Surface syntax of English: a formal model within the meaning-text framework (Linguistic & Literary Studies in Eastern Europe)

SURFACE SYNTAX OF ENGLISH

LINGUISTIC & LITERARY STUDIES IN EASTERN EUROPE (LLSEE) The emphasis of this scholarly series is on recent developments in Linguistic and Literary Research in Eastern Europe; it includes analysis, translations and syntheses of current research as well as studies in the history of linguistic and literary scholarship.

Founding Editor: John Odmark General Editor: Philip A. Luelsdorff

Volume 13

Igor A. Mel'cuk & Nikolaj V. Pertsov Surface Syntax of English  Formal Model within the Meaning-Text Framework

SURFACE SYNTAX OF ENGLISH A FORMAL MODEL WITHIN THE MEANING-TEXT FRAMEWORK

Igor A. Mel'čuk and Nikolaj V. Pertsov

Editor: Richard Kittredge

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA 1987

Library of Congress Cataloging in Publication Data Mel'cuk, I.A. (Igor' Aleksandrovič), 1932Surface syntax of English. (Linguistic & literary studies in Eastern Europe (LLSEE), ISSN 0165-7712; v. 13) Bibliography: p. Includes indexes. 1. English language - Syntax. I. Pertsov, Nikolaj V. (Nikolaj Viktorovic), 1944Title. III. Series. PE1361.M45 1987 425 86-6884 ISBN 90 272 1515 4 (alk. paper)

. II.

© Copyright 1987 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.

To all our American friends. (No names need be mentioned, need they?)

"Farewell,  Muses! And you, Grammar, be gone with them as well Lest that damned Syntax of yours hasten me into my grave!" Palladius, a Greek poet of Alexandria, IV-V centuries A.D.

CONTENTS Editor's Preface

xiii

Foreword

xv

Acknowledgements Introduction

xvii 1

§1.

Three raisons d'être of this Book

§2.

Five Main Characteristics of the Syntactic Approach in this Book . 1. Framework of analysis 2. Utterance representations at different levels 3. A static viewpoint 4. Dependency trees 5. Labeled surface-syntactic relations

4 5 6 6 7 8

§3.

Organization of this Book

9

Chapter I.

A Brief Outline of the Meaning-Text Theory and the Corres­ ponding Linguistic Model

1

12

§1.

General Remarks

12

§2.

Levels of Utterance Representation in the Meaning-Text Model .

15

§3.

General Design of the Meaning-Text Model

34

§4.

Five Basic Principles of the Meaning-Text Approach

43

Chapter II.

Surface-Syntactic Representation for English Sentences . . .

46

§ 1.

General Characterization of the Surface-Syntactic Representation ..

46

§2.

Surface-Syntactic Structure 1. Dependency trees 2. Correspondence between the nodes of a D-tree and the wordforms of the sentence it represents 3. Deep-morphological representation of a wordform 4. Surface-syntactic relation

48 48 57 58 61

x

CONTENTS 5. §3. §4.

Grouping in dependency structures

72

Conditions for Grammatical Correctness of English Surface-Syntac­ tic Structures

78

Surface-Syntactic Relations in Modern English

85

Specimens of Surface-Syntactic Structures for English Sentences . . . . Chapter III.

157

Deep-Morphological Representation for English Sentences .

163

General Characterization of the Deep-Morphological Representa­ tion (of a Sentence)

163

§2.

Deep-Morphological Structure

165

§3.

Conditions for Grammatical Correctness of English Deep-Morpho­ logical Structures

166

§ 1.

§4.

Morphological Variables and Values of Wordforms in Modern Eng­ lish 167

Chapter IV.

Surface-Syntactic Component of English

178

§1.

General Characterization of the Surface-Syntactic Component

.

178

§2.

The Notion of Syntagm 1. Characterization of the syntagm 2. Means and devices for a more compact syntagm notation . . . Standard subtrees Standard functions General conditions for groups of syntagms Compounding syntagms Variables in syntagms 3. Classification of syntagms

186 187 191 192 193 195 195 196 198

§3.

Syntagms of Modern English Guide to the List of Modern English Syntagms List of Modern English Syntagms

198 200 216

Appendix I.

Parts of Speech and Syntactic Features of English Lexemes .

Appendix II. Standard Subtrees A.

Internal Structure of Standard Subtrees

471 485 485

CONTENTS

xi

B.

External Connections of Standard Subtrees

490

C.

Distribution (among the Individual Nodes of the Subtree) of Labels that are Attached to the Subtree Symbol as a Whole

491

Appendix III. List of English Lexemes Mentioned in Syntagms ('Syntactic' Lexemes)

493

Appendix IV. Samples of English Syntactic Phenomena Unaccounted for in the Proposed Description 501 References

507

Subject and Term Index

521

EDITOR'S PREFACE

The publication of Surface Syntax of English (SSE) by Igor Mel'cuk and Nikolaj Pertsov marks a significant event in the continuing dialogue between theoretical and computational linguistics. Here we have one of the most complete formalized grammars of English ever published. This is an important book for at least three reasons. First of all, the SSE consti­ tutes a major contribution to the syntactic description of English. By its broad inventory of syntactic structures and use of more nuanced and explicitly labeled syntactic relations, it sets a high standard for completeness of description. The relatively few constructions not included in the formal description are nevertheless listed with a discussion of the problems they raise. The SSE's breadth of coverage is equalled by the wealth of syntactic and lexical detail given for each construction, often in support of new and insightful analyses of the phenomena of English grammar. Because of this, it will surely take its place as one of the leading reference works on English. Second, beyond the breadth and depth of its coverage, the SSE is important for an entirely separate and particularly timely reason.  and Pertsov's grammatical description is very well suited to and inspired by the demands of computer processing of natural language. Although the descriptive framework does not impose a procedural view of grammatical analysis (and is in fact perfectly suited to the static or non-directional views of grammar now popular), the syntactic descrip­ tions presented here are expressed in a formal language that lends itself to a relatively direct translation into rules executable by a computer program. Such rules can serve both for automatic analysis and synthesis of text. Moreover, the grammatical insights contained in this book have profited from the more than two decades of experience of Mel'cuk and his colleagues in machine translation and text synthesis (or "genera­ tion"). This brings us to the third reason for this book's importance. The SSE is expres­ sed within the framework of the Meaning-Text Theory of language description. It presents a succinct overview of the MTT model, and then gives this model new life by showing exactly how it treats the important syntactic phenomena of English. The

XIV

EDITOR'S PREFACE

MTT model, developed over a period of twenty years by a group of Soviet linguists led by Mel'čuk, first became known to Western linguists in the late 1960's when it was used as a basis for machine translation systems. Although the MTT gained a number of adherents during the 1970's, its spread in the West has to date been hindered by the lack of sufficiently developed treatments of English. This book should go a long way towards making the MTT accessible to a broad community of linguists and computer scientists, who may now give it the attention it deserves. The SSE is also a book which makes serious demands on the reader. Many linguists, having been raised in the tradition of phrase structure grammar, will want to look carefully at the theoretical overview (chapter I). The use of a dependency formalism, which downplays phrases and their syntactic categories in favor of lexical items and explicit syntactic relations among them, requires nothing short of a re-thinking of syntax. Speakers of languages such as English may need some time to appreciate the fact that Mel'cuk's dependency trees, with their explicitly labeled syntactic relations, derive some of their merit from being far better for representing "word-order free" languages (including Russian). It is only fair to say that the preparation of this book also made heavy demands on the editors, typists, proofreaders and typesetters, and all others who played a role in bringing it into being. Very few of us foresaw what we were in for in 1980 as we planned the production of this essentially finished book. As for myself I must apologize for the time it has taken, and for not having succeeded in shortening and tidying all of Igor's herculean sentences. Those which remain should be seen as monuments to an irrepressible spirit, always ready to sneak in another example or improve the content by yet another parenthetical. Let it be known, however, that the hiker-author who always carries the heaviest pack was also the one who did more than all others combined to speed the production and make the SSE a reality.

R.K., September 11, 1986

FOREWORD

The present book has come into being as the result of joint work by Nikolaj Pertsov and myself. We started studying English surface syntax, with a system of automatic text processing in view, in Moscow — as early as 1970. Our first preliminary report was published as Mel'cuk and Percov 1973a, followed by Mel'cuk and Percov 1973b, which contained a tentative list of English surfacesyntactic relations. Then a complete list of English syntagms (roughly, surfacesyntactic rules) appeared: Mel'cuk and Percov 1975. We had another year and a half to check our description against texts, discuss it with colleagues, and collect more data. In May 1977 I was forced to leave the Soviet Union, and the close contact with my friend Pertsov was thereby broken. It became impossible to continue working on the book together with a distance of several thousand kilometers between us; so I had to take the entire responsibility upon myself and go ahead to the best of my personal abilities, without Dr. Pertsov at my side, deprived of his logical insights, his particular attention to important details, and his excellent knowledge of the whole system. In the final version, this book contains two papers written by us together: Mel'cuk and Percov 1973b (= § 4 of Chapter II) and 1975 (= § 3 of Chapter IV), as well as Appendices I through IV, all of them heavily corrected and in some places completely reworked by myself. These two sections make up, quantitatively and conceptually, the kernel of the book. The rest, i.e. Introduction, Chapters I and III, §§ 1—3 of Chapter II, and §§ 1—2 of Chapter IV have been written by myself, so that I am alone to blame for all the shortcomings to be found there. To conclude my foreword, I would like to express my most heartfelt gratitude to the following people, whose assistance has been vital in the preparation of the manuscript : — Leo Stern, who translated Mel'cuk and Percov 1975 into English and so laid the foundations of the book; — Leo Elnitsky, who went through the whole mansucript many times, hunting down errors, clumsy expressions, misprints, etc.; — Line Arè s, who kindly agreed to type the manuscript, and Marc Blain, who made all the necessary arrangements; t For bibliographic reasons the name of Nikolaj Pertsov is spelled Percov in the references through­ out; see also note 1 on p. 15.

XVI

FOREWORD

— Céline eauchemin, who was instrumental in typing into the final draft of the book an astronomically large number of corrections and additions; — Richard Kittredge, who has done for me everything a good friend can do for a friend, and a good editor must do for a book he is working on; — and the late John Odmark (1937—1980), who invited me to publish the book in a series he was editing. Before his untimely death of a heart attack, he encouraged me to continue my work on it and negotiated its publication with Benjamins Publishers.

Igor Mel'čuk, Montreal, Winter 1985

ACKNOWLEDGEMENTS

The first version of § 4, Chapter II of the present book (= list of surface-syn­ tactic relations), was read by Ju. D. Apresjan, A.S. Čexov, L.L. Iomdin, David Kilby, O.S. Kulagina, and Z.M. Saljapina; Sheldon Klein checked most of our examples. § 3 of Chapter IV (= list of syntagms) was carefully studied by Z.M. Saljapina. The final text of Mel'cuk and Percov 1975 was read by L.L. Iomdin, David Kilby, L.N. lordanskaja, Johanna Nichols, and N.N. Pertsova. Of particular value were the remarks and suggestions made by D. Kilby and J. Nichols, who saved us from many blunders. The completed manuscript of the book underwent a scrutinizing examination by Ju. D. Apresjan, L.L. Elnitsky, L.N. lordanskaja, and Ian Mackenzie; Chapters I through III were read by Alan Dench and J. Nichols. James Steele helped us with the proofreading. We wish to thank all the above-mentioned colleagues and friends for their criticisms and assistance in connection with the book. Of course, none of them should be held responsible for any inconsistencies or errors that may have survi­ ved.

INTRODUCTION

This Introduction is intended to answer the following three questions: 1. What have we written this book for? 2. What are the most salient characteristics of our syntactic approach — and what is our main concern in the book? 3. How is the book organized? Each question is taken up in a special section. §1 Three raisons d'être of this Book To write a fairly complete description of English syntax is an extremely dif­ ficult task; for foreigners, like ourselves, without a perfect command of the language, it may well be impossible. Why, then, try it at all? Our reasons are im­ portant, and they must be explained in order not only to justify our efforts, but also to help our reader understand better what we are after. • First, this book is written within a new theoretical framework called the Meaning-Text Model Theory. (Chapter I explains this theory in some detail.) Eng­ lish, so widely known and so well studied, offered us exceptional illustrative ma­ terial — in the sense that to prove the utility of a new linguistic theory one simp­ ly has to show how it works for English. Moreover, since semantics is relatively universal and English morphology is so poor, it is exactly the syntactic level that is best suited to our purpose. English syntax is richly developed and sufficiently 'exotic' to provide many challenging problems. So we needed this model of Eng­ lish syntax as a piece of evidence in favor of the Meaning-Text linguistic theory. • Second, strange as it may seem, in 1980, when this manuscript was being completed, there was, to the best of our knowledge, no description of English syntax that is simultaneously FORMAL, COMPLETE, and THEORETICALLYORIENTED. Such major contributions to the study of English syntax as Jespersen 1909— 49, Poutsma 1926—29, or Kruisinga 1931—32, with all their permanent relevance to the field, are in no way formal (which does not of course diminish their value); cf., however, Jespersen 1969, which tries, and with excellent results, to introduce a formalism into syntactic descriptions — but deals more with the theory of ge­ neral syntax than with English.

2

SURFACE SYNTAX OF ENGLISH

Stockwell et al. 1973, on the contrary, uses an elaborate transformational formalism, providing a coherent picture of many surface-syntactic phenomena of English, but it is very far from complete (an exhaustive description of English was not the authors' goal). Huddleston 1975, with all its clarity, is, as the author himself puts it, "rather an introduction to the general theory of transformationalgenerative grammar, with extensive and systematic exemplification from selected areas of English" (page IX). Likewise, Chomsky 1975, rich in profound insights and thought-provoking analyses, is devoted to sharpening the methodology of transformational-generative grammar and consolidating the foundations of the theory, not to a systematic description of English structure. Emonds 1976 is also somewhat in the same line. The author aims at creating "a sort of handbook of (English and French based) syntactic theory" (p. vii), not a description of Eng­ lish syntax. Therefore, his main thrust is ramified syntactic argumentation in fa­ vor of or against various transformational analyses of some particular construc­ tions. (The book does contain, however, interesting descriptive statements concerning the there is/are construction, for-to infinitive, parentheticals, place­ ment of adverbiais, etc.) Sager 1968 is both formal and fairly complete but was intended for computer text processing, and because of this is overloaded with technical details; more­ over, until recently it was not available as a regular publication (now see Sager 1981). There are several computer-oriented descriptions of the same kind, but they are virtually unavailable on the general market. As an example, Kittredge 1973 can be mentioned, including Annex I of the same progress report (pp. 117— 178). This is a fully formalized surface-syntactic description of English but it is oriented exclusively towards automatic text analysis and is presented in a com­ puter-encoded form. Huddleston 1971 and Quirk et al. 1972 both are very accurate descriptions, rigorous and consistent; both approach our ideal to some extent. However, the former is neither quite complete nor sufficiently formalized, while the latter in­ cludes some non-syntactic sections (morphology, word-formation, and the like) and is not formalized either. A very special place belongs to Nida 1966 (written as early as 1943, though first published in 1960). It still remains the most extensive description of English syntactic constructions in terms of immediate constituents, i.e. phrase-structure, — but with indication of the head constituent, where it is relevant from Nida's viewpoint (in so-called endocentric constructions). Moreover, Nida characterizes the prosody of every construction on the surface (the pitch and the pauses, as well as the stress); he is rather an exception in this respect. He also supplies lists

INTRODUCTION

3

of words capable of filling a given position in the given construction. Very de­ tailed and accurate, Nida's Synopsis served for us as a primary source of data (and examples) and a model of clear organization. However, with its venerable age, the volume cannot be deemed quite satisfactory from the viewpoint of to­ day's theoretical standards. In addition, a higher level of formalization seems achievable now. Thus we try, with the present book, to fill a gap existing in the literature, by providing a systematic and relatively complete description of English surface syntax which is, at the same time, fully formal yet not designed exclusively for computer processing — but rather for human readers. At this point, an important clarification seems in order. When speaking of this book as of a relatively complete description, we mean that we try to describe the WHOLE of English surface syntax, and not some selected parts of it, as is often done in many contemporary syntactic studies. Of course we do not claim exhaustiveness of our data or definitive coverage of all the problems we raise. On the contrary, the number of lacunae we are aware of is very large (some of them will be indicated below). But our intention has been to outline, within the limits of our possibilities, the entire surface-syntactic edifice of English from the foun­ dations to the rooftop, even if several partitions will be absent, many walls unfi­ nished, some toilets leaking, etc. • Third, nearly all the formalized presentations of English syntax we have heard of use phrase-structure formalism with the addition of transformations (in the sense of Chomsky). The only exception known to date is string grammar, launched in Harris 1962. But string grammar is still closely enough related to the phrase-structure approach; moreover, until recently it has not been very popular with specialists of English syntax (Sager 1968 and 1981, mentioned previously, are unique in using the string grammar formalism). Lytle 1974 should probably be mentioned in this connection, too. The book presents an inventory of syntactic rules in English (40 rules covering all major ty­ pes of English phrases) written in terms of a new formalism, devised by the author: so-called junction grammar. But as far as we can see, this is again a varie­ ty of the phrase-structure approach (although avoiding the use of transforma­ tions). In contrast to all these works, in the present book we put forth the pure formalism of dependency trees, a novelty for many Western linguists. Although syntactic representations in terms of dependencies have been discussed in Ameri­ can professional journals (e.g., Hays 1964b and Robinson 1970a, b) and even

4

SURFACE SYNTAX OF ENGLISH

more often in Europe (see references on page 78), a full-fledged description of English using the syntactic dependency formalism has never been attempted be­ fore, as far as we know. Yet, the dependency formalism seems to us to be the best one for the description of syntactic phenomena in any language. This con­ viction has prompted us to show its various advantages on the basis of English data. To summarize, we have written this book with three goals in mind: — to present a fully elaborated fragment of a specific Meaning-Text model, and do it for a well-known language, in order to make the model available to the widest possible range of linguists; — to provide a complete, formal and theory-oriented description of English surface syntax; — to apply the syntactic dependency formalism to English. Let it be stated that we are not attempting a systematic review of the enorm­ ous literature on English syntax. We cannot do more than indicate a few works that have a direct bearing on what we are discussing. A solid bibliography is now available for the serious student of English syntax: Smith and Johnsen 1981. §2 Five Main Characteristics of the Syntactic Approach in this Book In order to facilitate the reading of our syntactic description, it may be helpful to state how it compares with other descriptions known in the field. But even a cursory survey of the existing literature is out of the question for reasons of space. Therefore, we will limit ourselves to briefly outlining some of the basic elements of the approach adopted in this book, emphasizing what seems to us most relevant. To do this, though, we will need certain concepts and notions that are introduced and explained only later on. So we ask our reader to have patience and be satisfied for the moment with less than a full exposition. In this section, we aim at a very general picture, leaving out all the details. Our description of English surface syntax is characterized by the following five properties: 1. A maximally restrictive and narrow framework of analysis. 2. A stratificational viewpoint: emphasis on representations of utterances at dif­ ferent levels. 3. A static viewpoint: no generation, no transformations, no ordering of the rules.

INTRODUCTION

5

4. Dependency trees as the only descriptive apparatus for surface-syntactic structure. 5. Extensive use of labeled surface-syntactic relations. The first'three properties are by no means specific to syntax: they are characte­ ristic of the whole Meaning-Text Model, of which our surface-syntactic descrip­ tion is only one component. The last two, on the contrary, are exclusively syn­ tactic. 1. Framework of analysis In this book, we try to keep the domain of our description as narrow and restricted as possible in at least two respects. First, we deal only with the most neutral variety of Written Standard English, as used in the United States. Our main object of study is non-fiction prose (scientific, journalistic, and the like), although we also use sentences from novels and short stories provided they are not archaic, too informal or overly collo­ quial. Our slogan is 'No marginal phenomena'. What is covered by our model belongs to the common core of different styles, registers, and sublanguages of American English. Second, we concentrate on surface-syntactic structure only and exclude various related problems that often emerge in syntactic descriptions of English. For example, take such pairs of sentences as They quickly sold the book vs. The book sold quickly or Mary washed the clothes clean vs. The clothes washed clean. The use of the verbs in the second members of these pairs (called 'ergative verbs' in Huddleston 1971: 65ff. and analysed in van Oosten 1977), the nuances of their meaning, their cooccurrence with modais or adverbs may take up dozens of pages in a work on English syntax. But to us, all this belongs to lexicography or semantics, or, possibly, to deep syntax, rather than to surface syntax, which is our only concern. Similarly, we do not discuss such topics as subject- and ob­ ject-raising, the use of passive as opposed to active voice, selection of terms of address, and many other such questions. This approach follows one of the main principles of the Meaning-Text model theory: a strict separation of all linguistic levels and strata, each to be treated in the most autonomous way possible and then related to the others by special rules. Let it be emphasized that the restrictiveness of the approach we have adopted does not conflict with our objective of completeness. We intend to be complete and systematic WITHIN THE NARROW RANGE OF PHENOMENA we have set out for ourselves at the beginning. Our field is small: only the surface syntax

6

SURFACE SYNTAX OF ENGLISH

of only the most typical major sentence types in only the most neutral style of Written Standard American English. Yet this field is well delineated and we claim to have plowed it thoroughly. 2. Utterance representations at different levels Prior to formulating any linguistic model, we introduce representations of utterances at those levels that are of interest to us. In this particular case, these are the surface-syntactic representation and the deep-morphological representa­ tion of English sentences. Linguistic representations must be written in formali­ zed languages constructed by the researcher, and their justification or substantia­ tion is completely outside the linguistic model as such. The only task a linguistic model is supposed to perform is to ensure the correspondence between two representations of adjacent levels, and nothing beyond that. A simple comparison may help to grasp this idea. Suppose we are to translate a text from Japanese into English; then our problem is not to discuss Japanese or English, but rather to find the correspondences that would render each Japanese sentence in English. Simi­ larly, in what follows, our task as we see it lies in translating one representation into another, and not in discussing the representations. For precisely this reason, a rigorous and explicit formulation of the represenations used becomes an abso­ lute necessity. (Note that this simile is for us much more than just a figure of speech. We really do believe that the functioning of human language consists in translating between different levels of utterance representations. For more on this subject see Mel'cuk 1978.) A linguistic representation, in our opinion, need not be a homogeneous object. Rather, our representations are sets of different formal objects, called structures, each of which is devoted to one particular aspect of the phenomenon represen­ ted. In this we also follow our general principle of strict separation of levels. 3. A static viewpoint Our syntactic description is intentionally static: we see our task only as formulating all the necessary correspondences between surface-syntactic and deep-morphological representations in English, and not as describing the proce­ dure(s) that could actually ensure the transition from one level to the other. The compiler of a bilingual dictionary lists all the correspondences between the words of the two languages, and no more; the algorithm for looking up a word and then using its equivalent lies outside his competence. In much the same way,

INTRODUCTION

7

we content ourselves with describing static correspondences — without consider­ ing the dynamic mechanisms needed for their implementation. This means, first of all, that no generation is involved — at any point of our exposition. Nothing is generated. Both representations we are working with (= the surface-syntactic and deep-morphological representations of sentences) are given as postulates in this book; their well-formedness should be ensured by for­ mal rules of the filter type (for more see p. 78ff.). Furthermore, there are no transformations involved. Nothing is actually transformed. Returning to the above comparison, in a Japanese-English dic­ tionary no Japanese expression is transformed into an English one. Rather, an appropriate English expression is chosen under the control of the Japanese input, which does not undergo any change. The same happens in our syntactic model: an appropriate deep-morphological configuration is chosen under the control of the surface-syntactic input, but the latter remains intact. (Among other things, this allows us to avoid creating complex mechanisms to preserve information about transformed items, such as the recent theory of traces in generative gram­ mar.) Finally, there is no ordering of rules. Each rule is written so as to include all that is necessary for its proper functioning — in explicit form. Since we do not consider the process of transition from one representation to another, we do not have to determine which rule should apply before or after which other rule. In this respect, our description resembles the most traditional ones. 4. Dependency trees Almost all of the familiar approaches to English syntax are couched in terms of constituency, or phrase-structure, systems. The elements making up the surface-syntactic structure are linearly ordered and grouped into higher-order units, which are considered to be non-terminal syntactic items. To indicate major syntactic roles, such as 'the subject of' or 'the object of', special phrase-structure configurations are used. This book, on the contrary, uses dependency systems as the only surface-syn­ tactic formalism. The elements in the surface-syntactic structure are not ordered linearly and no higher-order units are shown (thus there are no non-terminal items). Instead, the surface-syntactic elements, which are images of wordforms, are connected by binary directed relations of dependency so that not only the major, but all of the syntactic roles are indicated in a straightforward manner. Much more will be said about dependency systems in Chapter II, § 2, p. 48 ff.

8

SURFACE SYNTAX OF ENGLISH

5. Labeled surface-syntactic relations Perhaps the least familiar notion extensively used in this book is LABELED SURFACE-SYNTACTIC RELATION (SSyntRel). We consider SSyntRel's to be linguistic units, more or less like phonemes, or morphs, or values of gram­ matical categories (= grammemes). Forty different SSyntRel's are proposed for English, starting with certain well-known ones, such as the predicative or comp­ letive SSyntRel's, and including some new ones, such as the colligative or appositive-adverbial SSyntRel's; see the tentative list in § 4 of Chapter II. The notion of SSyntRel is introduced and discussed at some length in §§ 1—2 of Chapter II. Here it will suffice to say that SSyntRel's occupy a central place in our presentation. It can be said that the book as a whole answers only one question: How can a given English surface-syntactic relation be expressed at the deep-morphological level (i.e., by what deep-morphological string)? Or, if viewed from the opposite direction: Which surface-syntactic relations can be expressed by a given deep-morphological string (if it expresses some)? The nearest analogy to this approach from the domain of morphology would be as follows: to list, for each grammatical category of a language, all the surface expressions of all its values, along with all the conditions for the proper use of each individual expression. (Or vice versa: to list, for each morphological means, all grammatical values it can manifest in the given context.) For a successful reading of this book, it is vital that our reader fully under­ stand our most important message: (1) This book proposes to set forth the correspondences between the English surface-synItactic relations and the deep-morphological strings that implement them.

This is what the main thrust of this presentation is directed at. Everything else plays an auxiliary, even if important, role. However, statement (1) needs to be qualified. We fully understand the imperfection of our work. It is clear that many of the solutions we have proposed can turn out to be wrong and that incorrect assertions about English will ine­ vitably be found in this book. Therefore, we cannot (and do not) claim definitive truth for our present description. We agree in advance that our SSyntRel's are not introduced in the best possible way, and that the deep-morphological for­ mations that implement them are not presented quite ideally. But what we do claim is the status of a preferred method for our approach as a whole: ( 2 ) T h i s book proposes to show that the postulation of SSyntRel's and the description of their surface realizations is a very efficient and promising method of syntactic study.

INTRODUCTION

9

We will be happy if it is felt that we have succeeded in doing this with a reasonable degree of convincingness. Our only ambition, thus, is to pose the right problems and stake out the relevant gaps, which must be closed through further empirical studies. §3 Organization of this Book We start with a general survey of the Meaning-Text linguistic approach and an outline description of Meaning-Text models, putting special emphasis on three of their components: the semantic, deep-syntactic and surface-syntactic com­ ponents. All this material is covered in Chapter I, which is composed of four sec­ tions : § 1 - General Remarks. § 2 — Levels of Utterance Representation in the Meaning-Text Model. § 3 — General Design of the Meaning-Text Model. § 4 — Five Basic Principles of the Meaning-Text Approach. On the first reading, Chapter I can be skipped altogether or leafed through quick­ ly, to be returned to later, when theoretical problems arise in connection with the specific analyses proposed in the following chapters. The next two chapters are devoted to two major types of utterance repre­ sentation that, as indicated above, occupy a central position in our syntactic approach. Chapter II examines the surface-syntactic representation (SSyntR) of English sentences. The material is arranged in four sections: § 1 — General Characterization of the Surface-Syntactic Representation. Here, the composition of the SSyntR is specified. § 2 — Surface-Syntactic Structure. In this section, dependency formalism is compared to phrase-structure for­ malism, and the notion of syntactic dependency is developed. Surface-syntactic relations are introduced and commented upon. Some criteria for SSyntRels are stated. § 3 — Conditions for Grammatical Correctness of English Surface-Syntactic Structures. § 4 — Surface-Syntactic Relations in Modern English (a tentative list, with many examples). Chapter III describes the deep-morphological representation (DMorphR) of English sentences in a manner parallel to that adopted in Chapter II:

10

SURFACE SYNTAX OF ENGLISH

§ 1 - General Characterization of the Deep-Morphological Representation. § 2 — Deep-Morphological Structure. § 3 — Conditions for Grammatical Correctness of English Deep-Morphological Structures. § 4 — Morphological Variables and Values of Wordforms in Modern English (a tentative list, with examples). When both representations are introduced, we proceed to the main part of the book: description of the correspondences between them. The component of the Meaning-Text model responsible for these correspondences is called the surface-syntactic component. This component, the focal point of our study, is dealt with in Chapter IV, divided into three sections: § 1 — General Characterization of the Surface-Syntactic Component. § 2 — The Notion of Syntagm. The notion of syntagm is defined and exemplified, which prepares the ground for the next section, the most important in the book: § 3 — Syntagms of Modern English. It can be said, without exaggeration, that the whole book has been written because of this section. The list of English syntagms is the core of the book and the main goal of our effort. Footnotes appear at the end of each section. Four appendices are added to the main body of the volume: Appendix I lists all the part-of-speech symbols and syntactic features used in the description. Appendix II shows standard subtrees appearing in the syntagms as variablės providing for more compact presentation. Appendix HI contains a list of lexemes mentioned in the syntagms. Appendix IV gives some 50 examples of English constructions not covered by our description. And a Subject and Term Index should help the reader to find his way through the text under the onslaught of so much new terminology. So much for the external, 'physical' organization of our study. As far as its internal or conceptual organization is concerned, it should be emphasized that it differs sharply from most American publications in the field. It is not our purpose to advance hypotheses, present systematic argumentation or make serious claims about English or about any linguistic universais. We simply state — from a logical viewpoint, very much as postulates — a surface-syntactic and a deep-morphological representation for English, and then formally describe pos-

INTRODUCTION

11

sible correspondences between them. Arguments in favor of our proposals and solutions are given quite sporadically and do not play an important role. We focus on correspondences, as stated in (1) and (2), p. 8, adding a few comments and examples to facilitate the exposition. Now that our reader has been properly warned, we wish him a safe and inte­ resting expedition through the dependency forests of Meaning-Text syntactic land.

Chapter I A Brief Outline of the Meaning-Text Theory and the Corresponding Linguistic Model

§1 General Remarks The Meaning-Text Theory (henceforth, MTT) was put forward by Alexander K. Zholkovsky 1 and one of the present writers in 1965 (Zolkovskij and Mel'cuk 1965); two years later, a major presentation of the theory appeared (Zolkovskij and Mel'cuk 1967) and was soon translated into English and then into French. Shortly afterwards, Jurij D. Apresjan joined us, and thus was formed the nucleus of what was (partly as a joke) called the Moscow Semantic Circle. Over a ten-year period, some 20 people contributed to the work on a Meaning-Text Model of Rus­ sian. Basic general readings on MTT, in addition to the two titles just mentioned, include Mel'cuk 1970, 1973, 1974a, 1974b, 1978, 1981, Mel'cuk and Zolkovskij 1970, Apresjan 1974, 1980. A number of papers dealing with some more specific topics in syntax will be indicated in the main body of this book. Conceived and developed as a general theory of human language, MeaningText Theory is based on the following two postulates: POSTULATE 1. Every speech event presupposes exactly three main compo­ nents: — content, or pieces of information to be communicated, which are called meaning(s); — certain forms, or physical phenomena to be perceived, which are called text(s) — a many-to-many correspondence between an infinite set of meanings and an infinite set of texts, which constitutes language proper (or 'language in the narrow sense of the term'). This postulate can be diagrammed as follows: (1)

0
A natural language is thus viewed as a logical device which establishes the corres­ pondence between the infinite set of all possible meanings and the infinite set of all possible texts and vice versa. This device ensures both speaking, i.e. the con­ struction of linguistic utterances which express a given meaning, and the under-

MEANING-TEXT MODELS

13

standing of speech, i.e. the comprehension of possible meaning expressed by a given utterance. POSTULATE 2. Hypotheses about devices of the type illustrated in (1) can be formulated as functional,2 or cybernetic, models, with the actual language considered a black box where only the inputs and outputs can be observed but not the internal structure. Such models are systems of rules approximating the meaning-text correspondence. The MTT proposes to describe languages with the aid of such models, which are naturally called Meaning-Text Models (MTM). Proceeding from Postulates 1 and 2, one can characterize the MTM by the following two important properties. First, the MTM is not a generative but rather a translative, or purely equational, system. It does not seek to generate (i.e. enumerate, or specify) all and only grammatically correct or meaningful texts, but merely to match ANY GIVEN meaning with all texts having this meaning and, conversely, ANY GIVEN text with all meanings the text can have. Second, the MTM is no more than a fragment of the full-fledged model of human linguistic behavior: (2) Only fragment II, i.e. Language, is the subject of linguistics proper and should be represented as an MTM. Fragment I is the -subject of various fields, such as philosophy and psychology, including what is called artificial intelligence; and fragment III is the subject of acoustic and articulatory phonetics. With all their significance for language and the study of language, neither falls within the scope of the MTT. However, even fragment II is not at present represented by an MTM in full. To simplify our task, we make abstraction from a number of relevant aspects, properties, and phenomena of natural language — in the hope that we will ob­ tain a clearer and more insightful picture of what remains under our lenses. This is the only justification for the many deliberate omissions that the reader may notice. Thus for practical reasons we impose on MTM's in this book the following six limitations (that can eventually be removed): (i) Functions of natural language other than the communicative function are not considered at all. Language is treated exclusively as a communication system.

14

SURFACE SYNTAX OF ENGLISH

(ii) The extremely important problem of language acquisition and development is deliberately ignored. (iii) No attempts have been made so far to relate the MTM experimentally with psychological or neurological reality. An MTM is no more than a model, or a handy logical means for describing observable correspondences. (iv) The correspondence between meanings and texts is presented statically, i.e., as a correspondence between elementary fragments of meaning and equally elementary fragments of text. The procedures for actually moving from meanings to texts and vice versa, or the operations that have to put together those elementary fragments to produce actual life-size representations, are left out of consideration (since we claim they lie outside the domain of linguistics 3). Therefore, the problem of rule ordering does not arise. (v) The feedback between meanings and texts in the actual process of speaking or listening (changes in the original message under the influence of the text already constructed and uttered, etc.) is not taken into account. (vi) We exclude from the scope of an MTM the analysis of meaning as such: the detection of semantic anomalies, i.e. contradictions, trivialities, absurdities, etc. requires a different type of device. Thus, if confronted with such de­ viant meanings as 'Colorless green ideas sleep furiously' or 'John ate up the sincerity of his car', the MTM should provide for them perfect English (or French, or Russian,. . .) sentences. So it should be able to detect the formal, or grammatical, anomalies of a text but it does not deal with the semantic ones. (This is done on purpose, in order to reflect the essential asymetry of meanings and texts.) Only a very sketchy description of the MTM can be given in this chapter. We do not mention any other feasible solutions and draw no parallels with the closely related models proposed by the theories of transformational grammar (Chomsky and his followers), generative semantics (G. Lakoff, McCawley), stratificational linguistics (Lamb), case grammar (Fillmore), and other more recent developments, including relational grammar. Suffice it to state that all these have significantly influenced the authors of the MTM; a qualified reader will easily recognize some common points and perceive the differences.

15

MEANING-TEXT MODELS

NOTES 1. (To page 12.) There are problems in transcribing some Russian names into English. For example, Zholkovsky's name will be written Zolkovskij when it is transliterated directly from Cyrillic type; moreover, various erratic spellings are found in American publications, e.g. Zholkovski , or Zolkovsky, which had to be adopted in the literature cited. The same applies to Apresjan (spelled sometimes as Apresyan), Mel'čuk (spelled also Melchuk and Mel'chuk) and Pertsov (Percov). 2. (To page 13.) The term functional, which is now a buzzword in linguistics, as it is used here, has nothing to do with functional sentence perspective (topic-comment problems), nor with grammatical functions (such as 'subject of', 'object of', etc.). We say that X is a func­ tional, or cybernetic, model of an object Y if and only if X is a system of rules simulating the functioning, the 'work', or the behavior of Y, with no claims as to its observable struc­ ture. 3. (To page 14.) But an MT model does of course enumerate the possible kinds of changes between adjacent levels.

§2 Levels of Utterance Representation in the Meaning-Text Model We assume that both meanings and texts can be represented in an explicit DISCRETE manner, using formal languages devised specially for this purpose. For texts this is more or less obvious: phonetic transcriptions — or, for that matter, conventional spelling systems — have existed for a long time. The ap­ proximation they ensure for the continuous acoustic flow of speech is generally deemed satisfactory. Yet, to the best of our knowledge, no semantic transcrip­ tion exists for meanings. Nothing, however, should prevent us from believing that such a transcription is possible and from trying to construct one, if only as a working hypothesis. A version of semantic transcription, or, as we call it, semantic language, has been proposed within the framework of MTT. In what follows, the existence of exactly this specific version of semantic language is presupposed. Thus our MTM deals in fact with FORMAL REPRESENTATIONS of meanings and texts, rather than with 'genuine', 'real' meanings and texts. (We prefer not to delve into the highly philosophical question of the reality of meanings and texts, etc.) Therefore, formula (1) of § 1 can be rewritten as (1) below: MTM

1)

Sem(antic) R(epresentation)і

Phon(etic) or Graph(ic)

R(epresentation)і

16

SURFACE SYNTAX OF ENGLISH

From here on, we speak only of representations, which play a very important role in MTT; more about them will be said in this section. One of the most basic facts about natural language is the following one: (2) A given meaning of sufficient complexity can normally be expressed by an astronomical­ l y large number of nearly synonymous texts.

Consider sentence (3), quoted (with slight modifications) from Newsweek maga­ zine (Sept. 15,1980): (3) The Food and Drug Administration has seriously cautioned expectant mothers to avoid one of life's simple pleasures: a cup of coffee. The meaning of this sentence can be expressed, e.g., by (4) as well: (4) Pregnant women have been strongly warned by the FDA against drinking coffee, one of the small pleasures of life. Except for function words (articles, auxiliaries, . . .) and 'technical terms' like FDA, coffee, pleasures and life, all the other words have been changed, as has the overall structure. Yet the meaning is preserved relatively intact. With additional variations, the number of English sentences nearly-synonymous with (3) and (4) runs into thousands.1 We see, then, that synonymy occupies a special place in language. By way of paradox one can say that natural language is a system capable of producing a great many synonymous texts for a given meaning; the mastery of a language boils down to the control of its synonymic means. (Note that homonymy is not given special attention within the MTM approach. Problems of disambiguating utterances, discovering the intended meaning, etc. are deliberately ignored since none of them pertains to language as such.) On the practical side, the MTM has to match, generally speaking, a given meaning with too many different texts — and a great many different texts have to be reduced to the same meaning representation. This makes it practically impossible to establish a DIRECT correspondence between semantic represen­ tations and phonetic representations, and so two intermediary levels of utterance representation have to be introduced. In complete accordance with the generally held view these are syntactic and morphological levels, the former aimed at the sentence as a structural object, and the latter dealing with the word. But the MTM approach is distinctive in that all the levels, except for the semantic level, are split into two sublevels each: a deep sublevei geared to the meaning, and a surface sublevei determined by the form. So, starting with meaning, we have a total of SEVEN representation levels in the MTM: 1.Semantic Representation (SemR), or the meaning. 2.Deep-Syntactic Representation (DSyntR).

MEANING-TEXT MODELS

17

3. 4. 5. 6.

Surface-Syntactic Representation (SSyntR). Deep-Morphological Representation (DMorphR). Surface-Morphological Representation (SMorphR). Deep-Phonetic Representation (DPhonR; this is what is commonly called phonemic or phonological representation). 7. Surface-Phonetic Representation (SPhonR, which is called in the literature simply phonetic representation), or the text. A representation is a set of formal objects called STRUCTURES, one of which is distinguished as the MAIN one, with all the others specifying some of its characteristics. Linguistically, each structure depicts a certain aspect of the item considered at the given level. Utterance representations are written in formalized languages defined by the researcher. With all this in mind, we can now present (1) in full: (5)

The top line in (5) is a sequence of the utterance representations at all the seven levels, with the correspondences between any two adjacent levels shown by twoheaded arrows. In the bottom line we see the components of the MTM and their functions: thus semantics provides for the correspondence between the semantic representation of an utterance and all the sequences of deep-syntactic represent­ ations carrying the same meaning, etc. (See § 3 for more information.) Let us now give examples of successive representations for sentence (3) and some near-synonymous sentences. We will start from the meaning and work upwards toward the text. Some very sketchy explanations will be provided, with heavy reliance, though, on the reader's intelligence and on references. But before we proceed with our examples, a very important clarification concerning the notion of synonymy, or identity of meaning, seems to be in order. Two theoretical approaches to synonymy must be distinguished. On the one hand, we need to consider synonymy of linguistic expressions within an 'absolute', or 'ideal', framework, when we take into consideration all potential semantic contrasts and differences which are present in the language and may be exploited by the speaker whenever he wishes. These differences are reflected in the dictionary as fully as possible, and no two expressions should be called synonymous if a semantic difference CAN be found among them.

18

SURFACE SYNTAX OF ENGLISH

On the other hand, however, it proves very useful to consider synonymy with­ in a 'relative', or 'approximate', framework, when we pay attention only to those semantic contrasts that are indeed observed by the speaker in actual dis­ course. As a matter of fact, many semantic differences well known to the speaker are ignored when he speaks. Assisted by the context, the speaker can afford being a bit careless and use less than exact semantic equivalences. It is this textual, i.e. actual rather than potential, synonymy that appears in the examples below. Thus we call two expressions synonymous if, for the purposes of a speci­ fic act of communication, NO semantic difference MUST be stated among them. That is, sometimes we are aware that some semantic distinctions can be detected between sentences we take to be synonymous; but we feel that we may safely make abstractions from them. Semantic Representation A sample Sem(antic) R(epresentation) simplified for illustrative purposes is given below as (6): Fig. 1.

Fig. 1 : The SemR of (3) and (4), as well as of all the other sentences (or sequences of sen­ tences) synonymous with them.

MEANING-TEXT MODELS

19

Notice that the letters , ,  and the numbers 1, 2, 3 that appear at the left side and top of the diagram are not part of the representation: they are used merely for referring to locations in the diagram. The shaggy lines mark the boundaries of theme and rheme, cf. below. SemR (6) can be worded approximately as follows: (6') The FDA has intensively attempted to communicate [A1—2] to all pregnant women [Bl ] that they should not [A2] drink coffee, which belongs to the things that constitute simple pleasures of life for them [B2-3, C2-3]; the reason for that is that drinking coffee is dangerous for pregnant women [B2]; this attempt took place before the moment of speech, yet its results persist until this moment [A2—3; this represents the English present perfect]'. A SemR of an utterance consists of two structures: the semantic structure and the semantic-communicative structure. The Sem(antic) S(tructure) specifies the meaning of the utterance under con­ sideration as independently of its linguistic form as possible. The distribution of meaning among the words, clauses or sentences is ignored; so are such linguistic features as the selection of specific syntactic constructions, and so on. At the same time the SemS tries to depict all aspects of meaning except for the speaker's intentions concerning the organization of his message; what should be presented first, what should be foregrounded, etc. These intentions are taken into account in the second (= semantic-communicative) structure of the SemR. Formally, a SemS is a simple connected directed graph without circuits, which we call a network. 2 A vertex, or node, of a SemS is labeled with a semantic unit, or SEMANTEME, of the language: a meaning which can be either elementary (= a SEME) or complex (see immediately below). The number seen to the right of certain words appearing in our SemR is an index identifying a particular sense of the correspon­ ding word. Thus 'attempt 1 (in [A1]) denotes 'a try', while 'attempt 2' is 'an assault' (as in attempt on one's life) etc. All numerical indices accompanying words used in the examples always have exactly this function. A complex semanteme consists of semes or less complex semantemes and can be represented by a semantic network, which then specifies its semantic decompo­ sition. For example, the semanteme 'earlier' found in (6) [see A3] can be decom­ posed as follows:

20

SURFACE SYNTAX OF ENGLISH 'earlier'

'moment

A

1

'moment 1

'less'

B

The index 1 in 'moment 1' indicates, as just stated, the intended sense of the word moment: 'a specific point in time'. The number 2 on the arrow leaving the node 'moment 1' specifies the 2nd argument of the predicate 'tі is the moment 1 of . . .'; for more details see below. The physical difference of the two nodes labeled 'moment 1' shows that these are two different t[, so that the right part of (7) can be read as follows: 't 1, the moment 1 of A, is less than t 2, the moment 1 of B'. Semantic decomposition can go on, as deep as we need or want it, until semes are reached. For the time being, a definitive list of semes, or semantic primitives, is not available ; some likely candidates are '(some)thing', 'more', 'say', 'this speech act', 'not', 'set' [in the mathematical sense], 'space', 'time', 'and', 'or'. (For a different view of semantic primitives see Wierzbicka 1972 and 1980, where ex­ actly 13 semantic primitives are proposed and argued for.) However, in our SemR's complex semantemes are used for the most part, since a complete analy­ sis into semes would make the semantic network unreadable. The names of the semantemes are word senses picked from a good dictionary. For the illustration in this chapter, we have taken English word senses from The American Heritage Illustrated Dictionary of the English Language ; thus 'persist 3' in A3 is 'continue in existence until. . .' (rather than 'be obstinately insistent in . . .' or 'hold stead­ fastly to . . . despite obstacles'). Two major classes of semantemes are distinguished: 1) FUNCTORS, further subdivided into PREDICATES (relations, properties, actions, states, events, etc.), LOGICAL CONNECTIVES ('if', 'and', 'or', 'not'), and QUANTIFIERS ('all', 'there exist', plus all numbers); 2) NAMES (OF CLASSES) OF OBJECTS. Both types of semantemes can receive arcs, but only a functor can be the starting point of an arc. The arrows on the arcs point from functors to their arguments.

MEANING-TEXT MODELS

21

An arc of a SemS is labeled with a distinctive number, which has no meaning of its own but only serves to differentiate the various arguments of the same functor. For instance, 'communicate' (8)

A

B

C

means that A is the first argument of 'communicate' (who communicates), B, its second argument (what is communicated), and C, the third argument (to whom the information is passed). The exact role of each argument is specified by further decomposition of the functor: 'A communicates B to ' = 'A, who is aware of B, directly causes  to become aware of B' A deeper decomposition (bearing, e.g., on 'aware' and 'cause') would reveal more subtle links between the functor 'communicate' and its arguments. The Sem(antic)-Comm(unicative) S(tructure)specifies the intentions of the speaker with respect to the organization of the message. It is a commonplace that the same meaning reflecting a given situation can be encoded in different messages according to what the speaker wants. So the Sem-CommS must show at least the following contrasts: — Theme (= topic, designated ) vs. rheme (= comment, designated ), i.e., the starting point of the utterance, as opposed to what is communicated about the topic. In (6), is 'the caution given by the FDA and aimed at pregnant women', and is 'that pregnant women should not drink coffee since this is dangerous for them'. If we interchange the symbols and n this SemR, the message becomes different: e.g., (9) Mothers-to-be had better keep clear of coffee, which can be harmful to them — that is the stern warning issued by the FDA. and show, so to speak, the itinerary through the situation, which is up to the speaker to choose. Note that there can be different layers of topic-comment division. For example, within in (6). and of second order can be indic­ ated: t = 'the FDA', = 'intensively attempted to communicate to all pregnant women'.

22

SURFACE SYNTAX OF ENGLISH

- Old, or given (= known to both interlocutors), vs. new, i.e., communicated by the speaker. - Foregrounded (expressed as a main predication) vs. backgrounded (relegated to an attribute). - Emphatically stressed vs. neutral. All these contrasts are not fully independent. There obviously exist some cor­ relations whose exact nature is not yet known. Semantic-communicative information stands in approximately the same rela­ tionship to the semantic network (= SemS) as do suprasegmental prosodic pheno­ mena to the segmental phonemic string that makes up a sentence. In the simplified examples given here only the first layer of the topic-comment contrast is shown, and that on the first stratum only. Deep-Syntactic Representation To make the contrast between a SemR and a D(eep)-Synt(actic) Representa­ tion) more vivid, we will give two different DSyntR's for two different but roughly synonymous sentences, (3) and (10), see below: (10) The FDA has issued a stern warning to pregnant women: they should not drink coffee, one of life's small pleasures. Both sentences can be assigned the same SemR, namely, (6). Their DSyntR's appear as Figs. 2 and 3, respectively. A D(eep)-Synt(actic)R(epresentation) [of a sentence] consists of four struc­ tures: the deep-syntactic structure, the deep-syntactic-communicative structure, the deep-syntactic-anaphoric structure, and the deep-syntactic-prosodic struc­ ture. The D(eep)-Synt(actic)S(tructure) is a dependency tree. (It is impossible to go into all the technicalities here in order to explain what a dependency tree is; the reader is referred to Chapter II, § 2, item 1 on p. 48 ff. for the relevant informa­ tion and further references.) It represents the syntactic organization of the sen­ tence in terms of its constituent words and relationships between them. A node of a DSyntS is labeled with a GENERALIZED LEXEME of the lan­ guage. A generalized lexeme is one of the following four items: 1) A full lexeme of the language. Semantically empty words, like (strongly) go­ verned prepositions and conjunctions or auxiliary verbs, are not represented: thus in (11) has [cautioned], to [avoid] and [cup] of [coffee] are absent. 3

MEANING-TEXT MODELS

Fig. 2 The DSyntR of sentence (3).

23

24

SURFACE SYNTAX OF ENGLISH

(12)

Fig. 3 The DSyntR of sentence (10).

MEANING-TEXT MODELS

25

2) A fictive lexeme. This can be a lexemic unit presupposed by the symmetry of derivational system, yet non-existent, or a unit representing a meaning which is expressed in the given language by purely syntactic means, i.e. by a construc­ tion. (There are no fictive lexemes in our examples.) 3) A multilexemic idiom, e.g., hit it off 'have good rapport' or pull a fast one [on somebody]' gain an advantage over an unsuspecting person by a subter­ fuge'. In (11) and (12) we find two examples of idioms: one is The Food and Drug Administration, shortened to FDA and represented by a single node; the other example is expectant mother. 4) A lexical function: see below. The symbol of a generalized lexeme must be subscripted for all the meaningbearing morphological values, or grammemes, such as number in nouns, or mood, tense and aspect in verbs. Syntactically-conditioned morphological values, such as person and number in verbs, are not shown in a DSyntS. At this point, two important notions must be introduced: — An individual morphological value is called a GRAMMEME, see below, p. 168 ff. — A lexeme name subscripted for all meaning-bearing grammemes is called a REDUCED DEEP-MORPHOLOGICAL REPRESENTATION (DMorphR) of the corresponding wordform (see § 2, item 3, Chapter II, for more about deepmorphological representations of wordforms). A (standard elementary) LEXICAL FUNCTION (LF) f is, roughly speaking, a semantico-syntactic relation which connects a word or phrase W — the argument of f — with a set f(W) of other words or phrases — the value of f — in such a way that: (i) for any Wl and W2, if f (W1) and f(W2) exist, both f(W1) and f(W2) bear an identical relationship with respect to the meaning and syntactic role to W1 and W2, respectively: 'f(W l) ':'W1'='f(W 2 )': 'W2'; (ii) there are many different W's which can be arguments of f [that is, the meaning of f is abstract enough for f to cooccur with a great number of meanings]; (iii) in most cases,f((W1)≠f(W2), which means that f(W) is phraseologically bound by W. In the MTM, about 50 standard elementary LF's are used. Some examples of LF's: Syn(shoot) = fire [synonym];

SURFACE SYNTAX OF ENGLISH

26

Syn ( shoot) = shell, machine-gun [narrower synonym]; Anti(victory) = defeat [antonym]; Conv21 (include) = belong to [conversive; as in This paradigm includes the locative case = The locative case belongs to this paradigm]; S0(despise) = contempt [derived substantival]; A0(SUN) = solar [derived adjectival]; Magn(need) = great, urgent, bad, Magn(settled [area]) = thickly, Magn(illustrate) = vividly, Magn(belief) = staunch [Very'; an intensifier]; Oper1 (analysis) = perform, Oper1 (attention) = pay, Operi (step) = take, Oper1(favor) = do ['be the subject of]; Oper2(analysis) = undergo, Oper2(attention) = receive, Oper2(control) = be under ['be the object of]; Real1 (promise) = keep, make good on, Real2(attack) =fall to ['fulfill the re­ quirements of X, which is the argument, being respectively the subject (with Real1) or the object (with Real2) of']; Son(cow) = low, moo, Son(windowpanes) = jingle, rattle [typical sound]. Furthermore, there are COMPLEX LF's, e.g., AntiReal1|(promise) = renege on, AntiReal2(attack) = beat back, IncepOper1|(fire) = open, FinOper2(controt) = get out of, etc. To avoid overburdening the exposition, we will not touch on socalled NON-STANDARD LF's. LF's play a crucial role in covering restricted lexical cooccurrence. In (11), we see a Magn(caution), i.e., seriously, in (12), there are three LF's: S0(warn) = warning,Oper1(warning) = issue, and Magn(warning) = stern. (For more about lexical functions see Mel'cuk 1982a.) A branch of a DSyntS is labeled with the name of a DEEP-SYNTACTIC RE­ LATION (=DSyntRel). A DSyntRel is one of the following nine binary relations: — 1, 2, . . ., 6 are actant, or predicational, relations, connecting a lexeme whose meaning corresponds to a functor, roughly a predicate, with its 1st, 2nd,. . ., 6th arguments, respectively; for instance: (13)

SELL

[who?] seller

[what?] [to whom?] [for how much?] merchan- buyer price dise

MEANING-TEXT MODELS

27

- ATTR is the attributive relation, which covers all kinds of modifiers and at­ tributes (in the broadest sense of the term): (14) a. TWO

ARGUMENTS

— COORD is a relation that accounts for all coordinate, or conjoined, construc­ tions : (15)

The number 2 on the arrow leaving a deep-syntactic node labeled with a pre­ position or a conjunction stands for '2nd place of the corresponding predicate'. Prepositions and conjunctions, being semantically two-place predicates, retain only their 2nd place at the DSynt-level: their DSynt-governor corresponds to their 1st semantic argument. — APPEND is an 'appenaancy' relation that subsumes all parentheticals, inter­ jections, direct addresses, etc.:

SURFACE SYNTAX OF ENGLISH

28

(16) a. CONSIDER imper ,

[Consider, for example, the category . . .] b. BY-GOLLY

[By golly, he might do it!]

[Look at me; Mart!]

MEANING-TEXT MODELS

29

It is presumed that these relations are sufficient for the description of any syn­ tactic construction of any language at the deep-syntactic level. In other words, the DSyntRel's are deemed to be UNIVERSAL. The DSyntRel's are strictly asemantic. A DSyntRel represents a family of syn­ tactic constructions belonging to the same general structural type, regardless of their semantic content. A meaning which is expressed in the given language only syntactically must be rendered in the DSyntS by a fictive lexeme (p. 25). E.g., to represent the Russian approximative-quantitative construction of the type kilometrov desjat 'about ten kilometers', where the postposing of the numeral desjat' 'ten' means 'approximately' (as opposed to desfaf kilometrov 'ten kilometers'), we will use, at the deep-syntactic level, the fictive lexeme «PRIBLIZITEL'NO» 'approximately': «PRIBLIZITEL'NO»

DESJAT'-

KILOMETRpl.

In fact, the universality of our DSyntRel's is guaranteed by the device of fictive lexemes in a syntactic capacity. Where the discriminatory power of the DSynt­ Rel's proves insufficient, fictive lexemes fill the gap. Let it be emphasized that there is no linear order of nodes within the DSyntS. Word order is taken to be a MEANS FOR ENCODING syntactic structure into speech strings and therefore it is banned from the syntactic structure as such. (This allows the MTM to avoid a number of problems connected with, e.g., dif­ ferent movement transformations which change an established order of consti­ tuents.) The D(eep), Synt(actic)-,Cornm(unicative)S(tructure) specifies the division of the sentence represented into topic and comment, old and new, etc. In our simpli­ fied examples the DSynt-CommS shows the first stratum of topic-comment divi­ sion only, in much the same manner as within the SemR. We will not discuss the differences between the DSynt-CommS and Sem-CommS any further. The D(eep) Synt(actic)-Anaph(oric)S(tructure) carries the information about coreferentiality; cf. the broken-line arrows in (11) and (12): e.g., the understood subject of avoid is the expectant mothers who have been cautioned, etc. The D(eep) Synt(actic)-Pros(odie)S(tructure) represents all of the meaningbearing prosodies that appear at this level: intonation contours, pauses, emphatic stresses, and the like, where these are not syntactically conditioned. In (11) and (12), it is the acronym NAP, standing for N(eutral) A(ssertive) P(rosody), that exemplifies the DSynt-ProsS.

30

SURFACE SYNTAX OF ENGLISH Surface-Syntactic Representation

To save space, only sentence (3) will be represented at the surface-syntactic level, since this level will be dealt with in more detail in the rest of the book: see Fig. 4.

NAP

Fig. 4 The SSyntR of the English sentence that corresponds to the DSyntR (11),i.e.,of(3).

MEANING-TEXT MODELS

31

The S(urface)-Synt(actic)R(epresentation) of a sentence consists of four struc­ tures corresponding to those of the DSyntR. The SSyntS is also a dependency tree, but its composition and labeling differ sharply from those of the DSyntS. A node of a SSyntS is labeled with an ACTUAL LEXEME of the language. There are four differences between the nodes of a SSyntS and the nodes of a DSyntS. First, in a SSyntS, all the lexemes of the sentence are represented, inclu­ ding the semantically empty ones. Second, all the idioms are expanded into ac­ tual surface trees. Third, the values of all the lexical functions are computed (on the basis of a special lexicon, see below) and spelled out in place of the LF's. Fourth, all pronominal replacements and deletions under lexical or referential identity are already carried out, so that a SSynt-node can be a zero or a pronoun. A lexeme labeling a node in a SSyntS is also subscripted (just as in a DSyntS) for all meaning-bearing grammemes. Thus the nodes of a SSyntS are reduced DMorphR's of the corresponding wordforms, i.e., the same formal objects as in a DSyntS. But the wordforms represented in a SSyntS are actual wordforms of the sentence, and not wordforms of generalized lexemes, which appear in a DSyntS. A branch of a SSyntS is labeled with the name of a SURFACE-SYNTACTIC RELATION(= SSyntRel). A SSyntRel belongs to a set of language-specific binary relations that obtain between the words of a sentence and describes a particular syntactic construc­ tion. The inventory of SSyntRel's for a language is established empirically, ac­ cording to the criteria discussed in Chapter II, § 2, item 4. In (17), the reader can see examples of English SSyntRel's; a tentative list of all SSyntRel's for English is found in § 4, Chapter II. As is the case with the DSyntS, the nodes of the SSyntS are not ordered linearly. This enables us to keep strictly apart two basically different 'orders' ob­ served in a sentence: the (non-linear) structural, or syntactic proper, hierarchy, on the one hand; and the linear ordering, which serves to express this hierarchy, on the other. The SSynt-CommS, the SSynt-AnaphS, and the SSynt-ProsS are analogous to their deep counterparts. A detailed discussion of important divergences can­ not be undertaken here, due to lack of space.

32

SURFACE SYNTAX OF ENGLISH

Deep-Morphological Representation Only one DMorphR will be offered — for sentence (3); see (18): (18)

The DMorphR of the sentence which corresponds to the SSyntR (17), i.e., of (3). The D(eep-)Morph(ological) R(epresentation) of a sentence consists of two structures: the deep-morphological structure and the deep-morphologico-prosodic structure. The Dfeep-JMorph(ological) S(tructure) is a string of D(eep-)Morph(ological) R(epresentations) of all the wordforms4 that compose the sentence. The DMorphR of a wordform W is the name of the lexeme to which W belongs subscripted for all its morphological values. Thus the DMorphR of W unambiguously specifies this particular W (up to wordform homonymy), while the DMorphS of the sentence unambiguously specifies its word order (for more about DMorphR's of wordforms see Chapter II, § 2, item 3). The D(eep-)Morph(ologico)-Pros( odie) S(tructure) indicates the pauses, intona­ tion contours, and the like. In (18), the vertical bars stand for pauses of different length, and the arrows represent the pitch contours. Representations at Other Levels Levels 5 through 7 — the Surface-Morphological [= Morphemic] Representa­ tion, the Deep-Phonetic [= Phonemic] Representation, and the Surface-Phonetic [= Phonetic proper] Representation — will be omitted here, since within the scope of the present book they are of minor interest. In a nutshell, the aim of all these representations, with different structures within each, is TO DESCRIBE ALL THE DIFFERENT ASPECTS OF AN AC­ TUAL UTTERANCE AS SEPARATELY AND AUTONOMOUSLY AS POS­ SIBLE. That is, a MTM does not aim at a homogeneous representation; on the

MEANING-TEXT MODELS

33

contrary, we try to keep apart, explicitly, all the phenomena that intuitively ap­ pear to differ.

NOTES 1. (To page 16.) To see how thousands of paraphrases for (3) and (4) can be produced, con­ sider the following. Sentence (3) can be broken into eight semantic fragments, each of which will be represented by one column in the chart below; each column contains some (nearly-)synonymous variants that express the fragment in question (the figure in boldface under a column represents the number of variants in it) : 1 FDA

3

2

caution warn forewarn counsel put on guard issue a warning advise

seriously earnestly sternly strongly

address a warning make public

1

4

4 to to to to

pregnant women expectant mothers mothers-to-be

women/ ladies

not to should not

7

10

7

6

►coffee

avoid shun refrain from abstain from

should

10

coffee (indulge in)

5

one of something

8 small simple

of life/ life's

cup of coffee

6

3

8

34

SURFACE SYNTAX OF ENGLISH

Any variant from one column combines with nearly any other variant from another column: (i) The FDA has addressed a stern caution to ladies to the effect that while expecting a baby they should abstain from coffee, something which belongs to life's small joys, (ii) The FDA made public a strong warning addressed to mothers-to-be: they should not indulge in consuming coffee, one of the simple pleasures of life; etc. This allows us to simply multiply the number of variants in each column: 1 x 4 x 10 x 7 x 10 x 6 x 3 x 8 = 403 200 True, some of the paraphrases will be sifted out by selectional restrictions and other con­ straints; however, a native speaker can easily think of new variants, so that it becomes evi­ dent that synonymy is really rich. 2. (To page 19.) For a review of the notions and terms from graph theory used in this book the reader may consult an elementary manual such as Preparata and Yeh 1973. 3. (To page 22.) Two clarifications seem in order here. First, by (strongly) governed preposition/conjunction we mean a preposition/conjuction which is predicted by the syntactically governing word: ask for, call on, approve of, refer to, part with, independent of, superior to, order that, etc. Second, words we call empty need not be devoid of semantic content in all possible con­ texts. A word may become empty in a particular context where it duplicates a part of the meaning of another word syntactically linked to it. 4. (To page 32.) A wordform is a minimal autonomous utterance possible in a given lan­ guage. Any wordform belongs to a lexeme (for lexeme see Note 3 on page 76), so a wordform can also be said to be a lexeme occurrence.

§3 General Design of the Meaning-Text Model As stated above, a MTM has the task of establishing correspondences between the semantic and the (surface-)phonetic representations of any given utterance through the five intermediate levels listed in § 2. Accordingly, the MTM consists of the following six basic components: 1. The Semantic Component, or semantics, for short. 2. The Deep-Syntactic Component, or deep syntax. 3. The Surface-Syntactic Component, or surface syntax. 4. The Deep-Morphological Component, or deep morphology. 5. The Surface-Morphological Component, or surface morphology. 6. The Deep-Phonetic Component, or phonemics. The Surface-Phonetic Component, or phonetics, which provides for the cor­ respondence between a surface-phonetic representation and actual acoustic

MEANING-TEXT MODELS

35

phenomena, falls outside the scope of the MTM in the strict sense. Each component of the MTM is a set of rules having the trivial form

x

.y|c,

where X stands for a fragment of utterance representation at level n,  stands for a fragment of utterance representation at level n +1, and  is a set of conditions (expressed by Boolean formulas) under which the correspondence X Y holds. The two-headed arrow must be interpreted as 'corresponds', not 'is transformed into'. Thus, when the transition from a meaning 'X' to a DSyntR Y is performed, 'X' itself is not changed: nothing happens to 'X' while Y is being constructed by semantic rules under the control of 'X'. The relation between a representation n and an 'adjacent' representation n + 1 is the same as that between the blueprint of a house and the house itself. The blueprint is by no means transformed into the house; but during construction, it is the blueprint that guides the workers. Another important property of the rules in the MTM is that they are logical­ ly unordered. All relevant information about the language is expressed explicit­ ly, i.e., by symbols within the rales rather than by the order of the latter. The philosophy behind this decision is that finding the best order of rule application in a specific situation goes far beyond the task of linguistics proper. The rules themselves are conceived of not as prescriptions, or instructions of an algorithm, but rather as permissions and prohibitions, or statements in a cal­ culus. Basically, each rule is a FILTER sifting out wrong correspondences. The idea of defining the structures in the MTM in terms of sets of rules representing well-formedness conditions was introduced into the Meaning-Text theory by L.N. Iordanskaja (1963, 1964) and is now firmly embedded in the research on MTM's. Now we will briefly characterize the first three components of the MTM, leaving aside the rest for the same reason as the three nearest-to-surface represen­ tations. Note that only the direction from meaning to text is considered. The Semantic Component of an MTM The semantic component establishes the correspondence between the SemR of an utterance and all the synonymous sequences of DSyntR's of the sentences that make up that utterance, cf. (6) on p. 18. To do this, it performs the fol­ lowing eight main operations: 1) It cuts the SemR into subnetworks such that each corresponds, in its semantic 'size', to a sentence. (In fact, the original SemR is not cut: simply, the seman-

36

SURFACE SYNTAX OF ENGLISH

tic component constructs alongside it another SemR, equivalent to the original one, but consisting of a sequence of smaller SemR's.) 2) It selects the corresponding lexemes by means of semantico-lexical rules of the types illustrated below: (1) 'attempt 1' WARN 1

(1) is a semantico-lexical rule in the strict sense; C1 and 2 stand for different sets of conditions that might determine the choice of the lexeme. Essentially, (1) represents two dictionary entries for two (roughly) synonymous verbs to warn 1 and to caution.

This is a semantico-idiomatic rule specifying two idioms for the meaning 'preg­ nant woman', namely, expectant mother and mother-to-be. (To see that these

MEANING-TEXT MODELS

37

two phrases are idioms it suffices to recall that an expectant mother or a motherto-be is not necessarily a mother.) This meaning can of course be expressed also by the free phrase pregnant woman, which is ensured by two separate straight­ forward semantico-lexical rules. (3)

This is a semantico-functional rule responsible for some lexical functions in the DSyntS. Cf. in (11) and (12), p. 23—24: seriously cautioned, stern warning. 3) It supplies meaning-bearing grammemes (i.e. morphological values of lexemes) by means of semantico-morphological rules like the following one:

Here, the meaning of the English present perfect is roughly rendered as 'the event X took place earlier than the moment of this speech but its results persist up to this moment'. 4) It forms a DSyntS (= a tree) out of the lexemes and meaning-bearing grammemes it has chosen. 5) It introduces the DSynt-AnaphS; that is, it indicates coreference, etc. for the lexical nodes that have appeared as a result of the duplication of certain se­ mantic nodes. 6) It computes the prosody of the sentence on the basis of semantico-prosodic rules, e.g.:

SURFACE SYNTAX OF ENGLISH

38

('The speaker asks the listener to communicate to him the truth value of the statement P' is rendered - along with other means, of course — by the specific interrogative prosodies.) As we see, all the meaning-bearing prosodic phenomena are represented in the SemR in exactly the same way as any other meanings, but in the DSyntR their representation is different since prosodies and words are so different. 7) It provides the DSynt-CommS (topic — comment, etc.) from the data con­ tained in the Sem-CommS. 8) For each DSyntR produced the semantic component constructs all the sy­ nonymous DSyntR's that can be exhaustively described in terms of lexical functions. (Other types of linguistic synonymy are accounted for by semantico-lexical rules that can group bundles of semantemes into lexemes in dif­ ferent ways.) This is achieved by means of a paraphrasing system that defines an algebra of correspondences on such DSyntR's where the DSyntS contains LF symbols. The paraphrasing system includes rules of two major classes: Lexical Paraphrasing Rules (about 60 of them, valid for any language) repre­ sent either semantic equivalences or semantic implications. Examples: Equivalences The set contains[W] the point M The point M belongs[Conv21(W)] to the set. He wamed[W] them He issued [Oper 1 (S 0 W)] a warning[S0(W)] to them. He followed[ReaÌ2(W)] her advice [ W] to enroll He enrolled on [Adv 1B Real 2 (w] her advice[W].

MEANING-TEXT MODELS

39

Implication (7)

PerfCaus(W)

Perflncep(W)

John started [PerfCau(run 11)] the motor The motor started [Perflncep(run 11)].

Syntactic Paraphrasing Rules indicate what restructuring of a DSyntS is needed when a particular lexical paraphrasing rule is applied. It is known (Gladkij and  1975: 67-68) that there are only four basic syntactic opera­ tions at the deep level: merger of two nodes, splitting of a node, transfer of a node to another governor, and renaming of a branch. Moreover, as stated above, only nine deep-syntactic relations are distinguished in DSyntS. Therefore, the set of elementary deep-tree processing rules is finite, and not very large. Any particular syntactic rule defined on DSyntS's can be represented in terms of these elementary rules. To operate lexical paraphrasing rules (6), the following syntactic rules are needed:

40

SURFACE SYNTAX OF ENGLISH

Generally speaking, a syntactic rule, because of its abstract character, may serve several different lexical rules. A formalism has been devised for describing the processing of unordered dependency trees — so-called A-grammar, see Gladkij and Mel'cuk 1969 and 1975. The Deep-Syntactic Component of an MTM The deep-syntactic component establishes the correspondence between the DSyntR of a sentence and all the alternative SSyntR's which correspond to it. To do this, it performs the following five main operations: 1) It computes the values of all lexical functions by means of rules like (9):

Such rules belong, of course, to the dictionary. 2) It expands the nodes of idioms into corresponding surface trees: (10)

FDAsg

Rules like (10) also belong to the dictionary. 3) It eliminates DSyntS nodes that occur in anaphoric relations and should not appear in the actual text; thus the occurrence of MOTHERpl under AVOID in (11), p. 23, is deleted by the following rule (a kind of EQUI):

MEANING-TEXT MODELS

41

(11)

4) It constructs the SSyntS by means of three types of correspondence rules exem­ plified below: Replacement of a DSyntRel oy a SSyntRel

42

SURFACE SYNTAX OF ENGLISH Replacement of a DSyntRel by a SSynt-node

The notation X(V, 3 [TO]) indicates a lexeme whose third DSynt-valence slot must be filled in the SSyntS by the infinitive marker TO ; the corresponding in­ formation is stored in the dictionary entry of X. Cf. CAUTION in (11) and (17) in § 2, pp..23 and 30. 5) It processes the other three structures of the SSyntR. A description of a small fragment of the DSynt-component for Russian is found in Mel'čuk 1974b: 237-259. The Surface-Syntactic Component of an MTM The surface-syntactic component of the MTM for Modern English is described at some length in § 1 of Chapter IV of this book. This allows us to limit ourselves, at this point, to some preliminary remarks of a highly sketchy nature. The SSynt-component establishes the correspondence between the SSyntR of a sentence and all the alternative DMorphR's that are realizations of it. It performs the following four main operations: 1) Morphologization of the SSyntS, i.e., determination of all the syntactically conditioned morphological values of all the words, such as the number and person of the verb. 2) Linearization of the SSyntS, i.e., determination of the actual word order of the sentence. 3) SSynt-ellipsis, i.e., various kinds of conjunction reductions and deletions that are prescribed by the language in question, e.g.: (15) Rolande will take the course, and Dick might take the course, too Rolande will take the course, and Dick might too. 4) Punctuation, i.e., determination, on the basis of the DSynt-ProsS as well as on the basis of the resulting SSyntS, of the correct prosody, which, in the case of printed texts, is rendered by punctuation.

MEANING-TEXT MODELS

43

The basic tool of morphologization and linearization of the SSyntS is the SYNTAGM, a type of SSynt-rule. A list of English syntagms is given in § 4 of Chapter IV. Besides syntagms, the surface-syntactic component uses at least four additio­ nal types of rules: — Word order patterns for elementary phrases; e.g., all those beautiful French magazines vs. *French those beautiful all magazines. — Global word order rules, which compute the best possible word order for the given SSyntS on the basis of various types of information: the syntactic properties of some words marked in the lexicon; the relative length of dif­ ferent parts of the sentence; topicalization, emphasis, and the like; possible ambiguities produced by specific word arrangements; etc. These rules try to minimize the value of a utility function that represents the 'penalties' assigned (by the linguist) to certain infelicitous arrangements. The rules do this by reshuffling the constituent phrases within the limits of what the syntagms al­ low. A fairly full description of such global word order rules for Russian, a language famous for its involved word order, was published in Mel'čuk 1965, 1967 and 1974b: 268-300. — Ellipsis rules. — Prosodic or punctuation rules. Other components of an MTM These will not be dealt with in this book/The interested reader is referred to Mel'čuk 1973 and 1978.

§4 Five Basic Principles of the Meaning-Text Approach For a deeper understanding of the general framework within which our spe­ cific surface-syntactic model of Modern English is formulated, it will be helpful to state explicitly the following five theoretical principles we abide by. 1. Description of specific languages rather than the quest for the best possible grammar. MTT concentrates on languages, not on grammars. The only legitimate question to be asked within this theory is "How do you express the meaning 'X' in the language L?" or "What does the expression X of the language L mean?" We consider it more important to write a good grammar (including the lexicon)

44

SURFACE SYNTAX OF ENGLISH

at least for a language than to discuss the properties of the best grammar of Language. 2. Establishing specific facts of a specific language rather than hunting for generalizations or discovering linguistic universais. To state a useful generaliza­ tion, one needs something to generalize from; therefore, well-organized collec­ tions of facts must perforce precede generalizations, lest the latter be scientifi­ cally void. Hence the particular interest of MTT in lexical entries, i.e., in the dic­ tionary, a grammar is considered to be nothing more than a set of generaliza­ tions over a good dictionary. MTT is thus a dictionary-oriented, rather than a grammar-oriented, theory. Not only every language, but every lexeme of a language is an entire world in itself. All this means that the primary object of MTT is particulars, not universais. 3. Cybernetic modeling of native linguistic intuition rather than a mathe­ matical study of formal systems representing language. The only goal of an MTT linguist is to exteriorize his or her own intuition about language (probably checked against other speakers' intuition). Language is viewed as a black box, not as a formal mathematical system (semi-Thue, or whatever). Hence no rigo­ rous proofs are possible within the MTT. You can show with good counter­ examples that a description is not sufficient but you can never prove — in the strict sense of the term prove — that a description is sufficient, let alone neces­ sary. One can of course invoke the arguments of simplicity, similarity, elegance, etc. but in so doing one does not PROVE anything. What an MTT-linguist really does is suggest practical names, useful labels, etc. to pass his own linguistic intuition on to others in the most unambiguous and graphic way. Based on all this, he builds a model, and as far as it functions correctly, he must be happy; but there are no arguments to prove the truth of his claims. 4. Formalisms serve the linguist rather than the other way around. The formality of ah representations is strongly emphasized in MTT — but no for­ malism is allowed to become a fetish. Nobody should be afraid of replacing 17 symbols by 439, if need be. Why should we be slaves to oversimplified and overly narrow mathematical formalisms? Formalisms are nothing more than necessary tools, and it is up to the linguist to create tools that he feels are neces­ sary. (Let the mathematicians accommodate OUR formalisms to their theories, if they so wish.) As far as the technical nature of MTM formalisms is concerned, we prefer dependency systems, with their emphasis on relations rather than constituency. One of the reasons for this is the wish to completely detach linear order from structural representation. Moreover, we claim that relation, rather

MEANING-TEXT MODELS

45

than grouping, is the pivot mechanism of natural languages, at all levels — except for the actual phonetic output. 5. Different formal languages for different levels of representation rather than the quest for a homogeneous representation. Language is a multidimensional, multifarious, highly heterogeneous formation, and there is no use in trying to squeeze it into the straitjacket of a single formalism. Therefore, in MTM a distinct formal language is devised to represent differently each of the different levels of natural languages; and at each level different formal languages are used for its different aspects and strata; see above, § 2 of this Chapter. There are, to be sure, further relevant points that could be discussed, but such a discussion could take us too far from our main topic: Modern English surface syntax. In lieu of a conclusion we will simply mention a review paper, Nichols 1979, where substantive parallels are drawn between the MTM approach and modern Western linguistics.

After these theoretical preliminaries we can proceed to our first goal: a des­ cription of Surface-Syntactic Representation in general and for English in parti­ cular. This will be done in Chapter II.

CHAPTER II Surface-Syntactic Representation for English Sentences

In this chapter, we concentrate on the SURFACE-SYNTACTIC STRUCTURE and some related notions that are indispensable for understanding our surfacesyntactic rules. The discussion will be carried out in four stages: § 1 — General Characterization of the Surface-Syntactic Representation. § 2 - Surface-Syntactic Structure. § 3 - Conditions for Grammatical Correctness of English Surface-Syntactic Structures. § 4 - Surface-Syntactic Relations in Modern English.

§1 General Characterization of the Surface-Syntactic Representation The S(urface)-Synt(actic) R(epresentation) of a given sentence, as indicated above (p. 31), is a set of four formal objects called STRUCTURES; symbolically, SSyntR = {SSyntS, SSynt-CommS, SSynt-ProsS, SSynt-AnaphS}, where : - SSyntS ist the SURFACE-SYNTACTIC STRUCTURE (of the sentence rep­ resented). It specifies the hierarchical relations between the wordforms of the sentence and is syntactic in the proper sense of the term. The SSyntS is called the MAIN, or BEARING, structure with respect to the three other components of the quadruple ; it is completed by them (in order to fully specify the sentence) in much the same way a phonemic string is completed by an intonation contour and other prosodie phenomena. The surface-syntactic structure is described in more detail below: § 2, pp. 48 ff. - SSynt-CommS is the SURFACE SYNTACTIC-COMMUNICATIVE STRUCTURE, dealing with the topic/comment division, the known (=given)/new contrast, emphasis, and the like. - SSynt-ProsS is the SURFACE SYNTACTIC-PROSODIC STRUCTURE; it shows all of the meaning-bearing prosody that appears at this level: the pro­ sodic opposition of an assertion vs. a question vs. an exclamation, prosodie

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

47

foregrounding or backgrounding, prosodic signals of the speaker's emotional attitude (irony, anxiety, doubt, conviction, . . .), etc. - SSynt-AnaphS is the SURFACE SYNTACTIC-ANAPHORIC STRUCTURE, dealing with various types of anaphora, in particular coreference and lexemic identity (the use of pronouns and the like). Let it be stressed that only the SSyntS will be considered in any detail in this book. This means that all observable syntactic phenomena linked to topicalization, prosody and anaphora are deliberately left out of our discussion; the rules given in Chapter IV are based on the surface-syntactic structure alone. (In a few instances reference is made to the SSynt-CommS, SSynt-ProsS or SSynt-AnaphS, but this is not done systematically and is strictly for illustrative purposes.) Suffice it to say that the SSynt-CommS must be shown by marking relevant portions of the surface-syntactic structure with symbols ■ (= theme, or topic) and (= rheme, or comment); the SSynt-ProsS presupposes a system of des­ criptive prosodic symbols (which does not yet exist in our formalism) ; and the SSynt-AnaphS is plotted by means of broken-line double-headed arrows con­ necting the anaphorically linked nodes of the SSyntS. As an illustration, we give here a SSyntR for sentence (1) — but with the three last-mentioned structures shown in a highly approximate way. (1)

I loved them until they loved me [D. Parker].

48

SURFACE SYNTAX OF ENGLISH

A solid scientific status is claimed in this book for the surface-syntactic struc­ ture only, From now on, we will be dealing solely with that structure and leave aside the rest of the SSyntR. After these preliminaries, we can proceed to the definition of the SSyntS.

§2 Surface-Syntactic Structure The S(urface-)Synt(actic) S(tructure) of a given sentence is a dependency tree such that: (i) there is a one-to-one correspondence between its nodes and the wordforms of the actual sentence (except in cases of ellipsis, where extra symbols — dummies marking the ellipsis — appear in the tree); (ii) each node is labeled with the reduced deep-morphological representation of the corresponding actual wordform [with all the syntactically-condi­ tioned inflectional characteristics removed], and each branch is labeled with the name of the corresponding surface-syntactic relation [obtaining between the two adjacent nodes]; (iii) certain nodes forming a full subtree may be marked as belonging to one syntactic group. At least the following five items in the definition call for an explanation: — Dependency trees. — Tree nodes and sentence wordforms. — Deep-morphological representation of a wordform. — Surface-syntactic relation. — Grouping in the SSyntS. We will take these in turn. 1. DEPENDENCY TREES. There are, to the best of our knowledge, two diametrically opposed ways of representing natural sentences at the syntactic level: phrase-structure (PS-) trees and dependency (D-) trees. Obviously, any combination of the two methods is possible (and some have in fact been suggested), with compromises found at dif­ ferent borderlines; but it is important to emphasize that there is no essentially different third possibility. We will compare both methods, explaining D-trees in

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

49

terms of their differences from PS-trees, which may be more familiar to our rea­ ders. (I) A PS-tree shows which items — wordforms or phrases — of the represented natural-language expression 'go together', or combine, with which other items to form tight units of a higher order. This approach concentrates on CONSTI­ TUENCY and is, therefore, rightly called the "(immediate) constituent method/ system". A D-tree is quite different in this respect: it shows rather which items are related to which other items, and in what way. This approach, in contrast to the PS-approach, concentrates on the RELATIONSHIPS between ultimate syntactic units, i.e. wordforms, and leaves undercharacterized the formation of higherorder syntactic units. (II) In a PS-tree, the syntactic class membership (= categorization) of the syn­ tactic units is specified as an integral part of the syntactic representation, while the relationships among the units are not stated explicitly. In a D-tree, on the contrary, the types of syntactic relations are easily specified but the syntactic class membership of the units involved (elementary as well as higher-order) is not shown immediately within the syntactic representation itself: for an elementary unit, i.e. a wordform, all such information is part of the dictionary entry for the corresponding lexeme, so that it remains behind the scenes; for a higher-order unit, such information is simply not needed in most cases. (In some rather ex­ ceptional cases, where syntactic information about a higher-order unit may be involved, it is readily derived on the basis of the information assigned to the unit's components.) A corollary of the above is that a D-tree contains TERMINAL nodes only, while most of the nodes in a PS-tree are NON-TERMINALS. (III) In a PS-tree, the terminal nodes must be linearly ordered. The order is not necessarily that of the actual wordforms in the sentence represented but some linear order is unavoidable. In a D-tree, there is NO linear order of the nodes at all, since the linear order of wordforms is considered a syntactic means (along, e.g., with inflections) and therefore it must be excluded from the SSyntS. Word order is used by natural language to encode the surface-syntactic structure (of a sentence), which — exactly because of that — should not be linearly ordered. This is an important aspect of our approach in general and of our SSyntS in particular: linear order of symbols is never allowed to carry information in syn­ tactic (and semantic) representations; everything must be expressed explicitly by means of symbols. We cannot pursue this subject any further. Let us mention, however, that such a view runs counter to most modern linguistic theories.

50

SURFACE SYNTAX OF ENGLISH

(One of the rare exceptions is Sanders 1972, where it is required that "all terminal semantic representations be free of ordering relations" and that "there be no reordering": p. 95.) A comprehensive discussion of the problem is found in Meisel and Pam 1979. (IV) A PS-tree does not specify the syntactic relations between the syntactic units, at least not in an explicit and straightforward manner. A D-tree, however, lays particular stress on specifying in detail the TYPE of any relation obtaining between any two syntactically related items. The labeling of all branches in a D-tree constitutes a very important feature of our D-approach. (See below for the inventory of surface-syntactic relations in Modern English: § 4, pp. 85 ff.) A schematic example would not be out of place here. Sentence (1) (1) A girl at whom no one made passes No longer resents wearing glasses [the two first lines of an American limerick, using the famous verse by D. Parker] can be assigned the PS- and D-structures shown on p. 51, (1') and (1"). Since there is no linear order of nodes in D-trees, the variants a and b, as well as all the other variants obtainable by linear permutations of the nodes, are COMPLETELY EQUIVALENT. From now on, we will not distinguish between different linear arrangements of nodes in D-trees, using indiscriminately any arrangement we feel is convenient for the reader. We will neither undertake a detailed comparison of PS-trees and D-trees nor elaborate on their (non-)equivalence and their respective merits and disadvan­ tages; instead, we refer the reader to (in particular) Paduceva 1964, Gaifman 1965, Gladkij 1966: 8-29, Beleckij 1967, Fitialov 1968, Robinson 1970a, b, and Hudson 1976, 1980a, b. Furthermore, we will not touch upon the promising attempts (Gladkij 1968, 1971, 1980-81) to devise a more flexible system (of syntactic representation) that embodies the best aspects of both formalisms and thus tries to circumvent their drawbacks. We simply outline the specific variant of the D-representation that we are using in this book. Perhaps some remarks about the general philosophy behind this syntactic approach would not be out of place. (People eager to get down to business as quickly as possible can skip this section and go straight to page 57.) This ap­ proach, developed in Mel'cuk 1974b: 208-219 and Mel'cuk 1979a, can be summarized as follows: 1) Any natural language uses in its sentences exactly two opposed types of lin­ guistic devices to transmit the information a sentence carries: lexical and nonlexical means.

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

51

SURFACE SYNTAX OF ENGLISH

52

The LEXICAL MEANS used in a sentence are the list of all wordforms (= lexeme occurrences) that constitute the sentence. The NON-LEXICAL MEANS used in a sentence are of three varieties: • linear order of the wordforms; • different prosodies (intonation contours, pauses, phrase and sentence stres­ ses, etc.); • inflectional markers on the wordforms. Both types of linguistic means (i.e. lexical and non-lexical) can be used in one of the following two ways: — Either in a SEMANTIC capacity, i.e., to convey some meaning immediately, which amounts to these means being DIRECTLY connected to a portion of the semantic representation (SemR) of the sentence. — Or in a SYNTACTIC capacity, i.e., to mark relationships between semantic units. Linguistic means so used are connected to the SemR only INDIRECTLY, through the syntactic structure. This may be summarized in the following table: Table 1 Semantic

Lexical means

Non-lexical means

Word order

Prosody

Inflection

Syntactic

'Full' words: sea, help, because (of), and, beau­ tiful, . . .

Function words: (strongly) governed prepositions and conjunctions, auxil­ iary verbs, . . .

Wordform arrangements marking communicative structure: topic vs. comment, known vs. new, . . .

Wordform arrangements marking specific constructions: N+N, Adj+N, Prep+N, Prosodies marking syntactic links, borders of constituents, . . .

Prosodies marking questions, exclamations, irony, menace, anxiety, . . ., or focus, emphasis, etc. | Inflections marking number Inflections marking in nouns, tense and aspect agreement in adjectives in verbs, . . . and verbs, government,. . . Linguistic Means and their Possible Usages

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

53

Non-lexical means in a syntactic capacity are called SYNTACTIC MEANS for short; syntactic means are enclosed within the box in the above table. There are no other syntactic means in natural languages: all possible links, or relationships, between wordforms in a sentence, which as a whole might be cal­ led syntactic information, are expressed either by word order, or by prosody, or by inflection, or by a combination of these. 2) The three above-mentioned varieties of syntactic means of natural languages share one important property: they are all highly ambiguous, interdepen­ dent and inexplicit. As a rule, they convey syntactic information in a very complicated and cumbersome way, since there is no one-to-one correspon­ dence between the syntactic means used and the syntactic relationships marked. ||We maintain that in creating a surface-syntactic formalism it is exactly this ambiguity and inexplicitness (induced by non-lexical means in their syntactic capacity) that must ||be banned from the surface-syntactic structure of a sentence.

The SURFACE-SYNTACTIC STRUCTURE (of a sentence) is, then, a con­ struct introduced by the analyst for describing how the sentence is built from "the syntactic viewpoint". We feel that such a construct must satisfy certain intuitively justifiable requirements, for instance, the following: We require that all the lexical means of a sentence be represented in its S(urface) Synt(actic) S(tructure) directly, i.e., by a list of all the wordforms. However, we also require that the non-lexical syntactic means of a sentence should not be represented in its SSyntS directly. Therefore, the SSyntS of a sentence must not contain any explicit indications as to the word order, the prosody, or the morphological characteristics in their syntactic capacity (such as the gender, number and case of agreeing adjectives; or the person and number of the main verb). The absence of ah these linguistic means from the SSyntS necessitates their replacement by a special formal device that is free of their major draw­ backs — ambiguity and inexplicitness. This device must indicate explicitly and unambiguously the INTERACTIONS, or MUTUAL INFLUENCE, among the wordforms, since all of these are ordered, intonated, and inflected EACH ONE IN RELATION TO, OR UNDER THE INFLUENCE OF, ANOTHER ('X stands to the right/left of Y'; 'X has a higher/lower contour than Y'; 'X has the same gender and number as Y'; etc.). Non-lexical means in a semantic capacity (word order phenomena used for topicalization or emphasis; prosody which marks questions; etc.) should be repre­ sented in separate and autonomous components of the S(urface) Synt(actic) R(epresentation): in the SSynt-CommS and in the SSynt-ProsS mentioned above, pp. 46-47.

54

SURFACE SYNTAX OF ENGLISH

3) The above requirement to remove all the syntactically-conditioned non-lexical means (those in the box, see the right bottom corner of the table on page 52) from the SSyntS leads to the conception of the SSyntS of a sentence as a pair consisting of: — the set of all the wordforms appearing in the sentence, or more precisely, of all the deep-morphological representations of wordforms, from which all syntactically-conditioned morphological characteristics are removed; — a set of n- relations defined over the first set. The more formal requirement of maximal simplicity and uniformity entails positing a set of BINARY relations. In principle, the reduction of n-ary relations (with arbitrary n) to a system of binary relations is always possible, and the question of whether such a reduction is motivated in the case of natural syntax will be left open. In our opinion, natural languages seem to lend themselves readily to syntactic description in terms of binary relations. Therefore, we pro­ pose using only binary relations for representing the SSyntS's of natural senten­ ces. This leads to the question of what kind of relations these should be. First, they must be DIRECTED, i.e., ANTISYMMETRIC: if X Y, then (Y X). This follows from the natural principle that the surface-syntactic re­ lations we introduce must preserve all of the information available [see Principle (13) on p. 73] and simultaneously from the existence of such current examples as Russ. učitel ' brata 'brother's teacher' vs. brat účitelja 'teacher's brother'. If, in our SSyntS, we eliminate the word order and grammatical cases (on the grounds that they are purely syntactic means) and do not introduce antisymmetric re­ lations, then vital items of information are lost, and an ambiguous surface-syn­ tactic representation, namely, UĆITEL'sg—BRATsg, is assigned to both un­ ambiguous phrases. A directed relation preserves unambiguity: UĆITEL'sg BRATsg = učitel ' brata as opposed to UĆITEL 'sg←BRATsg = brat učitelja. To take an English example: pairs like rule analysis vs. analysis rule, minority stu­ dent vs. student minority etc. require that we have RULE←ANALYSIS, MINORITY ←STUDENT etc. for the first phrase in such a pair, but RULE ANA­ LYSIS and MINORITY STUDENT for the second. Second, the relations in our syntactic description must be ANTIREFLEXIVE: ¬(X X), or¬XO, since no wordform can be linearly ordered or inflected under its own influence. (Moreover, antireflexivity of a binary relation auto­ matically follows from its antisymmetry.) Third, our syntactic relations must be ANTITRANSITIVE: if X Y and Y Z, then ¬(X Z), i.e. "The direct dependent of my direct dependent is never my

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

55

direct dependent". This follows from immediate linguistic intuition based on examples like [a] very interesting book, where very is obviously not directly re­ lated to book, or [to] read [an] interesting book, where no direct relationship is perceived between read and interesting, as well as from the principle of the uniqueness of the syntactic governor put forth below, see the next page. Interestingly enough, if we postulate COMPLETE ANTITRANSITIVITY of syntactic relations, i.e., if we accept that no dependent of a dependent of X can be a direct dependent of X (for any n, from X Y1,Y1 Y2 . . ., Yn-1 Yn it follows that ¬ (X Yn)), then both the antisymmetry and antireflexivity of syn­ tactic relations are entailed automatically: Kazincov 1978. Fourth, the syntactic relations must be distinguishable, or sorted, i.e., LA­ BELED: in Russian, for example, BRATsg UÇITEL 'sg may stand both for brat učitelja 'teacher's brother' and brat-uëitel "the-brother who-is-a-teacher'; similarly, KOMNATApl DESJAT' may represent komnaty desjat "rooms [number] 10', desjat'komnat '10 rooms', and komnat desjat' 'about 10 rooms'; etc. Therefore, to avoid this loss of information, i.e., to abide by the above-mentioned principle, it is necessary to introduce DIFFERENT TYPES of surface-syntactic relations. Thus, BRATsgr2 UÇITEL'sgwillwill represent brat učitelja, whileBRATsgr2UÇITEL'sg (with r\ f ri) will represent brat-ucitel'. An English example could be [to like somebody more] than he vs. [to like somebody more] than him: THAN r1HE and THAN r2HE, r1≠ r2. (We need not go info detail here since more will be said about the different types of SSyntRel's in Subsection 4: pp. 70ff.) Now let us turn to the formal nature of the SSyntS itself. We can conceive of a SSyntS as a directed GRAPH whose VERTICES, or NODES, are labeled with partial deep-morphological representations (of wordforms) and whose AR CS, or BRANCHES, depict antisymmetric, antireflexive and antitransitive binary re­ lations (cf. note 2 on page 34). This graph as a whole must fulfill the following three conditions: (a) It must be connected, since no component of a sentence can be fully de­ tached from the sentence without destroying its integrity. We need to know, at the very least, to what sentence a given item belongs, otherwise it cannot receive the proper linear arrangement, intonation, and/or inflection. (b) Every node in this graph can have only one governor, or dominating node: if X Y and Z≠X, then ¬(Z Y) (the principle of the uniqueness of the syn­ tactic governor). On the other hand, the number of syntactic dependencies stemming from a given node is not limited.

56

SURFACE SYNTAX OF ENGLISH

() The syntactic graph must contain exactly one node which has no governor at all (called ROOT, or TOP NODE):

The two last conditions — (b) and (c) — are difficult to substantiate on intui­ tive grounds, so let us accept them as postulates. 1 Note, however, that they seem natural, that they have been independently accepted in many different grammatical traditions, and that they bring about a drastic simplification of the formalism used. Every objection that we are aware of against the obligatory uniqueness of the syntactic governor or against the obligatory presence of a single root in the D-trees can be easily explained away: such objections are in­ variably based on the misleading confusion of SYNTACTIC DEPENDENCIES with SEMANTIC or MORPHOLOGICAL ones. Consider, e.g., the well-known case of a "double" dependency in Russian: (2) On kazalsja bol'nym 'He looked ill'

vs. Ona kazalas' bol'noj' 'She looked ill'

where the instrumental case of the adjective BOL'NOJ is determined by the copula verb but its gender (and number) depends on that of the grammatical subject. However, this double dependency is morphological rather than syntactic. True, the grammatical subject does control the gender-number inflection of the predicative adjective — but this fact by no means implies a direct (i.e., im­ mediate) syntactic relationship between them: a lexeme controlling the inflec­ tion of another lexeme does not imply more than morphological dependency. In (2), BOL'NOJ depends syntactically on the verb only. 2 It is not possible to discuss in detail here the three types of dependencies (i.e., the semantic, the syntactic and the morphological) and the precise nature of syntactic dependency. See, in this connection, Mel'cuk 1964b, Garde 1977, Nichols 1978 (giving special attention to cases of the type (2) above), and Mel'cuk 1981b. The three conditions stated above fully determine the form of the SSyntS, as we understand it. The SSyntS proves to be a ROOTED TREE (in the mathe­ matical sense), specifically a D-tree.

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

57

The use of this type of tree as a basic device for representing SSyntS's re­ ceives some support from the general conception of linguistic models. It seems natural to represent the Sem(antic) S(tructure) of a sentence by means of an arbitrary (multidimensional) connected graph; this graph is essentially a net­ work (nearly the most complex of all graphs) whose vertices are labeled with the symbols of semantic units and whose arcs show the relationships among them. On the other hand, the D(eep) Morph(ological) S(tructure) of a sentence (i.e., the string of all the complete deep-morphological representations of the wordforms that occur in the sentence) is a chain (in the mathematical sense), i.e., a unidimensional or linear graph—a string, which is the simplest of all graphs. Now, the SSyntS of a sentence bridges the gap between its SemS and its DMorphS and stands, so to speak, halfway between the two. It should, then, be the simplest of all non-linear graphs, and the simplest of non-linear graphs happens to be a tree, which is two-dimensional. (Note that the imposition of natural, i.e., general, constraints on a tree, in particular, prohibiting branching, converts it into a string.) This is a plausible explanation of why trees are — and have always been — so popular in syntactic description. 2. CORRESPONDENCE BETWEEN THE NODES OF A D-TREE AND THE WORDFORMS OF THE SENTENCE IT REPRESENTS. Requirement (i) in the definition of SSyntS, namely that there be a one-toone correspondence between the nodes of a SSyntS and the wordforms of the sentence represented, needs further clarification: it is correct as it stands, but the notion of wordform is itself not quite clear. Therefore, three apparent deviations from the said correspondence must be indicated: 1) All linearly indivisible and morphologically invariable PHRASEMES are represented in the SSyntS by only one node: of course, by all means, with respect to, no one, no longer, . . . [see the last two phrasemes in the D-structure on page 51]. 2) Any morphonological fusion of several wordforms into one is represented by several nodes, in accordance with the number of the wordforms fused: Fr./o/ au à + le, Germ. im in + dem, Eng./wil/ we'll we will, /wont/ won't will not, /hiz/ he's he has/is, etc. 3) There can be ZERO WORDFORMS in the sentence under analysis. We do not know of any zero wordforms in English, so we give two Russian examples:

SURFACE SYNTAX OF ENGLISH

58 (3)

a. Masa studentka, lit. 'Masha student' STUDENTKAsg [with a zero copula form]; b. Masu [očen'] uvažajut, lit. '[They] respect Masha [very much]'

UVAZAT' MAŠAsg [with a zero pronoun meaning roughly 'indefinite THEY', which is quite similar to Fr. on or Germ. man]. See Mel'čuk1979b. Except for these three cases plus the above-mentioned case of ellipsis, the correspondence between the wordforms of a sentence and the nodes of its SSyntS is straightforward. 3. DEEP-MORPHOLOGICAL REPRESENTATION OF A WORDFORM. The DMorphR of a wordform w is an abstract description that specifies w up to free morphological variation; thus, hoofs and hooves must have identical DMorphR's while brothers and brethren have different DMorphR's (since these two forms have different meanings). The DMorphR of w, written as DMorphR(w), has the form where L is the name of the corresponding lexeme (the lexeme to which w belongs), and X is the morphological characteristics of w, or the string of its grammemes — the values of all the morphological variables attached to this lexeme. (With lexemes that are morphologically invariable, such as adverbs, prepositions, etc., X must be empty.) The name of a lexeme is printed in CAPITALS and may be provided with a numerical index (to distinguish it from the other lexemes of the same vocable 3 ). Some examples: (4) a. DMorphR(am) b. DMorphR(sheep)

= BEind(icative), pres (ent), 1 sg = SHEEPsg/p1

. DMoxphR(distinguished) =

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

59

[The component "act/pass"' in the morphological characteristics of the participle reflects our analysis of the -ed and -en forms: we consider every such form to be homonymous between 'active past participle', which appears only in perfect structures with HAVE, and 'passive past participle'; for explanations see Chapter III, page 176.] d. DMorphR(shows) = e. DMorphR(yery)

= VERY

A tentative list of morphological values for Modern English is given in Chapter III, pp. 168-173. What is defined above is the FULL DMorphR of a wordform. If we remove from a DMorphR all the morphological values that are controlled by the syntac­ tic environment only, in particular, by agreement and government, we obtain the REDUCED DMorphR of the corresponding wordform. It is reduced DMorphR's that appear in the nodes of a SSyntS. For instance, the reduced DMorphR's for am and [it] shows are ind pres and S H O W i n d , respectively (with "1sg" dropped in the first case, and "3sg", in the second, because these are syntactical­ ly controlled). In English, there are six morphological categories used in a syntactic capacity (cf. § 4 of Chapter III): 1)—3) finitude (see p. 173, item 7), person and number in verbs; 4) grammatical case in personal pronouns (the nominative vs. the obli­ que); 5) number in pronominal adjectives like this / these; and 6) possessive form in nouns. These categories — except for finitude which has an intermediate status — are banned from the reduced DMorphR's of wordforms and thus do not appear in the SSyntS. All the other morphological categories of English — such as number in nouns, mood, tense and aspect in verbs, degree in adjectives and adverbs — are considered meaning-bearing and are therefore retained in the SSyntS, as well as finitude. However, it should be constantly kept in mind that no impenetrable barrier separates the two types of categories. On the one hand, syntactically conditioned categories may, in some special contexts, acquire semantic functions as well. Cf.: (5) vs.

a. Mary's and John's arriving on time b. Mary and John's arriving on time,

60

SURFACE SYNTAX OF ENGLISH

with (5a) meaning two separate arrivals while (5b) implies a joint arrival of the couple (the example is adapted from McCawley 1968: 161—163). In the general case, the possessive inflection -s is considered a morpheme in a purely syntactic capacity and is therefore banned from the SSyntS. But in (5) its presence/ab­ sence is directly connected with meaning. More examples of this kind can be quoted from Russian: (6) a. Daj mne xleb [accusative]! 'Give me the bread!' vs. Daj mne xleba [partitive]! 'Give me some bread!' b. rajonnaja [sg] i oblastnaja [sg] gazety 'the district newspaper and the regional newspaper' vs. rajonnye [p1] i oblastnye [pl] gazety 'district and regional newspapers'; . Ja lëg [mase] 'I [= male] lay down' vs. Ja legla [fem] 'I [= female] lay down'. Here, the grammatical case of a noun, the number of adjectives, and the gender of a verb, which, as a general rule, are purely syntactic markers in Russian, turn out to be semantically relevant. In all such cases, which must be precisely cha­ racterized, the corresponding morphological values should somehow be reflected in the SSyntS — despite their presumably non-semantic character. This can be done either directly, by subscripting the grammemes in question to the lexeme names, or indirectly, by using some additional formal device in our SSyntS (see below, item 5 in § 2, Chapter II: p. 72ff.). A substantive discussion of the problem is found in Iomdin 1979: 34-37. On the other hand, meaning-bearing categories may, in some contexts, become redundant and be syntactically controlled. Such is the case of nominal number in phrases with numerals (in ten pencils the plural of pencils is automatic), or of verbal mood after some conjunctions (I insist that he be expelled). All such cases are simply ignored in our description; that is, if an English morphological value is normally a meaning-bearing one, it is always represented in the SSyntS, regardless of its actual status in a specific context. (To justify this solution, we might point out that the duplication of semantic information by various lin­ guistic devices in an utterance is highly typical of natural language.) Now let us turn to the presentation of all the necessary syntactic properties of English lexemes. This is done by means of a dictionary which plays a vital role within the framework of the Meaning-Text model. In this dictionary, each lexeme is assigned a full-fledged dictionary entry which consists of three major zones:

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

61

— Description of the physical aspect of the lexeme (its spelling and pronuncia­ tion), which need not interest us here. That is the SIGNIFIANT of the lexeme. — Description of its meaning, or its Sem(antic) R(epresentation), shown by inverted commas: 'X'. That is the SIGNIFIE of the lexeme. Generally spea­ king, our surface-syntax rules do not deal with the SemR's of lexemes. How­ ever, in some rather exceptional cases, the SemR of a lexeme may be involved, e.g., in the condition part of a rule. Then the necessary semantic feature is mentioned. — Description of all the combinatorial properties of the lexeme: its syntactic class membership (roughly, its part of speech), its government, its restricted lexical cooccurrence (= lexical functions), etc. This zone is called SYNTAC­ TICS', it is the most important part of a dictionary entry from the viewpoint of surface syntax since it includes all of the information indispensable for the successful operation of syntactic rules. The syntactics of a lexeme is shown, when necessary, as a right subscript in parentheses to the name of the lexeme (preceding the morphological characteristics); cf.: (7)

BE(Verb, copula,...)ind, pres,

1sg

[=am];

SHOW(Verb, 2 [N], 3 [to N],...)ind, pres, 3sg [ = shows];

NEVER(Adv, neg, pron,...). [= never]. (On a particular type of syntactics components in English — syntactic featu­ res — see below, pp. 470ff. ; for more details about the three components of a linguistic sign: signifiant (signans, signifier), signifié (signatum, signified) and syntactics — see Mel'cuk 1982b: 24—28.) Since the syntactics is only one of the three components of a lexeme, and in the DMorphR of a wordform the lexeme is represented by its name as a whole, the syntactics should not and does not explicitly appear in the DMorphR's of wordforms. Nevertheless, the surface-syntactic rules can refer to the wordform syntactics: the latter is 'everpresent' being unambiguously specified for each lexeme in the dictionary. 4. SURFACE-SYNTACTIC RELATION (SSyntRel). A SSyntRel is a binary relation connecting two lexical nodes (roughly, two wordforms) of a sentence; it is antireflexive, antisymmetric (= directed), and antitransitive. Linguistically, it represents a surface-syntactic role, or function such as 'be the (grammatical) subject of, 'be a modifier of, 'be a circumstantial

62

SURFACE SYNTAX OF ENGLISH

of', etc. Thus the notion of a SSyntRel is closely related to the familiar notion of an 'element ‹part› of the sentence', the keystone of traditional syntax. There are, however, two important differences: First, the traditional notion focuses on the DEPENDENT TERM of a syntactic relation, while our approach puts the stress on the RELATION itself. Second, we use much finer distinctions between surface-syntax roles than is traditionally done. While traditional grammar operates with about dozen syntac­ tic roles, we need between 30 and 50 different SSyntRel's. As has always been acknowledged (cf., for example, Bazell 1950: 9), SSyntRel's must be strictly distinguished from both SEMANTIC and MORPHOLOGICAL (= formal) relations. On the one hand, a given SSyntRel can correspond to several semantic relations: the predicative SSvntRel in English implements the semantic relations 'agent — action' (A lady dispatches a postcard), 'undergoer — action (Maude vas examined by a doctor), 'object - process of its modification' (Work expands so as to fill the time available for its completion), 'subject — state' (She njoyed it), 'time - event' (Last year saw an increase in the interest in semantics), and a great deal of others. On the other hand, a given SSyntRel can correspond to several morphological relations: the predicative SSyntRel in Russian can be implemented in a number of ways, e.g., 1) the noun phrase is in the nominative and the main verb agrees with it in person and num­ ber if it is in the present, or in gender and number if it is in the past ; 2) the noun phrase is in the genitive and the verb is invariably in the 3sg form of the present tense or in the neuter singular form of the past ; 3) the grammatical subject is an infinitive and the verb is in the 3sg or neuter singular form; etc. We can find a SSyntRel that expresses only one semantic relation but whose morphological implementation is very complex: the numerical SSyntRel in Russian (the seman­ tic relation is 'number — objects counted' but the rules for the morphological marking of the corresponding lexemes are extremely involved). We can also have the opposite case: the modificative SSyntRel in English is morphologically straightforward (the adjective is placed before the noun, with no inflection at all) although its semantical aspect is very complicated ('property - object': hot ← water-, 'agent - action': the French ← attack; 'producer - product': American ← machinery ; 'relation - its term' : equal ←numbers ;etc.) In short, surface-syntactic relations are introduced by the analyst as specially designed INTERMEDIATE items, in order to simplify the description of the complex correspondences between 'genuine' semantic relations and the 'genuine'

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

63

morphological relations which encode them. SSyntRel's must be postulated for a given language in such a way as to ensure an easy formulation of certain laws, or regularities, observed at exactly this level of representation. For instance, many interesting things can be said about the dependent term of the predicative SSyntRel in the given language, i.e., about the grammatical subject, QUITE INDEPENDENTLY of its semantic functions and/or its morphological markings: all grammatical subjects in a language, whatever their semantic interpretation and morphological form, have some properties in common that no other item has. It is such properties that a SSyntRel must capture. A given SSyntRel, then, represents a correspondence between a bundle of alternative semantic relations and a bundle of alternative formal means such that this correspondence provides for interesting and useful generalization. (A SSyntRel can be readily compared to a grammatical meaning, or a grammeme, such as 'present', 'past', 'singular', 'plural', etc.: each such meaning is no more than a handy label for the correspond­ ence between a bundle of alternative 'genuine' semantic units and a bundle of alternative morphological means, such as affixes or alternations.4) It follows from the above that a SSyntRel is a construct, an abstraction that is fairly removed from both semantic and acoustic linguistic realities. This fact makes it unavoidable, when distinguishing and defining SSyntRel's for a given language, to have recourse to temporary, hypothetical and even arbitrary solu­ tions which cannot always be properly justified. Hence the obvious non-unique­ ness of any inventory of SSyntRel's proposed so far. This statement applies as well to our own inventory of SSyntRel's for English that we use in this book. Its adequacy can be evaluated and judged only within a full-fledged Meaning-Text model of English. Until at least one such model is operational, our inventory of SSyntRel's cannot claim more than the status of a working hypothesis. Tentative inventories of SSyntRel's have been compiled for various languages; it may be helpful to indicate here the relevant publications: - for Russian: Mel čuk 1963: 490-493, Mel'cuk 1964a: 20-24, Hays 1964a, Klimonov et al. 1969: 71-77, Gladkij and Mel'čuk 1974: 9-11,Mel'čuk 1974a: 221-234, Apresjan, Iomdin and Percov 1978: 265-269; - for French: Kulagina 1970: 40-44, Percov and Polovko 1979: 8-13; - for English: Vasilevskij et al 1972: 17-18, Mel'cuk and Percov 1973b [the list basically used in this book, although with many modifications]; - for German: Kunze 1965, 1972: 23-24; - for Somali : Zolkovskij 1971: 12-13.

SURFACE SYNTAX OF ENGLISH

64

To complete this section, we must discuss the following four points: sub­ ordination and coordination with respect to the notion of a SSyntRel; depen­ dency on a phrase vs. on its head; the language-specific nature of SSyntRel's; and the criteria for establishing an inventory of SSyntRel's. 1) Surface-syntactic relations as proposed cover all kinds of syntactic depen­ dencies, i.e. subordination as well as coordination. However, there exists in theoretical syntax a clear-cut tendency to view coordination as something radically different from dependency. It is believed by many that in coordina­ tion no hierarchy can be found, since John and Mary is basically identical to Mary and John ; therefore, two coordinated, or conjoined, items are on equal syntactic footing, with no dependency relation between them. Conse­ quently, the most common view has it that a coordinate construction has no head at all. Some add reluctantly: 'If I were forced at the point of a gun to pick a single item as the head of a coordinate construction, I would make it the conjunction — though I would strongly prefer to say that there is no head'. But we maintain that a coordinate construction must be represented in terms of SSyntRel's as follows: (8) where "coordin" and "coord-conjunct" stand for the coordinative and coor­ dinate-conjunctional SSyntRel's (see § 4 of this chapter, especially pages 153ff. and 127ff.), while AND represents any coordinate conjunction. That is, we make the first (= leftmost) conjunct the head of the whole construction, with the coordinate conjunction, if present, subordinated to the preceding item and subordinating the following one. To argue in favor of representation (8), it suffices to remind the reader that our dependencies are purely surface-syntactic, i.e. they have nothing to do with semantics. Yet, the idea of the symmetry of coordination ("No head in a coor­ dinate construction") and the pivotal role of the conjunction comes clearly from the semantic level — or even from a still deeper, logical representation. It is only logical conjunction (or disjunction) that is really symmetrical (X  Y = Y  X; X V Y = Y V X) and which, being a functor, dominates its arguments, i.e. the conjuncts (or the disjuncts). At the level of SSynt-structure the state of affairs is quite different. In fact:

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

65

(i) The linguistic (in our case, English) conjunction is by no means symmetrical: He pulled down his pants and took a shit contrasts drastically with He took a shit and pulled down his pants (the example used by the late Boris Deloné [= de Launay], professor of mathematics at Moscow University, to show that the group of operations is not commutative). Further relevant examples can be cited: (9) a. I met Dick after dinner and persuaded him to go skiing ≠ I persuaded Dick to go skiing and met him after dinner. b. C.L. Perkins and some other specialists. . . vs. *Some other specialists and C.L. Perkins . . .  Hands up, or I'll shoot! ≠ *I'll shoot, or hands up! d. not only an artist but a master  not only a master but an artist; e. once and for all vs. *for all and once. (In Malkiel 1959 and Cooper and Ross 1975 we find many conjoined phrases which are not, or are hardly, reversible.) (ii) A conjoined phrase can be used only in the same surface-syntactic roles as its conjuncts, never in the role of the conjunction. According to our under­ standing of SSynt-dependency, this indicates unmistakably that the head of a conjoined phrase is one of the conjuncts (cf. the criterion of passive valence: B.I, page 69). On the other hand, AND + Y is a constituent, while X + AND is not, and in the constituent AND + Y it is the conjunction that determines, to a greater degree, its passive syntactic valence (no matter what word-class the item Y belongs to, the phrase AND + Y must enter into a coordination relation, i.e., must be used within a conjoined phrase). This makes the conjunction the head of the phrase AND + Y and leaves us with representations of the type (8) as the only possibility for all kinds of con­ joined phrases. To sum up, we claim that SSyntRel's are equally fit for describing both sub­ ordination and coordination, so that in this respect they are sufficient as a means of syntactic representation.5 2) In terms of SSyntRel's the dependence of an item on a multilexemic phrase can be expressed in a natural way only as a dependence on its head (since SSyntRel's obtain only between individual wordforms). 6 However, in most

66

SURFACE SYNTAX OF ENGLISH cases that we know of in human languages dependence on a phrase is not contrasted semantically with dependence on its head. Because of this, the above-mentioned property of dependency formalism is actually an asset: it allows us to avoid the parasitic structural homonymy induced by the phrasestructure representation and to circumvent the controversy over whether, for example, an adverbial is subordinated to the main verb, to the whole verb phrase, or to the whole sentence. Let us consider the following phrase : (10) the gentle curve of her thighs [H. Miller].

It can be represented by two different PS-structures: (10') a. ((the (gentle curve)) (of (her thighs))); b. (the (gentle (curve (of (her thighs))))). Under the PS-approach, the question is whether the gentle modifies the phrase curve of her thighs or the noun curve only (in which case it is of her thighs that modifies the whole phrase the gentle curve). However, with SSyntRel's we can get only one representation: and the problem disappears. Our opinion is that this is good since there is no semantic contrast between (10 ' a) and (10 ' b). Similarly, using SSyntRel's only we cannot distinguish in a natural way between the subordination of a whole phrase treated as a closely-knit constitu­ ent and that of the head of the phrase. Thus for (11): (11) [The test was]





to write elegantly

we cannot even discuss — if we limit ourselves to the language of SSyntRel's — whether the preposition in subordinates the whole phrase an ability to write ele­ gantly or the noun ability only, which is, in its turn, modified by the rest. We can get nothing but a unique representation (as in (11)). This, too, is an advantage. However, some very special constructions do exist where the distinction between subordination to a phrase and subordination to an element of the phrase becomes linguistically relevant from the viewpoint of SSyntRel's. This necessitates using the device of GROUPING as a supplementary means of SSynt-representation (see below, page 72ff.).

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

67

3) Surface-syntactic relations are not and cannot be universal. Any particular SSyntRel represents a correspondence between a set of specific formal means of a language L and a set of semantic relations which is quite specific for L as well. Therefore, a SSyntRel is in principle 'language-specific', or idio­ syncratic, i.e., characteristic of its language only. The SSyntRel's in this respect share the nature of grammemes (= elements of grammatical categories) or phonemes, which also cannot be universal. Take, for example, the dative case. A dative exists in Russian, German, Georgian and a number of other languages. But in spite of the common name, these datives are not commensurable, much less identical. The German dative is opposed to the nominative, the accusative, and the genitive; in Russian, the dative contrasts with the same three cases plus the instrumental and the prepositional; while in Georgian it contrasts with the three cases plus the instrumental, the ergative and the circumstantial. The syntactic roles marked by the datives and the meaning content thereof are quite different in these languages. (In Georgian, for instance, the dative marks the grammatical subject with transitive verbs in perfect tenses or the experience/perceiver with verbs of feeling and perception, which is ab­ solutely impossible in German or Russian, etc.) There is, to be sure, a common function for all these different datives: namely, to mark the complement of ad­ dressee (as in give to smb); this function is responsible for their common name. Nevertheless, a common name does not entail identity. It is clear that a gram­ matical case in a given language is quite specific: it is determined, fully and ex­ clusively, by the case SYSTEM to which it belongs and therefore it cannot have, as a general rule, an exact equivalent in a different language. Moreover, a lan­ guage may possess a case for which there is no corresponding case at all in a second language: e.g., Russian has an instrumental while German has none. Similarly, the English phoneme /t/ does not stand in one-to-one correspond­ ence to Fr./t/, Sp./t/ or Chinese /t/: all these phonemes belong to different phonemic systems and are therefore different. (The physical similarity of some of their allophones, though, allows us to use the same designation — /t/, which refers to such distinctive features of all /t/'s as [-voiced], [+dental], [-nasal], etc.) A phoneme can also be present in one language without any equivalent whatsoever in another: /n/ exists in English and German but not in Russian or French, while, e.g., nasal vowels, so typical of French, are not found in English or Russian, etc.

68

SURFACE SYNTAX OF ENGLISH

In much the same way, any SSyntRel of any language is determined by the whole SYSTEM of SSyntRel's to which it belongs - and therefore it cannot be found in a different language. The common names, while being of some practical significance, should not mislead us. Thus the predicative SSyntRel's in English, French, Russian or Georgian are quite different with respect to their morpholo­ gical implementation as well as their semantic interpretation. (However, there is a kind of intersection that justifies the common name: all these predicative SSyntRel's subordinate to the main verb the noun phrase which is syntactically most priviledged in the languages in question.) A SSyntRel may appear in one language with no equivalent or even rough parallel in another: the approximatequantitative SSyntRel (kilometrov dejat' 'about 10 kilometers') is known in Russian but not in other European languages, while the proleptic SSyntRel (Cette femme, on la connaît très bien 'As for this woman, everybody knows her pretty well') is typical of French (or Japanese, for that matter) but absent from Russian and English (Apresjan, Iomdin and Percov 1978: 259-2Ó0). Therefore, an inventory of SSyntRel's must be established individual­ ly for every language: this is what we have done for English. 4) When working on a language and trying to detect and list its SSyntRel's, certain criteria may be used as "crutches" for the analyst's intuition. There are three groups of criteria (see Mel'cuk 1981b): A. Criteria of connectedness, i.e., criteria which provide for establishing the PRESENCE of a SSyntRel between two given wordforms of a sentence. B. Criteria of direction, i.e., criteria which provide for establishing the DIREC­ TION of the SSyntRel obtaining between two syntactically linked wordforms. In other words, these criteria allow us to answer whether w1 de­ pends on w2 or vice versa. C. Criteria of SSyntRel type, i.e., criteria which provide for establishing the TYPE of the SSyntRel subordinating w1 to W2- These criteria allow us to say whether in two structures, w1- w2 and W3 W4, the SSyntRel's r1 and r2 are identical or not. We give here all the criteria we know at present, with no substantiation and almost no explanation. The reader who is really interested is referred to Mel 'cuk 1981b, where these criteria are dealt with at some length.

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH A.

69

Criteria of connectedness. A.I (word order). (A.I) If wordforms w1 and W2 are (directly) connected by a SSyntRel, then the linear position of one of them cannot be determined without reference to the other.

A.II (prosody). (A.II) If wordforms w1 and W2 are (directly) connected by a SSyntRel, then two prosodic possibilities exist: - either w1 and w2 can form a prosodie unity; - or one of the two wordforms, say, w1, can form a prosodie unity with an autonomous utterance consisting of a string φ of wordforms which are syntactically connected, such that in φ, the wordform w2 is the top node.

The second condition of Criterion A.II covers the dependency between a full word and a function word [a preposition, a conjunction, etc.], the latter intro­ ducing the string φ - some other full words — and being the top node in . In the song of canaries, the wordforms song and of do not form a prosodie unity; but song does form a prosodie unity with of canaries [= φ], where of is the top node, and therefore Criterion A.II allows for a SSyntRel between song and of: theB.

Criteria of direction. B.I (imposition of passive surface-syntactic valence). (B.I)

Let w1 and W2 be two wordforms connected by a SSyntRel. The head of the phrase w1–w2 is the wordform that determines the passive surfacesyntactic valence of the phrase to a greater degree than the other wordform.

B.I means, roughly speaking, that the ability of a phrase to be subordinated to an item or to remain independent (= to be the top node of the whole sen­ tence) is essentially conditioned by the head of the phrase. For instance, the surface-syntactic behavior of a prepositional phrase in English is determined N. almost completely by the preposition; therefore, we posit Prep B.II (point of morphological contact). (B.II) Let w1 and w2 be two wordforms of the same word class connected by a SSyntRel. The head of the phrase w1–w2 is the wordform whose morpho-

SURFACE SYNTAX OF ENGLISH

70

logical ties with the external context of the phrase are stronger than those of the other wordform.

For instance, in Russian we have Jubka-štany, sštaja Mašej, vsem ponravilas ' 'Everybody liked the divided skirt [lit. 'skirt-pants'] sewed by Masha', where the participle sSitaja and the verb ponravilas ' agree with jubka [feminine singular], and not with itany [plural]. Therefore, in the Russian appositive construction of two nouns in the same grammatical case Criterion B.II singles out as the head the first one: jubka→štany. B.II is valid for languages with a richly developed morphology (not for English) and can be applied only in cases where B.I fails, e.g., if w1 and w2 belong to the same word class and have ipso facto the same passive valence. In other words, B.II is subordinated to B.I. 

Criteria of the SSyntRel type (cf. Mel'čuk 1979: 99-136). C.I (semantic contrast).

Let P1 and P2 be two binary phrases composed of the same lexemes X and Y, which are provided with the same meaning-bearing morphological characteristics and connected by surface-syntactic dependencies in the same direction. The DMorphR's of P1 and P2 are different, the difference being in purely syntactic means. Moreover, P1 and P2 can appear in the same surface-syntactic role;then the language in question possesses pairs of u t t e r a n c e s s u c h that contains P1, :contains P2 and there are no other formal differences between 1 2i, E iand E .Pairs of utterances are called COMPRISING for P1 and P 2 . (C.I)

Two different SSyntRel's r1 and r2 are used to represent P1 and P2 if and only if, among all the pairs omprising for P1 and P2 there exists at least one pair such that E 1 and E 2 are semantically dif­ ferent: ' E 1 ≠ ' E 2 ' .

Take, for instance, P1 = than he, P2 = than him; E1 = I like you more than E 2 = I like you more than him. In accordance with C.I, r1 ≠ r2.

-he,

 (reciprocal substitutability of subtrees). Let X and Y be wordform types, and ZΔ and WΔ complete terminal subtrees (having as their top nodes the wordforms of the types Z and W). A SSyntRel r will be said to posses the Kunze property if and only if, for any pair of correct surface-syntactic structures that contain. and respectively, the replacement o f ( a n d vice versa) does not affect the correctness of these structures.

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

71

(II) Any SSyntRel used in the description of SSyntS's of a given language must possess the Kunze property.

If, for a SSyntRel r this is not the case, r must be split into two (or more) SSyntRel's. Thus, in English we cannot use the same SSyntRel riage and promising =promising; the phrase a promising← method has a correct SSyntS; but the replacement of promising by a leads to an incorrect SSyntS: *a a method. The­ refore, two different SSyntRel's are needed: a that a promising promising. . . method is considered to be a correct SSyntS since several adjectives are possible with the same noun, but not several deter­ miners.) C.III (repeatability of SSyntRel's stemming from the same node). If we can have a configuration of the form in a correct SSyntS, the SSyntRel r is called REPEATABLE. (C.III)

Any SSyntRel used in the description of SSyntS's of a given language must be either non-repeatable or repeatable with any dependent elements.

If, for a SSyntRel r this is not the case, r must be split into two (or more) SSyntRel's. Thus in English we cannot use the same r to describe an attributive complement and a floating quantifier: (12) [My friends, who had gone to Canada three years before,]

since this SSyntRel r would prove repeatable when at least one of its dependents is a pronominal adjective but non-repeatable otherwise: (120 Therefore, we use the subjective-copredicative SSyntRel for return rich men or return heroes and the pronominal-subjective-copredicative SSyntRel for return-all (see § 4 of this chapter, pp. 102—105). Criteria A.I—II, B.I—II and C.I—III must be observed when establishing and elaborating the set of SSyntRel's for the surface-syntactic description of a lan­ guage. We tried our best to comply with them when working on the inventory of

72

SURFACE SYNTAX OF ENGLISH

SSyntRel's for English. However, we have not made an over-all check of our SSyntRePs against these criteria, so we cannot be sure of the complete consis­ tency of our inventory. Even worse: we admit, for the time being, SSyntRePs that do not meet some of the criteria A—C; e.g., we have SSyntRePs that are restrictedly repeatable. We do so since we do not yet know whether we should modify our criteria and leave the "disturbing" SSyntRePs on the list or follow the criteria and as a result strike all such SSyntRePs out (split them or replace them). Notwithstanding, we propose our inventory of English surface-syntactic relations on the assumption that a defective inventory of SSyntRePs is better than no inventory at all. 5. GROUPING IN DEPENDENCY STRUCTURES. As was indicated above (p. 66), the following situation is possible (although by no means frequent) in natural languages. Consider a phrase X→Y. Theoretical­ ly, two complications may arise when D-trees are used to represent a sentence including this phrase. First, suppose that either this phrase as a whole is modified by an element Â, or only its head, i.e. X, is modified by A "; the difference between the A -modifica­ tion and the A "-modification, shown on the surface by some purely syntactic means, corresponds to a semantic difference:

Since the formal difference between A ' and A" is purely syntactic, it cannot be retained as such in the SSyntS; then for both A 'X  Y and A "X  Y we in­ variably get the SSynt-representation A←X Y, which is ambiguous. Second, suppose that this phrase modifies an element A either as a whole or as two separate items and that the difference, shown in the phrase by a purely syn­ tactic means, also expresses a semantic contrast:

For the same reason as above, the only correct D-representation of both struc­ tures will be with the difference lost and the ambiguity increased. It is obvious, on the one hand, that our syntactic formalism should be able to handle both of the above complications. On the other hand, we should abide by the following methodological principle:

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

73

(13) " A linguistic representation of level n - 1 must be such as to guarantee that no expression of level n will become more ambiguous at level n - \ than it is at level n.

In other words, no information present at one level should be lost at a deeper level. And notice, however, that in constructions like

that is exactly what happens. Let us illustrate these two theoretical cases. First, in Russian, we have both (14a) and (14b): (14) a. oborvannyj starik i rebënok 'an old man in rags and a child'; b. oborvannye starik i rebenok 'an old man and a child, both in rags' [in (14a), the adjective is in the singular and agrees with starik while in (14b) it is in the plural and thus agrees with the whole conjoined phrase ; example (14) is very similar to (6b) on.page 60]. If the inflection of the adjective is not represented in the SSyntS, then the semantically contrasting (14a) and (14b) are both assigned an identical D-tree:

which means a loss of important surface information. So if we do not want the agreement grammemes of the adjectives to appear in the SSyntS, we must add something to our dependency formalism. Second, compare (15a) and (15b): (15) a. Mary's and John's arriving on time; b. Mary and John's arriving on time [repeating example (5) from page 59]. Once again, if we don't represent the possessive marker in the SSyntS because of its purely syntactic status, then the two expressions must have identical re­ presentations at the surface-syntactic level:

This is another loss of important surface information.

74

SURFACE SYNTAX OF ENGLISH

Both problematic cases are related to conjunction reduction. We do not know of any other type of structure where the SSyntRel formalism proves insuf­ ficient. 7 In view of facts like (14) and (15) and in order to comply with Principle (13), we suggest the following measure: we will indicate groupings (in D-trees) where they are relevant, i.e., where the joint interpretation is imposed by observable surface devices; otherwise (i.e., if the grouping is not indicated), the disjunct interpretation will be accepted. Thus for (14a) and (14b) the SSyntS's will appear as

respectively, the big bracket indicating the grouping. Similarly, for (15a) and (15b) we will write:

All this amounts to saying that we suggest supplementing the language of SSyntRel's by something resembling a 'constituency' device, allowed, however, in a strictly controlled context only: conjoined structures of the types illustrated above. In all other cases, to the best of our knowledge, the SSyntRel's seem to be sufficient (except perhaps in some cases with negation of the type indicated in Note 7, p. 77). Moreover, grouping becomes unavoidable only if we refuse to represent such grammemes as Russian adjective agreement and English posses­ sive directly in the SSyntS. Thus grouping is by no means the only solution pos­ sible: it forms part of one alternative ; the other solution, i.e. admitting syntactical­ ly-conditioned grammemes into the SSyntS when they become meaning-bearing, is convincingly argued for in Iomdin 1979. We choose grouping in this book mostly for methodological reasons: in order to expand our formalism and to show the reader how the expressive power of our apparatus can be enhanced. Cf., in this connection, Hudson 1980a, b and Dahl 1980. Note that we deliberately speak of grouping, avoiding the more familiar no­ tion of constituent. The fact is that the full D-subtrees that must be marked as

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

75

groupings need not be constituents in the strict sense of the term. 'Constituent' refers to a purely surface phenomenon and presupposes linear contiguity and specific 'uninterrupted' prosody while what we consider a grouping need not surface as a constituent. A trivial example might be the grammatical subject phrase in English. In (16), certain types of equations which possess the following three important properties . . . is a grouping but not a constituent (because of are introduced inserted between its two parts): (16) Certain types of equations are introduced which possess the following three important properties. Let it be emphasized that, contrary to the viewpoints of the proponents of COMBINED syntactic representations, where groupings are extensively used along with dependencies (like syntactic groups in Gladkij 1968), we admit oc­ casional use of groupings if and only if the dependency language proves genuinely insufficient. The main idea is to exploit the possibilities of dependency represen­ tation to the fullest extent. 8

Now we can take the next step and characterize the well-formedness condi­ tions for SSyntS's.

NOTES 1. (To page 56.) To formalize the current notion of 'to be syntactically dependent on', we use the following criterion: the syntactic governor imposes its passive syntactic valence upon the whole phrase "governor + dependent" (see p. 69, Criterion B.I). If this approach is accepted, then the uniqueness of the syntactic governor and the obligatory presence of exactly one top node will follow from the definition of syntactic dependency. 2. (To page 56.) Morphological dependencies are not (and should not be) shown in the surface-syntactic representation. Rather, they are computed on the basis of syntactic de­ pendencies (when going from meaning to text) and used by the surface-syntactic rules that morphologize the SSyntS, i.e., provide all the syntactically-conditioned morphological characteristics of the wordforms. In the case of (2), the corresponding Russian syntagm says that the adjectival complement of a copula verb must be in the instrumental, while its gender and number must be copied from the grammatical subject of the verb. 3. (To page 58.) The» concepts of VOCABLE (a term coined after Lat. vocabulum 'word; name') and LEXEME are important and must be defined here.

76

SURFACE SYNTAX OF ENGLISH

A lexeme is a minimal unit of lexicographic description; it corresponds, in a one-to-one fashion, to a dictionary entry. Informally, a lexeme is a word in a specific sense, with all its syntactic and morphological properties (characteristic precisely of this sense). In more rigorous terms, a lexeme is the set of all the wordforms having the same stem in a specific sense. The signifiés of two lexemes may share a non-trivial semantic component; we say then that they are directly linked. Two signifiés connected by a sequence of direct links are said to be linked. (For example, let lexeme X have the signifié 'abc', lexeme Y the signifié 'ade', and lexeme Z, the signifié 'efg'; then 'efg' is linked with 'abc' since 'efg' is directly linked with 'ade', and 'ade' is directly linked with 'abc'.) A vocable is a set of all the lexemes such that: (i) their signifiants (signantia.) are identical; (ii) the signifié (signaturn) of any lexeme belonging to the set is linked with the signifie of any other lexeme of the same set. Any two lexemes belonging to the same vocable stand in the relation of polysemy; any two lexemes having identical signifiants but belonging to two different vocables are con­ nected by the relation of homonomy. Examples: - HEAD 1 'the upper or anterior vertebrate extremity, containing brain . . .', HEAD 2 'the hair on the human head' and HEAD 3 'a rounded, compact mass of leaves . . ., as of lettuce etc., reminiscent of the form of the human head' are three different lexemes of the same vocable, since HEAD 2 and HEAD 3 include the signifié of HEAD 1. This is a typical case of polysemy (metonymy in HEAD 2 and metaphor in HEAD 3). - STUNT 1 'to check the growth or development of' and STUNT 2 'a feat displaying unusual strength, skill, or daring' belong to two different vocables: they do not share non-trivial semantic components and are homonyms. 4. (To page 63.) We might mention in this connection Jespersen's famous dictum about grammatical categories - such as, e.g., the preterit - 'which Janus-like face both ways, towards form, and towards notion' (1969: 98). Jespersen calls grammatical categories 'the meeting-ground of outer and inner' (idem , 97; see pages 95ff.), i.e. of morphological markers and meanings. 5. (To page 65.) In his insightful paper (1979-80), V.Z. Sannikov thoroughly analyzes a set of Russian constructions illustrating a syntactic phenomenon which he calls 'lexicosemantic coordination': (i) Nikto i ni dlja kogo ne staraetsja, lit. 'Nobody and for nobody not makes-any-effort', i.e. 'Nobody will make any effort for anybody', (ii) Kto i kogo tuda posylal?, lit. 'Who and whom sent there?', i.e. 'Who sent whom there?' Sannikov feels that such constructions throw doubt upon the universal sufficiency of D-trees for the representation of coordination (and of comparative constructions, which he - quite rightly, in our opinion - prefers to consider together with coordinate con­ structions). Sannikov's proposal is to reject the principle of the uniqueness of the syntactic governor and to admit double dependency in 'difficult' cases, i.e. in coordination and comparison. However, although his arguments are very cogent, the facts he uses seem to us to be too marginal to justify such a deep and far-reaching revision of our descriptive ap­ paratus. We hope that a less radical solution can be found for the problems he raises, but for

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

77

the time being we are not in a position to offer one. So we limit ourselves to staking out the difficulty. 6. (To page 65.) Of course, a different label could be used on a SSynt-dependency arrow to show the dependence on the whole of the phrase, as distinct from the dependence on its head. For example: (i)

men and women 'old men, as well as women of all ages' vs.

(ii)

men and women 'old men and old women';

'phrase-modif' stands for 'phrase-modificative', i.e., a SSyntRel that covers the modification of the whole phrase. However, we find such a method highly unnatural, and it will not be considered in what follows. For cases where the distinction between the dependence on the phrase as a whole and the dependence on its head only is linguistically relevant, a device called 'grouping' (roughly speaking, bracketing) is used, see below. In old men and women it appears as fol­ lows: (iii) (iv) 7. (To page 74.) To be sure, there are some other cases believed to call for the use of groupings in the SSyntS. Consider, for example, the following Russian example (Apresjan, Iomdin and Percov 1978: 273-276): (i) Ustanovka ne rabotala tri goda 'The device did not work for three years' = 'The period of its idleness was equal to three years'. vs. (ii) Ustanovka ne rabotala trex let 'The device did not work three years' = 'The period of its functioning was less than three years'. Since the formal mark of the contrast is the grammatical case of the circumstantial ad­ verbial (the accusative in (i), the genitive in (ii)) and the case, as a purely syntactic means, is not retained in the SSyntS, the authors suggest using grouping to avoid'loss of surface in­ formation: roughly, (iii) (ne rabotala)  (tri goda) vs. ne (rabotala  (trëx let)). But we feel that the major difference lies here in the communicative structure: in (i), ne rabotala is a second degree theme and tri goda, the second degree r h e m e ( u s t a n o v k a is and ne rabotala tri goda, while in (ii), rabotala →(trëx let) is a singlefirstdegree rheme negated as a whole and with emphatic stress on rabotala. What we are saying, then, boils down to reflecting the difference observed on the surface in the SSyntCommS, rather than in the surface-syntactic structure as such. Our guess is that most of the cases where the scope of a particle (negation, only, even, and the like) is relevant do not presuppose groupings in the SSyntS but should be represented by means of the SSyntCommS or the SSynt-ProsS. (For a discussion of the corresponding problems see, e.g., Boguslavskij 1978.) Nevertheless, since we have not carried out the necessary research we by no means insist on our proposal, so that grouping in syntactic structures with negation and other scope-toting particles remains a moot question.

78

SURFACE SYNTAX OF ENGLISH

8. (To page 75.) Since dependency formalism is still not sufficiently known in theoretical linguistics, it might be a good idea to offer here a list of works in which the notion of syn­ tactic dependency is introduced, explained, or used extensively. The order of references is (roughly) chronological. Tesniere 1959 Hays 1960,1961,1964b Lecerf 1960 Hirschberg 1961 Fitialov 1962, 1968 Mel'čuk 1962, 1963, 1964a (especially, 1 7 - 2 7 ) , 1970,1973,1974b (especially, 208-235), 1979a, 1981b Mel'čuk and Žolkovskij 1970 Mel'čuk and Percov 1973a, b, 1975 Iordanskaja 1963, 1964, 1967 Marcus 1965a, b Gladkij 1966,1968,1973:

Beleckij, Grigorjan and Zaslavskij 1963 Beleckij 1967 Kunze and Priess 1967-1971 Kunze 1975 Robinson 1970a, b Heringer 1970 Goralčková 1973 Jelitte 1973 Vater 1975 Machová 1975 Hudson 1976, 1980a, b Happ 1976,1977,1978 Dönnges and Happ 1977 Garde 1977 Matthews 1981: 7 1 - 9 5

2 8 2 - 3 1 0 , 1980-1981 Add to this list the publications on SSyntRel's mentioned on p. 63.

§3 Conditions for Grammatical Correctness of English Surface-Syntactic Structures As stated above (page 48), only one component of the surface-syntactic re­ presentation of English sentences will be considered at length in this book: surface-syntactic structure. Having characterized the formalism used (= D-trees) and having explained to some extent the notion of surface-syntactic relation, we will now briefly touch on the problem of formally specifying the set of permit­ ted, or grammatical, English SSyntS's. Though important, this problem is rather marginal for the concerns of this book. We therefore limit ourselves to sketchy remarks and examples. There are basically two classes of conditions that must be satisfied by any well-formed SSyntS:

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

79

— Conditions for grammatical correctness of individual items in the SSyntS; in fact, these are conditions on the well-formedness of a given reduced DMorphR of a wordform. — Conditions on the cooccurrence of different items in the SSyntS. 1.The correctness of individual reduced DMorphR's appearing in a SSyntS should not interest us too much in this particular context. The problem is what meaning-bearing morphological values (= grammemes) must necessarily appear with what lexemes and which of these values can or cannot co-occur within the same partial DMorphR of a wordform. For instance, in a SSyntS an English noun must be subscripted for number, a finite verb — for tense and mood, and so on. We will not occupy ourselves with this matter any longer; cf. item 1 in § 3 of Chapter III, page 166ff. 2.Conditions of the second class are twofold: ■ Element X cannot occur with element Y. ■ Element X cannot occur without element Y. Both formulas are applicable either 1) to the cooccurrence of SSyntRel's with other SSyntRel's, or 2) to the cooccurrence of SSyntRel's with lexemes, or 3) to the cooccurrence of lexemes with other lexemes. Note that when talking about the cooccurrence of elements within a SSyntS, we don't simply mean 'cooccurrence anywhere in the SSyntS'. Cooccurrence in the sense intended here should be construed as 'contiguous cooccurrence only, the type of contiguity depending on the nature of the elements involved'. For example, for two SSyntRel's r1 and r2, to co-occur means 'to co-occur with a shared node (or nodes)', i.e. to form structures of the following form: or while for two nodes X and Y, to co-occur means 'to co-occur with a shared SSyntRel (or SSyntRel's)': X  Y ; e t c . 1) Conditions on the cooccurrence of different SSyntRel's in English. Example 1. For any non-repeatable (see above, p. 71) SSyntRel r, there can be no more than one branch labeled r stemming from the same node:

Example 2. As a general rule, a branch labeled "determinative" and a branch labeled "possessive" cannot simultaneously stem from the same node. In other words, an English noun cannot simultaneously have both a determiner and a possessive modifier:

80

SURFACE SYNTAX OF ENGLISH

(An exception involving the so-called qualifying, or non-referential, possessive form, is mentioned under the corresponding syntagm, page 406.) Example 3. A configuration of the form

is impossible: a noun [= Y] which is the dependent member of the compositive SSyntRel cannot subordinate a determiner rela­ tion) . 2) Conditions on the cooccurrence of SSyntRel's with lexemes in English. The two main fields covered here are the implementation of the passive valence of lexemes and the saturation of their active valence. The PASSIVE (SSynt-)VALENCE of a lexeme is its ability to be subordinated, in a specified role, to lexemes of a certain class. For instance, high, as an adjective, can be subordinated to a noun as its modifier (high prices), to a copula verb (is etc., which means that an Adj can co-occur with the modificative SSyntRel, the 1st completive SSyntRel (if the head is a copula), etc., as the dependent member. The ACTIVE (SSynt-)VALENCE of a lexeme is its ability to subordinate, in a specified role, lexemes of a certain class. For instance, a transitive verb can, and sometimes must, have a direct object, i.e., co-occur with the 1st completive SSyntRel — as its governing member. The corresponding conditions refer to the surface-syntactic properties of lexemes. These properties are of three types: — part of speech, — syntactic features (roughly, intersecting syntactic subclasses), — government pattern, and they are systematically recorded in lexemic dictionary entries (constituting the syntactics of lexemes, cf. above, p. 61). Most of the conditions dealing with SSynt-valence are built directly into the syntagms. In fact, the left-hand part of a syntagm is nothing more than a state­ ment about the possible cooccurrence of two types of lexemes with a specific SSyntRel: e.g.,

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

81

(cf. Syntagm 1.3, page 221),

which means that a gerund can be the dependent member of the predicative SSyntRel (on condition that the governor is a copula), and so on. There is, however, a special subclass of conditions responsible for saturating OBLIGATORY active valence slots (we mean here any active valence rather than actant valence only). These conditions refer to two kinds of information: (a) Lexicographic specification of the obligatory saturation of a given valence slot (for a given lexeme). For instance, the verb MEET has in its dictionary entry the feature (2 [oblig]), i.e., the second deep-syntactic actant of this verb implemented by its first surface-syntactic complement cannot be absent from the SSyntS (except on a reciprocal reading): (1) *I meet pretty often. Certain verbs cannot appear in a SSyntS without a prepositional object, as, e.g., RELY without ON + N(oun) P(hrase), CONSIST - without OF + NP in one sense and without IN + NP in its other sense, and so on. (This property is indi­ cated by the above-mentioned feature (2 [oblig]) in the dictionary entry of the corresponding verb, more precisely, in its government pattern.) For any SSyntS to be well-formed, no lexeme marked (i [oblig]) in its govern­ ment pattern can appear in this SSyntS without the (i'-l)th completive SSyntRel stemming from the corresponding node (except for the situations described in item (b) immediately below). (b) List of permitted or obligatory non-saturation situations. There are SSyntconfigurations that allow a given active valence slot which otherwise must obligatorily be filled to remain unsaturated, or even require that it be so. Example 1. A finite (= tensed) English verb cannot appear without heading the predicative SSyntRel (i.e., without a grammatical subject). However, there are a number of specific cases where a finite verb node does not head a predica­ tive SSyntRel: — conjoined verbs having the same grammatical subject, e.g., left in (2):

SURFACE SYNTAX OF ENGLISH

82

— imperative forms (with YOU deleted); — the main verb of a parenthetical clause introduced by AS: (3)

As is widely known, . . .;

— etc. Example 2. A preposition must have a dependent NP, i.e., any node labeled with a preposition must head a prepositional SSyntRel. Nevertheless, there are some situations of permitted non-saturation of prepositional nodes that create stranded and dangling prepositions (see pages 124 and 131): — In relative clauses with the dependent of the preposition relativized this dependent may be omitted: (4)

a. The table (which/that) the boy crawled under had been bought long ago. b. This is the university (that) he works at.

. The girl (who(m)/that) I glanced at was Gloria. — In some infinitive constructions (cf. (5)) and with passives (cf. (6)) the dependent of the preposition must be omitted: (5)

a. The paper is too flimsy to write on (*it). b. Joan seems too arrogant to speak with (*her).

(6)

Nobody likes being made fun of (*him).

— In constructions with several conjoined prepositions all the prepositions but the rightmost one can remain without a dependent NP (in in (7)): (7) "Essays in and  on Machine Translation" [title of a book]. A full-fledged model of surface syntax must include a list of all such SSyntconfigurations; in other words, of all situations observed in a SSynt-tree which admit of or entail non-saturation of strong valence slots. This list will be used by appropriate rules checking the conditions for grammatical correctness of English SSyntS's. Note that in all cases where not only SSyntRel's but lexemes as well are involved, it is not always easy to draw a line between cooccurrence restrictions concerning surface-syntactic structure (which are of interest to us here) and those operating at the level of deep-syntactic structure (and thus irrelevant for our purposes). Consider, e.g., the cooccurrence of nouns with determiners. This

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

83

seems to be mostly a problem of DSyntS; but there are some cases where the determiner is probably induced by the SSyntS, e.g., the indefinite article before a singular complement of certain copula verbs: (8)

a. John became a traitor. b. John turned (*a) traitor.

So our examples are no more than an attempt to indicate a direction for further research. 3) Conditions on the cooccurrence of different lexemes in English. We will briefly describe the most typical case of lexical cooccurrence, which is relevant to the grammaticality of a given SSyntS and cannot naturally be stated at the DSynt-level, and then indicate another, less obvious, case. Example 1. The most important instance is, beyond any doubt, GOVERN­ MENT. The cooccurrence of a specific syntactic governor with a specific func­ tion word it governs via a SSyntRel (a preposition or a conjunction) is given by the government pattern of the former, i.e., in the dictionary. For example, the verb COMPARE has in its syntactics the feature (3 [TO]), and this feature pro­ vides for the acceptance of (9)

a.

and the rejection of, e.g., b. A specific instance of government may be further restricted by certain condi­ tions. A typical example is the presence of another specified dependent of the governor in question; thus, KNOW can govern WHETHER only when negated or in a question (direct or indirect): (10) a. It is not known < I would like to know > whether he stays. b. *It is known whether he stays. For verbs like know, the supplementary condition is recorded in the government pattern: (2 [WHETHER] |

or in interrog. form)

(Many restrictions of the kind are found in Sager 1968.)

84

SURFACE SYNTAX OF ENGLISH

Example 2. More than one negative element cannot co-occur with the same verb (in Standard American English): (12) *I haven't never been on none of the big liners, neither. With a negated verb, all the pronouns must appear in non-assertive form (Quirk et al 1972: 376 ff.): (12*) I haven't ever been on any of the big liners, either. In general, there are fairly complicated rules for the use of assertive, non-as­ sertive and negative elements in English, all of them more or less automatic, i.e., syntactically relevant: (13) a. We've had some wine ~ We haven't had any wine. b. He's still there ~ He isn't there any longer. . She is coming, too ~ She isn't coming, either. d. with some delay ~ without any delay ; e. I often get some sleep after dinner ~ I seldom get any sleep after dinner. However, it is not at all clear whether all these rules should be relevant to the surface SyntS and not to the deep SyntS. Basically, what they determine is the choice of lexemes, and all the lexemic choices, except for function words, are supposed to be made at the deep-syntax level. Because of this, most of the selectional restrictions controlling lexemic cooccurrence must be relegated to the rules of well-formedness of DSyntS's and thus are not relevant for us here. One important point remains to be made in this section. Obviously, there are numerous restrictions that determine the cooccurrence of elements in SSyntS's of a given language. These restrictions are not only interesting but absolutely vital to a full-fledged syntactic description. This fact notwithstanding, they fall outside the narrow limits set by the authors in the present book. There are two reasons why this is so. First, we are not working with deep syntax, and many of the cooccurrence restrictions involving lexemes belong precisely to the deep-syntactic level. Much of what is impossible in the SSyntS is so because it is impossible already in the DSyntS. Yet all such cases are beyond our concern. Second, the only goal of this book, as has been stated repeatedly, are rules of correspondence, between SSyntS's and DMorphS's of English sentences - while the restrictions in question, even if they do bear on the surface-syntactic struc-

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

85

ture, belong to the rules for the well-formedness of English SSyntS's. We strong­ ly feel that both sets of rules should be kept apart as strictly as possible, and the latter type of rules is not included in this book. In view of this, many problems that are the staple diet of modern syntacticians and take up the lion's share in most American syntax-oriented works simply do not appear — and MUST NOT appear — in the exposition below (Chapter IV). In our opinion, a formal description of all the rules specifying the correct SSyntS's in Modern English is a first-priority task of syntax; but it is not our task in this book. What has just been said does not prevent us, though, from smuggling some of the above-mentioned information into the syntagms. We do so in order not to waste useful knowledge — in the hope that later it will be easier to extract all such indications from the syntagms and state them in an autonomous division of the SSynt-component. For instance, we formulate some restrictions on the cooccurrence of SSyntRel's when introducing a new syntagmeme ; a number of restrictions on lexemic cooccurrence, as well as certain indications as to permit­ ted or obligatory non-saturation of active valence slots, etc. are given within syntagms; and so on.

§4 Surface-Syntactic Relations in Modern English The present inventory of SSyntRel's for Modern English has been compiled as the result of several years of research. The first version was published in 1973 (Mel'cuk and Percov 1973b); it was heavily corrected later (cf. Mel'cuk and Percov 1975) and reworked once more for this publication. The following four preliminary remarks seem to be in order. 1. Our inventory does not claim completeness. We have compiled it mostly on the basis of scientific and journalistic prose, leaving aside many colloquial constructions. Moreover, no elliptical sentences have been considered, and thus no ellipses of any kind are covered by the SSyntRel's proposed. For instance, we do not know how to represent the SSyntS of such sentences as in (1): (1)

a. The simpler the grammar unambiguous structure.

, the easier it would be to obtain an

SURFACE SYNTAX OF ENGLISH

86 b.

One is predictive analysis, the second approach.

c.

We want the top nearly white but

the author's fulcrum the bottom a bronze.

Besides, we have left out of our study many so-called minor type sentences, in particular, sentences lacking a finite verb, as in (2): (2)

a. Why natural language? b. Britain to nationalize aircraft industry. c. And now to return to idioms.

An illustrative list of construction and sentence types not accounted for in our SSyntRel inventory nor in the syntagms (Chapter IV, § 4) is provided in Ap­ pendix IV, pp. 501-505. 2. Our inventory does not claim adequacy, either. There is no guarantee (nor can there be) that all the SSyntRel's are introduced in the best possible way or that the choices we have made to represent syntactic connections in a family of actual English phrases and sentences are the most felicitous ones. On the contra­ ry, we are sure that many very common constructions are not described satis­ factorily (cf. the remark on p. 72„ at the end of item 4). The problem is that we do not have ready solutions for the SSynt-description of comparative, conjoined, absolute, numerical, etc. phrases. Therefore, the selection of SSyntRel's for a number of phrase classes turns out to be an autonomous, and highly involved, problem of theoretical syntax. For obvious reasons, in all such cases we have had to be satisfied with less than ideal solutions, of whose temporary character we are fully aware. To avoid overloading our text, we have often been forced to forego considering and comparing some alternative solutions. 3. We do not require that our SSyntRel inventory be minimal. The inventory as a whole must be sufficient for describing any SSyntS (within the limits of the corpus chosen) but some SSyntRel's in it may not be necessary. Our SSyntRel's include some formally redundant relations that duplicate information contained in the labels of the nodes. However, such redundant SSyntRel's are deliberately admitted with the purpose of providing a better overview and a more natural, compact, and elegant presentation of the syntagms. 4. We do not provide a thorough validation for our choice of individual SSyntRel's or for their direction (i.e., for the choice of the syntactic governor). Hoping that in most cases the introduction of the SSyntRel considered is more or less obvious, we limit ourselves to sketching very briefly the reasoning involved

SURFACE-SYNTACTIC REPRESENTATION FOR ENGLISH

87

in a few problematic instances. Formally speaking, our SSyntRel's are postulated and must be accepted as a kind of working hypothesis. To describe the surface-syntactic structure of English sentences, 40 surfacesyntactic relations are proposed. We will list them in the following order: • First, all the subordinate SSyntRel's, then the coordinate one. • Among the subordinate SSyntRel's, the sentential SSyntRel's, i.e. the SSyntRel's linking sentence elements at the sentence level, precede the phrasal SSyntRel's, i.e. those linking individual wordforms within phrases. • The sentential SSyntRel's are classified into valence-controlled, or (strong­ ly) governed, and adverbial, or circumstantiative, the former being sub­ divided further into actantial, absolute, copredicative, and comparative. (Note, however, that the term valence-controlled, as applied to the clas­ sification of SSyntRel's, is interpreted here and later on rather loosely. We use it, by extension, to cover two types of SSyntRel's: first, genuine valence-controlled SSyntRel's, whose dependent elements do fill strong valence slots of governing lexemes; second, SSyntRel's that are, strictly speaking, not valence-controlled but are closely related to the first type. The latter include, in our case, the absolute-predicative and the three copredicative SSyntRel's; see below. Maybe a more neutral term would be appropriate: something like major SSyntRel's) • The phrasal SSyntRel's are grouped according to the type of phrase they organize: general, verb phrase, word-like phrases, expletive and conjunc­ tional phrases, and noun phrase (in the broad sense, including prepositional phrases). Note that some groups of SSyntRel's may contain only one SSyntRel each, and that in such a case the name of the group does not necessarily coincide with the name of its only element. At the end of this section, i.e. immediately following the list of English SSyntRel's, an extended illustration is provided: four English sentences with their surface-syntactic structures (Figs. 1 througli 4, pp. 157 ff.) showing many of the SSyntRel's introduced.

88

SURFACE SYNTAX OF ENGLISH Guide to the List of Surface-Syntactic Relations in Modern English Pages

I. Subordinate SSyntRel's

91

A. Sentential SSyntRel's

91

1. Valence-controlled SSyntRel's — Actantial SSyntRel's 1. Predicative 2. Completive (1st, 2nd, 3rd, 4th completive) 3. Agentive — Absolute SSyntRel 4. Absolute-predicative — Copredicative SSyntRel's

91 91 91 93 99 100 100 102

5. Subjective-copredicative 6. Pronominal-subjective-copredicative

102 105

7. Objective-copredicative

105

— Comparative SSyntRel

106

8. Comparative

106

2. Adverbial SSyntRel's

108

9. 10. 11. 12. 13. 14.

Adverbial Modificative-adverbial Appositive-adverbial Attributive-adverbial Parenthetical Adjunctive

B. Phrasal SSyntRel's 1. General 15. Restrictive 2. Verb phrase 16. Auxiliary

108 112 114 115 116 118 118 118 118 120 120

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

89 Pages

3. Word-like phrases 17. 18. 19. 20.

Phrasal-junctive Numeral-junctive Binary-junctive Colligative

4. Expletive phrases 21. Expletive 5. Conjunctional phrases 22. 23. 24. 25. 26.

Subordinate-conjunctional Coordinate-conjunctional Predicative-conjunctional Completive-conjunctional Absolute-conjunctional

6. Noun phrase 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39.

Prepositional Determinative Quantitative Modificative Descriptive-modificative Possessive Compositive Elective Appositive Descriptive-appositive Sequential Attributive Descriptive-attributive

II. Coordinate SSyntRel 40. Coordinative

121 121 123 123 124 125 125 127 127 127 128 128 129 129 129 132 133 134 137 139 140 142 144 145 147 148 151 153 153

90

SURFACE SYNTAX OF ENGLISH

We will briefly characterize each SSyntRel, exemplifying its use with English sentences that manifest it. Let it be emphasized that we feel under no obligation to be exhaustive in our characterization. That is, we do not claim to list all the possible dependents or possible governors of each SSyntRel. Rather, we indicate only the most typical items that can fulfill the said roles; an exhaustive descrip­ tion is left to the surface-syntactic rules, more precisely, to the syntagms (see pp. 216ff). The format of the description is roughly as follows. First, we give the name of the SSyntRel at issue (italicized), then its abbrevia­ ted name (in square brackets) and then indicate that it connects such-and-such Y — its dependent member — to such-and-such X — its governing member. It is important to make clear that the name we choose for a SSyntRel is no more than a convenient label and should not be construed as bearing a semantic load connoted by the current sense of the corresponding word. Thus the agentive SSyntRel is not necessarily related to agency, the possessive SSyntRel — to pos­ session, etc. Second, informal comments may follow: concerning the direction of the SSyntRel, related SSyntRel's, special cases of its use, etc. Third, we enumerate the most typical items that can be the governing and de­ pendent members of this SSyntRel (with examples). It would be highly desirable to supplement our list with a pairwise comparisons of all the SSyntRel's in it. In this way, we could establish semantic and formal contrasts and thus gain more insight into the nature of the SSyntRel's. (Such a procedure is currently applied to phonemic inventories, in order to show how each phoneme is opposed to each other phoneme.) Unfortunately, we are in no position to do so systematically ; we thus limit ourselves to occasional remarks in this sense.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

91

I. Subordinate SSyntRel's A. Sentential SSyntRel's 7. Valence-controlled SSyntRel's - Actantial SSyntRel's. 1. The predicative SSyntRel [predic] connects the grammatical subject to its governor — the main verb (considered to be the top node of the clause it belongs to). The direction of the predicative SSyntRel — FROM the main verb TO the subject — is justified by the fact that the passive SSynt-valence of a clause is that of a finite verb rather than that of a noun phrase, so that the finite verb must be deemed the governor, or the head. (On the criterion of passive valence for the direction of SSyntRel's see above, § 3 of this chapter: B.I.) The predicative SSyntRel is privileged in English in that any simple nonelliptical clause features exactly one predicative SSyntRel (property shared by no other SSyntRel). Three possible complications should be pointed out: (i) If there are several conjoined subjects, each of these, except for the leftmost, is connected to the preceding one, which is its governor (via the coordinative SSyntRel). The leftmost grammatical subject is thus the head of the con­ joined subject phrase, and it is this head only that is subordinated to the main verb via the predicative SSyntRel: predic

(ii) If there are several conjoined main verbs, each of these, except for the left­ most, is also connected to the preceding one, which is its governor (via the co­ ordinative SSyntRel). The leftmost main verb is the head of the conjoined predicate phrase, and it is this head only that subordinates the grammatical subject (simplex or conjoined) via the predicative SSyntRel:

92

SURFACE SYNTAX OF ENGLISH

Thus conjunction of subjects and/or of main verbs does not interfere with the uniqueness of the predicative SSyntRel. (For more details about the coordinati­ ve SSyntRel, see pp. 153 ff.) (iii) If there is no grammatical subject or neither grammatical subject nor main verb, the clause in question is judged to be elliptical; cf. the italicized clauses in (5): (5) a. Note that only the second rule applies. b. Curious that these plurals are not derived from conjoined sentences. In (5a), the surface-syntactic subject YOU has been deleted, and in (5b), the whole predicative phrase The governing member of the predicative SSyntRel being invariably a finite verb, its dependent member can be one of the following four items: 1) A noun, pronoun, or any other syntactic equivalent of a noun (see the stan­ dard subtree ANP, pages 485-487, Appendix II):

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

93

Note that if the dependent member of any SSyntRel r is a clause that is not introduced by a subordinate conjunction, then its top node is its main verb and the arrow of enters this top node. Such is the case of (6 f). Cf. also (7), the adverbial SSyntRel:

2. The 1st, 2nd, 3rd, or 4th completive SSyntRel [1st, 2nd, 3rd, 4th compi] connects an object or a complement to its governor. Objects and complements are surface-syntactic actants as opposed to circum­ stantials. By SSynt-ACTANT we understand an item (= a phrase) which cor­ responds to, i.e. fills in, a semantic valence slot of its governor. Thus, the pre-

94

SURFACE SYNTAX OF ENGLISH

prepositional or adverbial location phrase depending on the verb STAY is its actant since the idea of location is implied by this verb: a location phrase is provided for by the lexicographic definition of STAY and is foreseen in its government pattern. Therefore, in Birmingham (with his mother, at the baker's, there) are actants, i.e. completive phrases, in sentences like John is staying in Birmingham etc. But with verbs BUILD, DIE, CREATE, etc. all such phrases are no more than circumstantials insofar as the semantics of the verbs mentioned does not necessitate the location. English, as some other languages, has verbs whose meanings feature up to six semantic variables, or semantic valence slots (Apresjan 1974: 135—137). Fourvariable verbs are, e.g., verbs of selling and exchange: SELL — [l]who, [2]what, [3]to whom, and [4]for how much; EXCHANGE - [l]who, [2]what, [3]with whom, and [4]for what. Five-variable verbs are, e.g., verbs of leasing: RENT — [1 ]who, [2]what, [3]to whom, [4]for how much, and [5] for how long (it is the term of renting that distinguishes renting from selling). An example of sixvariable verb is EXILE: [l]who, [2]whom, [3]from where, [4]to where, [5]for what, and [6]for how long. Since, however, such a verb seems rather a rarity in English we permitted ourselves to consider only verbs with no more than 5 semantic slots, which leaves us with 4 completive SSyntRel's (the grammatical subject plus four completives). Note that the 1st completive SSyntRel is used to describe some constructions where neither objects nor complements are traditionally recognized. We will mention here the following four such cases: (i) A parametric adjective plus the preposed numerical phrase which expresses the value of the parameter:

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

95

b. The extensional definition can be transcribed in such a way that it is made intensional. The dependency between SO/SUCH and the conjunction AS1 is represented, however, in a different way: in terms of the comparative SSyntRel. The main reason is the parallelism with other usages of AS1 ; cf. examples (3) on page 107. (iv) A modal or quasi-modal verb plus an infinitive:

The problem of multiple (= conjoined) completives or a completive to a multiple governor, i.e., to a series of conjoined verbs or other conjoined items, is solved in much the same way as that of multiple subjects and multiple predicates for the predicative SSyntRel: All the conjoined items are subordinated to the leftmost one, which represents the whole conjoined phrase at the SSynt-level. The completive SSyntRel in question leaves the rightmost governor and enters the leftmost completive; thus the uniqueness of each completive SSyntRel is preserved. Cf. the examples below:

The governing member of a completive SSyntRel can be a verb (in any form), an adjective, a noun, or an adverb, i.e., an item from any major word class. The dependent member of a completive SSyntRel can be any one of the same first three items as the dependent member of the predicative SSyntRel, or a conjunctionless clause, a prepositional phrase, an adverb, or an adjectival; all in all, any of seven different types:

SURFACE SYNTAX OF ENGLISH

96

1) A noun, a pronoun or any other syntactic equivalent of a noun (see ANP, pages 485-487): (6)

a. This term identifies some property of the object. b. A few nights later the workmen took c. It is very unlike

him aside.

Walsh to be late.

2) A verb in the form of an infinitive or a gerund:

(7)

a. An intelligent parser needs any point in the parse. b. The delegation is reported

to have left.

[The verb REPORT, like many other verbs, can govern a completive to-infinitive only when it is itself in the passive.] . It is

to forget.

[(7c) means 'This thing can be easily forgotten' and contains it = IT 1 , i.e. a substitute pronoun; forget, where we have a different it [= IT 5 , i.e. the anticipatory IT], a different SSyntS, and a different meaning: 'To forget [anything] is easy'.]

ƒ A policeman

the street in front of the bank.

From (7d-'f) we see that, unlike the predicative SSyntRel, completive SSyntRel's admit 'bare' infinitives (= without TO2) with some lexemes.

97

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

Notice also that in the so-called Accusative with Infinitive construction (ex­ ample (7f)) we treat the noun and the infinitive as the first and the second complements, respectively, of the main verb. g. We consider h. These goods are worth

-buying.

3) A subordinate clause introduced by THAT5, WHETHER1 or IF 2 , or by a wh-word (except what and the -ever-word): (

8

)

n

o

deep structure has become to point out the obvious truth

strike people as preposterous. b. The reader does not know whether he is reading about an employment or purchase situation. c. . . . more tests to

succeeded.

4) A conjunctionless subordinate clause, including direct speech: (9)

a. American voters fear the U.S. is not keeping pace with Soviet power.

b. "I called", he told Jane, "to speak with your husband". 5) A prepositional phrase or an (10) a. Others simply impose mother-tongue patternsbn English. b. The immigration service was not aggressively searching

for

them.  Only once did Rajai refer d. This assumption is true

e. It is possible to

-to the hostages. -of ordinary pronouns.

as words of various classes.

SURFACE SYNTAX OF ENGLISH

98

6) An adverb: (11)

He was sent

7) An adjective or a participle: (12) a. This theory was most influential.

[Examples (12c—d) are borrowed from Borkin 1973, where the relationship between this construction and a similar one with to be is analysed: The UPI reports his division to be crossing the border. ]

[(12e) means roughly 'Somebody handed us a note'; but the same string can have a different reading corresponding to a different SSynt-analysis of it: e'. We had a note landed to us [by John], with handed to us a postmodifier of a note and meaning 'We possessed a note handed to us before'.] ƒ g. She ordered Pete removed from the day room. An important remark. After several years, we feel that our classification of surface-syntactic completives is not quite satisfactory and, therefore, the comple­ tive SSyntRel's will probably have to be replaced. It seems, in particular, that our 1st, 2nd, . . . SSyntRePs follow too closely the deep-syntactic relations and thus do not represent faithfully enough the surface-syntactic features of cor­ responding constructions. Maybe traditional SSyntRePs — such as direct object,

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

99

indirect object, prepositional object, and complement — would be more ap­ propriate. However, this is not immediately obvious. For instance, what should we do then with examples of the (12c—g) type? And what about the construc­ Other difficulties will ap­ tion "modal verb + infinitive" (He can ■ pear as well. Because of all this, we chose to stick to our first solution, although we clearly see many of its shortcomings. Otherwise, we would face too drastic a modification of our description, with no guarantee of a definitive success. The same considerations apply for the complement of a copula verb: (to) be of interest, (to) seem in order, (to) prove important, (to) become a pilot, etc. This structure exhibits so many special characteristics that the problem arises as to whether it would not be advisable to introduce a new SSyntRel, say copula­ tive, to cover it. Yet we have preferred to avoid such radical revision of our inventory. 3. The agentive SSyntRel [agent] connects a prepositional phrase implemen­ ting either the first deep-syntactic actant of an infinitive or a noun or the second deep-syntactic actant of a passive form1 to its governor. Roughly speaking, the agentive complement denotes what is loosely called '(semantic) subject' and at the SSynt-level is a transform of the grammatical subject. Not every grammatical subject (even if we limit ourselves to active verbs) denotes a genuine semantic 'Agent'. Consequently, our agentive complement does not necessarily correspond to an 'Agent': it is no more semantically agentive than any arbitrary grammatical subject. The name of this SSyntRel should not therefore be taken to entail semantic associations of agency; as has been men­ tioned on page 90, it is nothing more than a handy label for a particular syntac­ tic construction. The agentive SSyntRel covers three different cases and accordingly admits as its dependent member three different prepositional phrases: 1) A by-phrase with a passive participle: (1)

a. His lecture was attended

by a great number of people.

b. I remain totally baffled

by tenses and quantifiers.

c. The telegram followed

by a letter disturbed everybody.

2) Various prepositional phrases (mostly, of- and by-phrases) implementing the first DSynt-actant (= 'semantic subject') of a noun: (2)

a. activities

of the committee; the paper

by Mr. Fowler;

100

SURFACE SYNTAX OF ENGLISH

b. It amounted to deliberate, cold-blooded murder nians.

by the Alba­

3) A for-phrase with a to -infinitive: (3)

a. It was not a night for a husband to be missing. b. I will be content for you to name your own fee. c. He stood up for me to sit down.

- Absolute SSyntRel. 4. The absolute-predicative SSyntRel [abs-predic] connects the predicatelike element of an absolute construction to its governor — the head of the noun phrase (or its syntactic equivalent) being part of the same construction. (Note that the head of the noun phrase, mostly a noun, is, so to speak, a 'quasi-gram­ matical subject', while the dependent member is a 'quasi-grammatical predicate'.) The absolute-predicative construction is considered to be a specific nominalization of a clause capable of filling the following four surface-syntactic roles: — an adverbial modifier (with or without one of the prepositions WITH or WITHOUT); — a complement of a verb (more specifically, the 1st complement); — a dependent of a preposition; — a descriptive appositive to a noun. The direction of the absolute-predicative SSyntRel — from the noun as the governing member of the construction to the predicate-like element as its dependent member — has been adopted because it is the noun that seems to represent the construction in its external SSynt-relationships. More specifically: • The surface-syntactic passive valence of the absolute-predicative construc­ tion is rather that of a noun than that of a participle, an adjective, or a preposi­ tion. This is obvious insofar as we consider the capacity of the construction to be a complement, a dependent of a preposition or an appositive. As for its capacity to be an adverbial modifier, nouns as such do not normally fill this role — but a few nouns do, if accompanied by appropriate modifiers (such as last May or three miles to the North). Therefore we feel justified in grossly

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

101

equating, with respect to surface-syntax, the absolute-predicative construction and the noun. • The nominal member of the construction is its morphological contact point: criterion B.II p. 69 (see also Mel' čuk 1981b: 42 ff). If it is a pronoun it responds by its oblique form, whenever required, to the governing verb or pre­ position: (1)

a. Do you mind him ‹*he› speaking to your father? b. Without me (*I) knowing the answer . . .

• When the second member of an absolute construction is a prepositional phrase, it would be too unnatural to consider the nominal member to be depen­ dent on it. Cf.: (2)

a. He came in, a stick b. He came in, a stick

In (2b), we get two noun phrases depending on the same preposition. But if we reject the analysis in (2b), then considerations of uniformity prompt us to take the nominal as the head in all absolute constructions. The dependent member of the absolute-predicative SSyntRel can be one of the following five items: 1) A noun:

(3)

His first shot a failure, Dick fired again

2) An adjective or a participle: (4)

a. There are more such types, some of them very specific. b. We are opposed to these men

being deprived of anything.

c.  walked upstairs, his thirst already gone. d. A single grammar consists of blocks of rules, each block ordered with respect to others.

SURFACE SYNTAX OF ENGLISH

102

3) A prepositional phrase:

(5)

Johnny, his eyes still toward the table, took one out.

4) An adverb : (6)

He was found dead on the floor, his gun

nearby.

5)  -infinitive: (7)

a. The sellers offered 500 tons, delivery October. b. Without a man of trouble.

to be made in

to see to this, there is likely to be all sorts

— Copredicative SSyntRel's (5—7). 5. The subjective-copredicative SSyntRel [subj-copr] connects the predica­ tive modifier that characterizes the grammatical subject to the main verb. Such a predicative modifier is also called secondary predicate (see, e.g., Nichols 1978: 114). The subjective-copredicative SSyntRel should be distinguished from the fol­ lowing four SSyntRel's which are closely related to it: First, from the 1st completive SSyntRel with copula and semi-copula verbs: (1)

a. He was a nero vs. He returned a hero. b. He seemed

nappy vs. He walked along happy.

 He remained a beggar vs. He died a beggar. The left-hand members of the pairs quoted show (strong) government: the verb semantically and syntactically predicts the complement, which is registered in the corresponding government pattern (in the dictionary). In the right-hand members the verbs do not have this specific government, so that the nominais and adjectivais are by no means complements, or SSynt-actants. (Nevertheless, these verbs must be marked in the dictionary for their capacity to take a secon-

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

103

dary predicate: they have the syntactic feature "scopr". It is also possible that such verbs could be specified by their semantic properties; they seem to be mostly verbs of motion, position, or existence, including appearance, but the positive evidence is insufficient to make a definitive statement.) Second, from the pronominal-subjective-copredicative SSyntRel: (2)

a. They all

returned heroes vs.

They returned

_ heroes.

b. They

vs. They

 They both They died

died in the garden vs. -beggars.

There are two main differences between pronominal and non-pronominal subjective-copredicative phrases: (i) A non-pronominal phrase is possible only with certain verbs (as indicated by the feature "scopr" in their dictionary entries), while a pronominal phrase, called also a floating quantifier, is possible with any verb. (ii) A non-pronominal phrase can only follow the main verb, while a pronominal one can only precede it, except for the pronouns of the -self-series, which may both precede and follow. More precisely, a pronominal subjectivecopredicative phrase must precede the lexical part of the main verb, but fol­ lows the auxiliary or the modal finite verb. Non-pronominal [= Z] and pronominal [= Y] subjective-copredicative phrases can co-occur with the same verb: (3)

These three clauses, as we have seen above, all[Y] appear[X] embed­ ded-[z].

Third, from the modificative-adverbial SSyntRel: (4)

a. The brothers

, with eyes closed, as if

drowsing vs. The brothers b. In 1867, he and respectable.

and respectable vs. In 1867, he

SURFACE SYNTAX OF ENGLISH

104

There are three obvious formal distinctions translating a semantic difference: (i) The modificative adverbial is set off prosodically (with a comma in writing) while the secondary predicate forms a prosodic unity with the verb, (ii) A modificative adverbial, but not a secondary predicate, can be preposed to the grammatical subject: (5) a. Dead silent, with eyes closed, as if drowsing, the brothers sat in front of him. b. *Dead silent the brothers sat [cf. (4a)]. (iii) A modificative adverbial characterizes the grammatical subject only, outside its action or state, but a secondary predicate bears semantically on both the subject and its action. Thus, in (4b) the second sentence implies that before coming back he was not rich and respectable, while the first one is noncommital with respect to this. Fourth, from the adverbial SSyntRel: (6)

He

vs. He

The manner adverbial characterizes the action only but not the subject (being thus opposed to the modificative adverbial); so the first sentence, but not the second, can be continued as follows: (7)

a. He walked along happily, but he wasn't really happy. b. He walked along happy, but he wasn't really happy.

(7b) is semantically odd since it contains a contradiction, while in (7a) no contradiction is felt. (For more see Nichols 1978: 118 ff.) The dependent member of the subjective-copredicative SSyntRel can be one of the following two items: 1) A noun: (8)

They parted

2) An adjective or a passive (= past) participle: (9)

a. I cannot work

hungry.

b. He was born c. The voiceless stop may also occur d. The trilobite was fossilized

curled up.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

105

6. The pronominal-subjective-copredicative SSyntRel [pron-subj-copr] con­ nects a floating quantifier, i.e., a pronoun of the type EACH, ALL, etc., or a reflexive pronoun to the main verb : (1)

a. These phonemes each

have one allophone

b. These sentences are

all ambiguous.

c. You must 7. The objective-copredicative SSyntRel [obj-copr] connects the secondary predicate that characterizes semantically the direct object to the main verb. The objective-copredicative SSyntRel should be distinguished from the close­ ly related 2nd completive SSyntRel (with verbs of the type CONSIDER, LIKE, etc.): (1)

a. b.

['Frank wishes that Mary become/be fat, without necessarily liking Mary herself' vs. 'Frank likes Mary when she is fat'].  Ma Blum

be diluted vs. Ma Blum

['Ma Blum hates for somebody to dilute gin, although she probably does not drink at all' vs. 'Ma Blum hates to drink diluted gin']. Examples (lb—c) are from Borkin 1973. The contrast is of the same nature as in the case of the subjective-copredicative vs. the 1st completive SSyntRel: in the left-hand members of the pairs quoted, the verb (strongly) governs the dependent, i.e. predicts it semantically and syn­ tactically, but not in the right-hand members, where the dependents are not SSynt-actants of the corresponding verbs. It remains a moot question whether an objective secondary predicate is possible with any transitive verb or with some particular verbs only. Further­ more, we do not know whether, if the second alternative is true, one can still specify all such verbs semantically. Therefore we propose to mark all the verbs

106

SURFACE SYNTAX OF ENGLISH

admitting objective secondary predicates in the dictionary with the syntactic feature "ocopr". It cannot do any harm; at worst, this feature may prove redun­ dant. The dependent member of the objective-copredicative SSyntRel can be an adjective or a past participle only: (2)

a b. 

The objective-copredicative SSyntRel is also used to describe the phrases built by a transitive verb meaning '(to) cause a change of state' and an adjective denoting the resulting state: (3)

pound it flat, push it open, shoot him dead, laugh oneself sick, . . .

Clearly, the semantic interpretation of the objective-copredicative SSyntRel is different in (2) and (3): drink it cold means 'drink it IT BEING cold', while wash it clean means 'wash it SO THAT IT BECOMES clean'. Yet we feel that this difference depends on the meaning of both the verb and the adjectival, and not on syntax; cf. They burned her alive vs. *They burned her dead, or else They beat him unconscious ['. . . so that he became . . .'] vs. They beat him day and night, they beat him conscious and unconscious [' . ..he being . . .']. Therefore we can use the same SSyntRel. It is interesting to note that the grammatical subject as a syntactically privile­ ged sentence element admits a wider variety of secondary predicates than the direct object does. All of the other nominal sentence elements do not take secondary predicates at all. — Comparative SSyntRel. 8. The comparative SSyntRel [compar] connects a phrase introduced by the comparative conjunction THAN or AS1 with its governor, or comparative item. The dependent member of the comparative SSyntRel being invariably a than/ as-phrase, the governing member can be one of the following three items:

107

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

1) A comparative degree (of an adjective or an adverb): (1)

a. This case can no grammars can handle surface structures. b. He is a

CKP type

I have ever known.

. Children

10 are not eligible.

d. The latest computers calculate about 20 times faster the earliest ones. 2) A specific word capable of governing THAN: (2)

They contain no expression form on their lexical expression list.

other

what appears

3) A specific word capable of governing AS1 : (3)

a. They shot everybody

1937 in Moscow.

7

[The first as in as . . . as is AS .] b. We should look for rical evidence. c.

might be supported by empi­

implication, conjunction, etc. are extensively used.

d. But the alligator turned out to be

it is ugly.

In the surface-syntactic representation of sentences containing an instance of the comparative SSyntRel the following semantic contrast is possible: (4)

a. Mary likes Joan more than (as much as) he vs.

b. Mary likes Joan more than (as much as) him. This contrast is described at the SSyntRel-level in terms of two different con­ junctional SSyntRel's, cf. below (the predicative-conjunctional and the completi­ ve-conjunctional, p. 128).

108

SURFACE SYNTAX OF ENGLISH 2. Adverbial SSyntRel's.

9. The adverbial SSyntRel [adverb] connects a circumstantiative modifier (manner, time, place, cause, purpose, . . .) with the verb or the deverbative noun. A circumstantiative modifier can be one of the following nine items: 1) An adverb (including idiomatic multilexemic adverbiais and verbal adjuncts):

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

c. They d. The century.

109

helping him. a standard retirement age goes back to the 19th

4) A present participle phrase: (4)

a.

b.

proceed in the following manner.

(4a—b) show the case of the so-called unattached, or unrelated, participle, the structure traditionally judged as 'fault of style'. Yet in scientific literature and journalistic prose its use has become almost acceptable, especially when the semantic 'subject' of the participle is indefinite ONE. c.

to kiss her goodbye.

d.

few social engagements.

e. Lees,

recourse to statistics.

ƒ The children,

the table.

On the syntactic homonymy of the adverbial vs. the modificative-adverbial construction with the present participle see below, page 113. 5) A to -infinitive phrase, with or without a conjunction (IN ORDER, SO AS,...)

(5) a. the following technique. b. I will

not to forget it.

SURFACE SYNTAX OF ENGLISH

no

c.  left the country, never to return again. d. He awoke at night In a terrible fright To find this was perfectly true. We see in (5) two basic types of infinitive circumstantials: goal and subsequent action. 6) An absolute construction, either an absolute-predicative or an absolute-con­ junctional: — Absolute-predicative: (6)

a.

becoming difficult for the British housewife to make both ends meet.

b. Extraposition is a dimension orthogonal to factivity, the latter —** ^being a highly redundant feature. — Absolute-conjunctional: (7)

a. The catalysts

in solution.

b.

the Tara brooch in shape.



-pulled, and contracts

7) A dependent clause introduced by a conjunction, or containing a w/z-word, or conjunctionless (= conditional): — With a conjunction: (8)

a.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

b.

111

a protégé of Prime Minister David Ben-Gurion.

c. This not be a simple one.

the experiment will

— With a wh-word: (9)

a. However this constraint constraint is at work in the last case, too.

__

b. These relationships be treated mechanically.

that the same

to

— Conjunctionless: (10)

The customers dissatisfied with the machine output.

8) A coordinate conjunction introducing an independent clause: (11)

just where the trouble lies.

Such a conjunction, although it fully keeps its semantic force, does not mark syntactic coordination. 9) An idiomatic construction of the form X or no X: (12)

got to finish my work.

The next three SSyntRel's (10 through 12) form a group: they are semantically and syntactically related to each other and to the adverbial SSyntRel. All three describe a specific circumstantial which depends syntactically on the main verb (as all adverbiais do) but bears semantically on the grammatical subject. These are the modificative-adverbial, the appositive-adverbial, and the attributi­ ve-adverbial SSyntRel's, paralleling, respectively, the modificative, the appositive, and the attributive SSyntRel's (see pages 134, 144 and 148).

112

SURFACE SYNTAX OF ENGLISH

10. The modificative-adverbial SSyntRel [modif-adv] connects an adjectival phrase (including a single adjective) to the main verb, the head of the correspon­ ding clause, such that this phrase qualifies the grammatical subject of the clause and serves, at the same time, as an adverbial of contingency, cause, concession, etc. The modificative-adverbial SSyntRel contrasts semantically with the ad­ verbial SSyntRel: (1)

a.

over to stop in front of me. vs.

b. Jim

as always, to stop in front of me.

(1a) says that Jim was elegant, not his walk, while (1b) characterizes the manner Jim walks rather than Jim himself. Furthermore, the modificative-adverbial SSyntRel contrasts with the modi­ ficative and the descriptive-modificative SSyntRel's (pp. 134ff. and 137ff.): (2)

a.

bring about many personal conflicts. vs.

b. This

man will bring about many personal conflicts.

vs. c. This man, personal conflicts.

unsatisfied and irritated, will bring about many

Technically speaking, a modificative adverbial can either precede or follow the subject NP, but it is not part of it: it depends syntacticallyon the verb. How­ ever, both restrictive and descriptive modifiers are part of the NP whose noun they modify (not necessarily the subject NP). A restrictive modifier can precede or follow its noun (as a function of different factors that need not be considered here), and if it modifies a subject, it also can be disconnected from it and moved to the right of the VP. But a descriptive modifier can only follow its noun and can never be disconnected from it. Thus in a sentence such as So Joan lived on, even more beautiful than ever before the adjectival phrase is a modificative ad-

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

113

verbial, not a disconnected descriptive modifier; it can be preposed as well without affecting the grammaticality of the sentence: So, even more beautiful than ever before, Joan lived on. The dependent member of the modificative-adverbial SSyntRel can be one of the following two items: 1) An adjective: (3)

a.

herself with a glare of defiance.

b. I :o see them again. 2) A participle, present or past: (4)

a.

as two electron-opaque layers.

b. c. d. Inside, you

his enormous pistol. not scared to compete again. only by peace and quiet.

Note that in our description, a present participial phrase preceding the gram­ matical subject may be syntactically ambiguous and must then be interpreted in one of the following two ways: either as an adverbial which qualifies semantically the main verb of the sentence, or as a modificative adverbial qualifying the grammatical subject. Cf. (5): (5) Brandishing his weapon and cursing loudly, the sailor rushed through the crowd. The italicized participial phrase in (5) depends syntactically on rushed, but it can be said to depend on rush either via the adverbial SSyntRel (and to character­ ize the action: rushed in what way?) or via the modificative-adverbial SSyntRel (and then to characterize the actor: the sailor). True, there is in practice almost NO SEMANTIC DIFFERENCE between these two surface-syntactic inter­ pretations.

SURFACE SYNTAX OF ENGLISH

114

The aforementioned ambiguity can be resolved within the context of the sen­ tence in at least the following two instances: (I) The preposed participial phrase should be identified as an adverbial: — If the participial phrase is separated from the grammatical subject by a finite verb (which, then, precedes the subject): (6)

Leafing hastily through my diary, there was John's wife Linda.

— If the present participle is conjoined to an adverb or a prepositional phrase: (7)

Still crying, but without making any noise, she went through the contents of the next drawer.

— If the present participle is a stative verb which cannot have a progressive form: (8)

Foreseeing the answer, my parents became nervous.

(II) The preposed participial phrase should be identified as a modificative ad­ verbial (qualifying the subject) if the present participle is conjoined to an ad­ jective or a past participle: (9)

Stumbling, wind-blown, and winded and wet to the bone, I managed a fast job of assembling her belongings [T. Capote].

Should the formal contextual clues be missing, in the course of analysis the corresponding syntagms (9.6 and 10.1) will assign the construction in question both possible SSynt-structures. 11. The appositive-adverbial SSyntRel [appos-adv] connects a preposed noun phrase to the main verb, the head of the corresponding clause, such that this phrase qualifies the grammatical subject of the clause and serves, at the same time, as an adverbial of cause, or concession, etc. The appositive-adverbial SSyntRel contrasts semantically with the adverbial SSyntRel: (1)

a. An utter

to our pressing for more information.

vs. b. He objected

cowardly to our pressing for more information.

(la) qualifies him as a person, (lb) — his way of objecting.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

115

It contrasts also with the appositive and the descriptive-appositive SSyntRel's: (2)

a.

not able to participate in the colloquium ['being the director . . .']. vs.

b. Van is not able to participate in the colloquium ['which Van Roggen?' — there are probably two or more people with this name]. vs. c. Van of the Institute, is not able to participate in the colloquium ['V.R., who is the director, . . .]. The dependent member of the appositive-adverbial SSyntRel can be only a noun phrase: (3)

An inveterate

up late.

12. The attributive-adverbial SSyntRel [attr-adv] connects a non-adjectival modifier, i.e. an attribute, to the main verb, such that this attribute phrase cha­ racterizes the grammatical subject of the clause and serves, at the same time, as an adverbial of contingency, cause, concession, etc. The dependent member of the attributive-adverbial SSyntRel can be one of the following four items: 1) A prepositional phrase: (1)

through the sleeping town.

2) An adverb : (2)

the eternal problem of life. 3

3) An as-phrase [= AS ]: (3)

sent to the other shore.

116

SURFACE SYNTAX OF ENGLISH

4) A to-infinitive phrase: (4)

to me the major reason of my depression.

In connection with the last three SSyntRel's, the following three remarks could be made: 1. All these SSyntRel's are "subject-oriented". That is, their dependent member is semantically related to the grammatical subject, to the exclusion of any other sentence element. This fact emphasizes once again the syntactically privileged status of the grammatical subject. (Two exceptions to this general rule are indicated in Syntagm 10.1, on p. 289.) 2. The dependent member of these three SSyntRel's is, more or less obliga­ torily, set off by heavy pauses and a specific prosodic contour (by commas in writing). It is in a sense detached from the rest of the clause, thus creating a particular type of sentence element: a disjunctive sentence element. 3. The modificative adverbial can be subordinated not only to the main verb of the sentence but to an infinitive as well: see example (12) on page 289. On the other hand, the appositive and the attributive adverbial can be subordinated to the main verb only. Syntagms 10—12 provide for this difference. 13. The parenthetical SSyntRel [parenth] connects a parenthetical phrase to the head of the expression (the whole sentence included) that contains it. Note the important semantic contrast between the parenthetical and the ad­ verbial SSyntRel's: (1)

a. He time. b.

me on

over my shoulder vs. over my shoulder.

Examples (1) are from Connors 1971. The dependent member of the parenthetical SSyntRel can be one of the fol­ lowing five items: 1) A special parenthetical adverb or a phraseological adverbial:

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

(2)

117

be defined in terms of phonemics. b.

not the case.

c. A particular paper in, on his investigations.

a Chinese journal bears

d. e.

2) A phrase of a particular type (Adv in -ly + enough, Adj + to +V inf , Adv in -ly + Vppres, to + Vinf + with, to + vinf + NP, as + Vpass [= AS2]): (3)

a.

a different view.

b.

not explored all the possibilities in detail.

c. d

.

t

h

i

r

e.

d

analysis. -ly adverbs in a count

3) A comment clause, introduced by AS2 or conjunctionless: (4)

a. This analogy has, discussion.

he admits, no relevance to the present

b. c. To find the appropriate rules aims a linguist can have.

been developed. one of the most worthy

SURFACE SYNTAX OF ENGLISH

118

4) An explicative what-clause : (5)

It is true for all the PS-grammars the existing T-grammars.

important, for

5) A full-fledged sentence set off by pauses and specific prosody (in writing — by parentheses or dashes): (6)

a. They never freely with almost any element). b. But the toxins away.

occur quite

few and poor — has sued to keep

14. The adjunctive SSyntRel [adjunct] connects an interjection, a sentential adverb or a direct address to the main verb as the top node of the sentence:

(D

a. b. 

feel bad about my not coming [J. Updike].

B. Phrasal SSRel's (15-39). 1. General. 15. The restrictive SSyntRel [restr] connects to (the head of) an expression a phrase which restricts, distinguishes, intensifies/attenuates it or characterizes it with respect to its extent. In most cases the restrictive phrase immediately precedes its governor; the exceptions are a few quite specific words, such as NOT, ONLY or EVEN. The dependent member of the restrictive SSyntRel can be one of the following five items: 1) A special restrictive adverb (or particle) marked as such in the dictionary (the syntactic feature "delim")· Negations, for instance, belong here:

119

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

(1)

a. The distributives b.

cooccur with the other quantifiers. are important, but the method used.

c. They make a new translation one. d. She came right e. immediately

as they revise a poor

when I was in the midst of everything. after the play;entirely

in the wrong;

2) An adverb of degree or extent modifying an adjective or another adverb: (2) Such an adverb can in principle modify a noun as the dependent of the com­ positive SSyntRel: (3)

purely

Some of the degree adverbs can modify comparatives, superlatives or numerals; this property is marked by the syntactic features "adcompar", "adsuperl" and "adnum": (4)

a.

intelligent in the class;

b. 3) An adjective modifying another adjective: (5)

burning

4) A quantity phrase modifying a comparative or a specific preposition: (6)

a. large as mine; b. five years

our meeting;

the hill.

120

SURFACE SYNTAX OF ENGLISH

5) A coordinate or a quasi-coordinate conjunction introducing a detached sen­ tence element:

(7)

This definition is found over and over on logic.

a g a i n , w o r k s

2. Verb phrase. 16. The auxiliary SSyntRel [auxil] connects a non-finite lexical verb to an auxiliary verb (marked "aux" in the dictionary) within analytical (= complex) verbal forms: the passive voice, the progressive, the perfect, the future tense, the conditional and the subjunctive moods, the emphatic; cf., e.g.: (1)

a. The meaning tion. b. It 72.

by the following transla­ to look at a diagram found in Jespersen 1938:

c. It must be repeated for each prefix that the first character of the input word. Any two SSynt-nodes connected by the auxiliary SSyntRel are represented in the DSyntS by a single node. That is, at the level of deep syntax an analytical form contracts into one deep-morphological representation of the correspon­ ding lexeme. For instance, the sentence (2): (2)

He did kill her.

is represented at the deep-syntactic and the surface-syntactic levels, respectively, as (2'):

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

121

3. Word-like phrases (17-20). The following three junctive SSyntRel's represent 'split words' — units which are syntactically complex, i.e., phrases, but which semantically constitute Simplexes. Such a unit is assigned a single node in the DSyntS in much the same manner as an analytical verb form. 17. The phrasal-junctive SSyntRel [phr-junct] connects a verbal adjunct of the type UP, OFF, DOWN, etc. to a verb, to form a phrasal verb: (1)

a. He ate

the whole stock.

b. Shut  The poor guy brought

his breakfast.

d. Don't give The verb + adjunct phrases, or phrasal verbs, can show a wide range of semantic 'cohesion', from almost free combinations to quite frozen (= idiomatic) ones: — The verb and the adjunct may retain (more or less) their individual meanings, like bring in/out, switch on/off [electricity], etc. — The verb alone may keep its meaning while the adjunct has an 'intensifying' function, like in find out 'discover', sweep up [the crumbs], etc.

122

SURFACE SYNTAX OF ENGLISH

— The verb and the adjunct may be fused into a new unit which shares no nontrival semantic components with its constituents, like put up 'tolerate', catch on 'understand', back up 'support', put across [an idea] 'make understand', etc. This, however, is not very relevant at the level of surface syntax: from the strictly syntactic viewpoint (word order, stress, etc.) all phrasal verbs behave more or less in the same way, independently of their idiomatic nature. Therefore we can describe them all in terms of only one phrasal-junctive SSyntRel. On the other hand, the said idiomaticity helps us to single out phrasal verbs phrases) as opposed to free combinations of verbs with adverbs. We can indicate here two rough criteria by which we judge a verb + ad­ junct phrase to be a.phrasal verb, representable by the phrasal-junctive SSyntRel. One of them is semantic, and the other syntactic: (i) A phrasal verb is semantically idiomatic, even if only slightly so; its meaning cannot be automatically computed from the meanings of its components. (All phrasal verbs must, then, be listed in the dictionary.) (ii) A phrasal verb does not admit the inversion of its components: (2)

a. *Up he gave! vs. Up he climbed! b. *In he broke on our conversation vs. In came the funniest-looking guyc. *Off it took! vs. Off.they ran!

Free-formed phrases of the type considered are subsumed under the adverbial SSyntRel: bring In addition to verb + ad­ junct phrases, the phrasal-junctive SSyntRel is also used to connect to the verb certain other elements, such as adjectives or nouns, which form with it an idiomatic phrase: An exhaustive and insightful analysis of syntactical and semantic properties of English phrasal verbs is found in Bolinger 1971 and Fraser 1976. As for the constructions of the verb + adjective type, such as bang shut or paint blue, which are often considered along with phrasal verbs, they are dealt with in this book under the objective-copredicative SSyntRel; see page 106 for explanations.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

123

18. The numeral-junctive SSyntRel [num-junct] connects any component of a compound numeral to its governor — the following component of the same numeral. The top node of a compound numeral is taken to be the rightmost integer number in it. The dependent member of the numeral-junctive SSyntRel can be one of the following four items: 1) A numeral: (1) 2) The lexeme AND: (2)

one hundred

3) The lexeme A (the indefinite article): (3)

pesetas.

4) The name of a fraction (a nominalized ordinal numeral): (4)

three

19. The binary-junctive SSyntRel [bin-junct] connects one part of a binary idiomatic word combination to the other part. The following three cases are provided for: 1) A paired conjunction, such as if. . . then, either... or or both . . . and. The direction of the dependence is so selected as to make the autonomous part of the conjunction (i.e., the part that can appear alone: if, or, and) be the head: 2) The construction with the . . . the followed by comparatives. Here, the head is the the [= THE3] that introduces the main clause:

(1)

the need for syn­ tactic analysis will be felt.

[The second the is THE2.]

124

SURFACE SYNTAX OF ENGLISH

3) The paired preposition from . . . to, with from the head: (2)

The color varies

20. The colligative SSyntRel [collig] connects the so-called STRANDED preposi­ tion to its governor — usually a verb but sometimes a semiidiomatic verbal phra­ se. It is in a kind of passive construction (= pseudopassive) where the dependent of the preposition becomes the grammatical subject that the colligative SSyntRel appears: it replaces the i-th completive SSyntRel that subordinates the preposi­ tion and its noun in the active construction. Compare:

(1) For instance: (2)

a. He deals with this problem in Chapter X => This problem is with by him in Chapter X. b. By the Burmese they are referred  These cigarettes go out in five minutes if not puffed

The colligative SSyntRel is necessary since the completive SSyntRel that sub­ ordinates the preposition in question in the active construction cannot be re­ tained in the passive: it is, as it were, converted into the predicative SSyntRel (as shown in diagram (1)). Let it be pointed out that a stranded preposition — the dependent member of the colligative SSyntRel — should not be confused with a dangling preposition, which keeps (at least in principle) its complement, the latter having been moved to the left or even deleted on the surface (but then easily recoverable) ; dangling prepositions are italicized in (3):

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

(3)

125

a. b. All the accountants

receivables are piling up.

c. It is the wrong place A dangling preposition, unlike a stranded one, retains the original SSyntRel that subordinates it (one of the completive SSyntRel's or the adverbial SSyntRel). For more about dangling prepositions see pages 131 and 365ff.

4. Expletive phrases. 21. The expletive SSyntRel [explet] connects a clause or a verbal or nominal phrase to an anticipatory or clefting it [= IT5 or IT6] or to an anticipatory THERE2. Thus the governing member of the expletive SSyntRel can be one of the following three items: anticipatory IT5 in the role of a dummy subject or dummy direct object; clefting (or emphatic) IT6 in the role of a dummy subject; and anticipatory THERE2 in the role of a dummy subject or dummy preposi­ tional dependent (with for). The dependent member of the expletive SSyntRel can be one of the following seven items: 1) A gerund phrase (with IT5 only): (1)

a. changed his mind all of a sudden.

b. I thought 2) A to-infinitive phrase (with IT5 only) : (2)

a.

have agreed to his proposal.

b. The fact is supposed to make neutralize the conse­ quence. 3) A dependent clause introduced by THAT5 , WHETHER1 or IF 2 (with IT5 only):

126

SURFACE SYNTAX OF ENGLISH

(3)

a.

tape services emanating from Europe will contain a wealth of otherwise inaccessible information.

b.

the embedded structure has the sam< or different subject.

4) A dependent clause or a to-infinitive phrase containing a wh-word (with IT5 only): (4)

a.

the country.

b.

go in such an emergency.

5) A dependent clause introduced by a temporal conjunction (with IT5 only): (5)

he began working here.

6) A dependent clause introduced by THAT5 or containing a wh-word (with IT6 only): (6)

a.

the computer has importance.

b.

I told him the news.

7) A noun phrase (with THERE2 only): (7)

a. b.

5 6

the steamer could not leave the port. 2

IT " and THERE , as semantically empty items, are not represented in the deep-syntactic structure. Correspondingly, the expletive SSyntRel disappears at the deep-syntactic level. Thus (la) becomes roughly (la '): (la')

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

127

(4b) becomes (4b'): (4b') and so forth. See Morgan 1968 for a discussion of the status of anticipatory IT and for pos­ sible representations of expletive constructions.

5. Conjunctional phrases (22—26). The next five SSyntRel's all describe the functioning of a conjunction as the head of the corresponding phrase (hence the term "conjunctional"). 22. The subordinate-conjunctional SSyntRel [sub-conjunct] connects the head of the expression introduced by a subordinate conjunction to this very conjunction. The governing member of the subordinate-conjunctional SSyntRel being in­ variably a subordinate conjunction, its dependent member can be one of the fol­ lowing three items: 1) A tensed verb (as the head of a clause): (1)

a. Suppose b.

by, the local peasants were secretly burying four American women.

2) A to -infinitive (with certain conjunctions only): (2)

So far, the Western allies had not produced a coordinated response, irritate the Russians.

3) Another subordinate conjunction (THAT5 or WHEN only), which forms with the governor a compound conjunction: (3)

They always agree, except

when they discuss jazz.

23. The coordinate-conjunctional SSyntRel [coord-conjunct] connects any sentence element — a word, a phrase, or a clause — to a coordinate, quasi-co­ ordinate or comparative conjunction that heads the corresponding expression:

128

SURFACE SYNTAX OF ENGLISH

(1)

a. These elements are paradigmatically established and —►stated as values in the structure. b. As a convenience of presentation, and in order to avoid repetitious statements, these patterns are analyzed in a special section. c.  neither accomplished his purpose, nor else be able to do it now. d. The plural is predictable by the same rules as inflection of verbs.

will anyone

for the

e. Why should we be envious if they can live a better There is an important exception: with a comparative conjunction as the governor, the coordinate-conjunctional SSyntRel is not used if the dependent member is a personal pronoun. For personal pronouns, the predicative-con­ junctional and the completive-conjunctional SSyntRel's are reserved, see im­ mediately below. 24. The predicative-conjunctional SSyntRel [pred-conjunct] connects the second term of a comparison, i.e. what an item is being compared with, to the comparative conjunction — if this term is correlative with the grammatical sub­ ject and if it is implemented by a personal pronoun (in any grammatical case, since the oblique [= objective] form is also capable of appearing as a correlate of the grammatical subject; cf. (lb)): (1)

a. But his son must live a better life than b. And what if we find out that they are better than

25. The completive-conjunctional SSyntRel [compl-conjunct] connects the second term of a comparison to the comparative conjunction — if this term is correlative with the direct object and if it is implemented by a personal pronoun in the oblique case : (1)

I like Dick more than

The nominative [= subjective] case is excluded in this role.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

129

26. The absolute-conjunctional SSyntRel [abs-conjunct] connects (the head of) a phrase that constitutes an absolute construction to the conjunction intro­ ducing it. The dependent member of the absolute-conjunctional SSyntRel can be one of the following five items: 1) A noun phrase: (1)

the subject can be marked by the ergative.

2) A prepositional phrase: (2)

While

on a flight, you shouldn't . . .

3) A participial phrase: (3)

a. A yogi who visited Yale, when turned pale [A. Zholkovsky].

offered an armchair,

b. If traveling by air, do not forget to bring along your Pan-Am Card. 4) An adjective: (4)

Although

smaller, it resembles the Tara brooch in shape.

5) An adverb: (5)

Once

there, he said: . . .

In conformity with the passive valence of its head (= a subordinate conjunc­ tion), the absolute-conjunctional construction appears always in the SSynt-role of a circumstantial, exactly as do full subordinate clauses introduced by the same conjunctions. Actually, this construction represents a reduced clause, with the subject — a pronoun — and the main verb — BE — omitted; cf. (1) vs. (1 '): (1 ')

If IT IS a personal pronoun, the subject can be marked by the erga­ tive. 6. Noun phrase (27-39).

27. The prepositional SSyntRel [preposit] covers all the prepositional phra­ ses. Its governing member is invariably a preposition or the particle TO2 intro-

SURFACE SYNTAX OF ENGLISH

130

ducing an infinitive (the latter being considered as a special case of prepositional structure). The dependent member of the prepositional SSyntRel can be one of the following ten items : 1) A noun phrase or any of its syntactic equivalents: (1)

a. The subject consists b. I would like to review the grounds somewhat more convincing manner. c. . . . the growing number of documents available them anywhere in the world.

need

d. Reliability decreases to Two special cases deserve mention, (i) If the dependent of the preposition is a wh-word, a noun modified by a whword or THAT4 [= 'which'], the preposition can be and often is moved to the right end of the clause introduced by this wh-word or THAT4 : (2)

a. An interesting field

must also be

mentioned. b. Which c. Let's take the following assumption Since WHICH or THAT4 can be omitted altogether the right-shifted preposition may remain with no surface dependent. This is one of the situations where the non-saturation of some strong active surface-syntactic valence slots is admitted: (3)

a. The MTA property was land the President has walked upon, smiled upon. b. The lake I swam across so many times will now be destroyed.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

131

The dependent of a preposition can also be moved to the front of a sentence for the purpose of emphasis: (4)

In addition, the dependent of a preposition is obligatorily deleted in some in­ finitival attributive constructions such as the following one: (5)

You've chosen a bad place to be lame in [R. Kipling, Rikki-Tikki-Tavi].

A preposition whose dependent is either positioned to the left of it or deleted by a surface-syntactic rule is called DANGLING. Dangling prepositions must be carefully distinguished from stranded prepositions, whose dependents have been converted into grammatical subjects; compare above the colligative SSyntRel (page 124). The difference is that a dangling preposition does not change the SSyntRel r which subordinates it to its governor, while with a stranded preposi­ tion this SSyntRel r turns always into the colligative SSyntRel. (ii) Some English prepositions can be or always are postposed to their depend­ ents: (6)

notwithstanding.

2) An adjective (with a few lexemes only): (7)

a. The color changes from b. temperatures above

3) A gerund: (7)

It gave a real stimulus to

4) A clause introduced by a subordinate conjunction or a relative (= wh-)word: (8)

a. This distinction has caused a great amount of confusion logical or syntactic. b. The iota operator is like every other linguistic device in to the situation.

132

SURFACE SYNTAX OF ENGLISH

. . . . to account

understood.

5) An infinitive phrase introduced by a relative (= wh-)word: (9) We are concerned

good this damage.

6) An adverb: (10) before 7) Another preposition: (11) a. . . . except b. since

inside the palace.

8) A numeral (in a sequential phrase, with TO1 or THROUGH only): (12) Chapters 3 through 9) Anticipatory THERE2(for-or-to-constructions and before a gerund): (13) a. language. b. We are interested in all the possibilities.

for calculating

10) An infinitive (if the main member is TO 2 , EXCEPT or BESIDES): (14) a. Rather than try b. The editors have been encouraged to experts in many fields.  I cannot do anything, except

such data, we will . . . consult widely with you some money.

28. The determinative SSyntRel [determ] connects a determiner to the deter­ mined noun or its syntactic equivalent (a numeral, an adjective, ONE2, etc.). Its dependent can be either an article or a pronominal adjective marked "det":

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

133

Since the dependent member of the determinative SSyntRel belongs to a closed and very small class, this SSyntRel could be deemed redundant. It would be logically feasible, for example, to use only one general modificative SSyntRel covering ALL possible modifiers; simply, sometimes we would have to specify the subclass of the modifier. However, informal considerations of greater natu­ ralness and better surveyability prompt us to postulate the determinative SSyntRel as a separate SSyntRel in its own right : (i) The determinative SSyntRel is non-repeatable (a noun cannot have more than one determiner) while the general modificative SSyntRel is in principle repeatable: there can be several assorted modifiers with the same noun, (ii) The presence/absence of a determiner controls the syntactic behavior of a noun, so that the determinative SSyntRel turns out to be an important element of tree context when formulating various surface-syntactic rules (unlike the general modificative SSRel). (iii) The determiner, in contrast to general modifiers, occupies a strictly specified linear position with respect to its noun (normally, it is the leftmost element in the string of noun premodifiers). Furthermore, Criterion  II for types of SSyntRel's (the 'Kunze property', page 70; and Mel'cuk 1979: 115ff., especially, 120-121) also requires that the determinative SSyntRel be set off as a further SSyntRel among all the SSyntRel's of English. In keeping with this, we have included it in the inventory of English SSyntRel's. 29. The quantitative SSyntRel [quant] connects a numeral or its substitute (a formulaic designation of a number, a quantitative pronominal lexeme) to the noun quantified :

SURFACE SYNTAX OF ENGLISH

134

In much the same way as the determinative SSyntRel the quantitative SSyntRel is redundant but the same three reasons (non-repeatability, a specific syntactic role, and an established linear position) argue for its admission into the inventory of English SSyntRel's. 30. The modificative SSyntRel [modif] connects an adjective or any adjec­ tive-like modifier to its modified noun or a syntactic equivalent thereof (a nominalized numeral, the pronoun ONE2). The dependent member of the modificative SSyntRel can be one of the fol­ lowing four items: 1) An adjective, preposed or postposed: — Preposed

— Postposed (2)

a. These etymologies explain the plethora of verbal anomalous in philological b. They accept something c. There is something

than good

translations.

amiss with this method of analysis.

Preposed adjectival modifiers also include the pronominal adjectives that can cooccur with articles (those which replace articles are connected to the noun by the determinative SSyntRel):

(3)

a.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

135

2) A participle, preposed or postposed: — Preposed ( 4 ) T h e m o s t c o n s e q u e n c e of accepting the proposed characterization concerns isolated — Postposed (5)

a. The invitations are being sent to all sides agreement. b. A summary of the phrases c. The l o g i c a l - m o r e or less important part of the papers in the field.

3) A restrictive relative clause belonging to one of the following three types: — introduced by a WH-word; — introduced by a temporal conjunction; — with no surface introductory element (i.e., with THAT4 deleted by syntac­ tic rules). In this case, the dependent member of the modificative SSyntRel is the top node of the relative clause: i.e., its main verb in the first and third types, the con­ junction in the second. — A relative clause introduced by a w/z-word (6)

a. The idea of constructing artificial tic to consider translating into by computer had been suggested. b.

not show this inflection are excluded.

 Some associated with different suf­ fix-sets. — A relative clause introduced by a temporal conjunction (7)

The day

she saw Jack she was filled with anxiety.

SURFACE SYNTAX OF ENGLISH

136

— A relative clause with no overt introductory element (8)

a. It is one of the worthy b. The listener will say of it is false.

as wildly improbable that

 The best-known positional Calculus.

to is the Pro-

4) An infinitival phrase, i.e., a to-infinitive introduced by a wh-word or with a wh-word deleted by surface rules: (9)

a. I have no b. I have no one

speak to.

Remark. As Teyssier 1968 points out, the preposing or postposing of a sin­ gle adjectival in English can differentiate meaning: (10) a. an insane man ['a madman'] vs. a man insane ['a man who behaves like a madman'], b. a sitting figure [as in a picture] vs. a figure sitting ['a figure that is actually in the sitting position'];  a written word ['a word which is written'] vs. a word written ['a word when being written']. According to J. Teyssier, a preposed adjectival denotes a more constant, more ordinary or inherent property, while a postposed one implies rather a temporal quality that is perceived at a particular moment (i.e., the postposed adjectival is more like a circumstantial). However, we do not know whether this phenomenon is sufficiently regular, i.e., first, whether adjectivais that change their meaning in accordance with their linear position are numerous enough, and second, whether the shift in meaning is quite predictable (e.g., always something like 'currently'). If this is the case, then two different SSyntRel's should be used to represent the members of each pair in (10): say, the modificative SSyntRel for the preposed

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

137

adjectival and a different SSyntRel for the postposed one (it could probably be one of the SSyntRel's already in our inventory, for example, the attributive SSyntRel — or a new one, to be added to the inventory). Since we lack positive evidence, we abstain from making a decision. 31. The descriptive-modiflcative SSyntRel [descr-modif] connects any quali­ fying, or descriptive, adjectival modifier to its governor, normally a noun (more rarely, an adverb, a clause, or an adjective). The semantic contrast between a restrictive and a non-restrictive (= quali­ fying) modifier is well-known: the first one restricts the set of objects specified by the modified governor to a narrower subset, while the second adds an ad­ ditional characterization of the thing denoted without changing the scope of the governor. Cf. the following examples: (1)

a. All his papers May were sent to CL ['only those of his papers that had been written in May were sent, but not necessarily any others, written earlier or later']. vs. b. All his papers, written in May, were sent to CL ['all his papers were sent, and they happened all to be written in May'].

(2)

a. I do not like always insulting people ['. . . those soldiers who are always insulting people']. vs. b. I do not like always insulting people [' . . . soldiers in general, since they all insult people'].

This contrast is expressed at the surface by prosody in postposed modifiers: the descriptive ones are set off by longer and more pronounced pauses, as well as by a specific contour; in writing, commas are obligatory. Moreover, the re­ lative THAT4 ('which') is possible in restrictive clauses only. This proves the necessity of distinguishing two SSyntRel's: the modificative and the descriptivemodificative. Note that this distinction is inapplicable to preposed modifiers. True, the same semantic contrast is found between preposed adjectives as well. Thus the

138

SURFACE SYNTAX OF ENGLISH

phrase dogmatic grammarians can be understood both ways: 'those [= some of] grammarians who are dogmatic' or 'all grammarians, all of them being dogmatic'; in a beautiful woman the adjectival modifier is strictly restrictive while in my beautiful wife it is qualifying. But when preposed, the adjectival phrase never marks the distinction formally at the syntactic level (it must be deduced from the context), and therefore a single modificative SSyntRel is sufficient. The dependent member of the descriptive-modificative SSyntRel can be one of the following two items: 1) An adjectival phrase, i.e., an adjective or a participle with at least one depend­ ent: (1)

a.

will be discussed later.

b. We turn our attention to this rule, ment of the thematic vowel.  The attack,

governing the develop­

aimed at throwing the Iraqis out of Kuzestan

Province, was hobbled by winter weather. Notice that a single adjectival is not acceptable in this role. 2) A relative clause: (2)

a. Richens presented alternative translations to the choose among them. b. The of transformational grammar.

discuss below, concerns the utility

 The tape reader transmits the combination to the buffer on cathode tubes. Three special cases should be mentioned: (i) A relative clause modifying an adverb of the shifter type :

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

(3)

a.

139

the men tried to build a spacious barn.

b. Tomorrow, works out.

everybody is in, we will see how it

(ii) A relative clause introduced by WHICH or WHERE and modifying the whole preceding clause or sentence (i.e., subordinated to its main verb): (4)

a. b. Omission of the preposition ral. c

rather natu­

lose his temper.

(iii) A relative clause introduced by WHICH and modifying an adjective or nonreferential noun (in the capacity of a complement to BE): (5)

Of these two phrases, the first is said to be

32. The possessive SSyntRel [possess] connects a noun in the possessive form to its governor — another noun or gerund: (1)

a. He analyses the first informant's b. Mr. Macklovich informed the firm of the buyer's insured his goods.

Let it be emphasized that the name of this SSyntRel should not be taken to imply the semantic notion of 'possession' (cf. the remark on page 90). In ac­ cordance with our general practice it is no more than a mnemonic designator.

SURFACE SYNTAX OF ENGLISH

140

A well-known complication is that the possessive form marker -s (or its or­ thographic variant —') can be linearly and prosodically attached not to the 'possessed' noun but rather to the rightmost element in its phrase:

(2) a.

last summer.

b. 33. The compositive SSyntRel [compos] connects a preposed noun phrase to its governor — a noun, an adjectival (an adjective or a participle), or a verb. The result behaves much like a compound word (= compositum) ; hence the name of this SSyntRel. For a clear and thorough description of English compound nominais see Levi 1978. A noun appearing as the dependent member of the compositive SSyntRel with another noun acquires, as is well known (cf., e.g., Jespersen 1965: 94, 98), some syntactic properties of an adjective: (i) it cannot have a determiner; (ii) it can be modified by an adverb: [a] purely *- family [affair]; (iii) it can be conjoined with a genuine adjective: business -> and -* professional

We]. Therefore, a logical possibility exists to treat all such nouns as converted into adjectives and to subordinate them to their governor via the modificative SSynt­ Rel. However, this would not eliminate all the problems since the dependent noun in the compositive SSyntRel preserves too many properties of a noun as well: (i') it can be modified by an adjective , by a numeral

by another compositive noun (Avia­

tion quite unlike an adjective; (ii ) it can be subordinated to an adjective or a participle in a typical nominal role : factory sitive;

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

141

(iii) it can be in plural (although only in a few special cases):

Therefore we have prefered to stick to a more traditional view of compositive nouns (= first elements of binary nominal compounds), i.e., to consider them as nouns. The dependent member of the compositive SSyntRel can be one of the fol­ lowing four items: lì A noun, including proper nouns:

[With adjectives and participles as governors, a hyphen is obligatorily used in a compositive phrase.] 2) A sequential noun phrase: (3)

3) A conjoined phrase:

SURFACE SYNTAX OF ENGLISH

142

Two special cases deserve mention: — Compound adjectivais of the bahuvrihi type, i.e., phrases having the form (Adj + N) + ed or (N + N) + ed, such as long-winged, pear-shaped, etc. — Compound verbs, such as to brain-wash, to Chomsky-adjiin, to frequencymodulate; cf. further: (6)

He also guest

conducts around the world.

34. The elective SSyntRel [elect] connects an of-phrase (or among-phrase) to a superlative form of an adjective or to a numeral as its governor. The elective phrase is a syntactic equivalent of a noun; at the same time, it is itself elliptical: (1)

the rightmost of these strings = the rightmost STRING of these strings,

this ellipsis being virtually obligatory. The dependent member of the elective SSyntRel is invariably a prepositional phrase with OF (or, rarely, AMONG); its governor can be one of the following four items: 1) An adjective in the superlative: (2)

a. The poorest

among these nations seems to be Namibia.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

b. Everybody will try to pick out the presented.

143

the solutions

2) A cardinal numeral: (3)

a. 45

256 sentences tested contain an error.

b. . . . including the following seven assumptions.

Lakoff s fundamental

3) An ordinal adjective: (4)

and second only have been implemented.

4) A quantitative word having the feature "elect": (5) a. Not all

the difficulty of learning a new language is due to

the difference. b. He has read most c.  spends some Remark 1. In the case of an analytical (= complex) superlative, we consider the of-phrase to be dependent on most/least (cf. example (2b) above) rather than on the adjective itself. This leads to non-projectivity. To avoid it, a different analysis should be prefered: (2b') . . . the most <- elegant - of the solutions. However, we leave this question open for the moment. Remark 2. A prepositional phrase modifying a superlative adjective and de­ noting the semantic scope of its superlativity, so to speak, is considered in this book to be the attribute of the adjective:

144

SURFACE SYNTAX OF ENGLISH

(6)

See page 149, the end of item 1. 35. The appositive SSyntRel [appos] connects a postpose d noun phrase to a common noun denoting the object of which this phrase is the name or a speci­ fication. Note that this is a restrictive appositive. In the domain of naming appositives, the dependent member of the appositive SSyntRel can be one of the following four items: 1) A proper noun: (i) 2) A number, a formula, or a notation symbol: (2) 3) An expression in quotation marks or in a different typeface: (3) 4) A nominalized adjective denoting a characterization etc. (of a proper noun): (4)

The domain of specifying appositives is, by its nature, more limited. We have here only two types of appositives: 1) An expression somehow set off from the rest of the text (by quotation marks, a different type face or a different script). This expression follows a noun which is its generic name: (5)

a. the logical form the term connotation;

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

b. The more familiar linguistic notion at that point.

145

should be considered

2) A common noun: (6) 36. The descriptive-appositive SSyntRel [descr-appos] connects a postposed noun phrase to a noun (including proper nouns) or a pronoun, or to a verb, such that the dependent noun phrase describes the same object, phenomenon, quality, or event as the main term, indicating some of its additional characteristics. Semantic contrasts between the appositive and the descriptive-appositive SSyntRel's are possible: (i) vs.

(2)

b.

will forget everything [i.e., 'Bill, being an alcoholic,

a. The suffix

-tion will be discussed later.

vs. b.

will be discussed later.

The dependent member of the descriptive-appositive SSyntRel can be one of the following four items: 1) A noun phrase headed by a common or proper noun and modifying — A noun: (3)

a. Content more difficult . . .

at best, becomes even

146

SURFACE SYNTAX OF ENGLISH

b. I am grateful to Dr. R. sing Division.

of the Language Proces­

c. I am most grateful to its Head,

Simmons.

— A pronoun: (4)

a. Have you forgotten b. Neither of us

professors is quitting.

 None of you

guys are any good.

d. And wipe off the lipstick, you

fool!

— A verb (in fact, here the appositive modifies a clause or a whole sentence): (5)

a. The translator would be bound to likely to be subconscious. b. The UDC may indeed become a suitable "switching language" parti­ cularly because its and also because its basically faceted structure makes it possible to not possessed by any other coding or classification system.

2) A pronominal phrase headed by THAT2 or ONE2 : (6)

a. A more recent possibility, that of obtaining a translation of an article from the Federal Translation Service, . . . descr-appos

b.

that she gave me, is as follows.

3) A noun phrase introduced by explicative or restrictive expressions (of the especially type) :

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

(7)

a.

147

believe that it is not the case.

b. Romance following features.

or Portuguese, possess the

4) A conjunctional phrase with an explanatory conjunction: (8)

a. A computer, to know this.

cannot be supposed

b. So'we wish to construct a formal g r a m m a r , t h a t is, a set of rewriting rules. 37. The sequential SSyntRel [sequent] connects: — Either a noun to the preceding noun, to form a sequential phrase that appears as the dependent member of the compositive SSyntRel (see page 141). Such a sequential phrase can denote the starting and ending points of a movement (sometimes including intermediary points), the main actants of an event, etc. — Or a numeral, alone or subordinated by the preposition TO or THROUGH, to the preceding numeral, to form a sequential phrase that denotes the extreme points of a numerical interval. — Or an adjective to the preceding adjectives, to denote some conjoined prop­ erties of the item modified. Accordingly, the dependent member of the sequential SSyntRel can be one of the following four items: 1) A noun: (i) 2) A numeral: (2) 3) A prepositional phrase with TO or THROUGH:

148

SURFACE SYNTAX OF ENGLISH

38. The attributive SSyntRel [attrib] connects a non-adjectival postmodifier, called an attribute, to its governor: a noun, an adjective, or an adverb. The dependent member of the attributive SSyntRel can be one of the follow­ ing eight items: 1) A prepositional phrase modifying — A noun: (1)

a. Learners

with different backgrounds do not necessarily have

the same interests. b. The words

in question are identifiable as adjectives,

 References are numbered by numerals in square brackets which refer to the bibliography at the end of the book. — A pronoun: (2)

Doesn't

me?

— An adjective, including the superlative adjective: (3)

a. This connection is strictly causal b. While advantageous drawbacks.

in nature.

in many respects, the procedure has many

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

.  is the

149

our department.

NB: note the non-projectivity normally entailed by the attribute of a superlative adjective, when it modifies a noun. 2) A prepositionless noun phrase denoting — A parameter: (4) — Time: (5) — Distance: (6)

the

beyond the hill.

— A measurement unit: (7)

a. 300 b.

— A part of the day: (8)

Friday

3) An absolute phrase, namely — An absolute-predicative phrase headed by a pronoun: (9)

The council includes three delegates, each representing a separate constituency: the student body, the teaching body and the graduates.

— An absolute-conjunctional phrase: (10) a. We are concerned here with water b. The gas, volume.

when in a pure state.

if under heavy pressure, can be reduced to minimal

150

SURFACE SYNTAX OF ENGLISH

4) A reflexive pronoun: himself . . .

(11) But even Chomsky 5) An adverb: (12) a. The discussion

above of ranks of construction also indicates

that it is something b. 6) An as-phrase: (13) For me

as President this day will remain unforgettable.

7) An infinitive phrase modifying — A noun or a pronoun: (14) a. a book

to read; something

b. The amount

to eat;

to be paid includes the cost of packing.

 The program generates material and the answers

to be presented to the student

to expect.

— An adjective: (15) a. CETIS is t h MT services.

e

p

r

o

v

i

d

e Russian-English

b. The Georgetown system was the

be started.

8) A distributive phrase: (16) a. His students have to read six volumes of about 300 pages b. They are selling sirloin for 3

(per pound).

151

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

39. The descriptive-attributive SSyntRel [descr-attr] connects a qualifying (= non-restrictive) attribute to its governor: a noun or an adverb (unlike the at­ tributive SSyntRel, the governing member of this SSyntRel cannot be an ad­ jective). The descriptive-attributive SSyntRel can be semantically contrasted with the attributive SSyntRel: (1)

a. a student

- from Cornell University

vs. b. Professor L. Babby,

from Cornell University, . . .

The dependent member of the descriptive-attributive SSyntRel can be one of the following six items (roughly paralleling these of the attributive SSyntRel): 1) A prepositional phrase modifying — A noun: (2)

the second section, have no such property.

— An adverb : (3)

He saw the lady earlier,

2) A prepositionless noun phrase denoting — A parameter: (4)

Her brother Mike,

— Distance: (5)

This southward from Victoria Square, is considered by some a masterpiece of Teipimeinen.

— Localization: (6)

The publication of this volume was made possible by a grant adminis­ tered by the fornia.

Angeles,

Cali­

152

SURFACE SYNTAX OF ENGLISH

— Quantity: (7)

Its overwhelming tination in less than three hours.

— allows it to reach its des­

3) An absolute-predicative phrase:

(8)

4) An adverb: (9)

knew nothing of the degree.

5) An as-phrase : (10) Prof. Fokker, budget cut.

as the principal researcher, objected to this

6) A to-infinitive phrase: (11) a. The subsidiary electrode, to step pulses of the applied potential, was a helix of platinum wire. b. Below 1°K, magnetic methods, chapter, are employed.

to be explained later in this

It is probably worthwile to indicate the following proportionality: 1

2

3

modif

descr-modif

: modif-adv

appos

descr-appos

: appos-adv =

= attrib

descr-attrib

: attrib-adv

9z

- The dependent member of the SSyntRel's in the first column specifies both the restrictive and the qualifying characteristics of the governing member (with the exception of the appositive SSyntRel: the appositive is limited to restrictive properties only).

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

153

— The dependent member of the three descriptive SSyntRel's (in the second column) specifies exclusively qualifying characteristics of their head. - And the dependent member of the three disjunctive adverbial SSyntRel's (in the third column) specifies qualifying characteristics, yet not of the cor­ responding head, but of the grammatical subject (see page 116, Remark 1). This is a detached sentence element. II. Coordinate SSyntRel 40. The coordinative SSyntRel [coordin] connects any two conjoined sentence elements, or CONJUNCTS, the governor being invariably the leftmost one. (For argumentation in favor of such a treatment of coordination see above, § 2 of this chapter, pp. 64-65.) Two types of coordinate constructions are distinguished: 1) Conjuncts are linked directly, with no conjunction (asyndetic coordination): (1)

a. . . . in the writings of Austin,

Zadeh, and

Searle. b. . . . a broad, c. Slowly,

and interesting field. he crept towards his victim.

d. The system one.

find the next

A particular case of asyndetic coordination is the adjoining of the second conjunct introduced by a specific restrictive dependent (such as NOT, LESS or NEVER): (2)

a. The spot is b. I will not be able relevant to the subject.

2) Conjuncts are linked with a coordinate conjunction: (3)

a. We examine utterances consisting of one

all questions

154

SURFACE SYNTAX OF ENGLISH

c. The paper gives a realistic assessment and

of the needs of their readers.

d. Scientists are now discussing an alternative to reliance on human translators. Three important remarks are called for at this point. 1. Some conjuncts may be linked by certain conjunctions that are currently considered subordinate: THOUGH, WHILE, IF. We single out this particular subclass of conjunctions under the name of quasi-coordinate: (4)

a. Now let me mention a parallel b. Few

later distinction.

of the services agree to use a thesaurus.

2. Conjuncts can belong to different parts of speech — or, more precisely, to different syntactic classes — provided they fill RELATED surface-syntactic roles: (5)

a. They occur too frequently to be useful. b.

c.It was a typical summer evening, the atmosphere that inanimate objects seemed to be endowed with two or three senses.

SURFACE-SYNTACTIC RELATIONS IN ENGLISH

155

[In (5d), the prepositional/br-phrase for a new . . . election and the infinitive for-to-phrase for the people to establish . . . fill the same syntactic role: both must be subordinated to calls via the 2nd completive SSyntRel] Generally speaking, the relatedness of surface-syntactic roles of X and Y is a necessary condition for their coordination. We say that in a given sentence, phrases X and Y fill related surface-syntactic roles if and only if X, when Y is absent, is subordinated to its governor via the SSyntRel r\, and Y, when X is absent, is subordinated to its governor via the SSyntRel r2 such that either r\ = r2, or the pair (r1, r2) belongs to a pre-established list of SSyntRel pairs, e.g., (modificative, compositive) or (1st completive, auxiliary): (6)

Note, however, the following interesting example of conjuncts belonging to different syntactic classes: (7)

But now let us ask the question of whether conditions it happens.

which

This example is not covered by our definition since the SSynt-roles of whether and the prepositional phrase under which conditions are not related: the former must subordinate the main verb of the clause, the latter must be subordinated to the same verb. Thus a broader definition of related SSynt-roles is obviously needed. 2 3. Coordination of adjectives each preceded by an article presents an addi­ tional difficulty for the surface-syntactic description; cf. the following examples:

SURFACE SYNTAX OF ENGLISH

156

(8)

a. . . . the mathematical

b. . . . analyzing the firstAt the deep-syntactic level, the deleted noun must be restored, so that (8a) ap­ pears as (8'): (8')

. . . the mathematical PROBLEM(S)

If the article is not repeated, the SSyntS of the construction in question is dif­ ferent: (8") . . . the mathematical

physical problems.

NOTES 1. (To page 99.) With the second passive form of a verb such as GIVE, SHOW, etc. the agentive complement can correspond to the third DSynt-actant:

(i) Peter [= 1st DSyntA]

[= 3rd DSyntA].

2. (To page 155.) We gratefully acknowledge here the assistance of Mrs. N. N. Pertsova: the corpus of English coordinative constructions she collected helped us to formulate the notion of relatedness of SSynt-roles. - An important discussion of conjuncts filling non-related SSynt-roles in Russian is found in Sannikov 1979—80.

SPECIMENS OF SURFACE-SYNTACTIC STRUCTURES FOR ENGLISH SENTENCES t (1) That this interest continues to increase is attested to by the growing number of papers.

157

SPECIMENS OF SURFAC STRUCTURES

158

SURFACE SYNTAX OF ENGLISH

(2) Here we have the first example of adjectives being tied to their subjects in a different manner. The SSyntS of(2):

(3) These experiments were done ten years ago, and were somewhat rudimentary by today's standards, yet they were surprisingly successful.





H

n



 H

> n m

TI

  Pö

TÍ





Ü tn

m n

 "d

(4) The task of devising mechanical procedures for determining the case structure, while undoubtedly difficult, does not appear to be any more difficult than it would be under any other theory.

SURFACE SYNTAX

SPECIMENS OF SURFACE STRUCTURES

161

NOTES

1. (Figure 2.) An alternative syntactic description is available for sentence (2): - Either we take the phrase being tied . . . to be a postmodifier of adjectives, and the meaning is '. . . example of adjectives . . . , the latter being restrictively modi­ fied by this phrase. - Or we understand the phrase being tied as a quasi-predicate of an absolutepredicative construction, with adjectives as its subject (= head, in our approach), and then the meaning is '. . . example of. . . being tied . . .'. The actual semantic difference is minimal. 2.

(Figure 3.) An alternative analysis is possible at this point. Namely, the phrase by today's standards can be considered as an adverbial depending on the verbal head, i.e. on BE2 (rather than as an attribute of RUDIMENTARY):

Cf. an alternative surface rendering: (3') a. . . . and, by today's standards, were somewhat rudimentary. b. . . . and were by today's standards somewhat rudimentary. This syntactic variance has no relevance for meaning, since be rudimentary represents ONE semantic unit and thus on the semantic level there is no difference between the modification of rudimentary or of were.

162

SURFACE SYNTAX OF ENGLISH

3. (Figure 4.) The phrase for determining the case structure can also be analysed as an object of DEVISE (rather than of PROCEDURE p i):

This alternative structure could be implemented, e.g., as follows: (4') . . . the task of devising, for determining the case structure, mechanical procedures that would lend themselves to easy programming, while undoubtedly difficult, does not appear . . . Note that in our SSynt-description of (4), we presuppose that both the verb DEVISE and the noun PROCEDURE take a/or-phrase as their syntactic object, which fills a semantic valence slot. 4. (Figure 4.) ANY modifying MORE 1 and ANY modifying THEORY below are two dif­ ferent lexical items and should be distinguished by numerical superscripts.

Chapter III Deep-Morphological Representation for English Sentences

The presentation in this chapter proceeds by the same stages as the discussion of the SSyntS in Chapter II: § 1 — General Characterization of the Deep-Morphological Representation (of a Sentence). § 2 — Deep-Morphological Structure. § 3 — Conditions for Grammatical Correctness of English Deep-Morphological Structures. § 4 — Morphological Variables and Values of Wordforms in Modern English. In fact, there is little to say about the Deep-Morphological Representation without meddling in the internal affairs of morphology. Since it is syntax we are now preoccupied with, we will limit ourselves to a few remarks which seem to be necessary for understanding the syntagms in Chapter IV, as well as the model as a whole. For that reason, the four sections below are rather short and could easily be collapsed into one. We prefer, however, that Chapter III should exhibit the same logical structure as Chapter II and that it should describe the DMorphR in a way parallel to that of the SSyntR. §1 General Characterization of the Deep-Morphological Representation (of a Sentence) The D(eep)-Morph(ological) R(epresentation) of a given sentence is a pair of formal objects called STRUCTURES; symbolically, DMorphR = where:

SURFACE SYNTAX OF ENGLISH

164

- DMorphS is the DEEP-MORPHOLOGICAL STRUCTURE (of the sentence represented). It specifies all the actual wordforms of the sentence and their linear order. The DMorphS is the MAIN, or BEARING, structure with respect to the other member of the pair, which completes it. For more details about the DMorphS of a sentence see below, § 2 of this chapter. - DMorph-ProsS is the DEEP MORPHOLOGICAL-PROSODIC STRUCTURE; it shows all the prosody patterns, both the meaning-bearing and syntacticallyconditioned types, present in the sentence at hand. In written texts, the DMorphProsS is represented by punctuation. Only the DMorphS of sentences is considered in this book. The surface-syn­ tactic rules (syntagms) given in Chapter IV are based on the deep-morphological structure of sentences alone, so that different prosodiC phenomena relevant at this level (interword junctures, pauses, contours, etc.) are not taken into account. Moreover, we have, not yet elaborated any formal means for representing the DMorph-ProsS, so that the illustration that follows is quite approximate in this respect. The DMorphR (1) below represents sentence (2) (which appeared on page 47):

(2)

I loved them until they loved me [D. Parker].

Ideally, the surface-syntactic component of a Meaning-Text model for English must, among other things, establish the correspondence between the SSyntR (2) on page 47 and the DMorphR (1) above. Actually, the syntagms in § 3 of Chapter IV do less: basically, they are limited to the correspondences between the SSyntS and the DMorphS of a sentence only. But even these correspondences are not covered completely: a great deal of word order information is not presented in the syntagms, but is described by global word order rules, which fall outside the scope of this book. Nevertheless, the syntagms take care of some punctuation marks, namely, those which are determined by the SSyntS. Since the punctuation reflects, to a certain degree, the prosody, one may say that at least part of the

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

165

DMorph-ProsS is covered by our syntagms. See a remark about this in § 1 of Chapter IV, p. 182, item (ii). §2 Deep-Morphological Structure The D(eep)-Morph(ological) S(tructure) of a sentence is a string (= a totally ordered set) of nodes such that: (i) there is a one-to-one correspondence between its nodes and the wordforms of the actual sentence it represents; (ii) each node is labeled with the full deep-morphological representation of the corresponding wordform; (iii) the order of nodes is exactly that of the corresponding wordforms in the sentence. Unlike the SSyntS, where three deviations from the one-to-one corresponden­ ce between the SSyntS nodes and the wordforms of the sentence are foreseen (see pages 57-58), in the DMorphS this correspondence is violated in only one case: Morphological fusions such as the Fr. au = à le or the Eng. /wil/ we'll [= we shall /will] are split into their constituent wordforms at this level, so that we have in the DMorphS À LE (for au) and WE SHALL or WE WÌLL (for we'll). Except for the indicated phenomenon, the correspondence between the wordforms of a sentence and the nodes of its DMorphS is absolute. Thus a multilexemic idiomatic unit (= a phraseme), irrespective of its syntactic properties and behavior, is represented in the DMorphS by as many nodes as there are wordforms in it :

OF COURSE; BY ALL MEANS;NO LONGER;. . . In the SSyntS, each such rigid phraseme is represented by a single node. Zero wordforms (page 57, item 3) do not appear in the DMorphS at all, since they are essentially syntactic units. A node of the DMorphS of a sentence is labeled with the full D(eep)-Morph(ological) R(epresentation) of the corresponding wordform. The DMorphR of a wordform was described in Chapter II, § 2, pages 58-59, and there is no need to repeat this description here.

166

SURFACE SYNTAX OF ENGLISH

Conditions for Grammatical Correctness of English Deep-Morphological Structures There are two classes of conditions that specify the well-formedness of a string of DMorphR's of wordforms making up a sentence: — Conditions for grammatical correctness of the individual items in the DMorphS, i.e. conditions for the well-formedness of a given full DMorphR (of a wordform). — Conditions on the cooccurrence of wordform DMorphR's. Since the only relation between DMorphR's of wordforms explicitly stated at this level is that of linear order ('A DMorphR immediately precedes/follows another DMorphR'), all conditions on the cooccurrence of wordform DMorphR's are in fact condi­ tions on their ordering with respect to one another. 1. Conditions concerning individual DMorphR's of wordforms state what morphological variables must appear with a given lexeme and what morphological values can/cannot co-occur within the DMorphR of a wordform. For example, some of these conditions specify that in the DMorphS an English finite verb must distinguish mood and (in the indicative) tense, while an infinitive, a participle, or a gerund must not; that a personal pronoun must be marked as to its grammati­ cal case; and that a noun must carry an indication of its number. From the viewpoint of surface syntax, the conditions for the well-formedness of individual DMorphR's operate as filters checking the output (in the direction from meaning to text) or the input (from text to meaning) of the SSynt-component of a linguistic model. Yet, conceptually they belong to morphology rather than to syntax, and we will not formulate these conditions here. 2. As for the conditions on the cooccurrence of wordform DMorphR's we can only give a few examples. — A coordinate or comparative conjunction cannot immediately precede another coordinate or comparative conjunction: *than and, *but or, *as than,. . . (An exception: a coordinate conjunction can precede the left part of a binary coordinate conjunction: and either. . . or, but both . . . and.) — A determiner cannot immediately precede another determiner, a wh-word, or a conjunction: *the a, *this the, *my when, *the what, . . . — The lexeme IF2 (= 'whether') cannot occupy the first position in a senten­ ce:

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

(1)

167

a. *If he will accept the nomination is still unknown [= Whether he will accept . . . ] ;

but cf.: b. It is still unknown if {= whether ) he will accept the nomination. In fact, we do not yet know to what extent the cooccurrence restrictions at the DMorph-level are relevant or useful. We are inclined to believe that they can be of particular importance in automatic text analysis, when a computer has to discover the SSyntS of a sentence starting from its DMorphS. Then such local cooccurrence restrictions can help sift out a host of ill-formed combinations and thus facilitate the computations. In any case, we mention these restrictions here for reasons of symmetry. §4 Morphological Variables and Values of Wordforms in Modern English Since the grammatical categories of English lexemes are not our basic concern in this book, we simply adopt the most traditional view of grammatical categories, as expressed in, e.g., Smirnickij 1959, Barxudarov 1975, Quirk et al. 1972, and, as far as verbal categories are concerned, in Jakobson 1957 and 1959. This view is summarized in Percov 1976, 1978a, and 1978b, and here v/e will closely fol­ low the three last-mentioned papers. Our only departures from the well-established view concern 1) grammatical case in nouns and 2) verbal moods; see below. First of all, we must emphasize the important difference between what we call the full GRAMMATICAL REPRESENTATION of a lexemic form, or a lex, and the full DEEP-MORPHOLOGICAL REPRESENTA TION of a wordform (cf. Percov 1978b: 6—7). A lex is a more general notion than a wordform: a lex can be either a wordform or a phrase, i.e., a surface-syntactic configuration of wordforms (= a dependency tree) implementing a single lexeme as its analytical form. A particularly striking feature of English is, as everybody knows, the plethora of analytical forms — phrases that express grammatical meanings. Thus a verbal lexeme like INDICATE can be actualized in a text either by phrases such as has indicated, will indicate or am indicating, or by wordforms such as indicate, indi­ cates or indicated. All such phrases and wordforms are lexes of the lexeme in question, and all of them express some grammatical meanings. Thus the lex has

168

SURFACE SYNTAX OF ENGLISH

indicated expresses the active voice, the perfect taxis, the non-progressive aspect, the indicative mood, the present tense, the non-potential perspective, the direct status, the finite finitude, and the 3rd person singular; the lex indicated [taken as a modifier participle] expresses the passive voice, the non-perfect taxis, the past tense, and the participial finitude. All these and similar characteristics make up the full grammatical representation of a lex, or an actual lexemic occurrence. We give below a list of all grammatical categories of English along with their grammatical meanings, or grammemes. The categories are grouped according to the part of speech of the corresponding lexeme. Grammatical Categories in Modern English Noun Number: singular(sg), pl(ural) {pencil - pencils, man - men, foot - feet) In our opinion, number is the only grammatical category of English nouns. As for case, we do not consider the possessive form (also called 'Saxon Genitive') to be a case form; therefore, in our framework the English noun does not have cases at all (in contrast to pronouns, which do). For more details, see Note 1, page 174. Pronoun (personal) Case: nom(inative), obl(ique) (I - me, he - him, . . ., who - whom). Note that number and person are not grammatical categories of English pronouns but rather components of their lexicographic description. For example, I and we belong to two different lexemes, as do he and they, or, for that matter, he and she (while pencil and pencils ox foot and feet are, respectively, lexes of PENCIL and FOOT). Verb The English verb has ten grammatical categories: seven meaning-bearing and three syntactically conditioned ones. — Meaning-bearing verb categories: 1. Voice: act(ive), pass(ive) (writes - is written, [to]love - [to] be loved, etc.)

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH 2. Taxis: 3. Aspect: 4. Mood:

169

non-perf(ect), perf(ect) (comes - has come, telling - having told, etc.) non-progr(essive), progr(essive) (comes - is coming, wrote was/were writing, etc.) ind(icative), cond(itional), irr(ealis), subj(unctive), imper(a­ tive) ([/] come - [I] would come [if. . .]- [If I] were [you . . .] - [He demands that I] (should) come ~ cornel)

While all the other English verbal categories on our list are generally accepted, the number and structure of English moods is an open question: some consider our subjunctive to be in fact imperative, some identify the irrealis with the past tense form of the indicative, and so on. Without attempting a full-fledged sub­ stantiation of our analysis, we will simply mention in Note 2, p. 175, some facts behind it (cf. Percov 1976: 44-52). 5. Tense: pres(ent), past ([I] come - [Γ] came) 6. Perspective: non-pot(ential), pot(ential) ([ƒ] come - [I] shall/will come). 7. Status: dir(ect), indir(ect) ([I] come - [I] do come); direct and indirect statuses are sometimes called 'non-assertorial' and 'assertorial', respectively. — Syntactically conditioned verb categories: 8. Finitude: part(iciple), ger(und), infinitive), fin(ite) (written/writing, writing, [to] write, [I] write/[he] writes, . . .) 9. Person: 1st, 3rd 10. Number: singular(sg), pl(ural) The English verb forms exhibit the following four person-number combina­ tions: — 1sg: [I] am, write, ask; — 3sg: [he] is, writes, asks; — sg: [I (he)] was, wrote, asked; — pi: [we (you, they)] are, were, write, wrote, ask, asked. Person is not distinguished in the plural and in the past tense; moreover, neither person nor number is distinguished in the non-indicative moods and nonfinite forms. The second person is not distinguished in the verb, since there is no singular 2nd person pronoun and no second person singular forms of the verb in

170

SURFACE SYNTAX OF ENGLISH

English (THOU and the forms in -est being obsolete and YOU invariably plural regardless of the real number of its referent(s)3,much like such pluralia tantum as SCISSORS or PANTS). A few verbs show no person and number distinction at all (modais and some auxiliaries). Adjective/adverb The adjective and the adverb have only one grammatical category: degree of comparison (with the exception of two pronominal adjectives: this - these and that - those, which also distinguish number). This category is meaning-bearing. Degree: positive, comparative, superlative (tall - taller - tallest). We avoid examining here the so-called analytical forms of degree, such as more beautiful and [the] most beautiful. Admitting them entails considering less beautiful [the] least beautiful and maybe as beautiful as as analytical forms of degree, too. Many of the complications involved are obvious. However, the problem is immaterial within the narrow framework of a purely morphological approach, which is the only thing of interest to us here. The forms more, most, less and least are the comparative and the superlative degrees of the lexemes MUCH/MANY and LITTLE/FEW, respectively. This is sufficient for our purpose. A crucial fact about the grammatical categories of English is that some of them are expressed morphologically (= synthetically), within the limits of a single wordform (by suffixes or apophonies, such as sing - sang - sung), while the others are expressed only syntactically (= analytically), by constructions with auxiliaries that comprise both two wordforms and the surface-syntactic relation which connects them. Because of this, the grammatical representation of an English lex is an abstraction that never appears as such in full at any level of linguistic representation. Let us consider an example: [I] am writing. — At the DSynt-level, this phrase will be represented by a single lexical node subscripted for all the meaning-bearing grammemes: WRITE ACT, non-perf, progr, ind, non-pot, dir

But the grammemes of finitude, person, and number are absent, since the syn­ tactically-controlled categories are not reflected in the DSyntS. — At the SSynt-level, the phrase will be represented by two lexical nodes, with different grammemic subscripts:

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

171

As we can see, the grammemes of taxis, aspect, perspective, and status do not appear: they are encoded at this level by the construction itself. Moreover, the three syntactically-controlled grammemes (fmitude, person, and number) do not appear with BE for the same reason as above. — At the deep-morphological level, our phrase will be represented as two nodes: BE1 ind, pres,

1sg

WRITEACT

pres, part

these DMorphR's completely specifying the actual wordforms am and writing. The full deep-morphological representation of a wordform specifies it up to free variations: HOOFP1=hoofs / hooves. Notice that the number and nature of morphological variables making up the DMorphR of an English wordform are, generally speaking, different from what appears in the grammatical representation of an English lex, and there are two reasons for this: First, some grammatical variables are never expressed morphologically in English: e.g., the verbal aspect or voice. Such variables do not appear in the DMorphR's of wordforms. Second, some variables that determine the actual shape of wordforms and appear in their DMorphR's do not pertain to the grammatical representation: in our opinion, they are not purely inflectional. Such is, e.g., the possessive form of nouns. These variables are adjoined (at the DMorph-level) to the gram­ matical representation. Let us now take a closer look at the composition of DMorphR's of English wordforms. Morphological Variables and Values of Wordforms in Modern English Noun 1. n(umber): sg (singular), pl(ural). 2. poss(essive form).

172

SURFACE SYNTAX OF ENGLISH

As stated above, the possessive form is not considered to be a case form. For us, it is rather a quasi-inflectional form induced either by the meaning (actual possession: my father's garden), the semantico-syntactic role (being the first argument of a predicate noun: John s departure), or the syntactic context. Practically, this means that 'common case forms', i.e., forms without the -s-suffix, are not marked for case at all. (See Note 1, p. 174, for arguments in support of our decision.) Examples pencil

= PENCILSg

pencils

=PENCILp1

sister

- SISTER sg

sister's

= SISTER sg, p o s s

sheep

=

SHEEP sg / p1

sisters

= SISTER p1

feet

-

FOOTp1

sisters'

= SISTER p1;

poss

Pronoun (personal) 3. c(ase): nom(inative), obl(ique) Examples me you

=Iobl =YOUnom/obl

we them

= WEnom = THEY o b l

Verb For establishing morphological variables and values of the English verb and for defining the corresponding paradigms, Huddleston 1975 is of great interest (although we by no means agree with him on all points). 4. v(oice):

act(ive), pass(ive).

Voice is morphologically expressed in participles only, never in finite or other infinite forms: written

=

writing

-

(the homonymous form writing represents also the gerund). Let it be stressed that we consider past participle forms of transitive verbs to be ambiguous between active and passive; see Note 4 for our reasons (page 176).

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

173

Note also that in presenting syntagms, we often use as an abbreviation the following two cover symbols: - "ppres", which stands for "act, pres, part" (present participle); - "ppas", which stands for the disjunction "pass, past, part/act, past, part" (past participle). 5. m(ood): ind(icative), cond(itional), irr(ealis), subj(unctive), imper(ative). Mood is expressed in finite forms only, so that the presence of a mood grammeme unambiguously signals the finiteness of the corresponding form. Note that the conditional is morphologically expressed in modal verbs only. In Could you give me a hand1 or You might be right we find CAN cond and MAYcond- In all other verbs, the conditional is realized syntactically, i.e. as a phrase lex (would like, would give, etc.). 6. t(ense): pres(ent),past. Tense is expressed in the indicative mood and in the participle only. (For cogent arguments against distinguishing present and past tenses within the Unreal Mood — roughly, our irrealis plus subjunctive — see Huddleston 1975: 159ff.) 7. f(initude): part(iciple), ger(und), inf(initive). The finiteness is not indicated as such in the DMorphR of a finite verbal wordform, since it becomes redundant in the presence of mood or person-number grammemes. 8. p(erson):

1st, 3rd (am - is, [I] go — [he] goes; distinguished only in the present tense and in the singular).

9. n(umber): sg, pl (was - were; are; takes — [they] take; distinguished only in the indicative mood). For a general survey of verbal paradigms see note 5, p. 177. Adjective and adverb 10. d(egree):

posit(ive), comp(arative), super(lative).

Examples tall

=

TALLposit

easier

= EASY comp /EASILY comp

taller

= TALL comp

ta

= TALLsuper

tallest

These ten groups of morphological values, and only these, appear in the DMorphR's of English wordforms.

174

SURFACE SYNTAX OF ENGLISH

NOTES

1. (To page 168.) We cannot fully justify here our rejection of the -'s-form as a grammatical case; yet a short summary of possible arguments would not be entirely out of place. We will proceed from the following three assumptions: (i) Grammatical case is an INFLECTIONAL category (whatever the difference between inflection and derivation might be), (ii) Grammatical case expresses, first of all. the SYNTACTIC ROLES of nouns - with respect to the GOVERNING LEXEMES, generally - verbs. (The lexeme X is said to govern the case of the noun Y if and only if X both semantically predicts a noun in a specific syntactic role with respect to X and controls its case.) (iii) At least some marked cases within a case system must express MAJOR SYNTACTIC ROLES, such as grammatical subject and objects. It is not sufficient that all these roles be expressed by only the unmarked case, which is taken to be a case solely by virtue of its opposition to the marked cases. Assumptions (i) —(iii) roughly specify the notion of grammatical case, as we would like to construe it. We want to model this notion after such typical case systems as, e.g., those in Latin, Sanskrit, or Russian. For,a formal definition of case see Mel'cuk 1977. We will try to show that the so-called English '-s-genitive' contradicts all three pro­ perties of typical case systems. 1) The -'s-form is different from all the other English inflectional categories in that: - It is applicable to only a small proportion of English nouns (mostly animate nouns and measure units, such as a month's leave, etc.); a typical inflectional category in Eng­ lish, such as number in nouns or tense in verbs, applies to most lexemes of the corre­ sponding part of speech. - It can co-occur with another inflectional category (as in childr-en-'s etc., i.e., with number); no other English inflectional category can. - It can characterize a whole phrase (as in the girl I danced with's brother), which is rather typical for English derivation (transformational grammar-ian, general paralyt-ic, physical chem-ist, etc., derived from transformational grammar, general paralysis and physical chemistry, respectively). No other inflectional category of English is capable of characterizing phrases. 2) The -'s-form is never governed by a verb. 3) The -'s-form never expresses a major syntactic role, and there is no other marked case form of a noun which does so. The only noun form capable of filling major syntacic roles is the unmarked 'common case', which may become a case only by contrast with the -s-form. So a two-case system in English (common case - genitive case) would be radically different from typical case systems, and we see no reason to use a general label of 'case' for a specific phenomenon we observe in English. That is why we call the -'s-form simply

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

175

a possessive form and see it as an instance of syntactic derivation, approximately parallel to denominal adjectival derivation {father - father's /paternal), etc. To be sure, this solution opens a terrible theoretical abyss. If, indeed, possessive forms pertain to derivation, then father and father's must belong to two different lexemes, which is strongly counter-intuitive. To alleviate this, perhaps the concept of 'different lexemes' should be sharpened, allowing for different types of lexemic distinctions. Or else the contrast of inflection vs. derivation should be resolved into three-term rather than two-term opposition: father vs. fathers will represent inflection , father vs. father's - inflection 2, and.father vs. paternal - derivation. (The idea that the borderline between inflection and derivation is not quite clear-cut is by no means novel in lin­ guistics.) Be that as it may, at present the situation with -s-forms is far from clear. Therefore we do not insist on our proposal: much more time and effort are needed to establish reliable conclusions. But all this confusion is tangential to our main topic, so that a different analysis of English possessive forms will not seriously change the proposed model. 2. (To page 169.) Let us indicate some reasons why we think that the number of English verbal moods cannot be reduced. (i) The irrealis cannot be identified with the past tense form because of cases such as if he were, where were, and not was, is obligatory. It is true that it is only the verb BE that distinguishes formally between the irrealis and the past tense. However, BE plays an exceptional role in English conjugation, forming passives and progressives and thus imposing this distinction upon all verbs. (ii)The subjunctive cannot be identified with the imperative (or the infinitive) for at least the following four reasons: - The subjunctive presupposes the compulsory use of an overt grammatical subject, while with the imperative a grammatical subject is rather exceptional, and with the infinitive impossible. - The synthetic form of the subjunctive we are discussing is synonymous and mutally replaceable with the analytical form SHOULD + V i n f : (1) a. FIDE rejected the suggestion that there be (= should be ) no limit on the number of games. b. The house should be left to all of us - with the understanding that it be <= should be) sold after both parents pass on and the proceeds be split three ways. Neither the imperative nor the infinitive allow replacement by the SHOULD + Vinf form. - The meaning of the subjunctive is quite particular and does not coincide with the meaning of T ask you to . . .' or 'I order that . . .' of the imperative: cf. (1) above. - The synthetic form of the subjunctive is negated by putting not immediately before it, while the imperative does not admit this kind of negation: (2) a. The chief physician ordered that John not examine them immediately. b. *Not examine them! [= Don't
176

SURFACE SYNTAX OF ENGLISH

(iii) The phrase WOULD + Vinf should be considered part of the system of moods (i.e. the analytical form of the conditional) for the following three reasons: - Its meaning is unmistakably modal. - It stands in a regular correlation with the irrealis : (3) If I were now in Paris I would go for a walk on the quais. - In modal verbs, as indicated above, the conditional is expressed morphologically (i.e., it has a synthetic form): (4) If I were now in Paris I could (= would be able to) go for a walk on the quais The form could in (4) is not replaceable by the form of any other mood of any other verb. (iv) The imperative cannot be identified with the infinitive for at least the following three reasons: - Their meanings are so different. - The imperative allows the replacement by the phrase DO + Vinf for emphasis or negation {Please, do (not) sit down!), but not the infinitive. - The imperative, but not the infinitive can be syntactically conjoined with a finite verb form, as in (5): (5) a. Have [imper] another beer and you'll have a hang-over tomorrow. b. Don't [imper] try that trick again, or I'll leave immediately. 3. (To page 170.) We think that in contrasts such as (1) You deceive yourself vs. You deceive yourselves and (2) You are a teacher vs. You are teachers the conditioning factor is the SEMANTIC number of referents, not the GRAMMATICAL number of YOU. The situation is very much like what we observe in (3)—(4): (3) My friend hurt himself vs. My friend hurt herself [will we say that FRIEND is gram­ matically ambiguous between masculine and feminine gender?]. (4) The scissors are a tool that . . . vs. Scissors are tools that . . . [will we say that SCIS­ SORS is grammatically ambiguous between singular and plural?]. Cf., however, Huddleston 1975: 171 for a different solution. 4. (To page 172.) We know of three arguments in favor of distinguishing the active and pas­ sive forms under the surface shape of an -ed or -en participle (cf. Percov 1976: 21ff.): (i) The past participle of a transitive verb features two alternative government patterns. For example, written.1 in a perfect-tense phrase has the same government as the verb WRITE; but written 2 in a passive phrase or used as a modifier has a converse government pattern. Written1 remains transitive while written 2 is clearly intransitive. To capture this property, we propose to treat written\ as an active past participle to be used in perfect phrases only, and written 2 as a passive past participle, with the conversion of the syntactic actants. (ii) The past participle of an intransitive verb is never passive. A few verbs, such as GO or COME, even show an obviously active use of their past participles:

DEEP-MORPHOLOGICAL REPRESENTATION FOR ENGLISH

177

(5) a. All his anger gone, he walked up the slope. b. Come to think of it, I see the inevitability of her failure. c. . . . a big selection of newly arrived accessories. (True, this construction is by no means productive.) It seems much more natural to describe the italicized forms as active past participles. (iii) Distinguishing voice in past participles makes the system of English voices look more symmetrical, since voice IS distinguished in present participles: writing vs. being written. 5. (To page 173.) As an aid to the reader we give here four sample paradigms - for verbs BE, TAKE, FINISH and CAN, as they appear under our analysis. In arranging our data we follow the lines of presentation found in Huddleston 1975.

Legend: - means 'not distinguished'; stands for 'has only syntactic [= phrasal] realiza­ tion'; and * means 'absent from the paradigm' [= in a defective verb]. As can be seen from the table, our proposal gives the paradigm of TAKE or FINISH thirteen wordforms (many of them homophonous); the same paradigm in Curme (1935: 327-333) includes twenty-nine wordforms, while that of Quirk et al. (1972: 70) only five (with no homophony) and that of Huddleston (1975: 175) seven (with the homophony of took 'past indicative' / took 'unreal mood' and take 'present indicative not 3sg' / take 'base' (roughly, infinitive + imperative + subjunctive). In connection with the composition of the English verb paradigm, in particular, with the problem of morphological values and zero endings, the paper Carden and Pesetsky 1977 should be mentioned.

Chapter IV Surface-Syntactic Component of English

This chapter discusses the notion of SYNTAGM and offers a list of basic English syntagms. It consists of three sections: § 1 — General Characterization of the Surface-Syntactic Component. § 2 — The Notion of Syntagm. § 3 — Syntagms of Modern English. Section 3 is the core of this book — the main course, so to speak, everything else being a kind of garnish provided for purposes of introduction or explana­ tion. §1 General Characterization of the Surface-Syntactic Component As stated earlier (pp. 42-43), the surface-syntactic component of a MeaningText linguistic model provides for the correspondence between the S(urface-) Synt(actic) R(epresentation) of a sentence and its D(eep-)Morph(ological) R e ­ presentation). Just to refresh our reader's memory:

The full-fledged SSynt-component of a linguistic model comprises three types of rules: 1. Well-formedness rules for SSyntR's in the given language. 2. Well-formedness rules for DMorphR's in the given language.

SURFACE-SYNTACTIC COMPONENT OF ENGLISH

179

3. Rules for establishing correspondence between any correct SSyntR in the given language and all the correct DMorphR's that manifest it — or between any correct DMorphR and all the correct SSyntR's manifested by it. The rules of type 3 are called surface-syntactic rules, and they constitute the SURFACE-SYNTACTIC COMPONENT PROPER (i.e., in the narrow sense of the term). Only the surface-syntactic rules will be our concern in this chapter; for the rules of types 1 and 2, see above, pp.78-85 and 166-167. The SSynt-component proper consists of four subcomponents (in accordance with the four structures of the SSyntR): SSyntComp 1 is responsible for the correspondence between a SSyntS and all the possible DMorphR's of the sentence in question. SSyntComp 2 is responsible for the correspondence between a SSynt-CommS and all the possible DMorphR's (i.e., it takes care of all the features of the com­ municative organization). SSyntComp 3 is responsible for the correspondence between a SSynt-ProsS and all the possible DMorphR's. All the meaning-bearing prosodies, which are present in the SSyntR, are dealt with: implemented either by the linear order of the items in the DMorphS or by specific prosodic phenomena in the DMorphProsS. When we are working with written English, this subcomponent provides for the correct punctuation. SSyntComp 4 is responsible for the correspondence between a SSynt-AnaphS and all the possible DMorphR's — from the viewpoint of pronominal usage, coreference, etc. This book deals only with the first subcomponent; the other three are not discussed even in outline. To be more precise, in both representations involved, i.e., in the SSyntR and the DMorphR, ONLY the MAIN STRUCTURES ARE TAKEN INTO CONSIDERATION: the surface-syntactic structure (= the labeled dependency tree) and the deep-morphological structure (= the string of the DMorphR's of the wordforms). The communicative organization of the sentence, its prosody, the anaphoric and similar relations in it are excluded from our discussion. Thus all the rules of SSynt-subcomponents 2 through 4 (implementa­ tion of the topic/comment or old/new division; processing of the question, exclamation, etc. prosodies; treatment of pronouns — as opposed to nouns — from the viewpoint of their anaphoric functions) are not mentioned in what fol­ lows. Moreover, we do not cite any of the rules that establish correspondences

180

SURFACE SYNTAX OF ENGLISH

between the SSynt-structure and the syntactically-induced prosodies (or punctua­ tions), i.e. rules of the t y p e A s a result, only part (but the most important part, tor that matter) ot the tirst subcomponent is our goal. Specifically, we will be talking only about SSynt-rules that describe a much simpler correspondence than (1), namely, the correspondence (2): (2) In fact, there is a further limitation. The SSyntS-to-DMorphS rules are not described in full, either. To be more specific, we do not consider here what we propose calling "global rules". At this point we should distinguish two types of SSyntS-to-DMorphS rules: — LOCAL rules, or syntagms; — GLOBAL rules for word order. A local rule specifies the correspondence between a minimal syntactic tree, i.e. a pair of nodes X and Y linked by a SSyntRel r, and its deep-morphological implementation: the mutual linear order of X and Y and the syntacticallyconditioned morphological values of these nodes. Such SSynt-rules are called SYNTAGMS, and they will be discussed in greater detail in § 2, p. 186ff. A global word order rule specifies the correspondence between an unlimited configuration within the syntactic tree, i.e. more than two nodes, and the linear order of the corresponding items in the DMorphS of the sentence. Generally speaking, some global word order rules concern whole sentences, and there is no finite way of expressing these rules as deterministic (algorithmic) processes; such rules are better stated in terms of (a) certain target functions, (b) ranks associated with specific word arrangements, and (c) procedures such as those used in linear or dynamic programming (see It might be helpful to distinguish two types of global rules: 1) Arrangement of more than two items. Here are three examples: • Positioning of several premodifiers in a noun phrase, like the following seventy-three thick French volumes, but not, e.g., *the thick French seventythree following volumes. •• Positioning of the different grammatical objects, complements and circum­ stantials of the same verb; for instance, a prepositionless nominal object must precede an infinitive: Captain Malone ordered my friend to stay, but not *... ordered to stay my friend.

SURFACE-SYNTACTIC COMPONENT OF ENGLISH

181

Positioning of a noun modifier with right dependents: the thirteen primi­ tives proposed earlier, but not *the thirteen proposed earlier primitives. (In English, in sharp contrast to German or Russian, a modifier that has its own right dependents cannot normally precede the noun modified; cf. page 386.) 2) Situations of permitted or obligatory non-projectivity. 1 Non-projectivity may, and sometimes must, occur in English sentences, but only in certain restricted and well-defined contexts. Let us give two examples: • A heavy (= very long) postmodifier of the grammatical subject tends to be moved to a position after the verb phrase, so that a non-projective construction arises: (3)

The heavy postmodifier, which causes non-projectivity, is printed in italics; the non-projectivity is created by the framing of the predicative arrow:

(4)

The non-projectivity is created by the intersection of dependency arrows. Both of these rule types, i.e. rules for the simultaneous mutual arrangement of several lexemic items and rules for permitted non-projectivity, should be described separately from the syntagms, thus constituting independent sections of the SSyntS-to-DMorphS subcomponent. But, on the one hand, we have not yet developed these sections; we have not even explored, to any significant extent, the possibilities of using them. On the other hand, the global rules cannot be completely detached from the syntagms,

182

SURFACE SYNTAX OF ENGLISH

since in numerous cases a specific linear arrangement of several lexical items or a specific non-projectivity situation is possible (or impossible) only in connection with a specific syntagm. This is not always the case, of course; but we do not yet know where to draw the line between 'syntagm-sensitive' and 'syntagm-independent' global rules. At the same time, we feel that it would be regrettable to throw away some interesting surface-syntactic data already in our possession only because we have no storeroom for them today. Therefore, we have decided to include all the information we have about global SSynt-regularities of English in our syntagms (in the condition part; see below, § 2, pp. 189ff). In some instances this will be the right solution; in others, our decision may be viewed as an interim measure, to be revised later, when all necessary data are available. It will be easier at some later time to separate what is to be separated than to start looking for it anew. In any event, even in an ideal syntactic description, where all different types of data are kept strictly apart and presented separately, a system of cross-references between the global word order rules and the syntagms will be needed. Following the same line of reasoning, we have squeezed into the syntagms two further types of information pertaining to the SSynt-component but not to the syntagms themselves: (i) Data on the well-formedness (= grammatical correctness) of the SSyntS in question. Here we find, e.g.: restrictions on the cooccurrence of SSyntRel's or of some 'syntactic' lexemes; situations of permitted non-saturation of otherwise obligatory active valence slots; etc. These data properly belong to the well-formedness rules for SSyntR's; but as these rules have not yet been written, we have put several such indications into the condition part of the syntagms for which they are relevant. Moreover, in a few cases we have included in the syntagms certain lexemic cooccurrence restrictions concern­ ing the deep-syntactic structure. This has been done in an effort to avoid any loss of information already available, (ii) Data on punctuation. Actually, any such data should be used in SSyntS-toDMorph-ProsS rules because punctuation basically reflects the prosody. Yet, for the same reason as before, we have put them with the syntagms. To sum up: Only one fragment of the surface-syntactic component of the English language is presented in this book, namely a list of syntagms. We think, however, that this is a fairly representative fragment, the syntagms being the

SURFACE-SYNTACTIC COMPONENT OF ENGLISH

183

backbone of any syntactic description. And all the more so, since ours are 'enriched' syntagms: they contain much information that does not belong there and that stays there merely as a refugee waiting to be resettled permanently. The main types of 'illegitimate' data smuggled into the syntagms have just been characterized, and we will return to them below, in the discussion of the condi­ tion part of a syntagm. Now that the necessary restrictions and reservations have been clarified we can proceed to a closer examination of the syntagm, the "leading persona" of this book.

NOTES

1. (To page 181.) Projectivity is a very important surface-syntactic notion introduced in­ dependently by Hays (1960) and Lecerf (1960) and intensively studied by Hirschberg (1961), Fitialov (1962), I ordanskaja (1963), Marcus (1965a), and many others. The property of projectivity characterizes pairs (a sentence, its SS-structure). Put in different terms, a sentence is called projective if and only if, taken together with its SSyntS, i.e. with its dependency tree, it satisfies the following two conditions (= Projectivity Condi­ tions): Let the (i) (ii)

all the nodes of the SSyntS be ordered in accordance with the word order of sentence. Then: No two branches of the tree cross one another. No vertical projection from a node onto a horizontal line drawn below the tree crosses a branch of the tree.

If SS-dependency arrows are inserted into an actual sentence. (PC) appears under a slightly different form : (i) (ii)

No two arrows cross one another: No arrow 'frames' another arrow so that the first covers the governing node of the second:

(the nodes shown are not necessarily contiguous; the formulation is given in Iordanskaja 1963).

184

SURFACE SYNTAX OF ENGLISH

In most (maybe all) languages, most sentences are projective. This fact reflects the natural tendency to put a syntactically dependent item linearly as close to its governor as possible, or at least, if the dependent must be distant, to avoid the intersection of the domains of two different governors. Metaphorically, an item which is not my dependent should not be nearer to me than one of my dependents, direct or indirect. (i) is an example of a projective sentence:

However, probably all natural languages admit, to a greater or lesser degree, non-projective sentences. Some non-projective sentences are only stylistic variants of projective ones; some represent the only way of expressing the given construction. Thus non-projectivity can be either optional or obligatory. Let us give three examples of non-projective sentences:

SURFACE-SYNTACTIC COMPONENT OF ENGLISH

185

186

SURFACE SYNTAX OF ENGLISH

The requirement of projectivity for all sentences EXCEPT some specified structures is a powerful filter that rejects many ungrammatical arrangements of sentence constituents. Therefore, it is included in the global word order rules. And precisely because of that, all cases of permitted non-projectivity have to be listed for the corresponding language.

§2 The Notion of Syntagm A syntagm is a surface-syntactic rule specifying the correspondence between a minimal SSynt-tree (a pair of nodes linked by a SSynt-relation) and a minimal string that implements it at the deep-morphological level. For example, a syntagm tells us that the predicative SSyntRel in English is manifested by the agreement of the tensed verb with the grammatical subject (in person and number), the verb normally following the subject (but vice versa, if some special conditions for this inversion are present). Another syntagm stipulates that the prepositional SSyntRel is shown, at the DMorph-level, by the dependent being placed after the preposition (and, if it is a personal pronoun, being given the inflection of the oblique case); and so on. From a different angle, one can say that a syntagm is an elementary linguistic sign of the surface-syntactic level: its signifié (= signatum) is the SSyntRel ex­ pressed; its signifiant (= signans) is the corresponding fragment of the deepmorphological string, i.e., a pair of sets of morphological characteristics provided with relevant syntactics components plus the specification of the mutual order of both corresponding wordforms; and its syntactics is the set of conditions describing the cooccurrence of the given syntagm with other syntagms. (On the notion of the linguistic sign see, in particular, Meľčuk 1982b: 24—28, 40—41.) Our discussion of the notion of syntagm is best presented in three steps: 1. Characterization of the syntagm as such. 2. Means and devices for a more compact syntagm notation. 3. Classification of syntagms.

SYNTAGM 1.

187

Characterization of the syntagm.

Formally speaking, a syntagm is a tripartite expression of the form where means 'corresponds' and: 1) , the left-hand part of the syntagm, stands for a fragment of the surface-syntactic structure (of a sentence), i.e. for a (sub)tree. 2) if (from string), the right-hand part of the syntagm, stands for a fragment of the deep-morphological structure (of the same sentence), i.e. for a (sub)string. 3) ' (from condition) is the set of conditions (imposed on or both) under which the given correspondence holds. Let us have a closer look at the three components of a syntagm. • The left-hand part of a syntagm is a dependency tree which obligatorily contains a minimal non-unit subtree of the form

where: 1) r is the surface-syntactic relation (SSyntRel) described by the syntagm under consideration, i.e. its signifié; 2) the nodes X and Y are, respectively, the governing and dependent members of the SSyntRel r; 3) are strings of components of the syntactics of the lexemes with which nodes X and Y are labeled; 4) are strings of meaning-bearing morphological characteristics of the lexemes with which nodes X and Y are labeled. To simplify the notation, we will, omit the subscripts and since this cannot result in ambiguity. Thus instead o f ' w h i c h means 'node Y labeled with a lexeme having the syntactic features (o) and the meaning-bear­ ing morphological values x'." If, in addition to fragment contains other nodes and branches, these make up the TREE CONTEXT, which appears unmodified in the right-hand part of the syntagm. The use of tree contexts in their present form is a concession to better surveyability: essentially, tree contexts are part of the syntagm's syntactics and should

SURFACE SYNTAX OF ENGLISH

188

appear in the condition part (cf. the discussion of however, since we are striving to facilitate the reading of conditions (which can sometimes be extremely involved) we choose to keep the tree context where it is now. • The right-hand part of a syntagm is a string of two nodes having the following form: (3)

a.

or b. plus, perhaps, the tree context carried over from . The string (3a) or (3b) is the signifiant of the syntagm. Here the symbols have the following meaning: 1) The " + " sign means 'precedes' (in a linear sequence from left to right), and the "..." sign denotes a gap, i.e. the possibility of appearance of other lexemic occurrences between the two nodes separated by it; parentheses mean optionality. Thus,X+Ymeans thatXimmediately precedesY;X+ ... +Y—thatXprecedes Y but not necessarily immediately; and X...Y — that X and Y can be arranged in any order with a possible gap between them. 2) have the same meaning as above: strings of components of the syntactics of the lexemes with which nodes X and Y are labeled. Since these strings are identical in both the left- and right-hand parts of a syntagm, we will omit in order to achieve a more compact presentation of the syntagms. 3) are strings of all the morphological characteristics of the lexemes in nodes X and include meaning-bearing values ( and ) as well as all the syntactically-conditioned values that manifest the SSyntRel r in this syntagm. As proposed above, we will omit the subscripts x and y in these strings, too. Taking into account the tree context and the convention about the omission of the subscripts and syntactics in , the more developed form of a syntagm might look as follows:

SYNTAGM

189

In the graphic representation of a syntagm, the two-headed correspondence arrow is positioned against the 'working' SSyntRel r, and the string is positioned against the arrow. • The condition part of a syntagm is, formally speaking, the set of restrictions imposed on some of the nodes in as well as on the correlations between some of the nodes in both parts of the syntagm. These restrictions are such that it is difficult or impossible to express them immediately in the tree or in the string; the nature of this difficulty will become clear from the following survey of the different types of conditions admitted in From the linguistic point of view, syntagm conditions express three kinds of information: — 'Internal' properties of the syntagm, i.e. the correlations among its components proper. — 'External' properties of the syntagm, i.e. its cooccurrence with other syntagms. — Information that actually belongs to different sections of SSyntComp 1 but is temporarily stored in the syntagms (see above, p. 182). Each condition is an independent statement, or an independent logical formula; for example, it can be an implication, possibly with other implications embedded in it. All the conditions in are numbered consecutively, and the logical relation between them is conjunction: that is, for a syntagm to hold, ALL the conditions in its -part must be satisfied simultaneously. There are four basic types of conditions, and they are listed in the -parts of syntagms in the order in which they are mentioned here. (Remember that the condition part of a syntagm can include supplementary information coming from some other parts of surface syntax: for instance, data about sentence prosody or punctuation.) The conditions of a syntagm are arranged in the following manner: I. Correlations between different properties of two or more nodes. The properties involved can include components of syntactics, morphological values, or participation in some specified tree configurations (e.g., being the dependent member of a given SSyntRel, and the like). The correlations must be expressed by Boolean formulas, sometimes with a considerable depth of embedding of logical connectives. There are two sources for conditions of type I. First, they may express some cooccurrence restrictions in the SSyntS, thus bearing on its well-formedness (or even on the well-formedness of the underlying DSyntS). These conditions do not in fact pertain to syntagms and are present there only on a temporary basis; in the future, they will have to be eliminated from the condition part of syntagms and transferred to where they belong — to the rules for the correctness of the SSyntS.

190

SURFACE SYNTAX OF ENGLISH

Second, the need to specify the logical correlations between certain nodes may be due to collapsing several similar syntagms into one. Consider, e.g., Syntagm 27.7, page 364. We could have written two different syntagms — one with temporal prepositions and adverbs, and the other with locative ones, that is:

But instead of this, in order to save space and enhance surveyability, we have chosen to collapse these two syntagms into one, with the disjunction of syntactic features:

(5')

This requires adding conditions on the correlation between the syntactic features: if X = (temp), then Y = (temp) and if X = (loc), then Y = (loc). The conditions are needed to avoid synthesizing (or analyzing) ungrammatical expressions such as * through then or '"before there. From a different angle, we may divide conditions of type I into two subtypes: "substantive" and "screening". A "substantive" condition further characterizes constructions covered by the syntagm in question, making their description, pre­ sented in and ', more precise and specific. But a "screening" condition serves only to reject phrases which otherwise could be improperly subsumed under the given syntagm and which are described by a different syntagm. In most cases, we identify "screening" conditions as such in our comments. As a typical example of "screening" conditions, see Condition 1 in Syntagm 2.10, p. 238. II. 'Negative' tree contexts. These conditions express requirements that some particular nodes should not participate in some particular tree configurations. This type of condition is logically necessary since it is impossible to indicate 'negative' contexts directly in the tree in any natural way. III. Word order information of a global nature. Two subtypes must be distin­ guished here:

SYNTAGM

191

a) The linear arrangement of more than two nodes. These conditions are twofold: — Requirements concerning the linear gap between X and Y, i.e. statements as to what can, cannot or must be placed between them, and under what specific circumstances. — Requirements concerning the linear order of X or Y with respect to some other nodes appearing in the tree context. b) Permitted or obligatory non-projectivity situations characteristic of the con­ struction in question. They are preceded by the traffic symbol Conditions of types I through III can describe the cooccurrence of the given syntagm with other syntagms and thus constitute its syntactics proper. As indicated above (§1, pp. 181-183), these conditions may contain some 'illegitimate' informa­ tion, which should in fact be given elsewhere, e.g., in special sections (other than syntagms) of the SSyntS-to-DMorphS subcomponent. IV. Prosodic/punctuation information: indications about the possible or obligatory use of punctuation marks in the given construction. Conditions of this last-mentioned type, as was already pointed out, do not belong here (see p. 182, item (ii)); their presence in the syntagms is temporary. To be sure, there are conditions that present different combinations of these "pure" types: e.g., a specific correlation between some properties of two nodes entails a specific linear arrangement of these nodes; a particular non-projectivity situation presupposes particular punctuations; etc. This concludes the general characterization of the syntagm as the basic type of surface-syntactic rule.

2.

Means and devices for a more compact syntagm notation.

In order to provide for a more compact and elegant presentation of syntagms and at the same time to ensure their better survey ability, it proves helpful to have recourse to three further notions: — Standard subtrees. — Standard functions. — General conditions for groups of syntagms. Apart from contributing to the abbreviation of the syntagm notation, these entities are of some interest in themselves, since they help capture some useful generaliza­ tions about English.

192

SURFACE SYNTAX OF ENGLISH Standard Subtrees

A standard subtree denoted by A (from Greek AENAPON 'tree') is a dependency tree specifying a construction that behaves syntactically as a unit and is roughly equivalent to a lexemic category. To take a simple example, English has n predicative constructions that differ only by what fills the role of the grammatical subject: it can be a noun, a pronoun, a cardinal numeral, an adjective in an elective construction (the most gifted of young researchers), a specific prepositional phrase (from ten to twenty dollars), a gerund, a free relative clause (of the type what we might call 'causative'), an infinitive etc. — all in all, n different surface-syntactic items. At the same time, there are n completive constructions where the role of a prepositionless obj ect can be played by any of the same n items. And that is not all: there are n types of preposi­ tional phrase with the same n items as the dependent of the preposition. Note that n is greater than 10. So, instead of writing down  different syntagms to cover all three above-mentioned constructions, we introduce the variable ANP, i.e. a standard noun phrase subtree, and, using it, formulate only 3 syntagms. The possible values of the variable ANP for English are specified elsewhere, outside the syntagms proper. Our English syntagms use four standard subtrees: 1)ΔANP,i.e., a tree representing any noun phrase or any of its syntactic equivalents. 2)ΔANUMP,i.e.,a tree representing any quantity phrase (like ten times ox fifty girls). 3) AAPPROX, i.e., a tree representing any approximate quantity phrase (such as from 10 to 20 days or over 10 days). ΔAAPPROX is a particular case of ANUMP, the latter being a particular case of ΔNP: ΔANP⊃ΔANUMP ⊃ΔAAPPROX. 4)ΔAVP,i.e.,a tree representing any analytical verb form (have been living, will be stored,...) or any of its syntactic equivalents, such as constructions with modal and phase verbs (has begun speaking,...) or modal adjectives (is likely (sure) to). Note that in our formalism standard subtrees are never used in syntactic represen­ tations. They are an exclusive means for abbreviating and generalizing syntactic rules, in particular, syntagms; Δ's appear in ℐ and ℐ. A full description of the four standard subtrees mentioned above is found in Appendix II, pages 485ff.

SYNTAGM

193

Standard Functions A standard function is a set of conditions that must be satisfied for a specific syntactic construction to occur. To be granted the status of a standard function, the set of conditions has to be rather involved and represent a general enough syntactic phenomenon of the language under description (that is, recur in a number of different contexts). In our description of English surface syntax, eight standard functions of three different types are used: (i) Standard functions that describe grammatical agreement, or concord. (ii) Standard functions that describe grammatical inversions. (iii) A standard function that describes coordination. Unlike the standard subtrees, the standard functions of English are not given in this book. We limit ourselves to the following informal and very brief sketch. (i) Agreement Functions 1) A G R E E D E T ( N ) ( X , Y ) : rules of agreement in number between an inflected determiner X and the noun Y it modifies; e.g., this view—these views, but not *that books or those book. 2) AGREE (X,Y): rules of agreement in number and some syntactic features between a noun X in apposition to another noun Y and the noun Y; e.g., John, an excellent football player—John and Ron, excellent football players, but not *John, excellent football players. 3) AGREE (X,Y): rules of agreement in number between a noun X in the role of an subjective-copredicative complement and a noun Y, the grammatical subject; e. g.,He died an old man ( *old men ) — They died old men (* an old man ). 4) A G R E E R E F L ( N ) ( X , Y ) : rules of agreement in person, number and gender between a reflexive pronoun X and the noun Y it refers to; e.g., the letter itself ( *himself) — Peter himself ( myself) — the schooner herself ( himself). 5) AGREEy( N )(X,Y): rules of agreement in person and number between the main verb X and the grammatical subject Y, which most often (but not always) is a noun; e.g., The numbers have ( *has) f eminine f orms —I am ( *is) running — What bothers me is ( *are) his indifference. Agreement is antisymmetrical, i.e., oriented: it is Y that imposes upon X certain values of the morphological variables of X, and not the other way around. Y is called CONTROLLER, and its own variables are independent in this respect, being either lexically inherent (i.e., components of Y's syntactics, like person and number in ƒ,

194

SURFACE SYNTAX OF ENGLISH

we, etc.) or semantically conditioned (e.g., number of the noun in A G R E E D E T ( N ) ) , or else syntactically conditioned by a third element. Note that the so-called sequence of tenses (Mary said she was<*is>feeling bad) or the concord in mood between the main and the subordinate clauses (If we are (*were) here they will be happy) are taken to pertain to the domain of deep syntax and therefore are not considered here. We believe that a correct surface-syntactic structure must feature a correct sequence of tenses and a correct sequence of moods within a complex sentence (since tense and mood are considered as meaning-bearing morphological values; see above, Chapter III), so that both phenomena fall outside the scope'of surface-syntactic rules, and in particular, of syntagms. It can also be the case that the agreement in number between two nouns (i.e., an appositive and its head noun; a subjective-copredicative complement and the grammatical subject) equally belongs to deep syntax, the nominal number being a meaning-bearing category as well. If so, then the functions and should be discarded. We prefer, however, not to engage in solving this fairly involved problem here, keeping both functions, if only as an illustration. (ii) Inversion Functions 6) INVERS S u b J _ v (X,Y), or (X,Y) : rules stating the conditions under which the grammatical subject X may (or must: if there is the superscript follow the main verb Y, e.g., Do[Y] you[X] understand?.' So absurd was[Y] his answer[X] that everyone started in surprise: Here comes[Y] the nurse[X]. (For a recent discussion of English inversions, with a representative sample of inversion types, see Green 1980; see also Emonds 1976:28-43.) 7) or : rules stating the conditions under which the object or complement X may (or must) precede the governing verb Y, e.g., The problem that[X] we are discussilig[Y]...; exhilarated[X] though Dick was[Y]...; etc. (iii) Coordination Function 8) COORD(X,Y): rules stating the conditions under which items X and Y may be conjoined (correspondence in semantico-syntactic class, in some contextual features, etc.). Cf., e.g., (6): (6)

Sets often figure not only in one's metalanguage for talking about propositions[X] and their truth conditions[Y] but also as elements of content in the propositions themselves.

SYNTAGM

195

In (6), the nouns propositions and conditions can be conjoined, but not *propositions[X] and content[Y1] or *propositiotis[X2] and the propositions[Y2] themselves. Standard functions are stated independently of syntagms. Instead of repeating all the entangled conditions of a specific case of agreement or inversion in each syntagm where they could prove necessary, we simply use the name of the corresponding function specified elsewhere. Although we do not actually describe the standard functions, the rules for agree­ ment and inversion in English, even if not formalized, are pretty well known; so we can hope that the absence of such a description will not impede the reading of our book.

General Conditions for Groups of Syntagms If a condition appears in the -part of more than one syntagm, it must not be repeated but factored out and stated only once, in front of the group of syntagms for which it is valid: see, e.g., General condition on all adverbial syntagms (p.272) or on all parenthetical syntagms (p.291), General condition on Syntagms 13.11 - 13.14 (p.301), etc. This is done both for abbreviation and for a more general formulation of some syntactic phenomena. To be sure, we do not claim that we have succeeded in extracting ALL the repeated conditions from ALL the syntagms, although we have tried to do our best. In addition to standard subtrees, standard functions and general conditions, yet another device of a different nature is used to enhance the over-all transparency of the syntagm system: compounding syntagms. Compounding Syntagms It often happens that the left-hand part of a syntagm appears in several syntagms with different right-hand parts , and usually with different condition parts as well. In such an event, we avoid repeating the same left-hand part in all the i syntagms but, instead, we collapse them all into one COMPOUND SYNTAGM, with one left-hand and several right-hand parts. The resulting syntagm has the following form:

196

SURFACE SYNTAX OF ENGLISH

(7) Rules of the form (7) are used almost exclusively for different word orders in the right-hand parts, so that in most cases i=2. In accordance with the definition of the syntagm as a linguistic sign, we should consider every triple to be a separate syntagm. But we permit ourselves a small digression from the dictates of logical rigor and, par abus de langage, we call expressions of the form (7) simply syntagms, since this cannot lead to any harmful confusion. It is in this broader sense that the term syntagm is used from here on. Concluding this subsection, a purely formal technique of using variables in syntagms must be mentioned and explained. Variables in Syntagms If a node or a branch in a syntagm must carry a disjunction of labels (whose number exceeds two in the case of a node), then, to make the syntagm more readable, the sequence of several disjunctive labels is replaced by a variable, with the values of the latter specified separately (under the graphic presentation of the syntagm). Let us give two examples. Example 1. In Syntagm 15.8 (p.315) the governor of the construction quite [an] X can be one of the following three items: a non-pronominal and non-ordinal adjective {quite a good paper, quite a tiring journey), a noun (quite a philosopher), or an adverb (quite a well prepared lecture). Instead of listing all these three possibilities in the left-hand part of the syntagm, i.e. attaching the disjunctive label to the node, like this: X ( A , not pron, not ord)/(N)/(Adv), we use a variable and write simply

with a specified immediately after the syntagm itself:

SYNTAGM

197

a = (A, not pron, not ord), (N), (Adv). Example 2. In Syntagm 4.3 (p.263), the SSyntRel appearing in the tree context can be either the 1st completive o, prepositional. Instead of writing directly:

specifying r afterwards: r - 1st completive, prepositional. In order to make the manipulation of variables and reference to them easier, we establish the following mnemonic notation: stand for fragments of syntactics; L stands for a set of lexemes; PREP and CONJ stand for a set of specific prepositions and a set of specific conjunc­ tions, respectively; r stands for a set of SSyntRel's; and i stands for the number of an actant. Let it be stressed that the use of variables does not change anything essential ; they are an exclusively notational device and thus stand in sharp contrast to the abovementioned standard subtrees, functions, etc., which are aimed at expressing substan­ tive generalizations. The last topic that remains to be discussed in this section is the classification of syntagms.

198

SURFACE SYNTAX OF ENGLISH 3.

Classification of syntagms.

All the syntagms manifesting the same surface-syntactic relation r are considered to belong to the same SYNTAGMEME and therefore are grouped together, under a common heading, which is the name of the corresponding SSyntRel. This heading appears as a page heading throughout the whole set of syntagms belonging to a syntagmeme. Within a syntagmeme, syntagms are grouped according to some informal consider­ ations: more current syntagms precede less frequent ones; those simpler in form precede the more complex ones; etc. One important criterion for syntagm order is also the word class (= part of speech) of the dependent or the governing member in the syntagms to be ordered. Syntagms describing the same linguistic phenomenon receive a common heading of second (and, if need be, of third) order. Thus we use groupings of syntagms and headings to emphasize all the substantive resemblances, which should help the reader to grasp more easily the linguistic content of our formal rules. The syntagmemes are ordered according to the order of SSyntRel's adopted in §4 of Chapter II (page 87): first, the subordinate syntagmemes, and then the coordinate one; within the subordinate syntagmemes, the sentential syntagmemes come first, followed by the phrasal ones; etc. Let it be noted that a syntagmeme stands to all the syntagms it contains — its ALLO(9-syntagms — in roughly the same relationship as a morpheme to its allomorphs, a lexeme to its allolexes, or a phoneme to its allophones. For more details see Mel'cuk 1982b: 120-122. For each syntagmeme, we indicate repeatability (see page71) of the corresponding SSyntRel: repeatable, non-repeatable, restrictedly repeatable. Very roughly speak­ ing, non-repeatable SSyntRel's include either valence-controlled (=strongly gov­ erned) or determinative relations plus several relations very close to these (such as copredicative ones etc.). Repeatable SSyntRel's are, mostly, modificative, attribu­ tive and adverbial relations.

Syntagms of Modern English As has already been repeatedly stated, this section is the central part of the book, the entire preceding part serving as an extended introduction to it. When compiling the list of syntagms presented below, we relied mostly on our own

SYNTAGMS OF MODERN ENGLISH

199

corpus of examples drawn mainly from scientific and technological texts (dealing with such subjects as linguistics. automatic text processing, computer science. etc.). in addition, numerous specimens of newspaper and magazine prose, as well as some works of fiction were also investigated, which added a number of new structures to our data pool. In our analyses, we have tried to follow Fries 1952, Jespersen 1969 and Nida 1966, as well as some other more recent studies (mentioned wherever relevant). Four Soviet English grammars were consulted quite often: Kacalova and Izrailevic 1957, Barxudarov 1966, Xajmovic and Rogovskaja 1967, and Iofik 1972. Especially useful proved to be the progress report Sager 1968, which contains a relatively complete string grammar of English. (Curiosly enough, in spite of the fact that our approach and that of Sager diverge considerably, both in the general theory and in the for­ malism used, our list of syntagms shows an essential similarity to Sager's grammar — in composition, in general organization. and in the treatment of certain constructions. This is all the more significant since we turned to Sager 1968 only after completing the work on our own list. Maybe the relative closeness of our results to those of Sager can be taken as a good sign indicating that we are on the right track.) The notion of syntagm was presented in the preceding section. We also explained there the principles for ordering syntagms, some conventions used in their presenta­ tions, and so forth. In order to facilitate for our reader the task of locating a syntagrn we do two different things. First, we give immediately below the contents of this section — with page references. Second, we use the names of the corresponding SSyntRel's as page headings throughout the section.

200

SURFACE SYNTAX OF ENGLISH

Guide to the List of Modern English Syntagms Basic Notations and Abbreviations (In what follows the equality sign stands for 'manifested by'.)

212

I. Syntagms covering subordination

216

A. Syntagms covering sentential

SSyntRel's

216

1. Syntagms covering valence-controlled SSyntRel's

216

1. Predicative syntagms 1.1.

1.2. 1.3.

1.4. 1.5. 1.6. 2.

216

Grammatical subject = a nominal phrase

216

Grammatical subject = a gerund, a to-infinitive, or a subordinate clause

219

Predicate = a special verb Predicate = a copula with a special complement . . Grammatical subject = a to-infinitive or a subordinate clause introduced by a wh-word

219 221

Predicate = a special verb Predicate = a copula with a special complement Grammatical subject = anticipatory there

223 224 225

. .

Completive syntagms (objects and complements) 2.1.

223

226

Direct object = a nominal phrase

227

Prepositionless indirect object or equative complement

229

2.2.

Prepositionless indirect object

229

2.3.

Equative complement

231

Retained direct object or equative complement 2.4. 2.5. 2.6. 2.7. 2.8.

. . .

231

Retained direct object Retained equative complement Direct object = a quantity phrase

231 232 233

Complement = a nominal phrase

233

Complement of a like-type adjective Complement of a parametric adjective Complement = a

to-infinitive

233 234 235

SYNTAGMS OF MODERN ENGLISH 2.9.

Complement tive 2.10. Complement 2.11 Complement = 2.12 Complement =

201

of anything except an easy-type adjec­ 235 238 239 239

of an easy-type adjective a bare infinitive a gerund

Complement = an infinitive or subordinate clause

. .

2.13. Complement introduced by a conjunction 2.14. Complement introduced by a wh-word 2.15. Complement with ellipsis of that 2.16. Complement = direct speech 2.17. Complement = a prepositional phrase 2.18. Object or complement = a pronominal-prepositional adverb 2.19. Complement = an adverb 2.20. Complement = an adjective or participle 2.21. Complement = an as-phrase Complement in copula constructions 2.22. 2.23. 3.

241 241 242 243 244 245 248 248 249 250 251

Complement of BE Complement in a construction of the type fool that he is

251 252

Agentive syntagms

253

3.1. Agentive by-complement of a verb in the passive voice 3.2. Adnominal agentive complement 3.3. Agentive for-complement of a to-infinitive

253 254 256

4. Absolute-predicative syntagms Absolute construction as an adverbial modifier 4.1. 4.2. 4.3. 4.4.

Prepositionless absolute construction Absolute with/without-comtruction Absolute construction as an object, a complement or the dependent of a preposition Absolute construction as a descriptive appositive . .

5. Subjective-copredicative syntagm 5.1.

258 . . .

Copredicative = a non-pronominal phrase

6. Pronominal-subjective-copredicative syntagm

258 258 261 263 264 265 265 266

202

SURFACE SYNTAX OF ENGLISH 6.1. 7.

Copredicative = a pronoun

267

Objective-copredicative syntagm

269

7.1.

269

Copredicate = an adjectival

8. Comparative syntagm 8.1.

270

Comparative than/as-phrase

270

2. Syntagms covering SSyntRel's which are not valence-controiled

.

9. Adverbial syntagms 9.1. 9.2.

272

Adverbial modifier = an adverb Adverbial modifier = a nominal phrase with temporal meaning Adverbial modifier = quantity phrase with spatial meaning

9.3. 9.4. 9.5. 9.6. 9.7. 9.8. 9.9.

272 276 277

Quantity phrase denoting location Quantity phrase denoting distance Adverbial modifier = a prepositional phrase, an abso­ lute with/witho-construction or an as-phrase . . . . Adverbial modifier = a present participle Adverbial modifier = a to-infinitive

277 277

Adverbial modifier = an absolute construction

281

. . .

Prepositionless absolute-predicative construction Absolute-conjunctional construction

. .

Adverbial modifier = a clause with an introductory ele­ ment 9.10. A clause with a conjunction 9.11. A clause with a wh-word 9.12. Adverbial modifier = a clause without an introductory element 9.13. Adverbial modifier = a conjunction 9.14. Adverbial modifier = phraseologized construction "X or no X' 10. Modificative-adverbial syntagm 10.1

272

Modificative adverbial = an adjectival phrase

278 279 280 281 282 284 284 285 285 286 287 287

. . . .

288

203

SYNTAGMS OF MODERN ENGLISH 11.

Appositive-adverbial syntagm

290

11.1

290

Appositive adverbial = a NP

12. Attributive-adverbial syntagm 12.1

291

Attributive adverbial = a prepositional phrase, an adverb, as an -phrase or a to-infinitive

13. Parenthetical syntagms

291

Parenthetical phrase = an adverb 13.1 13.2

292

With verbs and within sentential adj ectivals, absolute, coordinative and attributive constructions With nouns, adjectives, or adverbs Parenthetical phrase = a verbless comment clause

13.3. 13.4. 13.5. 13.6.

Of Of Of Of

the the the the

. .

type naturally enough type frankly speaking type to give an example type as shown above

With With With With

an intransitive or passive verb a clause-governing adjective a transitive verb an auxiliary or modal verb

Conjunctionless parenthetical clause 13.11. 13.12. 13.13. 13.14. 13.15. 13.16.

With an intransitive or passive verb With a clause-governing adjective With a transitive communication verb Of the type I am sure what/which-dausc Set off by parentheses or dashes

14. Adjunctive syntagms 14.1. An interjection or a sentential adverb 14.2. An address

292 294 294 294 295 295 296

Parenthetical clause Parenthetical ^-clause 13.7. 13.8. 13.9. 13.10.

291

297 297

.

297 299 300 301 301 302 302 303 304 305 305 306 306 307

204

SURFACE SYNTAX OF ENGLISH

B. Syntagms covering phrasal SSyntRel's

307

1. General

307

15. Restrictive syntagms 15.1. 15.2. 15.3 15.4. 15.5. 15.6. 15.7. 15.8. 15.9.

307

Negative adverb with a verb Restrictive adverb Adadjectival adverb Adverb modifying a comparative Adverb modifying a superlative Adverb modifying a numeral Adjective modifying an adjective Quite Conjunction introducing a disjunctive sentence element

2. Verb phrase

317

16. Auxiliary syntagms 16.1. 16.2. 16.3. 16.4. 16.5.

317

Passive voice Progressive aspect Perfect taxis Future tense; conditional and subjunctive moods Forms with do

. .

3. Word-like phrases

325

17.1. Verb plus verbal adjunct

325

18. Numeral-junctive syntagms (compound numerals) Compound Compound Compound Compound

318 322 323 324 325 325

17. Phrasal-junctive syntagm

18.1. 18.2. 18.3. 18.4.

308 309 311 312 313 314 314 315 316

numerals numerals numerals numerals

including including including including

only numerals . . . . and the indefinite article . fractions

19. Binary junctive syntagms 19.1. Paired conjunctions (both - and, etc.) 19.2. the . . t h e ... -construction 19.3. Paired preposition from- to 20. Colligative syntagms (pseudo-passives with stranded prepo-

327 327 328 329 330 330 331 331 332

SYNTAGMS OF MODERN ENGLISH

205

sitions)

333

20.1. With simple verbs 20.2. With semi-idiomatic verbal phrases

333 335

4. Expletive phrases 21.

336

Expletive syntagms

336

Dummy grammatical subject = anticipatory it 21.1 21.2. 21.3. 21.4. 21.5.

. . . .

The predicate = a verb taking verbal subjects . . . The predicate = a copula with a special complement (entailing verbal subjects) The predicate = a verb taking w/z-phrases as subjects The predicate = a copula with a special complement (entailing wh-phrases as subjects) Constructions of the type It was two months before he recovered Dummy direct object = anticipatory it

21.6. 21.7. 21.8.

Constructions of the type make it clear that Constructions of the type make it clear why Constructions of the type doubt it that Clefting

. . . .

it

339 340 341 342

346 347 348 348 349

Subordinate-conjunctional syntagms 22.1. Subordinate conjuction introducing a clause 22.2. Subordinate conjunction introducing an infinitive 22.3. Compound subordinate conjunction

343 344 345 345

5. Conjunctional phrases

23.

336

342

21.9. Constructions of the type it is NP that 21.10. Constructions of the type it is because ... that . . . 21.11. Dummy grammatical subject = anticipatory there . . 21.12. Dummy agentive complement = anticipatory there .

22.

336

349 . .

350 351 351

Coordinate-conjunctional syntagms

352

23.1. 23.2. 23.3. 23.4.

352 354 354 355

Coordinate conjunctions Quasi-coordinate conjunctions Conjunction but Comparative conjunctions

24. Predicative-conjunctional syntagm

356

206

SURFACE SYNTAX OF ENGLISH 24.1. Comparative conjunctions and conjunction but 25.

. . .

Completive-conjunctional syntagm

357

25.1. Comparative conjunctions and conjunction but

. . .

26. Absolute-conjunctional syntagm

6. Noun phrase

360

Prepositional phrases

360

Preposition governing an NP Preposition governing an adjective Preposition governing a gerund

360 361 361

Preposition governing a clause

362

Via whether Via that Via a finite verb or a

362 362 363

to-infinitive

Preposition governing an adverb or a preposition

. .

27.7. A temporal or locative preposition 27.8. Preposition instead of or except 27.9. Preposition/or governing anticipatory there "Right-shifted" preposition

28.

358 360

27. Prepositional syntagms

27.4. 27.5. 27.6.

357 358

26.1. Subordinate conjunctions introducing an absolute con­ struction ...

27.1. 27.2. 27.3.

356

364 364 364 365 365

27.10. With a wh-word 27.11. With a NP including a wh-adjective 27.12. With a free relative clause 27.13. With where 27.14. Preposition in a sequential phrase with numerals . . . 27.15. "Postpositive" prepositions (ago, notwithstanding, ...) 27.16. Particle to introducing an infinitive 27.17. Preposition besides or except introducing an infinitive

366 366 367 367 368 368 369 370

Determinative syntagms

371

Articles 28.1. 28.2.

Indefinite article Definite article

372 372 373

207

SYNTAGMS OF MODERN ENGLISH 28.3. Demonstrative pronoun

374

Other determiners 28.4. 28.5. 28.6. 28.7. 29.

375

Quantifying or possessive pronoun with a noun

. .

375

Possessive or demonstrative pronoun with a gerund

375

Recurrent determiners in conjoined noun premodifiers

376

Left conjunct Right conjunct

376 377

Quantitative syntagms 29.1. With cardinal numerals 29.2. Quantitative phrase as a compositive dependent

378 . . .

Cardinal numeral modifying a special item 29.3. 29.4. 29.5.

379

An abbreviation A currency unit symbol ( $  , ...) o'clock  percent

379 380 380

Pronominal etc. quantitative words 29.6. 29.7. 29.8. 29.9. 29.10.

381

all or both double, half, etc. before an NP double, half, etc. before a free relative many, more, less manya Quantitative phrase with a personal pronoun

378 379

381 381 382 383 384 . . . .

384

29.11.

With a cardinal numeral

384

29.12.

With all or both

385

30. Modificative syntagms

385

30.1. Standard preposed adjectival modifiers

385

30.2. too Ad] aN

389

Postposed modifiers 30.3. 30.4. 30.5. 30.6.

Expanded adjectival phrase Special single modifier Modifier of a pronoun Single modifier in a set phrase Restrictive relative clause 30.7. Relative clause introduced by a wh-word 30.8. Relative clause introduced by a temporal conjunction

390 390 392 393 394 395 395 397

208

SURFACE SYNTAX OF ENGLISH 30.9.

31.

Relative clause without an introductory element . .

30.10. Infinitival modifier

400

Descriptive-modificative syntagms

400

31.1. Descriptive modifier = an adjectival phrase Descriptive modifier = a relative clause

401 403

31.2. 31.3. 31.4. 31.5.

403 404 405 405

Modifying Modifying Modifying Modifying

a noun an adverb a whole clause a non-referential noun or an adjective

.

32. Possesive syntagms

406

32.1. Possessive marker attached directly to the "possessor" N 32.2. Possessive marker attached to the rightmost word in the phrase of the "possessor" N 33.

Compositive syntagms

33.3.

407 408 409

Compound nouns 33.1. 33.2.

409

Dependent component = an NP Dependent component = a hyphenated phrase or a phrase in quotation marks Dependent component = a sequential or attributive phrase composed of cardinal numerals

409 412 413

Compound adjectivais

413

33.4. N + Adj type (context-free) 33.5. Bahuvrihi type (fair-haired) 33.6. Compound verbs

413 414 415

34. Elective syntagm

415

34.1. Dependent component = an of/among-phrase 35.

398

. . . .

Appositive syntagms

417

Naming appositive 35.1. 35.2. 35.3. 35.4.

Appositive Appositive Appositive Appositive

= = = =

415

aproper noun a name of a man-made object a common noun a numeral or a formula

417 . : . .

417 419 419 420

SYNTAGMS OF MODERN ENGLISH Equative appositive 35.5. Appositive = a non-alphabetical or an autonymous expression 35.6. Appositive = a non-qualifying nomen agentis . . . 36. Descriptive-appositive syntagms

422

Modifying an NP Modifying a personal pronoun Modifying a sentence or an absolute construction

.

Appositive as an explanation etc

37.

421 421 422

Appositive as an additional characterization 36.1. 36.2. 36.3.

209 421

422 424 425 426

36.4. 36.5.

Introduced by an explanatory conjunction Not introduced by an explanatory conjunction . . .

426 427

36.6.

Appositive manifested by an inventory

428

Sequential syntagms

429

Nominal sequence 37.1. Without preposition 37.2. With preposition 37.3. Adjectival sequence Numeral sequence 37.4. 37.5. 37.6.

429 429 430 430 431

to

Without preposition With preposition to With preposition to or through

431 431 432

38. Attributive syntagms

432

Attribute = a prepositional phrase 38.1. 38.2. 38.3. 38.4.

Modifying Modifying Modifying Modifying

432

a noun an adjective or an adverb a superlative or an ordinal numeral . . . a cardinal numeral

Attribute = a prepositionless NP 38.5. 38.6. 38.7.

437

Denoting a parameter Denoting time Denoting distance Attribute = a distributive-measurement phrase

432 435 436 436 437 438 438

. . .

439

SURFACE SYNTAX OF ENGLISH

210

38.8. Modifying a noun 38.9. Modifying an adverb 38.10. Attribute = a reflexive pronoun 38.11. Attribute = a single noun denoting a part of the day . 38.12. Attribute = an absolute construction Attribute = an adverb 38.13. 38.14. 38.15. 38.16. 38.17.

. . . .

443

Adnominal adverb Adverbial modifier (of a noun) else

443 444 444

Attribute = a conjunctional phrase

445

With conjunction as and modifying a noun With conjunction as and modifying an adjective

. .

Attribute = an infinitive phrase

445 446 446

Modifying a noun

39.

439 440 440 441 442

446

38.18. Postposed to the noun 38.19. Preposed to the noun 38.20. Modifying an adjective 38.21. Attribute = a temporal abbreviation 38.22. Attribute (distributive) = the pronominal adjective each, the adverb apiece or the preposition per . . . .

446 448 449 449

Descriptive-attributive syntagms

451

Descriptive attribute = a prepositional phrase 39.1. 39.2.

Modifying a noun Modifying an adverb Descriptive attribute = a prepositionless NP

39.3. Denoting 39.4. Denoting 39.5. Denoting 39.6. Denoting 39.7. Descriptive 39.8. Descriptive 39.9. Descriptive 39.10. Descriptive

a parameter distributed quantity distance localization attribute = an absolute phrase attribute = an adverb attribute = an as-phrase attribute = an infinitive phrase

. . . .

450

451 451 453 453 453 454 454 455 455 456 457 457

211

SYNTAGMS OF MODERN ENGLISH II. Syntagms covering coordination 40.

Coordinative syntagms

458 ,

.

Conjunctionless coordination 40.1. 40.2.

Second conjunct with no restrictive element Second conjunct with a restrictive element

459 . . . .

Coordination with a conjunction 40.3. 40.4. 40.5. 40.6. 40.7. 40.8. 40.9.

458

459 461 462

With a noncorrelative conjunction

462

With no groupings in the conjoined string . . . . With groupings in the conjoined string With groupings in conjoined possessive modifiers . With groupings in conjoined compositive modifiers With a correlative conjunction (both - and, now - now, etc.) With a quasi-coordinate conjunction With a phrase of the type etc

462 465 466 467 468 469 470

212

SURFACE SYNTAX OF ENGLISH Basic notations and abbreviations used in the presentation of syntagms

The notations given here do not include the notations for syntactic features. A complete list of the latter is given, as indicated above, in Appendix I. Elements of the SSyntS (SSS) and the DMorphS (DMS) of a sentence X,Y,Z,... X

: items labeling SSS nodes or the corresponding DMS elements. : a phrase consisting of X and everything syntactically depending on X (= a full subtree of SSS whose top node is X) or the correspond­ ing string in DMS. : an item X which is a lexeme L with syntactics (ξ) and morphologi­ cal characteristics χ NB 1: L, (ξ) and χ may be empty. NB 2: The syntactics are not repeated in the right-hand part of the syntagm, i.e. in the elements of the DMS string, for purposes of abbreviation. NB 3: The meaning-bearing components (= values) of the morphological charac­ teristics χ are present in both parts of the syntagm while the purely syntactic ones are present only in the right-hand part.

: the top node of a full subtree of the Ψ type (A for the Greek ΔENΔPON 'tree'); a comprehensive characterisation of all the relevant types of subtrees is given in Appendix II. : Y depends directly on X in the SSS (via a SSRel r). : Y depends on X in the SSS but not necessarily directly (the SSRel stemming from X being r). : X and Y are coreferential, i.e. they denote the same object, fact, etc. : X linearly precedes Y. : a gap is possible between X and Y, i.e., some lexical material can be inserted between the two items. Thus, X + Y means 'X immediately precedes Y'; X ... Y means 'X and Y may be arranged in any order and be separated by any distance'; and X+ ... + Y means 'X precedes Y but not necessarily immediately'.

SYNTAGMS OF MODERN ENGLISH

213

: the gap is not empty, i.e., it is obligatory. : the gap, if any, contains Z only. : the gap, if any, contains Z, and possibly, some other items. Components of syntactics L(i[i])

: the z'-th deep-syntactic valence slot of the lexeme L can be filled by K, which may be: —either a specific function lexeme (a preposition or a conjunc­ tion), —or a combination of function lexemes, —or a lexeme with the syntactic feature K, —or a certain syntactic construction.

For instance, the notation (z [V + wh]) means that the z'-th deep-syntactic valence slot of the lexeme in question can be filled by a verb on which a wh-word (= a word with the syntactic feature "wh") is directly or indirectly dependent.

PREP CONJ

: the saturation of the z'-th deep-syntactic valence slot is obligatory, except in cases of permitted or obligatory non-saturation. : a particular preposition. : a particular conjunction. Components of the morphological characteristics Variables

n p

: number (sg or pl). : person (1 or 3). Meaning-bearing values

act comp ger inf pass pl posit

: : : : : : :

active voice. comparative degree. gerund. infinitive. passive voice. plural (of nouns). positive degree.

214 ppas ppres sg superi

SURFACE SYNTAX OF ENGLISH : past participle (gone, written, asked) [a cover symbol for both active, i.e. perfect, and passive participles], : present participle (going, writing, asking). : singular (of nouns). : superlative degree. Purely syntactic values

1,3 nom obl pl poss sbjnct sg (V)fin

: 1st, 3rd person. : nominative (direct) case of personal pronouns. : oblique (indirect) case of personal pronouns. : plural (of verbs or pronominal adjectives, cf. these). : possessive form of nouns. : subjunctive mood (Isuggest he be expelled). : singular (of verbs or pronominal adjectives, cf. this). : tensed (finite) verb, i.e. either a verbal node with a predicative SSRel stemming from it (in a SSS) or a verb in a finite form (in the corresponding DMS). Punctuation

cln comma dash hpn lp lqm rp rqm smcln

: colon.

: hyphen. : left-hand parenthesis. : left-hand quotation mark. : right-hand parenthesis. : right-hand quotation mark. : semicolon. Miscellaneous alphabetical abbreviations : rules of grammatical agreement between an inflected determiner and the noun (this view - these views). : rules of grammatical agreement in number and in certain syntactic features between an appositive and the modified noun phrase. : rules of grammatical agreement in number between a subjectivecopredicative noun and the grammatical subject (He died an old man - They died old men).

SYNTAGMS OF MODERN ENGLISH

215

AGREE REFL(N): rules of grammatical agreement between a reflexive pronoun and the corresponding nominal phrase (the letter itself, Pete?' himself). AGREE V ( N ) : rules of grammatical agreement between the main verb and the grammatical subject. COORD : rules of correspondence between the possible surface-syntactic roles, syntactic features, morphological characteristics and the dependents of two conjoined items (the robot and the eye. but not *the robot and eye; his professional and business life). : deep-morphological structure (of a sentence; see p. 165). : rules of inversion of the object with the main verb. : rules of inversion of the grammatical subject with the main verb. : a lexical function having X as its argument, or key word, and Y as its value. : surface-syntactic relation. : surface-syntactic structure. Other symbols : possible or obligatory non-projectivity situations. : the string of word forms X1, X2 X n constitutes a phraseolo­ gical unit. : beginning and end of informal comments. Logical connectives (not, and. if ... then, etc.) are printed in boldface; the slash "/" marks exclusive disjunction — while the comma, for reasons of greater compact­ ness, is used to denote both disjunction and conjunction (but only in contexts which rule out an ambiguous interpretation). Other symbols are either self-evident or of rare occurrence, and in the latter case they are commented upon under the rules in which they appear.

PREDICATIVE

216

I SYNTAGMS COVERING SUBORDINATION A. Syntagms covering sentential SSyntReVs 1. Syntagms covering valence-controlled SSyntRel's

1.. Predicative syntagms The predicative SSRel is non-repeatable. Only a tensed verb (= a finite verbal form) can be the governing member of the predicative SSRel — and from any tensed verb an arrow should stem labeled with the symbol of the predicative SSRel (except in cases of permitted non-saturation).

Grammatical subject manifested by a nominal phrase (1.1)

► The first subrule, i.e. Order (1) — "Subject + Main Verb" — is applicable if no obligatory inversion condition is present in the sentence under consideration (no yes-or-no question, etc.). The second subrule, or Order (2) — "Main Verb + Subj ect" — is possible if at least one of the inversion conditions (stated in the function INVERS S u b j - v ) holds. The expression "not obl" in the right-hand part of the syntagm is an abbreviation having the following meaning: Either Y is a personal pronoun and must be in the nominative; or Y is a noun or the syntactic equivalent of a noun. (In this case the question of case marking does not arise at all.) ◀

PREDICATIVE

217

Examples Order (1) Y = noun (including proper names) or pronoun (1)

Certain assumptions[Y] were[X] made in the beginning.

(2) (3) (4) (5)

Katz[Y]argues[X]that... These[Y] have[X] both substantive and verbal characteristics. EachfY] expressedfX] a different opinion. Oh, mine[Y] is[X] not that hot! f

► As stated in Syntagm 1.1, pronouns in the role of grammatical subject must not be in the oblique case: items like me, her or them cannot be subjects. At the same time, nouns must not be in the possessive form — since we deliberately choose not to describe such elliptical constructions as, e.g., (6): (6)

a. John's is not that hot. b. The speaker's processing of syntax is straightfoward, but the hearer's is generally partial and faulty. c. I do not try to prove that my truth is more true than somebody else's.

We do so in order to avoid unnecessary complications. The same treatment is reserved for nouns in the possessive form in completive and prepositonal syntagms, cf. below, pp. 227 and 360. Y = conjoined phrase (7)

Both a dictionary[Y] and an encyclopedia contributefX] to the explaining of semantic phenomena. Y = formula or abbreviation

(8) (9)

Here (151)[Y] is[X] particularly interesting. VESP[Y] asserts[X] that...

t This is the punch-line of the following joke interesting for its linguistic character (homonymy is used in an essential way): The bar in a hotel for newly-weds on their honeymoon. Early morning. Two young husbands are drinking their cocktails and chatting idly. — By the way. — says one. — where is your wife? — She's upstairs, smoking, — answers the other. — Oh, mine is not that hot!

218

PREDICATIVE Y = adjective (10) The easiest[Y] of all solutions is[X] to ignore the problem. (11) The second[Y] is[X] called CUE. Y = mimerai

(12) Twenty[Y] were[X] given a copy of the two passages. Y = quantitative adverb (13) Enough[Y] has[X] been said to show... (14) ... natural languages, about which relatively little[Y] is[X] known. Y = quantity phrase (15) Over[Y] half the budget is[X] spent on developing the economy. (16) Between[Y] ten (minutes) and twenty hours can[X] be needed. (17) Close[Y] to a million items seems[X] not enough. Y — free relative clause (equivalent to a noun) (18) What we might[Y] call causative actives appears[X] in the positive. (19) They naively assume that whoever is[Y] able to understand written English will[X] also be able to formulate queries in this language. Order (2) ► Basically, the same constituents are possible in the subject role here; we limit ourselves to only a few examples. (20) In any given household, does[X] the set[Y] run on unattended? (21) Only by adhering to a strict formalism can[X] a grammar][Y] be tested in a completely mechanical way. (22) Sandwiched between these studies are[X] seven studies[Y] of morphology. (23) Why was[X] this ratio[Y] the same? (24) Outside, hands thrust deep into the pockets of his jeans, making a bulge like a codpiece, was[X] standing Charles Boon[Y]. (25) Not for nothing had[X] he[Y] taken up residence in a slide area.

PREDICATIVE

219

Grammatical subject manifested by a gerund, an infinitive or a clause introduced by a conjunction (1.2. - 1.3) The predicate is a verb which can take any of the above items as its grammatical subject

the i-th deep-syntactic valence slot of the lexeme is filled on the surface by a clause with the conjunction T H A T 5 and the predicate verb in the subjunctive mood. Cf. : (1)

a. That there is no limit on the number of games is widely known [= (1 5 ]).

vs. b. That there be (should be) no limit on the number of games is urgently required [ = ( l [ T H A T 5 + sbjnct])]. Examples Y=(V)ger (2) (3) (4)

Smoking[Y]kills[X]. Giving[Y] the lines orientation makes[X] the graph directed. Never was[X] their signing[Y] the contract as delayed as last summer.

t The reader is reminded that the list of all lexemes involved in the syntagms is given in Appendix III.

220

PREDICATIVE Y = TO2

(5) (6) (7) (8)

To[Y] have been in the fighting forces merits[X] real rewards. To[Y] deprive them of their games may[X] plunge them into despair. Both to[Y] be a soldier and to be an academic constitute [X] desirable goals. For a bridge to[X] collapse like that is[X] unbelievable.  = CONJ

(9)

That[Y] this interest continues to increase is[X] attested to by the growing number of papers. (10) That[Y] he remain in office was[X] urgently required. (11) Whether[Y] the speaker presupposes that truth contributes[X] to deter­ mining the syntactic form of a complement. (12) Doesn't[X] whether[Y] he came or not interest you?

► It should be noted that Syntagms 1.2 and 1.3 also describe a special construction in which the grammatical subject is manifested by an infinitive introduced by the conjunction WHETHER: (13) Whether[Y] to do it doesn't[X] concern me. (14) Whether[Y] to do it is[X] a question[Z]. [Examples (13) and (14) are from Harris 1963: 8.] Syntagm 1.2 calls for the following four remarks. 1. There exist some verbs which can take a type ß phrase (a gerund, a to- infinitive or a clause introduced by one of the three above-listed conjunctions) as their gram­ matical subject only when they are in the passive voice. When in the active voice, such verbs take this type of phrase as their first (=direct) object. This fact is reflected in their government pattern: every such verb is marked in the dictionary by the feature (2[α]). The change in its government pattern, i.e. (2[α]) => (l[α]), results when the verb assumes the form of the passive voice, see pp. 318-320. 2. Among the verbs described above we can also distinguish those which allow a subject(2[α]-clause,provided that the predicate of this clause is in the subjunctive mood: be advised, be allowed, be arranged, be consented to (decided on,provided for, . . . ) , etc. (Alexander and Kunz 1964:17). This property is recorded in the government pattern of these verbs as (2[THATD + sbjnct]). As indicated above, such entries are transformed when the analytical form of the passive voice is established, resulting in (1[THAT 5 + sbjnct]); in other words, the valence slot number is changed from 2 to 1. Then the phenomenon in question is accounted for

PREDICATIVE

221

in Syntagm 1.2 by Condition 2 and in Syntagm 22.1, at the moment when the subordinate-conjunctional SSRel is established between the conjunction that and the verb it governs. 3. The inversion of the predicate and the subject clause (Order (2)) is ruled out if, as a result of this inversion, an auxiliary or a modal verb immediately precedes the conjunction that which introduces the clause: (15) *Did that John showed up please you? vs. Did the fact that John showed up please you? (16) * Agreed upon by everyone was thatwemust work harder vs. Agreed upon by everyone was the proposition that we must work harder. [Examples (15) and (16) are borrowed from Kuno 1973.] This should be taken care of by the function INVERS Sub _ v . 4. In describing constructions with the anticipatory it as dummy subject, which are similar to the construction provided for in Syntagm 1.2, we make use of the expletive SSRel:

PREDICATIVE

222

a = ger, TO 2 , THAT 5 , WHETHER 1 , IF 2 ; ß = (l[α]),(N,pred-inf),T0 2 Y=(V)ger (1) (2) (3)

Flying[Y] planes can[X] be dangerous[Z] [N. Chomsky]. Their signing[Y] the contract seems[X] probable[Z]. Identifying[Y] the symbols produces[X] lattices of the following form: ... Y = TO2

(4) (5) (6) (7)

To[Y] contrast this study with Saussure's first and greatest publication is[X] very instructive[Z]. Merely to[Y] read and write was[X] a distinction[Z = (N, pred-inf)] at first. For it to[Y] dawn upon her consciousness that she wished for something was[X] definitely to[Z] renounce that wish. To[Y] describe exactly how a semantic dictionary is set up is[X] to[Y] make clear what the form of that dictionary would be.

Y (8) (9)

= com

That[Y] some cells are interconnected randomly may[X] seem quite possible[Z], especially if we consider this possibility together with the fact that... That[Y] the use of a particular set of classifiers does not imply belief in a best set may[X] be obvious[Z]

► What was said in Remarks 2 through 4 in the comments on Syntagm 1.2 holds, mutatis mutandis, for this syntagm as well.

PREDICATIVE

223

Grammatical subject manifested by a to-infinitive or a subordinate clause (1.4-1.5)

The predicate is a verb which can take any of the above items as its grammatical subject

(1) (2)

Which[Z] way to[Y] choose[W] must[X] be decided by the algorithm itself. How[Z] to[Y] protect[W] him began[X] worrying his mother. Y = (V)fin

(3) (4) (5) (6)

How[Z] conditionals are[Y] presented[W] depends[X] upon the analysis. What[Z] he means[Y] is[X] specified in this way. What[Z] was[Y] usually said of him did[X] not deserve mention. Which[Z] way of erecting the intermediate image will[Y] be used in the experiment greatly influences[X] the results.

► Some verbs are distinguished by their ability to take a wh-phrase as their grammatical subject only in the passive voice. This case is similar to the one with gerund as subject, etc., see comments on Syntagm 1.2, Remark 1, p. 220.

224

PREDICATIVE

The predicate is a copula verb with a complement which takes any of the above items as its grammatical subject

Y =TO 2 (1)

Which[Z] possible chain of inferences to[Y] select[U] should[X] become the overall theme[W] of our work. Y = (V)fin

(2) (3) (4)

Exactly who[Z] is[Y] responsible for this proposal is[X] unclear[W]. Why[Z] a specific quantifier suddenly turns[Y] nonspecific because another 'wide scope' quantifier follows it immediately is[X] enigmatic[W]. Why[Z] the SCD system tackles[Y] only a subproblem of the resolution of word-sense ambiguity seems[X] quite explicable[W].

PREDICATIVE

225

Dummy grammatical subject manifested by the anticipatory THERE (1.6)

All verbs labeled (v-there) are intransitive verbs implying a vague idea of existence, coming into existence or manifesting existence. An informative study of the syntactic behavior of different constructions with THERE 2 is found in Ross 1974 (with further references). See also Allan 1971 and Sheintuch 1980. Examples (1) (2) (3) (4) (5)

There[Y] certainly[Z] are[X] rules for expressing propositions. There[Y] is[X], incidentally, now a certain amount of syntactic evidence to indicate that in the deep structure noun phrases are not that diverse. There[Y] are[X] plenty of people getting promotion. There[Y] was[X] a young plumber of Leigh, Who was plumbing a girl by the sea. And usually there[Y] just[Z] wasn't[X] anybody to send a bill to.

Note that the there-is/are-construction, as well as all the other verb phrases with the anticipatory there (i.e., all the examples below), are analysed here as follows: there itself is considered to be the dummy grammatical subj ect of the main verb while the actual subject (= the NP following the verb) is assumed to depend on there via the expletive SSRel, cf. Syntagm 21.11, 348. This proposed analysis aims, first, at providing a homogeneous description of all constructions with anticipatory it and

t The end of this nice limerick is as follows: She said, 'Finish your plumbing. There's somebody coming!' Said the plumber, still plumbing, It's me.'

PREDICATIVE / COMPLETIVE

226

there (including the dummy agent construction of the type for there to be any busi­ ness), and second, at avoiding the need to treat the NP following the verb as formal grammatical subject (since otherwise it would be the only type of grammatical subject in English admitted in postposition to the main verb without any concurrent inversion conditions). The agreement of the verb X with the NP following it is provided for in Syntagm 21.11. (6) (7) (8)

TherefY] seem[X] to exist [= (V, v-there)] some hairy problems. TherefY] has[X] long existed a need for a complete description of English. Over the edge there[Y] peeped[X] a boyish face [A. Conan Doyle].

Examples (9) and (10) below are less acceptable than the others because they are sensed more as verbs in the passive (evoking a missing "agent") than as a copula + adjective phrase, just stating the existence of their subject: (9) (10) (11) (12) (13)

There[Y] are[X] also needed suitable training materials for handicapped children. For a moment there[Y] was[X] written in her eyes the desire and the determination [Th. Dreiser]. There[Y], just inside[Z] the door, stood[X] a wide shallow tray full of pots of pink lilies. TherefY], way back[Z] in the shadows, was[X] a big chunk of snow and the idiotic smile of Vamenos. And there[Y], all over[Z] the town, were[X] the Armenians from Bitlis, Van, Moush, and Diarbakir. Order (2)

(14) What other reason is[X] therefY] for judging this sentence to be anomalous? 2. Completive syntagms All completive SSRel's (1st through 4th) are non-repeatable. The government pattern of a lexeme L may include an indication of the obligatory presence of the ( i - l ) t h object or complement of the lexeme, i.e., of the (i—l)th completive SSRel, namely, the feature (i [oblig]). If a SSS node is labeled by a lexeme L ( i ) [ o b l i g ) ] , t h e (i-l)th completive SSRel must stem from it, except in cases of permitted or obligatory non-saturation of the corresponding strong valence slots.

COMPLETIVE

227

Direct object manifested by a nommai phrase (2.1) 2.1.

► The expression "not nom" in the right-hand part of this syntagm is an abbrevia­ tion of the same type as in Syntagm 1.1. Its meaning is as follows: Either Y is a pronoun and must be in the oblique case. Or Y is a noun or the syntactic equivalent of a noun. (Then it is not marked for case at all.) Cf. the comment on page 216. Condition 2 says that if a pronoun in the role of a direct object has no dependent, it should not be separated from the governing verb by any wordform. Clearly, this statement is an oversimplification. First, there are some adverbs that can (in certain contexts, at least) separate the verb from its pronominal object: (1)

We saw first him, then Ian.

Second, there seems to be a special case of the general word order rule of English that allows the separation of direct objects (not necessarily pronominal ones) and verbs only if the direct object is 'heavy' enough relative to the separating phrase (cf. page 189, the second paragraph of item 2): (2)

a. He sold bonds very quickly. b. *He sold very quickly bonds. c. He sold very quickly the largest bonds issued last year.

This general rule is related both to the global word order rules of English and to the completive syntagms. Therefore, we have chosen to mention it in Syntagms 2.1 and 2.2, even though in a highly approximate form. On the other hand, Condition 2 might prove helpful for the purposes of automatic

228

COMPLETIVE

analysis. Since a pronoun normally has no premodifiers (except words such as ONLY), it will almost always be contiguous to the verb. But a noun can be separated from the verb by its own left-hand dependents. It should be clear that Condition 2 is nothing but a rule of thumb, which badly needs a more accurate reformulation. Examples Order (1) Y = noun or pronoun (3) (4) (5)

These rules incorporate[X] allegedly linguistic knowledge[Y]. There is no wall separating[X] the person's language[Y] from his world. He reads[X], although detective stories[Y] only.

Note that we consider detective stories in (5) to be the direct object of reads: a solution which is obviously open to discussion (wouldn't it be better, for instance, to treat (5) as a kind of elliptical construction?). (6)

You must know[X] her[Y]. Y = syntactic equivalent of a noun phrase

(7) (8)

Let us take[X] now the easiest[Y] (of these). In New York dogs deposit[X] more[Y] than 125 tons of embarrassment on private lawns, public parks and sidewalks daily.  = free relative clause

(9)

Yet judicious use of men and machines may accomplish[X] very well what neither is[Y] now able to do alone because of inherent limitations.

(10) He also associates[X] what he calls[Y] degree of grammaticality with competence. Order (2) (11) What symptoms[Y] can we find[X] that reflect our naive approach to language? (12) There are two other four-letter words which[Y] sober reflection demands that we bar[X] from this book.

COMPLETIVE

229

Prepositionless indirect object or equative complement manifested by a nominal phrase (2.2 - 2.3)t Prepositionless indirect object

► Condition 1 prohibits both direct and indirect objects being personal pronouns simultaneously: (1) but

a. *Mary sent you me [= 'Mary sent me to you']; b. Mary sent you something.

Notice that (la) is possible with a heavy stress on me, but we do not account for such subtleties. Condition 2 requires that a personal pronoun indirect object having no dependent follow the verb immediately: (2) but

a. *They gave, as you might know, him a new appointment in the cabinet; b. They gave, as you might know, this unscrupulous charlatan a new appointment in the cabinet.

Let it be remembered that our comments on Condition 2 of Syntagm 2.1 (page 227-228) fully apply here as well.

f See Fillmore 1965 and Brittain 1971.

COMPLETIVE

230

Condition 3 stipulates that the direct object cannot be placed between the verb and the indirect object: either the direct object follows the indirect one, or the direct object precedes the verb (under an inversion, as in (7)). Examples (3) (4) (5) (6) (7)

She has served[X] the children[Y] their dessert[Z]. It took[X] the IBM 704[Y] over[Z] half an hour. It takes[X] people[Y] between[Z] 10 and 15 minutes. I'll show[X] that arrogant fellow[Y] I am[Z] as good as any of his customers. The book that[Z] they have given[X] him[Y] turned out to be very expen­ sive. (8) He gave[X] me[Y] what he was[Z] eating. (9) Can you get[X] this guy[Y] a job[Z]? (10) He gave[X] the door[Y] a kick[Z] /... a couple[Z] of kicks. (11) John gave[X] the table[Y] a fresh coat[Z] of paint.

► Note that the inversion of a prepositionless indirect object is impossible (or highly undesirable) in Modern English—unlike a direct object or, for that matter, a prepositional object: (12) a. *Who(m) did he give an umbrella? b. ?*The guy who(m) they offered the scholarship failed the final exam. c. *Only me would he give an umbrella. Cf. possible inversion in (13): (13) a. b. c. d. e. d.

Whom did he see? To whom did he give an umbrella? Whom (Who) did he give an umbrella to? The guy to whom they offered the scholarship failed the final exam. Only me would he meet at the airport. Only to me would he give an umbrella.

See Kuroda 1968. (14) If someone gives[X] us[Y] the language segment[Z] 'Mr. Smith is angry' and asks what it is really about, then it seems reasonable to reply 'a human being under discussion is in a certain emotional state'.

COMPLETIVE

231

Equative complement

Examples (1) (2) (3) (4) (5) (6) (7)

We then appointed[X] Joan[Z] secretary[Y] to the committee. He has taken unfair advantage of her by keeping[X] his illness[Z] a sec­ ret[y]. They made[X] her son[Z] what he is[Y]. R. Kennedy whom[Z] they had elected[X] senator[Y] These[Z] we call[X] fusionai properties[Y]. They called[X] their daughter[Z] something[Y] outlandish. Yes — the Captain ; we never call[X] him[Z] anything[Y] else [H.Walpole].

Retained direct object or equative complement with passive verbs manifested by a nominal phrase (2.4-2.5) Retained direct object

232

COMPLETIVE

► Syntagms 2.4 and 2.5 describe the only case when a past participle which is not part of an analytical perfect form can govern a prepositionless nominal phrase. Examples (1) (2) (3) (4) (5)

After a three month delay, the family was awarded[X] the money[Y]. Alice will be given[X] John[Y] as a partner under the condition that she accepts him. The program GENERTree should be given[X] the symbol[Y] S as its input. The fight against rent increases was given[X] further stimulus[Y] and encouragement yesterday. My father, so long denied[X] the promotion[Y] he wanted, left the company.

► In fact, in certain contexts some speakers (of American English) admit retained prepositionless INDIRECT objects with (adres) verbs in passive, especially when these objects are pronominal: (6)

a. The money given us recently comes from the Ford foundation. b. The promotion, so long promised him, came at last.

But even for the sentences of type (6), many speakers who accept them prefer the variants with a preposition, so that sentences (6') (6') a. The money given to us recently comes from the Ford foundation. b. The promotion, so long promised to him, came at last. are judged to be superior, from the grammatical viewpoint, to (6). Since there is such strong disagreement among native speakers concerning the acceptability of constructions with retained indirect objects we choose not to consider them at all. That is why Syntagm 2.4 provides only for retained direct objects.  Retained equative complement

COMPLETIVE Example Joan, when appointed[X] secretary[Y] of the committee, was 22.

Direct object of a measure verb manifested by a quantity phrase (2.6)

Examples (1) (2) (3)

It will cost[X] five dollars[Y]. The line measures[X] under[Y] two inches in length. How much[Y] does it cost (weigh) [X]?

Adjective complement manifested by a nominal phrase (2.7-2.8) With like-type adjectives 2.7.

233

234

COMPLETIVE

Examples (1) (2) (3) (4) (5) (6)

The iota operator is like[X] every other linguistic device[Y] in that its interpretation is relative. It is worth[X] the trouble[Y]. Finally, thanks are due[X] the Department[Y] of Spanish at Vanderbilt University, especially Charles Vance, chairman of the department. ... Auntie Sadie, whom[Y] nobody would want to look like[X]. This system of analysis is not worth[X] any discussion[Y]. This is a feature of an on-line LISP system like[X] the one[Y] in use at System Development Corporation.

With parametric adjectives or adverbs

► Condition 2 stipulates that if the adjective X is itself a preposed modifier of a noun W, then its complement Y must be in the singular.  Examples (1) (2) (3) (4) (5)

three feet[Y] high[X]; a three-foot[Y]-high[X] device[W]; a few months[Y] old[X] ; a six-month[Y]-old[X] child[W] ; a 1000 ft[Y]-wide[X] telescope[W; here, ft = foot]; a stack three meters[Y] high[X] ; a three-meter[X]-high[X] stack[W] ; two framing pulses, 20[Z] seconds[Y] apart[X].

COMPLETIVE

235

Complement manifested by a to-infinitive (2.9 - 2.10) Ato-infinitivegoverned by anything except an easy-type adjective

► The requirement 'not (A, obj)' imposed on X excludes from this syntagm constructions of the type a problem easy [=( A, obj )] to solve, which are described by another syntagm — 2.10 — due to some special conditions which must be stated with regard to their infinitive. Cf. also constructions of the type it seems very important to consider..., etc., in which the adjective has the feature (1 [TO2]) in its government pattern; Syntagm 21.2, p. 339. Condition 1 reflects the fact that some verbs can govern ato-infinitiveonly when they are in the passive voice: (1) He is said to be ill, but not () *They say him to be ill. A special feature appears in the government pattern of such verbs; e.g., SAY is marked in the dictionary as (2 [TO2] | pass), which means "governs ato-infinitiveas its second actant but only when used in the passive voice". Condition 2 provides for expression of the form (2):

COMPLETIVE

236

(2)

a. b.

Here, nouns such as fool or dumbbell contain the feature in their government pattern indicating the ability to govern an infinitive when they themselves appear as complements with a copula verb. Condition 3 covers the for+N+to+Vinf construction in the role of a comple­ ment, cf. examples (11) and (16). Condition 4 permits taking into account the following regular word order pattern in English: in cases when a certain word has two actants simultaneously — one an infinitive and the other a prepositionless noun, the former always follows the latter, e.g.: (3)

We want[X] your friend[W] to[Y] finish it,

but never (3') *We want to[Y] finish it your friend[W]. As for constructions of type (4): (4)

a. To make so much money is clever of him. b. It is clever of him to make so much money. c. He is clever to make so much money.

Thus adjectives of the clever type in the constructions of (4) are considered to be auto-conversives, i.e. they have two different government patterns in the dictionary, namely, (1[T02], 2[OF]) and (2[T02), providing for the correct treatment of phrases of the clever to make type (see (4'c)) in Syntagm 2.9. (All the other surface-syntactic relations of clever in (4') are described elsewhere, by the corresponding syntagms.) Bolinger 1961 also mentions an open series of constructions with assorted adver­ bial intensifiers such as awfully, devilishly, excessively, overly, pretty, terribly, unduly, etc. which resemble the constructions with TOO or ENOUGH covered by this syntagm:

COMPLETIVE (5)

237

a. It's terribly late to go out now. b. It's pretty cold to expect the snow to melt today. c. You are awfully small to say things like that.

Cf. (5) with (6): (6)

It's too late to go out now.

We believe, however, that these two constructions are different with regard to syntax. While in (6) we have a genuine case of strong government (the presence of the infinitive phrase is semantically motivated: 'too (much)' — for what? ; 'enough' — for what?), (5) shows an infinitive phrase in the role of a free modifier, namely, a specific adverbial. Therefore, constructions of type (5) must be described by an adverbial syntagm; see Syntagm 9.7, p. 280. -4 Examples (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20)

A third kind of nucleus remains[X] to[Y] be examined. The connection asserted[X] by the speaker to[Y] hold between A and  is causal in nature. Inquiries are scheduled[X] to[Y] take place into other councils in Scotland. Condition (2) was made[X] instead to[Y] partially define the levels of semantic representation. She couldn't bear[X] for them not to[Y] be friends. I don't want[X] there to[Y] be any misunderstanding. His desire[X] to[Y] go away... The shortest superfixes are more likely[X] to[Y] be word-superfixes. She is too[X] shy to[Y] come. Such expressions are usually accompanied by enough[X] pragmatic clues for their semantic ambiguities to[Y] be easily resolved. It was the most unlikely[X] thing[U] in the world to[Y] happen. These are important[Q] enough[X] methodologically[R] to[Y] warrant examination. His desire was[X] to[Y] be sent to the conference. The best plan is[X] to[Y] engage in no transactions at all.

► In examples (19) and (20) the verb BE plays the roles of Func2 (desire) and Func 3 (plan), respectively, and borrows its government pattern (the ability to govern a io-infinitive) from its key word. Note that in the construction dealt with in Syntagm 2.9, the infinitive depending on to may be omitted under lexical identity (leaving only to behind):

238

COMPLETIVE

(21) Custom demands that the two men fight for her even if she does not want[X] them to[Y]. This fact should be accounted for in the general rules of permitted nonsaturation of obligatory active valence slots, see above, p. 81ff.

A -infinitive governed by an easy-type adjective

► Condition 1 eliminates sentences like It is not easy to make this distinction, since they are described in terms of the expletive SSyntRel (between it and to make) and are covered by Syntagm 21.1, p. 339.  Examples (1) (2) (3) (4)

The distinction is not easy[X] to[Y] make[Z] /... not an easy[X] one[T] to[Y] make[Z]. The easy[X] man[T] to[Y] convince[Z] is John. A journey through Iraq was difficult [X] to[Y] dream[Z] of[U]. a. Candlelight is impossible[X] for me to[Y] study[Z] by[U].

COMPLETIVE

239

b. This type of light is impossible[X] for us to[Y] study[Z] given our time and space limitations. ► The construction described by this syntagm (and known in transformational grammal as the result of Tough-Movement) is analyzed in more detail in Berman 1973. 

► The condition stipulates that a nominal or prepositional object of a verb necessarily precedes its infinitival obj ect (in this case, a bare infinitive). Cf. Condition 4 in Syntagm 2.9 and the corresponding comment on page 236.  (1) (2) (3)

Few people can[X1] afford[Y1] to go through life listening[X2] to[Z] the birds sing[Y2]. These phrases must[X] now be[Y] assigned a function. The condition of Harris need[X] not hold[Y] in an axiomatic syntax.

240

COMPLETIVE

► Condition 1 is necessary because of the following facts. Sentences (la, b) are ungrammatical: (1)

a. *John was continuing[X] drinking[Y] vodka all night. b. *John was starting[X] going[Y] to concerts frequently.

But (2) is fine, since enjoy[=X] is not a (phas)-verb: (2)

John was enjoying[X] going[Y] to concerts frequently.

See Milsark 1972. Under Condition 2 a gerund may appear as a complement of a copula only in cases when its grammatical subject or its for-agent is in the singular and is not the pronoun I. Without this condition sentences of the type (3)

Airlines shouldn't even be paying for them

could be wrongly analysed as (3')

Examples (4) (5) (6)

He began[X] reading[Y]. We pardoned[X] them having[Y] done it. This is a jolly game worth[X] encouraging[Y].

► Note that there are some verbs which govern the verbal noun ending in -ing but not the gerund: (7)

It would be hard to duplicate his tactful handling of the situation

but not (7') *It would be hard to duplicate his handling the situation tactfully [Sager 1968: 86]. (8) The biggest mistake[Z] was[X] my writing[Y] the letter. 

COMPLETIVE

241

Complement manifested by a clause or an infinitive phrase (2.12- 2.15) With a conjunction

► In this syntagm, it is usually a clause introduced by one of the above-listed conjunctions that manifests the complement. However, with WHETHER an infinitive phrase is also possible, see example (10).  Examples (1) (2) (3) (4) (5) (6) (7)

The listener will say[X] of anything he regards as wildly improbable that[Y] it is false. He was informed[X] that[Y] the battle had begun. The definition takes[X] for granted that[Y] this condition has only one kind of surface exponent. This approach requires[X] that[Y] all the prefixes be[Z] (should[Z] be) accounted for. T h e statement[X] has[W] been made many times that[Y] these techniques should not be extended indiscriminately. The professor finds her case so[X] interesting[W] that[Y] he presents her in a clinical demonstration. The extensional definition can be transcribed in such[X] a way[W] that[Y] it is made intensional.

242

COMPLETIVE

(8)

The question[X] arises whether[Y] rocks are natural combinations of mineral matter. (9) Everybody wonders[X] if[Y] he will do it. (10) Dick wonders[X] whether[Y] to do so would be acceptable to him.

With a wh-word

Examples (1) (2) (3) (4) (5) (6) (7)

Les us ask[X] what[Z] the advantages are[Y] of using a natural language. Simple transactional analysis is concerned with diagnosing[X] which[Z] state implemented[Y] the transactional stimulus. The task then is to discover[X] how[Z] deeply organized such a phenomenon might[Y]be. The sentence states that the reason[X] why[Z] he beats[Y] her is not that he loves her. The professional knows[X] when[Z] to[Y] stop. An expectation table tells[X] us what[Z] to[Y] look for in the next words. In choosing[X] how[Z] to[Y] approach a natural phenomenon like language, we define our field of study.

► The infinitive phrase in this syntagm is always nonprojective:

(8) the corresponding type of non-projectivity is accounted for by Syntagm 27.16, Condition 3. 

COMPLETIVE

243

Examples (1) (2) (3)

I hope[X] many others will[Y] see the red light. The workers said[X] they would[Y] not end their strike. I wasn't sure[X] I hadn't[Y] already met him.

► Note that a conjunctionless complement clause is also possible with some nouns: (4) (5) (6)

I have the impression[X] they are[Y] still waiting. One reason[X] pastimes are[Y] so stereotyped is that they serve such stereotyped purposes. The correlation of the document with the request is increased in the event[X] the document is[Y] relevant to the request and their correlation is decreased otherwise.

The feature (2[fin]) is essential in the government pattern along with (2[THAT]) because there are verbs which govern a that-clause but which do not permit ellipsis of THAT: acknowledge, choose, disbelieve, prefer,... [Alexander and Kunz 1964: 1-5]: (7)

I should prefer that you did not go there alone.

but not (7') *I should prefer you did not go there alone. In connection with completive clauses, the following particularly interesting syntactic phenomenon of English deserves mention: relativization and questioning are admitted out of an embedded object clause which is part of a relative or other subordinate clause — provided the former is conjunctionless. The relative or inter­ rogative word (wh-word) occupies the left-most position in the resulting two-clause string; therefore, if it is not deleted by a later rule, such a construction necessarily

244

COMPLETIVE

entails non-projectivity (stated in Condition 3). Cf. the following examples: (8)

He could hardly wait to come to the destroyed city (that[U]) he had never thought[X] he would[Y] see again. (9) They decide what[U] crime the state shall claim[X] occurred[Y]. (10) Each of her friends placed that small blue star where[U] he thoughtp[x1] it belonged[Y 1 ] on the Dorothy Parker he thought[X 2 ] he knew[Y 2 ]. (11) They may have lost their chance to convict the men they suspect[X] ordered[Y] the killing.

In (11), the w/z-word (= who) in the role of grammatical subject is deleted. However, outside the construction in point, this deletion is considered substandard: (IV)*... to convict the men ordered the killing. Cf. a similar case on page 399, example (5). (12) Troops occupied the university following what[U] the regime claimed[X] was[Y] a bomb explosion in the engineering faculty. (13) The analysis contains loose ends (which[U]) Mr. Katz has to admit[X] he cannot[Y] yet tie up. (14) What[U] the students were taught[X] was[Y] going to happen has not happened.

Direct speech (2.16)

► The feature (2["SENT"]) in the syntactics of a lexeme means that the lexeme may have direct speech as its first complement. We do not attempt to cover all existing possibilities of punctuating direct speech. For instance, sometimes dashes are used instead of quotation marks, etc. We roughly characterize only one of the current conventions.

COMPLETIVE

245

Order (1) (1) (2)

The commander shouted[X]: "Stopp[Y]! One could but would not ordinarily answer[X1], 'They are[Y1] flying saucers" to the question[X2], "What are[Y2] these things?" or even "What are they?" Order (2)

(3) "I"11[Y] shoot," the commander warned[X]. (4) "I heard[Y] the crash," he said[X] later, "and[Z] felt the crack of the ceiling."

Object or complement manifested by a governed prepositional phrase (2.17)

► The general condition of Syntagm 2.17 excludes from consideration construc­ about and He was taken good care tions of the type She was spoke described by Syntagms 20.1 and 20.2. Comprehensive data on prepositional government in English is given in Borkovskaja et al 1973. 

COMPLETIVE

246

Examples Order (1) (1) (2)

... the obvious concern[X1|] about[Y 1 ] the ease[X 2 ] of[Y2] use by casual users. This throws[X] light on[Y] why sociological studies of criminals have been unproductive.

► Example (2) shows that a phraseological expression like throw light etc. can have, as a whole, a specific prepositional government.  (3) (4)

The piece[X1] of[Y1] information thus acquired is the result[X 2 ], in this qualified sense, of[Y2] the experiment. The use[X1] of [Y1] dived is much less reacted to than that[X 2 ] of[Y2] struck.

► The syntactic relation between X 2 and Y 2 in example (4) is established according to a general rule of English syntax, which provides that the pronominal substitutes ONE 2 and THAT 3 borrow their government pattern from their antecedents.  (5) (6) (7) (8) (9)

They can be recognized independently[X] of[Y] their verbal content. One of the necessary distinctions is[X] between[Y] words which pattern with determiners and words which do not. The progress made[X] in[Y] resolving ambiguity ... / The most significant progress has been[X] in[Y] resolving ambiguity. He determines the use he will make[X] of [y] those opportunities. Britain's working-class movement has been asking itself what lessons there are[X] for [Y] Britain in what happened in Chile.

► Examples (6) through (9) indicate that Oper's, Func's, Labor's and complex LF's composed of these, as well as the phrase there is/are borrow their government pattern (GP) from their key words or from the grammatical subject, respectively. So, make = Oper1(progress) and remain = ContFunc 2 (progress) strongly govern the preposition in in conformity with the GP of the noun progress, while there are strongly governs the preposition for in conformity with the GP of the noun lesson. (10) A different[X] proposition[T] from[Y] this was unacceptable. (11) She bought the book for me to amuse[X] myself with[Y]. (12) These games are the most important ones for the professional analyst to be aware[X] of[Y]. (13) A finite number of features is still hard for anyone to conceive[X] of[Y]. ► Examples (11) - (13) serve to illustrate an obligatory ellipsis in English: a nominal phrase depending on a strongly governed preposition is deleted in the SSyntS

COMPLETIVE

247

under referential identity within certain infinitive phrases: (ir)*She bought the book for me to amuse myself with it => ...to amuse myself with. (12')*These games are the most important ones... to be aware of them => ... to be aware of. (13')*A finite number of features is still hard for anyone to conceive of it => ... to conceive of. The ellipsis in question might be described by a rule having roughly the following form:

Cf. *She bought[U] the book[W2] for me to amuse[X] myself with[Y] it [W1]. Let it be recalled that the dotted two-headed arrow in the left-hand part of Rule (A) represents coreferentiality; see above, p. 212. Obviously, rules of type (A) do not belong to the list of syntagms and should not be considered in this book. The following remark may not be out of place here: In structures like (11) through (13) a non-saturated preposition is subordinated to its governor via a completive SSRel rather than being subsumed under the colligative SSRel as in is collig with or being lied ^about, see Syntagm 20.1. The similarity between the two constructions is superficial while the difference between them is quite significant. In the first instance the insertion of a pronoun is possible (in the surface-syntactic structure, and not in the actual sentence — because of the above-mentioned obligatory ellipsis), but it is absolutely ruled out in the second instance. ^

COMPLETIVE

248

(14) It is important to investigate what the basic intention[X] is[T] of[Y] this proposal so vehemently put forward by so many of his followers. Order (2) (15) Of [Y] more importance are[X; Oper ¿importance)] informal rituals. (16) On[Y] everything she met with Mary had[R; Oper1 opinion)] a firm preconceived opinion[X]. Object or complement manifested by a governed pronominalprepositional adverb (2.18)

Example He uses an imaginative modification[X] thereo[Y]. Complement manifested by a governed adverb (2.19)

/ = 2,3,4,5; o = loc, temp, manner

COMPLETIVE

249

Examples (1) (2) (3) (4)

Hestays[X]there[Y]. It began[X]early[Y]. He behaves[X]badly[Y]. In (14) the experiencer is clearly well qualified to be called[X] so[Y].

Complement manifested by a governed adjective or participle (2.20) 2.20.

Examples Y=(A) (1) (2) (3)

Its two solid little chimneys were painted[X] red[Y] and white. He fell[X]sick[Y]. It rings[X]true[Y]. Y=(V)ppres

(4) (5)

We saw[X] the engine being[Y] carefully packed in cases. Japp's footsteps were heard[X] descending[Y] the stairs[A. Christie]. Y=(V)pass

(6) (7)

We hope to see[X] this issue raised[Y] in all trade councils. A lexeme which is inherently[- Time] has[X] the feature[+ Time] transfer­ red[y] to it.

250 (8) (9) (10) (11)

COMPLETIVE The psychological advantage comes from having[X] the body mutilated[Y]. If you would like[X] your paper published[Y] next week, send a copy now. How did this window get[X] opened[Y]? My packet of cannabis got[X] found[Y] by Fido, the police-dog.

► As the presence of examples (10) and (11) here suggests, phrases of the type get opened and got found are not taken to be instances of the grammatical category of voice in English. (Some include them in this category under the name of gei-passive.) Rather, they are considered as free combinations of GET with a participial comple­ ment in much the same way as, for instance.

Complement manifested by a governed conjunction AS (2.21.)

Examples (1) (2) (3)

It is possible to identify[X] these fractions by their morphology as[Y] words of various classes. Attempts to point out the obvious truth strike[X] generative semanticists as[Y] preposterous. ... its status[X] as[Y] a sentence adjunct.

COMPLETIVE

251

Complement in particular constructions with a copula verb (2.22 - 2.23) Complement of BE

α = (N), (V)inf, TO 2 , (Prep), (Adv), (Conj, subord) ► Constructions of the type be of great importance or seem in good condition are not described in this syntagm since the verbs be, seem, etc. appear in them as lexical functions and, as such, strongly govern the corresponding prepositions.  Examples Y=(N) (1) (2)

It[Z] is[X] they[Y] who are responsible for it. The surface system[Z] is[X] a mixed one[Y].

252

COMPLETIVE Y=(V)inf (3) (4)

Essentially what[W] a rule tester does[Z] is[X] enable[Y] one to trace the consequences... A11[Z] (The only thing[Z]) (that) he has done[U] is[X] take[Y] the other literally. Y=T02

(5) (6) (7)

To[Z] win the game is[X] to[Y] win the series. What[W] Dick dreams[Z] of is[X] to[Y] travel to Japan. The first thing[Z] (that) he wants[U] is[X] to[Y] win the series. Y = (Prep)

(8)

It[Z] is[X] since[Y] Tuesday that he has been working here. Y=(Conj,

(9)

subord)

If Black does nothing, it[Z] is[X] because[Z] he feels helpless.

Complement in inverted structures 2.23.

Examples (1) (2)

Strange[Y] as[Z] it may[X] seem, the rule proves very helpful. Tired[Y] as[Z] they were[X], they continued on their way.

AGENTIVE

(3) (4)

253

Unhappy[Y] about the disclosure though[Z] Nixon is[X], he shows no inclination to resign. Fool[Y] that[Z] he is[X], he can't do it.

3. Agentive syntagms The agentive SSRel is non-repeatable.

Agentive by-complement of a verb in the passive voice (3.1)

► In this syntagm, the feature (2/3[BY]) in the syntactics of the node X refers not to the basic government pattern of the verbal lexeme in the node X, i.e. not to the government pattern registered in the dictionary, but to the converse government pattern derived in accordance with the rules specifying the correspondence between the lexicographic government pattern of a verb and the derived government pattern of the same verb in the passive voice, see pp. 318-320. Note that the derived govern­ ment pattern is used in the case of verbs in the passive voice not only in this syntagm but in a number of other syntagms as well, e.g., in predicative (1.2,1.4) and expletive ones (21.1, 21.3). Two problems arise in the process of describing English agentive complements with verbs in the passive voice: 1. Distinguishing agentiveby-complements from instrumental and other comple­ ments (with various prepositions) of passive verbs as well as from complements of adjectives formed from past participles. The range is extremely wide, from the type (1)

The hole was dug by John vs. The hole was dug with a big spade,

which is obvious enough, to complicated ones, like (2) through (4): (2) (3)

The sky was covered with clouds {= The clouds covered the sky ). Mary was preoccupied with this idea ( = The idea preoccupied Mary).

254 (4)

AGENTIVE She was amazed at (by) his success {= His success amazed her).

2. Identification of verbs allowing / not allowing an agentive by -complement, when in passive. Certain restrictions do obviously exist. For instance, an agentive by-complement is ruled out for some verbs in the passive voice: (5) *The water was contained by the pool, as well as in some phrasemes: (6) *A good look at the exhibition was taken by him. This restriction should be taken into account in the dictionary entries of verbal lexemes by means of a special syntactic feature that would block the application of the corresponding rule, pp. 318-320, namely, the transformation N => BY[ + N]. Since a special investigation is required to solve the above two problems, we have opted to completely refrain from considering them.  Examples Order (1) (7) (8) (9)

This version of the grammar was written[X] by[Y] Mr. P. Culicover. A software package developed[X] by[Y] our library proved quite successful. I remain totally baffled[X] by[Y] tenses and quantifiers. Order (2)

(10) Now let us ask by[Y] whom can this language be described[X] in the first place. (11) This expression corresponds, in its form, to the rule by[Y] which both headless relatives have been converted[X] into lexically headed CNP.

Adnominal agentive complement (3.2) 3.2.

AGENTIVE

255

Examples (1) (2) (3)

(4) (5) (6) (7)

.. .the complete disregard[X] of the members by[Y] the union's leaders. .. .attemps[X] by[Y] the managements to keep their eating houses open... ...the advent[X] of[Y] transformational grammar; the stability[X] of[Y] government finances; strong support[X] at the match from[Y] the capital's big Australian colony; That silly dancingfX] of[Y] theirs[Z] was energetic. That constant piano playing[X] of[Y] his[Z] is driving me wild. *That silly dancingfX] of[Y] its[Z; e.g., of a marionette] was energetic. That constant piano playing[X] of[Y] Dick's ( Sally's ) [Z] is driving me wild.

► In examples (4) and (5), adapted from Fraser (1970: 88), data on governing the preposition of by the regular mg-nominalst are derived from the government pattern of the respective verbs by a special rule which can be tentatively formulated as follows (cf. the rules for modifying the government pattern in the passive voice, pp. 318-320): When a N ing (V) is formed (by attaching the suffix -ing to a verb stem), the feature (l[OF]) is to be introduced in its syntactics, i.e. in its government pattern.  The condition describes the possible use of an agentive of +- possessive phrase quite approximately: — It does not make precise the syntactic and semantic circumstances under which such a phrase is fully idiomatic. Thus it seems important that the ing-nominal be modified by that: (4')

· The silly dancing of theirs was energetic [cf. (4)].

— It does not reflect the fact that the possessive form of a first name is less acceptable than a possessive pronoun, a polysyllabic first name less acceptable than a monosyllabic one, a last name still less acceptable, etc.  (8) (9)

The shooting[X] of[Y] the hunters[Z] [= 'the hunters are shooting'] was loud. The chirping[X] of[Y] the birds[Z] kept everybody awake.

t We mean here the of which introduces an agent NP.

AGENTIVE

256

Agentive for-complement of a to-infinitive (3.3)

Examples (1) (2) (3) (4) (5) (6) (7) (8)

I don't like for[Y] Lizzie to[Z] play[X] in the street. It would have taken at least a month for[Y] the committee to[Z] come[X] up with their findings. The tendency is [= Func2(tendency)] for[Y] the laymen to[Z] call[X] profes­ sional procedures rituals. This chap is very anxious for[Y] me to[Z] see[X] you [E. Hemmingway]. She couldn't bear for[Y] Tom and Mary not to[Z] be[X] friends. I'd have given anything for[Y] that not to[Z] have[X] happened. It is highly improbable for[Y] there[U] to[Z] appear[X] any new popula­ t i o n s ] of the kind discussed in the two previous chapters. It is too late for [Y] there[U] to[Z] be[X] any business[T] now.

► In examples (7) and (8), there is taken to be a dummy agent depending on for and governing — via the expletive SSRel — a real agentive NP (which follows the verb X). The surface-syntactic structure of (7) is, then, as follows:

Cf. Syntagms 1.6, 21.11 and 27.9.4 (9) She had her arms wide apart, as if for[Y] the child to[Z] run[X] into them. (10) He brought these books for[Y] you to[Z] examine[X].

AGENTIVE

257

► The agentive construction of the form FOR + NP + TO 2 + (V)inf is possible in nearly all syntactic roles where a to-infinitive is possible: as a grammati­ cal subject (Syntagm 1.2, example (8)), as a complement (Syntagm 2.9, example (16)), and as an adverbial modifier indicating purpose, determination, etc. (Syntagm 9.7, example (6)). Of course, it is not admitted in contexts that require that the agent of the infinitive be identical with the subject of the governing item, as in (11): (11) *I am capable for John to do this. Here, the agentive /or-phrase is ruled out for semantic reasons, and these reasons must not concern us now. However, the construction in question is not used in Standard English with some verbs that do govern a to-infinitive and do not require identity of their subject with that of the infinitive: (12) *I don't want for Lizzie to play in the street; cf. example (1) above. This means that the property of taking a for + N + to + V inf construction as a complement must be stated explicitly in the government pattern of the respective lexemes and a special condition must be introduced into Syntagm 2.9 [= Condition 3], for cases like (12) to be rejected by our grammar. A list of verbs satisfying the formula V + for + N + to + Vinf, as opposed to verbs satisfying only the formula V + to + V inf is found in Alexander and Kunz 1964: 78-79. It should be strongly emphasized that there are considerable differences among dialects in regard to which verbs can govern a for-to construction. Therefore, some of our examples (e.g., (1), (5) or (12)) are felt by some speakers to be questionable. Notice the semantic difference between (13a) and (13b): (13) a. Mugabe has asked for the British to stay on a little longer [=*... that the British stay on...']. b. Mugabe has asked the British to stay on a little longer. (13a) does not imply that Mugabe has asked the British themselves: he might have asked, e.g., the U.N. This difference is shown at the surface-syntactic level as follows:

258

4.

AGENTIVE / ABSOLUTE-PREDICATIVE

Absolute-predicative syntagms The absolute-predicative SSRel is non-repeatable. Absolute construction as an adverbial modifier (4.1 - 4.2)

Prepositionless absolute construction

α = (Ν), (A), (V)ppres, (V)pass, (Prep), (Adv, adsub), TO 2

ABSOLUTE-PREDICATIVE

259

Examples Order (I) Y=(N) (1) (2)

And Gala, her body[X] a twisted black potato crisp[Y] amongst a million others, had already fainted[Z]. We were now launched[Z] into an unfriendly, dead world, its attractions[X] those[Y] of a lunar landscape.

Y=(A) (3) (4)

Punctually at seven, his mouth[X] dry[Y] with too many cigarettes the night before, he forced[Z] himself out of bed. The man laughed[Z], his voice[X] unnatural[Y]. Y = (V)ppres

(5) (6)

(7)

The semantic component provides[Z] input to the syntactic component, syntax[X] being[Y] interpretative. A daring liar, the Pretender contends[Z] with the murderer of an innocent child, the wheeling, perfidious, cowardly boyars[X] cringing[Y] to both sides. a. There[X] being[Y] nobody else to answer, Clare went[Z] out. b. *There[X] coming[Y] nobody else, Clare went[Z] out.  = (V)pass

(8)

This[X] integrated[Y] in the only possible way, the duplicates are[Z] removed.

(9)

The poor husband, his game[X] broken[Y] up, will[Z] go into a state of despair. Most people have[Z] both friends and enemies, both[X] perhaps justified[Y] in their feelings. A guarantee[X] of the poet's taming and good conduct afforded[Y] the tsar by his marriage proposal, the drama could[Z] appear in print at the end of 1830.

(10) (11)

Y = (Prep), (Adv) (12) Johnny, his eyes[X] still toward[Y] the table, took[Z] one cut.

260

ABSOLUTE-PREDICATIVE

(13) The robber was[Z] found dead on the floor, his gun[X] nearby[Y]. Y = TO2 (14) The company offered[Z] the buyers 500 tons of oil, delivery[X] to[Y] be made in October. Order (2) (15) A paired structural condition must[Z] meet conditions of compatibility, chief[Y] among them the condition[X] which belongs to the theme of recoverability of deletions. (16) Whatever[Y] the physiological origin[X] of the need to drink, in terms of game analysis imbibing is[Z] merely a move in a game. (17) And Mr. Stalin, however[U] horrendous[Y] his deeds[X], simply expanded[Z] what the apostle had already begun. (18) I cannot follow[Z] his defence of this as implied conjunction reduction, however[U] undramatic[Y] the action[X] described. (19) However[U] well prepared[Y] the track[X], the frequent movement of men and stores up and down it would[Z] always be a source of anxiety. ► As for constructions of the type Given the evidence of this sort, one can easily show ..., etc., we hold that given is a special preposition capable of governing either a noun or a that-clause: (20) Given two readings R1and R 2 , such that R 1 is assigned to a node N 1 and R 2 is assigned to a node N 2 , the derived reading R 3 is assigned to the node N. (21) Given that a phrase marker such as (2) represents the structure that our rules must describe we can ask what sort of rules can produce such phrase markers. We have come to treat such constructions in the above manner because, first, no other preposed participles could be found in the absolute-predicative construction, and, second, because given is capable of governing an entire clause via that. Cf. provided (that), traditionally defined as a conjunction. 

ABSOLUTE-PREDICATIVE

261

Absolute with/without-construction

α = (Ν), (A), (V)ppres, (V)pass, (Prep), (Adv, adsub) TO 2 Examples Y=(N) (1)

With[W] conscription[X] a law[Y], we must[Z] decide...

Y=(A) (2) (3)

She receives[Z] almost as much as her husband, with[W], of course, one[X] less[Y] in the family to feed and clothe. The most probable structure is[Z] always the simplest one, with[W] related phrases[X] adjacent[Y]. Y = (V)ppres

(4) (5)

With[W] Hamada[X] making[Y] his third appearance in Britain in seven years, this is[Z] a time for interesting comparisons. We left[Z] without[W] Mary (her)[X] noticing[Y].

262

ABSOLUTE-PREDICATIVE (6) (7) (8)

Without[W] NP[X] and VP being[Y] adjacent, an Achenese sentence cannot[Z] be correct. Clearly, without[W] all these organic-chemistry equations[X] being[Y] fully memorized, students do[Z] not receive good grades on a science exam. Without[W] there[X] being[Y] enough substantive evidence, we cannot[Z] proceed with our discussion.  = (V)pass

(9)

The town has[Z] been almost totally destroyed, with[W] hardly a buiLding[X] left[Y] unmarked by bombs, shells or rockets. (10) Pushkin constantly projects[Z] tropes into a poetic reality, with[W] the objects[X] creating[Y] the "content" of a poem, and with[W] metonymic and metaphoric relations[X] therefore reified[Y]. Y = (Prep) (11) With[W] his body[X] at[Y] the wheel of his car, his mind is[Z] at the door of his office. (12) Bunch was[Z] on the floor with[W] her arms[X] round[Y] her knees. (13) This is usually an enjoyable pastime; with[W] a few originals[X] in[Y] the group it can[Z] become quite amusing. Y = (Adv) (14) He entered[Z] the room with[W] his coat[X] off[Y]. Y = TO2 (15) Then with[W] no occupations[X] to[Y] fill in her time, she began[Z] to examine her behavior more closely. (16) With[W] all these observations[X] to[Y] be published next year, Tycho's heirs were[Z] understandably furious.

ABSOLUTE-PREDICATIVE

263

Absolute construction as an object, a complement or the dependent of a preposition (4.3)

r = 1st completive, prepositional Examples (1) "I hope you don't mind[Z] me[X] coming[Y]?" asked Kit [J. Lindsay]. (2) The problem is[Z] man[X] refusing[Y] to try. (3) The notion of[Z] theories[X] being[Y] variants of each other allows for equivalent theories to differ in the way that they say what they say about language. (4) Here the risk is from[Z] this game[X] being[Y] culturally syntonic. (5) The slightest circumstance may give rise to the most violent altercation which may end with[Z] the daughter[X] being[Y] expelled from the house in the middle of the night. (6) We insisted on[Z] the contract[X] being[Y] signed immediately. (7) This results in[Z] two speechless people[X] glaring[Y] at each other. ► When an absolute-predicative construction is an object, a complement or the depedent of a preposition (other than WITH or WITHOUT), its own dependent element, i.e. Y in this syntagm, is restricted to a present participle only (although when this construction is an adverbial, as in Syntagms 4.1 and 4.2, Y can also be a noun, a past participle, an adverb, etc.). Of course, examples can be found in which Y is a past participle, an adverb or a preposition even in constructions described by Syntagm 4.3:

ABSOLUTE-PREDICATIVE

264

(8) Justification is[Z] man[X] declared[Y] righteous by God. (9) Do you mind[Z] me[X] here[Y]? (10) It resulted in[Z] two people[X] under[Y] police surveillance. However, first, all such sentences are elliptical, and second, they are limited to highly specific contexts; cf. (8') - (10'), which are less acceptable than their counterparts: 99

(8')

· · As their main point, they discussed man declared righteous by God [better: ... discussed man being declared righteous by God]. (9') a. · Do you object to me here? [better: ... to me being here?] b. · Do you mind him with Academic Press? [correct: ... mind him being (working, staying,...) with AP?] (10') · · Anybody will object to his friend under police surveillance [better: ... to his friend being under police surveillance]. Since we are unable to formulate grammaticality conditions for sentences of type (8) - (10) we have chosen to omit them from our description.  An absolute construction as a descriptive appositive (4.4) 4.4.

α = (Ν), (A), (V)ppres, (V)pass, (Prep), (Adv, adsub), TO 2

ABSOLUTE-PREDICATIVE / SUBJECTIVE-COPREDICATIVE

265

Examples (1)

(2)

5.

Two young women[Z] from Cairo University, one[X] a student[Y] of sociology and[W] the other a student of commerce, came collecting for their medical aid fund. We obtain three discontinuous suffixes[Z], the first[X] from[Y] Column 2, the second[W] from Column 4, and the third from Column 5.

Subjective-copredicative syntagm The subjective-copredicative SSRel is non-repeatable.

Non-pronominal copredicative phrase

α = (A, not pron), (V)pass, (N, not pron) not poss Examples Y=(A) (1)

Mr. White saw[X] the bills first[Y].

(2) (3) (4)

The flags in the square hung[X] wet[Y] from the white poles. Base prices have been pegged[X] high[Y] for years. She tried to jerk[X] loose[Y]. Y=(V)pass

(5)

This sense cannot occur[X] embedded[Y] as a complement of verbs like try

SUBJECTIVE-COPREDICATIVE

266

(6)

or force. She walks[X] through the bedroom half dressed[Y].

Y=(N) (7)

He[W] resigned[Z] his office and died[X] an old man[Y].

► Syntagm 5.1 also covers the use of verbs labeled (ocopr) in the passive voice. An (ocopr)-verb takes a secondary predicate that characterizes its direct object (see Syntagm 7.1); but if this verb is itself passive, then the secondary predicate becomes a characterization of its grammatical subject, so that instead of the objective-copredicative SSRel we otain the subjective-copredicative SSRel. Cf.:

(8)

The cigarette was smashed

See two further examples: (9)

The town has been almost totally destroyed, with hardly a building left[X] unmarked[Y] by bombs, shells or rockets. (10) With the cap on their flare container frozen[X] shut[Y], they weren't able to contact the plane. 

6.

Pronominal-sub]ective-copredicative

syntagm

The pronominal-subjective-copredicative SSRel is restrictedly repeatable: two arrows of this SSRel may simultaneously stem from one node on condition that one arrow enters a reflexive pronoun, and the other one, a quantitative pronoun. E.g.: (1)

The sentences which count as negative all[Y1] contain[X] either not or the negative words no, none, etc. themselves[Y 2 ].

The quantitative pronouns ALL, BOTH, EACH, EITHER and NEITHER as dependent members of the pronominal-subjective-copredicative SSRel are called floating quantifiers; reflexive pronouns MYSELF,..., THEMSELVES (in a capacity of an intensifying, or emphatic, pronoun) are called floating reflexives.

PRONOMINAL-SUBJECTIVE-COPREDICATIVE

267

Pronominal copredicative phrase (6.1) 6.1.

Examples Y = (quanti/) (1) (2) (3) (4) (5) (6) (7)

a. These links[Z] are[X], it seems, all[Y; = ALL1 ] of the same class. b. These links[Z] all[Y] seem[X] to be of the same class. This[Z] is[X] all[Y; = ALL 3 ] nonsense. LBC[Z] and Capital have[X] each[Y] (each[Y] have[X]> an annual budget of 5 million. They[Z] were[X] neither[Y] of them trying to put it over on us. The two dismissed players[Z] — Bob McCarthy and Gary Stevens — both[Y] easily escaped[X] suspension. John[Z] and Dick each[Y] immediately bought[X] a car. My students[Z] all[Y] passed[X] this exam.

Y=(N,refl) (8) (9)

Not playing[Z] games may[X] itself[Y] be a game ( may[X] be itself[Y] a game/may[X] be a game itsel[Y] ). A clause[Z] containing subject, predicator and object (as specified above) itself[Y] functions[X] (functions[X] itsel[Y] ) as a single clause element.

268

PRONOMINAL-SUBJECTIVE-COPREDICATIVE

Conditions 4 and 5 are necessary because of ungrammaticalities such as (10): (10) a. *We enjoyed[X] all[Y] this movie / * We enjoyed this movie all / *We have enjoyed this movie all. b. *They buy[X] each[Y] lots of books / *They buy lots of books each / *They may buy a lot of books each. Sentences (10'), with X an auxiliary or a modal verb, are quite normal: (10') a. We have[X] all[Y] enyoyed this movie. b. They might[X] each[Y] buy lots of books. So are (11): (11) a. We all[Y] fully enjoyed[X] this movie. b. They each[Y] buy[X] lots of books.  We enjoyed[X] this movie ourselves[Y]. See Emonds 1976: 215, 239-241 and Stockwell et al. 406 ff., for more details. Cf. also (12), where a floating quantifier appears after a non-finite form of an auxiliary verb [Condition 5]: (12) a. · Your friends should[X] have each bought that book/*... should have bought that book each, a'. Your friends should each have bought that book. b. * Your parents must[X] be being both severely criticized/*... must be being severely criticized both. b, Your parents must both be being severely criticized. Notice that X need not be a finite verb form: (13) (14) (15)

Each[Y] provided[X] with a particular marker, the four chunks[Z] are[W] stored in the rapid-access memory. Being[X] ourselves[Y] of Greek origin, we[Z] can[W] easily understand the plight of the Soteriadis family. The five interpreters[Z], each[Y] accompanying[X] an officer on a mission, have[W] already left for the Sinai.

The possibility of X not being a finite verb form is provided for by Condition 1. As it happens, the conditions of Syntagm 6.1 allow more positional freedom for reflexive pronouns (than for floating quantifiers ALL, EACH, etc.): the conditions stipulate that a pronoun of theSelf-seriescan either precede or follow its verbal head, independently of the latter's nature. But this freedom is not absolute, and possible positions for reflexive pronouns should be specified with sufficient precision, which

OBJECTIVE-COPREDICATIVE

269

is not done in this version of the syntagm. Moreover, the quantifiers NEITHER (of them) and EITHER are also different from ALL and EACH in their positional abilities, a fact not accounted for. Thus Syntagm 6.1 does no more than scratch the surface of the problem. It looks like it would be advisable to split it into two or even more syntagms (e.g., one for floating quantifiers and one for floating reflexives).

7. Objective-edicative syntagm The objective-copredictive SSRel is non-repeatable. 7.1.

α = (A, not pron, not ord), (V)pass Examples Y=(A) (1) (2) (3) (4)

Frank likes[X] Mary[Z] fat[Y]. They burned[X] her[Z] alive[Y]. Mr. Blum hates[X] gin[Z] diluted[Y]. They showed[X] it[Z] cracked[Y].

► Cf. (4) with a related example (5):

(5)

They

quoted and analysed in Sager 1968: 93. 

COMPARATIVE

270 8. Comparative syηtagm

The comparative SSRel is non-repeatabkle. 8.1.

α = (A/Adv)comp, (2[c-than]), (2[c-as]) ► The features (2[c-than]) and (2[c-as]) in the government pattern of lexemes of the type other [than] or the same [as] denote their ability to govern, via the compara­ tive SSRel, a phrase which is introduced by the conjunctions THAN and AS 1 respectively, and which fills the second deep-syntactic valence slot of the above lexemes.  Examples X = (A/Adv) comp (1)

This distinction is easier[X] to feel than[Y] to express.

(2) (3)

It was much smaller[X] than[Y] the cement industry wanted. He gave general secretary Terry Parry much more[X] time[Z] to speak than[Y] was afforded the Glasgow representative. The opposite case can[W] no more[X] be[U] handled in such grammar than[Y] CKP type grammar can handle surface structures.

(4)

X = (2[c-than]) (5) (6) (7)

They contain no expression form other[X] than[Y] what appears on their lexical expression list. Seeing and hearing have a different[X] quality[Z] for infants than[Y] for grownups. These rules express empirical generalizations about the syntactic structure of English, rather[X] than[Y] describe syntactic structure of specific English sentences.

COMPARATIVE

271

► In accordance with tradition, the word rather in example (7) is taken to be an adverb with the feature (2[c-than]) in its government pattern, serving as the depenent component of the adverbial SSRel.  X = (2[c-as]) (8)

As [X; = AS ] many[Z] samples as[Y] possible of the game are then collected. (9) Many criminals seem to get as[X] much[Z] satisfaction from outwitting the police as[Y] from their criminal games. (10) The order of modifiers is such[X] as[Y] to require a terminal. (11) This relation is given by such[X] relations[Z] as[Y] consist in the correspon­ dence of small and capital letters. (12) The time of the main clause must be the same[X1] as[Y1] or later[X 2 ] than[Y 2 ] that of the if-clause. ► Examples (11) and (12) provide an instance of permitted non-saturation: example (11) shows the finite verb consist after as without grammatical subject, and example (12) demonstrates non-saturation of the node Y1 within a phrase with coordination.  (13) Structures of this sort are so[X] frequent and well-known as[Y] hardly to need illustration. ► In the surface-syntactic representation of semantically contrasting phrases of the type "a comparative conjunction + a personal pronoun", as exemplified in (14) and (15): (14) [Mary likes Joan more] than he and (15) [Mary likes Joan more] than him we consider it logical to use two different SSRel's:

So two SSRel's, predicative-conjunctional and completive-conjunctional, have to be introduced; see below, Syntagms 24 and 25. 

ADVERBIAL

272

2. Syntagms covering SSyntRel's which are not valence-controlled

9. Adverbial syntagms The adverbial SSRel is repeatable. General condition on all adverbial syntagms An adverbial modifier Y cannot be placed between a preposition and the nominal phrase the preposition governs (while a parenthetical modifier or a restrictive adverb can; see examples below, Syntagms 13.1, examples (14) - (15), and 15.2, examples (5)and (9)):

Adverbial modifier manifested by an adverb (9.1) 9.1.

► 1. Y can be manifested not only by individual adverbs but also by set adverbial

ADVERBIAL

273

phrases, such as etc. 2. The linear position of several adverbial modifiers of the same verb with regard to each other is determined by certain complicated rules discussed in Jackendoffl972:87-93. E.g., (1)

Often Max has carefully been trying to decide whether to climb the walls,

but not (Γ) *Carefully Max has often been trying ... The corresponding regularities should be accounted for by the general word-order rules of English (see p. 180ff). The syntax of English adverbs and adverbial phrases has been thoroughly studied by Nilsen (1972); cf. also Connors 1971 and Jacobson 1964. Further references concerning English adverbs can be found in Sabourin 1977.  Examples X=(V) Order (1) (2)

Let us go[X] into this more concretely[Y].

(3) (4) (5)

Such verbs can co-occur[X] with manner adverbiais rather freely[Y]. We are[X] now[Y] in a position to explain ... He hit[X] her very hard[Y].

(6) (7)

a full semantic theory. Using[X] a dictionary with part-of-speech information rather[Y] than a part-of-speech algorithm would help in some cases. The information content Η1 estimated[X] therefrom[Y] is about 13 bits/ allophone.

(8)

Order (2) (9) Such verbs can freely[Y] co-occur[X] with manner adverbiais. (10) ... the now[Y] much[Y2] publicized[X] system; artificially[Y] restricted[X] set; (11) Normally[Y] time adverbs have[X] the same environments as manner adverbs. (12) Quickly[Y] though[Z] I walked[X], I could hardly catch up with them. (13) Hitherto[Y], few of these efforts have[X] involved an attack on obviously linguistic problems.

ADVERBIAL

274 (14)

linguists are[X] devoting more attention to problems of meaning.

( 1 5 ) W e watched him to Russia's Aeroflot for the flight to Leningrad. X — quasi-V÷ (16) (17) He will take results.

[X] disconcerting

(18) Natural '[X] of logic to the linguistic patterns, should not be taken as a sound base for our study. (19) The concept is defined for independent lexemes for[X] lexical functions. (20) The adventures are listened to with fascination and[G] often[Y] with[X] enjoyment. (21) Exocentric patterns could be handled by analyzing the first and[G] then[Y] the second constituents[X]. (22) While[G] at[X] our place yesterday[Y] he refused everything. (23) In the later stages[G], due[X] sometimes[Y] to organic deterioration, the persecution can be dispensed with. ► The following explanations regarding Syntagm 9.1 and all the other adverbial syntagms seem appropriate at this point. The regular, basic function of an adverbial, or circumstantial, modifier consists in modifying a verb; accordingly, it is usually a verb which is the governor in adverbial syntagms (i.e., the master of the adverbial SSRel). However, many adverbial modifiers with various inner structures, such as individual adverbs, prepositional phrases, certain adverbial clauses, etc., may form part of certain non-verbal construc­ tions (absolute-conjunctional phrases; various sentential, or disjunctive, modifiers; and attributive phrases) as well as of strings of conjoined constituents, e.g.:

t For the variable G in examples (16)—(23) see comments on the abbreviation "quasi-V", p. 275.

ADVERBIAL

275

(24) The president and, from the last week on, his lawyers have been continuing their effort to slow down the inquiry. (25) Many new features never before in any almanac! [in an advertisement] (the adverbial phrase never before forms part of the attributive phrase never before in any almanac). Obviously, in these and in all similar cases the adverbial modifier can by no means be subordinated to a verb. Two alternative solutions are possible: (a) Examples of type (24) - (25) are regarded as instances of surface-syntactic ellipsis, i.e., such sentences are considered to be syntactically incomplete. The "missing" verb, recovered from the preceding context, is restored in their SSyntS, and the adverbial modifier under consideration is attached to this verb. (b) Examples of type (24) - (25) are considered to be syntactically complete. In this case, for the purposes of automatic analysis it would be convenient to subordinate the adverbial modifier (even if provisionally) to the head of the phrase; these surface-syn­ tactic links would be used to compute the actual, more semantically oriented relation­ ships to be established in the deep-syntactic or the semantic representation. We have opted for solution (b), since the cited sentences, intuitively, seem to be syntactically complete. The non-verbal subordination of adverbial modifiers in our syntagms is provided for in the following manner: To describe compactly the characteristics of node X, which is capable of subordinating adverbial modifiers and parenthetical phrases without being a verbal form, the abbreviated notation "quasi-V'' is used. The expression

points to one of the following three instances: (i) Node X is the dependent member of an absolute construction:

where r = abs-predic, abs-conjunct. (ii) Node X is the head of a sentential adjectival, adverbial, etc. :

where r = i-th compi, agent, sub-copred, obj-copred, mod-adverb,

276

ADVERBIAL

appos-adverb, descr-modif, descr-appos, descr-attr. (iii) Node X is the dependent member in a construction in which Y is located between X and X's governor:

where r = preposit, attrib, appos, adverb, coord-conj, coordin; if r = coordin, then comma Y. In all these cases, Χ Φ (Conj, coord), (Adv, delim). What has been said about the non-verbal subordination of adverbial modifiers also holds for parenthetical phrases; the abbreviation "quasi-V" frequently appears in parenthetical syntagms as well. 

Adverbial modifier manifested by a nominal phrase with temporal meaning (9.2)

Examples X=(V) (1) (2)

(3) (4)

We[W] finished[X] the paper last[Z] week[Y]. a. They[W] have been watching[X] him all[Z] night[Y] (three[Z] hours[Y]). b. All night[Y] they[W] have been watching[X] him. c. *They[W] all night[Y] have been watching[X] him. The next[Z] day[Y] we are[X] going to Cambridge. When speaking[X] to her Friday[Y], I found it difficult to make myself understood.

ADVERBIAL

277

Χ = quasi-V (4)

While[G] in[X] Poland this[Z] summer[Y], I read it with the help of my

(5)

friend. It displays the same symbolic relationship[G], this[Z] time[Y] between[X] the veterans and the administration.

Adverbial modifier manifested by a quantity nominal phrase with spatial meaning (9.3-9.4) Quantity phrase denoting location 9.3.

Examples (1) (2) (3)

He lives[X] over[Y] three miles away[Z] (from[Z] here). He is dying[X] just a block[Y] from[Z] here. Only three miles[Y] from[Z] here (there) lives[X] a man who can change all that. Quantity phrase denoting distance covered

9.4.

ADVERBIAL

278

Example He walked[X] four miles[Y].

Adverbial modifier manifested by a prepositional phrase, an absolute with/without-construction or an as-phrase (9.5) 9.5.

Examples X=(V) Order (1) (1) (2) (3) (4) (5) (6) (7)

This law is[X] only true under[Y] certain restrictions. However, information is[X] prediction in[Y] that it communicates what is unknown. But ultimately the two seem to be[X] inseparable, with[Y] meaning derived from syntax and syntax determined by meaning. Postwar Italian governments have fallen[X] apart for[Y] almost every conceivable reason. It can be shown[X], but only in[Y] outline, that room has to be made for some interpretation at the surface. Mr. Cosell does not take[X] the floor except[Y] to rank himself as the best sports commentator. Weyl rejects an attempt to base[X] simplicity on probability, without[Y] the afore-mentioned class of functions being taken into account. Order (2)

(8) (9)

Like[Y] a normal code, spoken language has[X] redundant features. In[Y] the original materials, the sentences on the left bear[X] no relation to

ADVERBIAL

279

those on the right. (10) To[Y] some scholars, the study of meaning has[X] been effectively identified with the study of informal logic. (11) Of more significance as[Y] an introduction to game analysis are[X] informal rituals. (12) They are intended as an understanding system, and, as[Y] such, justify[X] a new attack on this problem. X = quasi-V (13) While[G] on[X] a flight over[Y] France last December, he met one of his classmates. (14) She may use[G] the affair, with[X] the help of a lawyer in[Y] some cases.

Adverbial modifier manifested by a present participle (9.6) 9.6.

Examples (1) (2) (3) (4) (5) (6) (7) (8)

I looked back and saw Rinaldi standing[X] watching[Y] me. At each stage, ambiguities on that level are[X] resolved, finally yielding[Y] just one structure for each sentence. Following[Y] his line we are[X] developing an encyclopaedia. He leaves[X] the house, feeling[Y] cheerful and relieved. Feeling[Y] sick, I tried[X] to think of some constructive action. "Look", said[X] Mrs. Niesthal, noticing[Y] Cassidy. Helen crouched[X] in a chair watching[Y] him. They laid[X] him on the sofa, undressing[Y] him together.

280 (9)

ADVERBIAL Emptying[Y] out the fluid, he set[X] the bowl on the floor.

► It is obvious that the mutual arrangement of the present participle and its governor is by no means arbitrary or optional. Condition 3, which requires that the present participle adverbial follow a non-finite verb form, as in (1), is obviously not sufficient. But we do not know the factors behind the right choice of position for a present participle adverbial and so we have to leave our formulation at that. It is not excluded that the preposing or postposing of a present participle adverbial is linked to semantic factors — in which case it would be necessary to split the adverbial SSyntRel into two different SSyntRel's in order to represent different meanings of the preposed and postposed -ing-adverbials.  (10) Returning home on Le Bourgogne in 1889, his ship collided with a British vessel off Nova Scotia. ► (10) exemplifies an unrelated participle, i.e., a participial phrase whose deep subject (= HE) does not coincide with the subject of the main clause (= HIS SHIP). See above, §4 of Chapter II, p. 109. 

Adverbial modifier manifested by a to -infinitive, with or without a conjunction (9.7) 9.7.

Examples (1) (2) (3)

He painted[X] the scenery to[Y] look like trees. These transformations apply[X] here as well to[Y] produce the correct surface structure. They may just let their feelings grow[X] naturally, to[Y] be plucked when

ADVERBIAL

(4) (5)

(6)

281

(Το [Υ] ) simplify the procedure, the following method has[X] been used. To [Y1] combat this problem, we have [X1] introduced [X2] control struc­ tures encourage the systematic use of heuristic knowledge. For Pakistan to[Y] finance so costly a program, public spending has[X] to be cut to the bone. X — quasi-V

(7)

We interpret these rules as rules of semantic interpretation or[G], to[Y] provide the basis for further discussion, as[X] rules mapping  onto Σ.

Adverbial modifier manifested by an absolute construction (9.8-9.9)

Prepositionless absolute-predicative construction

Examples X=(V) (1)

Many problems remain[X], the main one[Y] being[Z] that of determining the meaning of the request.

282

ADVERBIAL

(2)

Whatever[Z] the origin[Y] of the expression both terms must[X] be applica­ ble. X = quasi-V

(3)

Reflexivisation is applicable to the former tree and[G], everything[Y] else being[Z] equal, to[X] tree (81).

Absolute-conjunctional construction

Examples X=(V) Order (1) (1)

Level 4 will correct[X] the linkage when[Y] necessary[Z].

(2) (3)

He started[X] back surprise. It is a phenomenon well known[X], although[Y] not by[Z] that name, to professional linguists.

ADVERBIAL (4) (5) (6) (7) (8)

283

The iron filings appear[X 1 ] to be drawn[X2] to the magnet when[Y] placed[Z] in the vicinity. A significant feature to be borne[X] in mind when[Y] speculating[Z] on the nature of the earth's interior is the following one. Few local authorities have the funds to maintain[X] families deprived of benefits for many weeks while[Y] awaiting[Z] tribunal decisions. A reasonable range of phenomena will prove[X] tractable when[Y] dealt[Z] with under an extension of the parser. The importance and taxemic value of modulation are[X] also noted, though[Y] developed[Z] in less detail. Order (2)

(9) (10) (11) (12)

If[Y] necessary[Z], the wife can[X] start the divorce. If[Y] allowed[Z] to go unchecked, keen rivalry could[X] lead to bitterness. Once[Y] created[Z], they are[X], in principle, independent entitites. Keen rivalry, when[Y] reaching[Z] the point of open conflict, becomes[X] a serious drawback. (13) When[Y] informed[Z] that the tsar had made destructive and extensive deletions in "The Bronze Horseman", the poet abandoned thoughts of publication despite the fact that the unfavorable fiscal conditions in which he lived were precariously worsened by this decision. X = quasi-V (14) Rule (18) can apply to this kind of tree only, but [G], when[Y] properly extended[Z], to[X] the former PS-marker as well. ► The phrase Y of Syntagm 9.9 is sometimes called contingent adjective (or adverb) verbless clause (e.g., Quirk et al 1978: 256). Note, however, that this term is not synonymous with our absolute-conjunctional construction, since the former covers also disjunctive adjective and adverbial modifiers, sentential adjectivais and the like. <

284

ADVERBIAL Adverbial modifier manifested by a clause with an introductory element (9.10-9.11) A clause with a subordinate conjunction

Examples (1) (2)

Boasting has been tolerated[X] when[Y] it came[Z] from certain types: poets, entertainers, politicians. The true professional is said never to operate[X] until[Y] everything is[Z] in.

(3)

The entry is allowed[X] to stand in the verb phrase table, sequent revisions should[Z] be necessary.

(4)

Transactions tend to proceed[X] in chains, (postpos!)] each response is[Z] in turn a stimulus. The political metaphors which Plato used in this context, important though[Y] they are[Z], are[X] less striking. It should[X] be inappropriate, if[Y] indeed it were[Z] possible, to give a comprehensive survey of the work. If[Y] the graph had[Z] to serve to accept or reject the given sentence, a

(5) (6) (7)

ADVERBIAL

285

different version would[X] have to be constructed. (8)

Now that[Y] the coding has[Z] been completed, testing of the resolution logic has[X] been undertaken. (9) Whether[Y] a language is[Z] considered to be an object or[W] a process, it is[X] to be modeled in the way all physical entities are modeled. (10) Whether[Y] or[W] not this string is[Z] regarded as grammatical, it neverthe­ less is[X] a possible output of our grammar. (11) The program asks whether it is replaceable and[G], if[Y] it is[Z], actually replaces[X] it.

(1) (2) (3)

Examples Whatever[Z] orders he may[Y] give, they must[X] be obeyed. Whenever[Z] one patient visited[Y] his psychiatrist, he would[X] call himself all sorts of names. The attempt at blackmail failed[X], whoever[Z] had[Y] been behind it. Adverbial modifier manifested by a conditional clause without an introductory element (9.12)

9.12.

ADVERBIAL

286

α = HAVE irreal [Had I said ...], SHOULD 3 [Should he fail...], BE irreal [Were it done ...], BE sbjnct [... be it made or not] Examples (1) (2) (3) (4) (5)

My pitch mightp[X] have dropped a little, had[Y] I been playing since 2 p.m. this afternoon. Should[Y] all sets of proposed changes fail, the default CF heuristics is[X] invoked. The same process can[X] be applied, should[Y] the necessity of two levels above the morphemic level be examined. This shows that the indication of assertion, be[Y] it made by a sign or by accentuation, is[X] of a pragmatic nature. Anal intercourse, be[Y] it heterosexual or homosexual, was[X] quite common in ancient Israel.

Adverbial modifier manifested by a conjunction capable of introducing an independent sentence (9.13) 9.13.

Examples (1) (2) (3)

But[Y] we do[X] not in philosophy need to state what a language is. For[Y] we are[X] not particularly concerned with defining the term. Nor[Y] are[X] we concerned with a systematic classification of different

ADVERBIAL / MODIFICATIVE-ADVERBIAL

287

linguistic forms. Why? Because[Y] the trade union helps[X] in running the works.

(4)

Adverbial modifier manifested by the phraseologized construction "X or no X" (9.14)

Examples (1) (2)

10.

Liberation struggle is what it continues[X] to be in the eyes of all Iraqis, ceasefire[Y] or[Z] no[U] ceasefire[W]. However, negotiations[Y] or[Z] no[U] negotiations[W], the Air France plane was[X] refueled in Benghazi.

Modificative-adverbial syntagm

The modificative-adverbial SSRel is restrictedly repeatable: the main verb X can have at most two modificative adverbiais Y, one preposed and one postposed. For example:

MODIFICATIVE-ADVERBIAL

288 (1)

a. Self-righteous[Y1] and spiteful, McCormick left[X], full[Y2] of new suspicions. b. Melancholic[Yl] and depressed, he wandered[X] through the streets of the town, always silent[Y2] and alone.

α = (A, not pron), (V)ppres, (V) pass ► The label "not HAVE1" on node Y of this syntagm indicates that no perfect participle may assume the role of a modificative adverbial. Hence, a phrase of the type Having written his first book [,he left] can be interpreted only as a simple adverbial. Condition 1 requires that the grammatical subject separate the preposed [= left-hand] modificative adverbial and the main verb.  Examples Y=(A) (1) (2) (3) (4)

Suspicious[Y], self-righteous, spiteful, he[Z] was[X] like a wife that must be got rid of. He wandered[X] through the streets of the town, always silent[Y] and alone. Ashurst listened[X], silent[Y]. She appears[X] at the psychiatrist's office, ready[Y] to be hospitalized. Y = (V)ppres

(5)

Trembling[Y], cold, in ghastly fears, Ah! She[Z] did[X] depart [W. Blake].

MODIFICATIVE-ADVERBIAL

289

Y = (V)pass (6) (7) (8) (9)

Boxed[Y] in, Nixon[Z] has[X] been thrashing back and forth. Cheered[Y] by thousands of people, he[Z] leaped[X] from a steepler. He returned[X] after twelve, heavily loaded[Y] with newspapers and books. He got[X]up,scared[Y].

► On some rather rare occasions and under conditions not clear to us, a modifica­ tive adverbial is capable of being semantically related to a sentence element other than the grammatical subject: (10) Unable to shift his feet properly to defend himself, all ten inches of the blade pierced his side, and dying, he fell to the ground [I. Shulman, West Side Story]. The italicized phrase characterizes HE, realized on the surface as the determiner of an object (boldfaced his). (11) Actually named Metacom, the colonists gave him a royal nickname [= King Philip] for his proud manners. Here the modificative adverbial semantically characterizes the object. On the other hand, a modificative adverbial can semantically characterize a deepsyntactic subject which corresponds to no surface element of the sentence at all: (12) It is not normal to wander through streets on a Monday afternoon, obsessed. In (12), the modificative adverbial obsessed is semantically related to the deep-syntac­ tic item ONE [= It is not normal for ONE. to wander... ONE. being obsessed'], but to no actual sentence element. (By the way, this fact argues for the SSynt-subordination of the modificative adverbial to the main verb or to an infinitive, rather than to the grammatical subject of the sentence — which, on the surface, can be a dummy, as in (12).) (13) Two days later, a Soweto town councilor died[X] not far from his home, ambushed[Y] by a gang of township youths, who hacked him with knives and axes.

t This line is from the limerick which is quoted in full in the footnote onpage 399.

APPOSITIVE-ADVERBIAL

290 11.

Appositive-adverbial

syntagm

The appositive-adverbial SSRel is non-repeatable. 11.1.

Examples (1) (2) (3) (4)

(5) (6)

Editor[Y] of "Daedalus", Prof. Grobar[Z] devoted[X] long hours to the correction of manuscripts. A capable mechanic[Y], he[Z] earned[X] sixty dollars a week. A dictionary buff[Y], a devotee of obscure words, he[Z] had[X] been intent on improving his companion's grammar. An engineer[Y] who also has night classes and homework, he[Z] arranged[X] his schedule so that every night from 8 to 9 he and his girlfriend are free to be with each other. I went[X] to work everyday, an articulate business-woman[Y], liked by all my neighbors. The road came[X] on, an endless blistering ribbon[Y] between worn ruts [W.Faulkner].

► The postposition of an appositive adverbial to the main verb is restricted by a number of conditions which we are unable to formulate at the present time.

ATTRIBUTIVE-ADVERBIAL/PARENTHETICAL 12.

Attributive-adverbial

291

syntagm

The attributive-adverbial SSRel is non-repeatable. 12.1.

Examples (1) (2) (3) (4)

13.

With[Y] an open parasol, the girl strolled[X] beside him. Tautly[Y], but calmly, the Mayor explained[X] his position. As[Y] a kind of existential operator, the copula can[X] be easily accomo­ dated under this analysis. To[Y] be summarized later, this argument provides[X] a good starting point.

Parenthetical syntagms The parenthetical SSRel is non-repeatable. General condition on all parenthetical syntagms

The parenthetical construction must be set off by commas, parentheses or dashes except in the following case: Y is a single adverb without the syntactic feature (comma) or (comma!) [such as perhaps or probably]. The setting off of single parenthetical adverbs having the syntactic feature (comma) by commas is possible but not obligatory in all cases. (A more detailed study is needed for a more nuanced description.)

PARENTHETICAL

292

Parenthetical phrase manifested by a parenthetical adverb (13.1 -13.2) With verbs and within sentential adjectivais, as well as in absolute, coordinative and attributive constructions

► Y may be manifested by set (= phraseological) adverbiais, see examples (7), (12), (14), etc. 4 Examples Order (1) (1) (2) (3) (4)

We have[X], then[Y], the following results. One thing should be emphasized[X], however ( though ) [Y]. Those five men are[X] Polish, Irish, Armenian, Italian and Chinese, respectively[Y]. John and Harry love[X] Mary and Alice respectively[Y].

► The sentences of the type (4') *Respectively those five men are Polish, Irish, Armenian, Italian and Chinese. and (4") *John and Harry respectively love Mary and Alice. are ruled out since respectively = (postpos!).  (5)

Games are s a n d w i c h e d [ X ] , = (postpos!)], between pastimes and intimacy.

PARENTHETICAL

293

Order (2) (6)

Even under those circumstances, however[Y], it still keeps[X] its construc­ tive quality.

(7)

, sentence parsing programs produce[X] many different trees for a single sentence.

(8)

True[Y], there have[X] been investigations.

► Sentences of the type (8') *There have been, true, investigations..., etc. are ruled out since true = (9)

(prepos!)

Fortunately Sadly[Y], the train does[X] not stop in Wahoo.

(10) More surprisingly[Y] than wisely, E.T. A. chose[X] to snuff out the Prime Minister. (11) Honestly[Y], I'd[X] rather stay here. ► For a discussion of evaluative parenthetical adverbs, also called attitudinal disjuncts (clearly, obviously, amazingly, unquestionably, strangely, ...), see Corum 1974 and Greenbaum 1969. «1 X = quasi-V (12) He runs the risk of immediate reprisals or[G], , of[X] making an enemy. (13) He gives her a sandwich and[G] perhaps[Y] a cup[X] of coffee. (14) This specific intonation distinguishes a sentence we claim as true from[G], for instance[Y], a sentence[X] uttered as a question. (15) A particular paper in[G], say[Y], a Chinese journal[X] may bear on his investigations. (16) Such rules[G], for instance[Y], Equi-NP-Deletion[X], will be discussed further. (17) The explosions had all been prepared[G] to the same plan, perhaps[Y] even by[X] the same hand. (18) The modification structure will[G] not be established at this first level, indeed[Y] not until[X] level 4 of the analysis.

294

PARENTHETICAL With a noun, an adjective or an adverb

13.2.

Examples (1) (2)

consists of statements about static reality. it leans more heavily on the prosecutor.

Parenthetical phrases of various types: verbless comment clauses (13.3 -13.6) Of the type naturally enough 13.3.

Examples (1) (2)

Curiously[Y] enough[Z], the childhood prototype of Cops and Robbers is[X] not cops and robbers. They arrived[X], surprisingly[Y] enough[Z], before we did.

PARENTHETICAL

295

Of the type frankly speaking 13.4.

Examples (1) (2) (3)

Strictly[Z] speaking[Y], this is[X] a heuristic, rather than an algorithm. Her situation requires that she be proficient in ten different occupations, or[G], stated[Y] otherwise[Z], that[X] she fill ten different roles. Put[Y] frankly[Z] (Putting[Y] it frankly[Z]), he hasn't[X] a chance.

Of the type to give an example 13.5.

296

PARENTHETICAL Examples

(1) (2)

To[Y] return[Z] to[W] features, we can[X] take Halliday's view. The word 'open', to[Y] take[Z] a familiar example[W], can[X] occur in the

(3)

They were still the sole repositories of justice, which (to [Y] judge[Z] from[W] Hesiod's frequent complaints) they were[X] liable to administer. To[Y] quote[Z] from[W] a familiar source, 'Language is[X] a system where everything is interrelated' [F. de Saussure].

(4)

Of the type as shown above 13.6.

Examples (1) (2)

(3)

The word list was[X] then checked against the dictionary, as[Y] reported[Z] above. The little girl induces the boy to get dirty and then sneers[X 1 ] at him, as[Y1] classically described[Z 1 ] by Maugham, and[G], as[Y2] already noted[Z 2 ], by [X2] Dickens in "Great Expectations". As[Y] stated[Z] by de Saussure, the arbitrariness principle drew[X] fire from various quarters.

► There are a few verbs which are not (communic) and do not govern THAT 5 but which can nevertheless participate in this construction: (4)

a. As provided for in Section 8, the variables in this procedure will be automatically filled in. b. A merchant may add it to his retail price, as permitted by By-law 101.

PARENTHETICAL

297

To cover these verbs, one must list them in Syntagm 13.6 as lexical labels of Z. 

Parenthetical clauses: comment clauses with a tensed verb (13.7 - 13.16)t Parenthetical as-clause (13.7 - 13.10)

With an intransitive or passive verb 13.7.

Examples (1) (2)

This was[X] understandable, as[Y] was[Z] reported, because most of them had been welfare clients. But, as[Y] so often happens[Z] in the development of ideas, the fundamental error of the old position is[X] being carried over into the new one.

► 1. There exist several interesting constraints on the linear placement of parenthetical clauses of all types (not just of the type considered in this syntagm). Namely, they can separate certain sentence elements in certain specified contexts but not other elements or in other contexts: (3)

f

a. *A donation, it seems to me, to their campaign fund would be preferred. a' . They would prefer a donation, it seems to me, to their campaign fund. b. *These claims will make, I have no doubt, many people quite angry.

See Emonds 1976: 43-56.

PARENTHETICAL

298

b'. These claims, I have no doubt, will make many people quite angry [or: These claims will make many people, I have no doubt, quite angry]. (For more see Emonds 1976: 46ff.) These facts must be accounted for in the global word order rules mentioned on page 180. 2. The conditions of Syntagm 13.7 should be sharpened in many respects: - There are a few verbs which are not (communic) and do not govern THAT 5 but which can nevertheless participate in this syntagm, e.g. OCCUR: (4)

And, as often occurs,...

- There are a few verbs which do govern THAT 5 but cannot participate in this syntagm: (5) *And, as (it) appears (seems),... - IT 5 as dummy subject is necessary with some verbs and impossible with others: (6)

a.

And, as it turns out,...

a *And, as (often) turns out,... b. *And, as it occurs,... b And, as often occurs,... - Several verbs when used in this construction require the presence of additional elements while several others do not admit further modification: (7)

a. a'. b. b

And, *And, And, *And,

as occurs in daily life,... as (it) occurs,... as it happens,... as it happens in the development of ideas,...

There might be other important constraints. (The problems of the internal structure of English parenthetical clauses has been drawn to the authors' attention by I. Mackenzie.)

PARENTHETICAL

299

With a clause-governing adjective 13.8.

Examples (1) (2)

(3)

Hence such data, as[Y] is[Z] now[T] obvious[W], must[X] be stored in computer memory. a. *As[Y] is[Z] appropriate[W], we consider[X]... b. As[Y] is[Z] appropriate[W] for[T] this kind of problem, we consider[X]... And, as[Y] was[Z] clear[W] to[T] everyone, workers had[X] to begin restoring the land around Chernobyl.

300

PARENTHETICAL With a transitive verb

13.9.

Examples (1) (2) (3) (4) (5)

(6) (7)

As[Y] the title[W] suggests[Z], the emphasis here will[X] fall upon the formal aspects of language. The claim is[X], as[Y] he[W] asserts[Z], unsupported. As[Y] we[W] see[Z], nominais turn[X] up again. The claim is[X], as[Y] I[W] understand (imagine)[Z] (it (things, the facts,...)[U]), unsupported. a. The claim is[X], as[Y] we[W] view (perceive)[Z] the matter (the question)[U], unsupported. b. *The claim is, as we view (perceive), unsupported. This seems[X] to be true even for experienced translators, as[Y] I[W] shall[Z] later show. As[Y] we[W] have[Z] pointed out above, context-free languages are[X] produced by rewriting one symbol at each derivation step.

PARENTHETICAL

301

With an auxiliary or modal verb 13.10.

Examples (1)

(2)

To prevent the idea of games from becoming vulgarized[X], as[Y] so many terms are[Z], it should be emphasized once more that this is a very precise idea. These rules express empirical generalizations about the syntactic structure of English, rather than describe[X] the syntactic structure of specific English sentences, as[Y] do[Z] phrase markers.

Conjunctionless parenthetical clauses (13.11 - 13.14)÷ General condition on Syntagms 13.11 through 13.14

That is, if the parenthetical clause Y precedes the main verb X, then there should be a component Q of the main clause which precedes Y: no conjunctionless paren­ thetical clause can be sentence-initial.

t See Warne 1974.

302

PARENTHETICAL With an intransitive or passive verb

13.11.

Examples (1)

This technique[Q], it[Z] seems[Y], may[X] have an important bearing on our investigation. (2) But[Q1] even here[Q2], it[Z] should[Y] be noted, our choice removed[X] some of this difficulty. (3) Igor was[X] happy to live in Australia, it[Z] seemed[Y].

With a clause-governing adjective 13.12.

► The condition states that if the parenthetical clause Y follows the main verb X, then there should be a component Q of the main clause which follows Y: parenthetical

PARENTHETICAL

303

clauses of this type cannot be sentence-final: see example (4). (Parenthetical clauses of all different types can.) Examples (1)

The unsolved problems also require[X], it[Z] remains[Y] evident[W], considerable sophistication in the choice of strategies to be used. (2) The guard, it[Z] is[Y] true[W], told[X] me that smoking was not permitted. ► Not all adjectives with the feature (1 [THAT 5 ]) may appear as W in this syntagm:

(3) *The guard, it is right, told me that smoking was not permitted. The conditions for the proper use of adjectives in this parenthetical construction must be stated in the DSyntR = SSyntR component of English syntax, and not in the syntagms. (4)

a.

They are[X] a function, it[Z] is[Y] apparent[W], of[U] the state-control­ led system of education. a' *They are a function of the state-controlled system of education, it is apparent.

[Compare (4a') with (3) under Syntagm 13.11.]

With a transitive communication/thinking verb 13.13.

(1)

There are[X] only fourteen allowable actions, he[Z] said[Y].

304 (2) (3) (4)

PARENTHETICAL John should run[X], I[Z] think[Y], down the street. Mr. Drake, they[Z] claimed[Y], pushed[X] a child into the street. Linguists in France, you[Z] know[Y], take[X] Chomsky very seriously.

Of the type / am sure 13.14.

Examples (1)

(2)

a. b. c.

Your assertion is[X], I[Z] am[Y] afraid (sure)[W], unsupported. Your assertion, I[Z] am[Y] quite sure[W], is[X] unsupported. Your assertion is[X] unsupported, I[Z] am[Y] afraid[W]. His brother was[X], I[Z] have[Y] forgotten to mention[W] (it (this, thi fact)[U]), the same doctor who operated on Mr. Mansion.

PARENTHETICAL

305

Parenthetical what- clauses (13.15) 13.15.

Examples (1) (2)

It is true for all the PS-grammars and[X], what[Z] is[Y] more important, for the existing T-grammar. What[Z] is[Y] even more remarkable, corresponding particle constructions are[X] available as style disjuncts.

Parenthetical clauses set off by parentheses or dashes (13.16) 13.16.

► This syntagm describes the corresponding constructions in a fairly approximate manner: a parenthetical clause set off by parentheses or dashes is regarded as always subordinated to the immediately preceding word. In this way our surface-syntactic structure provides the information concerning the linear position of the parenthetical clause within the string of words of the actual sentence. This information is instrumen­ tal in automatic text analysis: it permits computing the semantic connection between the parenthetical clause and the corresponding component of the matrix sentence. 

PARENTHETICAL / ADJUNCTIVE

306

Examples (1)

(2)

(3) (4)

(5)

14.

Only the use of few categories of data[X] (each category may[Y] contain a large volume of data) enables the computer to operate as a high-speed processor. There are blind men who practice law and hold political office[X] (one such is[Y] currently mayor of the writer's home town) and deaf men who practice psychiatry. The highest percentage made was in sentence 2[X] (see[Y] English version, Appendix 2) where eighteen post-editors misinterpreted our output. If a biochemical or physiological abnormality is the prime mover in excessive drinking[X] — and that is[Y] still open to some question — then its study belongs in the field of internal medicine. If one[X] — let[Y] us say the husband — is playing this game, the other is taken into individual treatment.

Adjunctive syntagms The adjunctive SSRel is repeatable.

An interjection or a sentential adverb (14.1) 14.1.

Examples Order (1) (1)

Answer[X] my question, please[Y], and do not worry.

ADJUNCTIVE

307

Order (2) (2) Oh [Y1], no [Y2], I am[X] the housekeeper. (3) "Well[Y], what was[X] it?" demanded Bland. (4) Hm[Y], I'd [= I would[X]] like to talk to him, Ma'am. (5) Yes[Y], we all admire[X] your conscientiousness.

Address (14.2) 14.2.

► See Zwicky 1974 for the internal structure of English addresses and for restric­ tions on the nouns that can appear as Y in this syntagm. Examples (1) (2) (3)

And wipe[X] off the lipstick, you[Y] fool! But, Sir[Y] George, I've[= I have[X]] no idea where she is. Search[X] him, some[Y] of you shirking lubbers!

B. Syntagms covering phrasal SSyntRel's 1. General 15. Restrictive syntagms The restrictive SSRel is repeatable.

RESTRICTIVE

308

Negative adverb with a finite verb (15.1) ► 1. Negative adverbs with infinite verbal forms or with a word-class other than the verb are subsumed under Syntagm 15.2 (since all (Adv, neg) are considered to be (Adv, delim) as well; i.e. the syntactic feature "neg" implies "delim"). 2. In Syntagm 15.1, we have intentionally avoided a number of complicated problems related to the syntax of negation in English, namely, the use of pronouns of the type some vs. any within the scope of negation, double negation {Jones doesn't not pay taxes for nothing; That child hasn't exactly not been called upon; They didn't merely not do well, they failed miserably: see Sager 1968: 6), etc. We feel that all these phenomena belong rather to the domain of deep syntax which need not concern us now. For a comprehensive analysis of the related problems see Jackendoff 1972: 321-369 and Stockwell et al. 1973: 230-293.  15.1.

Examples

(2) (3)

Order (1) These results should[X] not[Y] obscure the existence of numerous unsolved problems. This is[X] not[Y] the case in AT. Do[X] you[Z] not[Y] think so?

(4)

These tools have

(5) (6)

He is[X] playing for fun. It is[X] also[Z1] probably[Z 2 ] not[Y] true that...

(1)

not[Y] yielded the desired solution.

Order (2) (7)

She requests that they not[Y] be[X] (* be not ) examined / . . . that they

RESTRICTIVE

309

not[Y] examine[X] her. (8) (9)

a. b. c.

He no longer[Y] is[X] playing for fun. He never[Y] can[X] make his decision alone. He can[X] never[Y] make his decision alone. NeverfY] can[X] he make his decision alone.

(10)

These tentative findings in no way[Y] confirm[X] or deny Schachter's hypothesis.

(11)

He by no means[Y] is[X] considered an authority on grammar.

Restrictive adverb (15.2) 15.2.

► 1. For more details concerning condition 4, see Emonds 1976: 174-175. 2. Condition 6 means that no dependent of the modified head X may stand to the left of the restricting adverb Y. 3. Condition 7 was inspired by Sager 1968: 26. 

RESTRICTIVE

310

Examples Order (1) (1)

It is just[Y] what[X] I wanted.

(2) (3)

Nearly[Y] all[X] of them explain how to play this role. This role is typically played by a member of the opposite sex, usually[Y] the spouse[X]. The usual order for end adverbiais is probably manner-place-time, not[Y] place[X]-manner-time. The town has been almost[y 1 ] totally[X1] destroyed, with hardly[Y 2 ] a building[X2] left unmarked by bombs or shells. They could not have afforded a cable address even[Y] had[X] there been anyone who might have wished to send them a cable. Our theory also[Y] supports[X] this claim. They know the proper greeting rituals and also[Y] how[X] to carry on topical conversation.

(4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)

As we have just[Y1] seenp[x1], school is ambiguous ways. They just[Y] can't[X] make it. ( *Just[Y] can't[X] they[W] make it?) He almost[Y] got[X] a job. <*Almost[Y] did[X] he[W] get a job?) I asked them not[Y] to[X] leave the room. All passengers are to raise their hands above their heads and not[Y] move[X]. You'd better, I think, not[Y] go[X] there. Mr. Dodgson neither claims nor acknowledges any connection with the books not[Y] published[X] under his name. The threshold test ban failed not[Y] by[X] attracting bitter animosity, but by indifference. Not[Y] many[X] people saw him. Not yet; not now; not me; not this year. Order (2)

(19) Britain[X] especially[Y] was getting into serious trouble. (20) The label is applied to a specific kind of unit[X] only[Y]. (21) The translator should not regard a word as a dictionary entry[X] only[Y], which can be subdivided into a number of definitions or senses.

RESTRICTIVE

311

Adadjectival adverb (15.3) 15.3.

Examples Order (1) X= (A, not pron, not,ord)jposit (1) (2) (3) (4) (5)

highly[Y] suspicious[X]; very[Y] interesting[X]; fairly[Y] clear-cut[X]; equally[Y]good[X]; ... devoting that[Y = THAT 6 ] much[X] time to her. definitions with about as[Y] many[X] attributes; an unfailingly[Y] and[Z] poignantly interesting[X] work; Oh, mine is not that[Y = THAT 6 ] hot[X]! [Cf. example (5) on p. 217]. X = (Adv, not pron)posit

sufficiently[Y] well[X]; a rather[Y] recently[X] published paper;

X=(N) (7)

the exclusively[Y] California[X] fruit[W]; a purely[Y] and[Z] strictly ballad[X]form[W]; X = (Prep)

(8) (9)

The action is really[X] in[Y] progress.

Landmark may have perceived it more[Y] out of[X] a desire to produce some neat statement than as a result of her own conclusions. (10) The reassuring function is more[Y] or[Z] less in[X] accord with the security

312

RESTRICTIVE needs of his wife. Order (2)

(11) a poor[X] enough[Y] garment; good[X] enough[Y] arguments; (12) Independent evidence is strong[X] enough[Y] to prevent the abandonment of the assumption.

Restrictive phrase modifying a comparative degree form or a "confr"-word (15.4) 15.4.

α = (Adv, adcompar), ANUMP, (N, temp)pl, THE 2 Examples Y = (Adv, adcompar) (1)

a still[Y] sharper[X] division; a much[Y] bigger[X] navy; twice[Y] as[X] much paper; Y = ANUMP

(2) (3) (4) (5) (6)

There are many times[Y] more[X] nouns that can be qualified by red than by comfortable. This is £2,000 million[Y] more[X] than in 1972. more[X] money; stronger[X]; four times[Y] as[Y] big as the first file; A minute[Y] later[X] her husband muttered an excuse and followed. Seven months[Y] after[X = (Prep, confr)] that the industry took the opportunity of slapping on another 60 tons.

313

RESTRICTIVE (7)

... one year[Y] too[X] soon.

► Cf. the representation of the following two synonymous constructions which differ syntacticaly: (8)

[I have bought]

(9)

[I have bought]

five[Y]

vs. [X] than you. 2

We distinguish between more = MORE , the comparative degree of much/many, and more = MORE3, the adverb with the syntactic feature (delim), cf. two examples more examples, both featuring MORE3, where the SSRel's shown are established by Syntagm 15.2(2). MORE2 can be replaced in all contexts by LESS2, the comparative degree of little/'few, and a parallel /less-construction (with LESS3) exists for MORE3. Cf. pp. 383-384. (In fact, MORE2 and LESS2 are handy 2 abbreviations for MUCH c o m p MANY c o m p and LITTLE c o m p / FEW c o m p, respectively, rather than separate lexemes.)  Y = (Ν, temp)pl (10) My memory comes to the surface again years[Y] later[X]. (11) The moves came hours[Y] after[X = (Conj,subord,confr)] troops occupied the university. Y=THE2 (12) The more ambitious is the undertaking, the[Y] more[X] the need for syntactic analysis will be felt.

Adverb modifying a superlative degree form (15.5) 15.5.

RESTRICTIVE

314

Example He was

strongest[X] boy[W].

Adverb modifying a numeral (15.6) 15.6.

Examples (1) (2)

approximately[Y] 300[X] people; precisely (exactly, just, almost)[Y] five[X]; some[Y] twenty[X] pounds; It costs anywhere[Y] from[X] five to seven dollars (... anywhere[Y] over[X] ten dollars).

Adjective modifying an adjective (15.7) 15.7.

RESTRICTIVE

315

Examples (1) (2)

wide[Y] open[X] window; perfect (extreme) old age; soft[Y] red[X] color; the second[Y] best[X] student; the third[Y] largest[X] power in the world.

► Note that constructions of the type the simplest possible solution have either of the following two readings: 'the simplest of all possible solutions' or 'the maximally simple solution'. Each reading apparently presupposes a different syntactic structure: (3) VS.

(4) or (4')

The second structure is described by Subrule 15.7(2). (L.L. Iomdin should have the credit for bringing the above construction to the attention of the authors.)

Constructions of the type quite a large house (15.8) 15.8.

α = (A, not pron, not ord), (N), (Adv) Examples (1) (2) (3) (4)

quite[Y] a[Z] large[X] house[W] ; quite[Y] the[Z] same[X] kind[W] ; The meeting was quite[Y] a[Z] heated[X] one[W]. Twenty men rescued in one day is quite[Y] a[Z] record[X] for the Coast Patrol. You are quite[Y] a[Z] humorist[X] !

316

RESTRICTIVE Constructions with conjunctions which introduce disjunctive sentence elements (15.9)

► The term disjunctive is used here to characterize any sentence element that is somehow set off, within the sentence, by considerable pauses and specific intonation (marked in writing by commas), sometimes by a particular linear position or even — such is the case in this syntagm — by certain conjunctions. Disjunctive elements carry a kind of semantic emphasis, which is clearly seen from the following pair of sen­ tences: (1)

She read detective stories only.

(2)

She read, but detective stories only.

vs.

In (2), the direct object detective stories is made disjunctive by the conjunction but. Disjunctive adjective and adverb modifiers are sometimes called adjective (adverbial) verbless clauses, as in the examples that follow: (3) (4) (5)

Cleared, this site will be very valuable. Single-handed, he ran the shop. The manager approached us, smiling.

For more details see Quirk et al. 1978:725-727 and passim. The other term used in the literature for the same phenomenon is detached. 

15.9.

r = ith compi, agent, sub-copred, obj-copred, adverb, appos-adverb, descrappos, attrib, descr-attrib. ► We have listed above all the SSRel's which allow their dependent component to be disjunctive. Cf. the list of SSRel's in connection with the feature quasi-V, pp. 275-276.

RESTRICTIVE / AUXILIARY

317

Examples (1) (2) (3) (4) (5)

She smoked[Z], but[Y] the purest cigarettes[X] only. The oscillator is working[Z], although[Y] not necessarily at[X] the right frequency. We can find[Z] it, but[Y] not here[X]. Come[Z] on, and[Y] quickly[X]. This model also fails[Z] to account for the differential behavior of above and below, and[Y] for[X] similar reasons.

► An alternative description of the construction in question could be as follows: The conjunction (and, but or although) in the examples (1) through (5) might be regarded as the head of the disjunctive phrase, which, in its turn, would be taken to be a sentence adverbial, and then this entire "adverbial", conjunction included, would be subordinated to its governor via the adverbial SSRel, irrespective of the initial (= inherent) syntactic role of the basic phrase. However, such a description appears somewhat awkward to us, since, in particular, it could eventually result in considerable duplication of syntagms. Hence, following the traditional approach, we find it more appropriate to consider any disjunctive item to be connected with its governor via the same SSRel that must be used when this item is not disjunctive (and therefore not preceded by a conjunction). 

2. Verb phrase 16.

Auxiliary syntagms The auxiliary SSRel is non-repeatable. General condition on all auxiliary syntagms with the exception of Syntagm 16.2(2)

if... ⊃ Q,then Q ⊂ R, and... ⊃ R, and

► The above condition allows us to account for the fact that the auxiliary and the lexical verb in an analytical verb form (= complex verb phrase) can be separated only by one of the following four types of syntactic items: - the subject phrase (in the case of inversion, such as occurs in a question); - an adverbial modifier;

318

AUXILIARY

- a parenthetical expression; - a restrictive phrase. The form of the progressive with the inversion of its components (Syntagm 16.2(2)) constitutes an exception to this condition since it allows more types of syntactic items between its components. 

Analytical forms of the passive voice (16.1) 16.1.

► This condition covers cases like the following: (1) His bed hadn't been slept in that night. (2) This land was smiled upon by the President. If the verb under analysis has the feature (pass), which means that it is not normally used in the passive voice (like sleep or smile), then for there to be a passive form it is necessary for this verb to govern a stranded prepostion. See below, Syntagm 20.1, which provides for such constructions.  Examples (3) (4) (5)

is[X] broken[Y]; will be broken; to be broken; is being broken; Some obscurities are[X] indeed[Q = R] cleared[Y] away. Since these solutions are[X], with[Q1] = R] rare[Q 2 ] exceptions[Q 3 ], rejected[Y], it is apparent that we need to resume our research.

► 1. The passive past participle which enters as a constituent into the analytical form of the passive voice must change its dictionary government pattern in a systema­ tic way. This is so because the government of the passive form as a whole is converse

÷ The variables Q and R in the examples below refer to the general condition on all auxiliary syntagms, see page 317. The same holds for all the other examples in this section.

319

AUXILIARY with respect to that of the basic (i.e., active) verb. Cf. : (6)

Mr. Schank has created a more accurate picture of it = A more accurate picture of it has been created by Mr. Schank.

To make provision for this it is necessary to specify, in an explicit manner, the regular correspondence between the basic (= lexicographic) government pattern of the verb in question and its actual government pattern in the passive form. For instance, a transtive verb (not of the give type) in the passive form cannot govern a prepositionless noun via the 1st completive SSRel, etc. The correspondence is given below: Basic, or government pattern

dictionary.

Actual government pattern of the passive form

The transformation into passive form does not entail alterations in the other government pattern columns, i.e. those with numbers >2. Ditransitive verbs (= (adres)-verbs), of the type give, show, assign, etc., apart from the above-mentioned derived government pattern in the passive (i.e., The book

320

AUXILIARY

was given by John to Mary), have one more pattern: Basic, or dictionary, government pattern

Actual government pattern of the passive form

Cf.: 1

(7)

3

2

1

2

3

John gave Mary the book = Mary was given the book by John.

The above correspondence is valid whenever past participles are used outside the perfect forms — i.e., when they are passive: 1) in the passive voice, 2) as modifiers of a noun, 3) in absolute-predicative construction, etc. 2. We have chosen not to consider the so-called gei-passive as a member of the grammatical category of voice in English. In other words, we propose to describe the phrase get opened in (8) : (8)

How did this window get opened?

by the 1st completive SSRel, treating GET in such constructions as any verb govern­ ing a past participle and the past participle as any kind of complement. Therefore, phrases like get opened, get wounded, got beaten, etc. are covered by Syntagm 2.20, pp. 249-250. Our reason is that the gei-passive adds some extra meaning to the basic meaning of the verb, which is not normal for the grammatical category of voice (the latter should be limited to the purely syntactic demotion/promotion of different actants of the verb). Cf. the remark on page 250. 3. An interesting problem is created by examples such as the following three sen­ tences: (9)

a. Sandwiched between these studies are seven studies of morphology. b. The task is to discover how deeply organized such a phenomenon might

AUXILIARY

321

be. . Associated with each iteration are (i) a function from V to the set of the nonnegative real numbers and (ii) a special subset of V. Do the italicized phrases represent (analytical) forms of the passive? If so, then the inverted o r d e r p a s t participle + BE 1 ", is possible, and the corresponding subrule (= Order (2)) should be added to Syntagm 16.1. However, the phrases in (9) are hardly passives; rather, they seem to be free combinations of a copula verb with a past participle in the role of an adjective: First, native speakers do not perceive these expressions as passives. Second, no agent by-phrase is possible in these three sentences: (9')

a. *Sandwiched by Mr. Johnson between these studies are seven studies of morphology. b. *The task is to discover how deeply organized by the researcher such a phenomenon might be.  * Associated by this procedure with each iteration are (i) a function ... and (ii) a special subset of V.

Third, the verb BE can be replaced by a different copula (or copula-like) verb, which is not characteristic of the passive: (9") a. b.

Sandwiched between these studies appear seven studies of morphology. The task is to discover how deeply organized such a phenomenon might look.

Fourth, no other analytical verb form — except the progressive — admits the inversion of the lexical verb component with the auxiliary. Thus the non-passive analysis of (9) is in line with a useful and very natural generalization. To be sure, our hasty remarks are far from conclusive. The problem of distinguish­ ing between passive analytical forms and copula + adjective (= past participle) phrases is much more complicated than indicated here. An exhaustive treatment is beyond the scope of this book; we will refer the reader to a paper by T. Wasow (1977) where some relevant facts are analysed and further references are given (in particular, pp. 338-352).

322

AUXILIARY Analytical forms of the progressive aspect (16.2)

16.2.

Examples Order (1) (1)

— What arep[x1] you[Q = R] d o i n g 4 ] in my bed? — I am[X 2 ] just[Q = R] trying[Y2] to find out what kind of person you are. Order (2)

(2) (3)

Throwing[Y] the hammer is[X] at present Ray McTavish[Z], our local champion. Walking[Y] over there on the tarmac was[X] nobody[Z] other than Mrs. Susan Thomas, former president of the Ontario Women's League.

► The subrule 16.2(2) calls for the following four remarks. 1. Curiously enough, among all the analytical verb forms in English it is only the progressive that admits the inversion of its components, with the lexical verb (in the form of present participle) preceding the auxiliary BE 1 . True, this inversion is possible under specific conditions only, namely, if Y, i.e. the lexical verb, is the theme, or topic, of the sentence and if, at the same time, it is emphasized. Moreover, the inversion of the main verb and the grammatical subject becomes then compulsory. Conditions 1 and 2 in the subrule (2) take care of these facts. (A series of examples with the inverted progressive form is found in Green 1980:584.)

t See the footnote on page 318 for the variables Q and R in the examples below.

AUXILIARY

323

2. The general restriction on the linear gap between both components of an analytical form (page 317) does not apply in case of the inversion of present participle and BE1 in the progressive: they may be separated by the object of the participle (see example (2)) or by various adverbiais (example (3)). 3. With some verbs (stative verbs, for instance) only the inverted form of progressive is possible: sentence (4) (4)

Underlying[Y] this study is[X] an attempt[Z] to clarify what language is and how it works in human communication.

is fine while (4') is excluded: (4') * An attempt to clarify what language is and how it works is underlying this study, completed some years ago. The fact is that the verb UNDERLY does not generally admit the progressive form, although this form seems to become admissible with the inversion described in Syntagm 16.2(2). 4. The syntactic properties of the progressive aspect agree rather well with its morphological feature that sets it apart from all other analytical verb forms: the perfect, the future, the negative, interrogative and emphatic forms (with DO), and even the passive are possible, roughly speaking, for ALL English verbst while the progressive is limited to DYNAMIC verbs. One can say that the progressive is a grammatical form of the verb to a lesser extent than the other phrases with auxiliaries; it seems to be nearer to the so-called periphrastic verbal constructions, like be about to, be going to, etc. 

Analytical forms of the perfect (16.3) 16.3.

t With the obvious reservation concerning the passive; it is formed normally from transitive verbs alone.

324

AUXILIARY Examples

(1) (2) (3)

We have [X1] treated[Y 1 ] that as an example of what we had[X 2 ] called[Y2] "reason causation". The statement must have[X] caused[Y] him to think about the fact. We have[X] always[Q =R] avoided[Y] ambiguous conceptual diagrams.

Analytical f orms of the future tense, conditional mood and subjunctive mood (16.4) 16.4.

► The verbs shall1 and will and their past forms should1 and would1 serve to build the analytical forms of the future tense: simple future and future-in-the-past. The verbs should2 and would2, regarded in the present study as separate lexemes (different from should1 and would1), serve to build conditional forms: (1)

He would do it if you came.

The verb should? (still a different lexeme) is used to build subjunctive forms: (2)

Should he come in time she would be [= conditional] very pleased.  Examples

(3) (4) (5)

The representation of these sentences will[X] now[Q = R] be[Y] more compact. It should[X], as[Q1] = R] is [Q 2 ] clear[Q 3 ], be[Y] able to direct the search for a solution in the best way possible. The second meaning of bedroom would[X] in different contexts be[Y] translated into German as Kinderzimmer, Jugendzimmer, etc.

AUXILIARY / PHRASAL-JUNCTIVE

325

Negative, interrogative and emphatic forms with D O (16.5) 16.5.

Examples (1) (2) (3) (4)

The hydraulic method does[X] not[Q = R] contribute[Y] to the gain in speed. Intonation patterns do[X] have[Y] meanings. What dop[X] the[Q1] words[Q2 =R] should[Q3] and[Q4] must[Q5] indi­ cate[Y]? Do[X] we[Q1 = R1] really[Q2 = R2] need[Y] to perform such an extensive analysis?

3. Word-like phrases 17. Phrasal-junctive syntagm The phrasal-junctive SSRel is non-repeatable. ► All junctive syntagms (17 through 19) describe surface-syntactic constructions that are, in a certain sense, equivalent to lexical items, i.e., "split" words, so to speak: phrasal verbs, compound numerals, and paired, or correlative, conjunctions and prepositions.  17.1.

326

PHRASAL-JUNCTIVE

► The notation /(Z) stands for the length of the phrase Ζ measured in the number of word forms. Condition 1 sets limits on the length of the direct obj ect admissible between a verb and its adjunct, while Condition 2 rules out all other positions for the direct object manifested by a pronoun. Condition 3 forbids a gap between the head of the phrase and the adjunct if the head is not a verb: cf. examples (12) - (15). Conditions 1 through 3 specify the relative order of the verbal adjunct and the direct object in a rather approximate manner. First, this order depends not only on the obj ect but on the nature of the adjunct as well; cf.: (1)

a. He took Mary prisoner. a'. *He took prisoner Mary. b. He knocked Mary up. b He knocked up Mary.

Second, the order depends on the information value of the object (Bolinger 1971 : 51ff.): (2)

a. ?He brought up the subject. a'. He brought the subject up. b. He brought up the funeral. b He brought the funeral up.

Third, it depends on the degree of idiomaticity of a particular phrasal verb, which must be marked in the dictionary: (3)

a. a b. b'.

He put up a fight. *He put a fight up. Dick found out the truth. *Dick found the truth out.

There may be other factors as well.  Examples (4) (5) (6) (7)

Little Red Riding Hood took[X] out[Y] her automatic and shot the Wolf dead. Israeli forces shot[X] down[Y] three Iraqi helicopters. His influence has to be thrown[X] off[Y]. Our problems of finding storage and logging[X] on[Y]...

PHRASAL-JUNCTIVE / NUMERAL-JUNCTIVE (8) (9) (10) (11)

a. b.

327

The cosmetics industry wants to cash[X] in[Y] on ties between smell and sex. Write[X] off[Y] the losses! Write[X] the losses[Z] off[Y]! They took[X] him[Z] prisoner[Y] /... took[X] prisoner[Y] the captain of the enemy squadron. He makes[X] good[Y] on all his promises.

► 1. This syntagm also describes "noun plus adjunct" and "adjective plus adjunct" constructions: (12) (13) (14) (15)

the shift[X] away[Y] from the machine-like language; the weighing[X] out[Y] of the different oxides; the setting[X] up[Y] of a National Corporation; to be well[X]off[Y].

2. We regard adjuncts in constructions of the above type not as separate manner adverbiais characterizing the action described by the verb but rather as parts of "compound" words, i.e., we suggest that phrases of the type get up, take off, write down, come true, take prisoner, etc., be regarded as single lexical units, idiomatic to a greater or lesser degree, ranging from relatively free ones, on the type clear up or find out, to fully idiomatic ones, of the type give up. Cf. pp. 121-122.  18.

Numeral-junctive syntagms The numeral-junctive SSRel is non-repeatable. Compound numerals (18.1 -18.4) Compound numerals including only numerals

18.1.

328

NUMERAL- JUNCTIVE Examples

(1)

(2) (3)

twenty[Y] five[X]; three[Y] million[X] workers; the twenty[Y] fifth[X] line; five[Y1] hundred[X1] twenty[Y2] seven[X2] [see the next syntagm for the treatment of AND]; sixty[Y]-seven[X]; a four[Y]-hundred[X]-dollar job

[the hyphens*in (3) are prescribed by Syntagm 33.1].

Compound numerals including AND 18.2.

L = HUNDRED, THOUSAND, MILLION, BILLION, TRILLION,... Examples (1) (2)

five hundred[Z] and[Y] twenty[X] seven; five hundredfZ] and[Y]first[X]; five hundred[Z] and[Y] twenty[X] first.

► 1. Since an expression like twenty-third is an ordinal numeral, i.e., the whole expression belongs to the word-class of its right-most component, we regard the latter as the surface-syntactic head and represent compound ordinal numerals as follows: (3)

twenty

third.

Such compound ordinals are covered by Syntagm 18.1. By analogy, we adopt the same direction of surface syntactic links for compound numerals of other types as well, including those with AND: (4)

a. twenty b. hundred

three; three;

NUMERAL- JUNCTIVE c. hundred d.fivehundred e.

329

third; twenty

three;

Thus, AND is treated here not as a coordinate conjunction (in its typical configura­ tion cf. Syntagms 40) but rather as a special particle used to build a subclass of compound numerals. 2. Strictly speaking, Syntagm 18.2 should be presented as two different syntagms: oneforAND + Num (in the context of Num ), and the other for Num + AND (in the context of Num ). But we chose to override logical rigor for the sake of better perspicuity. The same is true for Syntagm 18.4. 

Compound numerals including the indefinite article or a pronominal quantitative word 18.3.

L1 = HUNDRED, THOUSAND, MILLION, BILLION, TRILLION,..., DOZEN, GROSS; L2 = A/AN, SEVERAL, HOW MANY Examples (1) (2) (3) (4)

a[Y] million[X] items; a[Y] dozen[X] soldiers; several[Y] thousand[X] armed soldiers; dozen[X] eggs?

► Notice that for each lexeme listed inL1and considered a numeral in this book there are several nouns that are homophonous to it. These nouns can be pluralized, take OF with the name of the objects counted, etc.:

330

NUMERAL- JUNCTIVE

(5)

a. b.  d.

The hundred [= the name of the figure] you have drawn is crooked. I've told you dozens [= lots] of times. Millions [= an indefinitely large number] of people all over the world yearn for peace. The last thousand [= the name of the set] was okay. 

Compound numerals including fractions 18.4.

Examples (1) (2) (3) (4) (5) 19.

three[X] and[Y] one[W]-sixth[Z] tons; three[X] and[Y] five[W]-sixths[Z] tons; a. three[X] tons[U] and[Y] five[W]-sixths[Z]; b. *three[X] battalions[U] and[Y] three[W]-fifths[Z] [Condition 4]; one[X] and[Y] a[W] half[Z] hours; one[X] full hour[U] and[Y] a[W] half[Z].

Binary-junctive syntagms

The binary-junctive SSRel is non-repeatable. It links all kinds of paired words that form compound lexical items.

BINARY-JUNCTIVE

331

Paired, or correlative, conjunction (19.1) 19.1.

Examples Order (1) (1)

If[X] the first conceptualization occurs, then[Y] the result is highly likely. Order (2)

(2) (3)

The essential feature of both[Y] procedure[Z] and[X] rituals is that they are stereotyped. Underneath she is either[Y] bewildered[Z] by him or[X] laughing at him.

► The direction of dependence between the components of a paired conjunction is so selected as to make the independent, or autonomous, part of the conjunction (i.e., the part that can appear alone) the main member of the binary-junctive SSRel

THE... THE... construction (19.2) 19.2.

BINARY-JUNCTIVE

332

Examples Order (1) (1) (2) (3) (4) (5)

The[Y] more accessible Zelda is[Z], the[X] more satisfaction John can[W] obtain. The[Y] less probable the occurrence of an element is[Z], the[X] richer is[W] its information content. Anyone is eligible to present a case, the[Y] more horrifying the[X] better. Do it, and the[Y] sooner, the[X] better. The[Y] less he learns[Z], the[X] more effectively he can[W] play. Order (2)

(6)

W(h) has a value that is[W] the[X] greater[U] the[Y] more alternative arrangements it permits[Z].

Paired preposition FROM-TO (19.3) 19.3.

= (Num),(A),(N) Examples W = (Num) (1)

It costs anywhere from[X] five[Z] to[Y] seven[W] dollars.

BINARY-JUNCTIVE / COLLIGATIVE

333

W= (A) (2) (3) (4)

The mathematical models range from[X] very elementary[Z] to[Y] highly formal[W] ones. ... the change from[X] a lower[Z] to[Y] a higher[W] pitch. Noun phrases are examined from[X] the last[Z] to[Y] the first[W] of the sentence. W= (N)

(5)

20.

From[X] cell[Z] to[Y] cell[W] of his brain crept the one thought ...[O.Wilde].

Colligative syntagms The colligative SSRel is non-repeatable. Pseudo-passive construction with stranded prepositions — of the type (be) operated on (20.1 - 20.2)

With simple verbs 20.1.

Examples (1) (2) (3) (4) (5)

He will be listened[X] to[Y] by the authorities. ... Or being lied[X] about[Y], don't deal in lies [R. Kipling]. The games against the tourists were looked[X] upon[Y] as a separate category. He was done[X] away[U] with[Y]. Her bed this morning had not been slept[X] in[Y] [A. Conan Doyle].

334 (6)

COLLIGATIVE The room seemed as though it were being just moved[X] into[Y] [T. Capote].

► The problem of whether only governed prepositions can appear in this pseudopassive construction, as stipulated by Syntagm 20.1, remains, in our view, unsolved. Examples (5) and (6) make us favor the idea that this is not necessarily the case, despite the fact that sleep = Real, (bed) and, hence, sleep may be deemed to strongly govern in while move, meaning 'relocate', may also be considered a strong governor of into. T. Langendoen (1970: 38-39) points out that the sentence (7)

This house was lived in by George Washington,

is grammatically correct, while (7') * Virginia was lived in by George Washington. and, likewise, (7'')This house was remained in by George Washington. are not. One possible explanation of the above discrepancy is that in live in a house the verb LIVE is the value of the lexical function Real of house, and the preposition in in that phrase is not only governed but also semantically empty. As for the phrases live in Virginia and remain in a house, in which the verbs are not values of any lexical function, the preposition in that occurs in them, while also being governed by the verb, is not semantically empty, cf. live near (far away from ) Virginia, remain behind a house, etc. Thus one might easily conclude that for a preposition to be admissible in the construction described in Syntagms 20.1-2, it is not enough that it be governed; it is also necessary that it be semantically empty. However, it turns out, first, that this condition is not always necessary, and second, that it is far from sufficient. The usage of a stranded preposition seems to be regulated by a number of subtle semantic properties, such as the following three: - It has been emphasized many times that in a pseudo-passive construction the stranded preposition not governed by the verb must indicate a spatial relation­ ship: a localization or a direction. - It has been stated repeatedly that pseudo-passives with stranded prepositions are more felicitous as general statements than as descriptions of specific facts. Cf.: (8) (9) -

This table has been eaten at by many famous people. That lake has been gone across so many times that no guide is necessary.

Pseudo-passives with stranded prepositions are possible only with the verbs denoting conscious (roughly, voluntary) actions (Emonds 1976: 103):

COLLIGATIVE (10) a. a. b. b

335

The closet was slipped into by the thief. *The closet was slipped into by the piece of soap. The room was flown across by the bird. *The room was flown across by the dictionary.

While sentences (10a-b) are not very good, sentences (10a'-b") are obviously worse. There are possible other factors, too, that determine the acceptability of a specific instance of the construction in question. A lack of relevant data forces us to content ourselves here with a rather rough formulation of the pseudo-passive construction in English. 

With semi-idiomatic verbal phrases of the type "verb + noun" 20.2.

Examples (1) (2) (3) (4)

At last the ship was lost[Z] sight[X] of[Y] on the horizon. Her child was taken[Z] good care[X] of[Y]. Bill was taken[Z] advantage[X] of [Y]. The spy was constantly kept[Z] track[X] of[Y].

► It seems that the pseudo-passive with verb + noun phrases is in fact much more restricted that Syntagm 20.2 implies. First, all the examples that we have seen have only OF in the role of Y. Thus, e.g., (5) seems to be impossible to many speakers: (5) *This state of affairs should be put[Z] an end[X] to[Y]. Second, even with OF some pseudo-passives of the type discussed are question­ able:

336

COLLIGATIVE / EXPLETIVE

(6)

· His words were not taken[Z] notice[X] of[Y].

By the way, (6) becomes slighty better and more acceptable with a long enough (= sufficiently heavy) adverbial following OF: (6)

His words were not taken notice of before the accident.

Since we do not possess sufficient data we have limited ourselves, as in the case of the preceding syntagm, to an approximate formulation. 

4. 21.

Expletive phrases

Expletive syntagms The expletive SSRel is non-repeatable. Dummy grammatical subject manifested by anticipatory IT (21.1 - 21.5) General condition on all the Syntagms 21.1-21.5

(an instance of obligatory non-projectivity)

The predicate is a verb which can have as its grammatical subject a gerund, an infinitive or a clause introduced by a subordinate conjunction 21.1.

EXPLETIVE

337

► Condition 1 rules out incorrect sentences of type (1): (1)

a. *It[X] means[Z] that[U] it is not likely to rain that[Y] the sun is shining [Langendoen 1970:172]. b. *It[X] will show[Z] whether[U] or not he's been nipping at the silver nitrate whether[Y] or not his mouth turns black [adapted from Bresnan 1970:311]. c. *It[X] implies[Z] that[U] he does not want the job talking[Y] so much about terms and conditions. d. *It[X] requires[Z] that[U] the expenses be kept to a minimum to[Y] be ready for all the opportunities.

Compare the correct variants in (2): (2)

a. b. c. d.

That the sun is shining means that it is not likely to rain. Whether or not his mouth turns black will show whether or not he's been nipping at the silver nitrate. Talking so much about terms and conditions implies that he does not want the job. To be ready for all the opportunities requires that the expenses be kept to a minimum.  Examples

Y = (V)ger (3)

It[X] is[Z] of no interest talking[Y] about it.

► Note that the verb BE in this sentence is Oper2 (interest) and is therefore marked (l[ger]) in its government pattern. (4)

It[X] was[Z] a mistake waiting[Y] out in the cold.

► Note that was = Oper2 (mistake) in this sentence, too  (5) (6)

It[X] is[Z] no use telling[Y] White that he is not really stupid. [The expres­ sion be no use is a phraseme.] It[X] was[Z] a shame our waiting[Y] out in the cold. Y =TO2

(7)

It[X] remains[Z] to[Y] distinguish games from other types of social actions.

► The verbs remain, seem, appear, etc. do not allow a grammatical subject other than anticipatory IT 5 to precede them if they do not have a complement. Cf. :

338 (8)

EXPLETIVE a.

It seems that the wind has shifted direction

b.

That the wind has shifted direction seems highly probable,

or

but not (8') *That the wind has shifted direction seems.  (9) ... the length of time it[X] takes[Z] to[Y] learn to read a new language. (10) It[X] should[Z] be clearly an easy matter to[Y] convert from one collection strategy to the other. ► In (10), be = Oper1 (matter) and therefore it is labeled( 1 ( T  1 ] )in its govern­ ment pattern.  (11) It[X] will[Z] be up to the semantic calculator (It[X] entails[Z] a loss of time) to[Y] construe an interpretation out of this contradictory string. Y = CONJ (12) It[X] turned[Z] out that[Y] he was a good engineer. (13) Then it[X] may[Z] be that[Y] in one area /hænd/ is found, while... (14) It[X]is[Z] to be emphasized (It[X]satisfies[Z] us completely) that[Y] these correlations are purely empirical in nature. (15) Clearly it[X] is[Z] not the case that[Y] there have to be two clauses ... [be the case is a phraseme]. (16) It[X] was[Z] ordered that[Y] he be hanged. (17) It[X] disturbs[Z] the whole picture that[Y] multiple particles feature much more gradience than single particles. (18) It[X] had[Z] been observed that[Y] a diet of porridge helped to regulate blood sugar and fats. (19) It[X] is[Z] very likely that[Y] traditionally-minded teachers of translation courses when faced with translations such as (5b) will object: There is a word missing here!

EXPLETIVE

339

The predicate is a copula with a special complement which can have the above items as its grammatical subject 21.2.

r = 1st completive, 2nd completive; α = ger, TO 2 , THAT 5 , WHETHER 1 , IF 2 , (V)fin; β = (i[α]), (Ν, pred-inf), TO 2 , (Prep) ► Conditions 5 and 6 ensure the correct synthesis and analysis of phrases of the type It

that... lit was

that...◀ Examples

2

Y = TO (1) (2) (3)

It[X] is[Z] more useful[W] and enlightening to[Y] investigate social transac­ tions. It[X] seems[Z] more instructive[W] to[Y] attempt an adequate theory that fails than to let an inadequate theory get away with false claims of success. It[X] will[Z] perhaps not be thought presumptuous[W] of him to[Y] suggest that students new to linguistics may have similar difficulties.

► In sentence (4): (4)

It [= this formalism] would be utterly unable to accomodate this grammar.

340

EXPLETIVE

the adjective unable has no (1[T0 2 ]) feature in its government pattern, and this makes possible the correct automatic analysis: to accomodate is related to unable and not to it, the intended reading [= IT1] is selected for it, etc. ◀ (5)

It[X] was[Z] easy[W] to[Y] growl that the food was late.  = CONJ

(6) (7)

It[X] is[Z] clear[W] that[Y] neither of the forms is derived from the other. It[X] must[Z] be left [leave = (2[N], 3[A])] open[W] whether[Y] this property constitutes a primitive element. Y = (V)fin

(8) (9)

It[X] is[Z] too bad[W] we can't[Y] find the answer. It[X] is[Z] a shame[W] she left[Y] so early.

The predicate is a verb which can have as its grammatical subject an item introduced by a wh-word (an infinitive phrase or a clause) 21.3.

Examples Y = T02 (1) (2)

It[X] must[Z] be decided which[W] way to[Y] choose. It[X] has[Z] not yet been decided where[W] to[Y] go.

EXPLETIVE

341

Y = (V)fin (3)

It[X] should[Z] be established what[W] members of this highly structured microsystem enter[Y] the said paradigm.

The predicate is a copula with a special complement which can have the above item as its grammatical subject 21.4.

r = 1st completive, 2nd completive Examples 2

Y = TO (1) (2)

It[X] may[Z] be very important[U] which[W] method to[Y] employ. It[X] may[Z] be considered important[U] which[W] method to[Y] employ.

Y=(V)fin (3) (4) (5)

It[X] remains[Z] questionable (a question) [U] what[W] kind of person he is[Y]. It[X] is[Z] not yet clear[U] what[W] connections they have[Y] with the processes going on in the human mind. But it[X] is[Z] not at all clear[U] why[W] there should[Y] be two occurrences

342

(6)

EXPLETIVE of the lexical item man. It[X] was[Z] left unclear[U] by Mr. Stevenson where[W] such reserves would[Y] be found.

Constructions of the type It was two months before he recovered (21.5)

21.5.

CONJ = SINCE 2 , AFTER 2 , BEFORE 2 Example It[X] is[Z] now almost two years[W] since[Y] Albion agreed to sell the talented midfield man to Leeds for £ 177,000.

Direct object manifested by anticipatory IT (21.6- 21.8) General condition on Syntagms 21.6 and 21.7

EXPLETIVE

343

Constructions of the type make it clear that... 21.6.

α = ger, TO 2 , THAT 5 , WHETHER 1 , IF 2 , (V)fin Examples Y = TO2 (1)

I find[Z] it[X] very promising[W] (of[W] great importance[Q]) to[Y] properly analyze these data.  = CONJ

(2)

Repeated clinical experience makes[Z] it[X] clear[W] that[Y] games are initiative in nature.

Y=(V)fin (3)

His reaction makes[Z] it[X] clear[W] a game is[Y] in progress.

EXPLETIVE

344

Constructions of the type make it clear why .. 21.7.

Y = TO2 (1)

Everybody finds[Z] it[X] quite puzzling[W] what[U] method to[Y] adopt. Y=(V)fin

(2)

The preceeding remarks make[Z] it[X] clear[W] why[U] we shall[Y] base the system of logic on propositions.

EXPLETIVE

345

Constructions of the type doubt it that... 21.8.

► The feature (2[it-THAT]) marks, in the GP of a verb, its capacity to take the constructions illustrated below. ◀ Examples (1) I doubt[Z] it[X] that[Y] he is here. (2) He regrets[Z] it[X] that[Y] the project is almost over. (3) *He grasped[grasp ≠ (2[it-THAT])] it that the project was almost over. ► For an analysis of these constructions, see Stockwell et al. 1973: 551-553. ◀

Clefting IT (21.9 -21.10) General condition on Syntagms 21.9 and 21.10

EXPLETIVE

346

Constructions of the type it is NP (that) 21.9.

► Parentheses surrounding the Y o - · ►oW branch in the syntagm indicate the optionality of that contextual component. This means that Syntagm 21.9 holds even if there is no wh-word etc., cf. example (2) below. ◀ Examples U= ΔΝΡ (1) (2)

It[X] is[Z] these clusters[U] that (which) [W] enable[Y] the parser to build up networks. It[X] was[Z] a tax break[U] I counted[Y] on.

EXPLETIVE

347

U = (V)ger (3)

It[X] was[Z] buying[U] a new hat that[W] I enjoyed[Y]. U = (Prep)

(4) (5)

It[X] must[Z] have been in[U] the war[T] (where (that)[W]) I came[Y] to understand human nature better. It[X] is[Z] to[U] John[T] (that[W]) I spoke[Y].

Constructions of the type it is because ... that... 21.10.

α = (Conj, subord), (Prep), (Adv), TO 2 Examples W = (Conj, subord) (1) (2)

It[X] is[Z] only when[W] one of them disturbs the healthy balance that[Y] analysis and reorganization are indicated. Perhaps it[X] is[Z] just because[W] straight professionals are not playing this game that[Y] they are so seldom caught. W = (Prep)

(3)

It[X] was[Z] thanks to[W] his help only that[Y] I had managed to obtain these results.

348

EXPLETIVE W = (Adv)

(4)

It[X] seemed[Z] to be there[W] that[Y] Chuck first noticed he was alone. W= TO2

(5)

It[X] must[Z] have been to[W] ensure better readability that[Y] our editor chose this particular type. Dummy grammatical subject manifested by anticipatory THERE (21.11)

21.11.

Examples (1) (2)

There[X] seem[Z] to exist some hairy problems[Y]. Over the edge there[X] peeped[Z] a boyish face[Y]. Dummy grammatical agent manifested by anticipatory THERE (21.12)

21.12.

CONJUNCTIONAL

349

Example It was too late for[Z] there[X] to[W] be[U] any business[Y].

5.

Conjunctional phrases

All the syntagms from 22 through 26 describe constructions headed by different conjunctions. There are five SSRel's appearing in these syntagms: - the subordinate-conjunctional SSRel; - the coordinate-conjunctional SSRel; - the predicative-conjunctional SSRel; - the completive-conjunctional SSRel; -

the absolute-conjunctional SSRel; Conditions on occurrence and cooccurrence of conjunctional SSRel's

1) 2) These conditions are interpreted as follows: 1) If a node in a SSyntS is labeled by a lexeme with the feature (Conj), then one of the above-listed conjunctional SSRel's must stem from it. 2) No two conjunctional SSRel's may simultaneously stem from the same node.

22.

Subordinate-conjunctional

syntagms

The subordinate-conjunctional SSRel is non-repeatable.

SUBORDINATE-CONJUNCTIONAL

350

Subordinate conjunction introducing a clause (22.1) 22.1.

► Conditions 1 and 2 express the dependence between the subordinate conjunc­ tion in question and the mood of the verb it governs. Thus AS IF = (2[irreal]) takes only the irrealis; LEST = (2[sbjnct]) requires the subjunctive or the auxiliary verb SHOULD 3 ; WHEN permits only the indicative; etc. Condition 3 states the impossibility of the future tense forms after the temporal and conditional subordinate conjunctions. In much the same way, Condi­ tions 4 and 5 cover some minor restrictions on the co-occurrence of the conjunction with the specified elements in the clause introduced. Finally, Condition 6 permits any subordinate clause to be set off by commas, unless the conjunction that introduces it is T H A T 5 . ◀ Examples (1) (2) (3) (4) (5) (6)

When[X] we consider[Y] chlorine atoms, it appears that... Whether[X] this is[Y] done by producing one structure initially or by eliminating anomalous structures is unimportant. The above sentence insists on factuality, whereas[X] the corresponding construction merely announces[Y] the fact. He has successfully shown that[X] this relationship should[Y] be re­ examined. It is all right for people to accept help providing[X] it takes[Y] them a long time to do so. To learn a Pidgin is clearly simpler because[X] the task of learning lots of

SUBORDINATE-CONJUNCTIONAL

351

exceptions is[Y] mostly removed. (7) (8)

This feeling is described, or rather presented before the group it were[Y] a rare flower which should be regarded with awe. Lest[X] the date be[Y] thought of as involving exceptional properties, we can show here ...

how[Z] large the probability is[Y], the expectation decreases at a fixed rate. (10) The[X = THE 3 ] less[W = comp] effectively White can[Y] play, the more he learns.

(9)

Subordinate conjunction introducing an infinitive phrase (22.2) 22.2.

Examples (1)

for him to[Y] have a secure win, his means must be honourable.

(2)

I'll write down the number

not to[Y] forget it.

(3) (4)

She crouched down to[Y] warm herself. We should know whether[X] to[Y] apply these restrictions.

Compound subordinate conjunction (22.3) 22.3.

352 SUBORDINATE-CONJUNCTIONAL / COORDINATE-CONJUNCTIONAL ► This syntagm covers such compound subordinate conjunctions as except that, except when, provided that, etc◀ Example It is unimportant, except[X] that[Y] from the standpoint of efficiency and time the first alternative seems more attractive. 23.

Coordinate-conjunctional syntagms

The coordinate-conjunctional SSRel is non-repeatable. Moreover, it is incompatible with the subordinate-conjunctional, the predicativeconjunctional, the completive-conjunctional and the absolute-conjunctional SSRel's: see the general statement on page 349.

Coordinate conjunctions in constructions with conjoined constituents (23.1) 23.1.

r = coordinative, descriptive-appositive Examples (1) (2)

Digital techniques will be used to automatically accept, store[Z] and[X] update[Y] aircraft flight data. Its brevity[Z] and[X] paucity[Y] of exemplification made the "Outline" an unsatisfactory work.

COORDINATE-CONJUNCTIONAL (3)

353

(7)

It does not explain why a definite object[Z] was singled out for such a construction, but[X] not an indefinite one[Y]. It is the first[Z] and[X] the only[Y] example. There is no single policy[Z] but[X] rather a multiplicity[Y] of ad hoc programs. It is not stated whether[Z] and[X] under[Y] which conditions this sequence should be followed. The structure is contained in, depends[Z] on or[X] modifies[Y] this string.

(8)

Let's try to test now the above principle[Z], Boyle's

(9)

The new gas[Z] (new[Y], properties.

(4) (5) (6)

to the scientific world) had unusual

► Constructions with a coordinate conjunction introducing an independent sentence of the type you did it are described by Syntagm 9.13. Constructions with a coordinate conjunction introducing a sentential object (or some other disjunctive constituent), as in (10): (10) She smoked,

only.

are described by Syntagm 15.9. Parallel to its inability to be a postposed dependent of a preposition (*on that, *with that, etc. ), THAT 4 cannot be a dependent of a coordinate conjunction [Condi­ tion 1]: (11) *The troop transports that keep a loose formation, remaining in comfortable radarscope view of the leader, but that try to follow well-established sky­ ways, have to reduce altitude before turning inland. In (11), both that must be replaced by which: the second one because that cannot depend on but, and the first one because that cannot be coordinated with which: *... that keep ...,but which try... (the latter fact must be accounted for in the standard function COORD). ◀

354

COORDINATE-CONJUNCTIONAL Quasi-coordinate conjunctions in constructions with conjoined constituents (23.2)

23.2.

α Φ (Art), (partic), (coord), (quasi-coord), (V)fin Example Such imagery may be used to describe the relations[Z] between the powers which control certain world a r e a s , t h o s e [ Y ] between the world areas themselves. Conjunction BUT 2 in constructions of the type nothing left to do but to wait (23.3) 23.3

r = attributive, adverbial; α Φ (Art), (partic), (coord), (quasi-coord), (V)fin, (N,pron,pers)

COORDINATE-CONJUNCTIONAL

355

► But 2 cannot be regarded as a preposition since it admits the nominative case of personal pronouns as Y; cf., for instance, (1): (1)

Nobody could have done it, but he.

Cf. also the comments on p. 356. ◀ Examples (2) (3) (4) (5)

There was nothing[Z] left but[X] to[Y] wait. You could not[W] help[Z] but[X] tell[Y] him everything. He works every[W] day[Z] but[X] Sunday[Y]. It is anything[Z] but[X] interesting[Y].

Comparative conjunctions (23.4) 23.4

α ≠ ( A r t ) , (partic), (coord), (quasi-coord), (N,pron,pers) Examples (1) (2) (3) (4) (5) (6)

... a distinction that is easier to feel than[X] to[Y] express. None of his proposals went further than[X] the demands[Y] for a fifty per cent rise in tax rates. These cannot occur embedded as[X] a complement[Y]. These data can be thought of as[X] involving[Y] exceptional properties (... as[X] in[Y] the following examples). Some of the errors are very minor, as[X] when[Y] an adverb between two verb phrases is attached to the wrong phrase. Examples here include coordination gapping and pronominalization

PREDICATIVE-CONJUNCTIONAL

356

(7)

24.

almost every syntacti phenomenon[Y] which has been studied to some extent. He begins to see the child as[X1] a person[Y] who needs him rather than[X 2 ] as[Y 2 ;X 3 ] a rival[Y 3 ].

Predicative-conjunctional syntagm See page 128 for comments about the predicative-conjunctional and the comple­ tive-conjunctional SSRel's. The predicative-conjunctional SSRel is non-repeatable.

24.1.

► The condition of this syntagm represents the following fact of contemporary English. After comparative conjunctions (as, than), as well as after BUT 2 there is vacillation between nominative and oblique case forms of personal pronouns in the SUBJECT complement function. Normative grammar insists on the nominative case form but the oblique case form is felt to be more natural, particularly in informal style (especially so in American English). This usage is covered by our syntagms, which means that in a predicative-conjunctional construction with a personal pronoun the latter can be assigned either of the two cases: nominative or oblique. But if the personal pronoun appearing after a comparative conjunction (or after but) is in the oblique case, then under analysis this conjunctional construction may be ambiguous. This happens, however, only when the clause analyzed has a subject and a direct object BOTH of which can be taken to be the comparand of Y from the purely grammatical viewpoint. (By 'comparand of Y' we mean an item to which Y is compared in a comparative construction: in (1) below, the comparand of she is no one, and in (5), the comparand of her is I or Carol.) But anticipatory or clefting IT cannot be a comparand of anything, and if the subject of the clause is IT 5,6 , then Y in the oblique case cannot be analyzed as subject complement; in this case it ceases to be ambiguous. That is what the condition of Syntagm 24.1 says.

PREDICATIVE-CONJUNCTIONAL / COMPLETIVE-CONJUNCTIONAL

357

Examples (1) - (4) all have the pronoun in the nominative case, i.e. they are unambiguous (though slightly bookish). Example (5), on the contrary, with the oblique case form pronoun, is ambiguous; therefore, it is described by both this and the next syntagm. ◀ Examples (1) (2) (3) (4) (5) (6)

25.

No one[W] could[Z] make a better omelette soufflée than[X] she[Y] [S. Maugham]. They[W] really are[Z] more disinterested, virtuous, and spiritual than[X] I[Y] [S. Maugham]. Other people[W] are[Z] even more stupid than[X] he[Y]. Everybody[W] but[X] he[Y] saw[Z] all of them. I[W] will[Z] take more pleasure in seeing Carol than[X] her[Y]. The book had become more important to me than[X] her[Y], than all that had happened to us [H. Miller].

Completive-conjunctional

syntagm

The completive-conjunctional SSRel is non-repeatable. 25.1.

Examples (1) (2) (3)

I like[Z] you[W] more than[X] him[Y]/... It is easier to catch[Z] a mouse[W] than[X] him[Y]. Everybody saw[Z] all[W] of them but[X] him[Y].

► As stated in the comments to Syntagm 24.1, examples (1) and (3) are ambigu­ ous: the oblique case form pronoun can be analyzed as object complement by Syntagm 25.1 or as subject complement by Syntagm 24.1. That is, we can have:

COMPLETIVE-CONJUNCTIONAL / ABSOLUTE-CONJUNCTIONAL

358 (Γ)

I like you more than

[according to 24.1 and meaning 'I like you more than he does' ; the comparand of him is I] and (1") I like you more than [according to 25.1 and meaning 'I like you more than I like him'; the comparand of him is you]. It should be noted that the ambiguity in question is characteristic only of written text. In actual speech, the proper prosody — stress, pauses, and pitch — disambiguate the construction in most instances. Example (2) is unambiguous because the sentence has no appropriate subject: the grammatical subject is the anticipatory IT, which is excluded as the comparand of him. Example (4) below is also unambiguous because of word order (according to the : condition of this syntagm the comparand W of Y must precede (4)

Everybody but[X] him[Y] saw[Z] all[W] of them.

However, (4) will be analyzed by Syntagm 24.1, so that him will be treated as subject complement (and not as object complement). ◀ 26.

Absolute-conjunctional

syntagm

The absolute-conjunctional SSRel is non-repeatable.

26.1.

α = ΔΝΡ (not pron, not prop), (A, not pron), (V)ppres, (V)pass, (Prep), (Adv, not partic, not delim)

ABSOLUTE-CONJUNCTIONAL

359

Examples Y=

ΔΝΡ

(1)

The subject noun phrase, if[X] a pronoun[Y], may become possessivized.

(2)

The weight, when[X] above[Y] three tons and a half, may cause serious disturbances. Y = (A, not pr on)

(3)

Where[X] possible[Y], Friedman leaves the choice to the reader. Y = (V)ppres

(4)

While[X] discussing[Y] electrical conductivity, it is of interest... Y = (V)pass

(5)

... to allow the user the full power if[X] only desired[Y]. Y = (Prep)

(6) (7)

The catalysts are most active when[X] in[Y] solution. A eunuch who came from Port Said Had a jolly good time when[X] in[Y] bedt. Y = (Adv)

(8)

Once[X] abroad (there) [Y], he decided...

► The absolute-conjunctional SSRel has been introduced as a separate SSRel partly due to the fact that its dependent component, which is a kind of reduced clause, can take various types of adverbial modifiers. This property can be more easily expressed in terms of a specific SSRel. ◀

t The end of this limerick is as follows: Nor could any sultana Detect from his manner That he used a banana instead.

PREPOSITIONAL

360

6. Noun phrase 27. Prepositional syntagms The prepositional SSRel is non-repeatable. If a node in a SSyntS is labeled by a lexeme with the feature (Prep) or the lexeme TO2, then the prepositional SSRel must stem from it, except in well-defined cases of permitted or obligatory non-saturation of the corresponding valence slot. Prepositional phrases (27.1 - 27.8)

Preposition governing a nominal phrase {21 1) 21.1.

Examples (1) (2) (3) (4) (5) (6)

A review of[X1] the progress[Y] of[X2] science[Y2], published in[X3] four languages[Y3]by[X4]ICI[Y4],... a mixture of[X] the converted[Y] and the indifferent; The theory itself was seized on by[X1] W. Thompson[Y1] in.company with[X2] either Newton[Y2] or Pascal. Given[X1] the two readings[Y1], the derived reading is assigned to[X2] the node[Y2] N [see above, p. 260]. In[x1] f r o m [ X 2 ] seven[Y2] to[X3] nine[Y3] of[X4] these ten instances[Y4] the exact time is prominent. For[X] text data[Y]one would need definitions with[X2] about as many attributes[Y2].

PREPOSITIONAL (7)

361

It seems necessary to make the growing number of documents known and available to[X] whoever may[Y] need them. (8) The methods are derived from, and[W] answerable to[X], this theory[Y]. (9) And each has influenced, or[W] been influenced by[X], languages[Y] outside his group. (10) ... from[X] a true-to[Z]-life picture[Y].

Preposition governing an adjective and governed by a word of a specific subclass (27.2) 27.2.

Examples (1) (2) (3)

The color of the target varies[Z] from[X1] yellow[y1] to[X 2 ] blue[Y 2 ]. The target varies[Z] in color from[X1] yelow[Y1] to[X2] blue[Y2]. With[Y 2 ] the temperature[Z] above (below)[X] normal[Y] ...

Preposition governing a gerund (27.3) 27.3.

362

PREPOSITIONAL Examples

(1) (2)

The process of[X] entering[Y] structures into a single-entry file... Atomic oxygen was produced by[X] subjecting[Y] molecular oxygen to a microwave discharge. Preposition governing a clause (27.4 - 27.6) Via WHETHER

27.4.

Examples (1) (2) (3)

It answers the question[Z] asked by m a n y w h e t h e r [ Y ] this approach eliminates all of the roles of semantics. Whether or not the message follows the rules is not a reliable indicator[Z] of[X] whether[Y] these processes will break down during its interpretation. In Glasgow the inquiry[Z] by the mayor into[X] whether[Y] the corporation had failed to discharge its duty was held.

PREPOSITIONAL

363

Examples (1)

(2) (3)

For Empdeedocles Love is[Z] no mere metaphor, in[X] that[Y] the sexual attraction of living creatures to one another in ordinary experience is an actual instance of the basic vital force. Blacke's method is sound[Z] in[X] that[Y] he does not claim any formal parallelism for the classification. Given[X] that[Y] speakers are fluent because of their knowldge of the rules of the language, it seems[Z] necessary to conclude...

Via a finite verb or a to-infinitive 27.6.

Examples Y = (V)fin (1) (2) (3) (4) (5)

There is indecision[Z] about[X] what[W] stand the theory should[Y] take. Kleptomaniacs are examples[Z] of[X] how[W] widely this trivial game is[Y] played. ... the limitation[Z] on[X] what[W] deep structure sentences can[Y] be nominalized. Mrs. Oliver was looking somewhat different[Z] from[X] when[W] Poirot had[Y] last seen her. Two California condors knocked their only egg off a cliff while fighting[Z] over[X] which[W] bird should[Y] take care of the rare egg.

364

PREPOSITIONAL Y = TO2

(6)

The question[Z] of[X] what[W] to[Y] do first...

(7)

The question[Z] asked by the Under-secretary of State yesterday exactly what[W] to[Y] do first was unavoidably the most difficult one. Preposition governing an adverb or a preposition (27.7 - 27.8) A temporal or locative preposition

27.7.

Examples Y = (Adv) (1) (2)

from[X] there[Y]; through here; from afar; up to now; before then; until recently;  = ( Prep)

(3)

from over[Y] the hill; since[X] before[Y] Christmas; until[X] a few months ago[Y].

INSTEAD OF and EXCEPT 2 27.8.

365

PREPOSITIONAL Examples (1)

Little Suzan always runs to her daddy

to[Y] me.

(2)

Her move was met with silence

(3)

Some verbs allow only the participle construction, except[X] under[Y] very specific conditions.

with[Y] excited corroboration.

Preposition FOR governing anticipatory (existential) THERE in for-to constructions (27.9) 27.9.

Example We are not close enough for[X] there[Y] to be any intimacy. ► For more details about the SSynt-representation of the for-to construction see pp. 256-257. ◀ "Right-shifted" (= dangling) preposition (27.10 - 27.13) General conditions on Syntagms 27.10 through 27.13

366

PREPOSITIONAL With a wh-word

27.10.

Examples (1) (2) (3)

There is a certain transformation which[Y] I will[W] refer to[X] as SubjectRaising. Whatever[Y] Evans could[W] get away with[X] seems now immaterial. The analysis that[Y] pupils have[W] struggled for years with[X] is finally abandoned.

With an NP including a wh-adjective 27.11.

Example People who claim to be neutral soon show whose (which) [Z] side[Y] they are[W] neutral on[X].

PREPOSITIONAL

367

With a free relative clause 27.12.

Example What[Z] I couldn't[Y] reach with my hands I threw[W] snowballs at[X].

With WHERE 27.13.

Examples (1) (2)

Where[Y] does[W] he come from[X]? ... the country where[Y] he came[W] from[X] ...

368

PREPOSITIONAL Preposition in a sequential phrase with numerals (27.14)

27.14.

Examples (1) (2)

five[Z] to[X] eight[Y] hours; Levels 1[Z] through[X] 3[Y] define structural units.

''(

Postpositive" prepositions (27.15)

27.15.

Examples (1) (2)

three years[Y] and[Z] five months ago[X]; a little while[Y] ago[X]; an hour[Y] or so back[X]; a couple[Y] of months ago[X]; These blemishes[Y] notwithstanding[X], the book contains much that is well worth reading.

► The phrase long ago is unique, with an adverb and not a noun being dependent on ago, and is therefore considered to be a phraseological adverb phrase. Phrases of the type a week {three hours) before are described via the restrictive

PREPOSITIONAL

369

SSRel(Syntagm l5.4,p.312): (3)

a week

before,

in view of the two possible types of usage: firstly, in constructions such as a week before his departure, with the prepo­ sition before governing departure; and secondly, the independent use of before as an adverb, as, e.g., in I have met him before. ◀

Particle TO 2 introducing an infinitive (27.16) 27.16.

► Condition 1 tries to state, in a rather approximate manner, the restrictions imposed on so-called split infinitives. In particular, this splitting is impossible or undesirable with auxiliaries: η

(1) *to immediately have followed, · to immediately be stopped. In fact, native speakers find the splitting of infinitives much worse in the case of the perfect HAVE than in that of the passive BE. We chose, however, to disregard this subtlety here. For split infinitives, cf. also examples (2) through (4) below. A (2) (3)

to[X] immediately[Z] follow[Y] ; to[X] frequency[Z] modulate[Y].

► In (3), what we have is a compound lexical verb. Nevertheless, since such formations are rather free and unpredictable, we are forced to treat them, on the surface-syntactic level, as two separate nodes, and this brings us to subsume construc­ tions of type (3) under split infinitives. 4 (4)

I have tried to[X] consciously[Z] stop[Y] worrying about it.

370 (5) (6)

PREPOSITIONAL The program decides for itself with[W] which strings to[X] deal[Y] first. In order for[W] a deep generalization to[X] be[Y] stated, the constructions have to be the same at some level.

Preposition BESIDES or EXCEPT 2 introducing an infinitive (27.17) 27.17.

Examples (1) (2) (3) (4) (5)

I hardly remember what[Z] I did besides[X] read[Y]. Fanny had nothing[Z] to do except[X] go[Y] to parties and give them. What[Z] could I do then, except (besides)[X] read[Y]? The subjects did nothing[Z] except[X] lie[Y] on the bed. Jihan Sadat has nothing[Z] to do except[X] (to)[Y] prepare for a Ph.D.

► The distribution of bare vs. ίο-infinitives after the prepositions in question is fairly complicated and varies considerably from speaker to speaker. It is contingent, among other things, on the syntactic governor of the head of the Prep + V inf phrase, i.e. of Ζ in Syntagm 27.17 (more precisely, on whether this governor is the verb DO 2 ) and on the syntactic dependent of Ζ (on whether this dependent is the phrase to do) ; a host of other factors also intervene. However, we need not concern ourselves with all these problems here: the conditions that determine the selection or absence of TO 2 (in the Prep + V inf construction) belong to deep syntax and need not be stated in this syntagm. ◀

DETERMINATIVE

371

* From here on, until the end of the section on subordinate syntagms, the noun phrase in the strict sense is being described. In other words, we deal below with structures of nominal modification found in contemporary English. In this connec­ tion , Hough 1971 should be mentioned: this book contains a useful survey of all types of modification (including modification of verbs, adjectives, and adverbs) partly summarizing Francis 1958 and Nida 1966 and partly adding new data from a represen­ tative sample of American newspaper texts. 28.

Determinative syntagms The determinative SSRel is non-repeatable. Conditions on the cooccurrence of the determinative SSRel with other SSRel's

i.t., a state change:

but not *

Cf.,

however, (1)

a finite-state or a phrase-structure grammar

with the following analysis:

(1') As example () shows, a determiner can be dependent on a dependent member of the compositive SSRel, but only in case of a conjoined composite dependent; this case is taken care of by the condition introduced by if not; — · · •-►is explained on page 376. For more details, see Syntagms 28.6 (page 376) and 28.7 (page 377); cf. also p. 406.

2)

if Ζ is used referentially,

i.e., if Ζ denotes a concrete object or phenomenon: proposal,

372

DETERMINATIVE

but not

for children, not 'a room belonging to certain children', or = 'a doctorate' and not 'a degree of a certain doctor', etc., i.e., utterances in which children's and Doctor's are used nonreferentially: this is the so-called qualifying, or descriptive, genitive. In many instances, although by no means in all, such combina­ tions are, to a greater or lesser extent, phraseologized, and can be described by means of non-standard lexical functions.

Constructions with an article (28.1 - 28.2)

Indefinite article 28.1.

► A/AN stands for a specific lexeme — the indefinite article. Phrases of the type a dozen, a hundred, a million are described by Syntagm 18.3, page 329. ◀ Examples (1) (2) (3) (4) (5) (6)

The antithesis is a[Y] more complicated problem[X]. ... into a[Y] longer, more detailed report[X]. ... a[Y] restrictive natural language system[X]. The first of these cases seems an[Y] easy[Z] one[X] to substantiate. ... a[Y] very good[Z] and useful one[X]. a[Y] hard day's[X] work.

► The cooccurrence of the indefinite article with proper nouns, as in a certain

DETERMINATIVE

373

Jones, a Mary she had never met before, etc., as well as with many abstract and mass nouns is restricted by a number of fairly complicated contextual and semantic conditions. But since the whole range of lexical cooccurrence phenomena (of which cooccur­ rence with articles is but a particular case) is described at the level of DEEP-SYNTAC­ TIC REPRESENTATION, i.e., in the DSyntS of the sentence, we need not explore and formulate the corresponding conditions in Syntagm 28.1. For the use of English articles see, in particular, Hewson 1972. ◄

Definite article 28.2.

α = (Ν, not pron), (A, not pron), (V)pass, (Num), ONE 2 ►Nominalized adjectives and numerals cannot be directly dependent on nouns, except in appositive constructions of the type Gottfried the Beautiful, etc., which are covered by Conditions 1 and 2. 4 Examples X = (N, not pron) (1) (2) (3) (4)

To challenge the[Y 1 ]direct effects[X1]is to challenge the[Y 2 ] theory[X 2 ] of deep grammar. ThefY^ proponents[x 1 ] of the[Y 2 ] key-word systems[X 2 ]. ·. I do not mention the[Y] many others[X]. ...the[Y]poor Mary[X].

X=(A) (5) (6) (7)

The[Y] firstfX] of the two models... This project, the[Y] first[X] in this country to provide raw translations... This description will be acceptable as the[Y] best[X] currently available.

DETERMINATIVE

374 (8)

The[Y] rich[X] always gain. Χ = (V)pass

(9)

They sent a copter to rush the[Y] wounded[X] to the hospital. Y = (Num)

(10) In the[Y] final two[X] the following questions are considered. X = ONE2 (11) The more flexible formulation was the[Y] one[X] proposed. ►As with Syntagm 28.1, the complicated conditions restricting the cooccurrence of the definite article the with proper nouns and certain other nouns have been intentionally avoided here because they belong to a deeper level of representation.◄

Constructions with demonstrative pronouns (28.3) 28.3.

X=(N) (1) (2) (3)

I use that[Y] distinction[X] in discussing the other markers. Our ability to identify these[Y] effects[X]... It works for those[Y] Montreal children[X] that... X = (Num)

(4)

Aside from the different interpretation of these[Y] two[X], the factive nominal permits no insertion of the preposition of

DETERMINATIVE

375

Constructions with other determiners (28.4 - 28.5) Quantifying or possessive pronoun with a noun 28.4.

Examples (1) (2) (3)

Every[Y = (adsg!)] serious student[X] of the brain... He visits me every[Y] ten[Z] days[X]. We cannot say that any[Y1] one[X1] of them is embedded inside of any[Y2]

(4) (5)

other one[X 2 ]. He seeks employment in whatever[Y] field[X] he chooses. Its[Y] members[X] have been chosen...

Possessive or demonstrative pronoun with a gerund 28.5.

Examples (1) (2) (3)

his[Y] criticizing[X] the book before he read it; Poor Mrs. Allaby never looked at a young man without an eye to his[Y] being[X] a future son-in-law [S. Butler]. This[Y] showing[X] visitors around wastes time.

DETERMINATIVE

376

Constructions with recurrent determiners modifying conjoined preposed noun dependents (28.6 - 28.7)ϯ

Left conjunct 28.6.

ι

Examples Y = (Art) (1) (2)

a[Y] mathematical[X] and a [U] physical[W] problem[Z] ; a[Y] professional[X] or a[U] business[W] journal[Z]; a[Y] businessfX] or

t Cf. Men'kova 1969.

DETERMINATIVE

(3) (4)

377

a[U] professional[W] journal[Z]; thet[Y1]firstp[X1], the[U 1 ,Y 2 ] second[W 1, X 2 ] and the [U2 ] fourth[W2] levels[Z]; Its selection is determined by the[Y] second[X], not the[U] first[W] NP[Z]. Y = (det)

(5) (6)

my[Y] first[X] and your[U] third[W] child[Z] ; We need some[Y] left-handed[X] and some[U] right-handed[W] pitchers[Z] on this team.

Right conjunct 28.7.

Example My argument does not depend on any particular choice between a[U] phrase

►Our approach to the respective constructions in Syntagms 28.6 and 28.7 is determined by the syntactic pattern exemplified in the above example. Any other

378

QUANTITATIVE

arrangement of arrows would fail to reflect the distinctions between the two syntactic interpretations possible in such sentences; namely, noun W is conjoined either with noun Z, which configuration has the meaning '... between the phrase structure and the transformational introduction', or with adjective X, the meaning being '...be­ tween the phrase-structural introduction and the transformational introduction'. ◄ 29.

Quantitative syntagms

The quantitative SSRel is non-repeatable, except when a noun is simultaneously modified by the quantitative word ALL and a cardinal numeral, cf.:

(1)

Constructions with a cardinal numeral or a symbolic denotation of number (29.1) 29.1.

Examples (1) (2) (3) (4) (5)

thirty[Y] strings[X]; a hundred[Y] characters[Y]t; three[Y] and a half years[X]; a dozen[Y]bottles[X]t; Our talkative lady can say 128 billion[Y] sentences[X] of this type. There is but one[Y] clause[X] left. At one[Y] end[X], human language is connected with the world of human experience...

t The numeral-junctive SSRel between the indefinite article a, on the one hand, and hundred, dozen etc., on the other, is described by Syntagm 18.3, page 329.

QUANTITATIVE (6) (7) (8)

379

k[Y] letters[X] ; (m-2)2[Y] steps[X] ; five[Y] sixths[X] of the weight... one[Y] and[W] a half tons[X].

The quantitative phrase as a compositive dependent (29.2) 29.2.

Examples (1) (2)

four[Y]-line[X] poem[Z] ; the 98 million[Y] dollar[X] observatory[Z] ; forty[Y]-page[X] paper[Z] ; 180[Y]-ft[X] reflector[Z] ; two[Y]-sentence[X] text[Z]; two[Y]-response[X] designs[Z]. Constructions with abbreviations and words of the type o'clock (29.3 - 29.5)

With abbreviations

380

QUANTITATIVE Examples

(1) (2)

1000[Y]ft[X];35[Y]m[X]; 1000[Y] ft[X]-wide telescope.

With currency unit symbols 29.4.

Examples £[X] 60[Y]; annual salary: $[X] 44,100[Y].

With O'CLOCK or PERCENT 29.5.

Examples (1) (2)

at three c'clock; The later system provided alternatives for 5.3[Y] percent[X].

QUANTITATIVE

381

Constructions with quantitative words of the type ALL, MANY, HALF, etc. (29.6 - 29.10) With ALL or BOTH 29.6.

[ALL 1 means 'everybody', ALL 3 means 'whole, in toto'.] Examples (1) (2)

In this calculus all[Y] conceivable cases[X] are foreseen. The machine is told to stack all[Y] these[Z] blocks[X] on top of one another.

(3)

Transformational rules are applied to both[Y] these[Z] deep structures[X]. With DOUBLE, HALF, etc. before an NP

29.7.

α = (Ν, not pron, not prop, not abbr)sg, (Num, subst); r = determinative, possessive, numeral-iunctive;

382

QUANTITATIVE Examples X=(N)

(1)

double[Y] the[Z] mentioned sum[X] ; double[Y] her[Z] age[X] ; half [Y] this[Z] amount[X]; half [Y] a[Z] mile[X]; (2) ... withfive[W]times[Y] the[Z] initial force[X] ; twice[Y] her mother'sfZ] weight [X] ; (3) . The Harvard Syntactic Analyzer delivered the results only about half[Y] the[Z]time[X]. X = (Num) (4)

half[Y] a[Z] dozen[X] tons; half[Y] a[Z] million[X] blocks.

With DOUBLE, HALF, etc. before a free relative clause with WHAT 29.8.

Examples (1) (2)

double[Y] what[Z] it was[X] ; half[Y] what[Z] he has[X] had before.

QUANTITATIVE

383

With MANY etc.

29.9.

Examples (1)

many[Y] rather startling observations[X];

(2)

the demand for more[Y] papers[X]...

► In constructions of types (3) and (4): (3) What we've been licking for more than ten years...ϯ (4) In less than twenty days the entire stock of radio sets was bought up. the words more and less are considered to be the heads of the respective phrases (ΔΝΡ):

Constructions of the type (5)

five more tanks = five tanks more,

t Example (3) is borrowed from the translation, by Alexander Zholkovsky, of a jocular Russian song very popular in Moscow immediately after Khrushchev had been toppled from power and proclaimed both a "voluntarisf and non-existent at the same time. The full and unexpurgated text of the translation goes like this: We surprised the whole of Europe, Bourgeoisie and working class: What we've been licking for more than ten years Has just turned out to be the wrong ass... But the people are not snarling, Wait for the congress to be run — Know that the Party, our darling. Will supply another one!

384

QUANTITATIVE

where we have more = MORE 3 (i.e., a lexeme different from MORE 2 involved in this syntagm), are represented as follows: (5') or (5") They are accounted for by Syntagm 15.2(2), p. 309; see also comments on p. 313.

With MANYA... 29.10.

Examples many[Y] a[Z] rich man[X]; many a time. Constructions with a personal pronoun as the head (29.11 -29.12)

With a cardinal numeral 29.11.

QUANTITATIVE / MODIFICATIVE

385

Examples (1)

we four; for you five; with us six.

► Interestingly enough, the third person pronoun THEY is not admitted in this syntagm: (2) *they three; *for them five; *with them ten; the only right way to express the corresponding meaning is by using the elective syntagm: (2')

(all) three of them; for (all) five of them; with (all) ten of them.

The same restriction applies to the descriptive-appositive SSRel, see Syntagm 36.2, page 424. ◄ With ALL or BOTH 29.12.

Examples we all; from you both; with them all (both). 30.

Modificative syntagms The modificative SSRel is repeatable.

Standard preposed adjectival modifiers (30.1) 30.1.

MODIFICATIVE

386

α = (Ν, not pr n), (Num), ONE 2 ; β = (Α, not det, not pred!, not postpos!), (V)ppres, (V, not postpos!) pass ► Condition 1 indicates that in English it is impossible to have a preposed adjectival modifier with right-hand dependents, the latter being inserted between the modifier itself and the word it modifies. The only exceptions Rule 30.1 permits are the following three: (i) An adjective with enough: (1)

a good  enough argument.

(ii) An adjective with right conjuncts: (2) (iii) A hyphenated adj ective phrase, particularly, an adj ective in a sequential phrase or an adjective with either a right-hand attribute or an object: (3) In a sentence like (4)

There are current in popular literature allusions to "Brooklynese" or "honey-chile" accents.

MODIFICATIVE

387

the string current in popular literature is by no means a preposed modifier of allusions but rather belongs to the verbal phrase following the anticipatory there:

(4') Note that preposed modifiers with left-hand dependents are quite common: (5) Conditions 2 and 3 stipulate that no determinative or quantitative dependents of the modified word can be inserted between it and the adjective modifier, except in constructions of the type such a ( what a ) day, as well as in constructions described in Syntagm 30.2. \ Examples Y=(A) (6)

(7) (8) (9)

the same[Y1] prepositional[Y 2 ] phrase[X 1,2 ]; a syntacticfY] selection[X'] restriction[X"]; the first[Y] chapter[X]; asecond[Y 1 ] such[Y2] class[X 1 , 2 ]; a better[Y] friendfX] than I've ever had... Standard languages arise from urbanfY], upper-class[R] dialects[X] [cf. Condition 5]. In the final[Y] two[X] we shall... A model of language should avoid extralinguistic[Y1] explanations[X 1 ] where intralinguistic[Y2] ones[X 2 ] are readily available. Y = (V) ρpres

(10) the preceding[Y] three chapters[X]; the surrounding[Y] walls[X]; scan­ ning[Y] radar[X'] device[X"]. Y=(V)pass (11) the captured[Y] troops[X]; the intended[Y] referent[X]; (12) These are not only the referentially-based[Y] features[X]. (13) We are in poorly explored[Y] territory[X] here. ► Mention should be made of certain modificative constructions in which a dependent of the preposed adjectival modifier is located to the right of the word the latter modifies, thus creating non-projectivity:

388

MODIFICATIVE

The special character of such constructions is determined, however, not by the nature of the modifier itself but rather by the type of the surface-syntactic relation between it and its disconnected dependent. Three different SSRel's are possible here: the 1st completive, the comparative or the attributive SSRel. As no specific feature of the modificative SSRel appears in these possible cases of non-projectivity, the latter are described in the syntagms corresponding to the above-mentioned SSRel's, namely, in completive, comparative and attributive syntagms. We will confine ourselves here to only a few examples: (14) through (21). ◄ (14) This rule can be system will be able to produce 'A'.

can know that the

(15) It is quite draw. (16) That would be quite a different[Y] matter[XJ from any attempt to divide utterances into the meaningful and the meaningless. (17) This relation is given by such[Y] relations[X] as consist in the correspon­ dence of small and capital letters. (18) ... to the same[Y] level[X] as was achieved before. (19) They contain no other[Y] expression form[X] than what appears on their lexical expression list. (20) ... a higher[Y] level[X] than it had been presupposed; (21) ... the cheapest[Y] car[X] in this country. ►The relative ordering of several preposed adjectives is a challenging problem, which, however, cannot be tackled here. The fact that we say a large juicy apple (*a juicy large apple) or principal foreign expenditure (*foreign principal expenditure) must be accounted for, in the first place, by a semantic classification of adjectives: 'size', 'color', 'nationality', 'general evaluation', etc. (E.g., usually an adjective of size precedes one of color: a small white kitten rather than *a white small kitten.) But such a classification is not sufficient for determining the linear order of adjectives, since in many cases the relative arrangement of two adjectives depends on their semantic relation to the head and on the communicative organization of the phrase; cf. (22): (22) a. b.

American liberal opinion vs. a liberal American society; an American religious convention vs. a religious American family.

MODIFICATIVE

389

See Annear 1964, Vendier 1967, and Farsi 1975. Rules for the ordering of several preposed adjectives are part of global word order rules, mentioned in §1 of this chapter, see p. 180ff ◄

Constructions of the type too serious a matter (30.2) 30.2.

L = HOW, HOWEVER, SO, TOO 2 , AS 7 , NO Examples (1) (2) (3) (4)

How[Z] good[Y] a[W] hostess[X] are you? However[Z] clever[Y] a[W] retort[X] it may have been, it was disrespectful. .. .so[Z] tempting[Y] an[W] opportunity[X] ; so[Z] very nice[Y] a[W] girl[X] ; too[Z] formal[Y] a[W] solution[X] ;

(5)

as[Z] big[Y] a[W] house[X] as[T] you could[U] wish.

► However, we do not get (5'): (5') *as big a house as a palace;

MODIFICATIVE

390

the right way to express the meaning of (5') must be (5"): (5") a house as big as a palace, i.e., with different word order. This fact is accounted for by Condition 1, which excludes (5") from Syntagm 30.2. ◄ (6)

No[Z] less[Y] an[W] authority[X] than the US Secretary of State has answered the question. Postposed modifiers of various types (30.3 - 30.6)

Expanded adjectival phrase 30.3.

α = (Ν, not pron), (A)superl, ONE 2 ; β = (A, not det, not pron, not pred!), (V)ppres. (V)pass ►Condition 1 accounts for the phenomenon referred to in Smith 1961, namely, that a postposed non-descriptive adjectival can modify only a noun with an indefinite article or with no determiner at all: (1)

I bought a book yellow with age.

but (2) *I bought the book yellow with age. Cf., however, (3)

I bought the book, yellow with age.

MODIFICATIVE

391

where the postposed modifier is descriptive. (4)

John has a car too old to sell,

but (5) *John has the car too old to sell. (6) Here is a man easy to please. but (7) "Here is the man easy to please. [Examples (1) through (7) are borrowed from Smith 1961.] As a rule, cooccurrence of nouns with articles must not be dealt with in syntagms: it is the object of a deeper level of representation, namely the deep-syntactic level. (Cf. remarks to this effect on pages 372-374.) Nonetheless, we feel that the above-men­ tioned case should be described here since it involves an obviously surface phenome­ non: word order (only POSTPOSED modifiers are covered by Condition 1). Condition 2 pinpoints a fairly common phenomenon in English: an excessively 'heavy' modifying phrase in postposition may be disconnected from the grammatical subject which (or a constituent of which) it modifies and moved to the right of the verb phrase — under the condition, however, that the verb phrase does not include any object. Cf. a similar case with the relative clauses in Syntagm 30.7. Note, however, that we abstain from specifying the exact conditions under which such a shift of a modifying adjectival is possible, recommendable or necessary. In other words, for lack of relevant data, we are not in a postition to define what precisely 'heavy' means when referring to an expanded adjectival modifier. We suppose that being heavy might imply containing a specific number of wordforms and/or a specific construction (e.g., a subordinate clause); for the moment we are unable to pursue this matter any further.◄ Examples Y=(A) (8)

He offered his students a problem[X] difficult[Y] enough[Z] to challenge their imagination. (9) This discovery opened up many more fields of inquiry[X] both experimental[Y] and[Z] theoretical. (10) A package[X] such[Y] that[Z] its parameters vary very little... (11) This description will be accepted as the best[X] currently[Z] available[Y].

MODIFICATIVE

392 Y = (V) ppres

(12) ... the unconscious processes[X] underlying[Y] the listener's ability[Z] to understand sentences of his language. (13) A class[U] of grammarsfX] is[W] presented having[Y] the following properties[Z]. (14) A far-reaching agreement[X] has[W] been proposed covering[Y] social as well as production questions[Z]. Y = (V)pass (15) Similar types of evidence point to the active nature of the operations[X] involved[Y] in[Z] grouping words into phrases.

Special single postposed modifier 30.4.

Examples Y = (Λ, ρostpos) (1)

... with phrase-structure rules[X] alone[Y].

(2) (3) (4)

linguistics proper; the people present; the only solution[X] possible[Y]; the best book available; the poorest girl alive.  = (V)pass

(5)

the meaning of all the elements[X] classified[Y] ; the countries concerned; the issues involved; the text used.

► Syntagm 30.4 does not include various contextual constraints that actually influence the appropriateness of postposition for any given single participle. For one

MODIFICATIVE

393

thing, postposition is preferred and may even become obligatory if the participle in question has a homophonous adjective with a different meaning, the latter being obligatorily preposed: cf. in example (5) the elements classified vs. classified [= "wanf] ads /classified[≈ 'secret'] documents, the countries concerned vs. a concerned [≈ 'worried] friend, issues involved vs. an involved [≈ 'complicated'] method, the text used vs. used [≈ 'old1] cars. Then the emphasis on the participle must play a role. Finally, some subtle semantic differences seem to be linked to postposing as opposed to preposing single participial modifiers. Our impression is that different verbs, as a function of their meaning, tend to admit the postposing of their participles more or less readily, and we have tried to make a note of this property (rather than capture it) by using our feature "postpos" as label on some verbs. Serious research is obviously needed in order to describe the construction in question more accurately. (Cf. Tessier 1968 and our discussion of it on pp. 136-137.) Λ

Postposed modifiers of pronouns 30.5.

Examples Y=(A) (1) (2)

something[X] quite[Z] new[Y] (outlandish, amiss); special restrictions on the responses of those[X] present[Y].  = (V) ρpres

(3) (4)

The sense of something[X] needing[Y] to be completed grew strong in him. Those[X] wanting[Y] to go can leave now.

MODIFICATIVE

394 Y = (V)pass (5) (6)

A person has the ability to predict whether something[X] not previously[Z] encountered[Y] belongs to his language Those[X] investigated[Y] refused to answer.

Postposed single modifiers in set phrases 30.6.

Examples secretary-general, mother superior, notary public, surgeon (attorney) general, court-martial ► This syntagm presents an interesting problem. According to the norm, all such expressions pluralize their first element: secretaries-general, mothers superior, ..., courts-martial (see, e.g., the indications provided in The American Heritage Dictio­ nary of the English Language or in Quirk et al. 1972:174). But in informal style (especially, in the United States and in speech rather than in writing) they pluralize the second component: secretary-generals, mother superiors,... (Because of the clash between the norm and the actual usage people tend to avoid the plural forms of the above expressions in writing.) If, however, we are to describe this usage, what is then the SSynt-structure of a phrase of the secretary-general type? Our guess is that if we admit forms like secretary-generals as regular, we must consider the phrase a syntacti­ cally unanalyzable idiom, represented in SSyntS by a single node. ◄

MODIFICATIVE

395

Restrictive relative clauses (30.7 - 30.9)t Relative clause introduced by a wh-word 30.7.

α = (N, not pron, not prop), (N, pron, indef, not wh), ONE2, THAT', WHAT1, WHO t Thorne 1972 discusses some interesting properties of English relative clauses in connection with the reflexive pronouns and parentheticals they may contain.

396

MODIFICATIVE

► Conditions 1-4 reflect certain regular patterns in the cooccurrence of wh-words with the nouns they modify (which are, in fact, far more complex than is implied here). NB: strictly speaking, these conditions need not be stated in Syntagm 30.7. Much like the rest of lexical cooccurrence regularities, they belong to a deeper level of description (correspondence between the deep-syntactic and the surface-syntactic structures). We give them here in order to stake out a relevant problem. A noun modified by a relative clause introduced by a w/?-word, i.e., node X in Syntagm 30.7, is often accompanied by the demonstrative pronoun that/those [=THAT 2 ], which serves as a correlative. This fact may be used (in automatic text analysis) for resolving ambiguity in cases where there are several possible governors of the relative clause in question. This ambiguity will be resolved by preference rules (i.e., outside syntagms as such) which allow the analyzer to select the most plausible analysis out of several formally possible choices: namely, take as the governor for the relative clause being processed the noun on which that or those are dependent. Condition 6 rules out the possibility of a restrictive relative clause following a descriptive relative clause, if they are related to the same noun; see Smith 1964. ◄

Examples X = (N) (1) (2) (3) (4) (5) (6) (7)

Note sentential objects[X] which[Z] may[Y] appear as sentences. The English word unacceptable is composed of three morphemes[X], each one [R1] of[R2] which[Z] has[Y] a particular distribution [cf. Condition 7]. There has long existed a need for a description[X] of English that[Z] will[Y] be acceptable. All methodsfX] of structuring that[Z] I know[Y] about have been defective. Anybody[X] who[Z] has[Y] studied Amerindian languages... There was a time[X] when[Z] you gave[Y] me reason to understand that you did care about him. This difference is seen at that point[X] of the description where[Z] words are[Y] embodied in phonological substance. Disconnected relative clauses, cf. Condition 8

(8)

Numerous systems[X] have[U] been constructed that[Z] rely[Y] on syntactic and semantic analysis.

MODIFICATIVE

397

(9)

An explanation[X] is[U] here offered which[Z] is[Y] adequate for all the cases above. (10) A version[X] of the system was[U] assembled in the latter half of 1971 that[Z] is[Y] intended to take over the purely routine work. (11) The doctor took[P] the bandages[X] off[R] which[Z] had[Y] become encrusted with blood. X=WHAT! (12) When we try to spell out precisely what[X] the first-grader is doing that[Z] the computer is[Y] not, we find it very difficult. X = ONE2 (13) An isolating language is defined as one[X] in which[Z] all words are[Y] invariable. X=

THAT3

(14) The syntactic rules are of two types: thosep[X1] that[Z 1 ] generate[Y 1 ] simple sentences, and those[X 2 ] that[Z 2 ] generate[Y 2 ] complex sentences. ► An English noun can be modified by several (at least, two) non-conjoined restrictive relative clauses, although the result is often far from stylistically elegant: (15) The men [X12] who[Z 1 ] were [Y1] exposed to radiation who[Z 2 ] worked[Y 2 ] part-time are still healthy, but those[X 3 ] who[Z 3 ] worked[Y 3 ] full-time are not [Smith 1964]. ◄

Relative clause introduced by a temporal conjunction 30.8.

398

MODIFICATIVE Example The time[X] since[Y] he left has dragged.

Relative clause without introductory element 30.9.

α = (Ν, not pron, not prop), (N, pron, indef), ONE 2 , THAT 3 ► Condition 2 describes the obligatory non-saturation of one active valence slot in the relative clause. Four cases are distinguished: (i) the main verb lacks one object or complement; (ii) the main verb governs an infinitive which lacks one object or complement; (iii) the main verb governs a preposition which lacks its dependent; (iv) the main verb governs another finite verb which lacks an overt grammatical subject. ◄

MODIFICATIVE

399

Examples Χ = (Ν, not pron) (1) (2) (3) (4) (5)

The transformational mappings[Xj] we [Z1] allow[Y1] are the same as those[X 2 ] he[Z 9 ] desires[Y9] to have[U] available. In order to clarify the line[X] of argument we[Z] intend[Y] to follow[U], certain things will be taken as axiomatic. I shall never forget the drive[X] he[Z] took[Y] me on[R] [J. Galsworthy]. ... the subject[X] A. Zwicky[Z] made[Y] A report on[R]...; Guthrie held himself in readiness for the call[X] he[Z] believed[Y] would[S] one day return him from exile to the service of his country. X = (N, temp)

(6) (7)

The program keeps a tally on the number of times[X] each word[Z] is[Y] encountered. His tomb states the date[X] it[Z] occurred[Y]t. X = (N, manner)

(8) (9)

The world is the way[X] the sentence[Z] says[Y] it is. She knew the reason[X] he[Z] spoke[Y] about it. X = (Ν, ρron, indef)

(10) A11[X] we[Z] can[Y] think[U] is that this is a provoked confrontation. (11) We should think about what[X] it is we[Z] do[Y] when we understand and respond to natural language. X=

THAT3

(12) It may seem that such subtle distinctions as those[X] we[Z] have[Y] drawn[U] between morpheme and morph are something of an idle, scholas­ tic pastime.

t The limerick from which this line is borrowed runs as follows: There was an old man who averred He was sure he could fly like a bird. Cheered by thousands of people, He leaped from a steeple... His tomb states the date it occurred.

MODIFICATIVE / DESCRIPTIVE-MODIFICATIVE

400

Infinitival modifying phrase (= postmodification by infinitive clauses) (30.10) 30.10.

α = (Ν, not pron, not prop), (N, pron, indef, not wh), ONE 2 , THAT 3 Examples (1) (2)

31.

Selectional restrictions are the appropriate terms[X] in which[W] to[Y] state[Z] generalizations about co-occurrence restrictions. ... to provide an appropriate framework[X] within which[W] to[Y] discuss[Z] the status of passive constructions.

Descriptive-modificative

syntagms

The descriptive-modificative SSRel is restrictedly repeatable, i.e., two different descriptive modifiers can appear simultaneously with the item modified, namely: - a postpositive descriptive-modificative adjectival phrase; - a descriptive relative clause. Cf. the following example:

DESCRIPTIVE-MODIFICATIVE

401

(1) Heisenberg'S Principle of Uncertainty, known since 1927, which states that events at the atomic level cannot be observed with certainty, can be explained in terms of everyday experience. Such cooccurrence of two descriptive modifiers is not very current and is furthermore restricted to a noun head only. If the governing member of the descriptive-modificative SSRel is, e.g., an adverb, a verb, etc. (see below, Syntagms 31.3 - 5), this SSRel cannot be repeated.

General condition on descriptive-modificative syntagms If Χ Φ (Ν, pers) or Y = (V) f i n , then comma Y comma. This means than any descriptive modifier must be set off with commas, except a relative clause depending on a personal pronoun (it can, but need not: cf. example (9), p. 404).

Postposed adjectival descriptive-modificative phrase (31.1) 31.1.

► The condition stipulates that an adjectival descriptive modifier cannot be disconnected from its head noun and moved to the right of the main verb (while a restrictive modifier can: cf. Syntagm 30.3, Condition 2). Let it be recalled that adjectival phrases following the verb are considered in this book to be linked to the verb by the modificative-adverbial SSRel, cf. Syntagm 10.1, pp. 288ff. M Examples Y=(A) (1)

Some provision must be made for another kind of shift[X], optional[Y] in

402

(2) (3) (4)

DESCRIPTIVE-MODIFICATIVE some cases but obligatory in others. In Paraguay, all formal publication is in a variety of standard Spanish[X], essentially identical[Y] with that used in other countries. Some of the "satisfactions"[X] obtained under this programming, such[Y] as self-destructing ones, are difficult to recognize. But Brubaker[X], impassive[Y], kept reading the document. y = (V)ppres

(5)

The suffix -ke is characteristic of the synthetic future[X], persisting[Y] alongside the periphrastic future in Souletin. Y = (V)pass

(6) (7)

The promotion[X], so long denied[Y] him, came at last. Anyone who asks this question comes face to face with a mountain of words[X], said[Y] and written over centuries, and representing a myriad of conflicting theories and ideas.

► It might be interesting to give here a series of contrasting examples (suggested by I. Mackenzie): (8)

a.

The garbage, messy beyond belief, was reeking awfully

[the italicized phrase is a descriptive modifier]. b. -The garbage was reeking awfully, messy beyond belief./ • Messy beyond belief, the garbage was reeking awfully. [A descriptive modifier cannot be disconnected from its head; at the same time, the italicized phrase cannot be taken to be a modificative adverbial, see immediately below.] But (9) is fine: (9)

a. b.

The garbage was reeking awfully, putrid beyond belief/Putrid beyond belief the garbage was reeking awfully. The garbage lay everywhere, messy beyond belief. I Messy beyond belief, the garbage lay everywhere.

Here, italicized phrases are modificative adverbiais, which from the semantic viewpoint must be somehow logically related to the state or action described by the verb. It is the lack of this logical link, i.e. semantic incongruence, that makes (8b) unacceptable: it is difficult to see the connection between 'messy' and 'reeking1 , therefore messy beyond belief in (8) does not lend itself to an interpretation as a

DESCRIPTIVE-MODIFICATIVE

403

modificative adverbial.◄

Descriptive relative clauses (31.2 - 31.5) Modifying a noun 31.2.

► Conditions 1 and 2 account for the following features of English syntax described in Smith 1964: (i) A descriptive clause cannot modify a noun already modified by an indefinite pronoun, cf.: (1) *Any (every, each) book, which is about linguistics, is interesting. (ii) A descriptive clause cannot modify a grammatical object or an adverbial modifier used with a verb in the negative form: (2) *He didn't eat the mango, which was overripe. (3) *I did not buy the rug, which would have been perfect for the hall. Some speakers, however, do not agree with Smith's evaluation and find (1) - (3) perfectly acceptable. For their dialects Conditions 1 and 2 must be dropped. Condition 3 bars from this syntagm constructions of the type (4): (4)

He seems an idiot, which he is not. These are accounted for in Syntagm 31.5. ◄

404

DESCRIPTIVE-MODIFICATIVE Examples Χ = (Ν, not ρron )

(5) (6) (7) (8)

El-Quneitra, a town[X], which[Z] once had[Y] a population of 200,000, has been almost totally destroyed. This sentence has a construction with a fifth level of nesting[X], only four levels of which[Z] were[Y] identified originally. Identifying phrases are joined in a series[X] of phrases of the same kind, wherein[Z] each of the joined elements is[Y] of equal importance. The favoured pastime of prisoners is escape[X], some of whose[Z] prac­ titioners, such as Casanova and Baron Tranck, have[Y] become famous. X = (N, pron, pers)

(9)

We[X] who[Z] are[Y] suffering call for the further intensification of interna­ tional solidarity.

Modifying an adverb 31.3.

Examples (1) (2)

Here[X], where[Z] he had[Y] fallen,... Now[X], when[Z] Prof. Hsuy Huy-Zhay is[Y] with Bunker-Ramo,...

DESCRIPTIVE-MODIFICATIVE

405

Modifying a whole clause 31.4.

Examples (1) (2) (3) (4) (5)

The Republicans join[X] the Democrats in referring the question of impeachment to a committee for study, which[Z] means weeks of delay. That he wrote[X] me, which[Z] was[Y] quite surprising, became known to all. Your letter must[X] have come by ship, in which[Z] case it was[Y] probably held up by the strike in the port of Montreal [R. Kittredge]. Daughter may make the first move by being[X] imprudent, whereupon[Z] father finds[Y] fault. He calls[X] New York every five days (which[Z] is[Y] already a strain on his budget). Modifying a non-referential noun or an adjective

31.5.

α = (Ν, not pron, not prop), (A, not pron, not ord, not quant)

406

POSSESSIVE Examples

(1) (2) (3) 32.

It was considered a semi-group[X], which[Z] it is[Y] not. The sentence would be marked as ungrammatical[X], which[Z] it is[Y] not. From then on it would be different[X] — which[Z] it was[Y].

Possessive syntagms

The possessive SSRel is restrictedly repeatable: one noun can be simultaneously modified by two possessive dependents on condition that the latter of the two is used non-referentially, i.e. it must be the so-called 'qualifying' genitive. Cf. [Harris 1963: 4], i.e., ka bicycle made for little girls belonging to the boy ; cf. also: Boston University's women's varsity crew boarding shells [a caption]. 1

Conditions on the cooccurrence of the possessive SSRel with other SSRel's 1)

if Y is used referentially.

This condition coincides with Condition 2 on the cooccurrence of the determina­ tive SSRel, p. 371. 2)

if Y is used referentially.

Thus, the phrase (1)

Chapin's phrase-structure parsing

should be analyzed as follows:

here little girl's is used non-referentially.

POSSESSIVE

407

Possessive marker attacched directly to the dependent component of the possessive SS Rel (32.1) 32.1.

α = (Ν, not pron), (V)ger Examples X=(N) (1) Chomsky's[Y1] reply[X1] to Lamb's[Y 2 ] then unpublished criticism[X2] of Halle's[Y 3 ] argument[X 3 ] against a phonemic level is that Lamb's[Y 4 ] proposals[X 4 ] have no bearing on Halle's[Y 5 ] argument[X 5 ]. (2) ...engineer's[Y] lengthy[X] experience; (3) A man says /ne:hkha:h/ of his father's[Y 1 ] brother's[X1, Y 2 ] son[X 2 ], and so on. (4) The committee's[Y] appointment[X] of John (=John's[Y] appointment[X] by the committee); (5) The summer's[Y] rainless heat[X] ... broke at last [W. Faulkner]. (6) ...to-day's (tomorrow's)[Y] papers[X]; (7) John's[Y] and Mary's committing[X] of adultery shocked the city. ► See Syntagms40.3,p. 462, and 40.5, p. 466, for a representation of the semantic contrast between (7)

John's and Mary's committing of adultery = 'John and Mary committed adultery separately, with third parties', and

(8) (9)

John and Mary's committing of adultery = 'John and Mary committed adultery with each other'. ◄ England's[Y 1 ] Thatcher[X 1 ] and France's[Y 2 ] Mitterand[X 2 ] ·.. X = (V)ger

(10) Bill's[Y] marrying[X] the girl didn't make it easier.

408

POSSESSIVE / COMPOSITIVE Possessive marker attached to the rightmost word in the phrase of the dependent component of the possessive SSRel (32.2)

32.2.

α = (N, not pron), (V)ger Examples (1) (2) (3) (4) (5) (6) 33.

... the man[Y] in the street's[Z] arguments[X]. Details of Lunnik[Y] IIFs[Z] design[X] are published in today's Soviet newspapers. Weston said after a moment[Y] or two's[Z] thought[X]... The girl[Y] I danced with's[Z] brother[X] [F. S. Fitzgerald] = 'the brother of the girl with whom I danced'. Catherine[Y] the Great's[Z] Russia[X]; the day[Y] after's[Z] paper[X].

Compositive syntagms The compositive SSRel is repeatable. Conditions on the cooccurrence of the compositive SSRel with other SSRel's

1) not This condition coincides with Condition 1 on the cooccurrence of the determinative SSRel, p. 371. 2)

if Ζ is used referentially.

This condition coincides with Condition 2 on the cooccurrence of the possessive SSRel, p. 406.

COMPOSITIVE Compound nouns (33.1 - 33.3)

409

410

COMPOSITIVE

► Condition 1 deals with the grammatical number of the dependent component, i.e. Y. In most cases Y must be in the singular regardless of its semantic number: student power 'power of students (*student)\ bird sanctuary 'sanctuary for birds (*bird)\ tax policy 'policy with respect to taxes (*tax)\ etc. Y must be in the plural in the following three cases: - Y is Si plurale tantum which does not denote a physical object, e.g., sports magazine, arms race, customs duties; - Y is the lexeme WOMAN or a lexeme ending in -MAN and X is plural, e.g., women detectives, businessmen soldiers, etc; - Y or X has a special feature in the dictionary to the effect that Y must be plural or else there are special semantic considerations which we are unable to formalize now, e.g., promotions committee, translations editor, etc. Y can be in either the singular or plural if it is a plurale tantum denoting a physical object, e.g., scissor/scissors sharpener, trouser/trousers leg, etc. Condition 2 indicates that of all the potential left-hand dependents of the main component X in a composite phrase only modificative and compositive ones may occur in the gap between X and Y. For more detals about the mutual linear arrange­ ment of compositive and other left-hand noun dependents see Farsi 1975. By introducing Conditions 3 and 4 we try to reflect, if only partially, the regularities of hyphen and quotation mark usage within composite phrases. These regularities, which are fairly complex and too often disregarded in actual texts, may be summed up as follows: (i) There are at least four cases of a hyphen being used in strictly binary composite phrases when Y has no dependents: - X is a noun with the suffix -ing, e.g., the mess-making is...; - X is itself a compositive dependent, e.g., phrase-structure[X] grammar; - Y is an abbreviation or a formula, e.g., Lp-approximation or TGG-adepts; - X and Y are semantically coordinate; they denote two sets of properties jointly possessed by the item of which the composite phrase is the name: e.g., secretarytreasurer (somebody who is both a secretary and a treasurer), fighter-bomber, panty-girdle, sofa-bed, actor-pro ducer, ... Notice that in a composite phrase of the latter type only X pluralizes and thus shows that syntactically it is the head: secretary-treasurers (*secretaries-treasurers), fighter-bombers (*fighters-bombers), etc. In addition, there occur phraseme-like binary composite phrases in which the use of a hyphen, mandatory or optional, is determined by their dictionary entries, e.g., air-conditioning, window-cleaner, boat-ride, etc. (at least, according to British

COMPOSITIVE

411

punctuation norms). (ii) If Y has right-hand dependents which must be located between Y and X and which include items other than an appositive (as in a dress) or a conjunct, then all the constituents of the phrase headed by Y must be hyphenated, or the phrase must be set off by quotation marks, or both, e.g., a stem[Y]-affix and part[Y2]-ofspeech algorithm[X]. (iii) If Y has left-hand dependents or right-hand conjoined dependents (including coordinate conjunctions), then all the constituents of the phrase headed by Y may (but need not) be hyphenated, cf. a floating-point[Y] resistor vs. the first passage[Y] problem for Wiener process; an ear[Y]-nose-and-throat specialist vs. vertices[Y], outsets and paths enumeration. Note that in a composite phrase a hyphen may be used to express semantic contrasts, cf. precedence-and-dependency grammars = 'each grammar is based both on precedence and dependency', vs. precedence and dependency grammars = precedence grammars and dependency grammars = 'certain grammars are based on precedence and others on dependency1 . This phenomenon is discussed in greater detail below, in the section dealing with coordinative syntagms, pp. 466ff. Let it be emphasized that a hyphen is also widely used in English to join not only words but also parts of words, i.e. prefixes and suffixes with stems, cf. quasi-rational, non-regular, -Poisson [sequences], multi-dimensional, pseudo-subordinate, semi-algebra, out-patients, by-elections, tree-like [graphs], penny-wise [program­ ming]. This usage, however, should not be covered by compositive syntagms nor by any other syntagms since it is a question of word-formation. (To process such items during automatic analysis, a special device is required which will ensure their re­ presentation in the form of single syntactic nodes, prior to syntactic analysis.) We would also like to point out that this syntagm describes, among others, the following two phrase types: - a noun preceded by an identifying conventional label, as in PS grammar, Ψ-waves, etc.; - a classifier noun preceded by a proper noun, which is its name, as in New York state, the Hudson River, the Yamal Peninsula. ◄

412

COMPOSITIVE

(2) (3) (4) (5) (6) (7)

non֊lattice[Y] random variables[X]; an object[Yı] r é c o g n i t i o n ^ , Y2 ] computer[Y3] system[X? 3]; a life-size[Y] picture[X]; the VolgafY] River[X]; a large[Uı]-scale[Yı] generairiM-üurposeIY^ digital compuintake[Y 2 ] means[Xı շ]; ter[Y 3 ] s y s t e m ^ շ 3]; engine[Yļ] ah "ProblemIYı] solvingfXl] and decision[Y2] making[X 2 ] by computer" [title of an article]. an eight man delegation ; the Cambridge Language Research Unit members ; verb[U]-head[Y] expression[X]; a first[U]-year[Y] student[X]; phrase[U]-structure[Y] and transformational grammars[X]; The mess[Y 1 ]-making[X 1 = (N, -ing)] is merely a pleasure[U]-giving[Y 2] way[X2] for them to lead up to the crux. A kitchen[Yļ] pot[Xļ] is in the kitchen, a soup[Y 9 ] pot[X 2 ] is used for making soup, and a cast-iron[Y3] pot[X 3 ] is made of cast iron. their star[Y] of David flag[X] ; the University[Y] of Wisconsin students[X] ; a bachelor[Y] of arts degree[X];

(8)

The trip to Russia was our c o m b i n e d U ] gift[X] to him. (9) λ-functions; the PS-marker; FS-languages; a new Δ-grammar; (10) The Altamira[Y] Caves[X]; the Kalahari Desert; the Sinai Peninsula.

The dependent component is a hyphenated phrase or a phrase in quotation marks 33.2.

Examples (1) (2)

up[Y]-and-down method[X]; a "wait[Y]-and-see" approach[X]; an I-don't[Y]-care-if-something-does֊happen attitude[X].

COMPOSITIVE

413

The dependent component is a sequential or attributive phrase composed of cardinal numerals 33.3.

r — sequential, attributive Examples 100[Y] to[Z] 1 speed advantage[X]; one[Y]-one[Z] correspondence[X]; four[Y] by[Z] four matrix[X].

Compound adjectivais (33.4 - 33.5) The type ''modifying noun + adjectival" 33.4.

α = (A, notpron), (V)ppres, (V)pass

414

COMPOSITIVE Examples

Χ=(Λ) graph[Y]-independent[X] program; life[Y]-long[X] career; a two[Z]inch[Y]-long[X] line; context[Y]-free[X] grammar; stone[Y]-grey[X] sky; (2) ... the symbols are Turing[Y] observable[X].

(1)

X = (V)ppres (3)

question[Y]-answering[X] systems; a new heat[Y]-seeking[X] missile; the policy[Y]-making[X] executive. X = (V)pass

(4) (5) (6) (7) (8)

a shoulder[Y]-fired[X] rocket; a propeller[Y]-driven[X] aircraft. London-based; temperature-controlled; war-born; problem-oriented language; Scottish television part[Y]-owned[X] by the Thompson Organization cannot... This conductor cable assembly is shipped factory-installed. a problem-and-answer oriented technique; a liquid-hydrogen fueled engine.

Compound adjectivais of bahuvrihi type, such as blue-eyed 33.5.

Examples (1) (2) (3)

some clumsy[Y]-fingered[X] clod; the diamond[Y]-shaped[X] nodes; five[Y]-handed[X] game; two-valued logic; one-sided intimacy.

COMPOSITIVE / ELECTIVE

415

"Compound verbs" (33.6) 33.6.

Examples (1) (2) (3) (4) (5) 34.

The pulse does not frequency[Y]-modulate[X] the magnetron. A copying transformation Chomsky[Y]-adjoins[X] the experiencer to the basic sentence. Mr. Brick and I have been star[Y]-gazing[X], I don't know how long [W. Congreve]. Her husband had been holiday[Y]-taking[X] away from home [Th. Hardy]. The IPL grammar was machine[Y]-tested[X] by Morris Salkoff.

Elective syntagm The elective SSRel is non-repeatable. Condition on the cooccurrence of the elective SSRel with other SSRel's

416

ELECTIVE

α = (A)superl, (A)compar, (A, ord), (elect), (Num), (Ν, indef) ► Condition 2, relating to both subrules, indicates that an adjective in the com­ parative (but not in the superlative) degree is possible only if the item modified is one of exactly two items: the younger of the girls means that there are only two girls; but not *the younger of'all these girls. For more on the elective construction, especially on its semantic interpretation, see Lee 1971. ◄ Examples Order (1) (1) (2) (3) (4) (5) (6)

the easiest (second)[X] of[Y] the solutions[Z] available; the most[X] important[U'] and far-reaching[U"] of[Y] the Government's decisions[Z]; the cleverer[X] of[Y] the girls[Z] ; the first three[X] of [Y] R. Schank's premises[Z] ; both[X] of[Y] us[Z] ; all[X] of [Y] them[Z] ; Several[X] of[Y] them[Z] almost immediately found employment. Order (2)

(7) (8)

Of[Y] those[Z] who have made other significant contributions, Palmer[P], more than anyone[X] else, has[Q] emphasized the significance of phonetics. Of[Y] six samples[Z] tested, five[X] were pure.

APPOSITIVE 35.

417

Appositive synlagms

The appositive SSRel is non-repeatable. ► A detailed and insightful analysis of English apposition (with many further references) is found in Burton-Roberts 1975. Cf. the discussion of Russian apposition in Uryson 1981, which offers a complete list of corresponding syntagms and intro­ duces various SSyntRel's to represent apposition as well. ◄ Naming appositives (35.1 - 35.4) Appositive manifested by a proper noun 35.1.

► Condition 3 bars ungrammatical phrases of the type *the city Limassol, *the Republic Poland, * the state South Dakota, '"the island Madagascar, *the desert Gobi, * the peninsula Taymyr, etc. The correct phrases of the type the city of Limassol, the Republic of Poland, etc. are covered by Syntagm 38.1 (p. 432), and those of the type the Gobi Desert, the Tay myr Peninsula, etc. by Syntagm 33.1 (page 409ff.). The following general remark seems to be in order: rules for the formation of compound geographic names, consisting of a geographic feature and a proper name, are much more complicated in English than our condition implies. For example, we get both Lake George and Schroon Lake in the same community, the Mississippi River and the

418

APPOSITIVE

river Kwai, Mount Everest and Montblanc Mountain, etc. We do not have *the Lake George but we do have the Kamouraska River, and so on. Condition 3 does not even scratch the surface of the problem. ◄ Examples (1) (2) (3) (4) (5) (6) (7) (8)

Jurij[X] Apresjan[Y]; Noam[X] Chomsky[Y]; the article by ProfessorfX] Lehrberger[Y] ; Soviet U.N. Delegate[X 1 ]Yakov[Y 1 ,X 2 ]Malik[Y 2 ]; U.S. Secretary[X 1 ] of State Henry[Y 1 ], X2] Kissinger[Y2]; the downfall of CoupLeaderp[X 1 ]Nikos[Y1,X2] Sampson[Y 2 ] ; Mr.[X 1 ] Bill[Y1 X 2 ] Powell[Y 2 ]; Dr.[X] Potts[Y]; Shaw plays the obsessed shark killerfX] Reynolds[Y]. mypet-dog[X]Fido[Y]; the river[X] Thames[Y]; 'The bridge over the river Kwai" [title of P. Boule's novel].

► Examples (8) reflect rather British usage.◄ (9)

High on Mount[X] Kyparissovouno[Y] mortar shells ignited massive forest fires. (10) Lake Erie; Fort Istokpoga; Cape Gomez; Port Apopka.

► Actually, Syntagm 35.1 covers phrases of several different types, and in a sharper description of English syntax it will, in all probability, be split to form two or even three syntagms. (This fact has been drawn to the authors' attention by I. Mackenzie.) First, there are constructions formed by apposing a proper noun (= a name in the current sence of the term) to a common noun or noun phrase: my pet dog Figo, the eminent Scottish physicist Maxwell, etc. The result is a phrase which is not itself a proper name. Second, there are phrases which themselves are proper names compounded out of two proper names or of a common noun plus a proper name. These include several subtypes; for instance: - A human first name + human last name, as in Alexander Bell, with well-known complications (e.g., middle names, etc.). - A geographic feature + geographic name, as Lake Superior (notice that it does not take the definite article, which shows its proper name status). There are also other types, which exhibit certain surface-syntactic differences and would be better described by separate rules. This is, however, a problem for further research. ◄

APPOSITIVE

419

Appositive manifested by the name of a man-made object

Examples (1) (2) (3)

the U.S. carrier[X] Saratoga[Y] the series[X] "Working Papers[Y] in Linguistics"; In a profound paper[X] of 1956, 'Begründung einer strengen Implikation'[Y], W. Ackermann laid the foundation of this theory.

►Unlike the title in (2), the German title in (3) is considered unanalyzable and taken as a single syntactic unit, i.e. a single node in the surface-syntactic structure of the corresponding English sentence. ◄ (4) (5) (6)

...his co-star in the TV movie[X] Brief Encounter[Y]; In his essayfX] The Subjection[Y] of Women, Mill protested that... HMS[X] Bounty[Y]; M.V.[X] Queen[Y] ofSaanich.

Appositive manifested by a common noun

APPOSITIVE

420

Examples (1) (2) (3)

Dennis[X] the[Z] Menace[Y]; Peter[X] the[Z] Great[Y]; Rabin[X] the[Z] logician[Y] recalled that there were precedents for a prisoner exchange. Flipper[X] the[Z] trained seal[Y].

Appositive manifested by a numeral or a formula (as an identifier or a measurement unit) 35.4.

Examples (1) (2) (3) (4) (5) (6) (7) (8)

The FS-language[X] L[Y], considered here, may seem too artificial. the derivation[X] D[Y]; the rules[Y] R[Y] ; We use the notations of Sections[X] 10[Y], 27, 29 and 32. sections[X] 5[Y] through'(to) 10; cf. note[X] 2[Y]; As for theorem[X] 4[Y], it can be found in Chapter[X] VI[Y]. ... mentioned in Section[X] Three[Y] ; Lemma 1; corollary 3; definition 8; a derivation of length[X] 4[Y].

► The standard usage of the appositive numeral precludes a determinative dependent modifying its governing noun: (9) * ...in the section 3, *for this definition 8, etc., cf. examples (3) through (9). However, determiners modifying the noun X may still appear in certain special contexts, cf.: (10) The World War II you imagine is quite different from the World War II I have seen.

APPOSITIVE

421

Yet the cooccurrence of determiners and nouns in this construction is restricted by very complex and specific rules. Therefore we have left the problem of determiners out of the discussion here, as in the case of determiners and proper nouns, see comments on Syntagms 28.1 and 28.2, pages 372 and 374. -4 Equative or specifying appostitives (35.5 - 35.6)

Appositive manifested by a non-alphabetical or an autonymous phrase modifying the classifier name of the phrase in question 35.5.

Examples (1) (2)

I don't know the word[X] "steeple[Y]" the morpheme[X] -ism[Y]: the string[X] -  [Y]; the term[X] cb[Y].

Appositive manifested by a non-qualifying nomen agentis 35.6.

422

DESCRIPTIVE-APPOSITIVE

Examples (1) (4) (3)

my friend[X] the[Z] captain[Y] ; Bill[Y] the[Z] alcoholic[Y] ; the good lady[X] her[Z] mother[Y].

36. Descriptive-appositive syntagms The descriptive-appositive SSRel is non-repeatable.

Appositive serving as an additional characterization (36.1 - 36.3) Appositive modifying a noun phrase 36.1.

►We remind the reader (cf. ρ. 193) that is a function which ensures agreement between the two components of the descriptive-appositive syntagm, primarily in number and, possibly, in their determiners and certain syntactic and

DESCRIPTIVE-APPOSITIVE

423

semantic features. ◄ Examples Χ ≠ (prop), Υ ≠ (prop) (1) (2) (3) (4) (5) (6) (7) (8)

A body[X] of routines (in effect, a language[Y]) is used to express the logical operations. There appears an article[X] in the Sunday paper, a ten-item test[Y] to determine 'How good a hostess are you?' ... the reactions with a biphylic type[X] of reagent, the Grignard reagent[Y], can be said ... The experience[X] itself, the hospitalization^] and surgery, brings its own advantage. Only during 1941-43[X], the most difficult period[Y] of the Second World War, was there a deficit. The other derivation is another kind[X] of reduction of our construction, this time a deletion[Y] of the relative and a following verb. The system[X] — a simulated robot[Y] with a hand and eye and the ability to manipulate toy blocks on a table — deals with all aspects of language. There he[X] sat[Z], a giant[Y] among dwarfs. Χ Φ (prop), Y = (prop)

(9)

In Athens Sisko found the junta's strongman[X], Dimitrios[Y] Ioannides, enraged at the Turkish government's stalling. ( 1 0 ) T h e two dismissed players[X] — Bob[Y] McCarthy and Gary Sevens — escaped suspension. X = (prop), Υ ≠ (prop) (11) Richardson[X] and Ruckelschaus, the so-called paragons[Y] of Republican virtue, ... (12) There is[Z] a How'm[X] I doing? in[W] the Sunday paper, a ten-item test[Y] to determine this game. (13) The London Broadcasting Company Radio[X] is[Z] now fully operational — the first[Y] of 60 stations to be set up throughout the country. Χ≠ (pron), Y = (pron) (14) We operate with the system[X] of ten cases, that[Y] of Ch. Fillmore.

424

DESCRIPTIVE-APPOSITIVE (15) They actually do earn a living[X], usually a very good one[Y]. (16) All human languages[X], even those[Y] used by supposedly primitive cultures, have a complicated system of grammatical devices. (17) One[X = ONE 1 ] of the possible structures must[Z] be chosen, hopefully the correct one[Y = ONE 2 (N, pron)]. (18) The general meeting was addressed by the director[X] — myself[Y] — and by the chairman. X - (pron), Υ Φ (pron) (19) I[X],a (the) girl[Y] who Bill likes, would never... (20) We[X], the committee[Y], shall meet in December... (21) Game analysis is interested in something[X] quite different — the kinds[Y] of social transactions that are related to such excesses.

Appositive modifying a personal pronoun of 1st or 2nd person ► This construction and that of the type We, the men; I, a woman; etc., described in the preceding syntagm, are discussed in Delorme and Dougherty 1972. ◄ 36.2.

not form) ► Note that the third person pronouns HE, SHE, IT and THEY are not admitted in this syntagm: (1)

*he thief, *she whore, *they philosophers.

Compare Syntagm 29.11, on pp. 384-385. ◄ Examples (2) (3) (4)

we (you) boys; we (you) girls who(m) Bill likes; We[X] Canadians[Y] like whiskey. You[X] troops[Y] will go aboard.

DESCRIPTIVE-APPOSITIVE (5) (6) (7)

425

You've got to watch every damn little word with you][X] people[Y]. Go away at once, you[X] thief[Y]. You[X] M.P.'s[Y] can afford things a university professor cannot.

Appositive modifying a full sentence, an absolute or a sentential phrase, etc.

r Φ determinative

(1) (2) (3) (4)

Examples These phonemes show[X] no symmetry, a fact[Y] easily accounted[Z] for historically. Sales in the first six months of 1974 had[X] fallen 25%, a decline[Y] unpre­ cedented[Z] in the past two decades. Finally, Congressman Solarz asked[X] the US government to terminate all aid to the Marcos regime — a harsh[Z] but quite predictable reaction[Y]. Then a silent plume of colorless poisonous gas shot[X] up from the turbulent depths of Lake Nios, just inside Cameroon's northwest border — the worst[Z] natural calamity[Y] ever to strike the quiet African country.

DESCRIPTIVE-APPOSITIVE

426

Appositive serving as an explanation, an illustration, an interpretation, etc. (36.4 - 36.5) Appositive introduced by an explanatory conjunction 36.4.

► L(X) stand for a lexeme which labels node X; L(X) = L(Z) indicates that nodes X and Ζ should be labeled with identical lexemes. Thus, Condition 4 stipulates that in the construction in question non-projectivity is possible only if the adjectives or adverbs connected by the descriptive-appositive SSRel are lexemically identical. ◄ Examples

(2)

The individual starts off in[X] an autonomous s t a t e , c a p a b l e [ Z ] of awareness. It is possible to derive our constructions from the common[X] logical form

(3)

'A  B', common[Z] but for a different stylistic marker. ... theories expressed in traditional[X] (i.e.[Y], mathematical[Z]) terms.

(1)

427

DESCRIPTIVE-APPOSITIVE (4) (5)

... more complex relationships[X] — , games[Z]. Now they know[X] each other better, i.e. ,[Y] each knows[Z] the other is reliable.

► We have broadened considerably, as is evident from the above examples, the traditional concept of apposition. Our explanatory and illustrative appositives include, apart from noun phrases, also adjectival, prepositional and infinitive phrases, as well as full sentences. -4

Appositive not introduced by an explanatory conjunction 36.5.

a

Examples (1) (2) (3) (4) (5) (6) (7) (8)

They sent a squad[X] of eighteen — a coach[Y], a team manager and sixteen players. A major goal is to program computers to[X] do intelligent things — to[Y] converse with human beings, play chess, solve abstract problems, etc. He soon learned what the agency wanted[X] : clinical material[Y]. Let[X] me tell you a secret: I did[Y] in fact discuss it first with a union subcommittee. He does not accept the two-verb hypothesis[X] (namely[Z], the hypothesis[Y] proposed by G. Lee) either. Her married life had proved one thing[X] to her that she had always main­ tained: that[Y] all men were mean and tyrannical. A patient pointed out what the therapist had[X] overlooked: that[Y] in practice, waiting for Santa Claus and waiting for death are synonymous. The parents should find out two things[X] : which of them taught[Y] the child this game, and what they are doing to perpetuate it.

DESCRIPTIVE-APPOSITIVE

428

Appositive manifested by an inventory (36.6)

α = (Ν, not pron, not prop)pl, (A), (Num); β = (Ν), (A), ONE 1 Examples X=(N) (1) (2) (3) (4) (5) (6) (7)

He is pouring funds into two areas[X] where Iran is weakest, agriculture[Y] and[Z] education. There are 15 parts[X] of speech assigned: noun[Y], article[Z], adjective, verb, ... There are many emotions[X] reflected here — bitterness[Y], anger[Z] and occasional humor. Reality is defined as having two aspects[X]: static[Y] and[Z] dynamic. Here we have word groups of three types[X]. the prepositional[U], infinitive[Z] and[Q] participial phrases[Y]. Computer systems fall into two classes[X] — the very fast[Y] but expensive and[Z] the cheap but slow. There were two car-bomb explosions[X] — one[Y] in Great Scotland Yard and[Z] the other outside the Old Bailey.

X=(A) (8)

Games can be classified according to any[X] of[W] the other factors discus­ sed above: the aims[Y], the roles[Z], the most obvious advantages.

SEQUENTIAL

429

Χ = (Num) (9)

37.

All the players had folded except two[X], a research psychologist[Y] and[Z] a businessman.

Sequential syntagms

The sequential SSRel is non-repeatable and does not co-occur with any other SSRel, unless X is a compound proper noun: , then (r = determ, quant, modif, compos, appos, and Z, together with X, forms a part of a compound proper noun) [of the type the United Kingdom, Salt Lake City, etc.]. Nominal sequential constructions (37.1 - 37.2)

Without preposition 37.1.

r = compositive, sequential Examples (1) (2) (3) (4)

Chomsky[X]-Halle[Y] proposal[Z]; pressure[X]-temperature[Y] diagram[Z]; man-machine interaction; the Adult-Adult conversation; the Okamoto[X 1 , Z2]-Bey Khuy Day[Y 1 , X 2 ]-Andreotti[Y 2 ] t a l k s ^ ] ; The trip to Russia was our combined graduation[X]-birthday[Y] gift[Z] to him.

430

SEQUENTIAL

With preposition TO 37.2.

r = compositive, 1st completive Examples (1) (2) (3)

ground[X]-to[Y]-air[Z] missile[W] ; a computer-validated PortuguesefX] to[Y] EnglishfZ] transformational grammar; Those missiles are[W] ground-to-air.

Adjectival sequential construction (37.3) 37.3.

Examples the structural[X]-transactional[Y] classification; the red[X]-black[Y]-yellow[Z] banner.

SEQUENTIAL

431

Quantitative sequential constructions (37.4 - 37.6)

Without preposition 37.4.

Examples (1) (2)

Five[X], six[Y] minutes later the bomb exploded. 10-15 dollars; Sections[Z] 10-20.

With the preposition TO (in a quantitative phrase) 37.5.

Examples (1) (2)

five[X] to[Y] eight[Z] tracks[W] ; It will probably be another two[X] to[Y] three[Z] months[W] before the

432

SEQUENTIAL / ATTRIBUTIVE

(3)

ports are fully operational. ... the mosque seized last week by 200[X] to[Y] 500[Z] heavily armed fanatics[W]... With preposition TO or THROUGH (in an appositive phrase)

37.6.

Example ... levels[W]One[X] through[Y] Four[Z]. 38.

Attributive syntagms The attributive SSRel is repeatable. Attribute manifested by a prepositional phrase (38.1 - 38.4)

Attribute whose governor is a noun 38.1.

ATTRIBUTIVE

433

►Among the constructions described by this syntagm, Conditions 1 and 2 single out constructions of the type the town of Sommerville, the Federal Republic of Germany (example (5)), etc., and of the type that scoundrel of an uncle or a wretch of a ship (example (6)). We have chosen to describe these constructions by the attribu­ tive SSRel — contrary to the current practice of considering them appositive. We feel that despite their 'appositive' semantics they seem, in a purely syntactic sense, to be merely a specific case of the attributive construction. Cf. Syntagm 35.1, Condition 3. Note that the definite article is a prerequisite in constructions of the type the town of Sommerville, in which the proper noun designates a village, a town or a city: *a filthy town of Jampore is impossible, ?this town of Jampore sounds strange, etc. Naturally enough, this restriction does not hold when a geographic name Ζ is not the name of a town or a city, cf. a town of Scotland. Condition 3 accounts for constructions of the type this foolish wife of mine. Conditions 4 and 5 cover constructions of the type nothing [to do] besides sleep or What [did he want] except to see her1. Cf. Syntagm 27.17. ◄ Examples X = (N, not pron) (1) (2) (3)

the symbol[X] to[Y] the right of the arrow; a child[X] of[Y] the same age; the other verbs[X] of [Y] English; The words[X] in[Y] the Bible are simply statements. Beacon tracks are established to ranges[X] many times beyond[Y] the skin tracking capability.

434

ATTRIBUTIVE

(4) (5) (6)

The Earth is millions of kilometers[X] from[Y] the Sun. the[D] Republic[X] of [Y] Cuba[Z] ; the[D] State[X] of [Y] Ohio[Z] ; a. Try this[D] honey[X] of [Y] a[F] melon[Z]. b. What a[D] whopper[X] of [Y] a[F] lie[Z] ! . He's had a[D] nightmare[X] of [Y] a[F] day[Z]. d. He was a runt of a man and frail. (7) that[D] crazy girl[X] of [Y] yours[Z] ; (8) a boy [X]of[Y]eleven; (9) For the work[X] thereafter[Y] they employed two assistants. (10) Nixon's actions sounded like a Brownshirt operation[X] thirty years ago[Y]. X = (TV, pron, indef) (11) anything[X] interesting of [Y] this kind; (12) If nothing[X] but[Y] chaining paradigms prevailed, ...

(13) Those[X] at[Y] prayer were unmolested. (14) Superficial differences between the examples in (I) and those[X] in[Y] (III) can be ignored.

► Certainly, there must exist some restrictions on Y, i.e., there are prepositions in English that cannot introduce an attributive prepositional phrase — or. if they can, then only under highly specific contextual conditions: (15) *his death because of the wound, etc. [Emonds 1976: 175-176]. Yet since we do not know exactly what these restrictions are and are not in a position to undertake the necessary research, we must be satisfied in Syntagm 38.1 with an approximation. ◄

ATTRIBUTIVE

435

Attribute whose governor is an adjective or an adverb 38.2.

► An attributive prepositional phrase modifying an adjective may sometimes be placed to the left of this adjective in English: e.g., an in[Y] those days enormous [X] expenditure. However, our incomplete knowledge of the fairly intricate conditions of such preposing makes it impossible for us to list cases of this usage in Syntagm 38.2. ◄ Examples X=(A) (1)

The speaker intends this phrase to be interpreted as instrumental[x] in[Y] function. X = (Adv)

(2) (3) (4) (5)

early[X] in[Y] October; early in life; late on Sunday; soon[X] after[Y] breakfast; The distinction of modality is indicated elsewhere[X] in[Y] the sentence as well. Outside[X] under[Y] the window were some carts. The individual steps are buried somewhere[X] in[Y] the inner workings of our minds.

436

ATTRIBUTIVE

Attribute whose governor is a preposed adjective in the superlative degree, an ordinal numeral or one of the adverbs MOST1 or LEAST 1 38.3.

α = (A)superl, (A, ord), MOST1 , LEAST 1 ; Լ = IN, ON, AMONG, FROM Examples (1) (2) (3) (4)

the b e s t r ø friend[Z] in[Y] the world; the most[X] expensive[Z] car in[Y] England (on[Y] the island) ; They chose the smartest[X] studentfZ] in[Y] the class. ... the third[X] seat[Z] in[Y] the row (from[Y] the aisle).

Attribute whose governor is a cardinal numeral 38.4.

Լ = BY, TO1,...

ATTRIBUTIVE

437

Examples five[X] by [Y] five[Z] matrix; one[X] to[Y] many[Z] correspondence Attribute manifested by a prepositionless nominal phrase (38.5-38.11)

Attribute manifested by a noun phrase denoting a parameter 38.5.

r = determinative, possessive Examples (1) (2) (3) (4)

a[W] woman[X] my[Z] age[Y]; a[W] woman[X] my mother's[Z] age[Y]; a tree that height; women my age; trees that height; An[W] animal[X] that[Z] size[Y] is too big. He had a[W] blotched complexion[X] the colour[Y] of porridge. Montreal[X] the[Z] size[Y] of Los Angeles? Don't even think of such a nightmare!

438

ATTRIBUTIVE Attribute manifested by a temporal noun phrase

38.6.

Examples X=(N) (1) (2) (3)

his brother's classmate[X] last[ZJ year[Y] ; ... in connection with the murder[X] last[Z] week[Y] of Ch. Downey; London[X] the year[Y] before[Z Their efforts mirror the desperate bid[X] by the world community last[Z] week[Y] to restore the cease-fire in Cyprus. His mouth was dry with too many cigarettes[X] the night[Y] beforefZ]. X = (Adv)

(4)

late[X] that[Z] evening[Y]; early this year.

Attribute manifested by a noun phrase denoting distance 38.7.

ATTRIBUTIVE

439

Examples (1) (2) (3)

a small town[X] ten miles[Y] from[Z] here (down/up[Z] the river) ; He has a farm[X] half a dozen miles[Y] away[Z] (north/east[Z] of Adelphi). He ate his meals at a restaurant[X] only two blocks[Y] away[Z]. Distributive attribute manifested by a measurement noun phrase (38.8-38.9)

The governor is a noun, head of a quantitative phraset 38.8.

L = A/AN, EACH, EVERY Examples (1) (2) (3) (4) (5) (6)

... a[W] dollar[X] a[Z] day[Y] ; ... at over five hundred[W] miles[X] an[Z] hour[Y] ; fifty fiveIWl] strokesfXl] a[Zı] minuteIYı] six[W2] times[X2] every[Z2] night[Y2]; five[W] shillings[X] an[Z] inch[Y] (every[Z] two copies[Y]) ; He answered at least one[W] advertisement[X] a[Z] day[Y]. They should accept the government's income policy offer of 2.48[W] pounds[X]a[Z]week[Y].

է Cf. Syntagm 38.22, p. 450.

ATTRIBUTIVE

440

The governor is an adverb 38.9.

L1 = ONCE, TWICE (, THRICE); լշ = A/AN, EACH, EVERY Examples twice[X] a[Z] year[Y]; once[X] or[W] twice a[Z] year[Y].

Attribute manifested by a reflexive pronoun 38.10. oX

( N , not refl, not indef)

ļ

I attributive

<=^֊

x

n o t p o s s + ··· +

1) A G R E E R E F L ( N ) ( Y , X ) ; 2) ifX = (pron), then ... = A

Y

I oY

( N , pron, refl)

I Examples

(1) (2)

I myself; he himself; they themselves; The study[X] itself[Y] began under the supervision of Professor J. Dypstra.

(3) (4)

The expanded forms describe the actions[X] themselves[Y]. In the analysis of the game, drinking[X] itself[Y] is merely an incidental pleasure. A relative clause[X] itself[Y] can modify an already modified noun.

(5)

ATTRIBUTIVE (6)

441

Cases[X] themselves[Y] can be conceived of as syntactic roles.

►We consider a reflexive pronoun which immediately follows a noun to be an attribute rather than an appositive linked to the noun by the descriptive-appositive SSRel. Our judgement is based on the fact that the descriptive-appositive SSRel is non-repeatable, while reflexive pronouns cooccur freely with descriptive appositives: (7)

Mr. Quang Phuc Dong himself, a sharp observer, ... ◄

Attribute manifested by a single noun which denotes a part of the day 38.11.

α = (Ν, temp)sg, (Adv, temp); L1 = MONDAY, ..., SUNDAY, YESTERDAY, TOMORROW; L 2 = MORNING, AFTERNOON, EVENING, NIGHT ► X is taken to be the head since it is not omissible without destroying grammaticality, while Y is: (1)

a. I will come Monday (morning). b. *I will come morning. ◄ Examples

(2) (3)

Monday morning; tomorrow night; Sunday afternoon; Saturday[X] night[Y], at a quarter to eleven, I wheeled up to the door of "Sea Ledges", was received, and led into Mrs. Granberry's room.

ATTRIBUTIVE

442

Attribute manifested by an absolute construction (38.12) 38.12.

r = absolute-predicative, absolute-conjunctional Examples Y = ANP [r = absolute-predicative] (1)

The council includes three elected delegates[X], each[Y] from[Z] a separate constituency: the student body, the teaching body and the graduates. Three Greek statuettes[X], each[Y] worthfZ] over $ 10,000, were sold in London last week.

(2)

Y = (Conj, subord, absol) [r = absolute-conjunctional] (3) (4)

a. b.

(5)

a. b.

The PHRASE program uses the parts[X] of speech of the words as[Y] assigned[Z] by the part-of-speech program. Polyatomic molecules[X] if[Y] present[Z] in appreciable amounts affect the character of the discharge. The discharge affects polyatomic molecules[X] if[Y] presentfZ] in appreciable amounts. Iron filings[X], when[Y] placed[Z] in the vicinity of a magnet, appear to be drawn to it. This is characteristic of iron filings[X] when[Y] placed[Z] in the vicinity of a magnet.

► We are not quite sure of our analysis of examples (4) and (5). A different

ATTRIBUTIVE

443

solution could be proposed: namely, to consider the asolute-conjunctional phrase as a modificative adverbial depending on the main verb rather than as an adnominal attribute. One of the arguments is the fact that in (4a)-(5a) the absolute-conjunctional phrase can be readily preposed to the rest of the sentence with no perceptible semantic change: (4a') If[Y] present[Z] in appreciable amounts, polyatomic molecules affect the character of the discharge. (5a') When[Y] placed[Z] in the vicinity of a magnet, iron filings appear to be drawn to it. But, on the other hand, the absolute-conjunctional phrase in (4b)-(5b), i.e. in the sentences where its presumed head is not a grammatical subject, cannot be so preposed. Because of that, we prefer to leave sentences (4) and (5) in the realm of Syntagm 38.12. ◄ Attribute manifested by an adverb (38.13 - 38.15)

Adnominal adverb 38.13.

Examples (1) (2) (3)

for all the cases[X] immediately[Z] above[Y]; two lines[X] directly[Z] below[Y]; the night[X] before[Y]; The management is preparing a plan for the period[X] still[Z] ahead[Y]. We omit from consideration[X] here[Y] sentences of the following form.

444

ATTRIBUTIVE Adverbial modifier

38.14.

α = (Ν, not pron, not prop), THAT 3 , ONE 2 Examples (1) (2) (3)

From the assessment[X] of the process so far[Y], it can be seen... his arrival[X] a week-*֊ later[Y]; the activity[X] of more than two ego states simultaneously[Y].

Adverb ELSE 38.15. ^(pron, indef/wh) attributive

<=^>

Xnot

poss

+ Y

y  Y

ELSE Examples

(1)

anything else; nothing else; wherever else; who else; what else.

► The phrase or else is an idiom treated as a postpositive-only adverb appearing in the role of an adverbial modifier: (2)

Ratify SALT or else: Russia's Gromyko said last week. M

ATTRIBUTIVE

445

Attribute manifested by a conjunctional phrase (38.16 - 38.17)

With conjunction AS 3 and modifying a noun 38.16.

Examples (1) (2) (3) (4) (5) (6)

The reason applies to them[X] as[Y] a class[Z]. The included clause[X] as[Y] modifier[Z] always follows its head. ... with manganese oxide[X] as[Y] a major constituent[Z]. .. .for him[X] as[Y] the best swimmer[Z]. The order of words[X] as[Y] they stand[Z] in a sentence has become an important device. He may become more acutely aware of the things going on in the language[X] as[Y] he uses[Z] it and others use it and may derive from this understanding satisfactions of many kinds.

446

ATTRIBUTIVE With conjunction AS 4 and modifying an adjective

38.17.

Examples white as snow; red as blood; easy as pie; slow as molasses in January. Attribute manifested by an infinitive phrase (38.18 - 38.20) Infinitive phrase modifying a noun (38.18 - 38.19) Infinitive phrase postposed to the modified nount 38.18.

t See Berman 1974.

ATTRIBUTIVE

447

α = (Ν, not pron, not prop, not abbr), (N, pron, indef), ONE 2 , THAT 3 ► Condition 1 indicates that in this construction the noun valence slots of both the infinitive Ζ and the preposition U, which Ζ governs, may remain unsaturated (but need not: cf. examples (8) - (10) below). Such non-saturation is obligatory if the to find modified noun itself has an infinitive valence slot, cf. the decision such possibilities vs. a decisionto find: Condition 2. Notice that in fact Condition 2 must be restated in more general terms concerning the deep-syntactic structure: the deep object W of the infinitive Ζ or the deep dependent Q of the preposition U cannot be expressed in the surface-syntactic structure if they are coreferential with X, cf. examples (1) through (6); in this case, non-saturation of strong valence slots of Ζ or U is obligatory. However, the corres­ ponding rule, which deletes W or Q coreferential with X in the SSyntS, belongs rather to deep syntax, and therefore should not appear in this syntagm. (Condition 2 simply stakes out the problem, lest it be forgotten.) Condition 3 covers constructions of the type her husband-to-be, immigrants-to-be, etc. ◄ Examples (1) (2) (3) (4) (5) (6) (7)

(8)

... to continue the game in the absence of people[X] to[Y] rescue[Z] ; a set of data[X] to[Y] be[Z] accounted for[U] ; Yucatan is a wonderful place[X] to[Y] die[Z] in[U] [J. Lindsay]. There is[T] nobody[X] here[R] for him to[Y] play[Z] with[U]. She had nothing[X] to[Y] lose[Z]. ... the tin mug[X] with the hot water for Alexis to[Y] drink[Z] from[U]. Theproblemsp[X 1 ]to[Y 2 ] b e [ Z ] solved, the questions[X 2 ] to[Y 2 ] be[Z 2 ] asked, and the solutions[X] t o [ Y ] be[Z 3 ] sought are determined by what Th. Kuhn has called a 'paradigm' There is[T] nothing[X] in[R] this proposal to[Y] entail[Z] that all negative implicatives must have a positive counterpart.

448

ATTRIBUTIVE

(9)

A plantp[X] to [Y] produce[Z] X-ray apparatuses is under construction in Tartu. (10) The only class[X] in Britain to[Y] never lose[Z] an election is the ruling class.

Infinitive phrase preposed to the modified noun 38.19.

Examples (1) (2) (3)

a[U] never[W] to[Y] be[Z] forgotten[R] incident[X] ; An as yet[W] to[Y] be[Z] created[R] morphological[Q] analyzer[X] will make use of this node. In the late 1960's the[U] as yet[W] to[Y] be[Z] exposed[R] CIA[X] was working on a plan to bomb Castro's cigar.

► To the authors' knowledge, the construction in question is the only instance of a preposed non-adjectival attribute in English. Notice that this infinitival attribute is possible only with a passive infinitive: (4) *a never to be realistic dream. ◄

ATTRIBUTIVE Infinitive phrase modifying an adjective (38.20) 38.20.

α = (A, ord), (A)superl, ONLY 2 , LAST, NEXT Examples (1) (2) (3)

It was the only[X] thing[Z] to[Y] do. The secretary general was the first[X] to[Y] raise the question. Ian happened to be the youngest[X] to[Y] accept the challenge.

Attribute manifested by a temporal abbreviation (38.21) 38.21 X(Num) attributive <=¿> X + Y

Լ = A.M., P.M., B.C., A.D., O'CLOCK Examples (1) (2)

at two A.M.; till five p.m.; in 65 B.C.; at 3 p.m.; It is five o'clock.

449

ATTRIBUTIVE

450

Distributive attribute (of a quantitative phrase) manifested by the pronominal adjective EACH, the adverb APIECE or the preposition PER (38.22) 38.22.

Լ = EACH, APIECE, PER Examples (1) (2) (3) (4) (5) (6) (7)

A thousand programs of 1000 cards[X] are compiled per day. He bought the finest cigars, for (which cost) $5[X] apiece[Y]. These four columns, of twenty trucks[X] each[Y], left El-Quneitra at dawn. ... ten armoured divisions with one to seven reinforced battalions[X] each[Y]. Fifteen boxes, weighing 500 pounds[X] each[Y], were already aboard the plane. About 40 brochures, with 50 pages[X] and[Z] three color illustrations each[Y], are printed weekly. He sold caviar at $300[X] per[Y] pound.

► Contrast the different usages of each in (8) and (9): (8)

a. each—t 7

p

,

program of 1000 cards; descr-attr

, abs-predic

 

֊„„

,

b, for these programs, »-each -—»-of 1000 cards; c. 200 programs of 1000 cards »-each. (9) a. The two officers, ►each —»-accompanied by an interpreter, talked for 13 minutes. b. The two officers, accompanied by one interpreter ► each, talked for 13 minutes. In (9a), an alternative analysis is possible: each can be taken to be subordinated to accompanied by the pronominal-subjective-copredicative SSyntRel, and then the phrase headed by accompanied becomes a descriptive modifier; cf. (9a'):

ATTRIBUTIVE / DESCRIPTIVE-ATTRIBUTIVE

(9a') The two for 13 minutes.

451

by an interpreter, talked

There is, however, no semantic difference between sentences (9a) and (9b) nor between different surface-syntactic interpretations ((9a) vs. (9a')) of sentence (9a). Notice that each used as a distributive attribute of a quantity phrase with a noun head in the singular, selects ONE with this head rather than A/AN; see (9b) above. ◄ 39.

Descriptive-attributive syntagms

The descriptive-attributive SSRel is non-repeatable. ► This SSRel follows the pattern of the descriptive-modifying and descriptiveappositive SSRel's. Note that a contrast in meaning is possible between the attributive and descriptive-attributive SSRel's: ...a child of the same age as hers ... vs. a child, of the same age as hers, ...◄ General condition on descriptive-attributive Syntagms 39.1 - 39.9:

Descriptive attribute manifested by a prepositional phrase (39.1 - 39.2)

The governor is a noun 39.1.

► The condition provides for a particular type of descriptive-attributive phrase including a proper noun (name of a geographic entity like a country, a city, etc. or of an organization) and headed by the preposition OF: without a preposed restrictive dependent such a phrase can modify only a proper noun or a noun having another proper noun as its appositive. Cf.:

452 (1)

DESCRIPTIVE-ATTRIBUTIVE a. George Fairbanks[X], of[Y] Toronto[Z], writes us about his pet. b. *The first reader[X], of[Y] Toronto[Z], writes us about his pet. from[Y] Toronto[Z], writes us about his pet. of[Y] a Northern youth organization[Z], writes us about his pet.

More examples are found below. ◄ Examples (2) (3) (4)

Professor McCawley[X], from[Y] the University of Chicago, insists that.. Mr. Pucchi[X], of[Y] Alitalia[Z], averred he had learned to fly like a bird. Many linguists[X], especially of[Y] the transformational school[Z], believe that... (5) The ancient universities[X], particularly[U] of[Y] England, are very popu­ lar. (6) The sky[X], of[Y] a clear bright blue[Z], with white promising clouds, began to have the evening look [J. Galsworthy]. (7) Three FORTRAN packages[X] (of[Y] 80 cards[Z] each) were prepared in December. (8) But she[X], already in[Y] the water, was signaling us to join her. (9) The first visitor[X], from (* of)[Y] Harvard[Z], stepped to the podium. (10) With Luigi Elnizzi[X], of[Y] the Mafia[Z], having close connections with civic administration, you never know where you stand. (11) Let me present Professor Vitaly Jeaulpas[X], of[Y] Harvard[Z], who will speak about nostratics. (12) The schooner[X] Lady Anne[W], of [Y] Nantucket Shipping Co[Z], was not exactly the most seaworthy vessel afloat.

DESCRIPTIVE-ATTRIBUTIVE

453

The governor is an adverb 39.2.

Examples (1) (2)

Five months later[X], in[Y] April 1970, the PIB let the industry have another 33 ρ a ton. Three miles further[X], just[Z] outside[Y] the village, we met John return­ ing. Descriptive attribute manifested by a prepositionless nominal phrase (39.3 - 39.6)

Descriptive attribute manifested by a noun phrase denoting a parameter 39.3.

454

DESCRIPTIVE-ATTRIBUTIVE Examples (1) (2)

The shoes[X], size[Y] 12[Z], were still in their box. His little brother Mike[X], age[Y] three[Z], then said ...

Descriptive attribute manifested by a quantity phrase containing an attribute 39.4.

Example Three FORTRAN packages[X] (80 cards[Y] each pared in December.

[Z]) were pre­

► Cf. Syntagm 38.22, page 450. ◄

Descriptive attribute manifested by a noun phrase denoting distance 39.5.

DESCRIPTIVE-ATTRIBUTIVE

455

Examples (1)

Santa Ana[X], over[Y] twenty miles from[Z] Zurbagan, is quite a nice little town.

(2)

The station[X], three km[Y] farther ( away ) [Ζ], was bombed four times last week. Descriptive attribute manifested by a noun phrase denoting localization

39.6.

Examples (1) (2)

Prof .[X1 ] F. Ben-Dreizin, Bar-Dan University[Y 1 ,X 2 ], Israel[Y 2 ]; Available from the National Information Service[X1], Springfield[Y1 X 2 ], Virginia[Y 2 , X 3 ], U.S.A.[Y 3 ].

Descriptive attribute manifested by an absolute phrase (39.7)

39.7.

r = absolute-predicative, absolute-conjunctional

456

DESCRIPTIVE-ATTRIBUTIVE Examples Y = ΛΝΡ [r = absolute-predicative]

(1) (2) (3)

She wore a triangular shawl[X], its corners[Y] dragging[Z] on the stubble. Mrs. Folliot[X], a handkerchief[Y] clutched[Z] in her small hand, was staring up at the ceiling. The condition relationship should be taught as one of the fundamental logical relationships[X], each[Y] realized[Z] by various natural-language constructions. X = (Conj, subord, absol) [r = absolute-conjunctional]

(4) (5) (6) (7) (8)

PT3 and PT4[X], though[Y] not part[Z] of the verb, can take an object like a verb. Most of the social difficulties[X], whether[Y] in[Z] marriage, love, friend­ ship, or work, are caused just by this transaction. Some forms of intimacy[X] (especially if[Y] intense[Z]) are psychologically impossible for most people. Some[X] of the terms in the request, though[Y] similar[Z] in meaning to those in the document, are distinct from them. The keen rivalry[X], if[Y] allowed[Z] to go unchecked, could lead to bitterness.

Descriptive attribute manifested by an adverb (39.8) 39.8.

Examples (1)

Denny[X], now[Z] abroad[Y], knew nothing of the new degree [A. Kronin].

DESCRIPTIVE-ATTRIBUTIVE (2) (3) (4)

457

John[X], still[Z] there[Y], was unable to move. The river[X], further[Y] east[Z], still remained their cherished goal. We were approaching the city[X], directly[Z1] ahead[Y] of[Z2] us.

Descriptive attribute manifested by an AS3-phrase (39.9) 39.9.

Examples (1) (2)

This game[X], as[Y] an exploitative manoeuvre, is the basis for a large proportion of public relations in America. Mr. Arposo[X], as[Y] the best player in the League, had no difficulty winning.

Descriptive attribute manifested by an infinitive phrase (39.10)

► The constructions described in this syntagm are slightly preferable when the descriptive attribute is a passive infinitive, but an active infinitive is also possible. 4 Examples (1)

The music[X] of Bach, to[Y] achieve widespread popular recognition in our time, was known to only a handful of music scholars in the 18th century.

DESCRIPTIVE-ATTRIBUTIVE / COORDINATIVE

458 (2) (3) (4)

Claude Charron[X], to[Y] later attain notoriety as the result of a shoplifting episode, was at that time a rising light in the provincial cabinet. The program moves[T] this file[X] immediately [R], to[Y] be processed at a later stage. These data[X], to[Y] be dealt with at the next stage of processing, are placed in array 5.

II. SYNTAGMS COVERING COORDINATION 40.

Coordinative syntagms

The coordinative SSRel is restrictedly repeatable: no more than two arrows of the coordinative SSRel can stem from one node. ► Two coordinative arrows may stem from one node only when the node in question is the top node (= the leftmost component) of a conjoined phrase, which, in its turn, is embedded within another conjoined phrase, also as the leftmost compo-

General conditions on coordinative syntagms 1) X≠ (Art), (partic), (coord), (quasi-coord). An article, a particle, a coordinate or a quasi-coordinate conjunction cannot be governor in a coordinative construction. (Note that a subordinate conjunction can: see example (8) in Syntagm 40.1.) 2) COORD(X, Y) - for Syntagms 40.1 - 40.2; COORD(X, Z) - for Syntagms 40.3 - 40.8. The function COORD, see p. 194 above, accounts for the specific properties of the conjoined items and the contextual conditions which permit their coordination. Thus in the phrase (2)

... a robot with a hand and eye and the ability to manipulate toy blocks

the function COORD has the task of permitting coordination in (with a) hand and eye and (with a) hand ... and the ability, while ruling out the incorrect coordinations *eye and (the ability, *(a) robot... and eye, *a robot... and the ability. The analysis must be based on the distribution of articles and the ''agreement' 1 of both the syntactic features and morphological markings. Cf. Pospelova 1969. ◄

COORDINATIVE

459

Conjunctionless coordinative constructions (40.1 - 40.2)

With the dependent component not modified by a restrictive element

► Condition 1 deals with the following four instances of conjunctionless coordina­ tion characteristic of written English: (i) The coordinated pair in question forms part of a series of coordinated compo­ nents: (1)

...

, and adverbs,

(ii) Two independent sentences are conjoined: (2)

John

(iii) Two adjectives modifying the same noun or a modifying adjective and a preposed modifying noun are conjoined:

COORDINATIVE

460

(3)

a person with a

(4)

...from

face; dialects.

(iv) The first coordinate component forms a descriptive phrase: (5)

Her face,

, brightened.

Conditions 2 through 4 provide for the permitted non-saturation of strong active valence slots in non-final conjuncts. (The fact is that a conjoined phrase may be saturated, so to speak, as a whole.) For more detailed comments, see Condition 5 - 7 in Syntagm 40.3, pp. 462-464. Condition 5 describes coordination of the initial components of compound words of the type FS-[X],CF-[Y] and PS-grammars; three-, four- or five-handed game. ◄ Examples (6) (7) (8) (9)

his skill in procedures[X], rituals[Y] and[W] pastimes; Parents teach their children how to behave[X, Z 2 ], think[Y 1 , X 2 ], feel[W l Y2] and [W2] perceive. before[X], unless[Y] and[W] until they are corrupted, most infants seem to be loving. The management should keep the union informed of production[X], commercial[Y] and[W] financial matters.

► In (9), the coordinative relation between production and matters is ruled out by Condition 1, whereas this relation will be established between production and commercial because the function COORD stipulates the possibility of the coordina­ tive SSRel between a noun which is the dependent member of the compositive SSRel and an adjective, cf. also phrase (10)

(11) (12) a. b. (13)

transformational [grammars]. ◄

The CPUSA calls for[X1 Z 2 ] Nixon's impeachment, for[Y1 X 2 ] "a new emergency presidential election", for the people to[Y 2 , W1 ]establish their own watch-dog committees in Washington. She begins[X] to breathe hard, mother is[Y1, X 2 ] irritated, asthma sets[Y 2 , X3] in, mother apologizes[Y3]. before[X], during[Y] and after the meeting; on[X] the desk, under[Y] the sofa and in the cupboard; his knowledge[X] of[P], talent[Y] for and delight in exotic pastimes;

COORDINATIVE (14) (15) (16)

461

Louise wants[X] to[M], expects[Y] to, and therefore probably shall succeed in this enterprise. John left[X]; Mary stayed[Y]. ... the following four categories: verbs[Xı]; nouns[Yı; X 2 ]; adjectives[Y2]; and adverbs.

With the dependent component modified by a restrictive element 40.2.

Լ = NOT, NEVER. EVER, MUCH LESS, SELDOM, RARELY, SCARCELY,... Examples (1) (2) (3) (4)

... a way of describing the structure[X] of a language, not[Z] the human mechanismfY] for speaking. The firemen would weaken[X], not[Z] help[Y], their fight by breaking away from the union. In a grammatical description a linguist makes his classification on a for֊ mal[X], not (seldom)[Z] a semantic[Y], basis. I haven't read Chapter[X] 1, much less[Z] all[Y] of Part I.

462

COORDINATIVE Coordinative constructions with conjunctions (40.3 - 40.8) With noncorrelative coordinate conjunctions (40.3 - 40.6) No groupings in the conjoined string

40.3.

COORDINATIVE

463

► Condition 1 excludes from consideration the construction of the type John and Mary's walk ('John and Mary are walking together'). In this construction, in contrast to Johns and Mary's walk, it is necessary to mark the group John and Mary, see below, p. 466; therefore, it is described in Syntagm 40.5. Condition 2 rules out constructions of the type precedence-and-dependency grammar, in which the precedence-and-dependency phrase should also be marked as a single group, see Syntagm 40.6. Condition 3 permits a comma before a coordinate conjunction but not inside a prepositional phrase and not before AND conjoining two nouns, adjectives or adverbs, while Condition 4 permits commas in cases where three or more conjoined items are connected by conjunctions. Cf. the examples: (1)

(2)

a.This state may be described phenomenologically as[X] a coherent system of feelings, and[Y] operationally as[Z] a set of coherent behavior patterns. b. *He toyed publicly with the idea of lower interest rates[X], and[Y] a lower exchange value[Y] for the Canadian dollar. ... subject, and[U] object[X], and[Y] predicative[Z].

Note that a comma is not obligatory in cases like (2), cf. (3) and (4): (3) (4)

She was wearing a windbreaker[X]and[ Y] a pair[Z 1 , X 2 ] of blue jeans and[Y 2 ] tennis shoes[Z 2 ] [T. Capote]. But he wasn't dotty, just sweet[X] and[Y 1 ] vague[Z 1 X 2 ] and[Y 2 ] terribly slow[X2] [T. Capote].

Condition 5 sanctions the non-saturation of the non-final preposition among several that are conjoined, cf.: (5)

Essays in[X] and[Y] on[Z] Machine Translation.

Condition 6 allows similar non-saturation in constructions in which surface-syntac­ tic governors of prepositions (or comparative conjunctions) are conjoined; it also admits commas if X is a verb. Cf. : (6)

His knowledge[X] pf[P] and[Y] skill[Z] in[R] procedures ...

(7)

When you take[X] a book from[P], or[Y] return[Z] it to[R], a library you should... The time of the main clause must be the same[X] as[P] or[Y] later[Z] than[R] that of the if-clause.

vs.

(8)

COORDINATIVE

464

Conditions 7 and 8 provide for possible non-saturation and ensure the use of commas in conjoined phrases governing an infinitive: (9)

(10) (11)

a. b.

They can[X], and[Y] it[H] is[Z] expected that they will[F], apply[G]. They should have[Z], and[Y] it[H] is[Z] expected that they will have[F], applied[G]. They should[X], but[Y] are[Z] not necessarily expected to[F] attend[G] classes. They will[X], but[Y] have[Z] not yet begun td[F], choose[G] their own officers.

[Examples (9) through (11) are borrowed from Sager 1968: 7.] (12)

She always wants[X] to[M], and[Y] generally does[Z = F], avoid[G] occasions for quarrels.

We would like to make it clear that Conditions 3,4, 6 and 8 have been formulated only to a first degree of approximation. Whether a comma must, can or cannot be used in these and similar constructions depends, in fact, upon the specific coordinate conjunctions, the inner structure of the coordinated phrases and a number of other factors which we have not considered. In addition, in certain contexts dashes are used instead of commas, but this was not taken into account at all. The above conditions have been presented here in illustration of, and in order to call attention to, the extremely complicated, challenging but largely unexplored laws of coordination in English. Condition 9 describes coordination of the initial components in compound words, cf. Condition 5, Syntagm 40.1. ◄ Examples (13) 'The characters[X 1 ] of the English verb[X 2 ] and[Y 9 ] the expanded form[Z 2 ], and[Y x ] equivalent[X 3 ] or[Y 3 ] analogous[Z 3 ] constructions[Z] in English[X4] and[Y 4 ] cognate languages[Z 4 ]" [the title of a book by H.Poutsma, Groningen, 1921]. (14) She suggests[X] six such references for the present tense, and[Y] gives[Z] examples of each. (15) ... in the late forties[X] and[Y] the early fifties[Z] ; (16) Not[Q] the content[X] is important but[Y] the method[Z]. (17) These words have their own specialized[X] and[Y] in some cases quite complexfZ] meanings. (18) ... as an examination of Figures 16[X1] and[Y 1 ] 19[Z1 ]on pages 186[X2] and[Y 2 ]211[Z 2 ] will show.

COORDINATIVE (19) (20) (21) (22) (23) (24) (25) (26)

(27)

(28)

Our study is generative[X] descriptive[Z]. ... many selectional[X] or[Y] strict-subcategorization[Z] features. Time[X]- and[Y] tape[Z]-bounded automata' [the title of an article]. ... from words that occur[X] too frequently, but[Y] whose meanings are[Z] made precise by other words. He does not specify what the rules of this component are[X] to look like or[Y] how they interact[Z]. Thanks are due to Hood Roberts for somehow keeping the whole system up[X] and[Y] running[Z]. There is[X] no easy solution in this case, nor[Y] is[Z] there likely to be one forthcoming. There are dozens of journals devoted[X] to improving commercial manoeuvres, and[Y] which give[Z] accounts of outstanding players and games. Whether[X] this is permissible and[Y] what kind of limiting conditions must[Z] be taken into account, depends on the definition of the terms "language" and "code". Our competence in a foreign language depends on whether[X], and[Y] to[Z] what extent, the rules we have been taught are equivalent to those that speakers of the foreign language acquired naturally.

Groupings in the conjoined string 40.4.

465

466

COORDINATIVE

► The large bracket on the left-hand side of Syntagm 40.4 (and Syntagms 40.5 and 40.6 as well) indicates that the surface-syntactic structure of the construction in question should represent all the bracketed nodes as forming one grouping. This emphasizes the fact that dependency trees are not always sufficient to render faith­ fully all the syntactic relations observed between textual items: some surface-syntactic information cannot be described naturally in terms of dependecies alone. This is true in particular of the information expressed in Syntagms 40.4 - 40.6 by certain combina­ tions of conjunctions and commas, the presence/absence and the position of the possessive suffix-'s, and hyphens. Cf. pp. 72ff. ◄ Examples (1)

Let's now take verbs, but[W] not nouns[X] or[Y] adjectives[Z].

► Cf. the infelicitous sentence (2): (2) · Let's now take verbs, but not nouns, or adjectives; a stylistically preferble rendering would be (2'): (2') (3)

Let's now take verbs or adjectives, but not nouns. ◄ A person knows not only the structure of his language, but[W] how meaning is[X] conveyed and[Y] something[Z] about the subject of the conversation. Groupings in conjoined possessive modifiers

40.5.

COORDINATIVE

467

Examples (1)

John[X] and[Y] Mary'sfZ] committing[W] of adultery shocked the city

[meaning 'John and Mary committed adultery with each other'; cf. (2)

John's and Mary's committing of adultery,

implying rather that John committed adultery with a woman other than Mary, and Mary with a man other than John]. (3)

Boyde[X = U] and[Y] Thorne's[Z] analyses[W] of modais...

[speaking about a joint article by the two authors].

Groupings in the string of the first components of a compositive phrase

Examples (1)

precedence[X]-and[Y]-dependency[Z] grammars[W] = 'grammars using both precedence and dependency';

468

COORDINATIVE

cf. (2) (3)

precedence and dependency grammars = 'precedence grammars and depen­ dency grammars — two different types'; some set[U]-array[X]-and[Y]-vector[Z] algorithms[W].

With correlative coordinate conjunctions (40.7)

Examples (1) (2) (3) (4) (5)

Scientists need an alternative both[W] to[X] learning numerous languages and[Y] to[Z] reliance on human translators. ... a class which appears in close relation now[W] with[X] one class, now[Y] with[Z] another. He must either[W] divorce[X] her or[Y] send[Z] her to a hospital. Nixon will only go either[W] by the route of resignation[X] or[Y] impeachment[Z]. Attempts at using mathematical models either[W] digressed[X] into purely theoretical research, or[Y] (they) led[Z] to pseudoscience.

COORDINATIVE

With quasi-coordinate conjunctions (40.8) ► Similar French constructions are disscussed in Parun-Sarkisova 1973. A 40.8.

α = (A, not pron, not ord, not quant), (Adv), (V)ppres, (V)pass Examples (1) (2) (3)

a parallelpi] though[Y] later[Z] distinction; a brilliant [X], though[Y] lengthy [Ζ], novel ; heavily contaminatedpi], though[Y] fairly efficient[Z] procedures; an interesting[X], though[Y] sometimes sinister[Z] variant; (4) These explode suddenly[X] though[Y] not unexpectedly[Z]. (5) Repeated[X] though[Y] small-scale[Z] attempts have failed. (6) ... a capable[X] if[Y] frequently corrupt[Z] political administration; (7) few[X] if[Y] any[Z] of the services; (8) It has littlepi], if[Y] any[Z], effect... (9) most[X], if[Y] not all[Z], adj ectives ; (10) a good[X] if[Y] not infallible[Z] guide for differentiation; (11) a dramatic[X], as[Y] distinct[Z] from merely exciting, quality; (12) now[X] as[Y] always[Z]; now[X] even if[Y] not alwaysfZ].

469

470

COORDINATIVE Constructions with phrases of the type ETC. (40.9)

40.9.

Example It means that they know the proper greeting[Z], eating[X], etc.[Y] rituals.

APPENDIX I Parts of Speech and Syntactic Features of English Lexemes We apply the term parts of speech to disjoint subsets within the set of all (English) lexemes — such that each subset, i.e., each part of speech, is a maximal set of lexemes possessing a maximum of common properties that are relevant from the viewpoint of a linguistic description. A classification into parts of speech is thus a partition : no lexeme can simultaneously belong to two different parts of speech. (In cases like Eng. (to) kiss vs. (a) kiss we have two different lexemes related derivationally, and the derivational means involved here is conversion; for a formal definition of conver­ sion as a morphological means see Mel'cuk 1973 and 1982b: 102-104.) We apply the term syntactic features (= SF's) to those properties of (English) lexemes which, along with the part of speech, determine their syntactic behavior. Unlike parts of speech, the syntactic features do not induce a partition: they cross-classify the lexemes allowing for the intersection of their subsets. Thus, a lexeme may be, and normally is, marked by different SF's, and the same SF may mark lexemes of different parts of speech. (For more on parts of speech and syntactic features see Corbett 1981 and Apresjan 1985.) Both the part of speech and the SF's are assigned to a lexeme in the dictionary. Together with the government pattern, they constitute the main part of its syntactic description, the latter providing the only source of information for our surface-syntac­ tic rules (in particular, syntagms). Note that many SF's are closely related to the meaning of the corresponding lexemes, i.e., generally (at least) some SF's can be derived from the lexicographic definition of the lexeme. Thus, the SF's "prop" for proper nouns, "mod" for modal verbs, "communie" for verbs denoting processing of information, "geogr" for geographic features and "param" used to characterize quantities and parameters may serve as good examples. But insofar as such SF's determine the SYNTACTIC behavior of lexemes and, in particular, indicate their ability to participate in various syntactic constructions, we allow for a certain duplication of semantic information in the SF's. This permits us to present syntagms in a more compact and uniform manner, and also provides for their better survey ability. Parts of speech are denoted by obvious abbreviations which consist of or start

472

SURFACE SYNTAX OF ENGLISH

with a capital letter: N, V, Prep, ... Syntactic features are denoted by abbreviations consisting of between three and eight lower case letters: adres, mod, pron, ... Several SF's include the symbol "!", meaning 'only'. For example, SF "adadj" indicates adverbs which can depend on adjectives, participles or adverbs and on some other types of words as well; while the SF "adadj!1 ' marks adverbs which can depend only on adjectives, participles and adverbs, but not. e.g., on verbs (as distinct from (Adv, adadj)). We distinguish NINE parts of speech in English, and the syntactic feature inventory in this monograph includes 82 features. It should be mentioned that many syntactic features are closely correlated and are thus not independent, i.e., they may be deemed redundant. However, it is not our intention to make the list of SF's as short as possible. The introduction of various SF's is determined exclusively by considerations of convenience in formulating the syntagms and presenting them in the most transparent manner. The parts of speech and syntactic features are given in alphabetical order, and each is provided with a brief explanation as well as with examples of lexemes relegated to this part of speech or having this syntactic feature. In connection with this listing, there are two important remarks to be made. First, the lists of examples are not exhaustive, nor even sufficiently representative: their sole purpose is to facilitate the task of the reader. Second, the part of speech or syntactic feature illustrated may refer to only one sense of a polysemous word we use for illustration, without necessarily being valid for its other senses. Thus, for instance, when the SF"pred!" (page 481) is exemplified with fon d we mean only the fond found in be (seem ) fond of[something]', not the other fond as in a fond parent (embrace). The same is true of sorry, which is "pred!" in be sorry [= SORRY 1] but not in a sorry attempt (mess) = SORRY 2, etc. Apart from the most general SF's, like "pron", denoting pronominality, all other SF's are followed by the numbers of the syntagms using that SF, be it in a positive sense, i.e. when its presence is required, or in a negative sense, when it should be absent. The lists of syntagms in which the given SF's are mentioned are purely illustrative and not intended to be exhaustive. Parts of speech 1) A: adjectives in the broadest sense of the term, including 'pure', or 'genuine', adjectives (like blue, transitive, English, etc.), pronominal adjectives (my, all, this, which [in which case], etc.) and ordinal numerals (first, sixth, ...). NB: present and past participles are not considered adjectives; they are treated

PARTS OF SPEECH AND SYNTACTIC FEATURES

473

as verbs and designated (V)ppres and (V)ppas [or (V)pass]. 2) Adv: adverbs in the broadest sense of the term, including 'pure', or 'genuine', adverbs (like already, sometimes, axially, very, rather, only,...), phraseological adverbiais ...), pronominal adverbs (anyhow, somewhere, when, whenever, here, there,...), quantitative adverbs (much, once, twice,...), etc. All noninflected words and set phrases, apart from articles (Art), numerals (Num), interjections (Interj), prepositions (Prep) and conjunctions (Conj), are classified under Adv. 3)

Art: the articles A/AN and THE. See Syntagms 28.1, 28.2, 28.6, and 28.7.

4) Conj : all types of conjunctions, namely, coordinate (marked by the syntactic feature "coord") such as and, or, but, either ... or, etc.; subordinate (the syntactic feature "subord") such as that, though, if, because, when, after, while, etc. ; compara­ tive (SF "compar") such as as, than, etc.; and quasi-coordinate (SF "quasi-coord") such as though, if while, etc. The following relationship obtains between the above subclasses of conjunctions: the features "coord" and "subord" do not cooccur al­ though both are compatible with "compar" and "quasi-coor". 5) Interj: all types of interjections such as ah, oh, hum-hum, ouch! etc. See Syntagm 14.1. 6) N: nouns in the broadest sense of the term, including 'pure' or 'genuine', nouns (likeprogram, cork, ..., modesty, ..., movement, ..., redness, etc.), proper nouns (Mary, France, ...), abbreviations (USA, ft, km, PSG, ...), pronouns of all types such as /, ..., anybody, who, what, each other, that [that of Jakobson], etc. 7) Num: cardinal numerals, both in the form of words and in the form of figures, such as three, fifty, million [three million dollars], dozen [two dozen eggs]; 3.5; 1,000,000; 12; 3/4; etc. See Syntagms 18 and 29. 8)

Prep: prepositions such as about,

in, of, from ... to, on,

during, down [down the river], etc. 9) V: verbs in all forms, such as finite (= tensed) forms, infinitive, gerund, and participles.

474

SURFACE SYNTAX OF ENGLISH Syntactic featuresϯ

1) abbr: abbreviations of all types, i.e. abbreviated common and proper nouns, normally to be found in the dictionary, such as ft [feet], km, wff [well-formed formula], Mr., Prof., MT [Mechanical Translation], M. I. T. [Massachusetts Institute of Techno­ logy] IBM, FBI, Pa [Pennsylvania], etc. Note that abbreviations of the type e.g., i.e., cf., viz., vs., a.m., etc. do not have the feature "abbr" since their syntactic behavior is identical to that of the corresponding non-abridged phrases. See Syntagms 2.8 [Condition 1], 23.1. 29.3, 29.6-7, 29.9-10, 33.1-3. 33.6, 35.2. 35.4, 36.2-3, 38.5-7, 38.18-19,39.3-5,39.10. 2) absol: subordinate conjunctions capable of introducing absolute-conjunc­ tional constructions, such as when [His tool, when at ease, ...], if [if necesary], while [While an officer, he...], etc. See Syntagm 26.1. 3) action: nouns which denote actions, activities, events and processes, such as [a] kick, arrival, death, activity. See Syntagm 38.14 [Condition 2]. 4) adadj: adverbs that can depend on an adjective, a participle or an adverb such as very, rather [recently], so, quite, sufficiently, too1 [large],... See Syntagm 15.3. 5) adadj ! : adverbs that can depend only on an adjective, a participle or an adverb, such as very, too2, etc. See Syntagm 9.1. 6) adcompar: adverbs that can depend on an adjective or an adverb in the comparative degree, such as still [stronger], much [less], etc. See Syntagm 15.4. 7) adjunct: coordinate and subordinate conjunctions that can introduce 'dis­ junctive' sentence elements, such as and, but [She reads, but detective novels only], although [He will come, although late], etc. The conjunctions that have the feature "adjunct*" are considered to be dependent in such constructions on the heads of the phrase they introduce; see Syntagm 15.9. 8) adnum: adverbs that can depend on a cardinal numeral, such as approxi­ mately, precisely, exactly, just, almost, etc. See Syntagm 15.6. 9) adpl!: determiners that can modify only plural nouns, such as few, several, etc. See Syntagms 6.1, 28.4 [Condition 2]. Cf. "adsg!". 10) adres: verbs that take two prepositionless objects of which the first (in the order of appearance in the text) is an indirect object and denotes a beneficiary, such ϯ It is interesting to compare our list of English SF's with the list of semantic, syntactic and morpholog­ ical features of English words appearing in Kittredge 1973: 30ff.

PARTS OF SPEECH AND SYNTACTIC FEATURES

475

as assign, bring, furnish, give, serve [the children their dessert], etc. See Syntagms 2.2., 2.4. Cf. "equat". 11) adsg!: determiners that can modify only singular nouns (or plural nouns provided that the latter are modified by numerals), such as each, every, any, etc. See Syntagms 6.1 and 28.4 [Condition 1]. Cf. "adpir!" 12) adsub: adverbs that can be post-modifiers of a noun, such as above, ahead, here, etc. See Syntagms 4.1. 4.2, 4.4. 33.1 [Condition 1]. 38.13-14, 39.8. 13) adsuperl: adverbs that can depend on an adjective in the superlative degree, such as by far [the smartest], etc. See Syntagms 15.5. 14) approx: prepositions having the general meaning of quantitative approxima­ tion and capable of introducing a nominal phrase with a cardinal numeral or a quantitative noun (such as fortnight) in the role of grammatical subject, direct object or the dependent of a preposition. Some of these prepositions are about, above, around, below, between ... and, beyond, from ... to, over, under, [The dictionary contains from 200 to 250 thousand (between 200 and 250 thousands, close to a million) items], etc. The feature "approx" is used in specifying the standard subtree AAPPROX, see below, p. 487ff. 15) attr!: adjectives that can be used only attributively, i.e. as modifiers of a noun, but not predicatively, i.e. as the complement of a copula verb, such as lower, middle, upper, previous, prior, subsequent, etc. See Syntagm 2.20. Cf. "pred!". 16) auton: the absolute forms of possessive pronouns, such as mine [Mine is not that hot!], yours, ..., theirs. See Syntagms 3.2 [Condition] and 38.1 [Condition 4]. 17) aux: auxiliary verbs, namely, be, have, shall, will, should, would, do; see Syntagms 6.1, 13.11, 15.1. 15.2(l)[Condition], 27.16 [Condition 1,2]. 33.6. 40.1 [Condition 4], 40.3[Condition 7]. 18) bin: components of compound (= binary) prepositions and/or conjunctions. In fact, u bin" is a common 'heading' of several compound features which also include three other positions. The first one indicates — by g, for governor, or d, for depen­ dent, — whether the component in question is the main or the dependent one: the second, whether the component is a left-hand (/) or a right-hand (r) one ; and the third presents the lexeme which forms the other part of the binary combination, e.g., bin-g-l-THEN, that is, if in if... then, bin-d-l-AND [both in both ... and], etc. Apart from the above, the following lexemes are all marked by the in-feature: either, or, neither, nor, now [in now.... now...], see Syntagms 19.1 and 40.7: the" and the2 [in the more ... the better]. Syntagm 19.2; between, from, to, Syntagm 19.3. See Syntagms

SURFACE SYNTAX OF ENGLISH

476 40.3-4,40.7.

19) cifr: digital representations of cardinal numerals, i.e., numerical sequences such as 0, 15, 125847, 3.17, etc. See Syntagms 18.1-4, 29.3-4, 33.6. 20) comma/comma! : parenthetical word and phrases that can/must be set off by commas, such as of course, naturally, etc. See General condition on parenthetical syntagms, p 291. 21) communie: verbs that denote processing of information, such as tell, imagine, describe, etc. See Syntagms 13.4, 13.6-7. 22) compar: comparative conjunctions, such as as, than, but [Nobody knows them but he], unless, etc. See Syntagms 23.4, 24.1. 25.1, 40.3 [Condition 6]. 23) confr: adverbs, prepositions and conjunctions that can govern preposed quantitative phrases, such as after [three weeks after his arrival], before, as [= AS » three times as much], etc. See Syntagm 15.4. 24)

coord: coordinate conjunctions, such as and, but, neither ...nor,

either...or,

, etc. See Syntagms 4.4, 9.6-12 [Conditions], 9.14 [Condition], 15.2(1)-3(1) [Conditions], 23.1. 23.2-4. 30.1 [Condition], 33.1 [Condition 3], 36.4-5,40.3-4,40.7. Cf. ^subord". 25) copul: copula verbs, such as be, seem, become, grow, go [mad], get [pre­ gnant], turn [pale], appear, prove [impossible], remain, etc. See Syntagms 1.3,1.5, 2.9 [Condition 2], 2.12 [Condition 3], 2.22, 9.1 [Condition], 13.9, 13.13, 21.2, 21.4-5. 26) count: countable nouns, such as column, tool, ball, cock, screw, emperor, stroke, kiss, etc. (as distinct from uncountable nouns, such as water, sand, air, wisdom, ...). See Syntagms 29.1-2, 29.6 [Conditions 1], 29.9-10, 38.8. 27)

delim: a. Adverbs that can be syntactically governed by any right-hand adjacent word, except the articles and coordinate conjunctions, and that have a certain 'restrictive' (= 'delimitative , ) meaning, such as almost, even, hardly, just, merely, only, quite, rather, simply, etc. See Syntagms 9.1, 15.2, 26.1, 39.1 [Condition 1]. b. Parenthetical adverbs and set phrases that can be syntactically governed by any left-hand adjacent nominal or adjectival word, such as See Syntagm 13.2. 28) dem: demonstrative pronouns this, that, these and those. See Syntagms 28.3-5, 35.1 [Condition 4].

PARTS OF SPEECH AND SYNTACTIC FEATURES

477

29) det: determiners, namely: articles; demonstrative, negative, indefinite, attributive and interrogative-relative pronominal adjectivais, such as this, no, any, each, every, another, whatever, whichever, which, whose, etc. possessive pronominal adjectivais, such as my, your, ..., their: quantitative adjectivais such as few, much, several, some, etc. Determiners (= words with the feature "det") are incompatible with each other in the adnominal position. See Syntagms 28.4. 28.6-7, 30.1,30.3. 30) direct: prepositions and adverbs which denote direction, such as from, to, through, down/up [the road],... ; away, north, east, southwest, ... See Syntagms 38.7, 39.2,39.5. 31) -ed: an adjective which appears as the second component of a compound adjective of the bahuvrihi type and is derived from a noun by means of the suffix -ed. such as fair-haired, clumsy-fingered, two-handed [game], etc. See Syntagm 33.5. 32) elect: pronominal adjectives (in the broadest sense of the term) which can be the head of an elective construction, such as both [of us], all [of the difficulties], many, much, main, few, little, a little, each, either, neither, none, any, most, last, those, which, ... See Syntagm 34.1. 33) equat: verbs that take two prepositionless objects, of which the second (in the order of appearance in the text) is a complement, such as make [the boy a good swimmer], appoint, call, consider, elect, etc. See Syntagms 2.3 and 2.5. Cf. "adres". 34)

explic: 'explicative' words, i.e. coordinate conjunctions or adverbs that can

introduce sentential appositives. such as , i.e., or, viz., namely, especially, etc. See Syntagms 23.1 [Conditions], 36.4.

.e.g.,alias,

35) eval: nouns of evaluation that can be used in constructions of the type this scoundrel of a professor, that wretch of a ship, see Syntagm 38.1 [Condition 3]. 36) form: formulas, i.e., various non-alphabetical expressions, such as N [words], [the morpheme] -OB, (Vx)P(X), etc. See Syntagms 29.1-2, 33.1-3. 33.6, 35.1-2,35.4,36.2-3,38.5-7.38.19. 37) geogr: nouns that denote geographic features and can be used in construc­ tions of the type the city of New York but not *the city New York: town, state, village, republic, island, etc. See Syntagm 38.1 [Condition 2]. 38) human: nouns that denote human beings, such as man, daughter, program­ mer, scoundrel, ..., anybody, who, etc.; cf. the scoundrel who... but not '-'the scoundrel which... See Syntagms 30.7 [Conditions], 30.10 [Conditions 2 and 3], 35.1 [Condition

478

SURFACE SYNTAX OF ENGLISH

1], 36.6. 39)

indef: indefinite and negative pronouns, such as every, each, some, any, all,

everybody, everyone, everything, somebody, anybody, nobody, none, whatever, whichever, etc. See Syntagms 2.14, 2.22, 23.3 [Condition], 30.5, 30.7, 30.9-10,34.1,38.10,38.15,38.18. 40) -ing: verbal nouns derived by adding the suffix -ing. See Syntagms 3.2 [Condition], 33.1 [Condition 2]. 41) intr: both coordinate and subordinate conjunctions that can introduce independent sentences [= main clauses], such as and, because, but [But I know you did it], for, nor, etc. See Syntagm 9.13. 42) loc: prepositions and adverbs that have a locative meaning, such as in, under, on, here, where, etc. See Syntagms 2.19, 15.2(1) [Condition], 27.7, 31.4, 38.2 [Condition 2], 39.2. 43)

manner: a. Nouns of the type way, manner, reason, time that are capable of govern­ ing conjunctionless relative clauses in constructions of the type the way he did it. See Syntagm 2.19. b. Adverbs of manner, particularly -ly adverbs, such as methodologically, simultaneously, ... See Syntagms 9.1, 38.14 [Condition 2]. 44) mesur: nouns that denote measurement units, such as year, foot, ft, km, db, rouble, time [= TIME 2 ; ten times faster], inch, mile, moment, etc. See Syntagms 2.6, 2.8, 9.3, 9.4, 21.5, 29.1. 29.3. 29.6, 29.7 [Condition 1], 29.9-10, 38.7-8, 38.22, 39.5. 45) mobil: verbs that denote movement, such as go, walk, fly, swim, drag, ... See Syntagm 9.4. 46)

mod: a. Modal verbs can, may, must, need, shall, will, should, would. See Syntagms9.1 [Condition], 13.11,15.1,15.2(1) [Condition],33.6,40.1 [Condition4], 40.3 [Condition 7]. b. Modal adjectives of the type able, sure, likely, unlikely that govern ίο-infinitives [e.g., He is sure to do it]. See the definition of AVP in Appendix II, pp. 489-490. 47)

neg: negative words and phrases [N, A and Adv], such as nobody, nothing,

none, no, not, never, nowhere, 15.2(1) [Condition], 23.3 [Condition], 40.3 [Condition. 10].

... See Syntagms 15.1,

PARTS OF SPEECH AND SYNTACTIC FEATURES

479

48) obj: Object' adjectives, i.e., adjectives capable of governing an infinitive such that the semantic subject of the adjective is the semantic object of the infinitive (cf. This problem is hard to solve), such as difficult, easy, hard, useful, etc. See Syntagms 2.9 and 2.10. 49) ocopr: verbs that allow a non-pronominal object-oriented complement such as like [Olga slim], hate [gin diluted], burn, meet, see, etc. See Syntagm 7.1. Cf. "scopr". 50) ord: ordinal numerals (A, ord), such as first, second, ..., tenth, fifteenth, ... See Syntagms 7.1, 15.3, 15.7(1) [Condition], 15.8, 18.1-2, 18.4 [Condition], 30.2, 30.5-6,34.1,38.3,38.17,38.20. 51)

param: a. Nouns that denote parameters, i.e., physical quantitites and properties, such as color, height, quality, size, shape, ... See Syntagms 2.8, 29.7 [Condition 1], 38.5,39.3. b. Adjectives and adverbs that denote parameters and can have a preposed noun phrase complement, such as high, old, apart [... are a fixed distance apart], etc. See Syntagm 2.8. 52)

parent: parenthetical words and phrases, such as perhaps, probably, then,

respectivley, 15.2.

etc·

See Syntagms 9.1, 13.1-2,

53) partie: particles, i.e., adverbs that (unless used in some marginal expres­ sions) cannot be conjoined with anything, such as quite, rather, too, very, etc. See General condition 1 on coordinative syntagms, p.458, and also Syntagms 23.1-4, 26.1,36.4-5. 54) verbs that normally have no passive form. These are, in particular, such transitive verbs as cost, weigh, measure [It measures five meters], resemble [*Iam resembled by you], fit, escape, etc.; and all intransitive verbs. See Syntagm 16.1. 55)

pers: a. Personal and reflexive pronouns, such as I, me, you, he, him, ..., they, them, myself ..., themselves. See Syntagms 2.1, 2.2, 17.1 [Condition], 23.3, 23.4, 24.1, 25.1, 29.12, 30.7 [Condition], General conditions on Syntagms 31, Syntagms 35.6, 36.1 [Condition 3]. b. Possessive pronominal adjectivais, such as my, ..., their. See Syntagm 28.5. 56)

phas: phasic, or 'aspectual', verbs, such as begin, start, continue, finish,

480

SURFACE SYNTAX OF ENGLISH

cease, etc. See Syntagm 2.12, and also the definition of AVP in Appendix II, p. 489. 57)

postpos: a. Adjectives that can be used without dependents as post-modifiers of a noun, such as alone, available, imaginable, present, possible, ... See Syntagm 30.4. b. Verbs whose passive participles can be used without dependents as post-modifiers of a noun, such as use, classify [All the texts classified were...], give, concern, involve, ...t See Syntagm 30.4.  Adverbs that can be used as post-modifiers of a verb or an adjective, such as methodologically. See Syntagms 15.2(2) [Condition]. 15.3(2) [Condition]. d. Prepositions that can be used following the noun they govern, such as notwithstanding, etc. See Syntagm 27.15. 58)

postpos!: a. Adjectives that can be used only as post-modifiers of a noun, such as proper [Only after one has finished with linguistics proper, can one...], etc. See Syntagms 30.1, 30.2. b. Verbs whose past participles can be used without dependents only as post-modifiers of a noun.t See Syntagm 30.1 [neg], 30.4. c. Adverbs that can be used only in post-position to the verb governing them, such as too, respectively, etc. See Syntagms 9.1(2) [Condition], 13.1(2) [Condition], 14.1(2) [Condition]. d. Prepositions that can be used only following the noun they govern, such as ago, back, etc. See Syntagms 27.1, 27.3, 27.7, 27.13, 27.15. e. Subordinate conjunctions which allow the clause they introduce to be positioned only after the principal clause (in neutral style), such as for, etc. See Syntagm 9.10(2) [Condition 1]. Cf. "prepos!". 59) postv: post-verbs [Adv, A and N], such as up [give up], in [come in], off on, etc. ; shut [bang shut], dead [shoot dead], etc. ; prisoner [take prisoner], etc. See Syntagms 9.1, 20.1 [Condition 3], 33.1 [Condition 3]. 60) postv-L: verbs capable of taking a postverb L to form what might be a phraseme (= a phrasal verb). The following "postv-L" features must be recorded for

ϯ As indicated on page 393, both these features ( = postpos and postpos!) serve to temporarily stake out the problem of single participial post-modifiers. We do not a priori exclude the possibility that under favorable semantic and contextual conditions the passive participles of ALL English transitive verbs can participate in the construction in question (thepoints discussed ... etc.). But until these conditions are discovered we label some verbs whose single participles are at least preferably used in post-position to the modified noun.

PARTS OF SPEECH AND SYNTACTIC FEATURES

481

bring in its dictionary entry: postv-ABOUT, postv-BACK, postv-DOWN, postvFORTH, postv-FORWARD, postv-IN, postv-OFF, postv-ON, etc. See Syntagm 17.1. (This feature is, of course, nothing more than a reference to a separate dictionary entry for each phrasal verb. That is, when we write in the entry BRING "post-ABOUT", this means that there is an entry BRING ABOUT, which supplies all the information concerning the phrasal verb bring about.) 61) pred! : adjectives that in a given sense can be used only predicatively, i.e., as complements with copula and other similar verbs, and also as nominal postmodifiers (but then obligatorily with some dependents), such as afraid, alive, alone, asleep, awake, aware, fond, glad, ill, like, sorry, sure, well, worth, etc. See Syntagms 30.1-3.Cf. "attr!". 62) pred-inf : nouns which occur as copula complements with a grammatical subject manifested by an infinitive, such as delight [To speak with her was a delight], business [only if accompanied by an adjective: interesting (difficult, important, ... ) business], challenge, distinction, etc. See Syntagms 1.3 and 21.2. 63)

preposi: a. Adverbs that can be used as dependents of a verb or an adjective only

inpre-position,suchas,  sort oflaughed], etc. See Syntagms 9.1(1) [Condi­ tion], 13.1(1) [Condition], 14.1(1) [Condition]. Cf. feature [+ PREVERB] in Stockwell et al. 1973: 267. which labels adverbs that can be positioned before the predicate. b. Subordinate conjunctions which allow the clause they introduce to be positioned only before the main clause, see Syntagm 9.10(1) [Condition]. Cf. "postpos!". 64) verbs normally not used in progressive form, such as know, contain, own, etc. See Syntagm 16.2(1) [Condition]. 65) pron: pronominal words, i.e., in particular, words with features "auton" {mine, ours, ...), "dem" (this, that, ...), "det" (the, a, an, my, ..., each, this, ..., neither, ...), "indef" (everybody, something, ...), "neg" (no, nothing, ...), "pers" (I, me, ...; myself, ...my, ...), "ref;" (myself, ...), "rel" (which, that, ...), "wh" (who, whose, how, ...), as well as all, both1, many, much, few, etc. On many occasions the general feature "pron" is used to achieve greater compactness of syntagm formula­ tions and replaces lists of more specific pronominal features. 66) prop: proper nouns, such as Jim McCawley, Quang Phuc Dong, Yuck Foo, Washington, Rome, the Caucasus, the USA, ... See Syntagms 26.1, 29.6-7, 30.4,

482

SURFACE SYNTAX OF ENGLISH

30.6-7, 30.9-10, 32.1-2, 33.1-3, 33.4 [Conditions 2 and 3], 33.6, 35.2-4, 36.1 [Condi­ tion 2], 36.2-3,36.6,37.1 [Condition 2], 37.2 [Condition 2], 38.1 [Conditions 2 and 3], 38.5, 38.8, 38.13-14, 38.18, 39.1 [Condition], 39.4[neg]. 67)

quant: quantitative words, namely: a. Nouns such as dozen [dozens of eggs], score, hundred, thousand, million [millions of workers], ... Cf. Num, p. 473. b. Adjectives such as , all, both, double, enough, few, half, less, little, many, more, much, several, some, etc. See Syntagm 38.8.  Adverbs such as

enough, less, more, much, once, twice, etc.

68) quantif : pronominal quantifiers (in the logical sense) which can be floating, i.e. each, neither, either, all, and both. See Syntagm 6.1. 69)

quasi-coord: quasi-coordinate conjunctions, which cannot conjoin sen­

tences, such as as [here as elsewhere],

interesting if not brilliant paper],

though, etc. See Syntagms 23.1-4, 36.4-5, 40.8. 70) refi: pronouns myself, yourself, ..., themselves conventionally known as reflexives; see Syntagms 6.1,38.10. (The term "reflexive" seems somewhat inapprop­ riate since these pronouns have the reflexivity function only in some of their usages, e.g., in She washes herself, but not in He was himself critical of the attitude of the committee, where himself is used rather for emphasis.) 71) rel: conjunctive words that can introduce a relative clause being its con­ stituent, such as which, who, that4, where, how, when [the day when I first saw her], ...See Syntagms 21.9, 30.7, 30.9-10,31.2. 72) scopr: verbs that allow a non-pronominal subject-oriented complement, such as hang [high], occur [embedded], die [an old man], return [sick], etc. See Syntagm 5.1. (By the way, we are not sure that the introduction of the feature "scopr" is really necessary, since it is quite likely that this syntactic feature is characteristic of ALL English verbs. This is, however, a moot question.) Cf. "ocopr". 73) sent: adverbs which can serve as equivalents of sentences, such as yes, no, please, etc. See Syntagm 14.1. 74) subord: subordinate conjunctions, such as when, if, that5, whereas, etc. See Syntagms 2.22, 9.9-10, 15.2(1) [Condition], 21.10, 22.1-3, 30.8. Cf. "coord". 75) subst: cardinal numerals of they type dozen, gross, hundred, etc., that can be used in the half a.. .N-construction, such as half a dozen eggs. See Syntagm 29.7.

PARTS OF SPEECH AND SYNTACTIC FEATURES

483

76) symb: 'metalinguistic' nouns of the type symbol, word, term, phrase, expression, etc., that can be used in such appositive constructions as the suffix -ed, the phrase hot stuff, etc. See Syntagm 35.5. 77)

temp: words [N, Prep, Conj and Adv] denoting time, namely: a. Nouns that can be used without a preposition as adverbial modifiers answering the question 'When?' or 'For how long?', such as [the next] day, [the first] hour, [the whole] minute, Monday, ..., Sunday, week, year, ... See Syntagms 9.2, 15.4, 21.5, 30.7 [Conditions 3 and 4], 30.8, 30.9 [Condition 2], 30.10 [Condition 4], 38.6,38.9,38.11. b. Prepositions and conjunctions such as during, ago, before, after, since, when, ... See Syntagms 22.1 [Condition 3], 27.7, 27.15 [Condition 1], 30.8. c. Adverbs such as then, early, soon, etc. See Syntagms 2.19, 15.4, 27.7, 31.1,38.2,38.6,38.11,39.2. 78) there-PREP: pronominal adverbs derived from prepositions PREP, such as thereof thereupon, therefrom, etc. See Syntagms 2.18, 38.1. 79) 38.1.

-thing: pronouns something, anything, nothing. See Syntagms 2.3, 27.17,

80) v-there: verbs that can be used in a construction with the anticipatory there [= THERE 2 ], such as exist, arise, come, etc. A list of about 60 such verbs has been compiled by J. McCawley (unpublished). See Syntagm 1.6. 81)

wh: interrogative-relative words and phrasemes, such as who, whom, what,

where, when, how, , why, whose, which, ... See Syntagms2.14, 9.1 (1) [Condition], 15.2(1) [Condition 5], 21.3-4, 21.7, 21.9, 22.1 [Condition], 27.6, 27.10-12, 30.7, 30.9-10, 31.4, 38.15. 82) wh-ever: indefinite pronouns with the component -ever, such as whatever, whatsoever, whichever, whoever, whosoever, whenever, wherever, ... See Syntagms 4.1(2) [Condition], 9.11.

APPENDIX II Standard Subtrees÷ The description of a standard subtree includes three types of data: A. B. C.

A.

Internal structure of the subtree. External connections of the subtree within the SSyntS. Distribution (among the individual nodes of the subtree) of labels that are attached to the subtree symbol as a whole. Internal structure of standard subtrees

1. ΔΝΡ is a tree which represents a noun phrase or its syntactic equivalent. Formally speaking, the notation indicates that node Φ is the root (= the top node) of one of the following seven types or dependency trees (note that a tree may consist of only one node): (i) i.e., a noun such as computer(s), gas, movement, the racing [at Johnstown], New York, Babby; a pronoun such as I, me, ..., one, everybody, anything, ...; or an abbreviation:/f(t), .I.., ...

(ii)

(Num)

not

(N)

i.e., a cardinal numeral not depending on a noun: (1)

Said the two to the tutor...

(iii) not quant)/(V)ppres/(V)ppas

t For the notion of a standard subtree and some details about its usage see §2 of Chapter IV, in particular, p. 192.

SURFACE SYNTAX OF ENGLISH

486

i.e., an occasionally substantivized adjective or participle: (2)

a. b. c.

the brave; my poor; The second is ready. a mixture of the converted and the indifferent; Different secondary patterns impart different meanings to the primary.

(iv)

i.e., an occasionally substantivized cardinal adjective such as few, enough, ... not depending on a noun: (3)

a. b.

Attempting too much at once must be avoided. Far less is known about discourse analysis.

(v)

i.e., an elective construction: (4)

a. b.

I've seen most of them. Λ great many of the exercises were constructed by them.

(vi)

L = WHAT, WHO(M), WHATEVER, WHO(M)EVER, WHICHEVER, WHOSOEVER The tree in question represents subordinate clauses introduced by the above wh-words ; they are capable of filling most of the syntactic positions normally filled by nouns. Such clauses are currently known as free, or headless, relatives: (5)

a. b. c. d.

One's subjective judgments as to what[Ψ] goes[Φ] with what... To discard the syntactic argument is to abandon what[Ψ] objectivity there is [Φ] in linguistic research. The outcome depends on who[Ψ]had[Φ] the initiative. Failure to classify data properly has been one of the main weaknesses of much of what[Ψ] has[Φ ] been done in all social sciences to date.

STANDARD SUBTREES

487

In example (5d), we see a complex ΔΝΡ of type (v) — i.e., one of..., in which another ΔΝΡ, also of type (v) — much of... — is embedded; this latter includes, in its turn, a ΔΝΡ of type (vi) — what has been done ... 2. ΔNUMP is a tree which represents a quantitative phrase. is the top node of one of the following four types of trees:

(i)

i.e., a countable noun with a cardinal numeral or quantitative adjective: (6)

a. three tons; twenty trees; 35 columns; b. twenty to forty students; c. few papers; all girls; both the chapters.

(ii)

i.e., a singular countable noun with an indefinite article. This combination may be syntactically equivalent to a phrase with the numeral ONE1 (such as a mile; a line; It costs a dollar), particularly when the noun is the name of a measurement unit.

(iii) (iv) 3. ΔPPROX is a tree which represents an approximate quantity phrase, a particular variant of a quantitative phrase. is the top node of one of the following six types of trees:

(i)

488

SURFACE SYNTAX OF ENGLISH

i.e., a quantitative phrase introduced by an 'approximate' preposition of the type above, below, ...: (7)

The rally was attended by v[] three hundred[Σ] ρeople[Ψ].

i.e., a quantitative phrase with two nouns each introduced by a component of the binary preposition from - to: (8)

a seasonal reduction of fr[] 15[Σ] per nt[Ψ]tο[Τ] five[N] per cent[P]. ( )

STANDARD SUBTREES

489

i.e., a quantitative phrase with one noun and two conjoined numerals introduced by the binary prepositon from - to: (9)

a seasonal reduction of fr[] 15[N] to [T] 5[Σ] per cent[Ψ].

(vi)

r = quantitative, determinative i.e., a quantitative phrase introduced by the comparative adverb more/less: (12) a. in words of no r[] than[T] two[Σ] syllables[Ψ]; b. the usual analysis the schools have taught for r[] than[T] a hundred and fifty[Σ] years[Ψ]. 4. AVP is a tree without branching, i.e., a chain, representing a 'transparent' verb phrase, i.e., a sequence of consecutively subordinated lexemes such that its syntactic properties as a whole are fully determined by the syntactic properties of its top and terminal nodes (all the intermediary components being irrelevant for its external connections). With respect to its environment AVP behaves like a single verb, its top node agreeing with the grammatical subject, while the possible type of the latter or of the complements and objects following AVP is determined by the government pattern of the terminal node. The information concerning the possible type of the grammatical subject is conveyed along the chain from the terminal to the top node, which explains our referring to the AVP phrase as 'transparent'. E.g., may seem likely to contribute is syntactically equivalent, as regards the combinability with possible grammatical subjects, objects and complements, to the verb contribute, while it is may that implements the agreement, expresses the tense and mood, etc.

490

SURFACE SYNTAX OF ENGLISH is the top node of the chain having the following form:

it may also be a single node (then Φ1 = Φ η ). The terminal node Φ η is labeled with the symbol of an arbitrary verbal lexeme, and all the other nodes, if any, are labeled with lexeme symbols of the following six types: - auxiliary (aux) and modal (V, mod) verbs; - copula (copul) verbs; - phase (phas) verbs, such as START, CEASE, CONTINUE, etc.; - the verbs BE5[fo] and HAVE 2 [to], semantically close to modal verbs; - modal adjectives(A, mod); - the particle TO 2 , which introduces the infinitive. The top node Φ is a finite verbal form, and each link of the chain

must satisfy one of the English syntagms (more specifically, r1 can only be the auxiliary, the 1st completive or [with Φi = TO 2 ] the prepositional SSRel). Examples of AVP: handles, will handle, is handled, could have been handling, may cease to handle, had to handle, might have been sure to handle. B.

Externa connections of standard subtrees

As pointed out above, symbols of standard subtrees are used in the presentation of syntagms only and do not appear in actual surface-syntactic structures. This calls for an explanation as to how expressions of the form a)—» appearing in syntagms should be interpreted with regard to SSyntS. a) An arrow entering a node

is interpreted as entering the top node of A.

b) An arrow leaving a node may be interpreted in one of the following three ways: (i) For all types of ΔΝΡ and ΔNUMP, except ΔAPPROX, an arrow leaving a node is considered to leave the top node of the subtree ΔΝΡ or ANUMP. (ii) For ΔAPPROX, an arrow leaving a node the node labeled Ψ in the list of AAPPROX types.

is considered to leave

STANDARD SUBTREES

491

(iii) For AVP, an arrow leaving a node may be interpreted in the following manner: - the arrow of the predicative SSyntRel is interpreted as leaving the top node of the subtree AVP; - the arrow of any completive SSyntRel is interpreted as leaving the terminal node of the subtree AVP; - there are no arrows of any other SSyntRel's leaving the nodes of AVP in the presentation of syntagms. 

Distribution (among the individual nodes of the subtree) of labels that are attached to the subtree symbol as a whole

Parallel to the three above-mentioned ways of interpreting the arrow leaving a (i)

node , there are three possible kinds of label distribution: For all types of ΔΝΡ and ΔNUMP, except ΔAPPROX, any label attached to a node NP/ΔNUMP,such as lexeme L, syntactics (ξ) and morphological characteristics χ, characterizes the top node of the subtree ΔΝΡ or ΔNUMP.

(ii) For AAPPROX, any label L, (ξ) or χ attached to a node APPROX charac­ terizes the node Ψ of the subtree in question, (iii) For AVP, different labels are interpreted in different ways: - L and (ξ) characterize the terminal node of the subtree AVP; - χ characterizes the top node of the subtree AVP, except for "pass", which characterizes terminal node.

APPENDIX III List of English Lexemes Mentioned in Syntagms ('Syntactic' Lexemes) For each lexeme cited we indicate the numbers of all the syntagms where the given lexeme is mentioned. If it is mentioned within a 'negative' context only, i.e., if it belongs to an expression beginning with the logical operator not (or, sometimes, if it appears after the symbol ≠)the corresponding syntagms number is followed by a [neg] mark; thus ELSE 38.14 [neg] means that the lexeme ELSE appears somewhere in Syntagm 38.14 in its negative context. Superscripts distinguishhomographiClexemes, thereby helping to identify the intended sense. Where necessary, we supply examples to make the meaning clear. In connection with lexemic superscripts the following convention is observed: if in a syntagm all of η homographic lexemesL'must be mentioned, then instead of writing L1, L 2 ,.., L1-n,L"we omit the superscript altogether and put simply L. For example, AS in Syntagm 23.4 (Condition 1) means 'any AS' (p. 355). A/AN [indefinite article]

: 18.3, 18.4, 28.1, 29.7, 29.10, 30.1, 30.2, 30.3, 38.1, 38.8, 38.9, 38.22; ΔNUMP (p. 487), ΔAPPROX (p. 489) A.D. : 38.21 AFTER1 [preposition] : 38.2 AFTER2[temporal conjunction] : 21.5,30.8 AFTERNOON : 38.11 1 ALL [everybody] : 29.6,29.12 ALL 2 [in all directions] : 23.2,30.7 ALL 3[all his life] : 29.6,30.7 A.M. : 38.21 AMONG : 34.1,38.3 AND : 4.4, 15.9, 18.2, 18.4, 27.15, 36.6, 40.5, 40.6 AND SO ON : 40.9 AND THE LIKE : 40.9 AS1 [identify it as...] : 2.21, 8.1, 9.5, 23.4, 26.1, 30.2 AS2 [as reported] : 13.6,13.7,13.8, 13.9,13.10, 23.4, 26.1

494

SURFACE SYNTAX OF ENGLISH

AS3[for him as the best student] : 12.1, 23.4, 26.1, 38.16, 39.9 4 AS [white as snow] : 38.17 AS6 [quickly as he walked] : 2.23,9.1 AS7 [as many sticks [as]...] : 15.4,30.2 [the second as is taken to be AS1] B.C. : 38.21 BE1 [auxiliary: iswriting,etc.] : 9.6, 9.12, 16.1, 16.2, 38.19, 39.10 BE2[copula] : 2.22, 9.6, 9.12, 21.10, 38.18 BE3 [there is/are] : 4.1, 4.2, 4.3, 9.6, 9.12 BE5[He is to go there] : 9.6,9.12 BECOME : 9.1 BEFORE1 [preposition] : 38.2 BEFORE2 [temporal conjunction] : 21.5, 30.8 BESIDES : 27.17 BILLION : 18.2, 18.3 : 29.6, 29.12, 34.1 BOTH1 [both these structures] BOTH2[both... and...] : 40.7 : 23.2, 24.1, 25.1 BUT2 [nothing but wait] BY : 3.1,38.4 CAPE : 35.1 : 30.7,30.10 CASE[in the case when...] DO1 [auxiliary: do not sleep] : 16.5 DO2 : 2.22 3 DOUBLE [double this distance] : 29.7,29.8 DOZEN : 18.3 EACH : 30.7 [conditions], 38.8, 38.9, 38.22 ELSE : 38.14 [neg], 38.15 ENOUGH : 2.9, 13.3,30.1 ETC.,ETCETERA : 40.9 EVER : 40.2 EVENING : 38.11 EVERY : 38.8, 38.9 EXCEPT1 [subordinate conjunction] : 22.3 2 EXCEPT [preposition] : 27.8,27.17 FACT : 13.9, 13.14 FOR1 [for him] : 2.9, 2.12, 3.3, 21.12, 27.9, 38.22

ENGLISH LEXEMES MENTIONED IN SYNTAGMS FORT FRIDAY FROM GIVEN GROSS HALF 1 HALF 2 [half as big as] HALF 3 [half thesum] HAVE 1 [auxiliary: has told us] HAVE 2 [modal: have to] HOW HOWEVER HUNDRED I.E. IF 1 [if... then...— conditional if] IF 2 [whether] IN IN ORDER INSTEAD OF IT1 [neuter pronoun of 3sg] IT5 [It is easy to see — anticipatory it]

: : : : : : : : : : : : : : : : : : : :

495

40.2 38.11 19.3, 38.3 27.5 18.3 18.4 15.4 29.7,29.8 9.12, 10.1 [neg], 16.3 9.12 30.2 30.2 18.2, 18.3 36.4 19.1,22.1 1.2,1.3,2.13,21.1,21.2,21.6 27.5,38.2,38.3 9.7 27.8 24.1 [neg]

: 13.7,13.9,13.9 [neg], 13.11,13.12,13.13 [neg], 13.14, 13.14 [neg], 21.1, 21.2, 21.3, 21.4, 21.5, 21.6,21.7,21.8 IT6 [It is John who...— clefting it] : 2.22,21.9,21.10 LAKE : 35.1 LAST : 38.20 1 LEAST : 34.1,38.3 LESS 2 : 29.9 LESS 3 : 30.2 MANY : 29.9,29.10,38.4 MATTER : 13.9 MILLION : 18.2, 18.3 MONDAY : 38.11 MORE 2 [somemore books] : 15.4,29.9 MORE 3 [two books more / two more books] : 30.2 MORNING : 38.11

496

SURFACE SYNTAX OF ENGLISH

MOST1 MOUNT MUCH LESS NEVER NEXT NIGHT NO NO MATTER NOR NOT NOW2 [conjunction: now ..., now...] O'CLOCK OF ON ONE1 [number: 1] ONE2 [substitute pronoun: thisone]

ONCE ONLY2 [-unique] OR PERCENT P.M. PORT POSSIBLE QUESTION QUITE RARELY RIGHT RIVER SAME SATURDAY SCARCELY SELDOM SHALL1 SHOULD1 [the past tense form

: 34.1,38.3 : 35.1 : 40.2 : 40.2 : 38.20 : 38.11 : 2.14,30.2,30.7 : 22.1 : 40.7 : 40.2 : : : :

40.7 29.5, 38.21 3.2,33.1,34.1,38.1,38.22 38.2,38.3 : 4.1, 29.1, 36.6

: 28.1, 28.2, 28.3, 28.4, 29.1,29.9, 30.1, 30.3, 30.4, 30.7, 30.9, 30.10, 31.2, 36.1, 38.1, 38.13, 38.14,38.18 : 38.9 : 38.20 : 9.10,9.12,9.14,27.15,40.7 : 29.5 : 38.21 : 35.1 : 15.7 : 13.9 : 15.8 : 40.2 : 15.2 : 35.1 : 30.7 : 38.11 : 40.2 : 40.2 : 16.4,22.1

ENGLISH LEXEMES MENTIONED IN SYNTAGMS of SHALL 1 ] SHOULD [the auxiliary of the conditional: If you were there, I should leave] SHOULD 3 [the auxiliary of the subjunctive: He ordered that she should be released immediately]

497

: 16.4

2

: 16.4

: 1.2,1.3,2.13,2.15,9.12,16.4,21.1,21.2,21.6, 22.1 SINCE 2 [temporal conjunction] : 21.5, 30.8 SO : 2.13,2.19,9.1,30.2 SOAS : 9.7 2 SOME [some like it hot] : 14.2 SOME 4 [some more] : 15.4 SOMEBODY : 14.2 SOMETHING : 2.3, 2.5 SUNDAY : 38.11 SUCH : 2.13,30.1 THAN : 8.1, 38.16 [neg] THAT 2 [demonstrative pronoun] : 17.1 THAT 3 [substitute pronoun] : 30.5, 30.7, 30.9, 30.10, 31.2, 36.1, 38.1, 38.13, 38.18 THAT 4 [relative pronoun; which] : 27.1 [neg], 27.10. 30.10 [neg], 31.2 [neg] THAT 5 [subordinate conjunction] : 1.2,1.3,2.13,2.23,21.1,21.2,21.6,21.8,21.10, 22.1,22.3,27.5 THAT IS : 23.1,36.4 THE 1 [definite article] : 15.5, 28.2, 35.1, 35.3, 35.6, 38. THE 2 [the second the in the : 15.4,19.2 sooner, the better] THE 3 [the first the in the sooner, thebetter] : 9.10 [neg], 19.2, 22.1 2 THERE [anticipatory there] : 1.6, 3.3, 4.1, 4.3, 21.11, 21.12, 27.9, 30.9 THING : 2.22, 13.9 THIS [pronoun] : 13.14, 17.1

498

SURFACE SYNTAX OF ENGLISH

THOUGH THOUSAND THROUGH THURSDAY TIME2 [asinfour times] TO 1 [preposition] TO 2 [introducing the infinitive]

TOMORROW TOO 2 [-excessively'] TRILLION TUESDAY TWICE TWO UNLESS WE WEDNESDAY WHAT 1 [interrogative-relative pronoun] WHAT2 [as in What a day!] WHATEVER WHEN WHERE WHICH 1 [interrogative-re­ lative pronoun] WHICH 2 [as in She ignored him, which was un wise] WHICHEVER WHO WHO(M)EVER WHOSOEVER WILL 1 WITH WITHOUT WOULD 1 [the past tense form of WILL1]

: 2.23,9.1 : 18.2, 18.3 : 27.14,37.6 : 38.11 : 15.4, 29.7, 29.8 : 19.3, 27.14, 37.2. 37.5, 37.6, 38.4 : 1.2,1.3,1.4.1.5,2.9,2.10,2.14,2.22,3.3,4.1, 4.2, 9.1, 9.7, 12.1, 13.4, 13.6, 21.1, 21.2, 21.3, 21.4,21.6.21.7.21.10,21.12,22.2,27.6,27.16, 30.10,38.1,38.18,38.19,38.20,39.10,40.1,40.3 : 38.11 : 2.9,30.2 : 18.2, 18.3 : 38.11 : 15.4, 29.7, 29.8, 38.9 : 34.1 : 22.1 : 29.11,36.2 : 38.11 : 2.22,13.1,29.8,30.7 : 30.1 : AVP : 22.3, 30.7, 30.10, 38.16 [neg] : 1.2, 1.3, 2.13, 9.10,21.1, 21.2, 21.6, 27.4, 30.2 : 30.7,30.10 : : : : : : : :

31.2 [neg] ,31.4,31.5 ΔΝΡ 30.7,30.10 ΔΝΡ ΔΝΡ 16.4,22.1 4.2, 38.22 4.2

: 16.4

ENGLISH LEXEMES MENTIONED IN SYNTAGMS WOULD 2 [the auxiliary of the conditional] YESTERDAY YOU

: 16.4 : 38.11 : 14.2,29.11,36.2

499

APPENDIX IV Samples of English Syntactic Phenomena Unaccounted for in the Proposed Description We give here a list of English sentences featuring syntactic phenomena which we have not described in the present version of the model. However, these sentences are found in our corpus and their appearance in Appendix IV seems to be warranted for two reasons. First, even a perfunctory examination of these sentences enables the reader to perceive more clearly the scope of our model — and understanding the current limitations of the model is no less important than realizing its potential. Second, further elaboration of the model should be facilitated by our providing this list of unaccounted-for constructions which will serve as raw material. In particular, many of these constructions should definitely be covered in the course of further exploration in English syntax using other components of the full-fledged syntactic model, which have been foreseen but not yet prepared. We mean compo­ nents different from syntagms, such as, e.g., rules for restoration of surface-syntactic ellipsis, etc. To make the survey easier, all the sentences in the inventory have been arranged in four categories: A. Minor Type Sentences, or sentences without finite verb. B. Sentences with surface-syntactic ellipsis. C. Sentences with syntactic constructions unaccounted for in the present model (because of their rarity, purely colloquial character or for some other reason). D. Sentences containing constructions with unclear SSyntS. (The difference between the categories  and D is that in  we know what syntactic structure should be ascribed to each sentence — but our syntagms do not provide for this; while in D we do not have a clear idea about the structure we would like to assign to each specific sentence.) In the examples that follow we have tried to indicate in a natural manner the point at which each sentence presents problems for our description. Thus in the elliptical sentences of category  the position of the deleted item is shown by the mark In the examples of categories  and D undescribed phenomena are marked by italics.

SURFACE SYNTAX OF ENGLISH

502 A.

Minor Type Sentences (sentences with no finite verb).

We call examples in this category sentences without insisting on their theoretical status. What we mean is simply that these are well-formed fragments of English text that constitute 'full utterances' (whatever this might mean) and appear in coherent English discourse along with 'normal' sentences, even if in special roles (as titles, captions, etc.). With nominal phrase as top node. (1) (2) (3) (4) (5) (6) (7) (8) (9)

A first step in recycling. The formal analysis of natural languages. Britain to nationalize aircraft industry. White to play and win [caption to a chess problem]. Tempers up. And this, despite the fact that (again, as with ICTV) there had been no public demand for these commercial stations. Hence the fact of Sommelweis and other innovators. Why natural languages? So much for the notion that our society is becoming more democratic. With gerund as top node.

(10) Not listening. (11) Sizing up the program. With preposition as top node. (12) On the relation between disjunction and existential quantification. (13) Toward the logical description of languages. With infinitive as top node. (14) But why stop at instrumental adverbs? (15) And now to return to idioms. B.

Elliptical sentences. Ellipsis of grammatical subject.

(1) (2) (3) (4)

Says he is sorry. Didn't hurt a bit, did it? The form is dictated by the same kind of strict rules as apply to algebra. This task is not any more difficult t h a n w o u l d be required under any other theory.

ENGLISH SYNTACTIC PHENOMENA UNACCOUNTED FOR (5)

At the opposite side from hard players, such as workers, are good lawyers...

503

are found among social

Ellipsis of noun in possesive constructions. (6) (7)

The translations by Richens and Booth were surprisingly readable, and Masterman's were even more so. It was a whistling note like a bird's Ellipsis of copula verb.

(8) (9) (10) (11) (12) (13)

Teacher crisis not women's fault. No matter how large or small thefluctuationprobability , the expecta­ tion of M/N or L(0,t) decreases at a fixed rate. , the more the need for syntactic The more ambitious the undertaking analysis will be felt. The more important the relationship between sign and thing or event signified , the less important the phonetic part of the sign. The sooner a little boy starts his 'education' the better The next result of the preceding arguments : the monitor may record huge volumes of data... Ellipsis of strongly governed infinitive.

(14) No, he didn't (15) These words can be segmented into their constituent parts ; and so can the majority of English verbs. Ellipsis in coordinative and comparative constructions (entailing non-connexity of SSyntS) (16) White is more interested in the effects on his enemies than he is in success itself. (17) We want the top nearly white but the bottom a bronze. (18) He cherishes his humble hut more than his rich neighbor the palace on the hill. (19) Public spending must be carefully controlled and the increase kept below that for the current year. (20) Both can be brought into awareness only by appropriate procedures: the origin by some form of analytic therapy and the ulterior aspect by antithesis. (21) We'll try to establish the relative frequency of adjectives to nouns; of adverbs to verbs; of passive to active constructions; of abstract to concrete nouns; etc.

SURFACE SYNTAX OF ENGLISH

504 . (1) (2) (3) (4) (5) (6)

Sentences with syntactic constructions unaccounted for in the present model. How I was Robert McNamaraed into submission. There was nowhere else they could get cement in Britain. She liked where she was living. He resides where he likes. Shoot me down that devil! [I. Asimov] He could knock you off forty Latin verses in an hour.

► (5) and (6) feature the so-called Dativus ethicus. ◀ (7)

They are seldom 100% accurate.

► This is a special example of a compound adjectival, cf. Syntagm 33.4. 4◀ (8) (9) (10) (11) (12) (13) (14) (15)

In thirty-one out of the thirty-nine fewer than nine errors were made. Men lined the back of the room, three deep. Ten times the length of the mast. 14 trade associations cover most product categories from biscuit making to distilling to meat packing and starch. The receptor sites are similarly located in the brains of species ranging from fish to rats to the rhesus monkey. They killed a reported 300 black guerillas, 30 Mozambique soldiers and ten civilians. Albania and Corfu Island are a mere three miles apart. Elements like 1st Person, Singular Number, Present Tense, etc., I shall speak of as inflectional properties.

► In (15) we observe an emphatic inversion of the dependent of a preposition.◀ (16) The authorities built them a new house. ► The pronoun in (16) is the dependent element of a new SSyntRel, absent from our list; this relation could be called the benefactive SSyntRel. ◀ (17) Standing in the whirling whiteness near where I waited were an elderly man and woman. (18) Russian-born; French-backed; Chinese-supported; Indian-occupied; ... [For the construction illustrated in (18), now see I. Mackenzie and I. Mel'cuk, 'English Constructions of the Type French-built [widgets] American Speech, 1986, 61:2,99-120.] (19) I would appreciate it if you could send me a copy. (20) She asked me whether I wanted to cook instead of write.

ENGLISH SYNTACTIC PHENOMENA UNACCOUNTED FOR

505

► A bare infinitive introduced by the preposition INSTEAD OF is not covered by our syntagms; the conditions of its use are fairly involved: cf. the infelicitous expres­ sion *He intended to stay instead of go away [must be ... instead of going away], which is very much like (20). ◀ (21) I gave John and Bill each a present. D.

Sentences with unclear SSyntS.

Italicized in the examples below are portions of the sentences whose syntactic status is ambiguous to us: it is not at all clear on which word of the sentence they are syntactically dependent. (1) Which metal is heavier: copper or iron? (2) The widest unity against the dictatorship, mass organization, mass resistance, mass struggle — these are our aims. (3) 0 per cent to 100 per cent — that is the range of cooling capacity control. (4) They would be classified as follows: Class III, Order (d), Suborder B. (5) ... the author who, in this, the first of two articles, considers some of the aspects of computer-based text processing... (6) The options for structuring time are: 1) Rituals; 2) Pastimes; 3) Games; 4) Intimacy; and 5) Activity. (7) Within moments, the calls started coming in: from family, friends, relatives, and neighbors.

REFERENCES ALEXANDER, D., and W.J. KUNZ 1964 Some Classes of Verbs in English. Bloomington, Ind.: Indiana University. 81 pp. ALLAN, Keith 1971 A Note on the Source of THERE in Existential Sentences. Foundations of Language, 7:1,1-18. 1980 Nouns and Countability. Language, 56:3, 541-567. ANNEAR, Sandra S. 1964 The Ordering of Pre-Nominal Modifiers in English. In: Project on Linguis­ tic Analysis, no. 8, Columbus, Ohio: The Ohio State University, 95-121. APRESJAN,JurijD. 1974 Leksiceskaja semantika. Sinonimiceskie sredstva jazyka [Lexical Seman­ tics. Synonymic Linguistic Means]. Moscow: Nauka. 367 pp. 1980 Tipy informacii dlja poverxnostno-semanticeskogo komponenta modeli "Smysl <=>Tekst"[Types of Information for the Surface-Semantic Compo­ nent of the Meaning-Text Model]. Wien: Wiener Slawistischer Almanach. 119 pp. (Sonderband 1). 1985 Sintaksiceskie priznaki leksem [Syntactic Features of Lexemes]. Russian Linguistics, 9: 2-3,289-317. APRESJAN, Jurij D., Leonid L. IOMDIN, and Nikolaj V. PERCOV 1978 Ob"ekty i sredstva modeli poverxnostnogo sintaksisa russkogo jazyka [Objects and Means of a Surface Syntax Model for Russian]. International Review of Slavic Linguistics, 3:3, 249-312. BARXUDAROV, Leonid S. 1966 Struktur a prostogo predlozenija sovremennogo anglijskogo jazyka [The Structure of Simple Clauses in Modern Enlgish], Moscow: Vysšaja Škola. 200 pp. 1975 Ocerki po morfologii sovremennogo anglijskogo jazyka [Outline of Modern English Morphology]. Moscow: Vyssaja Skola. 156 pp. BAZELL, Charles E. 1950 The Fundamental Syntactic Relations. Časopis pro mo derní filolo gii, 33:1, 9-15.

508

SURFACE SYNTAX OF ENGLISH

BELECKIJ,Mixail I. 1967 Beskontekstnye i dominacionnye grammatiki i svjazannye s nimi algoritmiceskie problemy [CF-Grammars and Domination Grammars: Some Decision Problems]. Kibernetika, No 4, 90-97. BERMAN, Arlene 1973 A Constraint on Tough-Movement. In: Papers from the IXth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 34-43. 1974 Infinitival Relative Constructions. In: Papers from the Xth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 37-46. BOGUSLAVSKIJ, Igor' M. 1978 Otricanie v predloženijax s obstojatel'stvami v russkom jazyke [Negation in Sentences with Circumstantial Adverbiais in Russian]. In: R. Laskowski, Z. Topolinska, eds. Studia Grammatyczne II, Wroclaw etc. : Polish Academy of Sciences, 121-136. BOLINGER, Dwight 1971

The Phrasal Verb in English. Cambridge, Mass. : Harvard University Press. 187 pp. BORKIN, Ann 1973 To be and not to be. In: Papers from the IXth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 44-56. BORKOVSKAJA, LB, E.A. REJMAN, and N.V. USPENSKAJA 1973 Predlogi v angli)'shorn jazyke. Vyrazenie ob"ektnyx otnosenij [Prepositions in English. Marking of Object Relations]. Leningrad. 128 pp. BRESNAN, Joan 1970 On Complementizers: Toward a Syntactic Theory of Complement Types. Foundations of Language, 6:3, 297-321. BRITTAIN, Richard  1971 Indirect Observations about Indirect Objects. In: Ch. Fillmore, ed. Working Papers in Linguistics, vol. 10, Columbus, Ohio: The Ohio State University, 72-84. BURTON-ROBERTS, N. 1975 Nominal Apposition. Foundations of Language, 13:3, 391-419. CARDEN, Guy, and David PESETSKY 1977 Double-Verb Constructions, Markedness, and a Fake Co-ordination. In: Papers from theXIIIth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 82-93. CHOMSKY, Noam 1975 The Logical Structure of Linguistic Theory. New York - London: Plenum

REFERENCES

509

Press. 573 pp. CONNORS, Kathleen 1971 English Adverb Functions: Their Distribution and Relations. In: Etudes en linguistique appliquée à la traduction automatique, Montréal: Univer­ sité de Montréal, 29-49. COOPER, William E., and John R. ROSS 1975 World Order. In: Papers from the Parasession on Functionalism Chicago: Chicago Linguistic Society, 63-111. CORBETT, Greville G. 1981 Syntactic Features. Journal of Linguistics, 17:1, 55-76. CORUM, Claudia 1974 Adverbs... Long and Tangled Roots. In: Papers from the Xth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 90-102. CURME, George O. 1935 Parts of Speech and Accidence. Boston: Heath. XIII / 370 pp. DAHL, Östen 1980 Some Arguments for Higher Nodes in Syntax: A Reply to Hudson's 'Constituency and Dependency'. Linguistics, 18: 5/6, 485-488. DELORME, Evelyne, and Ray  DOUGHERTY 1972 Appositive NP Constructions: we, the men; I, a man; etc. Foundations of Language, 8:1, 2-29. DÖNNGES, U., and Heinz HAPP 1977 Dependenz-Grammatik und Latein-Unterricht. Göttingen: Vandenhoeck & Ruprecht. EMONDS, Joseph E. 1976  Transformational Approach to English Syntax (Root, Structure-Preser­ ving, and Local Transformations). New York etc.: Academic Press. 266 pp. FARSI, A.A. 1975 On the Relative Order of Prenominai Modifiers. Language Sciences, vol. 37, 37. FILLMORE, Charles J. 1965 Indirect Object Constructions in English and the Ordering of Transforma­ tions. The Hague: Mouton. 54 pp. FITIALOV, Sergej Ja. 1962  modelirovanii sintaksisa v strukturnoj lingvistike [On Syntactic Model­ ling in Structural Linguistics]. In: S.K. Saumjan, ed., Problemy strukturnoj lingvistiki, Moscow: Academy of Sciences, 110-114. 1968 Ob èkvivalentnosti grammatik NS i grammatik zavisimostej [On the

510

SURFACE SYNTAX OF ENGLISH

Equivalence of PS and Dependency Grammars]. In: S.K. Saumjan, ed., Problemy struktumoj lingvistiki -1967, Moscow: Nauka, 71-102. FRANCIS, Winthrop N. 1958 The Structure of American English. New York: Ronald. 614 pp. FRASER, Bruce 1970 Some Remarks on the Action Nominalizations in English. In: R. A. Jacobs, P.S. Rosenbaum, eds., Readings in English Transformational Grammar, Waltham, Mass. etc. Ginn. 83-98. 1976 The Verb-Particle Combination in English. Tokyo: Taishukan. 135 pp. FRIES, Charles 1952 The Structure of English. New York: Harcourt & Brace. 304 pp. GAIFMAN, Haim 1965 Dependency Systems and Phrase Structure Systems. Information and Control 8:3,304-337. GARDE. Paul 1977 Ordre linéaire et dépendance syntaxique: contribution à une typologie. Bulletin de la Société de linguistique de Paris, 72:1, 1-19. GLADKIJ, Aleksej V. 1966 Eekcii po matematičeskoj lingvistike dlja studentov NGU [Lectures on Mathematical Linguistics for Students of the Novosibirsk State Univer­ sity]. Novosibirsk: Novosibirsk University. 189 pp. 1968 Ob opisanii sintaksiceskoj struktury predlozenija [On Describing the Syntactic Structure of a Sentence]. Computational Linguistics, 8, 21-44 (Budapest: Computing Centre of the Hungarian Academy of Sciences). 1971 Opisanie sintaksiceskoj struktury predlozenija s pomošč'ju sistem sintaksiceskix grupp. I. Formal'nyj apparat [Describing the Syntactic Structure of a Sentence in Terms of Systems of Syntactic Groups. I. Formal Apparatus]. Naucno-texniceskaja informacija, serija 2, No. 9, 35-38. 1980-81 Opisanie sintaksiceskoj struktury predlozenija s pomošč'ju sistem sintaksiceskix grupp. II. Lingvisticeskaja interpretacija [Describing the Syntactic Structure of a Sentence in Terms of Systems of Syntactic Groups. II. Linguistic Interpretation]. Slavica (Debrecen), vol. 17, 5-38; vol. 18, 21-49. GLADKIJ, Aleksej V., and Igor A. MEL'ČUK 1969 Tree Grammars. Intern. Conf. on Comput. Linguistics [preprint no. 1]. Stockholm. 7 pp. 1974 Grammatiki derev'ev. II.  postroeniju Δ-grammatik dlja russkogo jazyka [Tree Grammars. II. Constructing Δ-Grammars for Russian].

REFERENCES

511

Informacionnye voprosy semiotiki, lingvistiki i avtomaticeskogo perevoda, vol. 4, 4-29. 1975 Tree Grammars: I. A Formalism for Syntactic Transformations in Natural Languages. Linguistics, vol. 150, 47-82. GORALČIKOVÁ, Alla 1973 On One Type of Dependency Grammar. In: Prager Autorengruppe, Functional Generative Grammar in Prague, Kronberg/Taunus: Skriptor, 64-81. GREEN, Georgia M. 1980 Some Wherefores of English Inversions. Language, 56:3, 582-602. GREENBAUM, Sydney 1969 Studies in English Adverbial Usage. Coral Gables, Florida: University of Miami Press. [Miami Linguistic Series, no. 5.] HAPP, Heinz 1976 Grundfragen einer Dendenzgrammatik des Griechischen. Göttingen: Vandenhoeck & Ruprecht. 597 SS. 1977 Syntaxe latine et théorie de la valence. Les études classiques, vol. 45, 337-366. 1978 Théorie de la valence et l'enseignement du français. Le français moderne, 46:2,97-134. HARRIS, Zellig S. 1962 String Analysis of Sentence Structure. The Hague: Mouton. 70 pp. 1963 Immediate-Constituent Formulation of English Syntax. Philadelphia: University of Pennsylvania. HAYS, David G. 1960 Basic Principles and Technical Variations in Sentence Structure Determina­ tion. Santa Monica, Calif.: RAND Corporation (Mathematical Division. P-1934). 19 pp. [Reprinted in:  Cherry, ed.. Information Theory. 1961. Washington: Butterworths, 367-374]. 1961 Grouping and Dependency Theories. In: H.P. Edmundson, ed., Proceed­ ings of the National Symposium on Machine Translation, Englewood Cliffs, N.J.: Prentice-Hall, 259-266. 1964a Connectability Calculations, Syntactic Functions and Russian Syntax. Mechanical Translation, v. 8, 32-51. [Reprinted in: D.G. Hays, ed. Readings in Automatic Language Processing, 1966. New York: American Elsevier, 107-125.] 1964b Dependency Theory: a Formalism and Some Observations. Language, 40:4,511-525.

512

SURFACE SYNTAX OF ENGLISH

HERRINGER, Hans-Jürgen 1970 Einige Ergebnisse und Probleme der Dependenzgrammatik. Der Deutschunterricht, 22:4, 42-98. HEWSON, John 1972 Article and Noun in English. The Hague-Paris: Mouton. 137 pp. HIRSCHBERG, Lydia 1961 Le relâchement conditionnel de l'hypothèse de projectivité. In: Discus­ sions sur l'hypothèse de projectivité [Rapport no. 35 du Centre de Traite­ ment de I'information scientifique], I spra: EURATOM, 1-15. HOUGH, George A. 1971 Structures of Modification in Contemporary American English. The Hague-Paris: Mouton. 125 pp. HUDDLESTON, Rodney D. 1971 The Sentence in Written English. Cambridge: University Press. 344 pp. 1975 Homonymy in the English Verbal Paradigm. Lingua, 37:1, 151-176. 1976 An Introduction to English Transformational Syntax. London: Longman. 273 pp. HUDSON, Richard A. 1976 Arguments for a Non-transformational Grammar. Chicago - London: University of Chicago Press. 214 pp. 1980a Constituency and Dependency. Linguistics, 18: 3/4, 179-198. 1980b A Second Attack on Constituency: A Reply to Dahl. Linguistics, 18: 5/4, 489-504. IOMDIN, Leonid L. 1979 Esce raz  sintaksiceskom soglasovanii v russhom jazyke [Once Again on Syntactic Agreement in Russian]. Moscow: Institut Russkogo Jazyka. 45 pp. (Predvarit. publ. PGÈPL. Vyp. 122). IORDANSKAJA, Lidija N. 1963  nekotoryx svojstvax pravil'noj sintaksiceskoj struktury (na materiale russkogo jazyka) [On some Properties of Correct Syntactic Structures (on the Basis of Russian)]. Voprosy jazykoznanija, No. 4, 102-112. 1964 Svojstva pravil'noj sintaksiceskoj struktury i algoritm  obnaruzenija (na materiale russkogo jazyka) [The Properties of Correct Syntactic Stuctures and an Algorithm for their Detection (on the B asis of Russian)]. Problemy kibernetiki, vol. 11, 215-244. JACKENDOFF, RayS. 1972 Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press. 400 pp.

REFERENCES

513

JACOBSON, Sven 1964 Adverbial Positions in English. Stockholm: AB Studentbok. 1971 Studies in English Transformational Grammar. Stockholm: Almqvist & Wiksell. 139 pp. [Acta Universitatis Stockholmiensis. Stockholm Studies in English. XXIII]. JAKOBSON, Roman 1957 Shifters, Verbal Categories, and the Russian Verb. Cambridge, Mass.: Harvard University. 14 pp. [Reprinted in: R. Jakobson, Selected Writings, vol. II, 1971, The Hague-Paris: Mouton, 130-147.] 1959 Boas' View of Grammatical Meaning. American Anthropologist, 61:5, part 2, 139-149. [Reprinted in: R. Jakobson, Selected Writings, vol. II, 1971, The Hague-Paris: Mouton, 489-496.] JELITTE, Herbert 1973 Modelle der Konstituenten- und Dependenzgrammatik im Russischen. Die Weltder Slaven, 18:1, 216-235. JESPERSEN, Otto 1954 A Modern English Grammar on Historical Principles, I-VII. London: Allen & Unwin. 1965 The Philosophy of Grammar. New York - London: Norton. 357 pp. 1969 Analytical Syntax. New York etc.: Holt, Reinhart and Winston. 160 pp. KAČALOVA,

1957

K.N., and E.E. IZRAILEVlČ

Prakticeskaja grammatika anglijskogo jazyka [Practical English Gram­ mar]. Moscow. KAZINCOV, A.I. 1978 Antisvoj stva binarnyx otnosenij [Anti-properties of Binary Relations]. In : A.V. Gladkij, ed., Matematičeskaja lingvistika i teorija algoritmov, Kalinin: Kalinin University Press, 62-68. KITTREDGE, Richard I. 1973 An English Analysis Grammar. In: TAUM 73, Montréal: Université de Montréal, 13-39. KLIMONOW, Gerta A., Stephan NÜNDEL, and Ingrid STARKE 1969 Syntaktische Analyse. In: Automatische Sprachübersetzung. Deutsch-Rus­ sisch, Berlin: Akademieverlag, 66-123. KRUISING A, Etsko 1931-32 A Handbook of Present-Day English. Groningen: Noordhoff. KULAGINA, OlgaS. 1970 Algoritm sintaksiceskogo analiza v sisteme francuzsko-russko go mašinnogo perevoda [A Syntactic Analysis Algorithm for a French-to-Russian

514

SURFACE SYNTAX OF ENGLISH

Machine Translation System]. Moscow: Inst. Prikl. Matem. AN SSSR. (Preprint no. 13). 47 pp. KUNZE, Jürgen 1965 Zur syntaktischen Synthese. Kybernetika, 1:1, 85-102. 1972 Die Komponenten der Darstellung syntaktischer Strukturen in einer Abhängigkeitsgrammatik. The Prague Bulletin of Mathematical Linguis­ tics, wol 18,15-27. KUNZE, Jürgen, and W. PRIESS' 1967-71 Versuch eines objektivierten Grammatikmodells. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung, B.20, 415-448; B.21, 421-466; B.23, 347-378; B.24, 373-402. KURODA, Sige-Yuki 1968 Review of Ch. J. Fillmore's Indirect Object Constructions in English and the Ordering of Transformations. Language, 44:2, 374-378. LANGENDOEN, Donald Terence 1970 Essentials of English Grammar. New York. LECERF, Yves 1960 Programme des conflits, modèle des conflits. Traduction automatique 1:4, 11-18;1:5,17-36. LEE,D.A. 1971 Quantifiers and Identity in Relativisation. Lingua, 27:1, 1-19. LEES, Robert . 1968 The Grammar of English Nominalizations. The Hague: Mouton. 205 pp. LEVI, Judith N. 1978 The Syntax and Semantics of Complex Nominals. New York etc. : Academic Press. 301 pp. LYTLE, Eldon G. 1974 A Grammar of Subordinate Structures in English. The Hague-Paris: Mouton. 139 pp. MACHOVÁ, Svatava 1975 Die Abhängigkeitsgrammatiken. In: Prager Autorengruppe, Einführung in die generative Grammatik, Kronberg/Taunus: Skriptor, 146-154. MALKIEL, Yakov 1959 Studies in Irreversible Binomials. Lingua, 8:2,113-160. [Reprinted in: Ya. Malkiel, Essays on Linguistic Themes, 1968, Oxford: B. Blackwell, 311-355.] MARCUS, Solomon 1965a Sur la notion de projectivité. Zeitschrift für mathematische Logik und

REFERENCES

515

Grundlagen der Mathematik, B.11, 181-192. 1965b Sur une description axiomatique des liens syntaxiques. Zeitschrift für mathematische Logik und Grundlagen der Mathematik, B. 11, 291-296. MATTHEWS, P.H. 1981 Syntax. Cambridge etc.: Cambridge University Press. 306 pp. McCAWLEY,James D. 1968 The Role of Semantics in a Grammar. In: E. Bach and R.T. Harms, eds., Universals in Linguistic Theory, New York etc.:Holt, Rinehart and Winston, 124-169. MEISEL, Jürgen M., and Martin D. PAM, eds. 1979 Linear Order and Generative Theory. Amsterdam: John Benjamins. 507 pp. MEL'ČUK,Igor A. 1963 Avtomaticeskij analiz tekstov (na materiale russkogo jazyka) [Automatic Text Analysis (on the Basis of Russian)]. In: Slavjanskoe jazy ko manie, Moscow: Nauka, 477-509. 1964a Avtomaticeskij sintaksiceskij analiz. 1. Obscie principy. Vnutrisegmentnyj sintaksiceskij analiz [Automatic Syntactic Analysis. 1. General Principles. Intrasegmental Syntactic Analysis]. Novosibirsk: Academy of Sciences, Siberian Branch. 359 pp. 1964b Tipy svj azej meždu èlementami teksta i tipologij a j azykov [Types of Links Between the Elements of a Text and Linguistic Typology]. In: Materialy Konferencii "Aktual'nye voprosy sovremennogo jazykoznanija i lingvisticeskoe nasledie E.D. Polivanova". Samarkand: Samarkand University, vol. 1,57-59. 1965 Porjadok slov pri avtomatičeskom sinteze russkogo teksta (predvaritel'noe soobscenie) [Word Order in Automatic Synthesis of Russian Texts (Preliminary Report)]. Naucno-texniceskaja informacija, no. 12, 36-44. 1967 Ordre des mots en synthèse automatique des textes russes. T.A. Informa­ tions, 8:2, 65-84. 1970 Towards a Functioning Model of Language. In: M. Bierwisch, Κ.Ε. Heidolph, eds., Progress in Linguistics, The Hague-Paris: Mouton, 198-207. 1973 Towards a Linguistic 'Meaning = Text' Model. In: F. Kiefer, ed., Trends in Soviet Theoretical Linguistics, Dordrecht: Reidel, 33-57. 1974a Esquisse d'un modèle linguistique du type "Sens = Texte". In: Problèmes actuels en psycho-linguistique. Colloques internationaux du CNRS, 206, Paris: Editions du CNRS, 291-317.

516

SURFACE SYNTAX OF ENGLISH

1974b Opyt teorii lingvisticeskix modelj"Smysl <=>Test Tekst" [Toward a Theory of Meaning-Text Linguistic Models]. Moscow: Nauka. 314 pp. 1977 Le cas. Revue des études slaves, 50:1, 5-36. 1978 Théorie de langage, théorie de traduction. Meta, 23:4, 271-301. 1979a Studies in Dependency Syntax. Ann Arbor, Mich. : Karoma. 163 pp. 1979b Syntactic, or Lexical, Zero in Natural Language. In: Proceedings of the Vth Annual Meeting of the Berkeley Linguistics Society, Berkeley, Calif. : BLS, 224-260. 1981a Meaning-Text Models: A Recent Trend in Soviet Linguistics. In: Annual Review of Anthropology, vol. 10, 27-62. 1981b Types de dépendance syntagmatique entre les mots-formes d'une phrase. Bulletin de la Société de linguistique de Paris, 76:1, 1-59. 1982a Lexical Functions in Lexicographic Description. In: Proceedings of the Vllth Annual Meeting of the Berkeley Linguistics Society, Berkeley: BLS, 427-444. 1982b Towards a Language of Linguistics. München: Fink. 160 pp. MEL'ČUK, Igor' Α., and Nikolaj V. PERCOV 1973a Fragment modeli anglijskogo poverxnostnogo sintaksisa (predvaritel'noe soobscenie) [A Fragment of an English Surface Syntax Model (Preliminary Report)]. In: PredvariteI'nye publikacii PGÈPL, Vyp. 35, Moscow: Inst. Russkogo Jazyka AN SSSR, 3-18. 1973b Poverxnostno-sintaksiceskie otnosenija v anglijskom jazyke [Surface-Syn­ tactic Relations in English]. Moscow: Inst. Russkogo Jazyka AN SSSR. 80 pp. (Predvaritel'nye publikacii PGÈPL. Vyp. 43). 1975 Model' anglijskogo poverxnostnogo sintaksisa. Časti I-II; Priloženija [A Model of English Surface Syntax. Parts I-II; Appendices]. Moscow: Inst. Russkogo Jazyka AN SSSR. 231 pp. (Predvaritel'nye publikacii PGÈPL, Vyp. 64-66). MEL'ČUK, Igor Α., and Aleksandr K. ŽOLKOVSKIJ 1970 Towards a Functioning Meaning-Text Model of Language. Linguistics, vol. 57,10-47. MEN'KOVA, V.V. 1969 Artikl' pri perecislenii [The Article in Enumerations]. In: LP. Ivanova, ed., Voprosy struktury anglijskogo jazyka v sinxronii i diaxronii, Lenin­ grad: University of Leningrad Press, 59-63. MILSARK, Gary 1972 Re: doubl-ing. Linguistic Inquiry, 3:4, 542-549.

REFERENCES

517

MORGAN, Jerry L. 1968 Some Strange Aspects of IT. In: Papers from the IVth Regional Meeting of the Chicago Linguistic Society, Chicago: University of Chicago, 81-92. NICHOLS, Johanna 1978 Secondary Predicates. In: Proceedings of the IVth Annual Meeting of the Berkeley Linguistics Society, Berkeley, Calif.: BLS, 114-127. 1979 The Meeting of East and West: Confrontation and Convergence in Contemporary Linguistics. In: Proceedings of the Vth Annual Meeting of the Berkeley Linguistics Society, Berkeley, Calif.: BLS, 261-276. NIDA, Eugene 1966 A Synopsis of English Syntax. The Hague-Paris: Mouton. 174 pp. NILSEN, Don Lee Fred 1972 English Adverbiais. The Hague-Paris: Mouton. 197 pp. OOSTEN, van, Jeanne 1977 Subjects and Agenthood in English. In: Papers from the XIIIth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 459-471. P A D U C E V A , Elena V.

1964 O sposobax predstavlenija sintaksiceskoj struktury predlozenija [On Methods of Representing the Syntactic Structure of a Sentence]. Voprosy jazykoznanija, No. 2, 99-113. PARUN-SARKISOVA, M.G. 1973  problEme konstrukcij s "podcinitel'nymi" sojuzami ν sovremennom francuzskom jazyke [The Problem of Constructions with "Subordinate" Conjunctions in Modern French]. In: Sbornik naucnyx trudov (MGPLLJa im. M. TOREZA). Vyp. 74. Moscow, 240-250. PERCOV, Nikolaj V. 1976 O grammaticeskix kategorijax anglijskogo glagola [On Grammatical Categories of the English Verb]. Moscow: Inst. Russkogo Jazyka AN SSSR [Predvaritel'nye publikacii PGÈPL, vyp. 90]. 62 pp. 1977 Grammaticeskoe predstavlenie anglijskoj glago'I noj formy dija celej avtomaticeskogo perevoda [The Grammatical Representation of English Verb Forms for Automatic Translation]. Linguistica Silesiana, vol. 2, 139-163. 1978a  ponjatii grammaticeskogo predstavlenija formy slova (na materiale anglijskogo jazyka) [On the Notion of Grammatical Representation of a Word Form (Using English Data)]. Moscow: Izdatel'stvo MGU [The abstract of the Ph.D. thesis]. 19 pp.

518

SURFACE SYNTAX OF ENGLISH

1978b K voprosu o morfologiceskom sostave anglijskix suščestvitel'nyx [On the Morphology of English Nouns]. In: PredvariteV'nye publikacii PGÈPL, vyp. 107, Moscow: Inst. Russkogo Jazyka AN SSSR, 28-31. PERCOV, Nikolaj V., and S.D. POLOVKO 1978 Naxozdenie aktantov nelicnyx glagoI'nyx form pri avtomaticeskom analize francuzskogo teksta [Identifying the Actants of Nonfinite Verb Forms during Automatic Analysis of French Texts]. Moscow: Institut Russkogo Jazyka AN SSSR. 38 pp. POSPELOVA, A.G. 1969 Sootriošenie ponjatij: sočinenie, sopodcinenie, toždestvo funkcij i odnorodnost' v prostom predlož enii (na materiale sovremennogo anglijskogo jazyka) [Correlations Between the Following Concepts: Coordination, Co-subordination, Function Identity and Conjoinability in a Clause (in Modern English)]. In: I.P. Ivanova, ed., Voprosy struktury anglijskogo jazyka v sinxronii i diaxronii, Leningrad: University of Lenin­ grad Press, 35-41. POUTSMA, H. 1926-29 A Grammar of Late Modern English. Groningen: P. Noordhoff. PREPARATA, F.P., and R.T. YEH 1973 Introduction to Discrete Structures for Computer Science and Engineering. Reading, Mass., etc.: Addison-Wesley. 354 pp. QUIRK, Randolf, Sydney GREENBAUM, Geoffrey LEECH, and Jan SVARTVIK 1972 A Grammar of Contemporary English. London: Longman. 1120 pp. ROBINSON, Jane J. 1970a A Dependency-based Transformational Grammar. In: Actes du X-ème Congrès International des Linguistes (Bucareste, I967), vol. 2, Bucharest, 807-813. 1970b Dependency Structures and Transformational Rules. Language, 46:2, 259-285. ROSS, John R. ("Haj") 1974 There, (There, (There, (There,...))). In: Papers from the Xth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 569-587. SABOURIN, Conrad 1977 Adverbs and Comparatives: An Analytical Bibliography. Amsterdam: John Benjamins. 208 pp. SAGER, Naomi 1968 A Computer String-Grammar of English. New York: New York University

REFERENCES

519

(String Program Report No. 4. New York University Linguistic String Project). 1981 Natural Language Information Processing. A Computer Grammar of English and Its Applications. Reading, Mass.: Addison-Wesley. 399 pp. SANDERS, Gerald A. 1972 Equational Grammar. The Hague-Paris: Mouton. 187 pp. SANNIKOV, Vladimir Z. 1979-80 Socinitelnye i sravnitel'nye konstrukcii: ix blizost', ix sintaksiceskoe predstavlenie. I-II [Coordinate and Comparative Constructions: Their Proximity and Their Representation]. Wiener Slawistischer Almanach, B. 4: 413-432; B. 5: 221-242. SHEINTUCH, Gloria 1980 The There-Insertion Construction in English — A Pragmatic Strategy for Promoting Certain Syntactic Structures. Glossa, 14:2, 168-188. SMIRNICKIJ, Aleksandr I. 1959 Morfologija anglijskogo jazyka [Morphology of English]. Moscow: Izdatel'stvo Literatury na Inostrannyx Jazykax. 440 pp. SMITH, Carlotta 1961 A Class of Complex Modifiers in English. Language, 37:3, p. I, 342-365. 1964 Determiners. Language, 40:1, 37-52. SMITH, G.R., and G. JOHNSEN 1981 A Bibliovect Guide to the Literature in English and Theoretical Syntax. New Canaan, Connect.: Information Reduction Research. 250 pp. STOCKWELL, Robert, Paul SCHACHTER, and Barbara PARTEE 1973 The Major Syntactic Structures of English. New York etc. : Holt, Rinehart and Winston. 847 pp. TESNIÈRE, Lucien 1959 Eléments de syntaxe structurale. Paris: Klincksieck. 670 pp. [Second edition, revised and corrected: 1969.] TEYSSIER, J. 1968 Notes on the Syntax of the Adjective in Modern English. Lingua, 20:3, 225-249. THORNE,James P. 1972 On Nonrestrictive Relative Clauses. Linguistic Inquiry, 3:4, 552-556. URYSON, Elena V. 1981 Poverxnostno-sintaksiceskoe predstavlenie russkix appozitivnyx konstrukcij [Surface-Syntactic Representation of Russian Appositive Construc­ tions]. Wiener Slawistischer Almanach, B.7, 155-215.

520

SURFACE SYNTAX OF ENGLISH

VASILEVSKIJ, A.L., L. Ja. VASIL'EVA, S.K. POVOLOCKAJA, M.V. SAMOJLOVlČ, G. A. TARASOVA, and Ju. M. ÈMDINA 1972 Algoritm avtomaticeskogo perevoda opisanij izobretenij s anglijskogo jazyka na russkij [An Algorithm for the Automatic Translation of Patent Descriptions from English into Russian]. Moscow. VATER, Heinz 1975 Toward a Generative Dependency Theory. Lingua, 36:2/3, 121-145. VENDLER, Zeno 1967 Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press. 203 pp. WARNE, Janet L. 1974 Parenthetical Expressions in English. Montreal Working Papers in Linguis­ tics, vol. 2, 51-65. WASOW, Thomas 1977 Transformations and the Lexicon. In: P. Culicover et al., eds., Formal Syntax, New York etc. : Academic Press, 327-360. WIERZBICKA, Anna 1972 Semantic Primitives. Frankfurt/M.: Athenäum. 235 pp. 1980 Lingua Mentalis. The Semantics of Natural Language. Sydney etc. : Academic Press. 367 pp. XAJMOVIC, B.S., and B.I. ROGOVSKAJA 1967 Teoreticeskaja grammatika angli]'skogo jazyka [A Theoretical Grammar of English]. Moscow. ZWICKY, Arnold M. 1974 Hey, Whatsyourname! In: Papers from the Xth Regional Meeting of the Chicago Linguistic Society, Chicago: CLS, 787-801. ZOLKOVSKIJ, Aleksandr K. 1971 Sintaksis somali [Somali Syntax]. Moscow: Nauka. 265 pp. ZOLKOVSKIJ, Aleksandr K., and Igor' A. MEL'ČUK 1965 O vozmožnom metode i instrumentax semanticeskogo sinteza [On a Possible Method and Some Tools for Semantic Synthesis]. Naucno-texničeskaja Informacija, No. 6, 23-28. 1967 O semanticeskom sinteze [On Semantic Synthesis]. Problemy kibernetiki, vol. 19, 177-238. [For a French translation, see T.A. Informations, 1970, No. 2,1-85; for an English translation, see Systems Theory Research, vol. 19,N.Y.].

SUBJECT AND TERM INDEX Accusative with Infinitive construction: 97 actant: 93 actant relation: see syntactic relations, actant active valence: see valence, active actual lexeme: see lexeme, actual adjunct (verbal): 121ff.,325ff. agent: 99 analytical form: 167 anaphora: 29 anticipatory it: see it, anticipatory anticipatory there: see there, anticipatory antisymmetry of syntactic relations: see syntactic relations, their antisymmetry antitransitivity of syntactic relations: see syntactic relations, their antitransitivity apposition: 144 -, arc (of a graph): 21 -, arguments of a functor: 20-21,64 asyndetic (=conjunctionless) coordination: 153, 459 attribute: 148ff. - , non-restrictive: 152-153 - , restrictive: 152-153 attributive relation: see syntactic relations, attributive auxiliaries: 120 'bare' infinitive: 370

-,

basic syntactic operations: 39 branch (of a syntactic tree): 26, 31 case (in English): 168,174 circumstantial: 93-94 clefting it: see it, clefting comment: see rheme comment clause: 117 communicative organization: 21-22, 29, 179 comparand: 356 completive: see SSyntRel, completive components of the MTM: 35ff. compound nominal: 140, 410 compound numeral: see numeral, compound compound syntagm: see syntagm, compound conjunct: 153ff. conjunction(s): 27 correlative: see paired governed: 22 paired: 123,325,331 conjunctionless coordination: see asyndetic coordination conjunctionless subordinate clause: 243 constituent: 49,75 contingent verbless clause: 283 controller: 193 cooccurrence: 79ff.,84 coordination (vs. subordination): 27,

522

SURFACE SYNTAX OF ENGLISH

64ff., 153, 194-195 see also subordinate conjunctions in coordination copula: 99 coreference: 29 correlative conjunctions: see conjunctions, correlative correlative prepositions: see prepositions, correlative correspondence (between meanings and texts): 12 criteria for SSyntRel's: 69-72 cybernetic model: 13, 15 D deepD-: dependency dangling preposition: see preposition, dangling deep morphological-prosodic structure: 32, 164 deep-morphological representation of a sentence: 32, 164 deep-morphological representation of a wordform: 32, 58, 167 -, full: 59 - , reduced: 59 deep-morphological structure: 32, 164-165 deep syntactic-anaphoric structure: 29 deep syntactic-communicative structure: 29 deep-syntactic component (of the MTM) : 40-42 deep syntactic-prosodic structure: 29 deep-syntactic relation: 26,29 deep-syntactic representation: 22 deep-syntactic structure: 22 degree adverb: 119

dependence on a phrase vs. dependence on the head of a phrase: 65ff. dependency, syntactic: 56, 62 dependency formalism: 7,48ff. detached (element of a sentence): see disjunctive determiner: 133, 371ff. dictionary (in a MTM): 40,60 dictionary entries of lexemes: 36 directed graph: see graph, directed directed ( = antisymmetric) relation: 54 disjunctive (element of a sentence): 316,353 ditransitive verb: 319-320 dummy: 48, 125 Δ-grammar: 40 element (=part) of the sentence: 62 ellipsis: 42,246 empty word: see lexeme, empty existential there: see there, anticipatory feedback between meanings and texts (in a MTM): 14 fictive lexeme: see lexeme, fictive floating quantifier: see quantifier, floating floating reflexive: 266 for + N + to + V inf phrase: 132,236, 257, 365 foregrounded vs. backgrounded: 22 free relative: 228,486 full DMophR: see DMorphR, full full lexeme: see lexeme, full functional model: see cybernetic model functor: 20 generalized lexeme: see lexeme,

INDEX generalized generation: 13 gei-passive: 250,320 given ( = old) vs. new: 22 global word order rules: 43,180 governed conjunction: see conjunction, governed governed preposition: see preposition, governed government: 83 government pattern: 94,226 grammatical categories: 168ff. grammatical representation (of a lexemic form): see representation, grammatical grammatical subject(s): 91, 106, 116 -, conjoined: 91-92 grammeme: 37, 168ff. graph: 19 - , directed: 19, 55 grouping (in dependency trees): 66, 72-75, 465ff. headless relative: see free relative heavy phrases: 181,227,391 homonymy: 16 idiom (see also phraseme): 25,36 intermediate levels (of linguistic representation): 16 inversion: 194, 322-323 it, anticipatory: 125, 132, 342ff. it, clefting: 125, 345ff. Kunze property: 70 labeled relation: 55 language: 12-13 language acquisition and development: 14 levels of utterance representation: 16-17

523

lex: 167 lexeme: 22,60-61.76 - , actual: 31 - , empty: 22 - , fictive: 25,29 - , full: 22 - , generalized: 22 lexical function: 25-26,40 lexical means (of linguistic expression): see means of linguistic expression, lexical lexical paraphrasing rule: see paraphrasing rule, lexical LF: see lexical function linear order (of linguistic items) : 29, 31, 49 logical connective: 20 main verb: 91 meaning-bearing vs. syntactically-controlled morphological categories: 59-60,168 Meaning-Text Model/Theory: 12 means of linguistic expression: 52 - , lexical: 52 - , non-lexical: 52 -,syntactic: 53 modifier: 134ff. - , non-restrictive: see qualifying - , qualifying: 137 - , restrictive: 137 moods (in English): 169, 175 morphological category: 59 morphological dependency: 62 morphological characteristics: 58, 212 morphological value: 25,36-37 morphological variable: 171ff. MTM: see Meaning-Text Model

524

SURFACE SYNTAX OF ENGLISH

MTT: see Meaning-Text Theory names (of classes) of objects: 20 NAP (Neutral Assertive Prosody): 29 negative adverb: 119, 308 network: 19 node: 19 - , terminal: 49 - , non-terminal: 49 non-lexical means (of linguistic expression): see means of linguistic expression, non-lexical non-projectivity: see projectivity non-restrictive attribute: see attribute, non-restrictive non-restrictive modifier: see modifier, non-restrictive non-saturation: 81,447,463-464 -, obligatory: 81-82,398 -, permitted: 82 non-terminal node: see node, nonterminal numeral, compound: 123ff., 327ff. obligatory valence: see valence, obligatory order of premodifiers: 43,180,388 ordering of rules (in a linguistic model): 7,35 paired conjunctions: see conjunctions, paired paired prepositions: see prepositions, paired paraphrasing rule: 38-39 - , lexical: 38 - , syntactic: 39 paraphrasing system: 38 parentheticals: 297-298 part of the sentence: see element of

the sentence passive vs. "copula + adjective" phrase: 253,321 passive valence: see valence, passive phrase-structure systems vs. dependency systems: 49-50 phrasal verb: 121ff., 325ff. phraseme: 57 possessive marker -s: 73, 174ff., 371372 predicate (in logical sense): 20 predicative SSyntRel: 91ff. premodifiers with right dependents: 386 preposition(s) - , correlative: see paired -, dangling: 124, 130, 365 -, governed: 34 -, paired: 332 -, stranded: 124,334 privileged status of the grammatical subject: see grammatical subject projectivity: 181, 183-186, 387-388 prosody: 29,69,179 pseudo-passive: 124,333ff. punctuation: 42, 179, 182, 410-411, 464 qualifying ( = non-restrictive) modifier: see modifier, qualifying "qualifying genitive" (=possessive form): 406 quantifiers: 20 - , floating: 20,71,105,266 quantitative/quantity phrase: 487ff. -R: representation reduced DMorphR: see DMorphR of a wordform, reduced Rel: relation

INDEX

related SSynt-roles: 154-155 relation: 49 repeatability (of SSyntRel's): 71 representation (in linguistics): 6,15 -, (of a lexemic form), grammatical: 167, 170 restrictive attribute: see attribute, restrictive restrictive modifier: see modifier, restrictive rheme: 21 right-shifted preposition: see preposition, dangling root (of a syntactic tree): see top node rules in the MTM, their nature and form: 35 cal R: see rheme ­­­ surface -S: structure saturation: 81 secondary predicate: 102-106 semanteme: 19 semantic component (of the MTM): 35ff. semantic primitives: 20 semantic representation: 18ff. semantic rules: 36ff. semantic structure: 19 semantic-communicative structure: 21 seme: 19 set phrase: see phraseme signifiant -, of the lexeme: 61 -, of the syntagm: 186 signifié -, of the lexeme: 61

525

-, of the syntagm: 186 single participial postmodifier:392-393 split infinitive: 369 "split word": 121ff. standard function: 193 standard subtree: 192, 485ff. static approach (to linguistic description): 6-7 stranded preposition: see preposition, stranded string grammar: 3 (strongly) governed prepositions/conjunctions: see prepositions/conjunctions, governed structure: 17 subject, grammatical: see grammatisubject subordinate conjunctions in coordina­ tion: 154 subordination (vs. coordination): 64ff. superlatives (their SyntS): 142-143 surface-syntactic component (of the MTM): 42-43, 179 surface-syntactic relation: 8,31,6163 -, adverbial: 87 -, completive: 98-99 -, coordinate: 87 -, phrasal: 87 -, sentential: 87 -, subordinate: 87 -, valence-controlled: 87 surface-syntactic relation, its non-universality: 67 surface-syntactic representation: 30, 46 surface-syntactic structure: 31,46,

526

SURFACE SYNTAX OF ENGLISH

48ff.,53 surface syntactic-anaphoric structure: 47 surface syntactic-communicative structure: 46 surface syntactic-prosodic structure: 46 synonymy: 16-18 syntactic features: 471 syntactic means (of linguistic exprEssion): see means of linguistic expression, syntactic syntactic paraphrasing rule: see paraphrasing rule, syntactic syntactic relations: 26,54-55 - , actant: 26 - , attributive: 27 - , their antisymmetry: 54 - , their antitransitivity: 54-55 syntactics of the lexeme: 61 syntagM: 43, 164, 186ff. - , compound: 195 syntagmeme: 198 terminal node: see node, terminal

theme: 21 there, anticipatory: 93, 225, 348ff. top node ( =root) (of a syntactic tree): 56 topic: see theme transformations: 7 translation, its role in language: 6 tree (in syntax): 55-57 1: see theme unattached (=unrelated) participle: 109, 280 valence slot: 94 valence: 69 - , syntactic: 69 - , active: 80 - , obligatory: 81 - , passive: 69,80 verbless comment clause: 294 vertex ( = node) (of a graph): 19 vocable: 76 wordform: 34 word order: 29, 43, 52, 166, 180ff. zero wordforms: 57