h e object position of the infinitive can be established by comparing these adjectives with verbs like want, try etc.: (88)

31

a. John b. John c. John d. John

tried to leave. tried it: i.e. to leave. is eager to leave. is eager for it: i.e. to leave.

For this type see Vendler (1968: 102-3); compare also Lees' analysis, cf. above 2.2.2. For the characterization of these adjectives as 'factive adjectives' taking an 'action complement' cf. Wilkinson (1970). For complement structures see footnote 29 above.

212

e. John f. John g. John h. John

hesitates to come. hesitates about it: i.e. to come. is hesitant to come. is hesitant about it: i.e. to come.

Here again, the object has to be introduced by a preposition (for it; but cf. willing it vs. hesitate about it) in the case of adjectives. If the infinitive as a whole is then of category N, what about its internal structure? One could, of course, propose the following analysis:

Or, taking over Montague's categorization in PTQ (cf. 3.2.1; cf. also Thomason, 1976: io2ff.):

(89) or possibly (90) seem reasonable analyses for adjectives like ready, willing, where the subject understood of the infinitive is identical to the subject of the sentence. In other cases, however, the infinitive can have its own subject, differing from the subject of the sentence: (91)

a. John is eager for Mary to pass her exams. b. John is anxious for Bill to get the job. 213

This is the reason why the analysis notably put forward by transformational grammar, i.e. the transformational derivation of the infinitive from a sentence (with Equi-NP deletion or F O R - T O transformation), is taken over here. This gives the following structures for (87a) and (87b): (92)

a.

John's

readiness [Gen. ob.]

[Inf]

John leaves

When specifying the infinitive-transformation (and F O R - T O transformation) it would have to be noted that certain adjectives demand a subject in the embedded sentence identical to the subject of the matrix sentence. It is typical of most of these adjectives with complement that they also have uses as one-place adjectives. In the latter case they ought to be categorized as P A , although one could also argue for a transformational object deletion as some kind of object is normally implied; cf.: (93)

a

- John wanted to marry, but he was far too hesitant (sc. about marrying someone).

But: b. John remained a bachelor; he is far too hesitant (as a person).

It is probably best to categorize these adjectives as S/N, (N), i.e. as twoplace adjectives with an optional second term. 214

According to this analysis of the infinitive, adjectives like aware and adjectives like eager differ in categorization-N 2 is a nominalized sentence or abstract in the case of aware, an infinitive in the case of eager-, while the corresponding abstracts occur in slightly different structures. If we amend (43 iv) in 4.2.1 to include the rule (94a), [N a ] N would have to be expanded as in (94b) for adjectives like aware and as in (94c) for adjectives like eager: (94)

a. N

-*•

.... |[NJ

b. [ N J N

. N

.N/N{

N [TA] . [[Gen. ob.l'(N/N) /N' . NJ]N / N l J LL J

where N, —>

where Nj

|J

/ the fact that Sj j S;. [Gerund] ([NJN Sj . [Infinitizer]

Note that in the case of infinitive structures the surface realization of the genitivus obiectivus is zero. There is a further group of adjectives with an infinitive complement; to these belong for instance easy, difficult and pleasant.*1 Leaving the infinitive complement out of consideration these adjectives are completely regular as one-place predicates: (95)

a. b. c. d. e. f.

This text is easy. The easiness of this text The exam was easy for John. The easiness of the exam for John The question is difficult (for us). The difficulty of the question (for us)

If the infinitive is added, however, certain irregularities are met with: (96)

32

a. Bill is easy to upset. b. It is easy to upset Bill.

C f . Vendler (1968: 97ff.) and, for a vigorous anti-transformationalist stance, Bolinger (1973). Some of the transformations involved are discussed under the heading of 'tough-movement' in the literature; cf. esp. Postal ( 1 9 7 1 : 17-31)» Postal, Ross (1971), Quirk (1977); for a treatment within the framework of Montague grammar cf. e.g. Partee (1973/76: 66).

"S

c. d. e. f. g. h.

*Bill's easiness to upset The easiness of upsetting Bill Homer is difficult to translate for me. It is difficult for me to translate Homer. Homer's difficulty to translate for me The difficulty of my translating Homer

Bolinger (1973) argues that in (96a) something is predicated of Bill-Bill is easy-to-upset-and that therefore, contrary to Chomsky's analysis (1970b), (96a) is not derived from (96b), where easy is predicated of upsetting Bill. Similarly, one might argue that in (96c) something is predicated of Homer, while in (96f) something is predicated of my translating Homer. This would account for the acceptability of (9d), corresponding to (96b), and of (96h), corresponding to (96f), as well as for the unacceptability of (96c), corresponding to (96a), and of (96g), corresponding to (96e). A s Bolinger has pointed out, however, structures like (96c) are possible (cf. Bolinger, 1973: 26): (97)

a. Dry clay is difficult to work with. b. The trouble is its difficulty to work with.

This means that the acceptability or inacceptability of (96c) is an idiosyncrasy on the level of surface structure; this being so, it cannot be accounted for by a deep structure analysis. I would therefore regard with most transformational grammarians (96a) and (96b), as well as (96c) and (96f), as 'generalizable', proposing for both the functional structure (98a), where an indefinite subject is assumed for (96a, b) analogous to the treatment of the infinitive in (66). The functional structure of both (96d) and (96)1) is then relatable to (98a) as (98b) (see p. 217):

[Pres]

i Someone upsets Bill i ( I translate Homer )

[I n f]

easy

(96c) and (96g), on the other hand, do not relate to (98a); they are only possible if structures like (96a, c) have become re-analysed analogous to 216

b. [pa r

(the) easiness [Gen. sub.] iSo. upsets Bill ï ( 7 translate Homer S (87a), with a 'passive infinite', a historical process accounting for grammatical idiosyncrasies here (cf. also Lightfoot, 1979: i86ff. for the reanalyses of the infinitive in English). There are, of course, also expressions like: (99)

a

- John's ease in solving the problem b. John's difficulty in solving the problem

As paraphrases such as (100a) suggest, the semantic relationship between ease/difficulty and John is not one of John's having a quality (John is easy/ difficult), but rather one of his possessing a faculty in respect to some action {John does something with ease/difficulty). When graphing the functional structure of (99a, b), in solving the problem would therefore have to be interpreted as an adverbial (more precisely a temporal) clause: (100)

a. Theiease ¡John haslwhemsolving the problem (difficulty (in i

John's

S

ease difficulty i

when John solves the problem

A last group of adjectives, of which likely and certain are typical examples, show similar behaviour. 33 The following sentences and nominalized adjective-structures are characteristic of these adjectives: 33

Cf. Vendler (1968: I04ff.), Chomsky (1970b: 191), Bowers (1975a: 3y4ff.). 217

(101) a. That John will marry Jane is certain/likely. b. It is certain/likely that John will marry Jane. c. John's marrying Jane is certain/likely. d. John is certain to marry Jane. e. John is likely to marry Jane. (102) a. The certainty/likelihood, that John will marry Jane b. The certainty/likelihood of John's/John marrying Jane c. John's certainty of marrying Jane d. *John's likelihood of marrying Jane The parallelism between (101) and (102) is somewhat misleading; while (ioia-e) can be considered as paraphrases of each other-as can be seen when substituting unambiguous expected for certain/likely-{ 1 o 1 d) is ambiguous. It can either mean that it is certain that John will marry Jane or that John feels certain that he will marry Jane. Accordingly, (102a, b) correspond to (ioia-e) with the first reading of (ioid), while (102c) corresponds to (10id) in the second reading, which is the reason why (i02d) is unacceptable. In the second reading of (ioid) certain of is a two-place adjective like fond of etc. We get consequently as functional structures for (101) and (10id) in the second reading:

The abstract structures in (102a, b) and (102c) are then represented by the functional structures (104a) and (104b) respectively:

'N/S' certainty

b.

John's

I

[Gen. sub.]

[Nom]

John will marry Jane

N

certainty

[Gen. ob.]

[Nom]

John will knarry Jane

Before turning to the semantic interpretation of the functional structures proposed for the de-adjectival abstract, some data on the syntactic context of abstract nouns will be given in this section. These data are based on the Brown University Corpus; they have been limited to abstract nouns in -ness and -ity. There are 1563 occurrences of lexemes ending in -ness and 4009 occurrences of lexemes ending in -ity in this corpus. Of these the following cases have not been taken into account: (1) Concrete nouns ending in -ness or -ity (e.g. business); (2) compounds with a word in -ness or -ity as first component (e.g. readiness exercises); (3) titles or phrases (e.g. for goodness' sake; pity, we...), where the words in -ness or -ity do not have their normal syntactic function.

This leaves then 948 occurrences of abstracts in -ness and 2525 occurrences of abstracts in -ity (including the use of abstracts with the indefi219

nite article and s o m e d o u b t f u l cases, w h i c h will, h o w e v e r , n o t significantly affect the ratios given below). In the formulas listed b e l o w ' M o d ' s y m b o l i z e s an optional expression f u n c t i o n i n g as a m o d i f i e r of its head n o u n (e.g. an attributive adjective, adverb plus adjective etc.); in rare cases it can be put behind its head n o u n (as in haughtiness personified). '(•••)' stands for b o t h an expression f u n c t i o n i n g as the s e c o n d argument of abstracts derived f r o m t w o - p l a c e adjectives (as that...

in my awareness that...)

and, in a f e w cases, also

for expressions functioning as a genitivus qualitatis (e.g. the flexibility of alignment w i t h a genitivus qualitatis vs. the flexibility of the alignment w i t h a genitivus subiectivus). 'Poss' s y m b o l i z e s a possessive p r o n o u n or the relative p r o n o u n whose, 'Rel' a restrictive relative clause and ' C o m p ' a functionally equivalent c o m p l e m e n t (as in an ability made possible = an ability which was made possible). T h e f o l l o w i n g syntactic constructions occur: (I)

The subject of the quality expressed by the abstract noun (Nab) is named (NP S or N s ). (a) the (Mod) N * ( . . . ) o / N P s Remarks: The cases where the definite article is implied by the context are also listed (e.g. badness in the goodness or badness of...). The definite article can be replaced by this (once in the case of nouns in -ity), that (once in the case of nouns in -ness) or zero (7 times in the case of nouns in -ness, 19 times in the case of nouns in -ity). (b)

.0 ] the

\ I

(Mod) N s 5 (Mod) Nab (. . .)

' Poss ' (c) P o s s ( M o d ) N a b (• • •) (II) to (IV) The subject of the quality expressed by the abstract noun is notnamed. named. (II) There is no definite or indefinite article with the abstract. (Mod) N a b (...) Remarks: Included are the cases with the following quantifying expressions preceding the abstract (the numbers given show the occurrences of nouns in -ness and -ity, in that order): no (5/18), all (1/9), much (4/2), some (1/10), any in negative sentences (2/12). (III)

The abstract is determined by the definite article, (a) the (Mod) N a b (• • •)

220

(b) t £ < ? ( M o d ) N a b ( . . . )

| Rel j ICompJ Remarks: The definite article can be replaced by this or that: in (Ilia) 9 occurrences with nouns in -ness, 33 with nouns in -ity, in (IHb) 3 occurrences with nouns in -ness, 4 occurrences with nouns in -ity. The noun is determined by the indefinite article. (a) a(n) (Mod) N * ( . . . ) (b) 4 » ; ( M o d ) N a b ( . . . ) i Rel 1 ICompJ

(IV)

-NESS

Type I

a)

17.5%) 2,4% [ i3>3%)

b) c) II III

IV

-ITY

33,2%

16,9%) 2,3% [ 10,5%*

48,9%

46,7% a)

8,1%)

b)

3.3%»

a)

6,9%) 1,8%
7,8%) 0,9% s

12,7%

8,7%

A s the chart s h o w s , there are no significant differences between the syntactic behaviour of abstracts in -ness and abstracts in -ity, although they are of different morphological complexity (cf. 4 . 1 ) . Furthermore, in about t w o thirds of the occurrences the 'deep structure subject' of the abstracts is not named: 66,8% in the case of abstracts in -ness and 7 0 , 3 % in the case of abstracts in -ity. 5.2.3 Abstract T y p e s and Lexicalization In his b o o k Subject and Predicate

in Logic and Grammar

P. F . Strawson

proposes a three-fold distinction concerning de-adjectival nouns like sincerity,

C h o m s k y ' s much discussed example f r o m his Aspects of the

Theory of Syntax: I have already remarked that nominal phrases derived from the same source may nevertheless stand for items of very different types. Compare, for exam221

pie, (i) 'Sincerity is dangerous' with (2) ' N o one doubts John's sincerity' and the latter again with (3) 'John's sincerity is intermittent (or an unstable quantity)'. There is a marked contrast between the first case, where 'sincerity' denotes a general quality, and the second, where 'John's sincerity' rather denotes something of the fact-like or propositional order and where the case for supplying a transformational history which includes the propositional clause 'John is sincere' seems very strong. The third case, in which 'John's sincerity' denotes a particularized quality, is different again. (Strawson, 1974: 134) W e have here three types of (de-adjectival) abstracts, generalized abstracts, factive abstracts and quality abstracts. Generalized abstracts, as w a s argued a b o v e (cf. 5.2.2), are transformationally derived (by F ) f r o m f u n c t i o n a l structures w i t h an indefinite N in the position of the (subjective) genitive. T h e y are quality abstracts, not factive abstracts as the f o l l o w i n g substitutions in a syntactic f r a m e (a 'container' in V e n d l e r ' s (1968) (105)

t e r m i n o l o g y ) w i t h a factive predicate s h o w : a. John's sincerity b. '''Sincerity

¡was soon realized by everyone.) I is a fact. S < was soon realized by everyone, i I is a fact. >

Factive predicates as in ( 1 0 5 ) , i.e. predicates w h i c h presuppose that the sentence(s) u n d e r l y i n g one o r m o r e of their a r g u m e n t - N s is (are) true, can be used to establish factive abstracts, standing in a paraphrase relationship to n o m i n a l i z e d sentences: (106)

a. I don't ignore that John is clever. b. I don't ignore the fact that John is clever. c. I don't ignore John's cleverness.

A b s t r a c t s stand in a paraphrase relationship to sentences o n l y w h e n the predicate of the matrix sentence is factive (cf. above (57)): (107)

a. She believes that John is clever. b. "'She believes the fact that John is clever. c. " She believes John's cleverness.

V e n d l e r ( 1 9 6 7 : i 4 7 f f . ) mentions as a parallel to fact also result as in: (108)

ill

a. John's cleverness is the result of his having been brought up in a good family. b. That John is clever is the result of his having been brought up in a good family.

Factive abstracts like cleverness in (108a) must not be confused with 'resultative nouns' like invention, which are not abstracts (cf. below for McCawley's analysis). The difference between factive abstracts and quality abstracts can best be brought out with 'containers' not allowing nominalized sentences: (109) a. John's cleverness is superior to Bill's. b. "'(The fact) That John is clever is superior to Bill's.

But how can this difference be accounted for in the semantic interpretation of abstracts? In whatever reading, the functional structure of John's cleverness is

'N' John

'(N/N) /N' [Gen. sub.]

cleverness

The category dominating cleverness in (110) can be reformulated as the 'application' of an expression of category PA-in the case of cleverness a free morpheme identical to the base of the abstract-to a functor of category N/PA. Let us call this functor an 'abstractor' and assign a denotation to it by the following rule: (m)

The denotation of [Abstractor] in relation to M and U is the function from the denotation of PA relative to M and U to the denotation of N relative to M and U such that S, S2, where S, = [ [ N i f ^ . t P A j f j N . S / N S 2 = [[Nom].[N i .PAj] s J N .S/N and it is presupposed that Sk = T, i.e. that S/N contains a factive predicate.

Rule ( I I I ) assigns a denotation to the abstractor-functor for the cases where the abstract is a factive abstract. It can be extended to abstracts derived from two-place adjectives; PA must be replaced by T A and S, and S2 will formulate the following structures:

223

(112)

/

S/N

N

[Njn/N

N

rrAj]N

\ S/N

N

[Nom]

Ni

TAi

Factive abstracts provide a strong case for a transformational analysis of abstracts as they are clearly equivalent to nominalized sentences. This is also the case when the abstract is derivationally opaque: (113)

a. b. c. d.

/ I I I

don't ignore John's joy. don't ignore (the fact) that John is joyful. regret Bill's sloth. regret (the fact) that Bill is slothful.

Here a transformationalist position would have to postulate derivational processes like joyful —» joy, which are clearly contrary to the derivational structure of these lexemes; joyful is not the morphological base of joy, but its 'semantic base'. Besides, abstracts like cleverness or joy are ambiguous in that they are both factive and quality abstracts, and it is in the latter case that a lexicalist position can capture the 'lexicalization processes' connected to this interpretation more adequately (cf. below). While (107c) is unacceptable, (114a) is grammatical: (114)

a. b. c. d.

Mary Mary Mary Mary

believes believes believes believes

in John's cleverness. in John. John's explanation. John.

Examples like (114a) do not contradict rule ( 1 1 1 ) and the stipulation made there that a factive interpretation of abstracts depends on a factive predicate, i.e. on the presupposition that the embedded sentence in S 2 is 224

true. The verbs believe in as in (144a, b) and believe as in (144c, d) are not synonymous with believe in (107) (cf. OED s.v. believe). Believing in someone or something (144a, b) means putting one's trust in someone or something, believing someone or something (144c, d) is supposing as true what someone says/explains/asserts etc. (there is, of course, the further use of believe (in) in theological discourse). What (114a) then states is not that Mary believes that John is clever, but that she has trust in John's cleverness. Cleverness is here therefore a quality abstract. According to ( i n ) the entities denoted by factive abstracts and by nominalized sentences are put on the same level. They are individuals belonging to I (cf. 4.2.3), just as the denotations of John or the man who owns this house are individuals of I. Yet there is, of course, an ontological difference between the individuals denoted by proper names or descriptions and the individuals denoted by expressions like the fact that John is clever or John's cleverness. Here the differentiation between the extension and the intension of an expression can be of use (cf. 3.2.1). As regards nominalized sentences, the objects of 'propositional attitudes' (expressed by predicates like believe, think, regret etc.) are the intensions of sentences, i.e. possible states-of-affairs regretted, believed, thought about etc. A sub-class of these possible states-of-affairs are the states-ofaffairs in w „ in the actual world, i.e. the states-of-affairs expressed by true sentences ('the facts'). It is the latter that both factive abstracts and nominalized fact-sentences denote.34 The denotations of quality abstracts are of a somewhat different sort. Here an intensional interpretation of predicates is necessary. Although both in (115a) and in (115b) the predicates H A V E and POSSESS are possible, there is a semantic difference between the possession of an object-alienable or inalienable-and the possession of a quality:35 (115)

z.John

ihas

)

intelligence,

(possesses > b. John

ihas

)

a house.

(possesses $ 34

35

Without entering into ontology, one could relativize the remarks made here by pointing out that synonymy between S, and S2 in ( 1 1 1 ) an (112) is a synonymy based upon the 'semantic code' of language. This is to say that the ontology hinted at here is the ontology encoded in language, i.e. that we have here a 'common sense theory of reality'. See in this connection Lyons (197/: II, 438ff.). For an attempt to define propositions as possible worlds (or rather possible heavens) cf. Cresswell (1973: 39ff-)For a detailed philosophical discussion of individuals cf. Strawson (1959). For a critical discussion of propositional objects cf. Quine (1969: 139-160). For the predicate H A V E cf. inter alia Bach (1967), Lyons (1967). 225

According to H . Bendix, ( n j a ) contains an 'inherent H A V E ' (cf. Bendix, 1966: 37-59). As he points out there is a large A has B class in which the given 5-noun is an abstract and is related to an adjective in a predicate-adjective verb phrase of which A is the subject: 3(5) (i) A has strength «-» A is strong (ii) A has (much) happiness A is (very) happy (iii) A has (no) fear «-» A is (not) afraid/fearful (iv) A has courage A is courageous (v) A has employment A is employed (vi) A has equality with C A is equal with C (Bendix, 1966: 47) Interpreting predicates like S T R O N G in (i), we get either the class of strong objects in an extensional interpretation of predicates or the property of being strong in an intensional interpretation. What A has strength expresses is not that A belongs to the class of strong objects but that A possesses the quality of being strong. We therefore get the following equations: (116)

a. A has strength

A has the quality of being strong.

Or more generally: b. A has [ P A ] N A X.x(x

PA)

In (116b) the lambda-operator is used, symbolizing the intensional interpretation of predicates ("has the quality of being an x such that x P A " ; cf. 3.2.1 (17)). It follows from (116) that the phrase A's strength or more generally A's [PA] N is paraphrasable as in (117): (117)

a. The quality such that being strong is a quality and A has the quality of being strong b. The quality such that strength is a quality and A has strength c. The quality such [ P A ] N is a quality and A has [PA] N .

This gives us the following semantic interpretation of the abstractor, when the noun formed is a quality abstract: (118)

226

The denotation of [Abstractor] in relation to M and U is the function from the denotation of PA (or TA) relative to M and U to the denotation of N relative to M and U such that S, S2, where

[N;] N / N

ifPAjf

b. John's 1 gayness\ increased, in the course of the evening. (gaiety £ c. John is looking forward to the gaieties of the Christmas season. d. John's pink hat certainly contributed to the gaiety of the scene.

The derived nouns in (131a) are factive abstracts, in (131b) quality abstracts; gaiety can also be interpreted as a quality abstract (the scene is gay), but it differs from gaiety or gayness in (131b) insofar as its derivational base is not gay, = "full of or disposed to joy and mirth...", but rather gay2 = "bright or lively-looking, esp. in colour..." (cf. SOED). Of a different kind is gaiety in (131c); what is characterized as being gay are not things but events. Action nominals were analysed by Reichenbach (1947: 267ff.) by paraphrasing expressions like (132a) as (132b), which in turn can be symbolized by (132c): (132)

a. The coronation of George VI b. The event of the coronation of George VI C. (IV) ([f(x,)]* (v))

In (132c) x, stands for George VI and f for the predicate is crowned; with the help of the 'asterisk-operator' (cf. 5.2.1) the sentence George VI is crowned is nominalized and used as argument for the predicate v ("is an event"). The sentence thus formed, (133a), is then further bound by the iota-operator, which gives (132c), paraphrasable by (133b): (133)

a. That George VI is crowned is an event. b. The event such that George VI is crowned

This analysis also lies at the basis of McCawley's derivation of action nominals such as invention in (134a) from structures like (134b): 41 41

Cf. McCawley (1968a: 75). McCawley stresses also that invention can also be a result noun as in John's invention is useful, logically equivalent to (ly) (John invented y). For

233

(134)

a- Edison's invention of the phonograph. b. (iv) (Edison inventv the phonograph)

Action nominals are encountered especially among the group of abstracts ending in -ance, -ancy, -ence, -ency: (13y)

a. John's compliance came suddenly. b. The event of John's compliance came suddenly. c. John's action of complying came suddenly. d. John's acceptance of the offer came as a shock. e. John's disturbance of the meeting upset the agenda. f. John's appearance at the meeting was unlooked for.

The reason for this is, of course, the historical fact that adjectives ending in -ant, -ent are present participles and that consequently their meaning incorporates components paraphrasable by verbs like "do". Compare also the discussion of -ance etc. as suffixes marking deverbal abstracts in 5.1.2. A further type of 'lexicalized abstracts' is constituted by the class of nomina acti (cf. Wik, 1973: 84-9, for the difference between nomina actionis and nomina acti). These nouns are countable and can be paraphrased as "the instance of being (adj)"; cf.: (136) a. John's first stupidity was that he fell in love with Mary, his second that he married her. b. John's last eccentricity was that he exchanged his pink hat for a green turban.

These nouns are similar in meaning to nouns like callosity with the difference that instead of a paraphrase like "the (adj.) thing which x has" we now have a paraphrase like "the (adj.) thing which x does". When looking at lexical entries for derived nouns falling into this group, we notice, however, that a great many idiosyncrasies are to be met with here which minimize the usefulness of general interpretive rules in this sector of the lexicon. Compare (from the ALD) for instance: (137)

adversity:

1 . " c o n d i t i o n of adverse f o r t u n e ; t r o u b l e " 2 . " u n f a v o u r a b l e thing or event; m i s f o r t u n e "

deficiency:

1. a) " t h e state of being short o f , less than, w h a t is c o r r e c t or n e e d e d " b) "instance of t h i s "

a similar analysis cf. Wagner (1971). For an analysis of action nouns cf. also Ross (1972), Fraser (1970), Newmeyer (1970), Kastovsky (1977b: 1 3 - 2 0 ) and, on a typological basis, Comrie (1976); for the terms 'action' and 'act' cf. also Menzel ( 1 9 7 $ : 4 6 - 5 2 ) .

234

2. "amount b y which sth. is short of what is correct or needed" «

3. "sth. imperfect"

death:

1. " d y i n g ; ending of life" 2. "killing or being killed" 3. "state of being killed" 4. "cause of death"

What is important to note here, however, is the fact that while abstracts ending in -ness are normally (though not necessarily always, cf. business) factive or quality abstracts, nouns ending in other suffixes tend to have lexical idiosyncrasies of the described kind. This explains why an abstract like stupidness can be judged as less 'lexically coherent' ('encapsulated') than an abstract like stupidity. This does not mean, that there is no process of lexicalization at work when we are dealing with factive or quality abstracts. As has been shown in this section quality abstracts can themselves be of different kinds. We have 'extent' abstracts, 'manner' abstracts etc. as subtypes, depending on the paraphrases possible; cf. (.38)

* '

'

'

There is then an intricate network of relationships between the meanings of formally abstract nouns and their formal and semantic base predicates. Some of these relationships are of a regular kind; they can be signalled in syntax by categories like [PA] N and interpreted by semantic rules like ( 1 1 1 ) and (118). Others are so idiosyncratic as to be best described in the lexical entry for individual lexemes. While in view of derivational irregularities and the process of lexicalization a lexicalist position has been favoured, derivational regularities and systematic relationships like that between nominalized /aci-sentences and factive abstracts are nevertheless formulated, using in a somewhat syncretistic manner both the insights of transformational analyses and lexemic decompositions as proposed in generative semantics.

235

6. Conclusion

In concluding it might be worth stressing again a few points which have been made in the course of the discussion in the previous chapters. It was the aim of this study to analyse both the morphological, syntactic and semantic structure of the de-adjectival abstract in English, investigating at the same time the possibility of a categorial framework for relating the various phenomena to one another. The derivation of an abstract implies a change of the part of speech, a change from an 'adjective' to a 'noun'. An attempt to come to grips with the complexity of this 'categorychange' lies at the core of the present investigation. What does that mean, to change a part of speech? And how can a part of speech be defined? Categorial grammars, especially where they give a 'notional definition' of the parts of speech, provide a basis for a formulation of this change. The parts of speech are here functionally defined and related to predication-theory. Their functional definition makes it possible to capture the syntactical dynamics of the sentence-its organization into ranks and dependency relationships-, and their orientation on predicationtheory takes account of the ultimate basis of syntax in sentential semantics. Abstracts, it has been shown, qua elements of a part of speech are terms denoting qualities. This seemingly simple characterization, superficially not very different from the one speculative grammarians gave, can be made precise in the framework of a deep structure categorial grammar as sketched here. In order to make the syntactic and semantic description of the abstract noun intelligible, Chapter Three-after having given a survey of the main approaches to derivational morphology in Chapter Two-was intended as an introduction into the basic notions of categorial grammar as developed by the 'standard version' of categorial grammar (3.1.1), applicative grammar (3.1.2), Montague grammar (3.2.1) and its variants (3.2.2). In Chapter Four a synthesis of these various versions was attempted in the form of a categorial deep structure grammar. It was felt that while 236

a logical orientation is indispensable for a grammatical theory also containing a semantic component-and hence a 'sentential semantics'-, the rigorousness of formalized semantics could be somewhat relaxed in a linguistic context. Although attention has been paid to the precise and unambiguous formulation of semantic interpretation rules both in 4.2.3 and later, these rules have been formulated with a minimum of mathematical and logical apparatus. The first part of Chapter Five treats the derivational morphology of the de-adjectival abstract. The corpus at the basis of Chapter Five was originally compiled with a view to applying Saumjan's derivational graphs to English. As 5.1 shows, the complexity of the English lexicon is such that derivational rules cannot be set up by simply plotting a derivational graph and assigning affixes to its nodes. On the other hand, derivational graphs can reveal systematic relationships in the lexicon, despite irregularities and idiosyncrasies. Thus a graph like

covers twelve lexemes according to the REWG (detribalize, delocalize, devocalize, defeudalize, delegalize, dematerialize, denaturalize, demoralize, decentralize, denationalize, devitalize, decasualize); the nominal base of the adjective is not always segmentable, it is therefore put between parentheses in the graph. As the affixation of -able to most transitive verbs is possible, and from this again the derivation of an abstract in (perferably) -ity and -ness, we get the following pattern (see (2), p. 238). We can now extend the analysis of 4.1 by saying that while one of the contexts for -ness and -ity is -able, one of the contexts for the affixation of -able is in turn -ize. Verbs in -ize can be prefixed by de- and themselves be derived from adjectives in -al. Graph (2) covers more than the items recorded in the REWG; it is the formulation of productive derivational processes. Taking any adjective in -al, say ducal, we can form ducalize —» deducalize —» deducalizable —» deducalizability. A speaker of

237

o

W

-al

de-

-able

-ity • -ity •

0 -ness

• -ness

English, though unfamiliar with the form deducalizability, will nevertheless be able to explicate its meaning, and he might even, should there be an occasion for doing so, use this word, i.e. when talking about the possibility of depriving someone or something of its ducal status. This investigation into the derivational structure of English-its possibilities and its limitations-had to be complemented by a semantic interpretation of the lexemes derived. For deducalizability rules (HI) and (118) in j.2.3 will give the interpretation of the lexeme as abstract (with additional information in its lexical entry). The rules interpreting -able, -ize (and other causative affixes) and the privative prefix de- will further simplify the lexical entry and express at the same time the systematic character of derivational relationships. It is when we get to -al that the sub-lexemic nature of derivation becomes apparent. While central or natural can be morphologically broken down into centre + -al or nature + -al, this is not the case with legal or vocal, although a similar semantic decomposition is possible: "concerning law", "concerning voice" (cf. Ljung, 1970, for adjectives in -ful; cf. also Kastovsky, 1974). It is observations of this kind, stressing the sub-lexemic character of derivation, that have led to the adoption of a lexicalist position in the syntactic analysis of abstracts (5.2.2), without, however, neglecting the insights of transformationalists and generative semanticists, insights which were made use of in the semantic component of the present model, when differentiating between several types of abstracts and formulating their meaning on the basis of semantico-syntactic paraphrases 238

(5.2.3)- It is not maintained that the dividing line between syntax and semantics is an absolute one (for such a claim, stressing the autonomy of syntax, cf. Chomsky, 1977: 25 ff.) nor that it must be drawn as it has been drawn here. Linguistic theory is after all, like all other theory, not a collection of 'facts' but a net of hypotheses about these 'facts'-which, in the case of language, are inextricably interwoven with the theorist's view of them, as Saussure has pointed out in his Cours de Linguistique Générale. It is hoped that by striking a relativist note at the end of this study its results are not invalidated but rather made acceptable to linguists of a different theoretical persuasion.

2

39

Bibliography

Abraham, W. (1970), „Passiv und Verbalableitung auf E. -able, D. -bar", Folia Linguistica, 4. 38-52Adams, V. (1973), An Introduction to Modern English Word-Formation, English Language Series, 7, London. Ajdukiewicz, K. (1935), „Die syntaktische Konnexität", Studio. Philosophica, 1, 1-27. Allerton, D . J . (1969), "The sentence as a linguistic unit", Lingua, 22, 27-46. Allwood, J., L.-G. Andersson, Ö.Dahl (1977), Logic in Linguistics, Cambridge. Aronoff, M. (1976), Word Formation in Generative Grammar, Linguistic Inquiry Monograph, i, Cambridge, Mass. Austin, J . L. (1962), How to Do Things with Words, ed. J . O . Urmson, Oxford. Bach, E. (1967), "Have and be in English syntax", L, 43, 462-485. Bach, E., R . T . Harms, eds. (1968), Universals in Linguistic Theory, New York. Baldinger, K. (1980), Semantic Theory. Towards a Modem Semantics, transi, and ed. W.C. Brown, R.Wright, Oxford. Bar-Hillel, Y . (1953), "A quasi-arithmetical notation for syntactic description", L, 29, 47-58- (1956-7), "Husserl's conception of a purely logical grammar", Philosophy and Phenomenological Research, 17, 362-369. - (1964), "Some linguistic obstacles to machine translation", Language and Information. Selected Essays on their Theory and Application, New York, p. 75-86. Bar-Hillel, Y., C. Gaifman, E. Shamir (i960), "On categorial and phrase-structure grammars", Bulletin of the Research Council of Israel, 9F, 1 - 1 6 . Barri, N. (1975), "Adjectifs antéposés et adjectifs postposés comme signes linguistiques différents", Folia Linguistica, 7, 209-220. Bartsch, R. (1976), The Grammar of Adverbials. A Study in the Semantics and Syntax of Adverbial Constructions, N H L S , 16, Amsterdam. Bartsch, R., T. Vennemann (1972), Semantic Structures. A Study in the Relation between Semantics and Syntax, Athenäum-Skripten Linguistik, 9, Frankfurt. Baum, R. (1976), „Dependenzgrammatik". Tesnières Modell der Sprachbeschreibung in wissenschaftsgeschichtlicher und kritischer Sicht, Beihefte zur Zeitschrift für Romanische Philologie, 1 5 1 , Tübingen. Beard, R. (1976a), " A semantically based model of a generative lexical word-formation rule for Russian adjectives", L, 52, 108-120. - (1976b), "Once more on the analysis of ed-adjectives", JL, 12, 155-157. Beljaeva, T.M. (1979), Slovoobrazovatel'naja valentnost' glagol'nyx osnov v anglijskom jazyke [The derivational valency of verbal stems in English], Moscow. Bendix, H. (1966), Componential Analysis of General Vocabulary. The Semantic Structure of a Set of Verbs in English, Hindi and Japanese, The Hague.

240

Bennett, M. (1976), " A variation and extension of a Montague fragment of English", in Partee (1976: 119-163). Bergener, C. (1928), A Contribution to the Study of the Conversion of Adjectives into Nouns in English, Diss. Uppsala, Lund. Berman, A. (1974), Adjectives and Adjective Complement Constructions in English, Cambridge, Mass. Berry, M. (1975-7), Introduction to Systemic Linguistics, 2 vols., London. Bierwisch, M. (1962), „Uber den theoretischen Status des Morphems", Studia Grammatica, 1, Berlin, p.51-89. Biese, Y . M . (1941), Origin and Development of Conversions in English, Annal. Acad. Scient. Fennicae, B X L V , 2, Helsinki. Bloomfield, L. (1935), Language, London. Boas, H . U . (1974), "On Halle's 'Prolegomena to a Theory of Word Formation' or what is a linguistic generalization?", Linguistics, 134, 5-8. Bochenski, I.M. (1949), "On the syntactical categories", The New Scholasticism, 23, 257-280. - (1973), Grundriß der Logistik, transl. and rev. by A. Menne, fourth ed., Paderborn. Bolinger, D. (1967), "Adjectives in English: attribution and predication", Lingua, 1 8 , 1 - 3 4 . - (1972), Degree Words, J L , Ser. Maior, 53, The Hague. - (1973), "''John's easiness to please", Special Issue of IRAL on the Occasion of Bertil Malmberg's 60th Birthday, ed. G. Nickel, Heidelberg, p. 17-28. Bondarko, A.V. (1975), "Klassifikacija morfologiceskix kategorij" [The classification of morphological categories], in Jarceva (1975a: 56-76). Bonney, W.L. (1976), "Problems in the Grammar and Logic of English Complementation", Diss. Oxford, I U L C . Botha, R.P. (1968), The Function of the Lexicon in Transformational Generative Grammar, J L , Ser. Maior, 38, The Hague. Bowers, F. (1969), "The deep structure of abstract nouns", FL, 5, 520-533. Bowers, J.S. (1975a), "Some adjectival nominalizations in English", Lingua, 37, 341-361. - (1975b), "Adjectives and adverbs in English", FL, 13, 529-562. Bradley, R., N . Swartz (1979), Possible Worlds. An Introduction to Logic and its Philosophy, Oxford. Brame, M . K . (1976), Conjectures and Refutations in Syntax and Semantics, Studies in Linguistic Analysis, 3, Amsterdam. Breide, H . E . (1968), "On the syntax of adjectives determining agent nouns in present-day English", in Brekle, Lipka (1968: 20-31). - (1975), „Zur Stellung der Wortbildung in der Grammatik", Flexion und Wortbildung. Akten der V. Fachtagung der Indogermanischen Gesellschaft, Regensburg, 9.-14. September 1973, ed. H. Rix, Wiesbaden, p. 26-39. - (1976), Generative Satzsemantik und transformationeile Syntax im System der englischen Nominalkomposition, second ed., IBAL, 4, München. Brekle, H . E . , L. Lipka, eds. (1968), Wortbildung, Syntax und Morphologie. Festschrift zum 60. Geburtstag von Hans Marchand am 1. Oktober 196/, The Hague. Brekle, H . E . , D. Kastovsky, eds. (1977), Perspektiven der Wortbildungsforschung. Beitrage zum Wuppertaler Wortbildungskolloquium vom 9.-10. Juli 19/6 anläßlich des 70. Geburtstags von Hans Marchand am 1. Oktober 1977, Bonn. Bresnan, J.W. (1979), Theory of Complementation in English Syntax, Diss. MIT, New York.

241

Broch, T. (1970), „Kategoriale Grammatik, strukturanalytische Algebra und Phänomenologie I", AL, 13, 33-44. - (1973), „Kategoriale Grammatik, strukturanalytische Algebra und Phänomenologie II: Die Verkettung", AL, 14, 181-200. Brockhaus, K. (1975), »Ein Syntaxformalismus für Flexion und Wortbildung", Linguistische Berichte, 40, 45-62. Brendal, V. (1948), Les Parties du Discours. Partes Orationis. Études sur les Catégories Linguistiques, Copenhagen. Brugmann, E. (1900), „Uber das Wesen der sogenannten Wortzusammensetzung. Eine sprachpsychologische Studie", Berichte über die Verh. d. königl. sächs. Gesell, der Wiss. zu Leipzig, Philol.-hist. Classe, 52, Leipzig, p.359-401. Buck, C . D . (1933), Comparative Grammar of Greek and Latin, Chicago. Carnap, R. (1956), Meaning and Necessity. A Study in Semantics and Modal Logic, rev. ed., Chicago. - (1958, Introduction to Symbolic Logic and its Applications, New York. Chafe, W. L. (1970), Meaning and the Structure of Language, Chicago. Chapin, P. (1967), "On the Syntax of Word-Derivation in English", Unpubl. Diss. MIT, Cambridge, Mass. Chomsky, N . (1957), Syntactic Structures, J L , Ser. Minor, 4, The Hague. - (1962), " A transformational approach to syntax", Proceedings of the Third Texas Conference on Problems of Linguistic Analysis in English, ¡$¡8, ed. A.A.Hill, Austin, Texas, p. 124-158 [also in Fodor, Katz (1964: 211-245)]. - (1963)1 "Formal properties of grammars", Handbook of Mathematical Psychology, ed. R . D . Luce, R . R . Bush, E. Galanter, 3 vols., New York, 1963-65, II, 323-418. - (1965), Aspects of the Theory of Syntax, Cambridge, Mass. - (1970a), "Deep structure, surface structure, and semantic interpretation", Studies in General and Oriental Linguistics Presented to Shirô Hattori on the Occasion of His Sixtieth Birthday, ed. R.Jakobson, Sh. Kawamoto, Tokyo, p. 52-91. - (1970b), "Remarks on nominalization", in Jacobs, Rosenbaum (1970: 184-221). - (1977), Essays on Form and Interpretation, New York. Chomsky, N . , M. Halle (1968), The Sound Pattern of English, New York. Christophersen, P. (1939), The Articles. A Study of their Theory and Use in English, Copenhagen. Church, A. (1956), Introduction to Mathematical Logic. I, Princeton, N . J . Clifford, J . E . (1975), Tense and Tense Logic, J L , Ser. Minor, 215, The Hague. Cohen, J . M . (1967), "The equivalence of two concepts of categorial grammar", Information and Control, 10, 475-484. Cole, P., J . M . Sadock, eds. (1977), Syntax and Semantics 8: Grammatical Relations, New York. Comrie, B. (1976), "The syntax of action nominals: a cross-language study", Lingua, 40, 177-201. - (1978), Aspect. An Introduction to the Study of Verbal Aspect and Related Problems, corr. ed., Cambridge. Cook, H.P., H.Tredennick, transi, and eds. (1938), Aristotle. The Categories, On Interpretation, Prior Analytics, London. Cook, W.A. (1969), Introduction to Tagmemic Analysis, New York. Cooper, D . E . (1974), Presupposition, J L , Ser. Minor, 203, The Hague.

242

Cooper, R., T. Parsons (1976), "Montague grammar, generative semantics and interpretive semantics", in Partee (1976: 311-362). Cresswell, M . J . (1973), Logics and Languages, London. - (1977), "Categorial Languages", I U L C . Crystal, D . (1967), "English", Lingua, 17, 24-56. Curry, H . B . (1961), "Some logical aspects of grammatical structure", in Jakobson (1961: $6-68). - (1977), Foundations of Mathematical Logic, rev. ed., New York. Danes, F., ed. (1974), Papers on Functional Sentence Perspective, J L , Ser. Minor, 147, The Hague. Davidson, D., G . Harman, eds. (1972), Semantics of Natural Language, Dordrecht. DeArmond, R . C . (1969), "The concept of word derivation", Lingua, 22, 329-361. Delacruz, E . B . (1976), "Factives and proposition level constructions in Montague grammar", in Partee (1976: 177-199). De Rijk, L. M . (1962-7), Logica Modernum. A Contribution to the History of the Early Terministic Logic, 3 vols., Assen. Dik, S . C . (1978), Functional Grammar, N H L S , 37, Amsterdam. Dillon, G . L . (1977), Introduction to Contemporary Linguistic Semantics, Englewood Cliffs, N . J . Dokulil, M. (1964), „Zum wechselseitigen Verhältnis zwischen Wortbildung und Syntax", Travaux Linguistiques de Prague, 1, 215-224. - (1968), „Zur Theorie der Wortbildung", Wiss.Zeitschr.d.Karl-Marx-Univ. Leipzig, Gesellschafts- u. sprachwiss. Reihe, 17, 2 0 3 - 2 1 1 . Donnellan, K . S . (1966), "Reference and definite descriptions", Philosophical Review, 75, 281-304. Dowty, D . R . (1976), "Montague grammar and the lexical decomposition of causative verbs", in Partee (1976: 201-245). - (1979), Word Meaning and Montague Grammar. The Semantics of Verbs and Times in Generative Semantics and in Montague's PTQ, Synthese Language Library, Texts and Studies in Linguistics and Philosophy, 7, Dordrecht. Dubois, J . ( 1962), Étude sur la Dérivation Suffixale en Français Moderne et Contemporain. Essai d'Interprétation des Mouvements Observés dans le Domaine de la Morphologie des Mots Construits, Paris. Dubois-Charlier, F. (1971), Éléments de Linguistique Anglaise. La Phrase Complexe et les Nominalisations, Paris. Dummett, M. (1973), Frege. Philosophy of Language, London. Eisenberg, P. (1976), Oberflächenstruktur und logische Struktur. Untersuchungen zur Syntax und Semantik des deutschen Prädikatadjektivs, Linguistische Arbeiten, 36, Tübingen. Elson, B., V. Pickett (1965), An Introduction to Morphology and Syntax, Santa Ana, Cal. Erben, J . (1975), Einführung in die deutsche Wortbildungslehre, Grundlagen der Germanistik, 17, Berlin. Esau, H . (1973). Nominalization and Complementation in Modem German, N H L S , 10, Amsterdam. Fillmore, C . J . (1968), "The case for case", in Bach, Harms (1968: 1-88). - (1970), "Subjects, speakers, and roles", Synthese, 2 1 , 251-274. - (1971a), "Some problems for case grammar", in O'Brien (1971: 35-56).

243

- (1971b), "Types of lexical information", in Steinberg, Jakobovits (1971: 370-392). ( 977)y "The case for case reopened", in Cole, Sadock (1977: 59-81). Fodor, J . A. (1970), "Three reasons for not deriving 'kill' from 'cause to die"', Linguistic Inquiry, 1, 429-438. Fodor, J. A., J . J . Katz, eds. (1964), The Structure of Language. Readings in the Philosophy of Language, Englewood Cliffs, N . J . Flury, R. (1964), Struktur- und Bedeutungsgeschichte des Adjektiv-Suffixes -bar, Diss. Zürich, Winterthur. Fraser, B. (1970), "Some remarks on the action nominalization in English", in Jacobs, Rosenbaum (1970: 83-98). Frege, G. (1891), Function und Begriff, Jena [also in Frege (1969: 18-39)]. - (1892), „Uber Sinn und Bedeutung", Zeitschrift für Philosophie und philosophische Kritik, N F , 100, 25-50 [also in Frege (1969: 40-65)]. ~

l

_

('969)1 Funktion, Begriff, Bedeutung. Fünf logische Studien, ed. G. Patzig, third ed., Göttingen. Fries, C. C. (1952), The Structure of English, New York.

Gawelko, M. (1975), "Sur la classification sémantique des adjectifs suffixés", Lingua, 36, 3°7-323-

Geach, P. T. (1970-1), " A program for syntax", Synthese, 22, 3-17. Geach, P. T., M. Black, eds. (i960), Translations from the Philosophical Writings of Gottlob Frege, Oxford. Ginsburg, S., B . H . Partee (1969), " A mathematical model of transformational grammars", Information and Control, 15, 297-334. Ginzburg, E . L . (1973), "Issledovanie struktury slovoobrazovatel'nyx gnezd" [A study of the structure of derivational nests], Prohlemy strukturnoj lingvistiki 1972 [Problems of Structural Linguistics 1972], ed. S.K. Saumjan, Moscow, p. 146-225. - (1979), Slovoobrazovanie i sintaksis [Word-formation and Syntax], Moscow. Givön, T. (1970), "Notes on the semantic structure of English adjectives", L, 46, 816-837. Gnutzmann, C. (1974), „Zur Graduierbarkeit von Adjektiven im Englischen", Linguistische Berichte, 31, 1 - 1 2 . Grebe, P., et al. (1973), Der Große Duden. 4. Grammatik der deutschen Gegenwartssprache, third ed., Mannheim. Greenbaum, S. (1969), Studies in English Adverbial Usage, Longman's Linguistic Library, 10, London. Grinder, J . T . , S.H. Elgin (1973), Guide to Transformational Grammar. History, Theory, Practice, New York. Groot, A.W. (1956-7), "Classification of the uses of a case illustrated on the genitive in Latin", Lingua, 6, 8-66. Gruber, J.S. (1976), Lexical Structures in Syntax and Semantics, NHLS, 25, Amsterdam. Günther, E. (1974), „Zur Frage des Verhältnisses der Wortbildung zur Grammatik", Linguistische Arbeitsberichte, 9, 30-35. Gunter, R. (1972), "English derivation", JL, 8, 1-19. Haack, S. (1978), Philosophy of Logics, Cambridge. Hall, F. (1877), On English Adjectives in -Able with Special Reference to Reliable, London. Hall, R. (1965), "Parts of speech", The Aristotelian Society, Supplementary Volume 39, 173-188.

244

Halle, M . (1973), "Prolegomena to a theory of word formation", Linguistic Inquiry,

4,

3-16. Halliday, M . A . K . (1961), "Categories of the theory of grammar", Word, 17, 241-292. Haimos, P . R . (i960), Naive Set Theory, Princeton, N . J . Hamlyn, D . W . (1970), The Theory of Knowledge, Harris, J . ( 1 7 5 1 ) , Hermes: sal Grammar,

London.

Or, A Philosophical Inquiry Concerning Language and

Univer-

London.

Harris, Z . S . (1951), Methods in Structural Linguistics,

Chicago.

- (1957), "Co-occurrence and transformation in linguistic structure", L, 33, 283-340 [also in Fodor, Katz (1964: 155-210)]. - (1968), Mathematical

Structures of Language,

Interscience Tracts in Pure and Applied

Mathematics, 2 1 , N e w York. Harwood, F . W . , A . M . Wright (1956), "Statistical study of English word-formation", L, 32, 260-273. Hawkins, J . A . (1978), Definiteness maticality Prediction,

and Indefiniteness.

A Study in Reference

and

Gram-

London.

Hays, D . G . (1964), "Dependency theory: a formalism and some observations", L , 40, 511-525. Helbig, G . (1968), „Zum Funktionsbegriff in der modernen Linguistik", Deutsch Fremdsprache,

- , ed. ( 1 9 7 1 ) , Beiträge zur Valenztheorie, -

als

5, 274-287. J L , Ser. Minor, 1 1 5 , The Hague.

(1974), Geschichte der neueren Sprachwissenschaft,

Henry, D . P . (1972), Medieval

Hamburg.

Logic and Metaphysics. A Modern Introduction,

Herdan, G . (i960), Type-Token

Mathematics.

A Texthook of Mathematical

London. Linguistics,

J L , Ser. Maior, 4, The Hague. Hewson, J . (1972), Article and Noun in English, J L , Ser. Practica, 104, The Hague. Hickey, R . (1980), Satzstrukturen im Rahmen

des Deutschen und Englischen, eine kontrastive

der Dependenzgrammatik,

Analyse

Diss. Kiel.

Hirtle, W. (1969), " - E d adjectives like 'verandahed' and 'blue-eyed'", JL, 6, 19-36. Hiz, H . (i960), "The intuitions of grammatical categories", Methodos,

12, 3 1 1 - 3 1 9 .

- (1961), "Congrammaticality, batteries of transformations and grammatical categories", in Jakobson ( 1 9 6 1 : 43-50). - (1967), "Grammar logicism", The Monist, 51, 1 1 0 - 1 2 7 . - (1968), "Computable and uncomputable elements of syntax", in Rootselaar, Staal (1968: Hockett, C . F . (1958), A Course in Modern Linguistics, N e w York. Hudson, R . A . (1975a), "The meaning of questions", L, 51, 1 - 3 1 . - (1975b), "Problems in the analysis of ed-adjectives", JL, Husserl, E. (1913—21), Logische Untersuchungen,

1 1 , 69-72.

3 vols., second ed., Halle.

Hust, J . R., M. K . Brame (1976), "Jackendoff on interpretive semantics", Linguistic

Analy-

sis, 2, 243-277. Ivanova, I.P., et al. (1975), Struktura anglijskogo imeni suscestvitel'nogo [The Structure of the English Noun],

Moscow.

Jackendoff, R . S . (1972), Semantic Interpretation

in Generative

Grammar,

Studies in Lin-

guistics Series, 2, Cambridge, Mass. - (1975), "Morphological and semantic regularities in the lexikon", L, 51, 639-671. - (1977), X Syntax. A Study of Phrase Structure,

Linguistic Inquiry Monographs, 2,

Cambridge, Mass.

2

45

Jackl, E. (1952), Von lateinischen Partizipiaperfecti abgeleitete Adjectiva, Substantiva und Verba im englischen "Wortschatz, Diss. Innsbruck. Jacobs, R.A., P.S. Rosenbaum, eds. (1970), Readings in English Transformational Grammar, Waltham, Mass. Jacobsson, B. (1961), "'An unexprected usage: ahead, alive, and the like, before nouns'", Moderna Sprdk, 55, 240-247. Jakobson, R., ed. (1961), Structure of Language and its Mathematical Aspects, Proceedings of Symposia in Applied Mathematics, 12, Providence, R.I. Jarceva, V . N . , et al., eds. (1975a), Tipologija grammaticeskix kategorij [The Typology of Grammatical Categories], Moscow. Jarceva, V . N . (1975b), "Ierarxija grammaticeskix kategorij i tipologiceskaja xarakteristika jazykov" [The hierarchy of grammatical categories and the typological characterization of languages], in Jarceva (1975a: 5-23). Jespersen, O . (1922), Language. Its Nature, Development and Origin, London. - (1924), The Philosophy of Grammar, London. - (1969), Analytic Syntax, second ed., N e w York. Johnson, D . E . (1977), "On relational constraints on grammars", in Cole, Sadock (1977: 151-178). Jones, L. K. (1977), Theme in English Expository Discourse, Edward Sapir Monograph Series in Language, Culture, and Cognition, 2, Lake Bluff, 111. Juilland, A., A.Roceric (1972), The Linguistic Concept of Word. Analytic Bibliography, JL, Ser. Minor, 130, The Hague. Kamp, J . A . W . (1975), "Two theories about adjectives", Formal Semantics of Natural Language. Papers from a Colloquium Sponsored by the King's College Research Centre, Cambridge, ed. E.L. Keenan, Cambridge, p. 123-155. Karlgren, H . (1974), "Categorial grammar calculus", Statistical Methods in Linguistics, 10, 1-126. Karttunen, L. (1970), " O n the semantics of complement sentences", CLS, 6, 328-339. Kastovsky, D. (1973), "Causatives", FL, 10, 255-315. - (1974), "Word-formation, case grammar, and denominal adjectives", Anglia, 92, 1-54. - (1977a), "Problems of word-formation", Grundbegriffe und Hauptströmungen der Linguistik, ed. C. Gutknecht, Hamburg, p. 301-335. - (1977b), "Word-formation, or: at the crossroads of morphology, syntax, semantics, and the lexicon", Folia Linguistica, 10, 1-33. Katz, J.J. (1966), The Philosophy of Language, New York. - (1972), Semantic Theory, N e w York. - (1977), Propositional Structure and Illocutionary Force. A Study of the Contribution of Sentence Meaning to Speech Acts, Hassocks, Sussex. Katz, J.J., J. A. Fodor (1963), "The structure of a semantic theory", L, 39, 170-210 [also in Fodor, Katz (1964: 479-518)]. Katz, J.J., P.M. Postal (1964), An Integrated Theory of Linguistic Descriptions, Cambridge, Mass. Keil, H., ed. (1855-1923), Grammatici Latini, 6 vols., Leipzig. Kempson, R . M . (1975), Presupposition and the Delimitation of Semantics, CSL, 15, Cambridge. - (1977), Semantic Theory, Cambridge. Khrakovsky, V.S. (1973), "Passive constructions (Definition, Calculus, Typology, Meaning)", in Kiefer (1973b: 59-75)-

24 6

Kiefer, F. (1973a), " A propos derivational morphology", Statistical Methods in Linguistics, 9. 4 2- 59- , ed. (1973b), Trends in Soviet Theoretical Linguistics, F L , Suppl. Series, 18, Dordrecht. Kiparsky, P., C . Kiparsky (1971), "Fact", in Steinberg, Jakobovits (1971: 345-369). Kneale, W., M. Kneale (1971), The Development of Logic, corr. ed., Oxford. König, E. (1971), "'Transitive' Adjektive", Linguistische Berichte, 14, 42-50. Krämsky, J . (1969), The 'Word as a Linguistic Unit, J L , Ser. Minor, 75, The Hague. - (1972), The Article and the Concept of Definiteness in Language, J L , Ser. Minor, 125, The Hague. Kubrjakova, E.S. (1974), "Derivacija, transpozicija, konversija" [Derivation, transposition, conversion], Voprosy Jazykoznanija, 1974.5, 64-76. Kurylowicz, J . (1936), "Dérivation lexicale et dérivation syntaxique (Contribution à la théorie des parties du discours)", BSL, 37, 79-92. Kutschera, F.v. (1971), Sprachphilosophie, München. - (1976), Einführung in die intensionale Semantik, Berlin. Lakoff, G . (1970a), "Natural logic and lexical decomposition", CLS, 6, 340-362. - (1970b), Irregularity in Syntax, N e w York. - (1970-1), "Linguistics and natural logic", Synthese, 22, 1 5 1 - 2 7 1 . Lakoff, R . T . (1968), Abstract Syntax and Latin Complementation, Research Monograph, 49, Cambridge, Mass. Lamb, S. M. (1966), Outline of Stratificational Grammar, Washington, D . C . Lambek, J . (1961), " O n the calculus of syntactic types", in Jakobson (1961: 166-178). Larkin, M . T . (1971), Language in the Philosophy of Aristotle, J L , Ser. Minor, 87, The Hague. Leech, G . (1974), Semantics, Harmondsworth, Middlesex. Leeman, D. (1973), "Les paraphrases", Langages, 29, 43-54. Lees, R . B . (i960), The Grammar of English Nominalizations, Publ. of the Indiana Univ. Research Center in Anthropology, Folklore, and Linguistics, 12, Bloomington, Ind. Lehrberger, J . (1974), Functor Analysis of Natural Language, J L , Ser. Minor, 197, The Hague. Lejnieks, V. (1967), "The system of English suffixes", Linguistics, 29, 80-103. Lemmon, E . J . (1968), Introduction to Axiomatic Set Theory, London. Lesniewski, S. (1929), „Grundzüge eines neuen Systems der Grundlagen der Mathematik", Fundamenta Mathematicae, 14, 1 - 8 1 . Levi, J . N . (1973), "Where do all those other adjectives come from?", CLS, 9, 332-345. Levkovskaja, K . A . (1952), " O slovoobrazovanii i ego otnosenii k grammatike" [On wordformation and its relationship to grammar], Voprosy teorii i istorii jazyka [Problems in the Theory and History of Language], Moscow, p.i 5 3 - 1 8 1 . Lewis, C . S . (i960), Studies in Words, Cambridge. Lewis, D. (1970-1), "General semantics", Synthese, 22, 18-67. Lewis, G . L. (1975), Turkish Grammar, corr. ed., Oxford. Lightfoot, D . W . (1979), Principles of Diachronic Syntax, C S L , 23, Cambridge. Lightner, T. M. (1975), "The role of derivational morphology in generative grammar, L, 51, 617-638. Link, G . (1976), Intensionale Semantik, Münchener Univ.-Schriften, Reihe der Philos. Fakultät, 17, München. Lipka, L. (1966), Die Wortbildungstypen WATERPROOF und GRASSGREEN und ihre Entsprechungen im Deutschen, Diss. Tübingen.

247

- (i97 l a )> „Grammatikalität, Akzeptabilität und Produktivität in der Sprache", in Stechow (1971: 142-151). - (1971b), "Grammatical categories, lexical items and word-formation", FL, 7, 211-238. - (1972), Semantic Structure and Word-Formation. Verb-Particle Constructions in Contemporary English, IBAL, 17, München. - (1975), "Prolegomena to 'Prolegomena to a Theory of Word Formation'. A reply to Morris Halle", The Transformational-Generative Paradigm and Modem Linguistic Theory, ed. E . F . K . Koerner, Amsterdam, p.175-184. - (1976), "Topicalization, case grammar, and lexical decomposition in English", Archivum Linguisticum, 7, 1 1 8 - 1 4 1 . - (1977), „Lexikalisierung, Idiomatisierung und Hypostasierung als Probleme einer synchronischen Wortbildungslehre", in Breide, Kastovsky (1977: 155-164). Ljung, M. (1970), English Denominal Adjectives. A Generative Study of the Semantics of a Group of High-Frequency Denominal Adjectives in English, Gothenburg Studies in English, 21, Lund. - (1976), "-ed adjectives revisited",//., 12, 159-168. Lloyd, P.M. (1964), "An analytical survey of studies in Romance word formation", Romance Philology, 17, 736-770. Lockwood, D . G . (1972), Introduction to Stratificational Linguistics, New York. Lucas, M.A. (1975), "The syntactic classes of antenominal adjectives in English", Lingua, 35. »55-I7ILuschei, E . C . (1962), The Logical Systems of Lesniewski, Amsterdam. Lyons, J. (1966), "Towards a 'notional' theory of the 'parts of speech'", JL, 2, 209-236. - (1967), "A note on possessive, existential and locative sentences", FL, 3, 390-396. - (1968), Introduction to Theoretical Linguistics, Cambridge. - (1977), Semantics, 2 vols., Cambridge. Maclay, H . (1971), "II. Linguistics: Overview", in Steinberg, Jakobovits (1971: 157-182). Mahn, L . H . (1971), 2ur Morphologie und Semantik englischer Verben auf -IFY mit Berücksichtigung französischer und deutscher Entsprechungen, TBL, 27, Tübingen. Marchand, H . (1964), "A set of criteria for the establishing of derivational relationship between words unmarked by derivational morphemes", IF, 69, 10-19. - (1966), " O n attributive and predicative derived adjectives and some problems related to the distinction", Anglia, 84, 131-149. - (1967), "Expansion, transposition, and derivation", La Linguistique, 1967.1, 13-26. Marcus, S. (1967), Algebraic Linguistics; Analytical Models, Mathematics in Science and Engineering, 29, New York. Martinet, A. (1962), A Functional View of Language, Oxford. Mates, B. (1968), "Leibniz on possible worlds", in Rootselaar, Staal (1968: 507-529). Matthews, P . H . (1974), Morphology. An Introduction

to the Theory of

Word-Structure,

Cambridge. McCall, S., ed. (1967), Polish Logic 1920-1939. Papers by Ajdukiewicz, Chwistek, Jaskowski, Jordan, Lesniewski, Lukasiewcicz, Slupecki, Sobocinski, and Wajsberg, Oxford. McCawley, J . D . (1968a), "Lexical insertion in a transformational grammar without deep structure", CLS, 4, 71-80. - (1968b), "The role of semantics in a grammar", in Bach, Harms (1968: 124-169). - (1971a) "Where do noun phrases come from?", in Steinberg, Jakobovits (1971: 217-231). - (1971b), "Prelexical syntax", in O'Brien (1971: 19—33). 248

Meid, W. (1967), Germanische Sprachwissenschaft. III. Wortbildungslehre, Berlin. Menzel, P. (1975), Semantics and Syntax in Complementation, J L , Ser. Minor, 176, The Hague. Meskov, O. D. (1976), Slovoobrazovanie sovremennogo anglijskogo jazyka [Word-Formation of Contemporary English], Moscow. Mohanty, I . N . (1964), Edmund Husserl's Theory of Meaning, Phaenomenologica, 14, The Hague. Montague, R. (1974), Formal Philosophy. Selected Papers, ed. R . H . Thomason, New Haven, Conn. Moravcsik, J . M . E . (1975), Understanding Language. A Study of Theories of Language in Linguistics and Philosophy, J L , Ser. Minor, 169, The Hague. Mötsch, W. (1962), „Zur Stellung der .Wortbildung' in einem formalen Sprachmodell", Studia Grammatica, 1, Berlin, p. 31-50. - (1964), Syntax des deutschen Adjektivs. Studia Grammatica, 3, Berlin. - (1967), „Können attributive Adjektive durch Transformationen erklärt werden?", Folia Linguistica, 1, 23-48. Mustanoja, T.F. (i960), A Middle English Syntax. I. Parts of Speech, Mémoires de la Société Néophilologique de Helsinki, 23, Helsinki. Neuhaus, H.J. (1971), Beschränkungen in der Grammatik der Wortableitungen im Englischen, Diss. Saarbrücken. - (1973), „Zur Theorie der Produktivität von Wortbildungssystemen", Linguistische Perspektiven. Referate des VII. Linguistischen Kolloquiums Nijmegen, 26.-30. September 1972, ed. A.P. ten Cate, P. Jordens, Linguistische Arbeiten, 5, Tübingen, p. 305-317. - (1977), „Wortbildungssemantik", in Breide, Kastovsky (1977: 203-209). Newman, S.S. (1948), "English suffixation: a descriptive approach", Word, 4, 24-36. Newmeyer, F.J. (1970), "The derivation of the English action nominalization", CLS, 6, 408-415. - (1971), "The source of derived nominals in English, L, 47, 786-796. Nida, E . A . (1949), Morphology. The Descriptive Analysis of Words, second ed., Ann Arbor, Mich. - (1966), A Synopsis of English Syntax, J L , Ser. Practica, 19, second ed., The Hague. Nolan, R. (1970), Foundations for an Adequate Criterion of Paraphrase, J L , Ser. Minor, 84, The Hague. Norrick, N . R . (1978), Factive Adjectives and the Theory of Factivity, Linguistische Arbeiten, 64, Tübingen. O'Brien, R.J., ed. (1971), Report of the Twenty-Second Annual Round Table Meeting on Linguistics and Language Studies. Linguistics: Developments of the Sixties - Viewpoints for the Seventies, Washington, D. C. Oomen, U. (1978), "Interrogative sentences as responses: structural properties and illocutionary meanings", Anglia, 96, 285-309. Palmer, F.R. (1979), Modality and the English Modals, Longman's Linguistic Library, 23, London. Partee, B . H . (1973/76), "Some transformational extensions of Montague grammar", Journal of Philosophical Logic, 2, (1973), 509-539 [quoted from Partee (1976: 51-76)]. - (1975), "Montague grammar and transformational grammar", Linguistic Inquiry, 6, 203-300. - , ed. (1976), Montague Grammar, New York. Paul, H. (1920), Prinzipien der Sprachgeschichte, fifth ed., Tübingen.

249

Pennanen, E. (1972), "Current views of word-formation", Neuphilologische 73, 292-308.

Mitteilungen,

Pike, K. L. (1958), "On tagmemes, née gramemes", International Journal of American Linguistics, 24, 273-278. - (1967), Language in Relation to a Unified Theory of the Structure of Human Behavior, J L , Ser. Maior, 24, second rev. ed., The Hague. Pike, K.L., E . G . Pike (1977), Grammatical Analysis, Summer Institute in Linguistics, Publ. in Linguistics, 53, Dallas, Texas. Pilch, H. (1968), „Modelle der englischen Wortbildung", in Brekle, Lipka (1968: 160-178). - (1970), Altenglische Grammatik. Dialektologie, Phonologie, Morphologie, Syntax, Commentationes Societatis Linguisticae Europae, 1 , 1 , München. Platts, M. de B. (1979), Ways of Meaning. An Introduction to a Philosophy of Language, London. Poldauf, J . (1959), „Die Bildung der englischen Adjektiva auf -hie. Ein Beitrag zur Theorie der synchronen Wortbildungslehre", Zeitschrift für Anglistik und Amerikanistik, 7, 229-245. Porzig, W. (1930), „Die Leistung der Abstrakta in der Sprache", Blätter für Deutsche Philosophie, 4, 66-77. Postal, P.M. (1964), "Limitations of phrase structure grammars", in Fodor, Katz (1964: '37-ijO- (1971), Cross-Over Phenomena, New York. - (1974), On Raising. One Rule of English Grammar and its Theoretical Implications, Current Studies in Linguistics Series, 5, Cambridge, Mass. Postal, P.M., J . R . Ross (1971), "¡Tough movement si, tough deletion no!", Linguistic Inquiry, 2, 544-546. Potts, T . C . (1973), "Fregean categorial grammar", Logic, Language and Probability, ed. R . J . Bogdan, I. Niiniluoto, Dordrecht, p.245-284. Pullum, G. K. (1977), "Word order universals and grammatical relations", in Cole, Sadock (1977: 249-277). Quine, W.V. (i960), Word and Object, Cambridge, Mass. - (1961), From a Logical Point of View. 9 Logico-Philosophical Essays, second rev. ed., New York. - (1969), Ontological Relativity and Other Essays, The John Dewey Essays in Philosophy, x, New York. - (1970), Philosophy of Logic, Englewood Cliffs, N . J . Quirk, R. (1977), " A tough object to trace", JL, 13, 99-102. Radford, A. (1977), Italian Syntax. Transformational and Relational Grammar, CSL, 21, Cambridge. Reich, P.A. (1969), "The finiteness of natural language", L, 45, 831-843. Reichenbach, H. (1947), Elements of Symbolic Logic, New York. Reichl, K. (1976), 'Tractatus de grammatical Eine fälschlich Robert Grosseteste zugeschriebene spekulative Grammatik. Edition und Kommentar, Veröffentlichungen des Grabmann-Instituts, N F , 28, Paderborn. - (forthcoming), "Derivational patterns in languages with a 'mixed lexicon'", Orbis. Robbins, B . L . (1968), The Definite Article in English Transformations, Papers on Formal Linguistics, 4, The Hague.

250

Robins, R . H . (1966), "The development of the word class system of the European grammatical tradition", FL, 2, 3-19. Robinson, J . J . (1970), "Dependency structures and transformational rules", L, 46, 259-285. Rogers, R. (1963), " A survey of formal semantics", Synthese, 15, 17-56. Rootselaar, B. van, J . F . Staal, eds. (1968), Logic, Methodology and Philosophy of Science III. Proceedings of the Third International Congress for Logic, Methodology and Philosophy of Science, Amsterdam 1967, Amsterdam. Rosenbaum, P.S. (1967), The Grammar of English Predicate Complement Constructions, Research Monograph, 47, Cambridge, Mass. Rosenberg, M.S. (1974-5), "Generative vs. interpretive semantics", FL, 12, 561-582. Ross, J . R . (1970), "On declarative sentences", in Jacobs, Rosenbaum (1970: 222-272). - (1972), "Act", in Davidson, Harman (1972: 70-126). Russell, B. (1912), The Problems of Philosophy, The Home Univ. Library of Modern Knowledge, 40, London. Sadock, J . M . (1974), Toward a Linguistic Theory of Speech Acts, New York. Saussure, F. de (1967-71), Cours de Linguistique Générale, ed. R. Engler, 4 vols., Wiesbaden. Schachter, P. (1973), "On syntactic categories", IULC. - (1976), " A nontransformational account of gerundive nominals in English", Linguistic Inquiry, 7, 205-241. Saumjan, S.K. (1965a), Strukturnaja lingvistika [Structural Linguistics], Moscow. - (1965b), "Outline of the applicational generative model for the description of language", FL, i, 189-222. - (1971a), Strukturale Linguistik, IBAL, 12, München. - (1971b), Principles of Structural Linguistics, J L , ser. maior, 45, The Hague. - ( I 97j)> "The genotype language and formal semantics", in Kiefer (1973b: 251-333). - (1974), Applikativnaja grammatika kak semanticeskaja teorija estestvennyx jazykov [Applicative Grammar as a Semantic Theory of Natural Languages], Moscow. - (Shaumyan) (1977), Applicational Grammar as a Semantic Theory of Natural Language, Edinburgh. Schön, E. (1905), Die Bildung des Adjektivs im Altenglischen, Kieler Studien zur engl. Phil., N F , 2, Kiel. Schopf, A. (1969), Untersuchungen zur Wechselbeziehung zwischen Grammatik und Lexik im Englischen, Quellen u. Forschungen zur Sprach- u. Kulturgeschichte d. german. Völker, N F , 32, Berlin. Searle, J . R . (1969), Speech Acts. An Essay in the Philosophy of Language, Cambridge. - (I979)> Expression and Meaning. Studies in the Theory of Speech Acts, Cambridge. Seiler, H. (i960), Relativsatz, Attribut und Apposition, Wiesbaden. Serebrennikov, B.A., ed. (1970-3), Obscee jazykoznanie [General Linguistics], 3 vols., Moscow. - (1973-6), Allgemeine Sprachwissenschaft, 3 vols., Berlin, München. Seuren, P . A . M . (1969), Operators and Nucleus. A Contribution to the Theory of Grammar, CSL, 2, Cambridge. - , ed. (1974), Semantic Syntax, London. Seymour, R . K . (1968), A Bibliography of Word Formation in the Germanic Languages, Durham, N . C . Siegel, M . E . A . (1980), Capturing the Adjective, Diss. Univ. of Mass., New York.

Smaby, R . M . (1971), Paraphrase Grammars, Formal Linguistics Series, 2, Dordrecht. Soboleva, P.A. (1973), "Derivational structure of the Russian lexicon", in Kiefer (1973b: 77-103)- (1976), "Derivational word-structure in the applicative grammar and the typology of homonyms", Theoretische Linguistik in Osteuropa. Originalbeiträge und Erstübersetzungen, ed. W. Girke, H. Jachnow, Konzepte der Sprach- u. Lit.wiss., 18, Tübingen, p.90-122. Southworth, F. C. (1967), "A model of semantic structure", L, 43, 342-361. Stankiewicz, E. (1962), "The interdependence of paradigmatic and derivational patterns", Word, 18, 1-22. Stechow, A. v., ed. (1971), Beiträge zur generativen Grammatik. Referate des f . linguistischen Kolloquiums Regensburg 1970, Braunschweig. Stegmüller, W. (1956), „Sprache und Logik", Studium Generale, 9, 57-77. - (1956-7), „Das Universalienproblem einst und jetzt", Archiv für Philosophie, 6, 192-225; 7, 45-81. - (1957), Das Wahrheitsproblem und die Idee der Semantik. Eine Einführung in die Theorien von A. Tarski und R. Carnap, Wien. - (1965-75), Hauptströmungen der Gegenwartsphilosophie. Eine kritische Einführung, z vols., Stuttgart. Stein, G. (1971), Primäre und sekundäre Adjektive im Französischen und Englischen, TBL, 22, Tübingen. - (1973), English Word-Formation Over Two Centuries. In Honour of Hans Marchand on the Occasion of his Sixty-Fifth Birthday, 1 October 1972, TBL, 34, Tübingen. Steinberg, D.D., L . A . Jakobovits, eds. (1971), Semantics. An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, Cambridge. Stepanova, M.D. (1968), Metody sinxronogo analyza leksiki (na materiale sovremennogo nemeckogo jazyka) [Methods of the Synchronic Analysis of the Lexicon (Illustrated from Contemporary German)], Moscow. - (Stepanowa) (1973), Methoden der synchronen Wortschatzanalyse, Halle, München. Stephany, U. (1970), Adjektivische Attributkonstruktionen des Französischen. Untersucht nach dem Modell der generativen Grammatik, Structura, 1, München. Strawson, P.F. (1950), "On referring", Mind, 59, 320-344. - (1952), Introduction to Logical Theory, London. - (1959), Individuals. An Essay in Descriptive Metaphysics, London. - (1974), Subject and Predicate in Logic and Grammar, London. Sussex, R. (1974), "The deep structure of adjectives in noun phrases", JL, 10, 1 1 1 - 1 3 1 . Svartvik, J. (1966), On Voice in the English Verb, J L , Ser. Practica, 63, The Hague. Sweet, H. (1892-8), A New English Grammar Logical and Historical, 2 vols., Oxford. Tesnière, L. (1966), Éléments de Syntaxe Structurale, second rev. ed., Paris. Thomason, R . H . (1976), "Some extensions of Montague grammar", in Partee (1976: 77-"7)Thorndike, E.L., I. Lorge, eds. (1944), The Teacher's Word Book of j 0,000 Words, New York. Ullmann, S. (1963), The Principles of Semantics, second ed., third impr., Oxford. Ullmer-Ehrich, V. (1977), Zur Syntax und Semantik von Substantivierungen im Deutschen, Monographien Linguistik u. Kom.wiss., 29, Kronberg. Ultan, R. (1978), "Toward a typology of substantival possession", Universals of Human

252

Language, ed. J . H . Greenberg, C . A . Ferguson, E . A . Moravcsik, 4 vols., Stanford, Cal., IV, 11-49. Ungeheuer, G. (1969), „Paraphrase und syntaktische Tiefenstruktur", Folia Linguistica, 3, 178-227. Vater, H. (1975), "Toward a generative dependency grammar", Lingua, 36, 121-145. Vendler, Z. (1967), Linguistics in Philosophy, Ithaca, N . Y . - (1968), Adjectives and Nominalizations, Papers on Formal Linguistics, 5, The Hague. Vennemann, T. (1976), "Categorial grammar and the order of meaningful elements", Linguistic Studies Offered to Joseph Greenberg on the Occasion of his Sixtieth Birthday, ed. A. Juilland, 3 vols., Saratoga, Cal., Ill, 615-634. Vennemann, T., R. Harlow (1977), "Categorial grammar and consistent basic VX serialization", Theoretical Linguistics, 4, 227-254. Wagner, K . H . (1971), „Die Nominalisierung im Englischen", in Stechow (1971: 264-272). Wall, R. (1972), Introduction to Mathematical Linguistics, Englewood Cliffs, N . J . Wasow, T., T. Roeper (1972), "On the subject of gerunds", FL, 8, 44-61. Weinreich, U. (1972), Explorations in Semantic Theory, J L , Ser. Minor, 89, The Hague. Wells, R.S. (1947), "Immediate constituents", L, 23, 8 1 - 1 1 7 . White, M. (1974), „Die 'ble'-Wörter im Englischen: Ihr Gebrauch zum Ausdruck von Möglichkeit und Notwendigkeit", Ratio, 16, 128-141. Whitehead, A . N . , B. Russell (1927), Principia Mathematica, second ed., Cambridge. Wierzbicka, A. (1972), Semantic Primitives, Linguistische Forschungen, 22, Frankfurt. - (1975), "Why 'kill' does not mean 'cause to die': the semantics of action sentences", FL, 13» 49I-5 2 8 Wik, B. (1973), English Nominalizations in -ing. Synchronic and Diachronic Aspects, Acta Univ. Upsaliensis, Studia Angl. Upsaliensia, 12, Uppsala. Wilkinson, R. (1970), "Factive complements and action complements", CLS, 6, 425-444. Winter, W. (1965), "Transforms without kernels?", L, 41, 484-489. Wittgenstein, L. (1953), Philosophische Untersuchungen/Philosophical Investigations, ed. G . E . M . Anscombe, R. Rhees, Oxford. - (1969), Preliminary Studies for the "Philosophical Investigations" Generally Known as the Blue and Brown Books, second ed., Oxford. Wlodarczyk, H. (1974), "La grammaire generative applicative de S.K. Saumjan", Langages, 33, 15-64. Wunderlich, D. (1971), „Warum die Darstellung von Nominalisierungen problematisch bleibt", Probleme und Fortschritte der Transformationsgrammatik. Referate des 4. Linguistischen Kolloquiums, Berlin 6.-/0. Oktober 1969, ed. D. Wunderlich, Linguistische Reihe, 8, München, p.189-218. Xolodovic, A . A . , ed. (1974), Tipologija passivnyx konstrukcij. Diatezy i zalogi. [The Typology of Passive Constructions. Voice and 'Genus Verbi'J, Leningrad. Yotsukura, S. (1970), The Articles in English. A Structural Analysis of Usage, J L , Ser. Practica, 49, The Hague. Zirmunskij, V . M . (1968), " O prirode castej reci i ix klassifikacii" [On the nature of the parts of speech and their classification], Voprosy teorii castej reci, na materiale jazykov razlicnyx tipov [Problems of the Parts of Speech, Illustrated from Languages of Various Types], ed. V.M. Zirmunskij, O.P. Sunik, Leningrad, p.7-32. Zimmer, K . E . (1964), Affixal Negation in English and Other Languages. An Investigation of Restricted Productivity, Suppl. to Word, 20, Monograph, 5, New York. Zuber, R. (1973), "La categorematicite et les adjectifs en polonais", Langages, 30, 1 2 5 - 1 3 1 .

253