Crucial Readings in Functional Grammar 9783110901191, 9783110176407

Citation preview

Crucial Readings in Functional Grammar

W G DE

Functional Grammar Series 26

Editors

Casper de Groot J. Lachlan Mackenzie

Mouton de Gruyter Berlin · New York

Crucial Readings in Functional Grammar

edited by

Matthew P. Anstey J. Lachlan Mackenzie

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication

Data

Crucial readings in functional grammar / edited by Matthew P. Anstey, J. Lachlan Mackenzie. p. cm. — (Functional grammar series ; 26) Includes bibliographical references and index. ISBN 3-11-017640-8 (alk. paper) - ISBN 3-11-017641-6 pbk. : alk. paper) 1. Functionalism (Linguistics) I. Anstey, Matthew P. II. Makkenzie, J. Lachlan. III. Series. P147.C78 2005 40Γ.Γ8—dc22 2004023126

Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .

ISBN 3-11-017640-8 he. ISBN 3-11-017641-6 pb. © Copyright 2005 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany.

Table of contents

Introduction Matthew P. Anstey and J. Lachlan Mackenzie Layers and operators in Functional Grammar Kees Hengeveld Further Reading Toward a unified analysis of terms and predications Jan Rijkhoff Further Reading Parts of speech Kees Hengeveld Further Reading Predicates as referring expressions M. Evelien Keizer

vii

ι 41

47 75

79 107

109

Further Reading

137

Places and things J. Lachlan Mackenzie

141

Further Reading

166

The hierarchical structure of the clause and the typology of adverbial satellites Simon C. Oik, Kees Hengeveld, Elseline Vester, and Co Vet Further Reading Semantic content and linguistic structure in Functional Grammar. On the semantics of 'nounhood' Peter Harder. Further Reading

169 216

219 246

vi

Table of contents

On assigning pragmatic functions in English J. Lachlan Mackenzie and M. Evelien Keizer Further Reading The utterance as unit of description: implications for Functional Grammar Mike Hannay Further Reading The multilayered structure of the utterance CoVet Further Reading Functional grammar and lexematics in lexicography Leocadio Martin Mingorance Further Reading Appendix - Further references Index of languages Index of names Index of subjects

249 276

281 296

299 323

325 344 349 365 367 373

Introduction Matthew P. Anstey and J. Lachlan Mackenzie The purpose of this book is to provide essential resources for the study of Functional Grammar (FG) and its newest version, Functional Discourse Grammar (FDG). It is designed for postgraduate students and prospective researchers and seeks to add a crucial component to the existing standard literature in this field. A complete introduction to F ( D ) G can be gained by reading this book together with the following central texts: 1.

Dik (1989), the major statement of the theory of Functional Grammar; this book was republished posthumously - Simon Dik died in 1995 - in a slightly updated form as Dik (1997a).

2.

Dik (1997b), the companion volume to Dik (1997a), containing many applica-

3.

Mackenzie and Gomez-Gonzalez (2004), in particular Hengeveld's (2004a,

tions of the theory. 2004b) presentation of FDG and Anstey's (2004) historical overview, which follows the development of FG into FDG. 4.

Hengeveld and Mackenzie (fc), the major book-length presentation of FDG. These four central texts are best followed up by four additional texts: 1

5.

Relevant sections of Butler (2003), a two-volume comparative study of FG,

6.

De Groot and Hengeveld (2005), with many examples of the implementation

7.

Mackenzie and Gomez-Gonzalez (fc), a collection of FDG articles.

8.

Siewierska (1991), a sympathetic but critical presentation of FG for the interested linguist.

Systemic Functional Grammar, and Role & Reference Grammar. of FDG in a wide range of typologically disparate languages.

It is our strong conviction that contemporary debates in any linguistic theory are only properly comprehended in the context of the history that lies behind each of the positions held in those debates. In particular, the emergence of F D G cannot be understood without a thorough knowledge of the rivalling positions of those F G researchers in the nineties who turned their attention to the relations both between grammar and discourse and between semantics and interpersonal pragmatics. The selection of articles, all but one

viii

Matthew P. Anstey and J. Lachlan Mackenzie

of which first appeared in the last decade of the twentieth century, has been made with this in mind. Each article chosen for inclusion can be seen as the starting-point - either chronologically or in terms of prominence in the F G community - for significant developments in Functional Grammar. The first article, for example, introduced the concept of layering into FG, a development which enormously increased the descriptive and explanatory power of the model. Layering was immediately adopted by Dik (1989) and subsequently by most researchers in the field. Analogously, Hannay's article on discourse moves (ch. 10) had a catalytic effect upon the emergence of Functional Discourse Grammar. Similar claims can be made for each of the articles in this book. After each article, there is a brief text by the editors of this volume that lists and summarizes the major publications sparked off by the article in question. The purpose in each case is to assess the article's effects upon the debate and to give an impression of the discussion by providing a bibliographical pathway from the date of appearance of the article to the present. These texts can thus be used as suggestions for further readings. Of course it is impossible for texts of such brevity to do any justice to the publications mentioned. The aim is merely to whet the reader's appetite. In addition, much more work has been omitted than could be included; to some extent, the bibliographies of the works recommended will point the reader to further relevant references. A s a further resource, this book also contains an Appendix with a list of books and doctoral dissertations that employ or discuss the framework of FG. 2 Further bibliographical assistance is available on the Functional Grammar website, at www.functionalgrammar.com. where there is a searchable F(D)G bibliography. This site also offers access to information about the Functional Grammar Series in which this book is appearing and the electronically accessible Web Papers in Functional Grammar (WPFG, previously known as Working Papers in Functional Grammar). In addition, there is access to a discussion list and information about courses, conferences, and so forth. The major geographical centre of F G is in Amsterdam. Simon Dik was Professor of Linguistics at the University of Amsterdam from 1967 to 1992, and has been succeeded by Kees Hengeveld. This university continues to be the powerhouse of FG, with active researchers at the Vrije Universiteit Amsterdam and the University of Groningen. There are also major centres of F G in Spain (Cordoba, Granada, La Laguna, La Rioja, Oviedo, UNED), Denmark (Copenhagen and Arhus), Belgium (Antwerp), Morocco (Rabat), and Brazil (Säo Jose do Rio Preto) as well as many individual researchers

Introduction

ix

across the globe. The international nature of F G is reflected in the biennial International Conferences on Functional Grammar (ICFG). Functional Grammar received its first major presentation in Dik (1978). It was to some extent an eclectic model, combining the explicitness and generativity of Chomskyan grammar, the semantics of Case Grammar, the syntax of Relational Grammar and (implicitly) the pragmatics of the Prague School (Anstey 2004). Yet Dik succeeded in binding the various elements together into a coherent theory that seeks to achieve adequacy in three domains: (i) consistency with theories of social interaction; (ii) compatibility with the major findings of psycholinguistics; and (iii) applicability to languages of all types. Work in the eighties was concerned above all with developing F G into a full-fledged theory, with a stream of publications from a growing number of researchers on such matters as word formation, the impact of pragmatics on constituent order, the syntax of the noun phrase, and the provision of F G accounts of hitherto undescribed languages. At the same time, there was ongoing reflection on the foundations of the model as well as many applications in language teaching, business communication, knowledge engineering, and so forth. Dik himself, together with several other researchers, became increasingly interested in the computer implementation of the model as an objective way of testing the generativity of the rules. Many of the results of the work done in the eighties were incorporated into Dik (1989), which retains the overall structure of the original model, but with sophistications at almost every point. The most radical innovation was the introduction of layering: whereas the underlying structures of 1978 had been monoplanar, work by Vet (1986) and Hengeveld (this vol. [1989]) had argued that interpersonal and ideational aspects of meaning should be represented separately, and that each of these is internally layered. The resultant structure in Dik (1989) shows four nested layers, the highest representing the Speech Act (an interpersonal feature) and the lowest the predicate (an ideational feature). Layering was to become the central issue of the nineties and is the focus of chapters 1 , 2, 4, 6, and 1 0 of this book. Chapter 2, for instance, shows how layering can be applied with profit to the analysis of the noun phrase. Although the concept of layering was generally accepted, many difficulties remained, both with implementing the layered structure in the analysis of utterances and with constraining the application of such a powerful mechanism; the suggested further readings will give a good impression of the debates that ensued.

χ

Matthew P. Anstey and J. Lachlan Mackenzie Another prominent aspect of work in the nineties was the increasing

alignment of F G with linguistic typology, with F G offering a framework that is particularly suitable for cross-linguistic generalizations (see chs. 1 - 3 ) . Furthermore, whereas F G as originally conceived recognized only verbs, nouns and adjectives, the nineties saw work on prepositions, adverbs, particles, etc. (ch. 5), and the development of research into the lexicon, which had lain fallow in the preceding decade (ch. 11). The growing maturity of the model inspired philosophical reflections on the theory, such as those presented in chapter 7. Perhaps the most characteristic property of F G in the nineties was a growing interest in discourse as researchers dissatisfied with the analysis of isolated sentences came under the influence of ever more persuasive functionalist work on the interface between grammar and discourse. More use was being made of corpora in doctoral research, and discourse became the theme of successive F G conferences and the resultant proceedings (e.g. Hannay and Bolkestein 1998). F G practitioners were divided between those w h o wished to extend the principles of layering upwards from the predication into discourse and those w h o insisted on the incommensurability of grammar and discourse and thus preferred a modular approach. These issues are central to chapters 8, 9, and 10. The major event of the nineties was the sad and untimely death of Simon Dik in 1995. Ill health had forced him to retire from active work in 1992, but in the remaining years of his life he worked with Kees Hengeveld on the revision of Dik (1989) as Dik (1997a) and on the production of Dik (1997b). These posthumous works have remained the standard texts on F G and contain many ideas that continue to influence research to this day. The lively discussions on the discourse-grammar interface in the nineties led in 2000, at the ninth I C F G , to a presentation of Functional Discourse Grammar ( F D G ) by Kees Hengeveld in an article later to be included (as Hengeveld 2004a) in Mackenzie and Gomez-Gonzalez (2004). This model is shown by Anstey (2004) to address the same issues as Dik's F G (and in his view has many of the same difficulties). However, F D G differs from F G in (i) starting with the speaker's intention and working through to articulation; (ii) analyzing units both larger and smaller than a single sentence; (iii) linking the grammar to three non-grammatical components dealing with conceptualization, context and output; and (iv) distinguishing four levels of analysis: the interpersonal, representational, structural and phonological levels, each internally layered. F D G (to be given full presentation in Hengeveld and

Introduction

xi

Mackenzie fc) in this way seeks respectively to (i) satisfy the requirement of psychological adequacy; (ii) integrate grammar with discourse analysis; (iii) reflect the impact of general cognitive factors on grammatical structure; and (iv) preserve layering while recognizing the different modes of organization of the linguistic message. In other words, F D G hopes to provide answers to many of the issues first raised in the nineties. A newcomer to F G may be surprised to discover, in comparison to other linguistic theories, such as Generative Grammar, Lexical Functional Grammar, Role and Reference Grammar, and so forth, that relatively little attention has been given to traditional areas such as phonology and morphosyntax, a trend that is changing (de Groot and Hengeveld 2005). However, w e hope that Crucial

Readings

in Functional

Grammar

as a

whole demonstrates that the considerable attention given in F G to semantic and pragmatic issues such as layering, speech acts, the interpretation of the model, and the like, leads to three conclusions: (i) grammatical models necessarily require discourse and pragmatic sophistication to account for language data; (ii) semantic and pragmatic factors to a very significant degree determine morphosyntactic expression; and (iii) the interpretation of one's grammatical model qua model influences both the development and application of the model, and the theory that gives rise to it. These conclusions are by no means original: in fact, they are taken for granted by almost all functional linguists. What Crucial Readings presents, therefore, testifies to the specific contribution that F G makes to such conclusions and provides a body of evidence to be reckoned with for those who deny them. 3

Notes ι.

For an excellent recent introduction in Spanish see Garcia Velasco (2003).

2.

Where the influence of FG on a book or dissertation is somewhat implicit (e.g. see in appendix van Staden 2000; Goedegebuure 2003) or the author is an established FG researcher writing in other functional frameworks that bear on discussions in FG (e.g. see in appendix Cornish 1999; Fortescue 2001), we have preferred to include such works. We have erred on the side of inclusivity and have striven to be exhaustive, so we apologize for any accidental omissions. We have, however, excluded the large number of FG masters' theses, many of which are worth consulting. The best places to search for these are the FG website, university library catalogues, and thesis/dissertation indexing services.

xii 3.

Matthew P. Anstey and J. Lachlan Mackenzie The particular focus of this book has meant that many major contributions to FG have been omitted. As a result, the book has had to remain silent on issues such as the expression rules of FG, modality in FG, the growing impact of corpus linguistics, the relation between FG and other models, and on applications of the theory in computational linguistics. The bibliographical resources mentioned throughout the book provide access to that literature. We wish to thank Niels Smit M A (doctoral researcher at the University of Amsterdam) for his invaluable assistance with the initial assembly and editing of the book. Lachlan Mackenzie acknowledges the financial support of the Spanish Ministry of Science and Technology (Ministerio de Ciencia y Tecnologia, M C Y T ) , the European Regional Development Fund (Fondo Europeo de Desarrollo Regional, FEDER) and the Xunta de Galicia (XUGA) under project BFF2002-02441, PGIDIT03PXIC20403PN, entitled "Discourse Analysis in English: Aspects of Cognition, Typology and Second Language Acquisition/ Anälisis del discurso en lengua inglesa: Aspectos cognitivos, contrastivos y de adquisicion". Matthew Anstey, academic associate of Charles Sturt University, School of Theology, Canberra, Australia, would like to thank the United Bible Societies, Reading, England, for their financial support.

References Anstey, Matthew P. 2004

Functional Grammar from its inception. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 23-71. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

Butler, Christopher S. 2003a

Structure and Function: A Guide to Three Major Structural-Functional Theories. Part 1: Approaches

to the Simplex Clause. (Studies in

Language Companion Series 63.) Amsterdam: John Benjamins. 2003b

Structure and Function: A Guide to Three Major Structural-Functional Theories. Part 2: From Clause to Discourse and Beyond. (Studies in Language Companion Series 64.) Amsterdam: John Benjamins.

Dik, Simon C. 1978

Functional Grammar. (North-Holland Linguistic Series 37.) Amsterdam: North-Holland.

Introduction 1989

xiii

The Theory of Functional Grammar. Part i: The Structure of the Clause. (Functional Grammar Series 9.) Dordrecht: Foris. 1997a The Theory of Functional Grammar. Part 1: The Structure of the Clause. 2nd rev. ed. Hengeveld, Kees (ed.). (Functional Grammar Series 20). Berlin/New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part 2: Complex and Derived Constructions. 2nd rev. ed. Hengeveld, Kees (ed.). (Functional Grammar Series 21). Berlin/New York: Mouton de Gruyter. Garcia Velasco, Daniel 2003 Funcionalismo y Lingiiistica: La Gramätica Funcional de S. C. Oik. Oviedo: Servicio de Publicaciones de la Universidad de Oviedo. de Groot, Casper, and Kees Hengeveld (eds.) 2005 Morphosyntactic Expression in Functional Grammar (Functional Grammar Series 27). Berlin/New York: Mouton de Gruyter. Hannay, Mike, and A. Machtelt Bolkestein (eds.) 1998 Functional Grammar and Verbal Interaction. (Studies in Language Companion Series 44.) Amsterdam: John Benjamins. Hengeveld, Kees 2004a The architecture of a Functional Discourse Grammar. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 1 - 2 1 . (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter. 2004b Epilogue. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 365-378. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter. this vol. Reprint. Layers and operators. Journal of Linguistics 25: 1 2 7 - 1 5 7 , 1989. Hengeveld, Kees, and J. Lachlan Mackenzie fc Functional Discourse Grammar. Oxford: Oxford University Press. Mackenzie, J. Lachlan, and Maria de los Angeles Gomez-Gonzälez (eds.) 2004 A New Architecture for Functional Grammar. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter. fc Studies in Functional Discourse Grammar. Berne: Peter Lang. Siewierska, Anna 1991 Functional Grammar. London: Routledge.

xiv

Matthew P. Anstey and J. Lachlan Mackenzie

Vet, Co 1986

A pragmatic approach to tense in Functional Grammar. Working Papers in Functional Grammar 16.

Layers and operators in Functional Grammar Kees

i.

Hengeveld

Introduction1

I have argued elsewhere (Hengeveld 1987b) that for a proper treatment of modality the clause model used in Functional Grammar (Dik 1978, 1980) should be adapted in such a way that a number of different layers can be distinguished. M y main argument there was that predications, used in Functional Grammar to represent linguistic expressions, have two different functions: a descriptive

function and a content function. A predication

not only gives a description of the external situation the speaker refers to within his speech act, it also represents the propositional content or message unit processed within that speech act. Subjective and evidential modalities, which express a propositional attitude, should take a predication in its content-representing function in their scope, whereas objective modalities, i.e. those modalities that are concerned with the actuality status of a State of Affairs (SoA), should take a predication in its SoA-designating function in their scope. A clause model should therefore be able to distinguish between these two functions of predications. In this article I want to refine the model in such a way that other categories of operators can be included and then proceed to explore the possibilities of the resulting structure. In doing so I will touch upon a number of related issues which have received much attention lately: 1.

Restrictions on the ordering of Tense, Mood, and Aspect (TMA) morphemes (Bybee 1985, Foley and Van Valin 1984).

2.

The representation of sentences in Functional Grammar, in relation to the treatment of Tense (Vet 1986).

3.

Restrictions on the selection and expression of TMA-morphemes in subordinate constructions, in relation to the degree of sententiality of the subordinate construction (Lehmann 1988).

4.

The representation of non-restrictive constructions and predication-combining in Functional Grammar (Hannay and Vester 1987).

2

Kees Hengeveld

In section 21 present a clause model which shares its layered structure with that proposed by Foley and Van Valin (1984), and uses the format proposed in Vet (1986) for the representation of individual layers. In section 3 I give a classification of operators according to their relative scope by associating them with the different layers distinguished within the clause model. Section 4 explores the implications of the model for the classification and treatment of subordinate constructions.

2.

A two-level analysis of the clause

The basic idea behind the clause model presented here is that every utterance can be analyzed at two levels: the representational (Bühler 1934) and the interpersonal (Halliday 1970) level. At the representational level a So A is described in such a way that the addressee is able to understand what real or hypothesized situation is referred to. At the interpersonal level this situation is presented in such a way that the addressee is able to recognize the communicative intention of the speaker. Thus the representational level is concerned with the narrated event, the interpersonal level with the speech event (see Jakobson 1971).

2.1.

The representational level

For the representation of narrated events I use Vet's (1986) proposal concerning the representation of sentences. Vet argues that sentences, like terms, are referring expressions. The entities they refer to are SoAs, which take place in some time-space region. To arrive at a unified account of terms and sentences, he proposes a general schema for the representation of sentences, which is exemplified in (1): (i)

{&•·. [Writev (x,: JohnN (Xi))Ag (xj: bookN (Xj))Go] (e,))

The predication, between square brackets, conforms to FG-conventions. It is built on the basis of a predicate frame, which contains a number of argument positions (xn Xj), each provided with a semantic function (Agent, Goal2), and a predicate which specifies the relation between these arguments. In the argument positions of the predicate frame terms, expressions with referential

Layers and operators

3

potential, are inserted. The predication as a whole is presented by Vet as a restrictor of the sentence variable e, symbolizing some time-space region. What is relevant here is that I interpret this variable as a variable representing the narrated event. I will return to some details and advantages of Vet's analysis in the sections on operators and satellites.

2.2.

The interpersonal level

Speech events may be analyzed in an analogous way: the participants in a speech event are the speaker, the addressee, and the transmitted content or message. The relation between these three participants is expressed by an abstract illocutionary frame, which represents the basic illocution of the linguistic expression. An example of an illocutionary frame with its paraphrase, based on Dik (1989), is given in (2): (2)

D E C L (S) (A) (X,) Speaker (S) wishes the Addressee (A) to add the content (X,) to his pragmatic information

The analysis presented here, which can be considered an F G equivalent of Ross' (1970) performative analysis, leads to the general schema for the representation of speech events given in (3), where the speech event variable E, symbolizing the current time-space region, is restricted by a clause designating the speech act which occupies this time-space region: (3)

(E i : [ILL (S) (A) (X,: [proposition] (X,))] (E,))

Here an abstract illocutionary frame (ILL) specifies the relation between the speaker (S), the addressee (A), and the content of the utterance (X,).

2.3.

The clause

The representations of narrated event and speech event may be combined in a single representation of the utterance, as in (4):

4

Kees Hengeveld

(4)

The representation of utterances clause (E,: [ILL (S) (A) (X, : [proposition] (X,))] (E,) (e,: [Predß (x,) (x 2 )... (x„)] (e,)) predication

In this analysis of the clause a predication fulfils two different functions: it designates a S o A at the representational level, and it represents the content of a speech act at the interpersonal level. To distinguish these two uses of predications I use the term 'predication' to refer to the former and 'proposition' to refer to the latter function. Starting from the innermost layer, the predication, the functions of the different layers distinguished in (4) can be understood in the following way. A predication gives a description of a set of possible S o A s . B y inserting a predication into a narrated event slot it becomes a referring expression. The entity it refers to is the real or hypothesized situation the speaker has in mind. B y inserting a fully specified predication into the proposition slot of an illocutionary frame it becomes an expression referring to the information unit or content transmitted in some speech act. The illocutionary frame contains instructions for the addressee about what the speaker wants him to do with this information unit. B y inserting a clause into a speech event slot it becomes an actual speech act or utterance-token (Lyons 1977: 35). The representation of a simple utterance thus contains four different variables, restricted by different linguistic units, resulting in expressions referring to different kinds of entity: (5) Var

Variables and restrictors Restrictor

Clause unit

Reference

X

PredN

(x,: PredN (x,))

Individual

e

Predication

(ei: [Predication] (e,))

State of Affairs

X

Proposition

(X,: [Proposition] (X,))

Ε

Clause

(E,: [Clause] (E,))

Potential Fact Speech Act

Layers and operators

5

The analysis therefore provides variables for first (x), second (e), third (X), and fourth (E) order entities. The E-variable is different from the other three kinds of variable in that it refers to the speech event itself rather than to one of the entities to which reference is made within that speech event. Ε-variables are created during the process of speaking. A speaker and his addressee(s) make use of them later when remembering or referring to a particular utterance. The Ε-variable can be seen as containing information about the time and place of, and the participants in the speech event. As such it provides clues as to the reference of what Jakobson ( 1 9 7 1 ) called shifters: grammatical units of which the general meaning cannot be defined without making reference to the speech event within which they are used, such as personal and demonstrative pronouns, and absolute tenses. The E-variable provides the deictic centre (see Comrie 1985: 36) on the basis of which the reference of these elements is determined.3 It will be of some use in defining tense distinctions (see 3.3).

3. 3.1.

Operators General outline

Leaving aside term operators, the clause model given in (4) provides four positions for operators, as indicated in (6): (6)

Operators (positions) (Ε,: [π4 ILL (S) (Α) (π3 X,: [proposition] (X,))] (E,) (π2 e,: [π, Predß (χ,) (x 2 )... (x„)] (e,)) π,: predicate operators π2: predication operators

π3: proposition operators π4: illocution operators

Definitions for these four classes of operators are given in (7), and a tentative classification of operators in terms of this fourfold distinction is given in

(8):

6 (7)

Kees Hengeveld Operators (definitions) a. Predicate operators capture the grammatical means which specify additional properties of the set of SoAs designated by a bare predication. b.

Predication operators capture the grammatical means which locate the SoAs designated by a predication in a real or imaginary world and thus restrict the set of potential referents of the predication to the external situation(s) the speaker has in mind.

c.

Proposition operators capture the grammatical means through which the speaker specifies his attitude towards the (truth of the) proposition he puts forward for consideration. Illocution operators capture the grammatical means through which the speaker modifies the force of the basic illocution of a linguistic expression so as to make it fit his communicative strategy.

d.

(8) Operators (classification) Semantic domain Predicate Internal temporal constituency Presence or absence of property or relation expressed by predicate Time of occurrence Frequency of occurrence Actuality of occurrence Source of proposition Commitment to proposition Weakening strategy Strengthening strategy

Grammatical category operators Imperfect/Perfective, Phasal aspect Predicate negation

Predication operators Tense Quantificational aspect Objective mood/Polarity Proposition operators Evidential mood Subjective mood Illocution operators Mitigating mode Reinforcing mode

The crucial difference between predicate operators and predication operators is that predication operators are concerned with the occurrence or setting of a S o A rather than with its properties. Thus predicate operators are related to the descriptive function of predications, whereas predication operators are related to the referring function of predications.

Layers and operators

7

The crucial difference between predication operators and proposition operators is that proposition operators are concerned with the attitude of the speaker towards the content of his speech act, rather than with the occurrence of the event to which reference is made within that speech act. Thus proposition operators are related to the content function of propositions. The crucial difference between proposition operators and illocution operators is that illocution operators are concerned with the possible perlocutionary effects of a speech act, rather than with its content. Thus illocution operators are related to the communicative intention with which the speaker presents this content. The general functions of the four operator types can now be defined as in (9): (9)

πι: Modification of the internal structure of the SoA π2: Qualification of the S o A as a whole π3: Qualification of the proposition τΐ4: Modification of the basic illocution

In what follows I give some examples of how distinctions in the field of Tense, Mood, Aspect, Polarity and Illocution can be interpreted in terms of this fourfold classification. It is not my intention to provide a full-fledged account of the typology of T M A systems, but rather to adduce some evidence for the distinction between four types of operator as it is made here.

3.2.

Operators at the representational level

The distinction between predicate operators and predication operators is motivated by the fact that at the representational level two functions should be fulfilled: first of all, the speaker should create a proper description of the situation he wishes to refer to; secondly, he should relate this description to the situation he has in mind. I will refer to these two functions of the predication as its predicating and referring functions. The same distinction may be applied to terms: within the predication they are referring expressions, but their internal make-up consists of the application of predicates to a term variable. When terms are inserted into argument slots they become expressions referring to individuals existing in some world. When predications are inserted into SoA slots they become expressions referring to

8

Kees Hengeveld

situations obtaining in some world. This correspondence between terms and predications is stressed in Vet (1986). The following figure tries to capture these two functions of predications and terms: (10)

Predicating and referring functions of predications and terms predication term (e,: [Predß (x,:Predß (χ,)) etc.] (e,)) I I predicating referring predicating referring

Predicate operators fulfil a function in building up a proper description of the situation the speaker wishes to refer to: without affecting the argument structure of a predicate frame, predicate operators specify properties of the S o A s designated by bare predications which are of sufficient generality to be coded grammatically rather than lexically within a given language. Predication operators are related to the referring function of predications: without affecting the properties of the S o A s designated by a predication they relate the description of a S o A to the occurrence of that S o A in a real or imaginary world.

3.2.1.

Predicate

operators

Given the property-assigning function of predicate operators, to make a grammatical category qualify for predicate operator status it should be such that the characteristics of the set of S o A s designated by a predication in which the grammatical category is applied are different from the characteristics of the set of S o A s designated by the same predication in which the

Layers and operators

9

predicate operator is not applied. The general properties expressed through predicate operators may include: 1. 2.

The 'internal temporal constituency' (Comrie 1976) of the situation referred to. The presence or absence of the relation or property expressed by the predicate.

Many, though not all, aspectual distinctions seem to express properties of the first type. This can most easily be demonstrated by looking at the typology of SoAs. What is generally called 'Aktionsart' is handled in FG in terms of a typology of SoAs (see Dik 1978; 1989; Vester 1983; de Groot 1985). The main parameters in the typology of SoAs are: [± dynamic], [± control], [± telic], [± momentaneous]. In different combinations these parameters define most of the SoA-types. Many aspectual distinctions are 'feature changing', in the sense that they change the value of one of the aforementioned parameters. For instance, the English Perfect, Progressive and Prospective turn a [+ dynamic] SoA into a [ - dynamic] one. Steedman (1977: 221, see also Goossens 1985) demonstrates that English sentences involving these aspectual categories are similar to states, since they cannot occur in pseudo-cleft constructions of the type given in (11): (11)

*What he did was be running

In a similar way, the perfective/imperfective distinction may affect the momentaneousness of a SoA, as in (12), where the imperfective value of the progressive cancels the momentaneousness of reach (see Comrie 1976: 43): (12)

a. b.

The soldiers reached the summit The soldiers were reaching the summit

[+Mom] [ - Mom]

I take it that this feature-changing property of many aspectual categories indicates that they can be analyzed as operating SoA-internally and should be taken into account in a typology of SoAs. This view is supported by the fact that dynamicness and momentaneousness are 'inherent' features of predicates, as opposed to, for instance, Control and Telicity, which may depend on features of arguments and satellites (see de Groot 1985: 75).

10

Kees Hengeveld

Apart from aspectual distinctions, negative polarity can be analyzed as operating on the predicate in sentences like the following (see Vet 1986): (13)

Charles is not unintelligent

On one interpretation of this sentence, it is claimed that it is the case that Charles has the property 'not-unintelligent'. On another, it is claimed that it is not the case that Charles has the property 'unintelligent'. Under the first interpretation the negative element can be seen as the expression of a predicate operator, under the second as the expression of a predication operator (see 3.2.2). It is a general property of the categories discussed here that they present borderline cases on the derivational-inflectional continuum (see Bybee 1985), in the sense that languages may differ with respect to the derivational or inflectional realization of these categories. A case in point is the perfective/ imperfective distinction in the Slavic languages, which seems to be analyzable as partly derivational, partly inflectional. Predicate negation in English, as illustrated in (14), is another illustration. Consider: (14)

Charles is not intelligent

This sentence imposes the 'it is not the case that...' interpretation, given the availability of (15): (15)

Charles is unintelligent

The use of not as the expression of a predicate operator remedies the ungrammaticality of *ununintelligent, but would be quite exceptional or unacceptable in other cases.

3.2.2.

Predication operators

The kinds of distinctions to be expressed through predication operators have been characterized as being concerned with the occurrence of a SoA rather than with its properties. A set of SoAs designated by a predication may be limited to the situation(s) the speaker has in mind by evaluating it with respect to:

Layers and operators ι. 2. 3.

11

time of occurrence frequency of occurrence actuality of occurrence

All tense distinctions can be analyzed as operating on predications. They locate the SoA designated by a predication on the time axis relative to the speech act or to other SoAs. Vet (1986) proposes to analyze tensed predications in the following way: (16)

(Tense e^ [predication] (eO)

Following Comrie (1985) a distinction can be made between absolute, relative, and absolute-relative tense. The time of occurrence of a S o A may be evaluated relative to the time of occurrence of the speech act, in which case the variable Ε provides the 'deictic centre' (Comrie 1985: 36) for absolute time reference. Or it may be evaluated relative to the occurrence of another S o A , in which case the narrated event variable e provides the reference point for relative time reference. An adaptation of Comrie's (1985: ch. 6) formalization in line with the variables used here leads to: (17)

a. b. c.

Absolute tense: Relative tense: Absolute-relative tense:

ei relative E[ ei relative e 2 ej relative e 2 relative E,

Examples are: (18)

a. b. c.

I crossed the street Having crossed the street,... I had crossed the street, ...

e, before E, e, before ej before e3 before E,

Thus the reference points needed to give a description of Tense distinctions are directly available within the clause model used here. Apart from Tense distinctions, some aspectual categories qualify for predication operator status. The aspectual categories I have in mind give a specification of the frequency of occurrence of a S o A , such as Semelfactive and Iterative aspect. Dik (1985: 9) uses the term Quantificational Aspect as a cover term for aspectual categories like these. Rather than modifying the internal structure of a S o A , these aspectual categories specify how many

12

Kees Hengeveld

times a S o A with a given internal structure occurs (occurred, will occur). Separating the frequency of occurrence of a S o A from the internal temporal constituency of that S o A implies that both categories can in principle be specified independently of each other. Comrie (1976: 30) showed this to be the case for English. The following example from Hidatsa confirms his analysis: (19)

Hidatsa (Siouan, Matthews 1964) Wio i hirawe ki ksa c Woman she sleep INGRITER MOOD 'The woman fell asleep again and again.'

The morpheme ksa indicates that the speaker refers to a situation that occurred frequently. The morpheme ki characterizes each individual occurrence of the situations referred to as being viewed from its starting point. The different status of the two aspectual categories can be demonstrated by the happen paraphrases given in (20): (20)

a. b.

It often happened that the woman started to sleep *It started to happen that the woman often slept

To distinguish between the two types of Aspect I will use the term 'Qualificational Aspect' to refer to those aspectual distinctions which can be thought of as expressed by predicate operators, and I will continue to use the term 'Quantificational Aspect' to refer to the aspectual distinctions which can be thought of as expressed by predication operators. Example (19) could be represented as in (21): (21)

(Iter e,: [Ingr hirawe v

wio N

(Xi))0]

fe))

But Quantificational Aspect can also be interpreted as a means of SoAquantification, 4 comparable to term-quantification. Under this analysis quantificational aspects can be represented as:5 (22)

Quantificational aspect a. (lei: [predication] (e,)) b. (me,: [predication] (ei))

semelfactive iterative

Layers and operators

13

The third T M A category which can be seen as operating on predications is objective modality.6 Linguistic means giving expression to objective modal distinctions can be regarded as the output of an evaluation process on the part of the speaker with regard to the actuality status of a SoA. The knowledge on which the speaker has to base this evaluation may be subdivided into the two types given in (23): (23)

Two types of knowledge a. Knowledge of possible situations obtaining in the speaker's conception of reality or of a hypothesized universe b. Knowledge of possible situations relative to some system of moral, legal or social conventions

The speaker bases his epistemic and deontic evaluations on (a) and (b), respectively. Depending on the degree of compatibility of a SoA designated by a predication with the speaker's knowledge of type (a) or (b) the two ranges of possibility and permissibility in (24) can be established: (24)

Objective modality a. Certain Probable Possible Improbable Impossible b. Obligatory Acceptable Permissible Unacceptable Forbidden

Elements giving expression to objective modal distinctions take a SoA as designated by a predication in their scope, as they represent the output of speaker's evaluation of the probability or desirability of occurrence of that SoA in terms of his knowledge. Polarity distinctions operating on the predication, paraphrasable as 'it is the case that' and 'it is not the case that', are closely related to objective modality. They differ from the objective modal distinctions discussed so far in that they characterize a SoA as simply actual or non-actual.

3.3.

Operators at the interpersonal level

The distinction between proposition operators and illocution operators is motivated by the fact that at the interpersonal level two functions should be fulfilled: firstly, the speaker should transmit some content, which may have different sources; secondly, he should indicate what he expects the addressee

14

Kees Hengeveld

to do with this content. Proposition operators are used by the speaker to specify to what degree he feels committed to the truth of the content he transmits, and they are therefore largely restricted to declarative sentences. Illocution operators capture the grammatical means which the speaker uses to modify the illocutionary force of his utterance in view of the possible perlocutionary effects of his speech act. Without affecting the basic illocution expressed by the abstract illocutionary frame, illocution operators specify strategical modifications of this basic illocution.

3.3.1.

Proposition operators

Following Chung and Timberlake (1985: 244) I use the term epistemological modality for those linguistic means through which the speaker expresses his commitment with regard to the truth of a proposition. Two subtypes are to be distinguished: subjective modality, through which the speaker specifies the kind and degree of his commitment; and evidentials (see e.g. Chafe and Nichols 1986; Willett 1988), through which the speaker specifies how the proposition came to his knowledge. What both subtypes have in common is the relevance of the source of the information contained in a proposition. In the case of evidentials this source is characterized as different from the speaker. In the case of subjective modality the speaker is the source. This source-revealing character of subjective modality is reflected in one of the differences between subjectively and objectively modalized sentences, as illustrated in (25): (25)

a. b.

A: It is possible that it will rain tomorrow B.· Who says so? A: Perhaps it will rain tomorrow B: *Who says so?

Questioning the source of the information contained in sentence (25b) is clearly out of place, as the modal adverb indicates that the speaker is expressing his personal opinion. The different subdistinctions to be made within the epistemological modality type are:

Layers and operators

15

(26) Epistemological modality Source Modality Speaker

Subjective

Epistemic

Certainty Probability Possibility

Boulomaic Evidence

Inferential

3rd person Experience

Quotative Experiential

Wishing, hoping

The following examples are from Hidatsa, a language making extensive use of elements expressing epistemological modality: (27)

Hidatsa (Siouan, Matthews 1964) waceo

ύίχί

α

äciwi

ski

man

antelope

he

track

CERT

'The man sure tracked an antelope.' (28)

wio

a

riiti

woman

she

hungry QUOT

rahe

'I've been told that the woman is hungry.' Distinguishing between two different types of modality again implies that both can be specified independently of one another. First, consider the lexical expression of both modalities in the English example (29): (29)

It is certainly possible

that John is ill

The subjective modal adverb expresses almost the opposite of the objective modal adjective. Yet there is nothing wrong with this sentence. The point is, as Lyons (1977: 49) puts it, that "logical probability can be defined, and measured, as a property of some system of propositions in abstraction from the beliefs of the users of that system". For an inflectional realization of both categories consider the Turkish example (30): (30)

Turkish (Altaic, Lewis 1967: 127) Gel-me-meli-ymi§-siniz COme-NEG-NEC-INF-2PL

'It seems you ought not to come.'

16

Kees Hengeveld

Here the subjective inferential follows the objective necessitative, thus reflecting the ordering assumed for predication and proposition operators. This sentence can therefore be represented as in (31): (31)

3.3.2.

(Inf. X;: [(Nec e^ [Neg gel- v (dmXi: p2 (χ,))] (e,))] (X,))

Illocution operators

Two of the strategies in which illocution operators can be used are mitigation and reinforcement (see Haverkate 1 9 7 9 ) . The general function of mitigation is to reduce the force of a speech act. The goals of mitigation can be more specific: to prevent loss of face, be polite, leave room for the addressee to refuse or disagree, make the addressee feel comfortable, etc. The general function of reinforcement is to impose the speech act more strongly upon the addressee. The goals of reinforcement, too, can be more specific: to convince the addressee, express impatience, show superiority, etc. The main reason to distinguish between proposition operators and illocution operators is that whereas the former are largely restricted to declarative sentences, i.e. operate inside the illocutionary layer, the latter can be applied to sentences with all kinds of basic illocution, i.e. operate outside the illocutionary layer. The sentences in (32) show the general applicability of the reinforcing use of the Spanish subordinator que, those in (33) the general applicability of the mitigating particle a/ya in Mandarin Chinese: (32)

Spanish (Indo-European, Hengeveld a. /Que no me gusta that

b.

c.

not me

1987a)

please.PRES.iND.3sG

nada

esa

nothing that

pelicula! movie

Ί don't like that movie at all!' J Que no te marches mananal That not yourself leave.PRES.suBj.2sG tomorrow 'Don't you leave tomorrow!' ι Que si vienes mananal That whether come.PRES.iND.2SG tomorrow 'Are you coming tomorrow?!'

Layers and operators (33)

17

Mandarin Chinese (Sino-Tibetan, Li and Thompson 1 9 8 1 ) a. Wo bing mei zuo-cuo a/ya I on.the.contrary not do-wrong MIT O n the contrary, I didn't do wrong' b. Chi-fan a/ya eat-food MIT 'Eat, OK?!' c. ni xiang bu xiang ta a/ya? You think not think he MIT 'Do you miss him?'

It is difficult to give a proper translation of the Mandarin examples. In general, the mitigating particle reduces the forcefulness of the speech act. Thus (33a) is 'less belligerent', (33b) 'much more friendly', and (33c) 'much softer' than their non-mitigated counterparts (Li and Thompson 1981: ch. 7-5).

3.4.

Some hypotheses

The ordering and classification of operators proposed here leads to the formulation of a number of hypotheses. All of these hypotheses have a provisional character and cannot be fully investigated here. Nevertheless they follow from the model proposed here and present promising guidelines for future research. Hypothesis 1 seems to be compatible with the results of Bybee (1985) 7 and Foley and Van Valin (1984), which inspired its present formulation. (34)

Hypothesis 1 — The preferred order of operators is π4 π3 π2 π, Predß or Predß τι] π2 π3 π4

I have not been able to find an example from any language in which all four categories are present in one sentence, but the order also holds for subsets of operators. Consider, for instance, the following examples from Hidatsa, Diegueno, and Quechua:

18 (35)

(36)

Kees Hengeveld Hidatsa (Siouan, Matthews 1964) a. wio i hirawe ki ksa c Predß π2 π3 woman she sleep INGR ITER PROB 'The woman fell asleep again and again.' b. Wira i άράαη ki stao wareac Predß πι π2 π3 tree it grow INGR REM.PAST QUOT 'They say the tree began to grow a long time ago.' Diegueno (Yuman, Gorbet 1976) W-a:-m-x-kx Predß π2 π3 3 SG-go-away-iRR-iNF

(37)

'It must be that he will go.' Quechua (Andean, Cole 1982) pay-ka shamu-nga-m-ärP he-TOP

Predß π2 π3 ττ4

COme-FUT.3-FIRSTHAND-REINF

'He will come!' The West Greenlandic examples given in (38) further illustrate the ordering of illocution and proposition operators relative to the indicator of the basic illocution: (38)

West Greenlandic (Inuit, Fortescue 1984) a. Qama-junnarsi-vuq

(π3 ILL)

be.0ut.hunting.seals-PR0B-3SG.LND

b.

'He's probably outside hunting seals' Aki-nngil-aanga luunniit reply-NEG-3SG.ISG.INDIC

(ILL π4)

REINF

'He didn't even reply to me!' Whereas in (38a) the subjective modal affix junnarsi precedes the Indicative, which marks declarative sentences, the emphatic particle luunniit follows it in ( 3 8 b ) , thus reflecting the differences in scope between the two categories of operator. (39)

Hypothesis 2 — Diachronic developments in the field of operators tend to follow the direction π1>π2>π3>π4

Again the hypothesis seems to be compatible with the data in Bybee and those in Foley and Van Valin ( 1 9 8 4 ) . 9

(1985)

Layers and operators (40)

19

Hypothesis 3 — Operators of the class πη may impose restrictions on the selection of operators of the class πη.ι

In other words, operators may impose selection restrictions on the next operators down and these are the only restrictions possible. If proved to be correct, this generalization considerably reduces the set of possible formulations of operator combinations. The following series may illustrate the hypothesis: (41)

a. b. c.

π4 > π3 Mitigation of a declarative sentence disallows strong commitment with respect to the truth of the proposition it contains. π3 > π2 Desiderative mood requires the situation referred to to be non-actual. π2 > τΐι Objective epistemic mood requires a phasal aspect operator to be applied if the SoAs designated by a predication are non-stative.

The first restriction is exemplified by: (42)

Spanish (Indo-European) a. Seguramente es/*sea posible certainly be.3sG.iND/suBj possible 'Certainly it's possible.' b. Quizäs es/sea posible maybe be.3sG.iND/suBJ possible 'It may/might be possible.'

Mitigation is expressed in Spanish by means of the subjunctive. Mitigation requires the speaker to be less than fully committed to the truth of the proposition, hence the ungrammaticality of the version of (42a) in which the subjunctive is combined with the modal adverb seguramente. The latter two restrictions can be demonstrated by means of the sentences in (43): (43)

a. b. c.

I wish I were travelling in France */ wish I am travelling in France *I wish I travelled in France10

20

Kees Hengeveld

Although wish cannot exactly be said to be the expression of an operator, these sentences may serve to illustrate the following two steps: (i) wish requires the situation referred to to be non-actual, hence the ungrammaticality of (43b); (ii) evaluating a situation as non-actual in an epistemic sense generally requires the SoAs to be non-dynamic, hence the ungrammaticality of (43c). Goossens (1985), following Steedman (1977), notes that dynamic SoAs provided with a Progr, Perf, or Hab operator count as non-dynamic SoAs (see 3.2.1), hence the grammaticality of (43a). Thus the restrictions illustrated by the ungrammaticality of (43b-c) can be stated as in (44): (44)

DES > NON-ACT > PROGR, PERF, HAB/[+ dyn]

The third restriction can be illustrated separately by the following Turkish examples: (45)

Turkish (Altaic, Lewis 1967) a. Selimiye camisini gör-meli-sin Selimiye

b.

mosque

See-NEC-2SG

'You must see the Selimiye mosque.' Sen-i gör-mü§ ol-mali you-ACC

See-RES

be-NEC

'He must have seen you.' Example (45a) has a deontic interpretation, whereas (45b), in which the resultative morpheme is added to the verbal base, has an epistemic interpretation. From hypothesis 3 we may deduce: (46)

Hypothesis 4 — Operators are specified most economically in the order π4 > π3 > π2 > πι

Given that according to hypothesis 3 higher operators may restrict the selection of lower operators the order given in hypothesis 4 ensures a selection procedure which consistently narrows down the options available during the creation of an underlying utterance.

Layers and operators 4.

21

Subordinate constructions

Assuming the validity of the layered model of the clause and the position of operators within that model, the question arises what the implications of the model are for the treatment of subordinate constructions. As in the case of operators, processes at term level are not dealt with here. By a subordinate construction I understand a construction that for its occurrence depends on another. The word construction is used here as a cover term ranging over predications, propositions, and clauses. It is the distinction between these three types of construction that leads to the first question to be asked with respect to complex constructions: should subordinate constructions be classified as predications, as propositions or as clauses? To put the same question in other words: what is the internal structure of subordinate constructions? Section 4.1 tries to answer this question with respect to constructions occupying an argument position. Section 4.2 is complementary to section 4.1, in that it is concerned with what I provisionally call the 'external structure' of embedded constructions, in particular with respect to satellites. Here the main question is to what layer of the clause satellites should be attached. Section 4.3, finally, looks at secondary predication.

4. ι .

Complement constructions

Assuming that arguments may refer to different kinds of entity, the following hypothesis seems to follow from the approach presented so far: (47)

Hypothesis 5 — Subordinate constructions can be classified according to the highest layer they contain.

This generalization should be understood in the following way. Within a narrated event reference can be made to speech events other than the one executed, propositions other than the one put forward for consideration, and narrated events other than the one referred to in the matrix clause. So, by peeling off layers from the general model for the simple clause, one encounters all types of subordinate construction. This means that: ι.

A subordinate construction cannot contain a layer of a certain level without at the same time containing all subsequent layers.

22

Kees Hengeveld

2.

The operators associated with the layers which a subordinate construction contains can be expressed within that subordinate construction.

The matrix predicate determines what kind of construction it may dominate. The examples given in (48) illustrate these points: (48)

Ε,: [π4 ILL (S) (Α) (π3 X;: [proposition] (X,))] (E,) a. b. c.

(n2 βί: (π2ε;: (π2 e;:

[π, Say v ( X i ) A g [π, Know v (x,) 0 [π, See v (x,)0

(Ε,: (n3Xf. (π2 ej:

[π4 ILL etc.] [n2 e^ etc.] [π! Predß etc.]

(Ej)) GO ] (e,))

(X,))Go] (e,)) (ej))Go] (e,))

The uppermost formula in (48) represents the interpersonal level of the matrix clause. The formulas in (48a-c) represent several matrix clause representational levels structured on the basis of predicate frames which take constructions of decreasing complexity as their Goal arguments. The verb say in (48a) takes a clausal Goal argument (E), the verb know in (48b) takes a propositional Goal argument (X), and the verb see in (48c) takes a predicational Goal argument (e). The representation given in (48a) is intended to capture direct speech reports. It is clear that all kinds of distinctions and modifications made by the original speaker can be repeated. The interesting thing about direct speech reports is that they involve a shift in the deictic centre, and this is precisely what the presence of the Ε variable accounts for. The representation in (48b) indicates that cognitive predicates embed propositions, not predications. One therefore expects the possibility of applying proposition operators to the complements of cognitive predicates. That this is in fact possible can be illustrated by the English examples in (49): (49)

a. b.

He didn't know that John would come He didn't know whether John would come

The complementizer that in (49a) indicates that the speaker considers John's coming to be a fact, whereas the complementizer whether in (49b) does not commit the speaker to the truth of the embedded proposition. Note that it is the actual speaker who expresses his commitment with respect to the embedded proposition in (49a).

Layers and operators

23

Indirect speech reports can be captured in a similar way. Unlike direct speech reports, these do not involve a deictic centre shift. However, it is generally possible to apply proposition operators in the subordinate clause, as in the case of cognitive predicates. The examples from Jacaltec given in (50) illustrate this point: (50)

Jacaltec (Mayan, Craig 1977) a. Xal naj tato chuluj naj presidente said he that will.come the president 'He said that the president would come.' b. Xal naj chubil chuluj naj presidente said he that will.come the president 'He said that the president would come.'

The complementizer tato in (50a) indicates that the actual speaker considers the original speaker unreliable, whereas the complementizer chubil in (50b) indicates that the actual speaker considers the original speaker reliable. Again the interpretation of the operators should be related to the actual speaker, not to the subject of the matrix clause. The representation in (48c), finally, is intended to show that a verb like see, in one of its readings,11 embeds predications, not propositions. The embedded predication can be specified for tense and aspect, but not for propositional attitude. This is not surprising, since we can witness SoAs, but not facts. I have restricted myself in (48) to verbal predicates, but comparable examples could be given for adjectival and nominal predicates (see Hengeveld 1990). Adjectives like true and undeniable have propositional complements, adjectives like certain and regrettable have predicational complements. Nouns like fact have a propositional complement, nouns like possibility have a predicational complement. These differences between predicates should be represented in the lexicon, for instance in the following way: (51)

Say v (xi) Ag (EI)GO (Direct speech) Say v (xi) Ag (X,)co (Indirect speech) Knowv (xi) 0Exp (XI)GO See v (Χι)0Εχρ (ei)Go

TrueA (X,) 0 FactN (XI) 0 Possible A (e,) 0 Possibility N (e,) 0

24

Kees Hengeveld

This approach not only accounts for the fact that the operators to be expressed in a complement are determined by the type of complement, but also provides the means to account for differences in the form and behaviour of complements. The former differences are illustrated in the following examples from Nama. In Nama, subordinate constructions may take three different forms. Direct quotation is achieved by repeating the original sentence and providing it with the quote particle tv. (52)

Nama Hottentot (Khoisan, Hagman 1974; Rust 1965) Όο-s ke //'tisä //χαάρά ke mtt Ζ'ύύ-ta then-3SG DECL

'a

ti

PRES

QUOTE

she

again

REM.PAST say

not.know-isG

'She said again: " I don't know'" Indirect quotation and other forms of propositional complementation may take the form illustrated in (53)-(54), containing the complementizer Ixäisa. (53)

(54)

//'iip ke 'am'a-se kere *om /'ae//amsa he DECL true-ADV REM.PAST believe Windhoek xuü-kxm /xii häa !xäisä from-1 DU come PF that 'He really believed that we had come from Windhoek.' !üu-ta ηιί Ixäis ke 'a 'am'a go-ISG FUT that DECL be.PRES true 'It's true that I will go.'

Predicational complements, finally, take the form of a nominalization, as in (55): (55)

fgü-s ke kaise a Igomba go-NMLZTN DECL very PRES difficult 'It's very difficult for me to go.'

te to.me

In terms of the present analysis, Nama has specialized forms for Ε-complementation, X-complementation, and e-complementation. With respect to the behaviour of complements the differences between action and fact nominalizations are illustrative. 12 Sentence (56) is "structur-

Layers and operators

25

ally ambiguous between a 'factive' sense and a 'manner' sense". It can mean "either that the speaker dislikes the fact that John drives or that the speaker dislikes the way in which John drives" (Katz and Postal 1964: I23ff.; cf. also Yendler 1967). (56)

I dislike John's driving

Specification of the goal {the car) and the manner (careless(ly)) of John's driving leads to (57) under the first and to (58) under the second interpretation: (57) (58)

I dislike John's carelessly driving the car I dislike John's careless driving of the car

The realizations of the goal and manner arguments in (57M58) are interrelated, as can be derived from: (59) (60)

*/ dislike John's carelessly driving of the car *I dislike John's careless driving the car

In terms of the present analysis, English treats X-nominalizations and e-nominalizations differently. With respect to the expression of operators in subordinate constructions the hypothesis given in (61) seems to be relevant: (61)

Hypothesis 6 Operators with higher scope may affect the expression of operators with lower scope

The scope of operators can be represented as in (62): (62)

The scope of operators π4 π3 π2

Ε,: [π4 ILL (S) (Α) (π3 Χ,: [(π2 e,: [π, Pred ß (α,) etc.] (e,))] (Χι))] (Ε,) (α = any kind of argument)

26

Kees Hengeveld

The representation in (62) shows that subordinate constructions occupying an argument position fall within the scope of most matrix clause operators. Two examples may serve to illustrate the effects this may have: (63)

(64)

Spanish (Indo-European, Hengeveld 1987b) Quizas es seguro que la ceguera puede maybe is.iND certain that the blindness can.iND ser vencida be cured 'Maybe it's certain that blindness can be cured.' Quizas sea seguro que la ceguera pueda maybe is.subj certain that the blindness can.suBJ vencida cured 'Maybe it's certain that blindness can be cured.'

ser be

Sentences (63)-(64) are a non-mitigated and a mitigated version respectively of a Spanish declarative sentence. Mitigation is expressed by means of the subjunctive, as in (64). Although the modal adjective seguro normally requires the indicative in its complement, as in (63), it takes a subjunctive complement in (64). The higher scope operator mitigation thus not only affects the matrix clause predicate, but also all other inflected forms in the clause. A second example is given in (65): (65)

I knew you would come

This sentence can be represented as in (66): (66)

(Past β;: [Know v (x;: pi (xDWp (Cert X;: [(Fut e,: [Come v (xj: p2 (xj))] (ej))] (Xi))Go] to))

Here the future tense operator of the complement of know falls within the scope of the past tense operator of the matrix clause. The result is a future in the past. The expression of lower operators need not be affected by the presence of higher operators, as can be demonstrated by means of the following Nama example:

Layers and operators (67)

Nama Hottentot (Khoisan, Hagman 1974) Siikxm ke ke //ηάύ //'iip kd Ι DU DECL REM.PAST hear he REC.PAST 'We heard that he had just gone.'

!uu go

27

Ixäisä that

Hagman (1974: 257) remarks that in this example the recent past (kd) in the complement is "recent relative to the context, i.e., the matrix sentence". The fact that the recent past operator in the embedded construction is correctly interpreted can be regarded as the result of its being within the scope of the remote past operator of the matrix clause, as predicted by (62).

4.2.

Adverbial constructions

In Functional Grammar adverbial constructions are generally referred to as satellites. Satellites are like arguments in that they contain a variable and are assigned a semantic function. They differ from arguments only in the fact that they are optional. Satellites can be classified according to their internal structure, as has been done for arguments in the preceding section. For instance, reasons are third order entities (Lyons 1977: 445), and should therefore be represented as propositions, causes are second order entities, and should be represented as predications, beneficiaries are first order entities, and should therefore be represented as simple terms. This, however, is not the main subject of this section. The question here is rather how satellites should be represented within the clause model developed so far. For a more detailed analysis the reader is referred to Dik et al. (this vol. [1990]). In this article I will concentrate on the interaction between satellites and operators. In 4.2.1 satellites are subdivided according to the layer at which they apply. Furthermore attention is given to the differences between restrictive (4.2.2) and non-restrictive (4.2.3) adverbials (see Hannay and Vester 1987).

4.2.1.

Layers and satellites

The functions to be fulfilled by satellites at the different layers are not very different from the ones postulated for operators. Stating that a certain S o A took place yesterday is applying a lexical strategy which is comparable to the application of the Past operator within a grammatical strategy. The

28

Kees Hengeveld

definitions given here for satellites at the different layers of the clause are therefore reminiscent of the definitions given for operators in section 3.1: (68)

Satellites (definitions) a. Predicate satellites capture the lexical means which specify additional properties of the set of SoAs designated by a bare predication. b. Predication satellites capture the lexical means which locate the SoAs designated by a predication in a real or imaginary world and thus restrict the set of potential referents of the predication to the external situation(s) the speaker has in mind. c. Proposition satellites capture the lexical means through which the speaker specifies his attitude towards the proposition he puts forward for consideration. d. Illocution satellites capture the lexical means through which the speaker modifies the force of the basic illocution of a linguistic expression so as to make it fit his communicative strategy. e. Clause satellites capture the lexical means through which the speaker locates the speech act designated by a clause within the context of discourse and thus restricts the set of potential perlocutions of the clause.

For the last category of satellites there is no corresponding operator type. This is because with this satellite type we are crossing the clause boundaries and entering the domain of interclausal relations at the discourse level. The main difference between satellites at the interpersonal level (proposition satellites and illocution satellites) and satellites at the representational level (predicate satellites and predication satellites) is that the former are speaker-, addressee- or speech act-oriented, whereas the latter are argumentor SoA-oriented. The classification proposed here can be compared very well with the classification of adverbial constructions proposed in Quirk et al. (1985), elaborating on Greenbaum (1969):

Layers and operators

29

(69) The classification of adverbial constructions Quirk et al. (1985) Corresponding satellite type Adjuncts Representational level Predication Predicate satellites Sentence Predication satellites Disjuncts Interpersonal level Attitudinal Propositional satellites Style Illocution Conjuncts Discourse level: Clause satellites For tests distinguishing between these different types of adverbial see Greenbaum (1969), Bartsch (1976), Quirk et al. (1985), and Dik et al. (this vol. [1990]).

4.2.2.

Restrictive

adverbials

Satellites have generally been represented in F G as in: (70)

Predß (arg,)... (arg„) (sat^ ... (satn)

In (70) satellites are represented as optional arguments of the predicate. This representation in fact captures the definition given for predicate satellites, the main function of which is to give additional information on the internal structure of the SoAs under consideration. Vet (1986) suggests an approach in which satellites specifying time and location are represented as secondary restrictors of the e-variable, as in: (71) (72)

I saw him yesterday (Past e ; : [/ see him] (e;): yesterday (e^)

(73) (74)

I saw him in the garden (Past e ; : [I see him] (e,): [(χ{: garden (Xi))^] fe))

In this approach satellites are considered to be optional secondary restrictors of the event variable e. This representation captures the definition given for predication satellites, the main function of which is to specify the setting within which a S o A occurs (occurred, will occur).

30

Kees Hengeveld

Perhaps Vet's analysis may be extended to capture restrictive adverbials at the propositional level, although truly restrictive adverbials seem to be exceptional there. An example could be sentence (75), 13 in which John is presented as the source (So) of the proposition presented by the speaker. It might be represented as in (76): (75) (76)

According to John there's a bull in the field (X,: [There's a bull in the field] (X,): [(x,: John (Xi))So] (Xi))

In this approach restrictive adverbials are within the scope of the operators associated with the level they restrict. Sentence (77), containing a restrictive predicational satellite, and its schematic representation (78) may illustrate this point: (77)

(78)

Nama Hottentot (Khoisan, Hagman 1974) tsii stäsisä Ι'οά-kxm ηιί Ζ/όά-s 'ai!'aa-p ke and station to-iPL FUT go.down-NMLZTN before-3SG DECL //näatse km kerkepa ke häa 'ii that.day big church.service REM.PAST COP COP 'And before we went down to the station that day there was a big church service.' (Rem.past e,: [predication] (e^: [(Fut [predication] (ej))Temp] (e,))

In (78) the second event falls within the scope of the Past operator of the first event. Due to a rule governing the sequence of tenses the temporal construction is in the past tense in the English translation, whereas in the Nama original the morpheme ηιί indicates future time reference.

4.2. j.

Non-restrictive

adverbials

For non-restrictive adverbial clauses I follow the proposal made by Hannay and Vester (1987), although it has to be adapted in such a way that it applies to all the different layers proposed here. Hannay and Vester suggest that nonrestrictive clauses be analyzed in the following way: (79)

Core predication (predication)Sem pragm synt

Layers and operators

31

Hannay and Vester compare the organisation of complex sentences with the organization of predications, in the sense that subordinate predications may entertain semantic, pragmatic and syntactic relations with core predications, just as arguments entertain these relations with predicates. Application of this model within a layered approach suggests a further subcategorization of non-restrictive adverbials according to the level at which they occur. In general terms, the approach to be adopted in this section can be summarized as follows: (80)

( < X i ) , ( a 2 ) s e r a Pragm Synt

where α represents any of the variables e, X, or E. The schema in (80) indicates that two constructions of the same level can be brought into a non-restrictive relation, where one of the two can be seen as subordinate to the other, and its semantic, pragmatic and/or syntactic functions indicate in what way it is subordinated. In this section I restrict myself to the semantic functions of non-restrictive adverbial constructions. The restriction that the two constructions be of the same level is in fact the restriction imposed on all kinds of clause linkage by Foley and Van Valin (1984: 188). Their approach runs into trouble where it has to deal with complementation (1984: 25iff.J, but works fine for the constructions dealt with here. The possible non-restrictive combinations of constructions in the present approach can be listed as in ( 8 1 H 8 3 ) : (81) (82) (83)

(eO, ( e ) s e m (XJ, ( X ) s e m 2

2

( E j ) , (E2)sem

predication combining proposition combining clause combining

The subordinate constructions in these combinations can be considered predication, proposition, and clause satellites, respectively. Examples of the intended constructions are:14 (84) (85) (86)

When he comes, I leave John is home, because that's his car outside Watch out, because there's a bull in the field

In (84) one S o A is presented as the circumstance (Circ) within which another takes place. In (85) one fact is presented as evidence (Evid) for

32

Kees Hengeveld

another fact. In (86) one speech act is presented as the motivation (Mot) for another speech act. Sentences (84H86) can be represented as in (87H89): (87) (88) (89)

(E;: [DECL (S) (A) (X,: [(Pres e;: [/ go out] (ej), (Pres ej : [He come] (e))aJ (X;))] (E,)) (Ei: [DECL (S) (A) (X,: [John is home] (X,)), (Χ,: [That's his car outside] (Xj))Evid] (Ei)) (Ε;: [IMP (S) (A) (e;: [Watch out] (ei))15] (E,)), (Ej! [DECL (Xj! [There's a bull in the field] ( X J ) (Ej))Mot

The most important aspect of these representations is that two constructions in a non-restrictive relation can be subordinate to shared higher operators, a phenomenon which Foley and Van Valin (1984), following Olson (1981), labeled cosubordination. The effects of this approach can be illustrated by means of the following examples.16 The coordinated predications in (84) cannot carry their own proposition operators (π3), as (90) demonstrates. (90)

*When he possibly comes, I must leave

Both can, however, carry their own predication operators (π2): (91)

When he comes, I will have left already

The coordinated propositions in (85) cannot carry their own illocution operators (π4): (92)

*John is home!, because that might be his car outside

But they can each carry their own proposition operator (π3): (93)

John must be home, because that certainly is his car outside

The coordinated clauses in (86) can carry their own illocution operators (ττ4): (94)

You had better watch out, because there might be a bull in the field!

Layers and operators

33

A final look at Nama shows an important difference between restrictive and non-restrictive satellites with respect to the expression of operators. In (77), discussed earlier, the restrictive temporal satellite is within the scope of the tense operator of the matrix predication, and the occurrence of the SoA expressed within the satellite predication is interpreted relative to the occurrence of the SoA expressed within the matrix predication. In (95), with its rough representation (96), the non-restrictive temporal satellite is not within the scope of the tense operator of the matrix predication and therefore has to be interpreted relative to the moment of speaking: (95)

Nama Hottentot (Khoisan, Hagman 1974) Tsii /'ae//ams 'άί-ta ke sü

'00-p

and

TEMP-3SG DECL

tu

Windhoek

LOC-ISG

/hdopä ke

poss.ISG friend

REM.PAST a r r i v e

I'aute

REM.PAST w a i t

häa.'ii

stasis

täpa

PF

station

LOC

ke

'And when I arrived in Windhoek, my friend had been waiting for me at the station.' (96)

(Rem.past e ; : [predication] (ej), (Rem.past e^ [predication] (ej))Temp

4.3.

Secondary predication

To complete the picture of subordinate constructions some final remarks should be made on the status of free secondary predication, as illustrated in: (97) / saw him walking down the street Here walking down the street is an optionally added secondary predication. Following Vester (1983), constructions like (97) may be represented schematically as in (98), where secondary predication is taken quite literally: (98)

(e,: [predication] (e;): [predication] (eO)

This representation predicts that the two predications cannot be specified independently for tense, but can be specified for aspect. That this is indeed the case can be illustrated by means of the following sentences (cf. Comrie 1 9 7 6 : 40):

34

Kees Hengeveld

(99)

I saw him cross the street

(100)

I saw him crossing the street

In the perfective variant (99) the use of the infinitive indicates that I witnessed the crossing from beginning to end, whereas in the imperfective (100) the use of the gerund indicates that I witnessed the crossing, but not necessarily from beginning to end.

5.

Conclusion

B y w a y of conclusion I summarize the main points that I have argued: 1.

Every main clause can be analyzed at two levels: the representational level and

2.

Each level can be analyzed as containing several layers.

the interpersonal level. 3.

Each layer has its own associated operators.

4.

Every subordinate construction can be classified according to the highest layer

5.

Every adverbial construction can in addition be classified according to the

it contains. layer to which it attaches.

Notes This contribution originally appeared in Journal of Linguistics 25 (1): 125-157, 1989. i.

This article has profited considerably from discussions with and the detailed comments of Machtelt Bolkestein and Simon Dik. I am furthermore grateful to Casper de Groot, Peter Kahrel, Lachlan Mackenzie, Hotze Mulder, Jan Rijkhoff, Nigel Vincent, and two anonymous referees of Journal of Linguistics for valuable comments on earlier versions of this article, a preliminary version of which appeared as Working Papers in Functional Grammar 27 (1988). Those unfamililar with the theory of Functional Grammar may refer to Dik (1978) or to the summary given in Dik (1980). Abbreviations used in this article - General·. FG, Functional Grammar; SoA, State of Affairs; A , Addressee; S, Speaker;

SG,

singular;

PL,

plural;

DU,

dual;

NMLTZN,

nominalization;

COP,

copula.

Layers and operators

35

Wordclasses: ß, any wordclass; N, noun; V, verb; A, adjective. Variables: a, any variable; E, speech act; X, potential fact; e, state of affairs; x, individual. Illocutionary frames: ILL, any illocutionary frame; DECL, declarative; INT, interrogative; IMP, imperative. Semantic functions: Sem, any semantic function; Ag, Agent; Go, Goal; 0 , Zero; Exp, Experience^ Poss, Possessor; Loc, Location; So, Source; Temp, Time; Circ, Circumstance; Evid, Evidence; Mot, Motivation. Predicate operators: Pf, Perfective; Ingr, Ingressive; Res, Resultative; Neg, Negation. Predication operators: Pres, Present; Fut, Future; Rem. past, Remote past; Rec.past, Recent past; Iter, Iterative; Nec, Necessitative; Irr, Irrealis. Proposition operators: Inf, Inferential; Quot, Quotative; Cert, Certainty; Prob, Probability. Elocution operators: Mit, Mitigation; Reinf, Reinforcement. 2. The semantic function Goal covers all arguments designating entities to which an action is applied. 3. Cf. Lyons' (1977: 170) remark that "We cannot say of a sentence like That man over there is my father that it expresses a true or false proposition unless we know who has uttered it and who is the person being referred to by means of the expression 'that man over there'." 4. This was suggested to me by Casper de Groot. 5. The difference between non-generic and generic predications can be accounted for in a similar way if the latter are represented, parallel to generic terms, in the following way, which was suggested to me by Simon Dik: (i)

(g e,: [predication] (e,))

The interpretation of the generic operator (g) in this context is something like: the statement is valid for any instantiation of the SoA designated by the predication. 6. See Hengeveld (1987b) for a more elaborate treatment of modality. 7. Bybee's results are not directly applicable to the model used here, as she investigated the expression of three categories, Tense, Mood, and Aspect, without further subdivisions being made. It seems, however, that even some of the exceptions to the preferred orders which she found (Pred Α Τ Μ or Μ Τ A Pred) could be explained within the model proposed here. In particular the possibility of objective mood distinctions occurring closer to the predicate than tense distinctions (as in several Creole languages, where the irrealis morpheme occurs within the scope of the anterior morpheme) is not excluded in the present model, since both categories can be seen as expressed by predication operators, with respect to which no claim is made as to their internal ordering.

36

Kees Hengeveld

8.

Cole (1982: 95) notes that "the primary device for the expression of sentence emphasis is the use of the highly emphatic validator -ma(ri) (from -mi 'first hand information' + ari 'affirmation')".

9.

Cf. also Traugott (1982: 253) who hypothesizes a development from "less personal to more personal".

10. Example (43c) is not ungrammatical under a habitual interpretation, as in the sentence (Lachlan Mackenzie, p. c.): I wish I travelled (for instance as a salesman) in France and not (as I actually do) in Finland. For an explanation see below. 1 1 . In fact there are at least three different senses of the verb see that are relevant in the present context. It can be used to designate the visual perception of individuals or first order entities, the visual perception of events or second order entities, and the mental perception of potential facts or third order entities. In the latter reading its meaning corresponds roughly to understand. In our representational system these different readings could be accounted for by assuming the following three lexical entries for the verb see, each with a different Goal argument: (i) See v (xi)0ExP (X2)go (ii) See v (xiWxp (Ci)c;„ (iii) See v (xiWxp (XI)GO

(visual perception of individual) (visual perception of event) (mental perception of potential fact)

These different Goal arguments are generally realized as nouns, non-finite complements and finite complements respectively (see Barwise and Perry 1983: 179; Woodbury 1986: 198). 12. I am indebted to Simon Dik for drawing my attention to the facts which follow. 13. The adverbial seems to be restrictive only if Focus is on the proposition as a whole. 14. Non-restrictive satellites should be distinguished from extra-clausal constituents such as Theme and Tail (see Dik 1989). The latter are necessarily outside the scope of all matrix clause operators, unlike some of the non-restrictive satellites treated here. 15. I assume that there is no intermediate propositional layer in imperative sentences. Imperative illocutionary frames could be assigned the following structure and paraphrase, adapted from Dik (1989): (i)

IMP (S) (Α) (ο,): Speaker (S) wishes the Addressee (A) to perform the action (ei).

Layers and operators

37

In this way the irrelevance of truth-commitment in imperative sentences is accounted for by means of the absence of the layer at which it applies. 16. Cf. also Bolkestein (1991) on the expression of modality in constructions expressing a causal relation in Latin.

References Bartsch, Renate 1976

The Grammar of Adverbials: A Study in the Semantics and Syntax of Adverbial Constructions. Amsterdam: North-Holland.

Barwise, Jon, and John Perry 1983

Situations and Attitudes. Cambridge, MA: The MIT Press.

Bolkestein, A. Machtelt 1991

Causally related predications and the choice between parataxis and hypotaxis in Latin. In New Studies in Latin Linguistics,

Robert

Coleman (ed.), 427-445. Amsterdam: John Benjamins. Bühler, Karl 1934

Sprachtheorie. Jena: Fischer.

Bybee, Joan L. 1985

Morphology: A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins.

Chafe, Wallace L., and Johanna Nichols (eds.) 1986

Evidentiality: The Linguistic Coding of Epistemology. Norwood, NJ: Ablex.

Chung, Sandra, and Alan Timberlake 1985

Tense, aspect, and mood. In Language Typology and Syntactic Description, Vol. Ill, Timothy Shopen (ed.), 202-258. Cambridge: Cambridge University Press.

Cole, Peter 1982

Imbabura Quechua. Amsterdam: North-Holland.

Comrie, Bernard 1976

Aspect. Cambridge: Cambridge University Press.

1985

Tense. Cambridge: Cambridge University Press.

Craig, Colette G. 1977

The Structure ofJacaltec. Austin: University of Texas Press.

38

Kees Hengeveld

Dik, Simon C. 1978

Functional Grammar. Amsterdam: North-Holland. Original edition,

1979

Raising in a Functional Grammar. Lingua 47. 119-140.

1980

Studies in Functional Grammar. London: Academic Press.

Dordrecht: Foris, 1978.

1989

The Theory of Functional Grammar. Part 1: The Structure of the Clause. (Functional Grammar Series 9.) Dordrecht: Foris.

Dik, Simon C., Kees Hengeveld, Elseline Vester, and Co Vet this vol.

Reprint. The hierarchical structure of the clause and the typology of adverbial satellites. Layers and Levels of Representation in Language Theory: A Functional View, Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 25-70. Amsterdam: John Benjamins, 1990.

Foley, William Α., and Robert D. Van Valin Jr. 1984

Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press.

Fortescue, Michael 1984

West Greenlandic. London: Croom Helm.

Goossens, Louis 1985

The auxiliarization of the English modals. Working Papers in Functional Grammar 7, University of Amsterdam.

Gorbet, Larry P. 1976

A Grammar ofDiegueno Nominals. New York: Garland.

Greenbaum, Sidney 1969

Studies in English Adverbial Usage. London: Longman.

de Groot, Casper 1985

Predicates and features. In Predicates

and Terms in

Functional

Grammar, A. Machtelt Bolkestein, Casper de Groot, and J. Lachlan Mackenzie (eds.), 71-84. Dordrecht: Foris. Hagman, Roy S. 1974

Nama Hottentot Grammar. Ann Arbor: UMI.

Halliday, Michael 1970

Functional diversity in language. Foundations of Language 6: 322361.

Hannay, Mike, and Elseline Vester 1987

Non-restrictive relatives and the representation of complex sentences. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 39-52. (Functional Grammar Series 6.) Dordrecht: Foris.

Layers and operators

39

Haverkate, Henk 1979 Impositive Sentences in Spanish. Theory and Description in Linguistic Pragmatics. Amsterdam: North-Holland. Hengeveld, Kees 1987a The Spanish mood system. Working Papers in Functional Grammar 22, University of Amsterdam. 1987b Clause structure and modality. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 53-66. (Functional Grammar Series 6.) Dordrecht: Foris. 1990 Semantic relations in non-verbal predication. In Layers and Levels of Representation in Language Theory: A Functional View, Jan Nuyts, A. Machtelt Bolkestein and Co Vet, (eds.) 1 0 1 - 1 2 2 . Amsterdam: John Benjamins. Jakobson, Roman 1971 Shifters, verbal categories, and the Russian verb. In Selected Writings, Vol. II, Roman Jakobson, 130-147. The Hague: Mouton. Katz, Jerrold J., and Paul M. Postal 1964 An Integrated Theory of Linguistic Descriptions. Cambridge, MA: The MIT Press. Lehmann, Christian 1988 Towards a typology of clause linkage. In Clause Combining in Grammar and Discourse, John Haiman and Sandra A. Thompson (eds.), 181-226. Amsterdam: John Benjamins. Lewis, Geoffrey 1967 Turkish Grammar. Oxford: Clarendon Press. Li, Charles N., and Sandra A. Thompson 1981 Mandarin Chinese. Berkeley: University of California Press. Lyons, John 1977 Semantics. 2 vols. Cambridge: Cambridge University Press. Matthews, George Hubert 1964 Hidatsa Syntax. The Hague: Mouton. Olson, Michael L. 1981 Barai clause junctures: Towards a functional theory of interclausal relations. Ph. D. diss., Australian National University. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman.

40

Kees Hengeveld

Ross, John R. 1970

On declarative sentences. In Readings in English Transformational Grammar, Roderick A. Jacobs and Peter S. Rosenbaum (eds.), 222-272. Waltham: Ginn.

Rust, Friedrich, Jr. 1965

Praktische Namagrammatik. Cape Town: Balkema.

Steedman, Mark J. 1977

Verbs, time and modality. Cognitive Science 1: 216-234.

Traugott, Elizabeth Closs 1982

From propositional to textual and expressive meanings: Some semanticpragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics, Winfried R Lehmann, and Yakov Malkiel (eds.), 245-271. Amsterdam: John Benjamins.

Vendler, Zeno 1967

Linguistics in Philosophy. Ithaca, NY: Cornell University Press.

Vester, Elseline 1983

Instrument and Manner Expressions in Latin. Assen: Van Gorcum.

Vet, Co 1986

A pragmatic approach to tense in Functional Grammar. Working Papers in Functional Grammar 16, University of Amsterdam.

Willett, Thomas L. 1988

A cross-linguistic survey of the grammaticization of evidentiality. Studies in Language 12 (1): 51-97.

Woodbury, Anthony C. 1986

Interactions of tense and evidentiality: A study of Sherpa and English. In Evidentiality: The Linguistic Coding of Epistemology, Wallace L. Chafe and Johanna Nichols (eds.), 188-202. Norwood, NJ: Ablex.

Further reading Matthew P. Anstey and J. Lachlan Mackenzie Hengeveld's (this vol. [1989]) proposals were immediately adopted by Dik (1989, 1997), who recasts Functional Grammar as a layered model of the clause. Dik introduces two modifications to Hengeveld's proposals, which were much discussed in the nineties: the abstract predicate ILL was reanalyzed as a n4 operator and π, was given scope over the predicate and its arguments. Dik et al. (this vol. [1990]) give further consideration to the implications of layering for satellites, providing detailed evidence that different satellites are attached to different layers in the structure. Layering was to become a major theme of F G in the nineties (for initial discussions, see Nuyts et al. 1990), and indeed there was analogous interest in this notion in other theories of grammar. Siewierska [1992] conducts a detailed comparison of F G and Government and Binding in this respect and Van Yalin and LaPolla [1997] argue for explicit parallels between F G layering and clause layers in Role and Reference Grammar (RRG). Thorough metatheoretical comparisons are undertaken by Butler (1996), who critiques layering from the viewpoint of Systemic Functional Grammar (SFG), and above all by Butler (2003), who provides an exhaustive assessment of layering in FG, R R G and SFG. Hengeveld's work sparked off a number of applications to grammatical problems. The implications of the four layers for complementation are pursued in Dik and Hengeveld (1991) with regard to perception verbs and similarly in Genee's (1998) monograph on Irish verb complements. Mackenzie (1996) shows that the various kinds of nominalization encountered in English can be insightfully classified in terms of those four layers. Vismans (1994) applies Hengeveld's system to the analysis of modal particles in Dutch, showing that it clarifies many aspects of their semantics and syntactic ordering. Layering has also been very significant in the discussion of modality in F G and in functional linguistics in general (Van der Auwera 1996, 2001; Nuyts 1993, 2001; Verstraete 2004). A t the same time, it was felt that the four-layered structure could be improved by introducing more variables. After the 1990 F G conference, largely devoted to issues of layering (Fortescue et al. 1992), Keizer (this vol. [1992]) argues for a variable ( f ) for predicates and Hengeveld (1992) for a

42

Matthew P. Anstey and J. Lachlan Mackenzie

corresponding variable (Fi) for the abstract illocutionary predicate. Others go further, pushing for the adoption of more layers. Cuvalay (1995), in a critical survey of the literature, argues that a six-layer system is necessary, three at the representational layer (predicate, core predication, and extended predication) and three at the interpersonal layer (proposition, clause, and expression, where the latter refer to the plain illocution and the illocution plus extra-illocutionary satellites respectively). Rijkhoff (this vol. [1990]) argues that layering applies as much to nominal terms as to clauses. The introduction of layering into F G also met various challenges. Bolkestein (1992) is generally sympathetic to the notion, but finds it difficult in practice to locate various phenomena. She shows that some of these appear to be acceptable in complement clauses, contrary to the layering hypothesis. In various publications, Nuyts ( 1 9 9 2 , 1 9 9 3 , 2 0 0 4 ) does not deny the reality of layering in meaning but argues that it should more properly be located in conceptual structure and not in grammar. Moutaouakil (1996) considers various non-declarative constructions, concluding that many of these require fewer than the full number of layers for their proper description. More recently, Anstey (2002) points out many differences in detail between various scholars' implementations of layering as part of a thorough-going revisitation of thirteen years of work on the topic. Like Nuyts, he proposes a redivision of labour between conceptual and grammatical structure, but unlike Nuyts, Anstey preserves the role of layering in grammar. To some extent, Anstey's proposals answer Harder's (1998) call for a conceptual structure to complement F G layering. Harder sees cognition and communication as having developed in tandem and requiring integrated treatment. This direction of thought has influenced, among others, Hengeveld's (2004) approach, w h o presents a new version of F G , Functional Discourse Grammar, in which the interpersonal and representational levels are no longer coalesced into one structure but still show internal layering, and in which the grammatical component is complemented by a conceptual component.

References Anstey, Matthew P. 2002

Layers and operators revisited. Working Papers in Functional Grammar. 77·

Further reading

43

Auwera, Johan van der 1996

Modality: The three-layered scalar square. Journal of Semantics. 13:

2001

Modality: Domains, layers, and parts of speech. Revista Canaria de

181-195. Estudios Ingleses. 42: 237-247. Bolkestein, A. Machtelt 1992

Limits to layering: Locatability and other problems. In Layered Structure and Reference

in a Functional

Perspective,

Michael

Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 387-407. Amsterdam: John Benjamins. Butler, Christopher S. 1996

Layering in functional grammars: A comparative survey. In Complex Structures: A Functionalist

Perspective,

Betty Devriendt, Louis

Goossens, and Johan van der Auwera (eds.), 1-27. (Functional Grammar Series 17.) Berlin/New York: Mouton de Gruyter. 2003

Structure and Function: A Guide to Three Major Structural-functional Theories. Two volumes. (Studies in Language Companion Series.) Amsterdam: John Benjamins.

Cuvalay, Martine 1995

The Ε-structure in Functional Grammar: towards a consistent treatment of Tense, Mood, Aspect and Illocutionary Force. Working Papers in Functional Grammar. 59.

Dik, Simon C. 1989

The Theory of Functional Grammar. Part 1: The Structure of the Clause. (Functional Grammar Series 9.) Dordrecht: Foris.

1997

The Theory of Functional Grammar. Part I: The Structure of the Clause. 2nd rev. ed. Kees Hengeveld (ed.). (Functional Grammar Series 20.) Berlin/New York: Mouton de Gruyter.

Dik, Simon C., and Kees Hengeveld 1991

The hierarchical structure of the clause and the typology of perception verb complements. Linguistics. 29 (2): 231-259.

Dik, Simon C., Kees Hengeveld, Elseline Vester, and Co Vet 1990

The hierarchical structure of the clause and the typology of adverbial satellites. In Layers and Levels of Representation in Language Theory: A Functional View, Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 25-70. (Pragmatics and Beyond 13.) Amsterdam: John Benjamins.

44

Matthew P. Anstey and J. Lachlan Mackenzie

Fortescue, Michael, Peter Harder, and Lars Kristoffersen (eds.) 1992

Layered Structure and Reference in a Functional Perspective. Amsterdam: John Benjamins.

Genee, Inge 1998

Sentential Complementation in a Functional Grammar of Irish. The Hague: Holland Academic Graphics.

Harder, Peter 1998

Function, cognition, and layered clause structure. In

Cognitive

Semantics: Meaning and Cognition, Jens Allwood and Peter Gärdenfors (eds.), 37-66. (Pragmatics and Beyond 55.) Amsterdam: John Benjamins. Hengeveld, Kees 2004

The architecture of a Functional Discourse Grammar. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 1-21. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

this vol.

Reprint. Layers and operators in Functional Grammar. Journal of Linguistics. 25 (1): 127-157. 1989.

this vol.

Reprint. Parts of speech. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 29-55. (Pragmatics & Beyond New Series 23.) Amsterdam: John Benjamins, 1992.

Keizer, M. Evelien this vol.

Reprint. Predicates as referring expressions. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder and Lars Kristoffersen (eds.), 1-27. Amsterdam: John Benjamins, 1992.

Mackenzie, J. Lachlan 1996

English nominalizations in the layered model of the sentence. In Complex Structures: A Functionalist

Perspective, Betty Devriendt,

Louis Goossens, and Johan van der Auwera (eds.), 325-355. (Functional Grammar Series 17.) Berlin/New York: Mouton de Gruyter. Moutaouakil, Ahmed 1996

On the layering of the underlying clause structure in Functional Grammar. In Complex Structures: A Functionalist Perspective, Betty Devriendt, Louis Goossens, and Johan van der Auwera (eds.), 201-227. (Functional Grammar Series 17.) Berlin/New York: Mouton de Gruyter.

Further Reading Nuyts, Jan 1992

1993 2001

2004

45

Aspects of a Cognitive-pragmatic Theory of Language. On Cognition, Functionalism and Grammar. (Pragmatics and Beyond 20.) Amsterdam: John Benjamins. Modality and the layered representation of conceptual and linguistic structure. Linguistics 31: 933-969. Epistemic Modality, Language and Conceptualization: A CognitivePragmatic Perspective. (Human Cognitive Processing.) Vol. 5. Amsterdam: John Benjamins. Remarks on layering in a cognitive-functional language production model. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 275298. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

Nuyts, Jan, A. Machtelt Bolkestein, and Co Vet (eds.) 1990 Layers and Levels of Representation in Language Theory: A Functional View. Amsterdam: John Benjamins. Rijkhoff, Jan this vol. Reprint. Toward a unified analysis of terms and predications. In Layers and levels of representation in language theory, Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 165-192. Amsterdam: John Benjamins. 1990. Siewierska, Anna 1992

Layers in Functional Grammar and Government Binding Theory. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 409-430. Amsterdam: John Benjamins. Van Valin, Robert D., Jr., and Randy J. LaPolla 1997

Syntax: Structure, Meaning and Function. (Cambridge Textbooks in Linguistics.) Cambridge: Cambridge University Press. Verstraete, Jean-Christophe 2004 The problem of subjective modality in the Functional Grammar model. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez, (eds.), 243-274. (Functional Grammar Series 24.) Berlin & New York: Mouton de Gruyter.

46

Matthew P. Anstey and. J. Lachlan

Mackenzie

Vismans, Roel 1994

Modal Particles in Dutch Directives: A Study in Functional Amsterdam: IFOTT.

Grammar.

Toward a unified analysis of terms and predications

Jan Rijkhoff

1.

Introduction1

Aristotle stated that movement or change "pertain[s] exclusively to quality, quantity, and locality, each of which embraces contrasts" (Physics, V. II; see also Rijksbaron 1989). I will argue that operators as well as satellites, both in the term phrase and in the predication, can be characterized by these three notions. Additionally I will attempt to show, firstly, that operators in the term phrase do not all have the same scope; and secondly, that these scope differences are formally reflected in the linguistic expression. In this respect this contribution can be regarded as an extension of Hengeveld (this vol. [1989]), which contains a proposal concerning the organization of operators in the predication and higher level domains. On the other hand, this article is also an amendment to his proposal, in that it argues that quantity operators have their own scope.

2.

Terms and predications

I will argue that terms can be represented in the following way: (ι)

(Ω

Xi :

Ω Φ, ( Xi ): Φ 2 (xD: ··.: Φ„ (x,))2

In (1) Ω stands for various term operators, which indicate, among other things, (in)definiteness and number. I use the phrase 'term operator' for any operator in the domain of the term phrase, including those operators that are supposed to have only the nominal predicate in their scope. Similarly, 'predication operator' (see below) is used for any kind of operator in the domain of the predication, including those operators that are supposed to have only the verbal predicate in their scope. Operators have a certain part (or 'layer') of some domain in their scope and may be expressed as grammatical (as opposed to lexical) elements; if expressed, term operators usually take the

48

Jan Rijkhoff

form of a determiner, a numeral/quantifier, or an inflectional element. This is discussed in more detail below. Xj is the term variable symbolizing the intended referent. Every Φ(χ ; ) is an open predication in x r Φ is a predicate, designating a property of the referent xi5 or a relation between Xj and one or maximally two other referents. These predicates are called restrictors: they progressively narrow down the set of potential referents of the term. Φι is the first restrictor; this is usually the head noun (N), which designates the most characteristic property of a referent: its name (which may be a proper name like Amsterdam or a type name like book); Φ 2 is the second restrictor, e.g. an adjectival predicate (A) designating such properties as colour (red), size (big), propensity (jealous) etc. For instance (i = indefinite, d = definite): (2) (2')

Two nice little houses in the country (i2x ; : house N (x,): littleA (x,): nice A (x,): {(dlxj: country N (Xj))LoC} (xi))

Here the fourth restrictor is a term predicate with the semantic function Location (see Dik 1989: ch. 8.5). Predications can be represented as follows: (3)

(πβί:[πΦν(Τ1)(Τ2)...](βί))

In (3) π stands for various predication operators, which indicate (among other things) aspect and tense. Predication operators, too, have a particular layer in their scope (Hengeveld this vol. [1989]; Dik 1989). So far two layers have been recognized in the predication, but I will contend below that three layers can be distinguished in this domain. e ; is the predication variable symbolizing the State of Affairs (SoA) (see Vet 1986 for the variable e). Φ ν symbolizes the main (verbal) predicate. Tn stands for term phrase. For instance: (4) (4')

The man gave the car to a friend (past ej: [give v (dlx;: manN (Xi))Agsubj (dlx,: carN (Xj))Goobj (ilx k : friend N (Xk))Rec] fe))

In sections 3 - 5 I will attempt to demonstrate that both in the term and in the predication, operators can be characterized in terms of the notions 'quality', 'quantity', and 'locality'.

Terms and predications 3.

49

Quality operators

By definition 'quality' relates to the essential or characteristic properties of an entity; if there is a quality operator in the term or predication, this operator must, therefore, concern essential or characteristic properties of the referents of these domains. Since referents are most crucially defined by the lexical head of the domain (prototypically a noun or a verb), I will assume that a quality operator has only the head of the domain in its scope. Qualitative operators should not affect the lexical meaning of a predicate, for then this predicate would designate a different property or relation. I assume that in addition to having a lexical meaning, predicates may be specified (overtly or covertly) for certain structural ('aspectual') characteristics of the designated property or relation which pertain to the spatial or temporal dimension. In other words, when a predicate is used in a referring expression (term or predication), it can also indicate something about a referent's representation in the dimension of time or space. Referents of terms and predications are mental entities; in other words, they are not entities in the physical, external world (Dik 1989; Seiler 1986: ch. 2). Furthermore, referents are mental constructs rather than mental copies of things as they occur in the external world. Because we construe and refer to mental constructs, the spatio-temporal characteristics of such a construct (the referent) need not necessarily coincide with those of its possible counterpart in the external world. What the referent 'looks like' mentally, partly depends on the spatio-temporal specifications coded on or in the predicate used in the linguistic expression. Thus, there can be certain discrepancies between referents and their real-world equivalents (if they exist) with respect to matters of time and space. For instance, in some languages one and the same real-world situation may be referred to by a predication containing a verb in the perfective or in the imperfective form. The English sentence / stood therefor an hour can be translated into Russian either with the verb in the imperfective form (ja stojal tarn cas) or in the perfective form (ja postojal tam cas\ from Comrie 1976: 17)· That we refer to mental constructs can also be demonstrated by the fact that we can refer to things that cannot be perceived in the immediate linguistic or non-linguistic context (e.g. lost or invisible objects, the Armada, oxygen) or entities that do not exist in the real world (Martians, unicorns).

50

Jan Rijkhoff

In short, nominal and verbal predicates may be marked for certain spatiotemporal specifications, i.e. for a particular verbal or nominal 'aspect'.

3.1.

Verbal aspect

In the temporal dimension a relation or property X (sleep, hit, go, arrive, etc.) can be represented as unmarked; or it can be represented as having a beginning, an ending, or as having both: (5) Tj> Tj>

Go > Ree >Ben > Instr > Loc > Temp

Ben and Instr are clear examples of σ,, with their own contribution to the definition of the SoA (see section 3.1 above). They can therefore also act as points of departure for perspectivizing the SoA. But Loc and Temp at first sight present a problem for this theory: if these are examples of σ 2 , they do not enter into the definition of the SoA as such, but rather serve to locate the SoA with respect to spatial and temporal dimensions. How could they then serve as a point of departure for perspectivizing the SoA? Closer scrutiny of the relevant data has revealed, however, that it is especially the 'inner' semantic functions which can be assigned Subject function: those semantic functions, that is, which do indeed contribute to the definition of the SoA as such. These inner semantic functions can therefore be interpreted as σ, rather than σ 2 satellites, and on that interpretation they do not provide counter-examples to the Subj/Obj assignment theory. Rather, this theory provides an additional criterion for distinguishing σ, and σ 2 satellites. Consider the following data. The evidence for Subj assignment to Temp is very slight indeed. It is reported for two Philippine languages, Kalagan (Keenan 1972) and Cebuano (Bell 1983). Keenan gives no examples. Bell (1983: 146) gives the following example from Cebuano:13 (74)

Mogikan ang barko sa alas act-leave SUBJ ship at clock 'The ship will leave at six o'clock.'

sayis six

The typology of adverbial satellites (75)

Igikan

sa barko ang alas INS-leave by ship SUBJ clock 'Six o'clock will be left by the ship'

189

sayis six

Bell adds, however, that constructions of type (75) are quite rare, except in relative constructions. In Cebuano (as in Kalagan) only Subjects can be relativized. Thus, the only way of expressing something like 'The time at which the ship will leave' is through embedding a construction of type (75). It may perhaps be assumed that it is this constraint on relativization which has occasioned Subj assignment to exceptionally go beyond its natural limits. The evidence for Subj/Obj assignment to Loc is much more extensive. Thus, Subj assignment to Loc is reported for Maguindanao, Tagalog, Kapampangan, Kalagan, and Cebuano (all Philippine languages); for Malagasy (which also belongs to the Malayo-Polynesian languages); and Subj and Obj assignment to Loc is reported for Luganda, Chimwi:ni, Swahili, and Kinyarwanda (all Bantu languages). Note, however, that Loc does not necessarily have the status of a σ 2 satellite, which simply locates the whole S o A in some spatial domain. Locative terms can also have argument status and, depending on the way the notion 'locative' is used, they might also cover one or more of the inner local or directional satellites which were assigned 0! status above. Could it be, then, that it is especially these inner locatives which can be assigned Subj function? That there may be something to this idea can be seen even in English. Compare the following pair: (76)

a.

b.

John was writing on the terrace = [a] 'John inscribed something on the terrace' = [b] 'John was writing something while being on the terrace' The terrace was written on by John = only fa]

In the [a] interpretation of (76a), the terrace can be considered as an argument of the predicate write on, or at least as a σ, satellite closely associated with the predicate; in the [b] interpretation it is a σ 2 satellite. Only in the former case can Subj be assigned to it, witness the unambiguous character of (76b). Something similar is involved in the following pair (Quirk et al. 1972: 804):

190 (77)

Simon C. Dik et al. a.

This problem was very carefully gone into by the engineers

b.

*The tunnel was very carefully gone into by the engineers

Again, it appears that Subj assignment is possible only when the term in question is close to being an argument of the nuclear predicate. There are similar indications in this direction for the languages mentioned above. Thus, in her description of Kapampangan Mirikitani (1972) makes a distinction between 'Terminus Locative', defined as "the case designating the place towards or from which an activity is directed" (= Source + Direction), and 'General Locative', comparable to our σ 2 locative satellite. It is the Terminus Locative, not the General Locative, which may receive Subj function. For example, in the Kapampangan equivalents of: (78)

a.

I will go to school (Terminus Loc)

b.

I will read in school (General Loc)

it is only the Terminus Locative in (78a) which can be assigned Subj. Examples of predicates which take a Terminus L o c are 'write on something', 'cook in something', and 'go, walk to some place'. Though we have not been able to check this for all the Philippine languages mentioned above, it is probable that the situation in these languages is similar. For example, Bell (1983: 209) notes that Loc in Cebuano includes Source and Direction, and the only example she gives of a Loc to which Subj is assigned is again 'cook something in a pot', where the L o c term is obviously closely associated with the predicate. A s for the Bantu languages, many of the examples in which Subj or Obj is assigned to Loc again concern 'inner Locatives' which either have the status of arguments of the predicate, or of σ, satellites. Gary and Keenan (1977: 114) provide a direct parallel to (76) in Kinyarwanda. Compare: (79)

a.

Yohani y-a-andits-e ku meza John he-past-write-ASP on table 'John wrote on the table with the pen.'

n-ikaramu with-pen

b.

Yohani y-a-andits-e-ho ameza John he-past-write-ASP-on table 'John wrote-on the table with the pen.'

n-ikaramu with-pen

The typology of adverbial satellites

191

(79a) has the same ambiguity as (76a); but (79b) can only be interpreted as saying that John inscribed something on the table with the pen. Many of the examples from Kinyarwanda given in Gary and Keenan (1977), Dryer (1983), and Perlmutter and Postal (1983) likewise concern inner locatives, as in 'throw something into the water', 'send someone to school', 'sit on a chair'. The same is true for the examples given from Chimwi:ni in Kisseberth and Abasheikh (1977): 'spill water on something', 'bring, send, write something to somebody'. On the basis of this evidence, though admittedly incomplete, it seems that there is a good basis for the following hypotheses: 1.

(Hi) In languages in which Subj/Obj can be assigned to Loc and Temp, this is in principle restricted to Loc and Temp arguments and σ, satellites.

2.

(H2) In the exceptional cases in which σ2 satellites can receive Subj/Obj function, this is under the external pressure of some rule (such as Relativization) which is constrained to Subj and/or Obj terms.

To the extent that these hypotheses are correct they can be used as an additional criterion for distinguishing σ 2 satellites from σι satellites and arguments with similar semantic functions. These hypotheses seem to offer a possible explanation for some 'difficult' passives in other languages as well. In Ancient Greek, for example, there are some cases which are traditionally described as passives, but whose Subjects are not evidently Goals (cf. also Metz 1988). A n example of a passive Subject with the semantic function of Duration is the following: (80)

ede

treis

menes

already

three.NOM.PL

months.NOM.PL

epitetrierarchento

moi

be-trierarch-beyond-the-legal-time.MP.3PL

me. DAT

Ί have been a trierarch already three months beyond the legal time.' (Demosth. 50,23.) A n active counterpart (with the same meaning) is found in Demosthenes 50, 36:

192

Simon C. Dik et al.

(81)

epitetrierarcheka tettaras be-trierarch-beyond-the-legal-time.ACT. 1 SG four.ACC.PL menas months.ACC.PL

'He had been a trierarch four months beyond his legal time.' The passive construction in (80) suggests that the Duration term treis menes is closely associated with the predicate, and has the status of either an argument or a σ, satellite. This close association could in this case have been reinforced by the fact that the Duration term has the accusative case form, which is also the normal case form for Goal arguments.

4.2.1.2.

Predicate

formation

There are strong indications that predicate formation rules may not only affect the arguments of the input predicate, but also the σι satellites associated with the nuclear predication. On the other hand, σ 2 and higher satellites seem to fall outside the scope of predicate formation rules. Consider the following three phenomena: (i) incorporation, (ii) valency reduction, and (iii) satellite absorption. (i) Incorporation. Incorporation can be described as a form of predicate formation through which nominal predicates are incorporated into (derived) verbal predicates. Compare: (82)

a. b.

John goes to school John school-goes

The formation of the derived predicate in (82b) can be described as a process through which a nominal predicate corresponding to a Direction satellite is incorporated into the predicate. Across languages, certain satellites can be more easily incorporated than others. We expect that those satellites can be most easily incorporated which have the closest relation to the nuclear predication. A s a working hypothesis we may assume that only σ, satellites can be incorporated. Across languages we find two types of incorporation: 14

The typology of adverbial satellites (83)

193

Incorporated nominal corresponds to argument John bird-catches (= catches birds) a. Goal man-drawn car (= drawn by men) b. Agent fuel-powered engine (powered by fuel) c. Force it is snow-falling (snow is falling) d. Processed Incorporated nominal corresponds to satellite John knife-cut the meat (= with a knife) a. Instrument John school-went (= to school) b. Direction c. Manner You must quiet-sit (= quietly) John fast-ran to the station d. Speed John chair-sits in the garden e. Location

Most of the examples in (84) clearly involve σ, satellites. Only the incorporation of nominals with Location function might be reason for some doubt. Note, however, that in a construction such as: (85)

John sits on a chair in the garden

it is the inner Location represented by chair which can be incorporated rather than the outer Location represented by garden. Thus, we do not expect constructions of the form: (86)

*John garden-sits on a chair

This is fully parallel to what we found in the case of Subj/Obj assignment: Locations can only be incorporated when they entertain a close bond with the predicate, such that the resulting compound predicate designates a specialized S o A rather than an unmodified SoA located at some place. This can also be seen in a number of verbs in Dutch which feature an incorporated locative nominal: a. b.

paalzitten waterskien

c. d. e.

ijszeilen wadlopen schoolblijven

pole sitting, a record breaking game water skiing, a special kind of skiing, not just 'skiing on water' ice sailing, a special kind of sailing (mud) flat walking, a special kind of walking school staying, not simply 'staying in school', but 'staying after school' (as a punishment)

194

Simon C. Dik et al.

(ii) Valency reduction,

σ, satellites can also be involved in predicate

formation when a valency reduction rule removes a σ, satellite instead of an argument. In Risselada (1987) it is argued that the middle-passive in Ancient Greek, which has a number of different uses (e.g. direct reflexive, indirect reflexive, pseudo-reflexive, pseudo-passive and passive) can be accounted for systematically by a set of valency reduction rules. In most cases, of course, the term which is removed by a valency reduction rule is an argument, but in the case of the Indirect Reflexive Predicate Formation rule the Beneficiary is removed and implied in the predicate frame of the new predicate: (88)

Indirect Reflexive Predicate Formation rule input:

pred V A ct

output: meaning:

PREDVMP ( Χ Ι ) Λ 8 (X2)G0

( X I ) A 8 (X2)GO

(ysVn

% performs pred v in his own interest'

MP (middle-passive) is the formal marker of the reduced nature of these derived predicates. A n example is: (89)

stratos

paraskeuazetai

tas

naus

the army.suBJ

prepares.MP

the

ships.GO

'The army is preparing the ships for i t s e l f Risselada (1987: 1 3 1 - 1 3 2 ) suggests that the French example (90)

Jean s' est casse la John R E F L is broken the 'John has broken his leg'

jambe leg

(cf. Vet 1985) can be analyzed in a similar way as involving reduction of the σι satellite Beneficiary, where se is the reduction marker involved in various types of valency-reducing predicate formation rules. These facts suggest that the domain of predicate formation rules is not the nuclear predication, but the core predication, i.e. the nuclear predication extended by σ, satellites. (iii) Satellite absorption. The term 'satellite absorption' has been used in two senses within the context of F G . First, it has been used to indicate the process through which a satellite gets absorbed into the predicate in some predicate formation rule (Kahrel 1985). Compare:

The typology of adverbial satellites

(91)

a. b. c.

195

Mary washes these clothes *These clothes wash These clothes wash easily

The valency reduction rule which removes the Agent argument of transitive wash at the same time requires that something is added to the derived predicate. Where this something is a satellite such as easily, we may say that this Manner satellite has been 'absorbed' into the derived predicate in the process of predicate formation, so as to become an argument. We assume that satellites will be more easily absorbed in this way when they are more closely associated with the nuclear predicate. The second sense in which satellite absorption has been used is to indicate the historical process through which an original satellite gets reinterpreted as an argument of the predicate to which it was originally more loosely attached. This process has been assumed (Pinkster 1988a, Mulder 1988) to underlie the phenomenon that in Latin and Greek certain two-place predicates have non-accusative second arguments. The facts are as follows. In Latin, with verbs like dolere ('to grieve for'), the constituent which refers to the Source or Cause of the emotion is either marked by an ablative case form or by an accusative case form (cf. Pinkster 1988a): (92)

a.

b.

qui

sociorum

iniuriis...

doleat

who

allies.GEN.PL

wrongs.ABLPL

deplores.3SG

'who deplores our allies' wrongs.' (Cie. Ver. 3,6) meum casum luctum-que

doluerunt

my

grieved.3PL

misfortune.ACC.SG

sorrow.ACC.SG-and

'They grieved for my misfortune and sorrow.' (Cie. Sest. 145) Although in Classical Latin both the ablative in (92a) and the accusative in (92b) are best considered second arguments, from a historical point of view iniuriis in (92a) could be considered as a satellite of Cause or Source, which normally has an ablative case form. The use of the ablative for coding a second argument could then be understood by assuming that the satellite has gradually been absorbed into the predicate frame, creating a two-place predicate dolere while retaining its original case form. The use of the accusative can then be seen as due to the Principle of Formal Adjustment (Dik 1985): the new two-place predicate adjusts its formal expression to the prototypical expression model for second arguments, the accusative case form.

196

Simon C. Dik et al.

A similar development can according to Pinkster be assumed for compound verbs like antecellere ('to surpass'), which occur with either a dative or an accusative second argument. In this case a Beneficiary would have been absorbed into the predicate frame. Mulder (1988: 235) defends a similar hypothetical scenario for the development of Ancient Greek non-accusative second arguments: (93) Phase 1 Phase 2 Phase 3

relation to predicate satellite 2nd argument 2nd argument

semantic function expression non-Goal non-accusative non-Goal non-accusative Goal accusative

A difference between Latin and Greek in this respect is that in Greek even some of the predicates which are in the presumed Phase 2 allow Subject assignment to the second argument, while in Latin such Subject assignment is not allowed. Again, it could be assumed that historically, those satellites which are most closely associated with the nucleus will be most easily absorbed into the nuclear predicate frame.

4.2.1.3.

Semantic constraints on the occurrence of satellites

A third fact which points to the argument-like behaviour of σ, satellites lies in the restrictions on their occurrence, which were already mentioned in 4.1.4 above. In general, satellites seem to be more constrained in their occurrence, the closer they are to the nucleus of the clause. Most σ2 satellites occur freely with any type of SoA. When they are sensitive to [+ control] SoAs (as is the case for Purpose and Reason (both σ2 satellites)), they may nevertheless occur freely with any such SoA. For the occurrence of σ, satellites more specific constraints must often be formulated. For example, Risselada (1987: 130) points out that intrinsically benefactive verbs in Greek, such as 'eat' or 'drink', which refer to [+ control] SoAs, can nevertheless not be extended by a Beneficiary. In Dik (1975: 97) a subcategorization of Manner adverbs is given according to the specific types of SoAs with which they may occur. Similar subcategorizations are given in Piatt and Piatt (1972) and Allerton and Cruttenden (1978). From these subcategorizations it is clear that rather specific semantic features of both

The typology of adverbial satellites

197

predicate and Manner adverb are essential for describing the privileges of occurrence of 'inner' satellites. A last example of such more specific restrictions is provided by satellites of Source, Direction and Path, which are mainly restricted to movement predicates. A further argument can be derived from the uses of the Latin ablative case. The ablative case is used for both 'outer' and 'inner' satellites with various semantic functions. Especially the occurrence of σ, satellites in the ablative is strongly dependent on the semantics of the nuclear predication.15 On the basis of such facts as these it could be argued that a predicate frame has a number of implied slots for 0! satellites which may but need not be filled. Moreover, the type of SoA is a property of the whole core predication (including the σ, satellites) rather than a property of only the nuclear predication. Thus, the following constructions have the same nucleus but describe two different SoA types ([+ con][+ dyn][- tel] in (94) and [+ con][+ dyn][+ tel] in (95)), as evidenced by the different possibilities of adding σ2 satellites: (94) (95)

John drove from Amsterdam to Paris *for hours/in five hours John drove along the highway for hours/*in five hours

We saw in this section that 'inner' (σ,) satellites in many ways behave differently from 'outer' (σ2) satellites, are closer to arguments, and are more strongly associated with the semantics of the nuclear predication. This corresponds to the overall semantic difference between σ, and σ2 satellites as conceptualized in this article: σ, satellites specify additional features of the nuclear SoA, while σ2 satellites serve to 'localize' the (specified) SoA in relation to temporal, spatial, and cognitive parameters.

4.2.2.

Ordering and position differences

Differences in the ordering and position of predicate and predication satellites are extensively discussed in Quirk et al. (1985: 5 1 1 - 5 1 2 ) . They note that what we call predication satellites are relatively free to occur in either sentence-initial or sentence-final position, as opposed to predicate satellites, which cannot freely occur in initial position:

198

Simon C. Dik et al.

(96)

a. b. a. b.

(97)

She kissed her mother on the cheek ?On the cheek, she kissed her mother She kissed her mother on the platform On the platform, she kissed her mother

These ordering differences can be interpreted as 'iconically' reflecting the relative scope differences between the satellites, in the sense that 'outer' or 'higher' satellites take 'inner' or 'lower' satellites in their scope. Further evidence for such a reflection of scope differences is found in the ordering of predicate and predication satellites relative to the predicate: (98)

4.2.3.

a. b. c. d.

She kissed her mother on the cheek on the platform On the platform, she kissed her mother on the cheek ?*She kissed her mother on the platform on the cheek ?*On the cheek, she kissed her mother on the platform

Paraphrase possibilities

In English, predication (σ2) satellites allow for a paraphrase by means of corresponding nouns, as in the following examples (from Mackenzie and Hannay 1982): (99)

a. b.

(100) a. b.

I met Sheila in the park The place that I met Sheila was the park I met Sheila at three ο 'clock The time that I met Sheila was three ο 'clock

Such paraphrases are not possible with predicate (σ,) satellites: (ι ο ι ) a. b.

/ approached the lion with great caution *The way that I approached the lion was great caution

(102) a. b.

I cut the meat with a knife *The instrument that I cut the meat was a knife

The typology of adverbial satellites Predication satellites allow a paraphrase with occur, happen, whereas predicate satellites do not: (103)

The event of John's travelling took place in Europe last summer frequently although he was ill because he hadn 't had a holiday for years

(104)

The event of John's travelling took place

take place,

199 or

*to Italy *by train *for Mary Again, these differences can be interpreted as reflecting the higher degree of independence of σ 2 satellites with respect to the content of the predication.

4.3.

Proposition (σ3) vs. Illocutionary (σ4) satellites

The difference between proposition satellites and illocutionary satellites resides in the fact that the latter specify or modify the illocutionary force of an utterance, whereas the former modify the propositional content. Since propositions are within the scope of the illocution, we may expect that the differences between the two groups of satellites centre around their behaviour with respect to the illocution. That this is indeed the case can be derived from some of the features which Greenbaum (1969) lists for illocutionary satellites: (i) They may appear in front of questions: (105)

Seriously, how do I look?

(ii) Many may appear in front of imperative and optative clauses: (106) Honestly, let's not tell him about it

200

Simon C. Oik et al.

In contrast, proposition satellites never occur in these positions: (107) (108)

*Hopefully, how do I look? *Probably, let's not tell him about it

Proposition satellites in general presuppose the speaker's positive commitment to the truth of the proposition he presents. These satellites are therefore largely restricted to declarative sentences. They can be said to operate inside the illocutionary layer. Illocutionary satellites, on the other hand, operate outside the illocutionary layer, which is shown in their relative freedom to occur with any kind of sentence type.

4.4.

Negation

As we saw in section 4.1.3, predicate and predication satellites (σ, and σ2 respectively) fall within the scope of the tense and (objective) modality operators as well as within the scope of (sentence) negation. In this section we examine more in detail the behaviour of the four satellite classes with respect to negation and show that negation provides useful criteria for distinguishing the different subcategories. Before dealing with the behaviour of the satellites proper it is useful for our discussion to make some observations about the interpretation of negation in sentences without satellites. In doing so, we use some elementary concepts of model-theoretical semantics. Compare: (109) a. b.

Mary danced Mary did not dance

In model-theoretical semantics (109a) is usually interpreted as follows: the referent denoted by Mary ([Mary]) belongs to the extension of the predicate dance ([dance]): in other words, [Mary] 0 [dance]. The negation (109b) indicates that Mary does not belong to the extension of the predicate dance: [Mary] 0 [dance]. For sentences containing a two place predicate: (no)

a. b.

Mary hit the cat Mary did not hit the cat

The typology of adverbial satellites

201

The interpretation is that in (110a) the pair belongs to the extension of the predicate hit, that is the set of all the pairs of individuals between which a hit-relation holds: 0 [hit]. Sentence (110b) can be interpreted as 0 [hit], that is, the pair consisting of the referents of Mary and the cat does not belong to the set of pairs between which a hit-relation holds. For our discussion it is important that while in nuclear predications sentence negation affects the relationship between the argument(s) and the predicate, this is not the case in sentences containing a predicate satellite (core predications). Compare: (HI)

a. b.

John died of pneumonia John did not die of pneumonia

In (11 ib), at least in the dominant reading, 16 John's dying is not negated. Rather, the negation affects the relationship between the nuclear predication {John died) as a whole and the Cause satellite. In ( m a ) the Addressee is invited to accept the truth of John died and to accept pneumonia as being the cause of this event. In ( n i b ) the Addressee can safely admit that John died, but the Speaker blocks the possibility of ascribing John's death to pneumonia. ( m a ) can be interpreted as follows: the state of affairs denoted by the nuclear predication ('John died') belongs to the set of events which are caused by pneumonia, whereas ( n i b ) states that the state of affairs ('John died') is not a member of the set of events which are caused by pneumonia. The other subcategories of predicate satellites can all be interpreted in the same way. For example: (112) Mary did not dance elegantly Here the negation indicates that the predicate elegant(ly) cannot be applied to the nuclear predication {Mary danced), or equivalently: the state of affairs denoted by Mary danced is not a member of the set of states of affairs to which the predicate elegant can be truthfully applied. The interpretation of (113) is analogous to that given for (11 ib) and (112): (113)

John did not accompany Mary as her lawyer

202

Simon C. Dik et al.

Here the interpretation is that John accompanied Mary, but not as her lawyer. A s in ( i i ib) and (112) the negation does not affect the nuclear predication, but the predicate satellite as her lawyer. In all the examples discussed so far the predicate satellites can be semantically interpreted as (second-order) predicates which take the nuclear predication of the sentence as their argument. 17 It is a general property of such predicate satellites that they are primarily affected by negation. Let us now turn to the second category distinguished above, that of predication (σ2) satellites. The point to be examined here is whether the behaviour of this type of satellite differs from that of predicate satellites. Compare the following examples: (114)

a. b. c.

John arrived at eight ο 'clock John did not arrive at eight ο 'clock At eight ο 'clock John did not arrive

The most striking difference from predicate satellites is that the (nuclear) predication can now be taken to be negated, i.e. fall within the scope of the negation: (114b) may mean that John arrived but that this event did not take place at eight o'clock, or that there was no event of John's arriving taking place at eight o'clock at all (the latter is the only interpretation of (114c)). In section 4.2.2 it was already pointed out that predication satellites have greater freedom of occurrence than predicate satellites; they can be placed in front of the nuclear predication, in which case they do not fall within the scope of the negation in negative sentences such as (114c). A different intonational pattern may have the same effect in sentences such as (114b). The fact that in (114) there are two negative sentences that correspond with one positive sentence has raised the question whether the positive sentence is structurally ambiguous. Kraak (1966: i56ff.J, for example, explains the existence of the two readings of the negative sentence by admitting that the corresponding positive sentence is structurally ambiguous. This ambiguity is supposedly due to a different distribution of focus. 18 Consider such a pair as the following two Dutch sentences: (115)

a.

Fred reisde met de treinvanwege Fred travelled with the train because.of 'Fred travelled by train because of the rain'

de the

regen rain

The typology of adverbial satellites b.

Fred reisde

niet

met

Fred travelled

not

with the

de

203

trein vanwege

de

regen

train because.of

the

rain

'Fred did not travel by train because of the rain' Sentence ( 1 1 5 b ) can be interpreted in at least two ways: (i) Fred travelled by train, but not because of the rain; (ii) Fred did not travel by train and this was because of the rain. Kraak admits two other readings: (iii) the 'strong' one w e mentioned in note 16 ('Fred did not travel at all, not by train and not because of the rain') and (iv) one in which Fred travelled but not by train and not because of the rain. In our opinion these two interpretations, and especially the latter one, are highly improbable. The analysis w e propose is as follows. For (115a) w e adopt the following structure: (116)

extended predication core predication I [[[nuclear predication]

I predicate sat]

1

predication sat]

1 II

Fred travelled

by train

because of the rain

The readings (i) and (ii) of (115b) correspond to the negation of the relations II and I, respectively. In the latter case it is said that Fred does not travel by train. If relation II is negated, it is claimed that Fred travels by train, but not because of the rain. The nuclear predication refers to a S o A ('Fred travelled') which is said to belong to the states of affairs which take place by train. The core predication refers to a S o A ('He travelled by train') which is said to pertain to the set of states of affairs which take place (or do not take place) because of the rain. We agree with Kraak that the different possible interpretations of (115a) and (115b) correspond with different focus assignments. ( 1 1 5 b ) could be used to answer the following questions: (117)

a. A : By which means of transport did Fred travel because of the rain ? Β: I do not know, but he did not travel by train

204

Simon C. Dik et al. b. A : Because of what did Fred (did Fred not) travel by train? Β: Because of the rain

The reason why predication satellites can combine with a positive or negative core predication may be that circumstances expressed by predication satellites such as because of the rain can influence somebody's behaviour in a positive or a negative sense: both cancelling activities which were previously planned and doing things which were not planned because of the rain creates a new situation. In the case of predicate satellites it is more difficult to imagine how these might specify something which would deny the validity of the nuclear predication. Consider an Instrument satellite such as that in: (118)

((He did not kill the duckling) with an axe)

Such a construction can hardly be interpreted as 'he used the axe for not killing the duckling'. This shows that with an axe, as a predicate satellite, is an integral part of the specification of the SoA, rather than providing some kind of setting for the S o A as already established in the predication. A l l predication satellites combine with negative or positive core predications, but also with positive or negative nuclear predications: (119)

a. b.

Mary left because of the rain Mary did not leave because of the rain

Depending on the distribution of focus, (119a) may answer the following questions: (120)

a.

What did Mary do because of the rain ? (left is Focus in (119a))

b.

Because of what did Mary leave? (because of the rain is Focus)

In the same way (119b) can answer the following questions: (121)

a. b.

What did Mary not do because of the rain ? (She did not leave) Because of what did Mary leave ? (I don't know, she did not leave because of the rain).

The typology of adverbial satellites

205

In these cases either the nuclear predication (or at least its predicate) carries focus, or the predication satellite. In core predications the nuclear predication (or its predicate, or one of its arguments) never carries Focus. As far as proposition satellites and illocutionary satellites are concerned we can be brief. Neither of them can fall within the scope of negation, but the (nuclear, core or extended) predication they combine with can be positive or negative (with the corresponding Focus distributions).19 For example: (122) a. b.

In my opinion we should do it/should not do it We should not do it in my opinion

(122b) does not allow the interpretation that we should do it, but not in my opinion. In the same way, illocutionary satellites (frankly, since you ask me) cannot fall within the scope of the negation. The same explanation can be given for the fact that proposition satellites such as possibly, probably do not have negative counterparts (*impossibly, *improbably). From these facts we can conclude that negation can be used as a criterion for distinguishing predicate satellites from predication satellites: the latter combine with negated predications, the former do not. The two satellite types have in common that they can both themselves be negated. Negation can also be used to distinguish proposition and illocutionary satellites from the rest (they never fall within the scope of the negation), but negation cannot serve as a criterion to distinguish between proposition satellites and illocutionary satellites since they show the same behaviour with respect to negation.

5.

Other parameters

In the preceding sections we have tried to demonstrate that the position satellites occupy within the hierarchical structure of the clause to a large extent determines their behaviour. The hosting layer of the clause was the parameter along which we tried to arrive at a typology of satellites. This is not the only parameter that is relevant for such a typology, and we will discuss two more in this section. The first of these concerns the internal complexity of satellites. The second concerns the restrictive/non-restrictive opposition.

206

Simon C. Dik et al.

5.1.

The internal structure of satellites

Satellites can have different degrees of internal complexity, as can easily be demonstrated by means of the following examples: (123) Mary danced beautifully (124) Mary danced because she didn 't want to talk to John The complexity of the Manner satellite beautifully in (123) is that of a mere predicate, while the complexity of the Reason satellite because she didn't want to talk to John in (124) is that of a finite sentence. Degrees of internal complexity of satellites can be defined in terms of the same hierarchical clause model (see section 3.1) as was used earlier to define the satellite types in terms of what may be called their external complexity, the layer of the clause they attach to. The construction with the highest degree of complexity is the whole clause structure. It contains a construction with a lower degree of complexity: a proposition. This proposition again contains a construction with a lower degree of complexity: a predication. Within the predication one finds two units of still lower complexity: predicates and terms. By peeling off layers from the hierarchical model of the clause one encounters all types of construction that can provide the structure of a satellite. It may be useful to give a formal definition of the several constructions mentioned here: (125)

Clause Proposition Predication Term Predicate

(E,: [ILL (S) (A) (X,: [etc.] (X,))] (E,)) (X,: [(e,: [predß (χ,: [etc.] (χ,))] (e,))] (X,)) fo: [predß (χ,: predN (χ,))] (e,)) (xi: predN (xO) predß

A predicate designates a property or relation, a term an individual, a predication designates a state of affairs, a proposition a potential fact, and a clause designates a speech act. Not only are these different layers, both formally and semantically, relevant for the construction of main clauses, they also constitute a typology of embedded constructions. 19 This typology can be applied to the satellite types we have been concerned with. Consider the following series of examples:

The typology of adverbial satellites (126) (127)

John met Peter repeatedly John met Peter on the platform

(128) (129)

John met Peter after leaving the train John met Peter because he wanted to talk to him

207

The satellites in ( i 2 6 ) - ( i 2 9 ) have all been classified as predication satellites (σ 2 ). They do not differ with respect to the layer that hosts them, but they do differ with respect to their internal complexity. The Frequency satellite repeatedly in (126) has the internal complexity of a predicate. It designates a property (repeated) of the state of affairs described in the main clause. The Location satellite on the platform in (127) has the internal complexity of a term. It designates an individual (the platform), on which the state of affairs described in the main clause is situated. The Time satellite after leaving the train in (128) has the internal complexity of a predication. It designates a state of affairs (John's leaving the train) with respect to which the state of affairs described in the main clause is located in time. The Reason satellite because he wanted to talk to him in (129) has the internal complexity of a proposition. It designates the propositional content ( Ί want to talk to Peter') that motivated the occurrence of the state of affairs described in the main clause. Just as w e can distinguish between several types of predication satellites by looking at their internal complexity, so can w e distinguish between several types of predicate satellites, proposition satellites, and illocutionary satellites. The following table gives an overview of the possibilities by crossclassifying satellites according to the two parameters discussed so far: external structure (ext) and internal structure (int). Note that this table is not intended to give an exhaustive listing of all satellites discussed in the preceding sections, but to give one illustrative example of each satellite type that results from combining the two parameters. (130) ext 1

Internal and external structure of satellites int —>

Predß

χ

e

predß e

Manner

Beneficiary

Force

Frequency

Location

Circumstance

Reason

X

Attitude Manner

Source

Evidence Circumstance

Motivation

ILL

Beneficiary

X —

Reason

Apart from providing a second parameter for the classification of satellites, the specification of the internal complexity of a satellite predicts the

2o8

Simon C. Dik et al.

kinds of distinctions that can be expressed within it, such as (in)definiteness in the case of satellites that have the internal complexity of a term, temporal distinctions in the case of satellites that have the internal complexity of a predication, and propositional attitudes in the case of satellites that have the internal complexity of a proposition.

5.2.

Restrictive and non-restrictive satellites

Sometimes a distinction is made between restrictive and non-restrictive satellites (e.g. Rutherford 1970; Quirk et al. 1985; Hannay and Vester 1987; Hengeveld this vol. [1989]). Rutherford (1970) as a matter of fact uses these terms to make the distinction between our σ 2 and σ3/σ4 satellites. Most of his tests have already been discussed in section 4.1 above. Quirk et al. (1985) only devote one paragraph to the topic, since, as they say, the distinction largely overlaps with their distinctions between Adjuncts, Disjuncts, etc. The opposition restrictive vs. non-restrictive is, in their opinion, only relevant for their Adjuncts (our σ, and σ 2 satellites). In fact all their examples of nonrestrictive Adjuncts are of the σ 2 type. Disjuncts (σ3 and σ 4 satellites) are necessarily non-restrictive. In Hannay and Vester (1987) the examples with either a restrictive or a non-restrictive adverbial clause are all of the σ 2 type; the same applies to Hengeveld (this vol. [1989]). The interaction between satellite type and restrictiveness may thus be represented as follows: (131) 0! σ2 σ3 σ4

restrictive

non-restrictive

nzeirtpredxiXzoJCeO π 2 βι:[predication]:σ 2 (e,) -

π 2 ε,:[Predication] (e,), σ 2 (e 2 ) π3 X,:[ext pred] (Χ,), σ 3 (Χ 2 ) ττ4 Ε,:[Proposition] (Ε,), σ 4 (Ε2)

Predicate satellites ( o j necessarily contribute to the specification of the state of affairs: they are necessarily restrictive. Predication satellites may either restrict the nature of the S o A through providing it with time/space coordinates, or provide additional information to the S o A as already defined. Higher-level satellites necessarily provide additional information, pertaining to the speaker's evaluation of the (nature or occurrence of) the SoA, or to the communicative intentions with which the S o A is presented to the addressee.

The typology of adverbial satellites

209

This is in accordance with the fact that the task of representing the S o A ends at the level of the extended predication. The distinction restrictive vs. non-restrictive can also be interpreted as a difference in scope: restrictive satellites fall under the scope of the corresponding operators; non-restrictive satellites do not fall under the scope of these operators. We thus find confirmation for the status of the negative operator discussed in 4.4 above: predicate satellites are necessarily within the scope of negation, predication satellites may be within or outside the scope of the negation, and higher-level satellites necessarily take the negation in their scope.

6.

Conclusion

We argued in this article that the layered structure of the clause, as proposed in Hengeveld (1989), provides a natural framework for the subcategorization of satellites. On mainly semantic grounds, we made a fourfold distinction between: (132)

(a) Predicate satellites (θ]) specify additional properties of the S o A designated by the nuclear predication. (b) Predication satellites (σ2) serve to localize the S o A as defined in the core predication with respect to temporal, local, and cognitive dimensions. (c) Propositional satellites (σ3) specify the attitude of the speaker vis-ä-vis the fact designated by the proposition. (d) Illocutionary satellites (σ4) specify or modify the illocutionary force of the speech act in which the proposition is presented.

We then showed that this division into four layers correlates with a variety of differential coding and behavioural properties of satellites, and provides a natural place for many observations which have been made in the literature on adverbial constituents. Negation was shown to interact in interesting ways with this layering of satellites. Finally, it was noted that satellites can be further subcategorized on the basis of their internal structure, and that the restrictive/non-restrictive contrast is distinctive only for predication satellites. Many problems concerning the place of satellites in the layered model of the clause remain to be explored. For one thing, we have hardly touched upon

2io

Simon C. Dik et al.

the problem of the relations between different satellites at the same layer (e.g., the relations between temporal and local satellites, both predication satellites). A l s o , certain theoretical issues have been left unresolved. For example, the question whether satellites can be analyzed as predicates over the units which they take in their scope (as proposed in Vet 1986), or should be regarded as some type of modifiers different from predicators (as in Dik 1989). In spite of these open ends, however, w e believe that w e have demonstrated that the layered clause model provides for a more adequate typology of satellites, and conversely, that the natural way in which a variety of distinctions can be captured in this conception of clause structure reinforces the validity of the clause model itself.

Notes This contribution originally appeared in Layers and Levels of Representation in Language Theory, Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds.), 25-70. 1990. ι.

We are grateful to Jan Nuyts and Machtelt Bolkestein for critical remarks leading

2.

See Hengeveld (this vol. [1989]) for a more elaborate description of this model.

to improvements of the text. The status of the illocutionary component in (4) is based on Hengeveld's analysis. Compare Dik (1989) for a slightly different view. In the present article we take the layered structure of the clause as described here as given, and consider what implications this would have for the analysis of satellites. Obviously, the layered model itself is under discussion as well (see, for example, Bolkestein 1989). 3.

Compare the distinctions made in Dik (1989).

4.

Bolkestein (1989) argues that predications may contain temporal satellites without having a Tense operator. This might lead to a reconsideration of the analysis presented here.

5.

In Hengeveld (this vol. [1989]) a fifth satellite type is distinguished at the level of the clause, which is used in order to account for the expression of interclausal relations. We restrict ourselves to the level of the clause here.

6.

Most of the definitions given in the following sections are taken from Dik (1989).

7.

Combination of two Spatial setting satellites is possible only when one of them specifies a subregion of the other, as in in Rome on the Forum, in Amsterdam

The typology of adverbial satellites near the Central

Station,

211

etc. These might well be analyzed as internally

complex single satellites. 8.

For formal differences between Reason and Cause satellites in Latin, compare Pinkster (1988b) and Bolkestein (1990).

9.

For the German data, see Bartsch (1972, 1976).

10. For discussion of this example, see Levinson (1983: 256). 11. This difference between quoniam on the one hand and quod/quia appears to be relevant only in classical literary Latin. 12. See Dik (1978) and the discussion in Dik (1989). 13. Note that the verb takes the 'instrumental' voice when Subj is assigned to Temp. This is also reported for Kalagan. 14. For general discussion of incorporation see Mardirussian (1975), Dik (1980), Anderson (1985). 15. See Vester (1983) for what this implies for the relations between satellites of Manner, Cause, and Instrument, which in Latin can all be expressed in the ablative. 16. It is sometimes claimed that sentences such as (112) have another, 'strong' reading, paraphrasable as: 'Mary did not dance elegantly, because she did not dance at all'. In this interpretation the negation does have the nuclear predication in its scope. This kind of reading, however, is quite exceptional in normal usage. We will return briefly to this point at the end of this section. 17. In Vet (1986) it is argued that (certain) satellites can also be analyzed as secondorder predicates in the underlying structure of the clause. In Dik (1989) the satellites are associated as modifiers or specifiers with the layer which they take in their scope. We leave this difference unresolved here, since it does not affect the main theme of this article. 18. On the interaction between negation and focus, see also Bossuyt (1983). 19. See Hengeveld (1990) and Bolkestein (1990) for applications of this typology.

References Allerton, David J., and Alan Cruttenden 1974

English sentence adverbials: Their syntax and their intonation in British

1978

Syntactic, illocutionary, thematic and attitudinal factors in the intona-

English. Lingua 44: 133-168. tion of adverbials. Journal of Pragmatics 2: 155-188.

212

Simon C. Dik et al.

Anderson, Stephen R. 1985

Typological distinctions in word formation. In Language Typology and Syntactic Description. Vol. Ill: Grammatical Categories and the Lexicon. Timothy Shopen (ed.), 3-56. Cambridge: Cambridge University Press.

Bartsch, Renate 1972

Adverbialsemantik. Frankfurt am Main: Athenäum.

1976

The Grammar of Adverbials. Amsterdam: North-Holland.

Barwise, Jon, and John Perry 1983

Situations and Attitudes. Cambridge, MA: The MIT Press.

Bell, Sarah J. 1983

Advancements and ascensions in Cebuano. In Studies in Relational Grammar 1, David M. Perlmutter (ed.), 143-218. Chicago: University of Chicago Press.

Bellert, Irene 1977

On semantic and distributional properties of sentential adverbs. Linguistic Inquiry 8: 337-351.

Bolkestein, A. Machtelt 1989

Latin sentential complements from a Functional Grammar perspective. In Amsterdam in Louvain. Papers read by members of the Department of Latin of the University of Amsterdam at the 5"1 colloquium on Latin Linguistics (April 1989). S. H. Rummens (ed.), 15-27. Preprints, University of Amsterdam.

1990

Sentential complements in Functional Grammar: Embedded predications, propositions, utterances in Latin. In Layers and Levels of Representation in Language Theory: A Functional View, Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 71-100. (Pragmatics and Beyond 13.) Amsterdam: John Benjamins.

1991

Causally related predications and the choice between parataxis and hypotaxis in Latin. In New Studies in Latin Linguistics,

Robert

Coleman (ed.), 427-445. Amsterdam: John Benjamins. Bolkestein, A. Machtelt, Casper de Groot, and J. Lachlan Mackenzie (eds.) 1985a

Syntax and Pragmatics in Functional Grammar. Dordrecht: Foris.

1985b

Predicates and Terms in Functional Grammar. Dordrecht: Foris.

Bossuyt, Alain 1983

Historical Functional Grammar: An outline of an integrated theory of language change. In Advances in Functional Grammar, Simon C. Dik (ed.), 300-325. Dordrecht: Foris.

The typology of adverbial satellites

213

Dik, Simon C. 1975 The semantic representation of manner adverbials. In Linguistics in The Netherlands 1972-1973, Albert Kraak, (ed.), 9 6 - 1 2 1 . Assen: Van Gorcum. 1978 Functional Grammar. Amsterdam: North-Holland. 1980 Studies in Functional Grammar. London/New York: Academic Press. 1985 Formal and semantic adjustment of derived constructions. In Predicates and Terms in Functional Grammar, A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds.), 1-28. Dordrecht: Foris. 1989 The Theory of Functional Grammar. Part 1: The Structure of the Clause. (Functional Grammar Series 9.) Dordrecht: Foris. Dryer, Matthew S. 1983 Indirect Objects in Kinyarwanda revisited. In Studies in Relational Grammar 1, David M. Perlmutter (ed.), 129-140. Chicago: University of Chicago Press. Gary, Judith Olmsted, and Edward L. Keenan 1977 On collapsing grammatical relations in universal grammar. In Grammatical Relations. Syntax and Semantics 8, Peter Cole and Jerrold M. Sadock (eds.), 83-120. New York: Academic Press. Greenbaum, Sidney 1969 Studies in English Adverbial Usage. London: Longman. Hannay, Michael, and Elseline Vester 1987 Non-restrictive relatives and the representation of complex sentences. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 39-52. Dordrecht: Foris. Harder, Peter 1989 Tense, semantics and Functional Grammar. Working Papers in Functional Grammar 30, University of Amsterdam. Hengeveld, Kees 1990 The hierarchical structure of utterances. In Layers and Levels of Representation in Language Theory, Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 1-24. Amsterdam: John Benjamins. this vol. Reprint. Layers and operators in Functional Grammar. Journal of Linguistics. 25 (1): 1 2 7 - 1 5 7 , 1989. Kahrel, Peter 1985 Some aspects of derived intransitivity. Working Papers in Functional Grammar 4, University of Amsterdam.

214

Simon C. Dik et al.

Keenan, Edward L. 1972

Relative clause formation in Malagasy (and some related and not so related languages). In The Chicago Which Hunt: Papers from the Relative Clause Festival, Paul Peranteau, Judith N. Levi, and Gloria C. Phares (eds.), 169-189. Chicago: Chicago Linguistic Society.

1976

Towards a universal definition of 'subject'. In Subject and Topic, Charles N. Li (ed.), 303-333. New York: Academic Press.

Kisseberth, Charles W., and Mohammad Imam Abasheikh 1977

The

object relationship

in Chimwi:ni,

a Bantu

language.

In

Grammatical Relations. Syntax and Semantics 8, Peter Cole and Jerrold M. Sadock (eds.), 179-218. New York: Academic Press. Kraak, Albert 1966

Negatieve zinnen; een methodologische

en grammatische

analyse

[Negative clauses: A methodological and grammatical analysis]. Hilversum: De Haan. Levinson, Stephen C. 1983

Pragmatics. Cambridge: Cambridge University Press.

Mackenzie, J. Lachlan, and Mike Hannay 1982

Prepositional predicates and Focus constructions in Functional Grammar. Lingua 56: 43-57.

Mardirussian, Galust 1975

Noun incorporation in universal grammar. Chicago Linguistic Society 11: 383-389·

Metz, Gerben 1988

Het Griekse passivum: Semantische en pragmatische aspecten van Subjecttoekenning in het Grieks. [The Greek passive: Semantic and pragmatic aspects of Subject assignment in Ancient Greek], M. A. Thesis, University of Amsterdam, The Netherlands.

Mirikitani, Leatrice 1972

Kapampangan Syntax. Hawaii: Oceanic Linguistics Publication 10.

Mulder, Hotze A. 1988

Non-accusative second arguments of two-place verbs in Ancient Greek. In In the Footsteps of Raphael Kühner, Albert Rijksbaron, Hotze A. Mulder, and Gerry C. Wakker (eds.), 219-236. Amsterdam: Gieben.

Perlmutter, David M., and Paul M. Postal 1983

Towards a universal characterization of passivization. In Studies in Relational Grammar, vol. 1., David M. Perlmutter (ed.), 3-29. Chicago: University of Chicago Press.

The typology of adverbial satellites

215

Pinkster, Harm 1984

Latijnse Syntaxis en Semantiek. Amsterdam: Grüner.

1988a

Non-accusative second arguments of two-place verbs in Latin.

1988b

Lateinische Syntax und Semantik. Tübingen: Francke Verlag.

Cuadernos de Filologia Cldsica 21: 235-245. Platt, John Τ., and Heidi Κ. Platt 1972

Orientation of manner adverbials. Paper, Monash University.

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1972

A Grammar of Contemporary English. London: Longman.

1985

A Comprehensive Grammar of the English Language. London and New York: Longman.

Rijksbaron, Albert 1976

Temporal and Causal Conjunctions

in Ancient Greek. Amsterdam:

1986

The pragmatics and semantics of conditional and temporal clauses.

Hakkert. Some evidence from Dutch and Classical Greek. Working Papers in Functional Grammar 13, University of Amsterdam. Risselada, Rodie 1987

Voice in Ancient Greek: Reflexives and passives. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 123-136. Dordrecht: Foris.

Rutherford, William E. 1970

Some

observations

concerning

subordinate

clauses

in

English.

Language 46: 97-115. Vester, Elseline 1983

Instrument and Manner Expressions in Latin. Assen: Van Gorcum.

1988

Enkele opmerkingen over de plaatsing van temporele bijzinnen. [Some remarks on the placement of temporal subclauses]. In Propemptikon, Afscheidsbundel

W. J. H. F. Kegel, A. Maria Van Ε φ Taalman Kip,

Daan den Hengst, and Johannes J. L. Smolenaars, (eds.), 116-120. Amsterdam: University of Amsterdam. Vet, Co 1985

Passive, reflexive, and causative predicate formation in French. In Predicates and Terms in Functional Grammar, A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds.), 49-69. Dordrecht: Foris.

1986

A pragmatic approach to Tense in Functional Grammar. Working Papers in Functional Grammar 16, University of Amsterdam.

Further reading

Matthew P. Anstey and J. Lachlan Mackenzie This major study of the typology of satellites in Functional Grammar gives an excellent example of the principles of layering in action. Whereas satellites in pre-layering Functional Grammar were distinguished above all by their semantic functions, they are now shown to differ in their position of attachment, with satellites being found at all four layers. Hengeveld (1992: 10) proposes a fifth type of satellite, the 'clause satellite', "through which the speaker locates his utterance within the context of the discourse and thus restricts the possible perlocutions of this utterance". A similar position is adopted by Cuvalay-Haak (1997: 79), who applies such satellites as first of all, in response to your question and to conclude my story to her highest layer, the Expression. In Functional Discourse Grammar (Hengeveld 2004), the rich layering of the representational and interpersonal levels makes it possible to introduce various refinements among such satellites: thus to conclude my story will be a satellite to an episode, in response to your question a satellite to a Move and first of all a satellite to an Act. For an application of the five-layer typology of satellites to concessive clauses, see Crevels (2000), and for an application to conditional clauses in Arabic, in which the distinction between expected and hypothetical conditions is related to attachment to the propositional and predicational layers respectively, see Cuvalay (1996). Hengeveld and Perez Quintero (2001) apply the F G adverbial satellite analysis to a sample of over 35 European languages. Hengeveld (1997) extended the analysis of adverbials in a typological direction, showing the existence of an implicational hierarchy across languages, such that the presence of the category of adverb in a language is most likely at layer 1 and least likely at layer 5. He also showed that there is generally a correlation between positioning at a lower layer and proximity to the main predicate in expression. Perez Quintero (2002) takes the research in another direction, that of a detailed corpus investigation of adverbial satellites in one language, English. Hengeveld (1998) had shown that subordinate clauses with dependent time reference are across languages more likely to be expressed by means of dependent verb forms than those with independent time reference and that subordinate clauses with a presupposed content are more likely to be expressed by means of dependent verb forms

Fu rthe r reading

217

than those with a non-presupposed content; Perez Quintero shows that this trend is reflected in the relative corpus frequency in English. Anstey (2002) demonstrates that F G scholars in the nineties were not mutually consistent in their assignment of phenomena such as satellites to layers, and tended to confuse syntax with semantics. He then proposes a radically new approach to satellites and to operators, their grammatical equivalents, by establishing a limited number of semantic domains. These are combined into a bottom-up layered semantic structure, which mirrors a layered cognitive structure. The resultant system solves many of the inconsistencies that had accreted from earlier work, leading for example to the claim that Layer-1 satellites are non-existent. K w e e (2003) also discusses various issues concerning adverbial clauses and their representation as satellites in FG. There has also been ongoing interest in adnominal satellites, such as the man with the limp. Keizer (2004) offers an excellent overview of this research and concludes with an argument "to simplify term structure by analysing all (syntactically) optional information (i.e. all modifiers) within the clause as satellites" (2004: 14).

References Anstey, Matthew P. 2002

Layers and Operators Revisited. Working Papers in Functional Grammar. 77.

Crevels, Mily 2000

Concession: A typological study. Ph. D. diss. University of Amsterdam, The Netherlands.

Cuvalay, Martine 1996

A classification of conditional satellites. In Complex Structures: A Functionalist Perspective, Betty Devriendt, Louis Goossens, and Johan van der Auwera (eds.), 149-175. (Functional Grammar Series 17.) Berlin/New York: Mouton de Gruyter.

Cuvalay-Haak, Martine 1997

The Arabic

Verb. A Functional

Grammar Approach

to Verbal

Expressions in Classical and Modern Arabic. (Functional Grammar Series 19.) Berlin/New York: Mouton de Gruyter.

218

Matthew P. Anstey and J. Lachlan Mackenzie

Hengeveld, Kees 1997

Adverbs in Functional Grammar. In Toward a Functional Hacia

una Lexicologia

Funcional,

Lexicology:

Gerd Wotjak (ed.),

121-136.

Frankfurt am Main: Peter Lang. 1998

Adverbial clauses in the languages of Europe. In Adverbial Constructions in the Languages of Europe, Johan van der Auwera (ed.), 335-419, (Empirical Approaches to Language Typology/Eurotyp 20-3). Berlin/ New York: Mouton de Gruyter.

Hengeveld, Kees, and Maria Jesus Perez Quintero 2001

Descriptive adequacy in Functional Grammar. Revista Canaria de Estudios Ingleses. 42: 103-117.

Keizer, M. Evelien 2004

Term structure in FG: A modest proposal. Working Papers in Functional Grammar. 78.

Kwee, Tjoe Liong 2003

Adverbial clauses: Scope, recursion, discourse, and Functional Grammar. Leuvense Bijdragen. 91 (3-4): 319-347·

Perez Quintero, Maria Jesüs 2002

Adverbial

Subordination

in

English:

A

Functional

Approach.

(Language and Computers: Studies in Practical Linguistics.) Amsterdam: Rodopi.

Semantic content and linguistic structure in Functional Grammar. On the semantics of 'nounhood' Peter Harder

i.

Semantics and underlying structure1

This article is based on some views on the format of linguistic description, which I shall begin by summarizing and placing in relation to FG. Part of the aim in doing so will be to try to clarify the position of Functional Grammar with respect to the relation between linguistic expression and semantic and pragmatic factors (cf. also Van der A u w e r a 1990; Bolkestein 1992; Hannay 1991; Mackenzie and Keizer this vol. [1991]). In this first section I shall take a critical look at the very influential descriptive practice, introduced by Chomsky, which is based on the notion of underlying form. With this practice goes a metaphor that has become part of the pattern of thinking of linguistics, in terms of which empirical findings are thought of as belonging on the 'surface', whereas the linguist's abstractions, located at a 'deeper' level, get us closer to the fundamental aspects of language. Description in terms of this pattern involves an initial level that is set up by the linguist through procedures that cannot be exactly specified (cf. the argument against discovery procedures in Chomsky 1957), and moves by rules towards the actual expressions. Theoretical description therefore starts at a level defined and selected by the theoretician himself, and ends up with his data. In the historical context, the liberation from the crippling restrictions of post-Bloomfieldianism was clearly necessary for American linguistics. However, it created a descriptive procedure that involved a permanent risk of circularity: explaining a 'surface form' (= an empirical datum) by means of an underlying form with no independent empirical status is descriptively empty except to the extent it involves what Chomsky called 'significant generalizations'. In other words, every time the linguist sets up an 'underlying' construct he incurs a sort of debt to the reader, which he has to make good again - and it is often very difficult to ascertain whether he succeeds in doing so.

220

Peter Harder

Instead of this descriptive pattern I think a functional grammar ought to be based on an initial level describing the semantic structure of the clause. The purpose of the derivational process would then be to show how this semantic structure is related to the actual morphosyntactic form. With this view goes a rejection of the tripartite view of linguistic structure that thinks of a phonological, a syntactic and a semantic 'level', where the same object, the same string, is analyzed in three different ways, one of which is syntactic. Instead, the fundamental distinction is between (semantic) content and (formal) expression. This means that the area of syntax is split between the two sides: a 'content syntax' describing how content elements (word meanings and grammatical meanings) form larger semantic wholes, and an 'expression syntax' that describes how expression elements (word forms and grammatical expressions) combine into complex expressions. The problem of the 'borderline' between syntax and semantics, according to this view, is fundamentally misconceived, since some syntactic relations (relations between sentence constituents) are semantic because they involve relations between content elements - and some are not semantic because they involve relations (e.g. linear order) between expression elements alone. 2 The question is now how this principle relates to the derivational procedure of F G . In terms of the picture I have tried to present, F G has taken over from generative grammar the weakness inherent in the 'underlying-tosurface' descriptive practice. A s far as I can see, the otherwise fundamental difference between the two approaches would therefore in fact be better expressed if the F G underlying structure were reinterpreted as describing the semantic structure of the clause. Such a reinterpretation would fit better than one might think with the existing F G apparatus, since the so-called underlying levels of the F G descriptive system (terms, predicates, predications and propositions, etc.) are almost exclusively semantic in nature. The preferred phrase in F G is 'underlying structure',

which is quite compatible with the

view I am taking, once it is understood that the underlying structure is a structure of semantic elements. It is semantic content that is layered, and expression which is linear, as indicated by the expression rules; these should be understood not as the last step from abstract form to surface form, but as assigning expression to content elements. What this means can be illustrated if w e look at the traditional diagram of F G , the 'conveyor belt' moving from the underlying clause structure up to linguistic expressions (cf. for example Dik 1989: 53). Under the accepted interpretation of F G , there is strictly speaking no semantics in

The semantics of 'nounhood'

221

this picture. Dik (1989: 46) speaks of the underlying clause structure as containing "formal and semantic organization", which is true in the sense that the underlying structure contains elements that reflect both semantic and formal properties of the clause. But this is not sufficient to make the underlying structure qualify as a semantic structure - since the motivation for postulating underlying structure is fundamentally to derive linguistic forms in a typologically and intralinguistically satisfactory way. What motivates the postulation of an element in the underlying clause structure is the necessity of abstractions to bring out common features in languages (cf. the introduction of the definiteness operator, Dik 1989: 15-16). Indeed, a number of elements in the underlying clause structure are explicitly said to be non-semantic (e.g. category labels of predicates, cf. Dik 1989: 161-162). Semantics, therefore, would only come in if we added an interpretation component which took the underlying clause structure as input and did something to it. In itself, the underlying structure says nothing about the meaning of the clause (cf. also Janssen 1981) - it is just a self-contained calculus of uninterpreted elements, just like generative grammar.3 This I think is unfortunate for a theory which explicitly rejects the autonomy of syntax. According to the principles of F G (cf. Dik 1989: 7), syntax is seen as instrumental with respect to semantics - and therefore I think this instrumental relationship should be part of the actual theory of grammar. If we assume that the constructs at underlying level are really labels for content elements, that is exactly what we get. We then begin with an account of semantic relations within the clause and go on to describe how this complex semantic structure is realized in terms of expression, i.e. morphosyntactic form. If 'underlying clause structure' is reinterpreted as 'semantic clause structure', it also becomes subject to direct empirical control: just as the 'expression structures' at the output end of the conveyor belt have to be recognizable as mirroring actual expressions of the language, the semantic (or 'content') structures at the input end have to mirror the way those expressions are understood. But it may be asked whether this does not turn the familiar underlying constituents of F G into something quite different. Why should abstract elements designed to account for form correspond to elements of meaning? The reason for this near-identity is simply that generally speaking, it is semantic intuitions which make plausible those abstractions on which the underlying structures are based. Asking why an aspect of form is present very often leads to an answer in terms of an aspect of meaning - whether this is explicitly recognized or not. Semantics is actually the major factor

222

Peter Harder

in justifying the underlying clause structure as it is presently understood, cf. also Siewierska (1992). What I suggest is basically that this semantic foundation of the grammar should be explicitly recognized. F G derivations would then be explicitly designed to do the central j o b of linguistics, namely relate the two basic empirical sides of language: content and expression, form and meaning. A s it is now, it takes us merely from a construct that the linguist has made up for himself to one end of linguistics, the formal end. Apart from the ever-present risk of circularity that is inherent in such a procedure, it makes the functionality of Functional Grammar less obvious than it should be. The two ways of understanding the derivational procedure of F G are depicted in figures ι and 2. THEORY

HEURISTICS

Figure 1. FG as conceived in terms of 'underlying-to-surface' description (The path of the arrows illustrates the risk of circularity)

The semantics of'nounhood'

223

LINGUISTIC EXPRESSION i1

Empirical control: Does the output of the grammar correspond to actual linguistic expressions (with the relevant semantic content)?

EXPRESSION RULES Specifying how the expression side of the content elements interact to form a complete clause expression

SEMANTIC CLAUSE STRUCTURE Consisting of content elements (=semantic instructions) organized into layers constituting the coded meaning of the clause

Empirical control: Does the organisation or content elements correspond to semantic intuitions (with respect to the relevant clause expression)?

Figure 2. Suggested reinterpretation: Semantically based FG

2.

Meaning as input to interpretation

A s a way of thinking about F G , the approach I am advocating puts coded meaning in a more central position in the theory, since the first stages of the derivation now consist of (labels for) coded meanings (= content elements). To give credibility to such an approach, we therefore need to be precise about the nature of such coded content elements. The view taken here is based on

224

Peter Harder

a pragmatic view of meaning, according to which coded semantic content consists in instructions designed as input to the interpretation process of the addressee (a view found in different forms in discourse representation theory (cf. Kamp 1979), procedural semantics (cf. Davies and Isard 1972; JohnsonLaird 1977) and relevance theory (cf. Sperber and Wilson 1986; cf. also Kirsner 1979). Again, this is in tune with some aspects of existing F G assumptions, such as reference (cf. Dik 1987 and 1989), which is understood as a process whereby the speaker gets the addressee to build and process a mental model in certain ways. This means that all aspects of linguistic meaning should be seen as instructions for mental, cognitive action by the addressee. What is needed as I see it is a full semantic theory based on this conception. There are two things I would like to stress about this view of meaning. The first is that in adopting such an approach, F G would offer an explicit place to the cognitive aspect of language that enjoys such favour in presentday linguistics. The crucial elements in grammatical description are good candidates for a semantic description based on cognitive activities involved in language understanding (cf. for example Wierzbicka 1988 and Langacker 1987a and b). The second thing is that if meanings are instructions for the addressee, semantics is defined through its place in relation to pragmatics. This, again, is fully in tune with the basic principles of F G (as expressed by Dik 1989: 7): just as syntax is instrumental with respect to semantics, semantics is instrumental with respect to pragmatics. The two instrumental relations can be paraphrased as follows from the 'productive' perspective that is built into FG: in making an utterance, the speaker begins with a communicative intention based on his goals in the speech situation (pragmatics); as a way of carrying out this intention he selects and combines some instructions coded in the language (semantics, including content syntax); in order to convey these instructions to the addressee he transforms them into linguistic expressions (expression syntax + phonetics). This pragmatic view of semantics, however, differs from a very ancient assumption about meaning, namely that meanings have to do with what expressions stand for. The most prestigious modern pattern of thinking within this tradition is the model-theoretic approach to semantics. The essential aspect of model theory is the existence of mapping relationships between expressions and states of a model world. Every time you change or add to expressions, the semantic aspects of the change are described with reference to what is the case in the model world. Putting it with commendable

The semantics of 'nounhood'

225

clarity in his 'Informal lectures on formal semantics', Bach (1989: 2) says: "1. Language has meaning, 2. Meanings are things that are not language... Meanings are the things that language is about."

walKvr

- -τ Λ τ ^ · Λ . λ?f\7\ Λ7\

Figure 3. The model-theoretic picture as drawn by Bach (1989) It is this picture that is disputed when we think about linguistic meaning as instructions for cognitive action. It does not imply that model-theoretic semantics is completely wrong, but it puts it in a position slightly outside the theory of natural language: not as a theory of linguistic meaning, but as a theory of correspondences between worlds and linguistic expressions. We need such a theory for independent reasons - but model theory is not a description of meaning as an aspect of language in relation to linguistic communication, model theory comes in at the stage after the instructions for cognitive action have been followed up by the addressee. When the addressee has grasped the instructions, and worked out a mental model on the basis of them, model theory describes what worlds that model is compatible with. Another way of saying the same thing is that linguistic semantics applies to instructions for processing, while model theory applies to the product stage. The scope relations between semantic elements then specify the 'input-output' relations in the production process: each instruction applies to what is inside its scope. According to this view, linguistic structures do not describe linguistic products, but potential process input (compare Hannay 1991, Mackenzie and Keizer this vol. [1991], and Bolkestein 1992). One argument for this is that at the product level one cannot distinguish between linguistic and contextual phenomena: all 'products', both on the content and expression side, are the result of applying linguistic instructions in the

226

Peter Harder

context at hand. One cannot distil either 'fortis' on the expression side or 'red' on the content side out of the linguistic product - but one can ask what contribution such elements make to the process of working out complete expressions and complete meanings.

•

CODED PROCESS INPUT EXPRESSION:

noun stem

put stem

put definite

General instruction:

'elephant'

'old' before elephant

article before 'old' and

Formulate...

pronounce [öi]

CONTENT:

entity

as charac-

and use this to

General instruc-

carrying

terized by

identify 'world-

tion:

elephant-

the property

Conceive of...

property

OLD

of-discourse' entity as being referred to

PRODUCT: —

f the old \ I elephant... I

>

(SYr?

ο>ΊΓ\ A

Figure 4. Content and expression as process versus product: (A very simplified illustration of the principle!) The difference between the model-theoretic conception of meaning and the conception argued here is schematically illustrated in figures 3 and 4. Figure 4 also illustrates the instrumental function of the expression rules in relation to the semantic structure: for every content element there is an expression rule specifying how the presence of the content element is reflected in the expression (cf. also Dik 1994). For discussion of the example analyzed in figure 4, see section 7 below.

3.

Terms and predicates

I shall now suggest a theory of 'nounhood' within an explicitly semantic version of FG. The distinction between noun-like expressions and verb-like expressions in F G has to do with the most basic distinction in the derivational procedure, that between 'term' and 'predicate'. That again is a version of the earliest distinction made in the history of semantic theory. Plato and Aristotle thought of a proposition, characterized as something that could be true or false, as divisible into 'onoma' and 'rhema'. These two concepts

The semantics of 'nounhood'

227

corresponded to either 'subject' or 'predicate' or 'noun' and 'verb', depending on the context. If w e look at the way Aristotle characterizes the notions in a celebrated passage in 'Peri hermeneias' w e find what I think is the central feature that motivates the distinction between terms and predicates, namely that a 'rhema' is "the sign of what is said of something else". A n 'onoma', by contrast, can stand alone without being said of something else (in which case, however, it does not say anything that is true or false - only the full proposition is truly complete). The semantic characteristic of a 'rhema' or in F G terminology a 'predicate' which renders it unable to stand alone is what gives rise to clausal semantic structure in the first place. The job of a predicate is to be applied to something that it does not say itself. Therefore something else must be there to say it - i.e. the term(s). Together, term and predicate constitute the nuclear predication - the basic structural configuration in the grammatical structure of the clause. I am not going to go further into the 'incompleteness' of predicates, the subject of the following being the relative completeness of the 'onoma'. One aspect of this property follows from the description of the incompleteness of the predicate: an 'onoma' can carry properties or relations as expressed by a predicate. Hence, an onoma is complete in designating, in itself, something which 'carries' the content it expresses; an onoma is not 'said of something else'. I propose to see this characterization as the crucial feature of what it is for a linguistic expression to characterize something as an entity. A n entity is a carrier of properties or relations. When a semantic element is, linguistically speaking, re-conceived as something which can carry a property or relation, it is 'entitled'. A property may be entified, when it is taken off the job of characterizing an entity and conceived of as standing on its own (Aristotle: a 'rhema', when spoken in itself, becomes an onoma). 'Onoma' corresponds less readily to F G 'term' than 'rhema' corresponds to F G 'predicate'; it corresponds better to what Lyons calls 'nominal expressions' or just noun phrases. A prototypical onoma is an expression referring to a person or thing, in other words a (prototypical) entity. 'Onoma' refers only to potential subject

terms, i.e. such expressions as designate entities

about which something is said. In F G the typical thing for a term to refer to is also an entity; but as pointed out by Mackenzie (this vol. [1992]), terms may also refer to places and times; and these are not obvious candidates for carrying properties and relations. Like everything else, they can be regarded as such - but then they are 'entified', i.e. conceived as objects in their own right

228

Peter Harder

(cf. Mackenzie 1987a). A s pointed out in Dik (1985), formal and semantic adjustment go hand in hand: turning an expression into a nominal expression goes with turning its content into something conceived as an entity. A s generally assumed, w e may look on entities as occupiers of locations or places these places are then defined in terms of some space, possibly also on a time axis. I think that the two ways of characterizing entities follow one from the other: a place-occupant is also a (potential) property-carrier and vice versa; but a full characterization of an entity would be to say that it is a 'place-holding property-carrier'. Thus, when you say 'the sky is red', 'red' is not a separate entity, but a property; when you say 'red is my favourite colour' you view 'red' as an entity, to which you can ascribe a property, and which has a place in a mental (sub)space occupied by colours. The Aristotelian categories were not always explicitly specified with respect to whether they applied to language or to those things the language was about. The continuing problems concerning the category of 'substance' may be seen as deriving from this ambiguity. What I am suggesting now is that if w e see linguistic meanings as instructions for cognitive action, the notion of 'substance' makes very good sense: when a nominal expression is used, the addressee is told to conceive of it as denoting a (potential) carrier of properties, in other words as a substance in this conceptual sense. This instruction is in contrast to the instruction associated with a predicate: the addressee is told to conceive of the content of the predicate as carried by something not denoted by the predicate itself. Predicates, therefore, do not denote substances, but must unload their content on a substance denoted elsewhere. On the other hand, if w e look for substance in the physical world, w e look in vain: when w e take all properties away, there is nothing left except space and time. With the cognitive trend in present-day linguistics, however, substantiality may be eligible for a revival. The lack of matching between the structure of what is out there and the structure of our way of talking and thinking about it points to a central feature of semantics: meanings are a way of interacting with the world, through the production of mental models; therefore having meaning does not imply having correlates in a world (or model) conceived of as independent of linguistic interaction. In interacting cognitively with the world, one of the basic abstractions w e make is the distinction between entities on the one hand and properties and relations on the other. This characterization is related to, but different from suggestions by Langacker (1987a and b) 4 and Wierzbicka (1988). Langacker's description

The semantics of 'nounhood'

229

focuses on the lack of a temporal dimension in the conceptualization associated with nouns; Wierzbicka (1988: 485) focuses on the fact that nouns create a category, a kind of thing, whereas adjectives add a feature without creating a new category. Both these characterizations of nouns go well with an assumption that a noun basically designates something as a carrier of properties/relations, something that has a local habitation (however derived and metaphorical), and a name. Being conceived of as an entity means borrowing the status of a thing-in-itself, as opposed to a process occurring in a thing or a feature that is predicable of it - hence the 'permanence' and 'category' aspects that distinguish nouns from adjectives and verbs.

4.

Term semantics: the FG apparatus and the instructional alternative

I am now going to take a critical look at the F G apparatus based on this view of meanings as constituting specifications for mental activity. M y basic view is going to be that in spite of the general pragmatic orientation of F G , there is some residual model theory in the framework, which ought to be reinterpreted in this pragmatic perspective. I see this as being related to the criticism offered in Mackenzie (1987b). Term semantics, as described in Dik (1987, 1989), operates with the notions variable,

restrictor, a typology of

entities, and finally with reference as tied to term semantics. A l l these belong, I think, in a pattern of thinking where correspondence with a (model) world is taken to be the stuff of which semantics is made. To begin with the notion 'variable': as I understand it, a variable is something that can potentially be replaced with various members of a set, as circumstances dictate. To say that a term has an indexed variable associated with it is therefore to say that the meaning of a term is defined in relation to a set from which the filler of the variable position is chosen. If such an assignment relation is not assumed, the notion 'variable' does not make sense. I think this description is perfectly true of typical terms. But it only applies once w e get to the stage where a linguistically designated entity is brought into a correspondence relation with some world - and as indicated above, I do not think this is the stage where linguistic meaning belongs. Instead of having a set already as part of the linguistic meaning, the linguistic instruction associated with a nominal expression involves an instruction to conceive of something as having substance, i.e. as a potential carrier of

230

Peter Harder

properties/relations. There is nothing variable about this. In order to comply with this instruction, the addressee has to set up a mental model, situated in an appropriate mental space and put an entity somewhere in it. I therefore propose that the omnipresent semantic element in term phrases should be the substance or entity component itself, seen as a way of conceiving - rather than any indexed entity that only arises as the product

of understanding.

Similarly, the notion of a 'restrictor' is based on the correspondence point of view. What is 'restricted' is the potential correspondence between the entity variable and the set of 'assignable' entities. In describing the potential reference of the term, we begin with an infinite number of entities, and with the help of the restrictor w e can throw some of them out. Again this is true, once w e assume a correspondence situation, but if w e persist in looking at it from the point of view of instructions for mental activity, what we get is a slightly different picture. If we have the 'entity' element and look at the descriptive (ideational) meaning, what it does is to add properties/relations to the entity - rather like a child filling in the colours in a picture. What the child does is not to restrict the picture, but to elaborate

it. Similarly, the job

of descriptive elements in the noun phrase is to add to the specifications for the mental model that the addressee is required to produce. The typology

of entities is similarly a typology of things in a model rather

than of instructions or specifications. The typology in Dik ( 1 9 8 7 and 1989) sets up a cover category of 'ensemble' which is divisible into sets and masses; individuals are regarded as a subtype of sets. The elimination of individuals is argued on the basis of simplification as well as on premises which reveal a set-theoretic bias inherited from model theory: since the subset relation is basically different from the membership relation, and language does not seem to respect this difference, we can account for the lack of linguistic distinctions by assuming that individuals stand in the same relation to a set as a subset does - in other words, individuals can be regarded as sets of cardinality one. I think there is an important truth in pointing out the linguistic affinity between sets and individuals. When you look at individuals, as denoted by linguistic expressions, from a logical perspective, the job of designating an individual is done via its properties, and properties are typically thought of as characteristic functions of sets rather than individuals. Yet linguistically speaking I do not see what sense it makes to say that 'a house' means a set of houses with cardinality one - as opposed to meaning an individual house. I think this issue can only be satisfactorily resolved from a linguistic point of view if, rather than presupposing the linguistic validity of logical distinctions

The semantics of 'nounhood'

231

in terms of subsets and members, w e focus on the distinctions made in terms of coded meaning. This, therefore, is what I shall try to do, based on the instructional view of meaning. The natural place to begin is with the super-category of 'ensemble'. I think the notion reflects a perfectly valid point, namely that in some classifier languages there are terms that are unspecified with respect to the count-mass distinction, so that e.g. Ί like banana' could mean either mass or unit(s) (cf. also Craig 1992). But as indicated in the diagram in Dik (1989: 123), the notion 'ensemble' really occupies the same place in the hierarchy as the notion of 'entity' itself - it covers everything that comes under the heading of 'entity'. Its usefulness is in providing a characterization that can be made to fit all the sorts of things w e may have inside our heads under one fairly fluid super-category. If w e look at it from the instructional point of view, it seems to me that this category becomes superfluous. According to the theory I have outlined above, all nominal expressions carry the instruction 'conceive of something that can carry properties'. If this is the only instruction that is associated with a nominal expression, w e are free to understand it as an individual, set or mass, as w e please. This is what w e find, as described by Dik, in the case illustrated with the 'banana' example. Nouns of that kind may then receive classifiers that make them either countable or uncountable; the noun itself is neither one nor the other. What happens when w e add the classifier, as I see it, is then the addition of further semantic instructions - among which is the one w e associate with countability. The distinction between 'count' and 'mass' nouns can be seen in instructional terms with 'count' as the marked term, embodying an instruction to conceive the substance in terms of units. If the choice [+unit] is made, one simultaneously has to choose between 'one' and 'more than one', singular and plural (on the elaboration of these choices, cf. Johnsen 1992). I prefer the term 'unit' to a description based on the notion 'bounded' (cf. also Langacker 1987) because four gallons need not be bounded, but they must constitute four units. The distinction between 'mass' nouns and 'ensemble' nouns then comes out as a distinction between 'substance' without any specification as between units or not (= 'ensemble'), and 'substance' with the unmarked choice 'minus unit' (= 'mass'). If you choose [+ unit], you simultaneously raise the possibility of the ambiguity between 'individual-units' and 'kindunits' : 'cats' may be used to refer to kinds of cat as well as to individual cats (some nouns may be compatible with a [+unit] interpretation with respect to

232

Peter Harder

kind-units but not individual-units (e.g. 'wine')). If you choose [-unit], everything coming under the category is denoted at the same time, regardless of potential division into either individual units or units denoting subcategories. Let me invoke at this point the familiar observation that there is no obligatory descriptive correlate to the count/mass distinction. This has been seen as indicating that this was not a semantic distinction, only an arbitrary grammatical boundary. Rather, as also underlined by Langacker, it means that the semantics has to do with cognitive activity, rather than features of the physical world. Like 'substance', the 'unit' instruction has to do with our way of playing around with things as opposed to being concerned with their actual properties (although the actual properties naturally favour certain conceptualizations, cf. Wierzbicka 1988). Thus, conceiving of something in terms of units is not the same thing as saying that it has the property of being divisible into units. 'Furniture' is clearly divisible into units, yet when we designate it with an uncountable term, we conceive of it in abstraction from this property of divisibility - a 'something with furniture-properties' rather than 'some units with furniture-properties'. The refusal to conceive of something in terms of units implies a certain conceptual distance, where the phenomenon in question actually does consist of reasonably salient units. The 'distance' effect of units fusing into a mass is especially striking when human beings are occasionally spoken of like this. When P. G. Wodehouse speaks of 'a seething mass of humanity', the non-unit conceptualization of 'humanity' goes together with the lexical meanings of 'mass' and 'seething' to convey a deliberate flouting of the dignity of the individual (the human unit!).

5.

Word classes

I would like to explore the implications of this view of term semantics for the description of word classes. This is a complicated question; Aristotle was by no means the last person to set up categories that were fuzzy with respect to the distinction between 'word class' and 'sentence constituent'. But I think the notion of word class has to do with the existence of a lexically specified relation between the two levels. One way of specifying this relation would be to say that the element of 'substance' or 'entity' is associated with the notion of 'head of a nominal expression' - a slot usually filled with nouns.

The semantics of'nounhood'

233

This would fit in with the basic distinction between 'onoma' and 'rhema' as reapplied within the domain of the nominal expression itself: modifiers are like 'rhemata' in that they unload their semantic content elsewhere, whereas the head designates the carrier as well as the property - or, more precisely, the 'carrier-with-the-property'. According to this view, nouns are not basically 'restrictors' or 'modifiers', but 'modifieds' - although they may function as modifiers. Nouns do designate properties, but in a noun the property co-exists with the 'entity' or 'substance' element, which means that they 'modify' themselves, as it were. In 'first-degree' substantives the 'property' and the 'entity' elements co-exist in such a way that the property is denoted only indirectly, as carried by the substance (as generally assumed in medieval grammar). For a language to have a word class of 'substantives' is then to have a class of lexemes which are lexically specified for use in nominal head position. The content of 'substance' or 'entity' is thus primarily associated with the 'head' position, but secondarily with all those predicates which belong in that position. That means that 'entity' is doubly coded: first in the head position, secondly in all those lexemes that are specialized for taking that position.5 This view does not contradict the 'functional' definition of predicate classes given in Dik (1989: 1 6 1 ) , but it adds a semantic component to it, so that we get a 'functional-semantic' characterization of word classes. In the cited passage, however, Dik contrasts his functional approach to what he calls a semantic description. In this context, however, it is clear that by a semantic definition Dik means a description in terms of correspondence with categories of the world. A s argued above, this may be a mistaken view of what semantics is, linguistically speaking. In relation to the typology proposed in Hengeveld (this vol. [1992]), the semantics proposed here fits in by showing different ways in which the 'entity' aspect of meaning may be associated with the descriptive content in individual languages. The traditional F G theory, inspired (as pointed out by Mackenzie 1 9 8 7 b ) by the position of Bach (1968), eliminates this formal and semantic variation between languages, seeing maximal flexibility as the basic structure - which means that all languages are seen as basically of Hengeveld's most flexible type (cf. Hengeveld, this vol. [1992]). This view of the semantics of nouns has some bearing on the discussion of whether nominal predicates are avalent or monovalent (cf. Mackenzie 1987b). A s I said in the beginning, I think linguistic structure begins with the

234

Peter Harder

semantic 'unsaturatedness' of predicates as opposed to terms. This means that the concept of 'valency', as expressing the relationship between a predicate and the terms that it belongs with, does not apply to terms. And if the description of substantives given above is correct, they are specially designed to serve the function of head of a nominal expression in term position - with the semantic correlate of designating a 'carrier' of a certain property. This 'saturatedness' or 'self-sufficiency' of substantives in their capacity as nominal heads means that they must, for semantic reasons, be understood as avalent, as claimed by Mackenzie for related reasons. Of course, nominal expressions, headed by nouns, can occur in predicative positions; but this must then be understood as a form of conversion. The 'substance' component then occurs twice in such predications: once in the subject and once in the predicate - but the relation of predication means that they are understood as one and the same substance. 'Speed kills' ascribes the action of killing to 'speed'; 'speed is a killer' views 'speed' as a substance which has the property of being (a unit of) another substance ('killer'). In relation to the point made by Mackenzie (this vol. [1992]), the question arises as to the relationship between designating places and entities. To attempt a thorough-going analysis of this very basic issue would take us beyond the scope of this article; I shall just indicate briefly how I see the relationship between the views taken above and those in Mackenzie's article. As pointed out by Mackenzie, there is a close philosophical affinity between an entity and the place it occupies, which makes it natural to use words for entities to refer to the place they occupy. Mackenzie's class of placedenoting nouns may be said to comprise all those nouns which lend themselves unproblematically to such usage. There is thus no necessary contradiction between saying that the 'entity' element is coded into the category 'noun' and saying that there is a category of place-denoting nouns. To the extent that a preposition is necessary to bring out this potential, it would also be natural to say that the 'entity' element is basic, and the 'place' element represents some sort of conversion. (Compare Thrane 1980: 5 iff for an account where the element of localization is common to the notions of 'category' [of entities] and 'place', and both are viewed from an instructional point of view.)

The semantics of 'nounhood'

6.

235

Reference

If w e assume that the instruction 'conceive of an entity' is part of nominal semantics, the question arises as to what the relationship is between 'entity' and 'referentiality'. A s suggested by Keizer (this vol. [1992]) and Mackenzie (this vol. [1992]) reference - understood as 'relation to correlates outside language' - cannot be understood as an exclusive property of nominal expressions or even terms. However, I shall argue that the concept of reference as currently used in F G depends on some questionable model-theoretic assumptions, and that a better account of the referential properties of linguistic expressions may be possible in the light of the discussion of instructional vs model-theoretic semantics in the beginning of this article. The prototypical form of reference is definite reference. A s illustrated by 'yesterday' and 'here', this is not limited to nominal expressions and entities. Definite reference is special in requiring the addressee (A) to perform an act of identification with something outside coded meaning, i.e. something in the hearer's general mental model of the world as conceived for the purposes of the ongoing communication ('world of discourse' will be used to cover this construct, to distinguish it from the physical world). To describe the working of definite reference, w e need to operate with a 'working model', one that is constructed as part of the j o b of understanding an individual utterance, as well as the world of discourse existing antecedently to the utterance. The term 'working model' is used on analogy with the notion of 'working memory'. Definite reference, in terms of this picture, is designed to establish a correspondence between a linguistic expression and something in the already existing world of discourse, via the working model. In the case of definite nominal reference, there is therefore a correspondence between an entity in the world of discourse existing independently of the linguistic expression used and the mental model worked out in understanding the expression. Or, more precisely, the mental model worked out on the basis of the linguistic instructions is merely a proxy for the entity to be identified. I would like to suggest that the term 'reference' is really only necessary for definite reference - because it alone involves the correspondence relation that is typical of the word 'reference' as used in the tradition, and the activities of 'restricting', 'picking out', etc., that go with that term. Dik's distinction between 'constructive' and 'identifying' reference recognizes this special quality of definite as opposed to indefinite reference, so there is little disagreement with the view argued here when it comes to definiteness. The

236

Peter Harder

question is, however, whether there is very much left for the word 'reference' to do, when one has put definite reference in a class of its own. The most obvious reason for having the term 'constructive' reference is that once a referent has been constructed, w e can refer identifyingly to it afterwards, so that identifying reference in a text may be seen as presupposing constructive reference, or 'referent introduction'. Whatever can be retrieved must be introduced first, one might say. However, this view implies that introducing something that can be referred to is a linguistically special j o b that needs a privileged place in our semantics. I do not think this is in fact the case. Meanings, as argued above, are instructions, not mental models. A s a result of those instructions, addressees work out models and relate them to the world in various ways which result in changes in their model of the world of discourse - and all aspects of the world of discourse are candidates for future identifying reference. 'Yesterday Jack killed a black sheep here' requires A to build a mental model involving correlates of all elements, including verb and adjective, and the addressee's picture of the situation is changed accordingly. But these are converted to entities only after the utterance. In the utterance itself, 'black' is not an entity but a property of the sheep - although one can go on to refer and say 'that colour is not represented among his own sheep'. Similarly, 'kill' is not an entity, but a relation between two entities - although one can go on to say 'that is a terrible thing to do to a sheep'. There are also correlates to higher syntactic levels up to and including the predication, the proposition and the speech act itself. If w e view those possible referents in pragmatic terms, however, they are merely aspects of the world of discourse after the utterance, not semantic units which are introduced as entities referred to. You can refer to a speech act just as you can refer to other events. For example, if you are knocked down from behind, you can say 'that's a cowardly thing to do!', where 'that' refers to an act of violence rather than a speech act. The principle is the same: the possibility of subsequent identifying reference does not imply that the assailant made a 'constructive reference' to the assault. He merely performed it, just as the speaker performed the speech act; and anything that has been performed can be referred to. To insist on the terminology of reference and restriction in the case of the whole speech act would take us back before Austin to the times when all utterances were supposed to describe an act, instead of performing it (cf. Van der Auwera, 1990: 28, w h o rightly describes this as an 'ontological confusion'; see also Van der A u w e r a 1992). The conclusion must be that

The semantics of 'nounhood'

237

w e do not need a notion of 'constructive reference' in order to account for 'identifying reference' - w e just need the ongoing process of constructing mental models that modify our picture of the world. But well-established patterns of thinking tempt us to think that at least in the case of prototypical entities, there must be a notion of reference that is common to definite and indefinite noun phrases. However, if w e are to accept a notion of reference over and above the notion 'entity', w e have to require something in the nature of a correspondence relationship, as far as I can see. Reference, if it is anything at all, must be a relation between an expression and something that it corresponds with. But seeing something as an entity does not mean that there is any necessary correspondence between the entity-denoting expression and something in the world of discourse. If w e take the utterance 'John did not see a policeman', the phrase 'a policeman' denotes something conceived as an entity, namely something belonging to the category 'policeman'. But there is nothing in any world that corresponds to this entity. It is part of the 'working mental model' that is rejected as constituting a picture of reality by virtue of the negation 'not'. Adding a notion of 'reference' to the notion 'entity' would not clearly add to our understanding of what is going on semantically. 6 In that case what w e have left are three different types of phenomena that w e need to distinguish. First, there is the model-building activity that constitutes the elementary groundwork of the process of understanding. The models subsume the whole 'product' aspect - the fact that in understanding and interpreting language you generate models which are there for future use. This is a basic fact about communicative interaction, more basic than reference and predication, more basic even than the distinction between linguistic and non-linguistic communication. Even dogs are able to form models of what is the case based on what they see, smell or are told by human beings. Secondly, w e have the coded conceptualization possibilities, aspects of the process of communication, feeding into the actual work of interpretation: setting up entities, ascribing properties, assigning illocutionary categories, etc. This is the realm of linguistic semantics, in that meanings are cues for doing this j o b properly. Thirdly, as a special aspect of the j o b of interpretation, there is definite reference, comprising those points at which the addressee is instructed to let the work of interpretation latch on to antecedently existing elements in the world of discourse.

238 7.

Peter Harder A reinterpretation of some examples

B y w a y o f c o n c l u s i o n I w o u l d like to g i v e a more concrete illustration o f h o w to set up semantic structures f o r terms, taking s o m e e x a m p l e s f r o m D i k and s h o w i n g h o w I think they should b e analyzed. I hope to show, on the one hand, that the d i f f e r e n c e is small, b e c a u s e I m e r e l y generalize a tendency w h i c h is already present; on the other hand, that the d i f f e r e n c e is significant, b e c a u s e it w i l l mean that semantics is promoted f r o m b e i n g an aspect o f structure to b e i n g the stuff o f w h i c h the basic structure is made. C o n s i d e r the discussion o f the phrase 'the old elephant' ( D i k 1989: 1 1 5 ) , w i t h the structure: (1)

(dlXj! elephant N (Xi) 0 : old A (Xi) 0 )

D i k then adds that the w h o l e term can b e spelled out as f o l l o w s : (2)

"definite singular entity x ; such that the property 'elephant' applies to x ; such that the property ' o l d ' applies to x "

A n d thirdly, it is g i v e n in the f o r m o f an instruction: (3)

Instruction f r o m S to A : Identify a single entity x, C l u e ι : Xj has the property 'elephant' C l u e 2: Xj has the property ' o l d '

T h e r e is v e r y little I disagree w i t h in the spelling-out and the instruction. W h a t I propose is essentially that an overtly semantic structure c o u l d take the place o f all three. T h i s structure, c o m i n g out r o u g h l y as f o l l o w s , w o u l d in itself s p e c i f y the task o f understanding f a c i n g the addressee: (4)

to-be-identified (one-unit (property 'old'-to-be-ascribed-to (elephantentity)))

In abbreviated form: (5)

d (1 (prop: ' o l d ' (ent: 'elephant')))

The semantics of 'nounhood'

239

Brackets indicate scope, corresponding to the structure proposed by Rijkhoff (1990); as discussed by Dik, interpretation of noun phrases starts from the head, working outwards, so that you take 'elephant-entity' first, 'old' afterwards, etc. In a sense, this follows from the entity-component: it is the entity which is being described, so to understand a property is to ascribe it to the entity, and not the other way round (cf. figure 4 above). Does this make a difference? Well, there are no indexed variables ranging over sets - because the instruction to conceive of an entity plus the instruction to identify a correlate to it is all that is needed to work out the right result. The indexed variable itself is not part of the instruction - it is something found as a result of successfully following the instruction. This means that co-reference, which is one reason for indexed variables, is also analyzed in an index-free way: it is not a matter of two noun phrases, an indefinite followed by a definite one, that both have the same index - it is an operation whereby the addressee locates a referent for the second, definite expression. Linguistic semantics deals only with the instruction, not with the two variables that are identified as a result of a successful follow-up by the addressee. As also suggested by Leech (1981), from a linguistic point of view co-reference is a one-way relation. There is no semantic function assigned to 'elephant', because it does not require something to unload its content on. What replaces the variable and the semantic function 0 is the notion 'entity' as input to interpretation.7 The set-theoretic bias in Dik (1989) comes out clearly in another example (1989: 117), specifying how to understand the phrase 'Buddhist Japanese': (6)

Instruction from S to A: Take the set {J} of Japanese Now single out that subset from {J}, whose members also have the property B.

In the picture argued in this article, the account would be as follows: (7)

Semantic (underlying) structure: >1 (prop: 'buddhist' (ent: Japanese))

Roughly, the corresponding instruction would be: form a model containing unit-entities with the property of being Japanese, with the property 'Buddhist', in the quantity of more than one.

240

Peter Harder

This is different from the instruction indicated by Dik, since there is nothing corresponding to the entity denoted by {J} = 'the set of Japanese'. Instead, there is a request to form a mental model with more than one unit of a certain kind in it. The criticism that I voice here is related to the criticism against the 'possible world' semantics for being psychologically implausible. Beginning with a model containing all actual and possible Japanese will not do for a semantics aiming at psychological adequacy, because it puts too great a load on the processing capacity of the mind (cf. Johnson-Laird 1986: 63). The essential point is that understanding an utterance with a term phrase does not begin with a universe of entities which is gradually cut down to size. It begins (apart from pragmatic expectations which may or may not be fulfilled) with a void, a 'building site', which is gradually filled out, as a result of guided mental activity.

8.

Conclusion

I have tried to show how an explicitly semantic structure as the basic level of F G , with linguistic meaning understood in terms of instructions for pragmatic interpretation, could make some of the basic principles of F G more visible in the descriptive practice itself. In the area of noun (or term) phrases, I have suggested some modifications in the understanding of semantic structure that makes it possible to simplify the description, e.g. by seeing indexing as an operation belonging outside the area of linguistic semantics - not as a matter of making sense, but as a matter of keeping track. In doing so, I have also tried to argue that this view of semantics may be a more plausible candidate for psychological adequacy than an approach based on set theory.

Notes This contribution originally appeared in Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.). Amsterdam: John Benjamins, 1992.

The semantics of 'nounhood' ι.

241

Thanks are due to Simon Dik, Johan van der Auwera and Lachlan Mackenzie for helpful comments on earlier versions of this article. Needless to say, they are not responsible for errors and imperfections in this version.

2.

I would like to make clear that 'content syntax' is not to be understood as a sort of Mentalese, constituting an abstract extra level over and above semantic content. Rather, it is to be understood as the discipline dealing with those aspects of semantic content that involve combinations of content elements. In terms of the instructional view of meaning it deals with the way the individual instructions interact to guide the whole process of interpretation. Semantics, on this view, can be divided into two subparts: an inventory of all the coded content elements of the language (whether grammatical or lexical), and the combinatorial

subpart,

which I have called 'content syntax', dealing with the construction of complex meanings out of simplex content elements. 3.

The present status of semantics in F G has actually given rise to some confusion. Janssen's picture can be contrasted with a diametrically opposite picture: the fact that the syntactic structures in F G are so clearly based on semantic criteria leads Van Valin (1990: 199) to say that the layered model has only semantics, and no syntax. The model I propose would make precise the place of semantics in relation to syntax and also make possible an investigation of the precise division of labour between clearly semantic structures and constituent structure such as those described by Van der Auwera (1990).

4.

Langacker's description of nounhood is that it profiles 'a region in some domain'; he uses the term 'entity' in a general sense, covering verbal and adjectival meanings as well. A s Langacker points out himself, the notion of 'region', however, is not predictive of what is designated by means of a noun (cf. Langacker 1987b), and therefore is not sufficient to single out the conceptualization that is unique to the class of nouns. But Langacker also speaks of nouns as designating 'things' in the most general sense. The characterization that I am suggesting, ('placeoccupying property-carrier'), is to be understood as my attempt to define thinghood in this general sense. This may be regarded as the resumption of a tradition going back to antiquity, cf. Harder (1990a).

5.

Determiners, as obligatory constituents of the noun phrase, also have this property. But it would take us beyond the scope of this article to go into a discussion of the relationship between determiners/pronouns and nouns as designators of entities.

6.

It would be natural to ask how to describe 'specific' noun phrases in a theory which lets go of referentiality; they are the most obvious cases of expressions that are referential without being definite, and also, in many languages, noun

242

7.

Peter Harder phrases which are interpreted as being correlated with a non-linguistic entity which the addressee is not instructed to identify; and on that reading 'specificity' is a second-order descriptive element, telling the addressee that only one among the potential designata of the phrase is meant. Again, this can be understood without any actual correlation being made by the addressee between language and the world - if he does, it is actually a bit like cheating. I realize that I am ignoring the problem of how to account for relational nouns, which may be seen as requiring more than one semantic function.

References Auwera, Johan van der 1990 Coming to terms. Habilitation thesis, University of Antwerp. 1992 Free relatives. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 329-354. Amsterdam/Philadelphia: John Benjamins. Bach, Emmon 1968 Nouns and noun phrases. In Universals in Linguistic Theory, Emmon Bach and Robert T. Harms (eds.), 90-122. New York: Holt, Rinehart, and Winston. 1989 Informal Lectures on Formal Semantics. Albany, NY: SUNY Press. Bolkestein, A. Machtelt 1992 Limits to layering: locatability and other problems. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 387-408. Amsterdam/Philadelphia: John Benjamins. Chomsky, Noam 1957 Syntactic Structures. Den Haag: Mouton. Craig, Colette 1992 Noun classifiers in a Functional Grammar perspective. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 277-302. Amsterdam/Philadelphia: John Benjamins. Davies, Julian, and Stephen Isard 1972 Utterances as programs. In Machine Intelligence, vol. 7, Donald Michie and Bernard Meitzer (eds.), 325-339. Edinburgh: Edinburgh University Press.

The semantics of 'nounhood'

243

Dik, Simon C. 1985

Formal

and

Predicates

semantic and

adjustment of

Terms

in Functional

derived

constructions.

Grammar,

A.

In

Machtelt

Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds.), 1-28. Dordrecht: Foris. 1987

A typology of entities. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 1-20. (Functional Grammar Series 6.) Dordrecht: Foris.

1989

The Theory of Functional Grammar. Part I: The Structure of the Clause. (Functional Grammar Series 9.) Dordrecht: Foris.

1994

Computational description of verbal complexes in English and Latin. In Function and Expression in Functional Grammar, Elisabeth EngbergPedersen, Lisbeth Falster Jakobsen, and Lone Schack Rasmussen (eds.), 353-383. Berlin/New York: Mouton de Gruyter.

Hannay, Mike 1991

Pragmatic function assignment and word order variation in a Functional Grammar of English. Journal of Pragmatics 16 ( 2 ) : i 3 i 155·

Harder, Peter 1990a

The semantics and pragmatics of reference. In Pragmatics

and

its Manifestations in Language, Lita Lundquist and Lone Schack Rasmussen (eds.). (Copenhagen Studies in Language 13.) Copenhagen: Handelsh0jskolens Forlag. 1990b

Tense, Semantics and Functional Grammar, In Layers and Levels of Representation in Language Theory: A Functional View, Jan Nuyts, Machtelt Bolkestein, and Co Vet (eds.), 139-164. Amsterdam: John Benjamins.

Hengeveld, Kees this vol.

Reprint. Parts of speech. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 29-55. (Pragmatics & Beyond New Series 23.) Amsterdam: John Benjamins, 1992.

Janssen, Theo 1981

Montague Grammar and Functional Grammar. In Perspectives

on

Functional Grammar, Teun Hoekstra, Harry van der Hulst and Michiel Moortgat (eds.), 73-97. Dordrecht: Foris.

244

Peter Harder

Johnsen, Lars 1992 Number, gender and countability in Danish: An FG approach. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 355-368. Amsterdam/Philadelphia: John Benjamins. Johnson-Laird, Philip N. 1977 Procedural semantics. Cognition 5: 189-214. 1986 Conditionals and mental models. In On Conditionals, Elizabeth Closs Traugott, Charles A. Ferguson, Judy Snitzer Reilly, and Alice ter Meulen (eds.), 55-75. Cambridge: Cambridge University Press. Kamp, Hans 1981 A theory of truth and semantic representation. In Truth, Interpretation, and Information, Jeroen Groenendijk, Theo Janssen, and Martin Stokhof (eds.), 1-42. Dordrecht: Foris. Keizer, M. Evelien this vol. Reprint. Predicates as referring expressions. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 1 - 2 7 . Amsterdam: John Benjamins, 1992. Kirsner, Robert 1979 Deixis in discourse: an exploratory quantitative study of the Modern Dutch demonstrative adjectives. In Discourse and Syntax, Talmy Givon (ed.), 355-375. (Syntax and Semantics 12.) New York: Academic Press. Langacker, Ronald W. 1987a Foundations of Cognitive Grammar. Vol I. Theoretical Prerequisites. Stanford: Stanford University Press. 1987b Foundations of Cognitive Grammar. Vol II. Descriptive Applications. Stanford: Stanford University Press. Leech, Geoffrey N. 1981 Semantics. Harmondsworth: Penguin. Lyons, John 1977 Semantics. 2 vols. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan 1987a Nominalization and basic constituent ordering. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 93-105. (Functional Grammar Series 6.) Dordrecht: Foris.

The semantics of 'nounhood'

245

1987b

The representation of nominal predicates in the fund. Working Papers

this vol.

Reprint. Places and things. In Layered Structure and Reference in a

in Functional Grammar 25, University of Amsterdam. Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 253-276. Amsterdam: John Benjamins, 1992. Mackenzie, J. Lachlan, and M. Evelien Keizer this vol.

On assigning pragmatic functions in English. Pragmatics

1 (2):

169-215. 1991. Rijkhoff, Jan 1990

Explaining word order in the noun phrase. Linguistics 28: 5-42.

Siewierska, Anna 1992

Layers in FG and GB. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds.), 409-432. Amsterdam: John Benjamins, 1992.

Sperber, Dan, and Deirdre Wilson 1986

Relevance: Communication and Cognition. Oxford: Blackwell.

Thrane, Torben 1980

Referential-semantic

Analysis: Aspects

of a Theory of

Linguistic

Reference. Cambridge: Cambridge University Press. Van Valin, Robert D „ Jr. 1990

Layered syntax in Role and Reference Grammar. In Layers and Levels of Representation in Language Theory, Jan A. Nuyts, A. Machtelt Bolkestein, and Co Vet (eds.), 193-231. Amsterdam: John Benjamins.

Wierzbicka, Anna 1988

The Semantics of Grammar. (Studies in Language Companion Series 18.) Amsterdam: John Benjamins.

Further reading Matthew P. Anstey and J. Lachlan Mackenzie The philosophical underpinnings of Functional Grammar are central to Harder's concerns. His article points out that the representations of FG, although claimed to be semantic, are actually a hybrid combination of semantic and non-semantic information, containing for example the syntactic category of the lexical items involved. He advocates a purely semantic underlying representation, the combinatorial aspect of which is dealt with by a 'content syntax' (cf. Anstey 2002). This syntax is opposed to 'expression syntax', which concerns the ordering of the items that are the output of the grammar. Moreover, Harder views meaning as a set of instructions to guide the addressee's interpretative process. This leads to a view of clause meanings as 'recipes' (Harder 1996: 214) - the layered representations are seen as being parallel to the instructions to a chef. The addressee's sense-making is comparable to the chef's food-making, just as interpretive skills are analogous to cooking skills; making sense is as "humanly unpredictable and context-dependent" (1996: 215) as is work in the kitchen. Harder brings the distinction between process and product into sharp focus. A s he writes (1996: 224), "the hierarchy may not be in the product, although it is necessary for the process". Throughout the history of FG, there have been different preferences among researchers: as Hengeveld (2004: 366) puts it, there is debate whether F G (and FDG) structures "should be interpreted as a representation of the communicative process itself, or as a representation of the linguistic units put to use in that process". Bakker and Siewierska (2004), for example, take the latter course in explicitly proposing a speaker model which seeks to follow the chronological sequence of the language user's activities. Mackenzie (2000) also takes some steps towards an incremental view of grammar. Hengeveld himself (2004: 366-367), however, while conceding that F D G interacts with dynamic conceptual and contextual components and aims to reflect the process of communication, emphasizes that F D G is a pattern model, representing static linguistic units in terms of their ideational and interpersonal functions. It should be noted that in these discussions, process refers to the formulation and articulation of an utterance, product refers to an actual utterance (or text, gesture, signing, etc.), and pat-

Further reading

247

tern refers to a grammatical model, such as F G , in an abstract, static sense (Fortescue 2004: 170 note 2). Further important contributions to the debate on the status of F G have been made by Nuyts, Fortescue, and Cornish. Nuyts's work, in an incremental cognitive model (called Functional Procedural Grammar, Nuyts 1992), is oriented to providing a "working hypothesis concerning the organization of the cognitive systems involved in a language user's production of discourse" (1994: 159). He emphasizes his belief that many phenomena, including layering itself, are preferably seen as cognitive than as grammatical (Nuyts 2004). Fortescue (2004; cf. also 2001) also contends that F G should take account of process, with due consideration to the complementary demands of process and pattern. For example, a complete account of Focus in Nootka, he argues, requires reference to Focus as an aspect of both language processing and language structure (Fortescue 2004). Cornish's (2002) proposals are akin to Harder's position in arguing that in anaphora it is the listener's perspective that must be understood as triggering the interpretative process. A l l four, Harder, Nuyts, Fortescue, and Cornish, are concerned with the interpretation of the F G / F D G model with respect to the classic triad of mind, language, and reality.

References Anstey, Matthew P. 2002

Layers and operators revisited. Working Papers in Functional Grammar. 77·

Bakker, Dik, and Anna Siewierska 2004

Towards a speaker model of Functional Grammar. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 325-364. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

Cornish, Francis 2002

Anaphora: Lexico-textual structure, or means for utterance integration within a discourse? A critique of the Functional Grammar account. Linguistics. 40 (3): 469-493.

Fortescue, Michael 2001

Pattern and Process: A Whiteheadian Perspective on Linguistics. (Human Cognitive Processing 6.) Amsterdam: John Benjamins.

248

Matthew P. Anstey and J. Lachlan Mackenzie

2004

The complementarity of the process and pattern interpretations of Functional Grammar. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 151-178. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

Harder, Peter 1996

Functional Semantics. A Theory of Meaning, Structure and Tense in English. Berlin/New York: Mouton de Gruyter.

Hengeveld, Kees 2004

Epilogue. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 365-378. (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

Mackenzie, J. Lachlan 2000

First things first. Towards an Incremental Functional Grammar. Acta Linguistica Hafniensia. 32: 23-44.

Nuyts, Jan 1992

Aspects of a Cognitive-pragmatic Theory of Language. On Cognition, Functionalism and Grammar. (Pragmatics and Beyond 20.) Amsterdam: John Benjamins.

1994

Epistemic modal qualifications. On their linguistic and conceptual

2004

Remarks on layering in a cognitive-functional language production

structure. Antwerp Papers in Linguistics 81. model. In A New Architecture for Functional Grammar, J. Lachlan Mackenzie and Maria de los Angeles Gomez-Gonzälez (eds.), 275-298 (Functional Grammar Series 24.) Berlin/New York: Mouton de Gruyter.

On assigning pragmatic functions in English

J. Lachlan Mackenzie and M. Evelien Keizer

i.

Introduction1

With the appearance of Dik (1989), linguists have acquired a cogent and lucid presentation of the current state of Functional Grammar (FG). The book provides a broad but also detailed coverage of all fundamental aspects of the model, so that it will be of value not only to theoreticians but also to those concerned with the practical description of languages. In the present article, w e have been inspired by the same duality of purpose: our argument is theoretical in nature, but is directed above all at practitioners in our ambition to test the extent to which one of the subsystems treated in Dik (1989) can be applied in the actual analysis of a language. The subsystem in question is pragmatic function assignment, and the language, English. F G is conceived as being embedded within a wider pragmatic theory of verbal interaction (Dik 1989: 12). Such a theory, the outlines of which are beginning to emerge in the pragmatic literature, will be designed to account for the regularities underlying the structured, cooperative, real-time activities of language users, and will be necessarily procedural in orientation. It will concern itself with language users' plans and goals, their motivations and strategies, their problem-solving techniques, their quest for contexts and for relevance. Contributions may be expected from specialists in cognitive psychology and artificial intelligence, from anthropology and micro-sociology, from semiotics and communication theory, indeed from pragmatics in the widest sense. A theory of verbal interaction will thus be a dynamic theory, a theory of operations and processes. A grammar, on the other hand, and F G is no exception in this respect, has a fundamentally different orientation: it seeks to describe and explain the outcome of operations and processes rather than those operations and processes themselves. It is static, not dynamic, and is essentially atemporal in its attempt to establish relationships between either virtual or attested expressions rather than follow the ongoing production or interpretation of utterances. If there appear to be certain analogies between paths through the grammar and plausible sequences of events in linguistic production, that may

250

J. Lachlan Mackenzie andM. Evelien Keizer

increase the 'psychological adequacy' of the model and its "modules", but the grammar can claim no more than to be written in a "quasi-productive" mode (Dik 1989: 52). The relationship between F G and the all-embracing theory of verbal interaction is thus problematic in nature, being an interface between a static and a dynamic system. Now, of all the various subcomponents of the grammar, there are two which obviously stand at the interface with the ambient theory of verbal interaction: one is the analysis of the linguistic expression as a speech act, with an indication of illocutionary force being incorporated into the representation of a clause (Dik 1989: 248, 254ff.J; the other, to be focused on in the present article, is the assignment of pragmatic functions. The very definition of pragmatic functions, as specifying "the informational value of different parts of the clause, in relation to the speaker's estimate of the pragmatic information of the addressee" (Dik 1989: 60), makes it clear that they are partially concerned with a dynamic quantity, i.e. one that can vary in time: the speaker's estimate of the pragmatic information of the addressee. In the following pages, we will submit Dik's (1989) proposals for the assignment of pragmatic functions (mainly to be found in ch. 1 3 , pp. 263-287, but also in chapters on constituent ordering and prosodic features) to a careful examination with respect to (i) the theoretical question of how the interface between the static grammar and the dynamic theory of verbal interaction is handled with regard to pragmatic functions; and (ii) the practical question whether the reader finds a set of proposals that can be operationalized in the analysis of linguistic material. In considering these matters, we will limit ourselves to the two intra-clausal pragmatic functions and their subtypes, i.e. Topic and Focus. The article will address three problems: firstly, the absence of a link between relevant aspects of the theory of verbal interaction and the pragmatic function assignment component of F G (section 2); secondly, the irreconcilability of the Given-New and the Topic-Focus distinctions (section 3); thirdly, the lack of any 'special treatment' for Topics in English (section 4). The final section offers new rules for PI-placement in English.

2.

The interface between F G and the theory of verbal interaction

The key notion common both to a grammar incorporating pragmatic function assignment and to a theory of verbal interaction is 'discourse'. This term,

Pragmatic fiinctions in English

251

however, is understood in two different ways, depending on which approach is taken. Firstly, a discourse (usually then a count noun) may be seen as the product of text-creating activity; alternatively, discourse (now usually a mass noun) may be seen as the ongoing text-creating process itself. Grammarians, with their interest in the outcome of real-time operations and processes, generally take the former view. So also Dik (1989: 266-267), who identifies stories, monologues, etc. as discourses. He cites the product of his own labours as an example, i.e. the book he has written, and its chapters, subsections, paragraphs and ultimately its individual clauses - all these are discourses, hierarchically organized with respect to one another. Through the identification of the clause as a mini-discourse, Dik paints a picture of a pyramidical structure, with an uninterrupted rank scale from clause to book. Each discourse so identified stands in a relation of 'aboutness' to an entity or group of entities in some 'mental world' (see also Dik 1989: 46). These entities are termed D-Topics, and have the property of topicality. Dik (1989: 266) further recognizes "topical elements" in discourses. These are not defined, but we take them to be the particular part or parts of each discourse that is/are referentially or denotationally linked to one or more D-Topics. Any of these topical elements may (or, importantly, may not) be singled out2 for special treatment with respect to form, order or prosodic properties. This singling-out is operationalized in F G by the assignment of one of the pragmatic functions Topic or Focus. These have corresponding effects on the operation of the expression rules. Note, as a difficulty, that topical elements are thus determined per discourse (which may be much longer than a clause), while pragmatic functions are assigned per clause. Alongside this grammarian's approach to discourses as the outcome of verbal activity, Dik (1989: ι β η ϊ ί . ) also adopts a view of discourse as an ongoing text-creating process. Here he suggests that as each stretch of discourse 3 unfolds, so a "topic store", in some unspecified abstract sense, is "gradually ... filled with D-Topics as these are introduced". We take it that the unmentioned agent of the two passives in the quotation is the speaker/writer, at work in real time. A s Dik's presentation continues, it becomes increasingly manifest that he does indeed intend the reader to imagine discourse production in real time: "... at some point... for the first time ... then ... go on to talk about..." (1989, ch. 1 3 . 3 , par. 3). Thus, in one and the same section 1 3 . 3 , Dik both claims that D-Topics are phenomena that can be recognized retrospectively by the grammarian and assumes that the speaker/writer has D-Topics at her disposal, ready to

252

J. Lachlan Mackenzie and M. Evelien Keizer

be marshalled in ongoing discourse. Here, then, the reader finds himself at the cutting edge, right up against the interface between F G and the ambient theory of verbal interaction. The notion of D-Topic is, to judge by Dik (1989: 267, n. 5), inspired above all by Brown and Yule (1983: 71). These writers, however, differ from Dik in taking an exclusively process view on discourse (always used as a mass noun in their book). Indeed, they insist that the notion of a speaker having D-Topics at her disposal offers an unrealistic outlook on discourse production. Rather, they are interested in "the general pretheoretical notion of 'topic'" as "what is being talked about" in a conversation. They further find it unlikely that their notion of a discourse topic will be identifiable with one part of a sentence. Brown and Yule's discourse topic is thus in essential respects different from Dik's entity-based notion with its implied correspondence with sentential constituents. 4 The question now arises whether the real-time approach to pragmatic function assignment taken by Dik on p. 267 is conceptually compatible with most of the rest of F G , which takes the grammatically respectable retrospective view of language structure. A s a model, F G is of course characterized by a 'quasi-productive mode', being biased towards production rather than comprehension, but not being designed "necessarily [to] simulate the various steps that a speaker takes in producing linguistic expressions" (Dik 1989: 52). In Dik's account of the assignment of pragmatic functions, however, he appears no longer to observe a gwasz'-productive mode but rather to provide the elements of a theory of ongoing production. To our view, this disturbs the conceptual unity of the model, a point to which w e will return in section 3, on the incompatibility of the pairs G i v e n - N e w and Topic-Focus. The central issue, from our perspective, concerns the relation between the range of available D-Topics and the instances of Topic assignment in the corresponding text. Given that topical elements are determined per discourse (in the clause-to-book sense adumbrated above), while pragmatic functions are assigned per clause, how is it determined which of the available topical elements is chosen as Topic? Let us consider whether either of the two perspectives taken, the retrospective view of the grammarian or the discourse analyst's interest in ongoing processes, offers an answer. Looking retrospectively at discourses as products rather than processes, and specifically at individual sentences, D i k (1989: 268ffJ provides a set of criteria with which to identify a Topic, and further to subclassify Topics into four classes. B y specifying various kinds of special treatment Topics may receive, he offers the practising linguist guidance in determining whether

Pragmatic fiinctions in English

253

to assign Topic or not. A s for the singling-out of topical elements referred to above, w e understand Dik's position to be that it is not the speaker/writer w h o singles out elements for pragmatic function assignment, but the analyst. It is he who allocates Topic, Focus, etc. to constituents in representations. But his criteria for so doing remain obscure. The other tack taken by Dik, which follows the real-time production of utterances in discourse, appears to offer above all a classification of D-Topics (i.e. entities) in terms of the result of their being mentioned. Not everything is fully clear (ch. 13.3, sec. 3): the first presentation of a D-Topic is termed a NewTop; 5 an entity so introduced is a GivTop (Given Topic); similarly, an entity that is mentioned, temporarily neglected and later revived is termed a ResTop (Resumed Topic). Let us conclude, however, that what is intended is a classification of the entities held in the store of D-Topics in terms of their history of use in corresponding discourse. What neither perspective offers the reader is understanding of the speaker/writer's motivation in choosing one term rather than another as Topic of a particular utterance in discourse. We believe that this is the missing and vital link between the two perspectives adopted, between the two views of language as product and language as process. On the one hand, the grammarian's backward-looking view takes us only as far back in time as the form that results from production. On the other, the discourse analyst's forward-looking view fails to make the jump from the arrangement of D-Topics to the actual choice of Topic. The only connection that is achieved is feedback from choice of Topic to the classification of the store of D-Topics. In outline, we see the interface as follows: speaker/

speaker/

writer's store of D-Topics

writer's GAP

choice

utterance 237, 239 inference, 16, 35, 180, 262-263 inferable, 128, 133, 259, 261, 263-265, 269, 272 inflection, 10, 15, 26, 48, 56, 62-63, 69, 181 instruction, 4, 223-225, 228-232, 234-236, 238-241, 246 intension, 112, 117, 1 3 1 - 1 3 2 , 141 intention, x, 2, 7, 208, 224, 276, 283284, 300, 308, 3 1 1 - 3 1 4 , 316, 323 interaction, ix, 27, 208, 211, 228, 237, 249-250, 252, 259-260, 276, 282-283, 308, 319. 328, 345 verbal interaction, 249-250, 252, 260, 282, 319, 328 interface, x, 250, 252-253, 281-282, 310, 312, 323, 325 interjection, 90-91 interpersonal, vii, ix-x, 2,4, 13, 22, 28, 34,42, 76, 80, 84, 166, 171, 183-184, 186-187, 216, 246, 277, 281, 296, 323, 345 interrogative, 35, 154-155. 187. 271, 285, 302-303, 306-307, 3 1 1 intonation, 180-181, 185, 202, 282287,313-314

irrealis, 35, 59-60, 69 knowledge, vii, ix, 13-14, 62, 256, 259-264,312, 3 3 1 - 3 3 2 language user, 246-247, 281, 293 layering, viii-xi, 1 - 3 7 , 4 1 - 4 2 , 4 7 - 4 8 , 64-65, 67, 75-76, 80, 84, 89-90, 99, 109, 129-130, 137, 160, 169-172, 200, 205-207, 209-211, 216-217, 220, 223, 240-241, 246-247, 276, 281, 290-293, 296, 299-320, 335 lexeme, 77, 108, 1 3 1 - 1 3 2 , 233, 327, 329-331, 333-338, 344-345 lexical, 15, 27-28, 36,47,49, 52-53, 56, 65, 76, 79, 85, 87, 90-91, 94, 96, 102, 108, 132, 152-154, 161, 166, 170, 172, 174, 176, 179, 182, 232, 241, 246, 284, 301, 309, 314, 325-327, 329-340, 344 lexical field, 330-331, 335-336, 340 Lexical Functional Grammar, xi lexicology, 325-340 lexicon, x, 23, 52, 54, 115, 119, 1 2 2 125, 129, 137-138, 144, 160, 166, 325-334, 337-340, 344-345 locality, 47-48, 58-62, 65-67, 189, 209-210, 229 locational, 29, 33, 35, 48, 57-61, 66, 69, 76, 82, 102, 1 3 1 - 1 3 2 , 141-142, 145-147, 149-151, 154-158, 161, 173, 176, 188-191, 193, 207, 292 marked, 50, 52, 54, 58-59, 63, 100, 131, 181, 184, 195, 231, 271, 333, 338 marker, 51, 55-56, 62, 64, 69, 76, 102, 153, 157-159, 194, 258, 268, 284, 290, 296 matrix clause, 21-23, 26-27, 33-34,

Index of subjects 36, 56-57, 66, 1 7 7 - 1 7 8 , 186, 206-207, 288, 290-292, 299, 309, 318 meaning definition, 3 2 9 - 3 3 1 , 344 memory, 235, 283 mental model, χ 16, 143-144, i66, 224-225, 230, 235-237, 240, 251 message, xi, 1, 3, 129, 262, 284, 287-290, 292-293, 305, 3 1 0 - 3 1 1 , 314,319 message management, 254, 276 metaphor, 60, 1 4 1 , 219, 229, 334 metonymy, 1 4 5 , 1 5 0 mitigation, 1 6 - 1 7 , 19, 26, 35, 274 modality, xii, 1 , 1 4 - 1 5 , 35, 3 7 , 4 1 . 180 boulomaic, 15 deontic, 13, 20 epistemic, 13, 19-20, 180 objective, 1, 1 3 - 1 5 , 35, 58, 107, 186, 200, 334 subjective, 1, 6, 1 4 - 1 6 , 18, 180 mode-theoretic semantics, 200, 224226, 229-230, 235, 327 modifier, 62, 64,76, 85, 87-88, 91, 93-94, 98, 100, 1 0 8 , 2 1 0 - 2 1 1 , 217, 233 module, x, 250, 281, 283, 293, 296, 300,310,312-316,319,323 Momentaneousness, 9, 68 mood, i, 7, 12, 19, 35, 56, 58-62, 1 1 8 , 170, 285 desiderative, 19 morphology, 1 , 1 2 , 20, 30, 35, 62, 328 morphosyntax, xi, 79, 107, 220-221 move, viii, 47, 1 1 4 , 143, 167, 175, 197, 216, 219, 281-285, 287, 292, 296-297,315-318,336 narrative, 2-4, 1 1 , 21, 300, 3 1 6 - 3 1 7

377

negation, 10, 35, 1 1 4 , 184, 186-187, 200-205, 209, 2 1 1 , 237, 292 nominalization, 24, 3 4 , 4 1 , 107, 1 1 1 , 124,137 non-restrictive, 1, 27, 30-33, 36, 169, 205, 208-209, 288, 290-291 noun, 219-242, passim collective noun, 52, 54-55, 63-64, 69, 75, 1 1 4 , 144 count noun, 54-55, 63, 75, 251 mass noun, 5 1 - 5 4 , 63, 68, 1 1 4 , 144, 231-232, 251-252 noun phrase, ix, 64, 75-76, 132, 227, 230, 237, 239, 241 set noun, 63-64, 77 nucleus, 196-197, 3 1 5 , 3 1 7 - 3 1 8 , 330 onomasiology, 226-227, 2 3 3 , 326-327, 331 ontology, 1 1 7 , 142-144, 159-160, 166, 236 operator, 1 - 3 7 , 4 1 , 4 7 , 49, 53-59, 61-67, 69, 1 0 9 - 1 1 0 , 1 1 3 - 1 1 4 , 1 1 8 , 122, 129, 132, 137, 1 5 1 - 1 5 2 , 1 6 1 , 1 6 6 - 1 6 7 , 1 7 1 - 1 7 2 , 184, 186, 200, 209-210, 217, 221, 274, 289, 296, 300-303, 305-309, 3 " , 3 1 3 , 318-320; see also specific operator types; satellites PI position, 1 2 7 - 1 2 8 , 1 8 1 , 197, 254, 256, 258, 263-264, 267, 269-272, 274, 287, 302-306, 308, 3 1 0 P2 position, 16, 287, 306-310 paradigmatic, 325, 3 3 1 , 335-336, 339 participant, 3, 5, 80, 95, 179-180, 260 participle, 3, 5, 80, 88, 95, 1 0 1 , 1 7 9 180, 260, 290, 303 particle, x, 1 6 - 1 8 , 2 4 , 4 1 , 100, 142 parts of speech, 79-103, 107-108,

378

Index of subjects

parts of speech (cont.), 232-233, 308, 310 perception, 3 6 , 4 1 , 160 perlocution, 14, 28, 216 phonology, x-xi, 220, 325, 328, 337, 339 place, 1 4 1 - 1 6 2 place-denotation, 144-145, 149-150, 1 6 0 - 1 6 1 , 234 place-reference, 1 4 1 , 145-146, 149-151 placement, 127, 180, 256, 269-270, 288 plural, 34, 54-57. 63-64, 69, 75, 122, 231 polarity, 7, 10, 13, 58, 308 potential fact, 4, 35-36, 109, 1 1 1 , 1 1 4 , 1 1 6 , 206 pragmatics, vii, ix, 130, 160, 224, 249, 272, 277, 296, 3 1 1 , 325, 328 pragmatic function, 69, 127-129, 133, 1 5 1 , 184-185, 249-274, 276-277, 286, 293, 338; see also Focus, Topic pragmatic information, 3, 250, 255, 259-266, 302, 338 pragmatic module, 293, 300, 310, 3 1 2 - 3 1 3 , 3 1 5 , 3 1 9 , 323 predicate, 109-133, passim nominal predicate, 23,47, 5 1 - 5 4 , 6 1 , 6 7 , 75' 7 9 . 8 1 , 8 7 , 92, 95, 1 0 0 - 1 0 1 , 107, 1 1 3 - 1 1 5 , 119, 1 2 2 - 1 2 4 , 126, 1 3 0 - 1 3 2 , 192, 233 non-verbal predicate, 80, 95, 99-100, 1 1 2 , 1 1 8 - 1 2 2 , 124-126, 128, 1 3 1 - 1 3 3 , 138 predicate formation, 52, 54, 68, 75, 82, 87, 107, 1 2 3 - 1 2 5 , 137, 188, 192, 194-195, 345; see also word

formation predicate frame; see frame predicate operator, 5 - 1 0 , 12, 35, 1 1 4 predicate phrase, 85, 100 predicate satellite; see satellite predicate term, 86, 1 2 0 - 1 2 1 , 126, 152 predicate variable, 81-84, 98-99, 1 0 1 , 137 term predicate; see term predication, 47-69, passim core predication, 3 0 - 3 1 , 4 2 , 56-57, 65, 1 7 1 , 184-185, 194, 197, 201, 203-205, 209, 300-302, 304, 3 1 1 , 319 extended predication, 42, 65, 1 7 1 , 184-185, 203, 205, 209, 300-302, 304 nuclear predication, 65, 8 1 - 8 3 , 9 1 » 95, 1 7 1 - 1 7 2 , 174, 176, 192, 194, 197, 201-205, 209, 2 1 1 , 227, 300-301, 304, 308, 3 1 0 - 3 1 1 , 319-320, 329 predication operator, 5-8, 1 0 - 1 2 , 32, 35, 47-48, 56, 67, 76 predication satellite; see satellite secondary predication, 2 1 , 3 3 predicator, 1 1 5 - 1 1 6 , 124, 157, 210 preposition, x, 88, 1 0 1 , 1 4 1 , 156-159, 162, 166-167, 234; see also adposition presentative, 1 2 8 , 1 4 8 , 270, 273 process, 1, 5, 13, 2 1 , 119, 153, 1 7 4 175, 178, 180, 192-195. 220, 224-226, 229, 237, 240-241, 246-247, 249, 251-254, 259-260, 282, 286-287, 3 1 2 - 3 1 3 . 323. 3 3 1 , 334. 337. 340, 345

Index of subjects production, χ, 119, 123, 133, 225, 228, 247, 249. 251-253, 281-282, 285, 288, 313, 326, 329 prominence, viii, 127, 255-258, 263-265, 267-269, 271 accentual prominence, 256-257, 264-265, 268-269 pronoun, 5, 60-61, 81, 84, 101, 107, n o , 1 1 2 - 1 1 3 , 130, 146, 154-155, 241, 256, 261, 269, 271 property, x, 6, 8-10, 15, 28,48-57, 68, 80, 83, 98, 100, 107,109, 1 1 2 1 1 7 , 120-123, I 2 6 , 128-129, Ϊ 3 1 , 148, 159, 161, 166, 169, 172-173, 184, 197, 202, 206-207, 209, 221, 227-239, 241, 251, 255, 257, 259, 262,264,311,325,329 proposition, 1, 3-7, 1 3 - 1 6 , 18-19, 2 1 24, 27-28, 30-32, 35-36,42, 69, 80, 84, 89-90, 109, h i , 114, 116, 124, 160, 166, 169-173, 179-182, 187, 199-200, 205-209, 216, 220, 226-227, 236, 288-290, 296, 300, 302, 304-305, 3 0 7 - 3 1 1 , 320 proposition operator, 5-7, 1 3 - 1 4 , 16, 18, 22-23, 32 proposition satellite; see satellite propositional content, 1, 80, 1 7 0 - 1 7 1 , 173, 179-182, 199, 207, 300, 302, 305, 308-311, 319 prototypicality, 49, 79-80, 88, 1 1 7 , 122, 138, 143, 195, 227, 235, 237 psychological adequacy, xi, 240, 250, 260 punctuation, 283-287 quantifier, 48, 55-56, 63, 302 quotative, 15, 18, 35 referent, 6, 28, 48-49, 5i~57, 60-61,

379

66, 83, 1 1 0 - 1 1 2 , 1 1 5 - 1 1 7 , 1 2 0 - 1 2 1 , 123-124, 126-127, 1 3 1 . 172, 200-201, 236, 239, 261-262, 283, 286 referential, 2, n o , 1 1 6 - 1 1 8 , 1 3 1 , 137, 159, 235, 241, 251, 323, 329, 334 referring expression, 2, 4, 7, 49, 109-133 reflexive, 107, 194 reinforcement, 16, 35, 192, 210 relative clause, 66, 69, 88,102, 1 1 1 , 154, 161, 288, 290-291 relativization, 81, 88, 92, 122, 130, 189, 191 relevance, 14, 58, 141, 169, 186, 224, 249, 268, 286, 333, 335 request, 127, 148, 240, 305, 3 1 2 - 3 1 4 restrictive, 27, 30, 33, 36, 83, 87, 101, 1 2 1 , 1 3 1 , 169, 205, 208-209, 235, 258, 334 restrictor, 3-4, 29, 48, 53, 75, 83, 85, 95, 101, 109, 111—117, I 2 I - I 2 2 , 124-126, 129-130, 132, 141, 145, 1 4 9 - 1 5 1 , 154, 158, 161, 229-230, 233 resultative, 20, 35, 301 rhema, 226-227, 233 rhetorical, 281, 283, 286-292, 296 rigid languages, 93-96, 99, 107-108 Role and Reference Grammar, xi, 41, 344 salience, 127, 133, 232, 259, 261, 266-267, 270, 273; see also prominence satellite, 3, 9, 21, 27-30, 33, 36,41-42, 61-62, 65, 67, 69, 89-90, 109110, 132, 153, 161, 166, 1 6 9 - 2 1 1 , 2 1 6 - 2 1 7 , 272, 274, 288, 291, 319

380

Index of subjects

satellite (cont.): manner satellite, 1 7 4 - 1 7 5 , 1 8 0 - 1 8 1 , 195, 206 predicate satellite, 28-29, 172, 174, 176, 187, 197, 199, 201-202, 204-205, 207-209, 291 predication satellite, 28-29, 65, 172, 176, 179, 185, 187, 197-200, 202, 204-205, 207-210 proposition satellite, 28, 172, 179, 182, 199-200, 205, 207 Quality operator/satellite, 47-49, 5 1 , 54-56, 65-67, 76, 160, 175, 235, 326 Quantity operator/satellite, 47-48, 55-58, 61-62, 65-67, 76, 239, 250 Reason satellite, 179, 183, 206-207 Speed satellite, 56, 65, 1 7 5 - 1 7 6 , 187,193 term satellite, 66, 69, 76, 194 schema, 2-3, 3 1 , 33, h i , 1 1 4 - 1 1 5 , 126, 266-267, 308, 323 scope, 1-2, 13, 18, 25-27, 30, 33, 3 5 3 6 , 4 1 , 47-49, 56-57. 59. 61-65, 67, 76, 83, 133, 137, 184-187, 192, 198-200, 202, 205, 2 0 9 - 2 1 1 , 225, 234, 239, 241, 281, 3 1 9 selection restriction, 19, 329, 336-338 semantic function, 2, 27, 3 1 , 35,48, 56, 66, 82, 1 0 0 - 1 0 1 , 126, 1 4 1 - 1 4 2 , 146,149-151. 153. 156-158, 166-167, 178, 182, 188, 191, 196-197, 216, 239, 242, 273, 301, 329, 337-338 Agent function, 2, 22-23, 35. 82, 84-85, 87, h i , 1 1 4 , 143, 152, 154, 1 6 1 , 193-195. 251, 265, 273,

288, 290-291, 301-306, 308-310, 338 Beneficiary function, 27, 174, 182, 1 8 7 , 1 9 4 , 196 Goal function, 22, 25, 35-36, 192-193, 196, 286, 301, 308-309, 338 Path function, 1 7 5 - 1 7 6 , 197, 222, 281 Patient function, 338; see also Goal function Possessor function, 33, 35, 66, 69, 94, 302-305 Purpose function, vii-viii, 154, 159, 178-179, 185, 187, 196, 220, 249, 253. 271, 328, 333 Zero function, 12, 22-23, 35. 81-82, 1 0 1 - 1 0 2 , h i , 1 1 3 , 119, 1 2 1 , 1 2 5 - 1 2 6 , 138, 146-147, 1 5 1 - 1 5 4 . 1 5 7 - 1 5 8 , 238-239, 291-292, 306, 308, 3 1 0 semantics, vii, ix, 41, 162, 197, 200, 217, 220-221, 224-225, 228-229, 232-233, 235-241, 277, 296, 325, 328, 3 3 0 - 3 3 1 , 340 shape, 51, 54, 77 sign, 227, 265, 327, 334 singular, 63, 69 spatial, 49, 5 1 , 53-54. 56-57. 60, 77, 1 0 1 - 1 0 2 , 1 4 1 , 1 5 7 - 1 5 8 , 166-167, 1 7 1 , 176, 188-189, 197, 210 speaker, x, 1-8, 10, 1 2 - 1 4 , 19. 22-23, 25, 28, 30, 34, 36, 52, 59, 61, 68, 80, 84, 90, h i , 123, 129, 1 4 1 , 148, 150, 152, 166, 169-173, 179, 200-201, 208-209, 216, 224, 236, 246, 250-254, 259-260, 263, 268, 273, 276, 282-287, 289, 292, 297,

Index of subjects

381

300, 302, 305, 3 0 7 - 3 1 4 , 3 1 6 - 3 1 7 .

2 1 7 , 2 2 0 - 2 2 1 , 224, 236, 241, 246,

319. 323. 326, 340

256, 269-270, 273, 276, 284, 3 1 4 ,

special treatment, 1 2 7 - 1 2 8 , 1 3 3 ,

325, 328-330, 334, 337-338, 344

250-252, 254-259, 263, 2 6 5 - 2 7 1 ,

Systemic Functional Grammar, vii, 41

273, 277

Tail, 36, 276

speech act, ix, xi, 1, 3-4, 7, 1 1 , 14, 1 6 17, 28, 32, 35, 80, 90, 109, I I I ,

telic, 9, 187 tense, 1, 5, 7, 1 1 , 23, 26, 30, 33, 35, 48,

1 1 4 , 1 1 6 , 166, 169, 173, 1 7 9 - 1 8 0 ,

58-60, 62, 65, 68, 95, 1 1 8 - 1 1 9 ,

1 8 2 - 1 8 3 , 206, 209, 236, 250, 281,

1 7 0 - 1 7 1 , 186, 200, 210, 300, 308,

283, 289, 296, 299-301, 303, 305,

3 1 1 , 317, 319-320

307, 309, 3 1 1 - 3 1 6 , 3 1 8 - 3 2 0 , 323

absolute tense, 1 1 , 6 1 , 68

speech event, 2 - 5 , 2 1 , 281 speech situation, 58-59, 224, 3 1 1 state of affairs, 1 - 2 , 4, 6 - 1 3 , 27, 29, 3 1 ,

past tense, 29-30, 33, 48, 81, 1 1 1 , 1 1 3 - 1 1 4 , 1 2 1 , 288, 290-291, 306, 308, 3 1 0

33-35. 48, 56-59. 68, 80-81, 90,

present tense, 16, 24, 32, 35, 81, 87,

95, 109, 1 1 6 , 1 4 1 , 166, 169, 1 7 1 ,

h i , 113, 119, 1 2 1 , 1 5 2 , 161,

1 7 3 - 1 7 9 , 1 8 1 - 1 8 2 , 188-189, 193,

3 0 1 - 3 0 6 , 308, 3 1 0

1 9 6 - 1 9 7 , 201, 203-204, 206-209,

relative tense, 1 1 , 68

286, 288-289, 297. 300-302, 3 1 1 ,

remote past tense, 18, 24, 27, 30, 33,

3 1 3 - 3 1 4 , 3 1 9 , 325,337. 344 stem, 62, 85-86, 89-90 stepwise lexical decomposition, 327, 330, 336, 344 story, 216, 2 5 1 , 256, 3 1 7 strategy, 6, 14, 16, 27-28, 84, 148, 170, 172, 249, 254, 262, 268, 273, 323 style, 160, 283, 325, 339 subcategorization, 3 1 , 144, 196, 209 subject, 23, 82, 93, 95, 1 1 9 - 1 2 0 , 126, 145, 148, 150, 158, 1 8 8 - 1 9 1 , 1 9 3 - 1 9 4 , 196, 2 1 1 , 227, 234, 270-274 subordination, 1 - 2 , 16, 2 1 - 2 6 , 3 1 - 3 4 , 56, 1 5 3 , 186, 216, 256, 269, 2 7 1 , 283-284 surface form, 1 3 1 , 2 1 9 - 2 2 0 , 222 syntagmatic, 325, 3 3 5 - 3 3 6 , 339 syntax, ix, xi, 3 1 , 4 1 , 69, 90, 1 5 8 - 1 5 9 ,

35 term, 47-69, passim term operator, 5 , 4 7 , 62, 66, 75, 122, 130 term predicate, 48, 66, 82, 85, 88, 100, 1 1 9 - 1 2 2 , 124 term satellite; see satellite text, viii, 2 1 0 , 236, 246, 252, 256, 283-284, 286, 300, 326, 3 3 7 - 3 3 8 Theme, x, 3 6 , 4 1 , 2 1 1 , 276 top-down, 282, 296, 323 Topic, 42, 1 2 7 - 1 2 9 , 1 3 3 , 184, 208, 250-259, 261, 263-265, 267-273, 276, 3 1 2 , 338 Given Topic, 128, 1 3 3 , 250, 2 5 2 257, 259-265, 267-269, 2 7 1 - 2 7 2 New Topic, 128, 253-256, 259, 262-265, 269-270, 272-273 Resumed Topic, 253-254, 257, 264,

382

Index of subjects

Topic (cont.): Resumed Topic (cont.), 269, 272 SubTopic, 128, 133, 254, 256-257, 262-265, 269, 272 topicality, 127, 133, 251, 254-255, 261-264, 267-269 translation, 17, 30, 284 typology, vii, x, xii, 7, 9, 75, 86, 91, 107, 132, 146-147. 149. 154, 158, 169, 205-206, 2 1 0 - 2 1 1 , 2 1 6 , 221-222, 229-230, 233, 270, 340, 344 underlying structure, ix, 20, 55, 61-62, 65, 81-83, 99. n o , 1 1 8 - 1 2 1 , 126, 130, 137, 2 1 1 , 219-222, 239, 246, 249, 284, 287-291, 299-300, 305-306, 3 1 0 - 3 1 2 , 314, 316, 318-320, 330, 337, 339 ungrammaticality, x, 10, 19-20, 36, 125, 148, 157, 289 universal, 62-64, 69, 132 unmarked, 50-51, 54, 69, 77, 231, 271-272, 274, 333, 338 upward layering, 281, 296 utterance, ix, 2-5, 14, 20, 79-80, 84, 90,

109-110, 114, 1 7 1 - 1 7 2 , 199, 216, 224, 235-237, 240, 246, 249, 253, 281-293, 296-297, 299-320 utterance content, 282, 308, 3 1 0 311,313-314,318 valency, 53, 119, 122-125, 132, I57> 195, 200, 233-234 variable, 3-5, 7, 1 1 , 22, 27, 29, 31, 35, 41-42, 48, 55, 57, 69, 76, 80-85, 90, 98-99, 101, 1 0 9 - 1 1 5 , 119, 1 2 1 - 1 2 5 , 129-130, 137, 145-146, 152-154, 159-160,166, 1 7 1 , 229-230, 239, 287-288, 292, 301-302, 305-308, 312, 320, 323 vocabulary, 160, 325-327, 3 3 1 - 3 3 3 , 335-336 word formation, ix, 328, 331, 334, 337, 340, 345; see also predicate formation word order, 1, 16-18, 35, 41, 62, 64, 67, 108, 187, 197-198, 246, 250, 276, 320; see also constituent order; PI position; P2 position working memory, 235 zero-level, n o , 1 1 2 - 1 1 5 , 1 1 7 , 1 3 1 - 1 3 2 , 137, 161