A New Architecture for Functional Grammar 9783110197112, 9783110173567

Citation preview

A New Architecture for Functional Grammar

≥

Functional Grammar Series 24

Editors

Casper de Groot J. Lachlan Mackenzie

Mouton de Gruyter Berlin · New York

A New Architecture for Functional Grammar

edited by

J. Lachlan Mackenzie ´ ngeles Go´mez-Gonza´lez Marı´a de los A

Mouton de Gruyter Berlin · New York

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

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

Library of Congress Cataloging-in-Publication Data A new architecture for functional grammar / edited by J. Lachlan ´ ngeles, Go´mez-Gonza´ lez. Mackenzie, Marı´a de los A p. cm. ⫺ (Functional grammar series ; 24) Includes bibliographical references and index. ISBN 3-11-017356-5 (pbk. : alk. paper) 1. Functionalism (Linguistics) 2. Discourse analysis. I. Macken´ ngeles (Marı´a zie, J. Lachlan II. Go´ mez-Gonza´ les, Ma. de los A ´ ngeles) III. Series. de los A P147.N478 2004 4011.118⫺dc22 2004049946

ISBN 3-11-017356-5 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 ⬍http://dnb.ddb.de⬎. ” Copyright 2004 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

Preface ......................................................................................... Abbreviations ................................................................................

vii xiii

The Architecture of a Functional Discourse Grammar Kees Hengeveld .......................................................................

1

Functional Grammar from its inception Matthew P. Anstey ................................................................... Behind the scenes: Cognition and Functional Discourse Grammar Carlos Inchaurralde ................................................................

23

73

The question of discourse representation in Functional Discourse Grammar John Connolly ..........................................................................

89

Focus of attention in discourse Francis Cornish .......................................................................

117

The complementarity of the process and pattern interpretations of Functional Grammar Michael Fortescue ....................................................................

151

Functional Discourse Grammar and language production J. Lachlan Mackenzie ...............................................................

179

Comment clauses, Functional Discourse Grammar and the grammar-discourse interface Peter Harder ............................................................................

197

Functional Grammar and the dynamics of Discourse María de los Ángeles Gómez-González .......................................

211

The problem of subjective modality in the Functional Grammar model Jean-Christophe Verstraete .....................................................

243

vi

Table of contents

Remarks on layering in a cognitive-functional language production model Jan Nuyts ..................................................................................

275

Discourse structure, the Generalized Parallelism Hypothesis and the architecture of Functional Grammar Ahmed Moutaouakil .................................................................

299

Towards a speaker model of Functional Grammar Dik Bakker and Anna Siewierska

.............................................

325

........................................................................

365

Index of names .............................................................................. Index of subjects ...........................................................................

379 384

Epilogue Kees Hengeveld

Preface

This volume arises indirectly from the Ninth International Conference on Functional Grammar (ICFG9), which was held at the Universidad National de Educación a Distancia (UNED), Madrid, from 20 to 23 September 2000. Some of the chapters derive from papers given there, others have been specially commissioned. This book addresses, and offers solutions to, a problem that was keenly felt at that conference: Functional Grammar (FG), the model of the organization of human languages proposed by the late Simon Dik, was failing to achieve one of its central ambitions, to account for regularities of form that are attributable to the use of language in discourse. The first chapter of this volume, by Kees Hengeveld of the University of Amsterdam (Netherlands), proposes that the solution to the problem lies in turning the original model on its head. He rejects the bottom-up approach, which started with the individual predicate and worked upwards to a fully specified underlying representation, in favour of a top-down approach, which takes the speaker’s intention to communicate as its starting point. At the same time, he develops suggestions in Dik’s posthumous work that a FG should have a modular architecture, proposing three levels of analysis for each discourse unit to be analysed: the interpersonal, the representational and the expression levels. Within each of these levels, the nesting of layers which came to be the trademark of FG in the nineties has been preserved. Hengeveld sketches various ways in which this architecture, to be known as Functional Discourse Grammar (FDG), is more adequate than the previous model, by showing how it can account for phenomena involving interactions between levels. All but the last of the remaining chapters offer reactions by other scholars to Hengeveld’s bold proposals. Matthew Anstey, Sessional Lecturer at the School of Theology, Charles Sturt University (Australia) and PhD candidate at the Vrije Universiteit, Amsterdam (Netherlands), offers a thoughtful, critical and thoroughly documented history of FG from its earliest beginnings to the present. He shows how FG, in its various guises, has struggled with the same seven problems throughout its existence and, even in its metamorphosis as FDG, continues to face many of the same difficulties.

viii

Preface

FDG brings grammar closer to cognitive science in introducing a Cognition module. In his chapter, Carlos Inchaurralde of the University of Zaragoza (Spain), assesses the cognitive adequacy of FDG. While stressing the impenetrability of cognitive processes, he points out various ways, for example in a language like Japanese with morphosyntactic marking of speakers’ awareness of social roles, that cognition impacts linguistic form in a manner that can be modelled within FDG. The following chapter, by John H. Connolly of the University of Loughborough (United Kingdom), turns to another module of FDG, that for the Communicative Context. He details the multiplicity of contextual factors that can influence the form of a discourse, and proposes a notation for representing contextual information. He distinguishes carefully between information which belongs to the contextual module and that which is more properly assigned to the interpersonal level. The chapter contains a detailed application of Connolly’s proposals to an invented mini-discourse. Discourse analysis is also central to the chapter by Francis Cornish, of the University of Toulouse (France), but his data are all taken from attested language use. His purpose is to show how the new FDG architecture provides a sound environment in which to account for phenomena of topic and focus in discourse. He achieves this aim by confronting FDG with work from the Columbia School of Linguistics (CS), which he finds to lack the tools needed to deal with the flow of discourse; on the other hand, incorporation of CS’s concern with attention and accessibility would further strengthen FDG treatments of language use. Michael Fortescue, of the University of Copenhagen (Denmark), observes a certain contradiction in the fact that FDG mimics language production while remaining a grammar. His position is that we need to address both Process and Pattern in a complementary fashion. He illustrates this thesis with data from Nootka, a language whose structures are heavily controlled by considerations of discourse pragmatics. Fortescue concentrates on Focus, showing that in languages like Nootka it belongs to both Process and Pattern; in a language where Focus is not encoded grammatically, it will be absent from the account of Pattern. For FG, the conclusion might be to retain the traditional architecture for Pattern and adopt the new for Process. The tension between static grammar and dynamic language production also underlies the contribution by J.Lachlan Mackenzie, of the Vrije Universiteit Amsterdam (Netherlands). He proposes a variant on Hengeveld’s proposals, dubbed Incremental Functional Grammar, in which the interpersonal and expression levels account for real-time processes. The

Preface

ix

interpersonal component analyses the Move as a sequence of Acts, each of which is composed of temporally successive Subacts. The role of the representational component, in this view, is that of a declarative grammar, providing semantic frames which organize focal and topical units into propositional form. Peter Harder, of the University of Copenhagen (Denmark), welcomes Hengeveld’s three-tier model, sharing Fortescue’s view that we need to recognize both the ‘cooking’ and the ‘recipe’, as Harder has it, even at the cost of a certain duplication of information. He concentrates on constructions in which there is a mismatch between the two, with especial reference to cases where the main clause serves to ‘hedge’ the content of the subordinate clause, as in I'm afraid John is ill. In the chapter by María de los Ángeles Gómez-González, of the University of Santiago de Compostela (Galicia, Spain), it is again FDG’s potential for treating the dynamics of discourse and of clause construction that is central. She confronts Hengeveld’s proposals with other recent developments in cognitively oriented linguistics, bringing them together in her model of Incremental Discourse Cognitive Grammar. She shows how clauses with complex beginnings result from the speaker’s incremental manipulation of the addressee’s focus of attention. Modality has long been a crucial issue in FG. Its treatment in FDG is the subject of the contribution by Jean-Christophe Verstraete, of the University of Leuven (Belgium). The chapter concludes that the new architecture offers an improvement in this respect, allowing – as Verstraete’s new data demand – a four-way distinction such that both epistemic and deontic modality (analysed as tensed and tenseless at the representational level) can be either subjective (located at the interpersonal level) or objective (and located at the representational level). The ambition of a Functional Discourse Grammar is comparable in many ways to the scope of Functional Procedural Grammar as developed by Jan Nuyts, of the University of Antwerp (Belgium). In a chapter which shares Verstraete’s concern with modality, he provides a critical review of FDG, arguing that layering is properly situated in cognition, and not in the grammar itself. Nuyts asks grammarian to be content with dealing with grammatically coded distinctions, and not to dabble in the study of conceptualization, which is non-linguistic, but hierarchical. Directly opposed to such functional minimalism, as Fortescue calls his analogous proposal, is the contribution by Ahmed Moutaouakil (University of Rabat, Morocco), who argues for an expansion of the layering principle upwards into the structure of discourse and downwards into the inner struc-

x

Preface

ture of the word. Moutaouakil proposes a Generalized Parallelism Hypothesis which entails that all products of linguistic activity, from text to word, are to be seen as richly layered discourses, with the proviso that the realization of the many layers is increasingly constrained as one descends to ever smaller units. The final reaction to Hengeveld’s proposal takes the form of a substantial chapter by Dik Bakker, of the University of Amsterdam (Netherlands), and Anna Siewierska, of the University of Lancaster (United Kingdom). Unlike their colleagues, they concentrate on the expression level, implementing proposals for an incremental, dynamic view of the morphosyntax of utterances. They propose a formalism that works top-down, left-to-right and depth-first and permits constrained inheritance and percolation. They then return to a central issue of this book, the relation between grammar and production, between pattern and process, arguing that FDG offers a model of the speaker’s actual processes of production, and backing this up with a detailed analysis of a few lines of transcribed speech. This volume closes with an Epilogue by Kees Hengeveld, in which he gives various responses to issues raised in the preceding chapters and then refines the model originally proposed. This chapter can thus be read as representing the current status quæstionis in FDG. Hengeveld begins by identifying FDG as a pattern model: although the structures of languages reflect the processes of communication, FDG does not model those processes. Yet he indicates his sympathy with dynamic implementations of the model. He goes on to welcome the various contributions to clarification of the cognitive and contextual components of the model and ends by offering an expanded version of FDG. The editors hope that this book will give a lively impression of the debates currently being conducted about the architecture of FDG. As Anstey says, now “FG could do with more engineers and builders”. The current volume can be seen as a blueprint for the work of those labourers. We hereby acknowledge the financial support provided by the Spanish Ministry of Science and Technology (Ministerio de Ciencia y Tecnología, MCYT), 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/Análisis del discurso en lengua inglesa: Aspectos cognitivos, contrastivos y de adquisición", and by the Netherlands Organization for Scientific Research, for making it possible for Dr Gómez- González to visit the Netherlands to work on the production of this book.

Preface

xi

Our gratitude goes to Susanna Hop MA for her invaluable assistance with word-processing. Last but not least, we wish to thank the authors of the various chapters for their contributions, and for their patience as they awaited the appearance of this book.

J. Lachlan Mackenzie Amsterdam

María de los Ángeles Gómez-González Santiago de Compostela

Abbreviations A ADS

Discourse Act; addressee Archetypal Discourse Structure Ass Assertive sentence C communicated content; clause c content CB Cobuild Corpus CDS current discourse space Cl Clause Config configuration ContrFoc Contrastive Focus CS Columbia School of Linguistics D discourse DCH Discourse Categories Hierarchy DECL declarative DM Discourse Model DRS Discourse Representation Structure DRT Discourse Representation Theory D-Topic Discourse Topic E State of Affairs ECC extraclausal constituent EL Expression Level EMT Extended Multiple Theme ER Expression rules ET Experiential Theme F Property or Relation FDG Functional Discourse Grammar FG Functional Grammar FncFtrs functional features

FPG

Functional Procedural Grammar FrmFtrs formal features GFT General Functional Theory GH Givenness Hierarchy GivTop Given Topic GPH Generalized Parallelism Hypothesis GPSG Generalized Phrase Structure Grammar H hearer HDPSG Head Driven Phrase Structure Grammar IDCG Incremental Discourse Cognitive Grammar IFG Incremental Functional Grammar IL Interpersonal Level ILL Illocution (of the expression) INFRM Informing move IS Information Structure ISO International Standardization Organization IT Interpersonal Theme L actual language Lab label general cognitive repreLcog sentation language Lex Lexeme LFG Lexical Functional Grammar LIBMSEC Lancaster IBM Spoken English Corpus

xiv

Abbreviations

LIPOC

Lkr Llog Lur M MNLU NewTop NLU NVC OSI p.c. P p P1 P2 Pa PAddr Para POS Pr PredFoc PrP

Language Independent Preferred Order of Constituents language of knowledge representation logical language for logical reasoning. language of underlying representations move; manner; message Grammatical Module of Natural Language User New Topic Natural Language User non-verbal communication Open Systems Interconnection Reference Model personal communication Participant in Speech Event propositional content Predication-internal clause-initial position External clause-initial position addressee addressee Paragraph positive Problem Predicate Focus Predicate Phrase

speaker speaker Referential Act; relationship ResTop Resumed Topic RL Representational Level RP Referential Phrase RST Rhetorical Structure Theory S Sentence; speaker SDRS Segmented Discourse Representation Structure SET Setting(s) SFG Systemic Functional Grammar SoA State of Affairs ST Situation Theory SubCat Subcategorization SubTop Sub Topic T Ascriptive Act; discourse type TT Textual Theme u utterance U utterance; discourse unit UC Underlying Clause UR Underlying Representations UtterFoc Utterance Focus WCF West Coast Functionalism X Individual; proposition x term variable

Ps PSp R

The Architecture of a Functional Discourse Grammar Kees Hengeveld 1. Introduction Since the beginning of the nineties, a significant part of the research carried out within the Functional Grammar framework has been directed at the expansion of Functional Grammar (FG) from a sentence grammar into a discourse grammar.1 There are several reasons why FG should aim at such a development. First of all, there are many linguistic phenomena that can only be explained in terms of units larger than the individual sentence: discourse particles, anaphorical chains, narrative verb forms, and many other aspects of grammar require an analysis which takes the wider linguistic context into consideration. Secondly, there are many linguistic expressions which are smaller than the individual sentence, yet function as complete and independent utterances within the discourse. This requires a conception of utterances as discourse acts rather than as sentences, as has been shown in Mackenzie (1998). Hannay and Bolkestein (1998) argue that the proposals2 which have been developed aiming at the expansion of FG into a grammar of discourse represent two different approaches. In the first, the discourse level is covered by additional hierarchically superordinate layers. This approach, called the upward layering approach in Hannay and Bolkestein (1998), is exemplified by Hengeveld (1997) and Moutaouakil (1998). In the second approach, the discourse level is handled by a separate component, linked to the grammatical component through an interface. Hannay and Bolkestein (1998) call this the modular approach, examples of which are Van den Berg (1998) and Vet (1998). In this chapter I want to claim that an adequate model of the grammar of discourse requires the integration of these two approaches, i.e. I will argue that both the application of extended layering and the recognition of vari-

2

Kees Hengeveld

ous levels of analysis are necessary. The model of a Functional Discourse Grammar (FDG) presented here is thus both hierarchical and modular. A major feature of the model is that it works in a top-down fashion, that is, decisions at higher levels and layers of analysis determine and restrict the possibilities at lower levels and layers of analysis. This feature of FDG will be treated first, in Section 2. Section 3 then presents a general outline of the model, focusing on the various levels of analysis and the complex interfaces linking them to one another. The layers to be distinguished at the various levels are presented in detail in Section 4. Section 5 looks at the dynamic top-down construction of basic linguistic expressions within FDG by analyzing a series of illustrative examples. More complex examples which involve intricate interactions between the various levels of analysis are discussed in Section 6. The paper is rounded off in Section 7.

2. Top down In Levelt (1989) the speech production process is described as a top-down process, running from intention to articulation. His analysis suggests that the speaker first decides on a communicative purpose, selects the information most suitable to achieve this purpose, then encodes this information grammatically and phonologically and finally moves on to articulation. Levelt shows that there is ample support in psycholinguistic research for this conception of speech production. The speech production model used in FG (Dik 1997a: 60) has a quite different orientation. It starts out with the selection of predicate frames that are gradually expanded into larger structures, which when complete are expressed through expression rules. In view of Levelt’s (and many other psycholinguists’) findings, this organization of the grammar runs counter to the standard of psychological adequacy that FG should live up to (Dik 1997a: 13–14). In the model defended here production is therefore described in terms of a top-down rather than a bottom-up model. This step, apart from having a higher degree of psychological adequacy, is crucial to the development of a grammar of discourse: in a top-down model, the generation of underlying structures, and in particular the interfaces between the various levels, can be described in terms of the communicative decisions a speaker takes when constructing an utterance, as will be illustrated in Sections 5 and 6.

The architecture of a FDG

3

3. Outline of the model Figure 1 gives the basic outline of FDG. It shows that FDG distinguishes three interacting levels: the interpersonal level, the representational level, and the expression level, in that hierarchical order. The presence of these three levels as separate modules within the model is the major difference from earlier upward layering proposals. Layering applies at each level separately, i.e. each level is organized hierarchically, as will be shown in the next section. This hierarchical organization of the model is the major difference from earlier modular proposals. Mapping rules link the interpersonal to the representational level, in those cases in which semantic content is necessary for the transmission of a certain communicative intention. Expression rules then link the interpersonal and representational levels to the expression level. In cases in which only pragmatic content has to be transmitted, expression rules directly link the interpersonal to the expression level. The various linking mechanisms may be interpreted as interfaces which define the possible correspondences between layers at different levels. The three levels interact with a cognitive component and with a communicative component. The cognitive component represents the (longterm) knowledge of the speaker, such as his communicative competence, his knowledge of the world, and his linguistic competence. The speaker draws on this component at each of the three levels. The communicative component represents the (short-term) linguistic information derivable from the preceding discourse and the non-linguistic, perceptual information derivable from the speech situation. As far as the linguistic information is concerned, the communicative component is fed by the interpersonal and expression levels, and feeds the representational level in order to enable later reference to earlier acts and expressions, as will be illustrated in Section 6. Of course, short-term information may be selected for long-term storage and is in that case passed on to the cognitive component.

Figure 1.

COGNITION

Outline of FDG

Mapping rules

GRAMMAR

EXPRESSION LEVEL

Expression rules

REPRESENTATIONAL LEVEL

INTERPERSONAL LEVEL

4 Kees Hengeveld

COMMUNICATIVE CONTEXT

The architecture of a FDG

5

4. Levels and layers 4.1. Introduction As was mentioned in the previous section, each level of analysis in Figure 1 is organized hierarchically. In this section I will first of all review each of the levels separately. Then I will present the full model and compare it to the earlier layered sentence model, as presented in Hengeveld (1989).

4.2. The interpersonal level The hierarchical structure of the interpersonal level is presented in Figure 2. (M1: [(A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))

Figure 2. The interpersonal level

Note that this representation is non-exhaustive, in the sense that there are higher levels of discourse organization which are not captured here. In 4.6 I will return to this point. At the interpersonal level a central unit of analysis is the move (M), defined in Kroon (1995: 66), following Sinclair and Coulthard (1975), as “the minimal free unit of discourse that is able to enter into an exchange structure”. As such, the move is the vehicle for the expression of a single communicative intention of the speaker. Examples of such communicative intentions are inviting, informing, questioning, threatening, warning, recommending, etc. In order to achieve his communicative intention, the speaker executes one or more discourse acts (A), defined in Kroon (1995: 65) as “the smallest identifiable units of communicative behaviour”. A move consists of one central act, which may be supported by one or more subsidiary acts. Every act may be characterized in terms of its illocution (ILL), by which I here mean the illocution as coded in the expression3 (Dik 1997b: ch. 11). Illocutions are represented as abstract illocutionary frames,4 which take the participants (PN) in the discourse act, i.e. the speaker (PS) and the addressee (PA), and the communicated content (C), i.e. the information transmitted in the discourse act, as their arguments. In order to build up the communi-

6

Kees Hengeveld

cated content the speaker may have to execute one or more ascriptive acts (T) and one or more referential acts (R): it is the speaker who refers to entities by using referring expressions,5 and it is the speaker who ascribes properties to entities by applying predicates to these referring expressions.6

4.3. The representational level The hierarchical structure of the representational level is presented in Figure 3. (p1: [(e1: [(f1) (x1)] (e1))] (p1)) Figure 3. The representational level

Note that, again, this representation is non-exhaustive. There are higher levels of semantic organization which are not captured here (see Section 4.6). In transmitting his communicative intention the speaker will in most cases have to fill his utterance with basic semantic content, i.e. with descriptions of entities as they occur in the non-linguistic world. These entities are of different orders: third-order entities or propositional contents (p); second-order entities or states of affairs (e); first-order entities or individuals (x); and zero-order entities or properties (f). Within the maximally hierarchical representation given in Figure 3 the propositional content (p1) contains the description of a state of affairs (ei), which contains the description of a property (fi) and the description of an individual (xi). Note, however, that all entity types may also be expressed directly, i.e. nonhierarchically, through lexical items.

4.4. The expression level The hierarchical structure of the expression level is presented in Figure 4. (Para1: [(S1: [(Cl1: [(PrP1: [(Lex1)] (PrP1)) (RP1: [(Lex2)] (RP1))] (Cl1))] (S1))] (Para1))

Figure 4. The expression level

This representation is just a simplified example of what the expression level might look like. It is an example, since every language has its own

The architecture of a FDG

7

expression possibilities, which lead to different expression units in their grammars. It is simplified, since the expression possibilities of a single language are generally much more refined than in the example I have given here. This simplified example will suffice, however, to illustrate the points I want to make in this chapter. It is a representation of constituent structure which starts at the level of the paragraph (Para), which may consist of one or more sentences (S), each of which may contain one or more clauses (Cl), which may contain one or more predicate phrases (PrP) and referential phrases (RP), each of which may contain one or more lexemes (Lex). It is important to note that the expression level corresponds with what in Levelt’s production is the product of grammatical and phonological encoding. Articulation, the final step, is not a level within the grammar, but the actual output of the grammar.

4.5. Integration The levels and layers discussed so far are given in Figure 5. The elements in boldface at the interpersonal and representational levels in this figure correspond to units that were present in the layered representation of clause structure defended in Hengeveld (1989) and its upward-layering elaboration in Hengeveld (1997). The correspondences may be listed as follows: M E ILL PN X e x f

Move Speech Act Illocution Speech Act Participant Propositional Content State of Affairs Individual Property or Relation

M A ILL PN p e x f

Move Discourse Act Illocution Discourse Act Participant Propositional Content State of Affairs Individual Property or Relation

Figure 5. Hengeveld (1989/1997)

The major differences at the interpersonal and representational levels, then, concern the presence of the variables C, T and R at the interpersonal level. In other words, a major feature characterizing the current proposal is downward layering at the interpersonal level. I will now briefly discuss each of these variables in contrast with their representational neighbours. Examples will follow in Section 5. The introduction of the variable T at the interpersonal level makes it possible to distinguish systematically between ascription as an act of the

8

Kees Hengeveld

speaker, and the entity type which is described within this act of ascription. Often the speaker will use the description of a zero-order entity (f) to give content to his ascriptive act, but he might also use, for instance, a firstorder entity (x) in a classifying or identifying construction. Similarly, the variable R allows for a systematic distinction between the act of referring on the one hand, and the entity type referred to on the other. Frequently the speaker will use the description of a first-order entity to give content to his referential act, but reference to other types of entity is equally possible. The introduction of the variable C opens up a way to distinguish the information communicated in a discourse act from the nature of the entity type the description of which is used to transmit that information. As a result, it is no longer necessary to assume that every discourse act contains a propositional content, i.e. a third-order entity. In many circumstances it is sufficient, for instance, to communicate information by simple reference to a first-order entity. A further difference between Hengeveld (1989, 1997) and the current proposal concerns the presence of the Expression Level in the model. The major motivating factor for the introduction of this level is the existence of meta-linguistic expressions (Sweetser 1990) or reflexive language (Lucy 1993). This phenomenon will be illustrated in Section 6.

4.6. Upward layering The previous sections have shown that each level or module in the proposed model has its own layered structure. Section 4.5 has stressed the relevance of further downward layering at the interpersonal level. Further upward layering is necessary too, but will not be dealt with here. A major point, however, is that upward layering is not restricted to the interpersonal level, but is a feature of all levels of analysis. Thus, at the interpersonal level there may be linguistic reasons to distinguish, for instance, between the layers of Turn and Exchange in dialogues; at the representational level languages may give special treatment to the layers of Episode and Story in narratives; and at the expression level there may be reasons to distinguish layers, for example Section and Chapter in written communication. In each case, the possible mappings of interpersonal to representational to expression categories have to be determined partly on a language-specific basis. Thus, a Move is mapped onto an Episode onto a single sentence in narratives in many languages of Papua New Guinea, whereas it is commonly expressed through a paragraph in most Western European languages.

Expression rules

(p1:[(e1:[(f1)(x1)](e1))](p1))

Mapping rules

GRAMMAR

((Para1:[(S1:[(Cl1:[(PrP1:[(Lex1)](PrP1))(RP1:[(Lex2)](RP1))](Cl1))](S1))](Para1)))

Figure 6. Levels and layers

COGNITION

(M1:[(A1:[ILL(P1)S(P2)A(C1:[...(T1)(R1)...](C1))](A1))](M1))

The architecture of a FDG 9

COMMUNICATIVE CONTEXT

10

Kees Hengeveld

5. Making choices at the interpersonal level Within the model proposed here, the construction of linguistic expressions can be interpreted as a decision-taking process on the part of the speaker. This process applies in a left-right and top-down fashion. Left-right decisions apply at the interpersonal level only. For instance, only after deciding on the communicative intention Warning for a certain move may the speaker select, for instance, the appropriate discourse acts Vocative, and Imperative. Only after deciding on the discourse act Vocative does the necessity to execute a referential act become obvious, etc. The result might then be something like (1): (1)

George, watch out for that tree!

Top-down decisions are of a more complex nature. These concern the decisions the speaker makes with respect to (i) the semantic content necessary to successfully execute an interpersonal act, and (ii) the expression category necessary to successfully transmit his communicative intentions. In what follows I will restrict myself to some examples of this decisionmaking process. In order to facilitate the interpretation of the representations I use tables in which the various levels correspond with rows: the first row contains the interpersonal units, the second row the representational units, and the third row the expression units. Furthermore, I concentrate on the representation of the relevant part of the example under consideration. Within the example this part is printed in italics; within the representation it is separated from the remaining part by means of vertical lines. Let us first consider a lexical, i.e. ready-made C. If the speaker wants to express his frustration about the way things are going he may select a lexical item which serves this purpose directly. An expression such as damn has pragmatic, not semantic content. Therefore, the speaker may move directly from the interpersonal to the expression level: (2)

Damn! (A1: [EXPR (P1)Sp (P2)Addr

(C1) ----(Lex1)

] (A1))

The expressive illocution takes care of the prosodic contour of this oneword expression.

The architecture of a FDG

11

The next example concerns a lexical R. If the speaker wants to draw the attention of someone present in the speech situation he may simply call his name. Here we have a referential act which makes use of a lexical item (Lex) which does not have semantic content, but only referential content. Therefore, the speaker may move directly from the interpersonal to the expression level again: (3)

John! (A1: [VOC (P1)Sp (P2)Addr (C1: [

(R1) ----(Lex1)

] (C1))] (A1))

Now consider a case in which the speaker draws on the representational level in order to transmit his communicative intention. The content communicated (C) here is the description of a third-order entity (p) expressed in a clause (Cl): (4)

The Plaza Santa Ana is the best place to go.

(A1: [DECL (P1)Sp (P2)Addr

(C1) (p1) (Cl1)

] (A1))

The same propositional content (p), expressed as a clause (Cl), may occur as the vehicle which the speaker uses to execute a referential act (R): (5)

I want to know whether the Plaza Santa Ana is the best place to go. (A1: [DECL (P1)Sp (P2)Addr (C1: [

(R1) (p1) (CL1)

] (C1))] (A1))

This may be contrasted with a case in which a referential act (R) again refers to a propositional content (p) but is expressed by means of a referential phrase: (6)

I want to know your opinion.

12

Kees Hengeveld (A1: [DECL (P1)Sp (P2)Addr (C1: [

(R1) (p1) (RP1)

] (C1))] (A1))

Ascriptive acts (T) often make use of the description of a zero-order entity (f) and are then expressed by means of a lexeme (Lex), as in the next example: (7)

The Plaza Santa Ana is wonderful, don’t you think? (A1: [DECL (P1)Sp (P2)Addr (C1: [

(T1) (f1) (Lex1)

] (C1))] (A1))

But the speaker may also decide on a first-order entity (x), expressed as a referential phrase (RP) to transmit the same kind of information, as in the following example: (8)

The Plaza Santa Ana is a wonderful place, don’t you think? (A1: [DECL (P1)Sp (P2)Addr (C1: [

(T1) (x1) (RP1)

] (C1))] (A1))

The examples just given serve to illustrate how linguistic expressions may be seen as the product of a top-down decision process on the part of the speaker, with a certain independence, within limits, of the three levels distinguished within the model.

6. Complex interactions between levels 6.1. Introduction Let me now turn to more complex interactions between the various levels. As Figure 1 already showed, the communicative context feeds into the representational level. The preceding discourse is of course part of this communicative context, and units within this discourse may be used for later reference. This is achieved in the model presented here by having

The architecture of a FDG

13

these units reappear within the representational level. In this way we may account for constructions like the following: (9) (10)

a. If you behave well, I’ll let you read my poems. b. Is that a threat or a promise? My brother-in-law, if that’s the right word for him, is a poet.

In (9b) the demonstrative that refers to the preceding move. In (10) it refers to the preceding lexeme brother-in-law. These examples thus illustrate that elements from both the interpersonal and the expression levels are available for later reference once they are produced, i.e. they may become part of the representational level in ensuing communication. Reference to elements from the interpersonal and expression levels will be studied separately below. In order to account for these cases the model presented here (Figure 7 below) allows for the copying of elements from the interpersonal and the expression levels to the representational level via the communicative context. In order to show their different status, variables from both levels are written with lower-case letters when they are used. In the next sections I will give some examples of how these variables are used in analyzing a variety of constructions which involve metacommunicative and metalinguistic expressions.

Expression rules

(m1: [(a1: [(c1: [...(t1) (r1)... ] (c1))] (a1))] (m1)) (p1: [(e1: [(f1) (x1)] (e1))] (p1)) (para1: [(s1: [(cl1: [(prp1: [(lex1)] (prp1)) (rp1: [(lex2)] (rp1))] (cl1))] (s1))] (para1))

Mapping rules

(M1: [IN (A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))

GRAMMAR

((Para1:[(S1:[(Cl1:[(PrP1:[(Lex1)](PrP1))(RP1:[(Lex2)](RP1))](Cl1))](S1))](Para1)))

Figure 7. Levels and layers

COGNITION

14 Kees Hengeveld

COMMUNICATIVE CONTEXT

The architecture of a FDG

15

6.2. Referents from the interpersonal at the representational level In this section I will discuss two construction types that can only be properly understood if we allow units from the interpersonal level to enter the representational level: hedged performatives and identity statements. Consider the following examples from Spanish: (11)

a. (Me) temo to.me I.am.afraid ‘I'm afraid that Juan is ill.’ b. Me temo to.me I.am.afraid ‘I'm afraid that Juan is ill.’

que Juan that Juan

esté is.SUB

enfermo. ill

que Juan that Juan

está is.IND

enfermo. ill

In (11a) the speaker simply expresses his state of mind. The embedded clause, in which the subjunctive is used, represents what he fears might be the case. In (11b) the indicative is used in the embedded clause. This sentence, unlike (11a), is an example of a so-called hedged performative (Fraser 1975), in which the embedded clause represents the actual information the speaker wants to transmit, but which he ‘hedges’ since he thinks the addressee might not like what he has to say. In cases like these the actual communicated content (c) is hidden in the embedded clause (Cl), which itself is the expression of a referential act (R), so that this sentence may be represented as follows: (A1: [DECL (P1)Sp (P2)Addr (C1: [...

(R1) (c2) (CL1)

...] (C1))] (A1))

A second case in which units from the interpersonal level figure at the representational level concerns so-called identity statements (Declerck 1988, Hengeveld 1992, Keizer 1992) as illustrated in (12): (12)

The Morning Star ís the Evening Star.

Sentences like (12), with a prosodically prominent copula, serve the purpose of stating that the act of referring to an object by using a certain name is equivalent to the act of referring to that same object by another name; hence they are statements about the validity of acts of reference. Therefore the representation of e.g. the Morning Star may be as follows:

16

Kees Hengeveld (A1: [DECL (P1)Sp (P2)Addr (C1: [...

(R1) (r2) (RP1)

...] (C1))] (A1))

This representation states that in (12) the speaker executes a referential act by making reference to a referential act (r), which is expressed as a referential phrase (RP).

6.3. Referents from the expression level at the representational level Units from the expression level may enter the representational level as well. I will discuss two cases here: metalinguistic conditionals and direct speech. An example of a metalinguistic conditional is given in (13): (13)

This concert, if you want to call it that, isn’t exactly what I was waiting for.

The word that refers to the preceding word concert, which is a case of reference to the code rather than to the message. Thus there is a referential act (R) in which reference is made to a lexeme (lex) which is expressed as a lexeme (Lex), as indicated in the following representation: (A1: [DECL (P1)Sp (P2)Addr (C1: [...

(R1) (lex1) (Lex2)

...] (C1))] (A1))

A second case in which reference is to the code rather than to the message concerns direct speech. Reporting direct speech may be interpreted as a form of mimicry (Clark and Gerrig 1990), where direct speech can be seen as imitated code. This is evident from the fact that direct speech reports respect the original language and/or dialect, as in: (14)

He said: “¿Cómo estás?”.

and that direct speech reports may contain meaningless noise, as in: (15)

He said: “gagugagugagu”.

The latter example furthermore shows that the imitated code can be any part of the expression level. In (14) the speaker refers (R) in the second argument of the verb say to a previous sentence (s) which in the actual expression is repeated through imitation. This example may thus be represented as follows:

The architecture of a FDG (A1: [DECL (P1)Sp (P2)Addr (C1: [...

(R1) (s1) (S1)

17

...] (C1))] (A1))

7. Conclusion In this chapter I have presented a basic outline of FDG and illustrated its appropriateness by analyzing a number of construction types that would have been difficult to handle in earlier versions of FG. Many aspects of FDG require further elaboration. These aspects can be grouped together into five categories: (a)

(b)

(c) (d)

(e)

What are the restrictions on left-right decisions within the production process, i.e. what are the systematic restrictions on the internal constitution of the interpersonal level? What are the restrictions on top-down decisions within the production process, i.e. what do the interfaces between the three levels of grammar look like? What is the internal structure of the cognitive component and how does it interact with the three levels of grammar? What is the internal structure of the contextual component, particularly with respect to the representation of the non-linguistic context, and how does it interact with the three levels of grammar? None of these questions is new to FG. I hope that the model of FDG presented here will provide the basis for an integrated approach to these central issues in linguistic theory.

Notes 1.

This chapter is the product of long and lively discussions with a great number of people. The Amsterdam FG-DISCO group has met at irregular intervals over the last few years, and has been a very inspiring environment for discussion of the topics dealt with in this chapter. I am indebted to the members of this group, Machtelt Bolkestein, Mike Hannay, Caroline Kroon, Lachlan Mackenzie, Rodie Risselada, and Co Vet, for the many openminded and inspiring discussions we have had. A special word of thanks goes to Mike Hannay, for a revival of the group’s activities when the time was there, and to Lachlan Mackenzie for joining this revival. Outside the FG-DISCO group, and extending over the same period, I have had countless

18

2.

3.

4. 5. 6.

Kees Hengeveld discussions with Gerry Wanders about the topic of this chapter, in which she has manifested herself as a critical and generous sparring partner. I am grateful to her for her help. Given the extensive interaction with all of these colleagues over a long period of time, it is hard to do justice to their individual contributions to the contents of this chapter. As a result, I do not want to claim originality for many of the ideas presented in this chapter, only for the way these are put together. See Van den Berg (1998), Connolly (1998), Connolly et al. (1997b), Crevels (1998), Gómez Soliño (1996), Hengeveld (1997), Jadir (1998), Kroon (1997), Liedtke (1998), Mackenzie (1998, 2000), Moutaouakil (1998), Rijkhoff (1995), Steuten (1997, 1998), Vet (1998). I take a broad view of coded illocution here, in that among the encoding possibilities I include not only sentences types, but also prosodic encoding, morphological encoding, and conventionalized lexicalization patterns. This slot may alternatively be occupied by a performatively used speech-act verb. Cf. Lyons (1977: 177): “... the speaker ... invests the expression with reference by the act of referring”. Cf. Lyons (1977: 161): “For example, in saying of a particular flower that it is red, we ascribe to it the property of redness, but we predicate of it the predicate ‘red’”.

References Berg, Marinus E. van den 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–106. Clark, Herbert H. and Richard J. Gerrig 1990 Quotations as demonstrations. Language 66: 764–805. Connolly, John H. 1998 Information, Situation Semantics and Functional Grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 167–190. Connolly, John H., Anthony A. Clarke, Steven W. Garner and Hilary K. Palmen 1997a A functionally oriented analysis of spoken dialogue between individuals linked up by a computer network. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 33–58. Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds) 1997b Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter.

The architecture of a FDG

19

Crevels, Mily 1998 Concession in Spanish. In: Mike Hannay and A. Machtelt Bolkestein (eds), 129–148. Declerck, Renaat 1988 Studies on Copular Sentences, Clefts and Pseudo-clefts. Leuven: Leuven University Press and Dordrecht: Foris. Dik, Simon C. 1997a The Theory of Functional Grammar. Part I: The Structure of the Clause (K. Hengeveld, ed.) [2d rev.]. Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part II: Complex and Derived Structures (K. Hengeveld, ed.). Berlin and New York: Mouton de Gruyter. Fraser, Bruce 1975 Hedged performatives. In: Peter Cole and Jerry L. Morgan (eds), Speech Acts (Syntax and Semantics 3). New York: Academic Press. Gómez Soliño, José S. 1996 La organización jerárquica de los textos desde una perspectiva funcional. In: M. Teresa Caneda Cabrera and Javier Pérez Guerra (eds), Os Estudios Ingleses no Contexto das Novas Tendencias. Vigo: Universidade de Vigo. Hannay, Mike and Machtelt Bolkestein 1998a Introduction. In: Mike Hannay and A. Machtelt Bolkestein (eds), vii–xii. Hannay, Mike and Machtelt Bolkestein (eds) 1998b Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. 1992 Non-verbal Predication: Theory, Typology, Diachrony. Berlin and New York: Mouton de Gruyter. 1997 Cohesion in Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 1–16. Jadir, Mohammed 1998 Textual cohesion and the notion of perception. In: Mike Hannay and A. Machtelt Bolkestein (eds), 43–58. Keizer, M. Evelien 1992 Reference, Predication and (In)definiteness in Functional Grammar. Dissertation, Free University Amsterdam.

20

Kees Hengeveld

Kroon, Caroline 1995 Discourse Particles in Latin: A Study of nam, enim, autem, vero and at. Amsterdam: Gieben. 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 17–32. Levelt, Willem J. M. 1989 Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Liedtke, Frank 1998 Illocution and grammar: a double level approach. In: Mike Hannay and A. Machtelt Bolkestein (eds), 107–128. Lucy, John A. (ed.) 1993 Reflexive Language: Reported Speech and Metapragmatics. Cambridge: Cambridge University Press. Lyons, John 1977 Semantics. 2 Vols. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–296. 2000 First things first: towards an Incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. Moutaouakil, Ahmed 1998 Benveniste's récit and discours as discourse operators in Functional Grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 25– 42. Rijkhoff, Jan 1995 Bystander and social deixis: some programmatic remarks on the grammar/pragmatics interface. Working Papers in Functional Grammar 58. Amsterdam: University of Amsterdam. Sinclair, John M. and Richard M. Coulthard 1975 Towards an Analysis of Discourse: The English Used by Teachers and Pupils. London: Oxford University Press. Steuten, Ans A. G. 1997 Business conversations from a conversation analytical and a Funcytional Grammar perspective. In: Connolly et al. (eds), 59–73. 1998 Structure and coherence in business conversations. In: Mike Hannay and A. Machtelt Bolkestein (eds), 59–75. Sweetser, Eve E. 1990 From etymology to pragmatics: metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press.

The architecture of a FDG Vet, Co 1998

21

The multilayered structure of the utterance: about illocution, modality and discourse moves. In: Mike Hannay and A. Machtelt Bolkestein (eds), 1–24.

Functional Grammar from its inception Matthew P. Anstey

1. Introduction Hengeveld’s chapter (this volume) was written with two purposes in mind.1 On the one hand, it is the latest attempt within Functional Grammar to formulate an explicit model for the detailing of natural language phenomena. On the other, it is a manifesto for those working within the Functional Grammar movement, a framework wherein scholars may locate their contribution to the general theory. This new architecture, therefore, is not only a design for a theoretical ‘space’. It is also, metaphorically speaking, a design for a collective, academic space, a conceptual meeting-place for those working in the many divergent subdisciplines of the linguistic sciences who wish to identify themselves to some degree with Functional Grammar. Accordingly, this chapter is an evaluation of Hengeveld’s proposal in light of these two purposes, as seen from the perspective of the evolution of Functional Grammar. Three important questions will be addressed. As a framework for Functional Grammar, what is its relationship to previous models? Does it address the objections raised against previous models? As a framework for Functional Grammarians, does it create sufficient space for the Functional Grammar academic community to continue to share one roof?

2. Preliminaries 2.1. Chronological Overview Clearly, such a historical overview needs to be highly selective. I have concentrated on the core of Functional Grammar, organizing the material

24

Matthew P. Anstey

into five periods. Each FGn refers not only to the period but also to its canonical FG publication. FG4, however, refers to both the period and Hengeveld’s Functional Discourse Grammar. (a) (b) (c) (d)

(e)

FG0 covers the period prior to 1978, the central work being Dik’s dissertation (1968) on coordination; FG1 covers 1978 to 1989, beginning with the publication of Functional Grammar (Dik 1978a); FG2 covers 1989 to 1997, beginning with the publication of The Theory of Functional Grammar (Dik 1989b); FG3 covers 1997 to September 2000, beginning with the publication of the two volumes of The Theory of Functional Grammar (Dik 1997a; 1997b); FG4 extends from the ninth International Conference on Functional Grammar (Madrid, Spain) onwards, where Hengeveld’s (2000) Functional Discourse Grammar was first presented.

Figures 1 and 2 provide an overview of these five periods.2

Total Publications 120 100 80 60 40 20

Figure 1. Functional Grammar publications from 1978 to 19983

98

96

94

92

90

88

86

84

82

80

78

0

FG from its inception

25

FG0 Dik 1968. Coordination 1968-1978 Dik 1977. Stepwise lexical decomposition Dik 1978. Functional Grammar Dik 1980. Studies in Functional Grammar Hoekstra et al. 1980. Special edition on Functional Grammar. Glot 3(3–4) Bolkestein et al. 1981. Predication and expression in Functional Grammar Martín Mingorance 1981. Translation of Dik 1978 into Spanish ’82: Functional Grammar workshop during 13th Int. Congress of Linguists, Tokyo, Japan Dik 1983. Advances in Functional Grammar FG1 ’84: 1st Functional Grammar conference, Amsterdam, The Netherlands 1978-1989 ’85: Functional Grammar symposium on predicate operators, Amsterdam, The Netherlands Bolkestein et al. 1985. Syntax and pragmatics in Functional Grammar Bolkestein et al. 1985. Predicates and terms in Functional Grammar ’86: 2nd Functional Grammar conference, Antwerp, Belgium ’87: Functional Grammar symposium on the computer, Amsterdam, The Netherlands Nuyts and de Schutter 1987. Getting one's words into line van der Auwera and Goossens 1987. Ins and outs of the predication ’88: 3rd Functional Grammar conference, Amsterdam, The Netherlands Dik 1989. The theory of Functional Grammar Moutaouakil 1989. Pragmatic functions in a Functional Grammar of Arabic ’90: 4th Functional Grammar conference, Copenhagen, Denmark Nuyts et al. 1990. Layers and levels of representation in language theory Hannay and Vester 1990. Working with Functional Grammar: descriptive and computational applications Siewierska 1991. Functional Grammar ’92: 5th Functional Grammar conference, Antwerp, Belgium Keizer 1992. Reference, predication and (in)definiteness in Functional Grammar Rijkhoff 1992. The noun phrase: a typological study of its form and structure FG2 Fortescue et al. 1992. Layered structure and reference in a functional perspective 1989-1997 ’94: 6th Functional Grammar conference, York, England Engberg-Pedersen et al. 1994. Function and expression in Functional Grammar ’95: Simon Dik dies after three years of illness Kroon 1995. Discourse particles in Latin ’96: 7th Functional Grammar conference, Córdoba, Spain Devriendt et al. 1996. Complex structures. A functionalist perspective Engberg-Pedersen et al. 1996. Content, expression and structure: Studies in Danish Functional Grammar Butler et al. 1997. A fund of ideas: recent developments in Functional Grammar Connolly et al. 1997. Discourse and pragmatics in Functional Grammar Dik 1997. The theory of FG. Part I: the structure of the clause Dik 1997. The theory of FG. Part II: complex and derived structures ’98: 8th Functional Grammar conference, Amsterdam, The Netherlands FG3 Olbertz et al. 1998. The structure of the lexicon in Functional Grammar 1997-2000 Hannay and Bolkestein 1998. Functional Grammar and verbal interaction Faber and Usón 1999. Constructing a lexicon of English verbs Martín Arista et al. 1999. Nuevas perspectivas en Gramática Funcional ’00: 9th Functional Grammar conference, Madrid, Spain Mackenzie 2000. First things first. Towards an Incremental Functional Grammar FG4 Pérez Quintero 2001. Special edition of Revista Canaria de Estudios Ingleses. Challenges 2000 and Developments in Functional Grammar. onwards ’02: 10th Functional Grammar conference, Amsterdam, The Netherlands Mackenzie and Gómez-González 2003. A new architecture for Functional Grammar

Figure 2. Historical Overview of Functional Grammar

26

Matthew P. Anstey

2.2. Central Problems In addition to the evolution of FG, we wish to pay attention to several recurrent problems that emerge. These provide various perspectives from which to evaluate Hengeveld’s proposal. It is beneficial to articulate them at this stage.4 (a) (b)

(c) (d) (e) (f) (g)

PR1. The problem of structure – what place, if any, does constituent structure have in the grammar? PR2. The problem of underlying representations – what do the underlying representations (URs) actually represent and how should they be interpreted? PR3. The problem of verbal interaction – how does FG relate to communication as process? PR4. The problem of functional primitives – how are the primitives defined and applied? How many are needed? PR5. The problem of discourse – how does FG account for linguistic phenomena beyond the sentence level? PR6. The problem of psychological adequacy – what does psychological adequacy mean and does FG fulfil it? PR7. The problem of formalization – how should the notation formally and explicitly represent language structure?

3. FG0 - prior to 1978 3.1. Publications prior to FG0 Two of Simon Dik’s earliest writings provide helpful background to FG0. Firstly, Dik (1966: 406; cf. 1967b) indicates that his first disagreement with generative-transformational grammar concerned the strict division between grammar and semantics. For example, the words manner, many, manifest, manly, manifold and so forth all contain the string ‘man’ but such a purely formal observation “can never tell us which of these [recurrences] are grammatically significant and which are not”. Therefore, some semantic criterion is necessary to determine this. Dik concludes that “no grammatically relevant element can be found in complete independence of meaning”. Nevertheless, he maintains, “[w]ithout being committed to recognize the exclusive rights to the title of ‘grammar’ claimed for it, one should not underrate the value of the contribution that especially the trans-

FG from its inception

27

formational component of generative grammar can make to general linguistic theory” (Dik 1966: 410). Secondly, the only published evidence of Dik’s familiarity with Prague School linguistics is in a book review wherein he first mentions “functionalism” (Dik 1967a). Dik approves of Jakobson’s view of the ‘means-ends model’ of language, writing “[t]his means that each fact of language is evaluated not only with respect to the system as a whole, but also with respect to the ultimate function it fulfils in the larger setting of extralinguistic reality” (Dik 1967a: 87). These two basic meanings of ‘functional’ appear in all of Dik’s writings. Of more interest is Dik’s criticism of the Prague School’s theme-rheme analysis. He argues that their analysis is “arrived at impressionistically” and lacks scientific rigour. “In the first place”, he writes, “it is rather misleading to subsume it under syntax and treat it as a part of grammar. Rather, it should be regarded as a phenomenon depending on the interpretation of the utterance in context and situation and thus already presupposing the grammatical (and the semantic) structure of the utterance. In the second place, if the distinction is really to be workable, there should be some principled basis for finding theme and rheme in any given sentence” (Dik 1967a: 86). This is the view now widely advocated in Functional Grammar (cf. Bolkestein 1998).

3.2. The theory Dik presents (proto-)Functional Grammar (FG0) as a chapter in his 1968 dissertation on coordination. He was motivated to provide an explanation of coordination that could overcome the problems in the approach advocated by generative grammar. Moreover, Dik thinks that the problems are not just with how generative grammar explains coordination but with the theory itself. Although Dik provides nine points of divergence of FG0 from “any version of ‘restricted’ constituent structure grammar” (1968: 199), two are central and will come to characterize all FG models. These are the “independent introduction of grammatical functions” and a “different notion of ‘derivation’”. Dik’s justification of grammatical functions as being basic to linguistic description consistently starts from the observation that formally equivalent constituents do not have equivalent linguistic properties. Therefore, constituents must have additional functional properties. Dik does not claim that this insight is original and is happy to assume Longacre’s (1965: 65) definition: “By function is meant the particular office or role of one distin-

28

Matthew P. Anstey

guishable part of a construction type in relation to other parts of the same construction”.5 What precisely then is a ‘grammatical function’? Dik writes that “… grammatical functions are irreducible aspects of grammatical structure, which can be partly correlated to formal features …, but cannot possibly be completely reduced to these” (1968: 154). His argument is essentially one of explanatory power: if we assume that a grammatical system has ‘grammatical functions’, then such a system provides a superior account of linguistic data. However, the axiom of functional primitives is itself not verifiable.6 It is clear, moreover, that psychological adequacy is already a motif in his thinking: “And we can explain the general features of human language if we can show them to be conditioned by general features of the human constitution, in particular, human mental and physiological properties. The latter kind of explanation, however, is not as such within the competence of the linguist, but should rather be explored by close collaboration between linguists and psychologists” (Dik 1968: 11–12). Dik’s “different notion of ‘derivation’” is the forerunner to the ‘avoid transformations’ constraint found in FG1 onwards. “The basic (negative) property of a functional grammar”, he writes, “… is that it does not include transformational rules” (1968: 163). To achieve this Dik posits a monostratal underlying representation of the functional, categorial, and constitutional elements of each linguistic expression. More specifically, FG0 has three basic components: constituents (that is, lexical categories) such as verb, noun, adjective, article; categories such as noun, pronoun, noun phrase; and functions such as subject, object, predicator, modifier, indirect object, question. How then does one select the necessary functions to describe a language? Dik provides the following stipulation, anticipating in many respects the central philosophy of FG1-4: “At least so many functions are set up for a language as there are grammatical (not semantic) differences between the linguistic expressions of that language which cannot be correlated to differences in constituency and/or in categorization” (1968: 176). FG0 distinguishes four sets of rules that relate constituents, categories, and functions together to form an ‘independent linguistic expression’. The application of these rules results in a single UR that combines functional (CAPITALS), categorial (normal) and constituent (italic) elements. For example, the sentence (1) has the underlying structure (2) (Dik 1968: 198):7 (1)

The man came.

FG from its inception (2)

29

ile(s (sdecl (SUBJ (np (npsg (DET(art(the) )+HEAD (nsg (man) ) ) ) )+PRED (fv (fvintr (fvintr past (fvintr past 3d ps sg(came) ) ) ) ) ) ) )

Dik considers the various possible relationships between semantics and grammar and concludes, along the lines of Weinrich (1966), “[i]n my opinion these various developments point to the possibility that [grammar is dependent on semantics] might finally prove to the right one” (1968: 293). This revealing quotation underlines the grammatical nature of the functions Dik posits. “Perhaps we could even go further”, he continues, “and include the full semantic description of a linguistic expression in its grammatical specification”. These are, however, “speculations which have little more than a programmatic value. They might, however, provide interesting possibilities for a further elaboration of the theory of functional grammar” (1968: 293). Thus, for Dik FG0 is an exclusively grammatical, non-semantic system, yet it includes ‘functional primitives’. The semantics of a language is a different matter. This bracketing out of semantics from the system of grammar will change in subsequent FG models, where the UR is interpreted as conveying grammar and meaning. Matthews’s (1969: 350) review of Dik’s dissertation is noteworthy. He views the dissertation positively but has grave doubts about FG0. Although modelled after tagmemics,8 FG0, unlike tagmemics, has no primitive that corresponds to syntagmeme (or to Halliday’s structure). Matthews takes this oversight as problematic to the whole concept of FG0: “But until he does so, his present formalization is literally not worth a moment’s consideration” (1969: 358). In other words, this is the first sign of the problem of structure (PR1). Matthews’s observation boils down to asserting that the URs of FG0 contain insufficient information to generate (via expression rules) the correct surface form. Bakker and Siewierska (2000), over thirty years later, conclusively demonstrate that this problem remains; but see Bakker and Siewierska (this volume).9

4. Dik’s writings between 1968 and 1978 The “interesting possibilities” that a semantically based grammar could offer occupied Dik for the rest of his life. In particular, the decade after his dissertation was spent on eclectically combining various components to assemble FG1. Three of his publications in this period are illustrative. Dik (1973b: 838) first refers to the predicate frame and operators when

30

Matthew P. Anstey

he writes, “In spite of differences in formulation and elaboration, many semantic theorists agree in that the basic form of semantic representations should look like the structure given.” prop

pred

arg 1

arg 2

arg 3

GIV E

JOH N

MAR Y

BO OK

Figure 3. Dik’s first predicate frame

Dik also examines briefly the possible basis for universal semantic categories. He rejects Chomsky’s ‘Innateness Hypothesis’ as a candidate and prefers what he calls the ‘Means-End Hypothesis’ (presumably from Jakobson 1963; see Dik 1967a) whereby language is an instrument of social interaction. The exploration of ‘communicative competence’10 across various cultures could hopefully lead to the positing of universal semantic categories. Dik (1975) provides the theoretical background to FG1’s States of Affairs typology. He adopts two criteria for distinguishing states of affairs: dynamism and control. For the first time Dik introduces explicit semantic functions into underlying structures, using the variable ‘M’ for ‘in the manner of’. Thus he suggests that sentence (3) is “correctly expressed” as (4): (3) (4)

Annette dances beautifully. s1(dance(Annette))s1 and beautiful(Ms1)

In 1977 Dik delivered a paper on stepwise lexical decomposition (Dik 1977). The central idea is that meaning definitions of predicates are given by a subset of simpler predicates from the lexicon of that same language (cf. Faber and Mairal Usón 1999: 58–65).

FG from its inception

31

5. FG1 - 1978 to 1989 5.1. The Theoretical Framework The first monograph devoted entirely to Functional Grammar (Dik 1978a) appeared ten years after Dik’s dissertation. Dik writes: In this book I develop a theory of Functional Grammar with the following main distinguishing properties: (i) it is conceived from a functional point of view on the nature of language: that is, one in which a language is regarded as an instrument of social interaction; (ii) it makes crucial use of functional notions of three different levels: the semantic, the syntactic, and the pragmatic levels; …

Consequently, the functional paradigm (1978a: 1) must deal with two rule systems: those that govern verbal interaction, ‘pragmatic rules’, and those that govern the formation of linguistic expressions, ‘semantic, syntactic, and phonological rules’. Both sets are “social in nature” (1978a: 1) with the linguistic rules being “instrumental with respect to” the verbal interaction rules (1978a: 2). We therefore discern three notions of functionalism: language as a whole is functional with respect to social interaction (the functional paradigm); individual linguistic expressions within a language are functional with respect to their uses in particular instances of verbal interaction; and the individual components of linguistic expressions are functional with respect to other components in the expression (cf. Nichols 1984). Dik (1978a: 6–9) argues that in addition to descriptive adequacy as found in generative grammar, FG1 adheres to three standards of explanatory adequacy: pragmatic, psychological, and typological (cf. Butler 1991). He writes (1978a: 6) that “[w]e want a Functional Grammar to reveal those properties of linguistic expressions which are relevant to the manner in which they are used, and to do this in such a way that they can be related to a description of the rules governing verbal interaction”. Unfortunately, to date there has been little formalization of such pragmatic ‘rules’. Inadvertently, Dik created the problem of verbal interaction (PR3) by setting an agenda for Functional Grammar that was never satisfied. Dik (1978a: 15) introduces three basic categories – nouns, verbs, and adjectives – which he describes as “basic predicates along the lines of Bach (1968)”.11 These three types of lexical predicates are embedded in a predication containing a predicate frame and optional satellites. The central

32

Matthew P. Anstey

concept of FG1 is essentially the combining of semantic, syntactic and pragmatic functions with predicate arguments. The semantic functions “are somewhat reminiscent of those distinguished in Fillmore’s (1968) Case Grammar” (Dik 1978a: 31). Fillmore writes that “[t]he case notions comprise a set of universal, presumably innate, concepts which identify certain types of judgements human beings are capable of making about the events that are going on around them…” (1968: 24). Dik (1978a: 39) is quite circumspect about semantic functions: “… the above proposal must remain tentative. … it may turn out that the distinctions made above are not, after all, as relevant as I think them to be. Much more research will be needed for a satisfactory solution of this question to be found.” The problem of primitives (PR4) is inherent in any grammar that has primitives, particularly semantic ones. The two syntactic functions, Subject and Object, are considered ‘primitive[s]’ of ‘grammatical relations’, in accord (partially) with Relational Grammar (Dik 1978a: 114; Fillmore 1968; Johnson 1977). Dik provides barely any references in his chapter on pragmatic functions, seeing the prevailing research as containing “much difference of opinion and much terminological confusion” (1978a: 129). Gebruers (1983; cf. Piťha 1985) helpfully observes that Bühler was the intellectual precursor both to Mathesius, founder of the Prague Linguistic Circle, and to Reichling, Dik’s doctoral supervisor. This connection possibly explains FG1’s (pragmatic) similarity to Prague.12

5.2. The Formal Notation The notation of FG1 has four components: the predicate frame, the term structure, the predication structure, and the ‘Outline of Functional Grammar’. The structure of the predicate frame is not given abstractly, but it may be clearly ascertained to be as follows (Dik 1978a: 29, 48; cf. Dik 1978c).13 (5)

[predicatetype (x1: )SF … (xn: )SF]SoA where n ≤ 4

The notation of the term structure is (1978a: 16; influenced by Dahl 1971): (6)

(ω xi : φ1 (xi) : φ2 (xi): … : φn (xi) )

FG from its inception

33

In this case, ω covers term operators, ωxi defines the universe of potential referents, restricted firstly by the subset of which φ1 (xi), the Head, is true. This subset is then restricted by a subsequent subset of which the Modifier φ2 (xi) is true, and so forth. The predication is as follows: { [ φ (x1) (x2) ... (xn)] (y1) (y2) ... (yn) } predicate arguments satellites terms

nuclear predication extended predication

Figure 4. The predication structure in FG1 (Dik 1978a: 26)

The overall model of FG is as follows: F U N D

P R E D I C A T IO N C O N S T R U C T IO N

P R E D I C A T IO N S

E X P R E S S IO N R U L ES

L IN G U IS T IC E X P R ES S IO N S

Figure 5. The overall layout of FG1 (from Dik 1980b: 2)

One could ask what the four arrows represent in the picture. Is this a productive model that mirrors a Speaker’s process, in which case the arrows

34

Matthew P. Anstey

represent the transition from intention to articulation? Alternatively, is this a procedural model? If so, of what procedure – the steps a computer simulation may go through in mimicking a language? That this same model gives rise to two different interpretations of Functional Grammar signals the beginning of the problem of psychological adequacy (PR6): how does such a standard apply to the model? To illustrate these different interpretations compare Van der Auwera (1983: 437 fn.5) who comments: “Another, rather implausible assumption is that the order of the FG formation would reflect the mental processes a speaker has to go through while forming a sentence.” Fortescue (1985: 113), on the other hand, writes that “FG seems to me interpretable as a model of real-time sentence production (this is of course not the only way it can be usefully envisaged)”. Finally, the formal notation intrinsically poses the problem of formalization (PR7), since it aims for explicitness and testability (see Bakker 1994; Hengeveld 1999). This problem became acute in the late midnineties when a plethora of notational variants competed for superiority.

5.3. Reviews Having outlined the salient points of FG1, we are now in a position to consider briefly the reviews it received, most of which were very positive. Comrie (1979: 275) is therefore representative when he writes that “[t]hese critical comments should not detract from the very positive promises that Functional Grammar holds out” (see also Comrie 1980; Dik 1980a). Most reviewers, however, noted the incompleteness of the theory and were “looking forward” to its explication. Almost all of the problems identified so far are observed by various people. Hymes (1979: 306), who clearly influenced Dik, notes the problem of verbal interaction (PR3): “It might be fair to say that Dik understands FG to be preferable as a way of analysing grammar as communicative means, but leaves the analysis of communicative ends, and the linkage between means and ends, to others, or for another time” (cf. Prideaux 1981; Hymes 1983). Bauer (1980: 52) picks up on the problem of primitives (PR4): “Objections of this kind – especially the difficulty of adequately motivating assignments [of semantic/syntactic functions] – were probably the most vital factor behind the flagging interest in case grammar, and it seems to me that in this respect, FG is no further ahead. Its viability depends on solutions to these problems”. Piťha (1980: 266) concentrates on the problem of underlying representations (PR2): “More problematic is the author’s term ‘semantic,’ denoting

FG from its inception

35

not only the semantics of nuclear predications, but also the semantics of fully developed linguistic expressions in their surface realizations. It is not clear whether the author … works with two notions: the semantics of predicate frames and the semantics of surface structures.”14 Watters (1980: 166) observes the problem of structure (PR1): “Perhaps the most glaring weakness is the lack of any extended discussion about the place and role of constituent structure and categorial information. Much of this is left implicit in the discussion of term formation and the ordering of constituents”.

5.4. Developments of FG1 FG1 sparked an enormous amount of interest, propelled along by Simon Dik’s charismatic leadership. In the next ten years, many publications were eventually to become assimilated into FG2, usually with some modifications by Dik. I have grouped them thematically.

5.4.1. The problem of structure (PR1) Dik never provides an explanation of how he constructed FG1 from the various components he appropriated and modified. But to understand the problem of structure (PR1) in contemporary FG work we need to consider why constituent structure seems to be absent from FG1. I submit the following ‘reconstruction’.15 The critical objective for Dik was to find a way for the URs of his grammatical theory to be independent of autonomous syntax. His approach was quite inventive. Firstly, he simply removed all references to constituent structure (such as noun phrase, head, and so forth) from the URs, reducing them to lexical items arranged in a combination of functional relationships – semantic, syntactic,16 and pragmatic relationships. Secondly, he ‘hides’ this stranded structural information in three separate locations, employing three strategies as follows: (a)

(b)

individuation – the predicate frames, which are stored in the lexicon, assign to every verb in a language a specific number of argument slots, each slot being specified for a semantic function, thereby obviating the need for an abstract (VP-type) structure in the UR; semanticization – the internal dependency relations of the components of noun phrases are isomorphically instantiated in the term

36

(c)

Matthew P. Anstey

structure as semantic relations, thereby eliminating the need for an abstract NP structure;17 deferment – sentential and clause structure, including linearization information, is reified into word-order templates, which, although ‘existing’ per se, are prohibited from contributing their structural information until after the UR is fully specified with functional information.

In other words, nouns and adjectives are assembled into ‘term frames’ (that is, term structures), which are assembled into verb-bearing ‘predicate frames’, which are assembled (with other ‘term frames’ as satellites) into ‘predication frames’ (that is, URs), which are mapped onto ‘sentence frames’ (that is, word-order templates). The genius of Functional Grammar is that this approach creates the illusion, in my opinion, that there is no autonomous constituent structure, when in fact the four frames have an a priori status in the model.18 The presence of covert structure in the model explains FG’s ability to describe certain (often distributionally dominant) types of language data and its difficulties with other types of language data. For the sake of argument, however, let us assume that my ‘reconstruction’ is incorrect and that there is genuinely no constituent structure in Functional Grammar. In this case, the success of Dik’s approach in describing linguistic data raises the obvious question: why does it work at all? Why does Functional Grammar demonstrate that (a large majority of) syntactic structure is apparently epiphenomenal, a mimicry of semantic structure? And why can systematic syntactic deviations from this isomorphism so often be tied to pragmatic features of the communicative context? Finding thorough answers to such questions is the quintessential task of the functional linguist, as it would demonstrate that the fundamental functional hierarchy of influence – pragmatics > semantics > syntax – is not just an a priori belief but a persuasively substantiated linguistic theorem.19

5.4.2. The problem of the underlying representations (PR2) According to Dik, the URs of FG1 supposedly only contain lexical items in functional relationships with one another. Such a view facilitated Dik’s shift to understanding the URs as bearers of meaning. No additional semantic interpretative module is necessary, since the UR contains all the information necessary to deduce the meaning of the sentence. As Dik

FG from its inception

37

(1973b: 835) himself remarks, the “semantic content of a linguistic expression” is “the information which, given the role of context and of extralinguistic knowledge, it is necessary and sufficient to assign to a linguistic expression in order to explain its different final interpretations”. This shift, I submit, lies at the origin of the problem of underlying representations (PR2), because Dik wants the underlying representation to do two different tasks: to structure the information that contributes to the interpretation of a linguistic expression (URs as ‘semantics’) and to structure the instructions (that is, the functional relations between the information units) that will generate a linguistic expression (URs as ‘syntax’).20 Dik (1986a: 11) clearly indicates this when he states that they “…are meant to contain everything that is needed to retrieve the semantic content of the predication on the one hand, and to specify the form of that expression on the other”. The problem is illustrated by Dik’s pragmatic explanations of linguistic phenomenon such as definiteness. Does the definiteness operator in underlying representations represent the intrinsic definiteness of a term (a semantic UR) and so is included for any intrinsically definite term,21 or does it represent a trigger to generate the ‘definite article’ where need be?22 Neither account seems sufficient even for English, with its fairly restrictive use of the definite article. Consider the following three sentences, annotated with both types of UR: (7)

a.

Mr Smith went home grumpy. intrinsic: (dsx: ‘Mr Smith’: ‘grumpy’) generative: (sx: ‘Mr Smith’: ‘grumpy’) b. The grumpy Mr Smith went home. intrinsic: (dsx: ‘Mr Smith’: ‘grumpy’) generative: (dsx: ‘Mr Smith’: ‘grumpy’) c. A grumpy Mr Smith went home. intrinsic: (dsx: ‘Mr Smith’: ‘grumpy’) generative: (isx: ‘Mr Smith’: ‘grumpy’)

The expression rules component would obviously prefer the ‘generative’ UR as input, but this is plainly at the expense of semantic accuracy. A further complication arises when we read Dik’s explanation of definiteness (1978a: 61; cf. 1973a): “The difference between ‘definite’ and ‘indefinite’ can in my opinion only be captured in pragmatic terms: by means of a definite term the Speaker expresses the fact that he acts on the presupposition that the Addressee can identify the particular intended refer-

38

Matthew P. Anstey

ent(s) of the term in question.” This suggests that neither type of UR presented above represents the pragmatic significance of the use/non-use of the article.23 Once we recognize from the above examples that ‘intrinsic’ and ‘generative’ can be understood as ‘semantic’ and ‘syntactic’ respectively, the problem of the URs (PR2) can be understood as reconciling the tripartite functional hierarchy of influence with a monolinear formal notation.24

5.4.3. Extreme Functionalism Strangely enough, in spite of the absence of constituent structure in FG1, in 1983 Dik is still willing to admit its theoretical existence. He writes (1983b: 75): “Does this mean that every structural property must somewhere find a functional explanation? No: as with all human instruments, the functional requirements put on a language leave quite a bit of leeway for alternative specifications of non-functional properties; historical developments unavoidably create rudimentary properties …; and conflicts between different functional requirements may even create dysfunctionality in given areas of linguistic organization”. Hence one could imagine that either FG1 would make room for such structure or Dik’s views would change. Three years later the latter occurs. Dik (1986b: 11) writes: “Saying that a certain feature of linguistic design or change cannot be functionally explained is tantamount to saying that we have not yet been able to find a functional explanation for that feature”. Autonomous, non-functional structure is completely excised.

5.4.4. The problem of psychological adequacy Nuyts (1983; cf. also 1985, 1992) is the first to attempt to graft FG into a broader cognitive model of grammar and to take seriously the allencompassing framework of pragmatics. Nuyts concludes, “It is obvious that Dik’s theory is not a complete functional language theory: (a) it does not deal with cognitive structures; (b) the pragmatic rules are well-nigh absent (apart from the pragmatic functions); (c) a discourse component is missing. … Thus FG is intended to be a grammar in the restricted sense of the word (a formal description of only the verbal aspects of language), and not in its wider sense” (1983: 384). From 1983 onwards it is clear that Dik is likewise increasingly concerned with the psychological adequacy of FG1. We can chart the

FG from its inception

39

problematic course that Dik sought to navigate in this regard. The problems begin when Dik makes a crucial redefinition of psychological adequacy. In FG1 psychological adequacy is defined weakly as “[a grammar] should not be incompatible with strongly validated psychological hypotheses about language processing” (Dik 1978a: 7 [emphasis mine]). But in 1983 he writes, “… FG would like to relate as closely as possible to psychological models of linguistic competence and linguistic behaviour. … [W]hy should a grammar be neutral as between producing and comprehending linguistic expressions, when producing and comprehending such expressions is just what the grammar is there for?” (1983b: 76 [emphasis mine]). Dik (1986a) attempts to integrate FG1 into a theory of verbal interaction, or more precisely, “an integrated model of linguistic interpretation”. Psychological adequacy reaches its most extreme (1986a: 6): “psychological adequacy implies that a grammar developed according to FG specifications should be a good candidate for incorporation into operational models of natural language users” (cf. Dik 1989b: 13, 1997a: 13). But clearly, for such a model to be accurate, it must represent humans and their participation in verbal interaction. Dik adopts a simple but hazardous strategy (1986a: 2): “the form of … knowledge representation structures can be derived from the theory of Functional Grammar.” In 1989, this equivalence between URs and human cognition is taken to its logical conclusion, where he proposes the following equation (Dik 1989a: 100): (8)

Lcog Lcog Lur Lkr Llog

=Lur = Lkr = Llog = general cognitive representation language = language of underlying representations = language of knowledge representation = logical language for logical reasoning.

Not surprisingly, a thorough critique was forthcoming. Hesp (1990) appraised Dik’s computational Functional Grammar model, raising serious doubts about the possibility of Dik’s interpretation of psychological adequacy. Due to the subsequent inattention to this issue, the role and formalization of conceptual representations in Functional Grammar remains unresolved. Finally, a double irony in the history of Functional Grammar’s relationship to psychology should be noted. One the one hand, since its inception it has been a theory that explicitly seeks psychological adequacy, and yet it is formal linguistic theories such as Chomsky’s that have been the most in-

40

Matthew P. Anstey

fluential in psycholinguistics. On the other hand, it is arguably the case that FG’s commitment to formalization (derived from the merging of predicate logic with functional primitives) that has led to its appropriation in the study of information sciences such as knowledge engineering (Weigand 1992), textual scenario analysis (Rolland and Ben Achour 1998), and organizational analysis (Steuten and Dietz 1998).

5.4.5. The layered ‘revolution’ Between FG1 and FG2 the most important development is the introduction of the layered structure for URs. This development was a gradual expansion of the predication structure.25 Dik (1979) presents tense and aspect (τ α) distinctions as ‘predicate operators’; Dik (1981b) is the first to include them in the formal notation:27 (9)

π1/2 predβ (ω x1)n

De Jong (1980) proposes indicating the illocutionary function of a sentence as a predication operator. For example, the Ass(ertive) sentence (10) could have the UR in (11): (10) (11)

JOHN opened the door. Ass openV (xi: JohnN (xi))AgSubj (xj: doorN (xj))GoObj

In 1986, Kees Hengeveld, inspired by Searle (1969), Lyons (1977), and Foley and Van Valin (1984),28 presents for the first time a (three-)layered UR that contains an abstract illocutionary frame and operators at each level. Predications are placed in the illocutionary frame, just as terms are placed in the predicate frame (Hengeveld, 1986: 77; cf. Hengeveld 1987):29 (12)

π4 ILL (π3 Xi: [π1 predβ (ω x1)n] (Xi))

Hengeveld (1988) uses the ‘E’ variable for speech acts and the ‘e’ variable for predications (Vet 1986), thereby creating the fourth layer:30 (13)

E1: [π4 ILL (S) (A) (π3 X1: [proposition31] (X1))] (E1) (π2 e1: [π1 predβ (ω x1)n] (e1))

FG from its inception

41

Not only is the clause layered horizontally, it is now also layered vertically, the two layers signifying two different aspects of language: the interpersonal and the representational. So sentence (14) is represented as (15):32 (14) (15)

It seems that it is possible that he can cure blindness. E1: [DECL (S) (A) (XI: [seemV (XJ: [ ] (XJ))] (XI))] (E1)

(Pres e1: [possibleA (ej: [canV cureVinf (xj: p3 (xj))Ag (xk: blindnessN (xk)Go] (ej))Ø] (e1))

Hengeveld (1989) brings his layered structure to full fruition in his article “Layers and Operators”. The structure is unchanged from 1988, but Hengeveld provides a clear explanation of each component of the model and many examples of various structures. More importantly, he lays out a method of handling verbal complementation and clause combining.

5.4.6. Functional Grammar in Spain In 1981 Leocadio Martín Mingorance translated FG1 into Spanish (Dik 1981b). This led to the establishment of the Spanish FG school, which has concentrated on lexicology. Martín Mingorance (most recently, 1998) also proposed a new synthesis, the Functional Lexematic Model, which has been the basis of the important work of Faber and Mairal Usón (1999) and others in recent years (see Mairal Usón and Pérez Quintero 2002).

5.5. Summary We see that FG1 generated much work and faced many challenges. The occasion arose to produce a new publication to pull these sometimesincompatible threads together.33 It is to this publication that I now turn.

6. FG2 and FG3 - 1989 to 2000 6.1. Introduction If read in isolation, FG2 (Dik 1989b) appears to represent a major overhaul of FG1, when actually it mainly incorporates developments of the model suggested in the preceding years. And for reasons explained below, FG3

42

Matthew P. Anstey

(Dik 1997a, 1997b) is not that different from FG2. So, assuming the reader is familiar with FG2/3, we will focus solely on those developments that led to Hengeveld’s FG4, except to notice that FG2/3 adopts Hengeveld’s layered structure of the clause but collapses the two tiers (interpersonal and representational) into a single tier.34 Thus, the structure is as follows. (16)

π4 Ei: [π3 Xi: [π2 ei: [π1 [predβ (ω x1)n] σ1] σ2] σ3] σ4

6.2. The impact of Simon Dik’s death Simon Dik fell ill in 1992 and died three years later, doing little new research in this period.35 To fulfil the goal of a second volume to FG2 (TFG2), it was decided to publish FG3 in such a way that “the text presented … is intended to reflect Simon Dik’s views as closely as possible” (Hengeveld in Dik 1997b: viii). And since between 1989 and 1992 he was preoccupied with computer implementations of FG2, FG3 is almost identical to FG2, except for the two chapters on term structure, which were revised to take into account (aspects of) the work of Rijkhoff (1992). TFG2 is of course all new material, but theoretically is mostly an application of TFG1. Xinzhang Yang (2000) and Verstraete (2000) provide reviews of TFG2.

6.2.1.

Hengeveld’s Layered Structure

Hengeveld continued after FG2 to develop his layered structure. In 1990, he for the first time presents the notation complete with operators and satellites,36 including a new σ5 ‘clause satellite’. The model is now (Hengeveld 1990a):37 (17)

E1: [π4 ILL:σ4 (S) (A) (π3 X1: [predication] (X1): σ3 (X1))] (E1):σ5 (E1) π2 e1: [π1 predβ:σ1 (ω x1: predN (x1))n] (e1): σ2 (e1)

Rijkhoff (1992) made a substantial contribution to the layered structure of the clause with his proposals for noun phrase structure. A fascinating observation he made was the parallel between NP structure and the structure of the extended predication, whereby nouns have operators (ω) and satellites (θ) in three layers, as represented in his (simplified) NP structure:

FG from its inception (18)

43

(ςi: [ ω3 [ ω2 [ ω1 predN (xi) θ1 ] θ2 ] θ3 ])

Hengeveld (1992) is an important article. Firstly, he incorporates Dik’s (1989: 50) zero-level ‘f’ variable into the representation for all predicates and an illocutionary ‘F’ variable. Thus the layered structure, with operators added, is: (19)

E1: [π4 F1: ILL (F1) (S) (A) (π3 X1: [predication] (X1))] (E1)) π2 e1: [(π1 f1: predβ (f1)) (ω x1: (f2: predN (f2)) (x1))n] (e1)

Hengeveld justifies the ‘f’ and ‘F’ variables using his characteristic argument: because a predicate and illocution may serve as an antecedent for anaphoric reference, a variable for the antecedent needs to be included in the UR. Secondly, Hengeveld clarifies the place of adverbs in Functional Grammar and provides functional definitions for parts of speech. Finally, we see the first hint that the representational level is not always necessary when Hengeveld briefly mentions that interjections such as Ouch! may have a structure as follows: (20)

E1: (fi: ouchInt (fi)) (EI)

Hengeveld (1997) introduces a rhetorical layer above the interpersonal layer, which consists of a discourse (D), a discourse type (T), and a move (M). Hengeveld changes (S) and (A) in the illocutionary frame to (P1)S and (P2)A and allows recursion at all levels to indicate the hierarchical structure of discourses. This approach to discourse became known as the ‘upward layering approach’. The layered structure is accordingly: (21)

D1: [(T1: TYP (T1)) (M1: [speech act] (M1))n] (D1) (E1: [π4 F1: ILL (F1) (P1)S (P2)A (π3 X1: [predication] (X1))n] (E1))n (π2 e1: [(π1 f1: predβ (f1)) (ω x1: (f2: predN (f2)) (x1))n] (e1))n

In FG4 the Move ‘M’ has moved down into the interpersonal level and the rhetorical level is missing. However, it would seem that Hengeveld still believes in such a level since he writes, “there are higher levels of discourse organization which are not captured here” (this volume: 5).

44

Matthew P. Anstey

6.2.2. Some additional problems Hengeveld’s 1997 model is still quite different from FG4. We need to review briefly the factors that motivated the other changes in the FG4 model. Bolkestein (1985a) is probably the first author to articulate clearly the problem of discourse (PR7) in FG. She notes how discourse-level considerations affect the code of language. Many other studies considered discourse within the FG framework, but it was probably Kroon (1995) who established this problem in the centre of contemporary FG debate. She provided the important definitions of ‘discourse move’ and ‘discourse act’ adopted in FG4. Harder (1989) is to my knowledge the first to point out that the ‘E’ variable cannot be interpreted as referring to a speech act in the way that the other variables refer to their respective referents, because an utterance is a speech act. Bolkestein (1992) also remarks that ‘E’ refers ambiguously both to a clause, which is a product of a speech act, and to a speech act itself. Vet (1998) demonstrates that the use of ‘E’ in direct speech implies that the speaker ‘says’ a speech act. Hengeveld (this volume) therefore makes two important changes: he changes ‘E’ to ‘A’ (a discourse act) and he introduces the expression layer to allow the UR to refer to the language code, as he explains in his Section 6.2. Van der Auwera (1992: 336) argues that Hengeveld’s use of variables is at times “an ontological mix-up” between reference and denotation. This criticism motivates Hengeveld’s (this volume) introduction of referential and ascriptive acts in the interpersonal level, indicated by ‘R’ and ‘T’ respectively. Moutaouakil (1996) and Mackenzie (1998) extend Hengeveld’s (1992) treatment of sentence fragments that omit certain layers in the layered structure altogether but still remain discourse acts. This explains why in FG4 the presence of a particular layer does not entail that all lower layers within its scope have to be present. Vet (1998) and Van den Berg (1998) suggest that discourse phenomena are better handled by a modular approach. Vet’s motivation for the modular approach is that the phenomena of verbal interaction “obey different rules” from those of the linguistic expression code. Vet also points out that such a module is necessary regardless of discourse phenomena, to account for pragmatic illocutionary conversion for instance. The introduction of the communicative context in FG4 is Hengeveld’s modular proposal.

FG from its inception

45

7. FG4: The New Architecture We are now in a position to consider our three introductory questions in the light of the evolution of Functional Grammar and the problems associated with it, with most attention being devoted to the second question.

7.1. FG4 as a framework for Functional Grammar 7.1.1. What is FG4’s relationship to previous models? A newcomer to FG4 who has only read FG2/3 would understandably struggle to see the connection between the two, but the above survey demonstrates the natural progression of Functional Grammar as envisioned by Hengeveld and as shaped by many people over the last 15 years. The different appearance of the notation should not be taken to mean that the core commitments of Functional Grammar have changed. On the contrary, FG4, for all its notational sophistication, is still basically a nontransformational theory of grammar whereby linguistic expressions are generated by expression rules operating on underlying representations signifying functional relationships between lexical items encased in various frames, hierarchically nested within each other. This continuity between Hengeveld’s model and FG0-3 should not overshadow the important discontinuities. As the discussion below will illustrate, there are two defining characteristics of Hengeveld’s model that are important progressions in the evolution of Functional Grammar: the shift from predicate-centricity to pragmatic-centricity and the trifurcation of the traditional Functional Grammar underlying representation into interpersonal, representational, and expression layers. Moreover, our survey amply documents the rapid and eclectic way in which FG has continually adapted to the stream of linguistic data, criticisms and suggestions in which all linguistic theories find themselves existing. FG4, which in the light of the survey above presents itself as a direct descendant of previous models, continues this tradition that began with “speculations which have little more than a programmatic value” (Dik 1968: 293). My suggestion that a few non-functional elements may have strayed into various FG models and that the theory has struggled with various problems does not entail that its core commitments are indefensible, only that their implementation has not been perfect.

46

Matthew P. Anstey

7.1.2. Does it address objections raised against previous models? Focusing on the problem of structure (PR1), we may concede that the expression level is an interesting addition to Hengeveld’s model, but his examples are not clear as to the precise nature of this level. Hengeveld (this volume: 7) characterizes it as “a representation of constituent structure”; the expression rules fill this constituent structures in generating linguistic expressions, since as Hengeveld observes, “the expression level … is the product of grammatical and phonological encoding” (this volume: 7) The problem of structure involves deciding if there is any nonfunctional structure in language, and if so, deciding where it belongs in FG4. From the perspective of speech production, the question is whether the Speaker himself contributes constituent structures from a sort of ‘syntacticon’ in the way he contributes lexemes from the lexicon (and functional primitives from the ‘semanticon’). Hengeveld hints at such a process when he writes, “Top-down decisions … concern the decisions the speaker takes with respect to (i) the semantic content …, and (ii) the expression category necessary to successfully transmit his communicative intentions” (this volume: 10) Yet consider his example (2), here renumbered as (22): (22)

Damn! (A1: [EXPR (P1)Sp (P2)Addr

(C1) ----(Lex1)

] (A1))

The choice of expression category, in this case EXPR, is not a matter of constituent structure but an abstract illocutionary type. This presumably results in an isomorphic function-to-form correspondence between the type and its (grammatical and prosodic) expression (cf. Van Buuren 1985; Liedtke 1998). Similarly, in footnote 2, Hengeveld (this volume: 18) lists among the “encoding possibilities” of the illocution “prosodic encoding, morphological encoding, and conventionalized lexicalization patterns”. So although the expression level is clearly a representation of constituent structure, it seems that, as in previous FG models, only the expression rules manipulate such structure – the Speaker can only influence the structure indirectly by choosing semantic and lexical primitives. On this interpretation of FG4 we would have to say that it does not resolve the problem of structure,38 but the absence of detailed examples

FG from its inception

47

illustrating the working of the expression level leaves open the possibility of alternative interpretations of the model. Let us now turn to PR2: the problem of the underlying representations. It is difficult to evaluate FG4’s treatment of this problem since Hengeveld does not discuss it. He only writes that “… the construction of linguistic expressions can be interpreted as a decision-making process on the part of the speaker” (this volume: 10). Nevertheless, one could argue that Hengeveld addresses this problem indirectly by introducing a third interpretation of underlying structures. Let me explain. The interpersonal level clearly represents the intention of the speaker, indicating discourse moves, discourse acts, ascriptive acts, and referential acts. The variables for these do not have ‘meaning’ per se, as they are outside C, “the information communicated in a discourse act” (Hengeveld this volume: 8). Neither could they be triggers for the expression rules to generate the correct linguistic expression. In a word, the interpersonal layer appears to be representing the pragmatics of communication, but pragmatics in a particular way. On the one hand, the interpersonal layer represents pragmatics in a very broad sense, including the illocutionary and intentional aspects of language use. On the other, it represents pragmatics in a rather narrow sense, as Hengeveld restricts the interpersonal layer to representing only grammatically coded reflections of communication. So although a language may be used for dozens if not hundreds of different types of discourse acts, illocutionary acts, intentions, and so forth, it is only those with a unique grammatical expression that will be represented. It should be clear from my historical survey that the solution to PR2 is not simply to clarify the status of URs, but in a principled manner to identify and untangle the components of URs and thus work out what belongs where. And although Hengeveld does not address PR2 explicitly, his clear separation of what can be understood as pragmatic, semantic, and syntactic layers paves the way, in my opinion, for a resolution of this problem. In other words, Hengeveld’s three layers correspond to the three possible interpretations of the single-layered UR of FG1-3. To illustrate, let us return to our three example sentences illustrating definiteness, this time with a possible FG4-like annotation: (23)

a. Mr Smith went home grumpy. interpersonal: Decl (R: ‘Mr Smith’: ‘grumpy’)39 representational: (dsx: Mr Smith: grumpy) expression: (NP: /mstə smθ/ … AdjP /grmpi /)

48

Matthew P. Anstey b. The grumpy Mr Smith went home. interpersonal: Decl (RFoc: ‘Mr Smith’: ‘grumpy’) representational: (dsx: Mr Smith: grumpy) expression: (NP: /ðə grmpi mstə smθ/) c. A grumpy Mr Smith went home. interpersonal: Decl (R: ‘Mr Smith’: ‘grumpy’) representational: (dsx: Mr Smith: grumpy) expression: (NP: /ə grmpi mstə smθ/)

This representation has several notable features. Firstly, the expression layer indicates the referential unity of the three utterances: in each the Speaker refers to Mr Smith. Secondly, the interpersonal layer can represent grammaticalized pragmatic differences if need be. Thus for sentence (b) we could argue that the speaker intends both to refer and to draw attention to what he is referring to, written formally as RFoc. Thirdly, the semantic, or intrinsic definiteness of ‘Mr Smith’ is shown in all three utterances in the representational layer, avoiding the dilemma that A grumpy Mr Smith… poses for FG1-3. Fourthly, the variation in syntactic structure of the three utterances is visible in the expression layer. So for the sake of argument, suppose that linguists agreed that (c) was simply a variation of (a) containing no semantic or pragmatic differences. In this hypothetical case the language provides the Speaker a choice between multiple syntactic realizations for the same pre-syntactic input. The point of this brief analysis is obviously to demonstrate that FG4 provides a framework to tackle such linguistic data, a framework I would suggest is eminently superior to FG1-3. The linguistic product emerging from the complex interplay between pragmatics, semantics, and syntax can now be more adequately described. More importantly, coupled with various linguistic, philosophical, and psycholinguistic arguments, such data, it is hoped, can be convincingly explained. In other words, Hengeveld’s contribution in FG4 is to introduce a tripartite formal notation corresponding to the tripartite functional hierarchy of influence: pragmatics > semantics > syntax becomes instantiated as interpersonal layer > representational layer > expression layer. It is my view that this change alone constitutes the most important development for Functional Grammar, for without it the theory can not unravel the Gordian knot that has arisen from the well-intended but ultimately self-defeating conflation of pragmatic, semantic, and syntactic information onto the monolinear URs of FG1-3.

FG from its inception

49

Our next concern is PR3, i.e. the problem of verbal interaction. FG4 attempts in part to solve the problem of verbal interaction by creating a pragmatic module that interacts with the grammar. While I would like to see a pragmatic module developed, I suggest that it may not work the way Hengeveld proposes, for two reasons. Firstly, let us consider Hengeveld’s example (9), here numbered (24): (24)

a. If you behave well, I’ll let you read my poems. b. Is that a threat or a promise?

Let us break this down into a rough approximation of an FG4 account of how this could work for our two participants, PA and PB, each with their Pragmatic Module, PMA and PMB respectively.40 (25) a. PA forms: M1: A1: [Decl (C: [letV (IPro) (e: [readV (youPro) (my poems)Go]) [(behaveV: wellAdv (youPro)]Cond]]] b. The ERs produce S1: “If you behave well, I’ll let you read my poems.” c. PA copies UR1 and its expression (as m1 and s1) into PMA d. PB hears: “If you behave well, I’ll let you read my poems.” e. PB copies s1 and m1 into PMB f. PB forms: M2: [A2: [Int (C: [{threatN or promiseN} (Am1)Ø]]] g. The ERs produce S2: “Is that a threat or a promise?”

The problem lies with the use of the ‘copies’ in step (c) and particularly (e). Such an approach to pragmatic interaction is too reductionistic: what feeds into the communicative context is not abstract underlying structures but interpretations and inferences based on linguistic expressions. And since each participant constructs their own communicative context, we must assume some Gricean principle of mutual trust whereby each participant is doing a faithful job in tracking participants, identifying discourse moves, establishing deictic reference points, and so forth. Much communication involves checking that everyone is ‘doing his or her job’ properly. But like interpretations and inferences, communicative principles and conventions resist formalization. Implementing these into the model is no simple task. Secondly, let us consider Hengeveld’s example (13), here number (26): (26)

This concert, if you want to call it that, isn’t exactly what I was waiting for.

Hengeveld writes that “[t]he word that refers to the preceding word concert, which is a case of reference to the code rather than the message”

50

Matthew P. Anstey

(this volume: 16). Again, let us break this down into stages to illustrate the problem, using Hengeveld’s three-line notation.41 1. A1: [Decl (C1: [(R1)]), 6. (C2:[(T1T2R2R3R4)])Cond 11. (C3: [(T3 R5]) 12. neg e: (exactly ..) 2. (prox sei: concert)Ø 7. (want (you)Ag [call (Ø)Ag (it)Go ( that)Ref]Go42 3. Ph1: /ðs knst/ 4. [ðs knst] 5. cl1, etc. copy to CC

8. Cl1: /f j wnt t k:l t ðæt/ 9. [f j wnt t k:l t ðæt] 10. cl2, etc. copy to CC

13. Cl2: /znt gzæktl .. / 14. [znt gzæktl ..] 15. cl3, etc. copy to CC

Figure 6. A representation of example (26)

The obvious problem is timing: step #5 must occur before step #7. Otherwise, the speaker cannot formulate C2, as he cannot refer (by using that, see note 42) to what R1 refers to (by using concert) until both he and the listener have encoded and decoded concert instantiated as [knst] in the communicative context.43 The objection to FG4 is not that this order of production is not a priori impossible, but that the notation records no ‘chronology of composition’ and thus is static, whereas verbal interaction is manifestly dynamic. Furthermore, it seems that such linking of underlying representations in a chain of verbal interaction is fraught with cognitive and notational problems (not least of which would be the size of the notation). Dik’s (1989b: 52) warnings about the ‘quasi-productive’ interpretation of FG need to be heeded. Therefore I suggest that FG4’s account of verbal interaction is problematic because on the one hand verbal interaction necessarily involves interpretation, inference, and co-operation, none of which FG4 accounts for, and on the other because it is dynamic while FG4 is not. Now let us move on to PR4: the problem of primitives. FG4 has nothing to say about (the problem of) functional primitives. This is unfortunate considering the problems that have been raised concerning these. Having a new architecture is of limited value if the building materials have a tendency to show stress under pressure.44 What we can surmise is that the division of the UR into three layers entails the movement (and perhaps redefinition) of the traditional Functional Grammar semantic, syntactic, and pragmatic primitives into new locations within the model.

FG from its inception

51

PR5, i.e. the problem of discourse, is specifically addressed by FG4. Hengeveld (this volume) seeks to deal with discourse phenomena by including discourse-sized units in the interpersonal (and possibly a higher rhetorical) layer and by feeding linguistic information into the modular ‘communicative context’. That is, he attempts to combine the upwardlayering approach with the modular approach (cf. Mackenzie 2000). The most significant improvement in FG4 is the notational severing of the model from FG’s sentence-centric shackles. I say ‘notational’ since it is ironic that in FG0, Dik (1968: 199) advocates “no restriction to sentence as highest unit” and in FG1 he writes (1978a: 15) that “FG is meant to cover any type of linguistic expression, to the extent that the internal structure of that expression is governed by grammatical rules. It is thus not restricted to the internal structure of sentences … ”. FG4 merely fulfils what Dik always envisioned for FG. The motivation for this change, however, arose through FG’s encounter with linguistic phenomena, not with a desire to align the notation with the vision. Further studies need to be done to see how beneficial the model is for the description of discourse. Our next concern is PR6, or the problem of psychological adequacy. Hengeveld suggests that FG4 has “a higher degree of psychological adequacy” (this volume: 2) than FG3. This is justified on the grounds that it starts from the interpersonal layer and works down towards the expression, while FG3 moves from the lexicon to predicates through to full expressions. It would seem that when Hengeveld talks of psychological adequacy he is actually addressing the problem of the underlying representations: how are we to interpret the model, what is it a model of? But this poses a problem: FG4 is strictly a model of grammar, verified and refined by linguistic data, and applied to linguistic data; it proves its worth by how good a job it does of explaining grammatical phenomena. It is not a model of speech production. Thus the model itself is agnostic with respect to psychological processes. Linguists may attach symbolic meanings to the model that are analogous to cognitive-linguistic phenomena but they are just that, symbolic meanings. Therefore I am not convinced that FG4 (or FG0-3) can be psychologically adequate in any meaningful sense. This may trouble some (and relieve others), so I will present a brief sketch of three ways of understanding the relationship between psychology and FG4. Firstly, we would do well to study Dik’s original use of psychological adequacy in FG1. Using LIPOC as an example (Dik 1978a: 192–212), we see that LIPOC derives from linguistic data as a general word-ordering

52

Matthew P. Anstey

principle and exists irrespective of any psychological theory. Dik notes furthermore that LIPOC is explanatory with respect to linguistic data, but what would explain LIPOC? He writes that “[w]e expect that such [explanatory] principles can be found in psychological mechanisms bearing on the perceptibility of constituents and on the human capacity for processing complex information” (Dik 1978a: 212). Note that we do not start with a psychological principle and then inject it into the FG model. By analogy, the proof of the pudding with respect to the interpersonal layer in FG4 (where the strongest claims for psychological adequacy are made) is whether it accounts for linguistic data. Do languages show clear evidence that discourse moves ‘affect’ illocutions, that illocutions ‘affect’ propositions, and so forth? If so, then irrespective of psychology, one can justify representing this ‘chain of command’, as it were, iconically in the linguistic model. Psychologists may have something to say about such phenomena, or they may not. In other words, I suggest that the way Dik used psychology to explain an already observed fact such as LIPOC, P1, and so forth, should be applied to layering. Secondly, an inverse situation obtains when the linguist encounters indisputable psycholinguistic phenomena, such as Hengeveld’s mention of the ‘intention to articulation’ processing pathway. Does this justify iconic representation in the model? Not necessarily, because the model is imitating linguistic phenomena and not psychological. A case in point is the ‘chronology of composition’ problem presented in example (27) above. The so-called ‘time course’ of speech production constitutes a sine qua non for psycholinguistic grammatical models (Van Turennout 1997) but is totally absent from almost every current linguistic theory. Similarly, with respect to the three-layered structure, I do not think that Hengeveld can justify this division on evidence from psycholinguistic research, but only on linguistic grounds. So then, what do we do with such research? It provides a conversation partner to engage with, one whose views may open up new approaches to knotty problems. Sometimes this partner may be called upon ‘to give evidence’ in a disputed linguistic case, but only alongside the witnesses of language. It is collaborative, not authoritative (cf. Berg 1998; Boland 1999). Thirdly, the very concept of psychological adequacy itself, as applied to linguistics, involves a confusion of categories. It is like saying that a theory of psychology should attain linguistic adequacy – one is simply talking about two non-contiguous academic disciplines, which nevertheless, like physics and biology, are in many ways deeply intertwined. The mistake is

FG from its inception

53

to make one academic discipline have evaluative authority over another. Even when psychology does bolster our case, we should beware of saying that our model has ‘psychological adequacy’. Perhaps we could say instead that the symbolic meaning we attribute to our model coheres well with psycholinguistic evidence. The last issue that needs to concern us here is PR7, that is, the problem of formalization. We are now in a position to evaluate the core of FG4: its formal structure. How does it line up with linguistic data? Irrespective of the hypothetical pathway taken by the Speaker in deriving URs, the notation can only be justified if it provides superior linguistic explanations. We should observe from the outset that the greater part of Hengeveld’s article is a justification of an extension of the formal notation to capture many subtle referential problems that “would have been difficult to handle in earlier versions of FG” (this volume: 17). Hengeveld is therefore motivated by problems of actual language data (hedged performatives, identity statements, metaconditionals, and direct speech). Consequently, a crucial question regarding FG4 is, in my opinion, whether the capacity of the model to describe referentiality exceeds the capacity of grammar. Rather than make a definitive judgement on this question, I wish to raise three basic issues that need to be considered. Firstly, Hengeveld clearly sides with the view that the URs are bearers of meaning: “the speaker refers to entities”, “the speaker … ascribes properties to entities” (this volume: 6), and since most of his problems are referential he has invested the model with referential precision. Whatever a proform refers to, Hengeveld needs a variable for it. So let us take Hengeeld’s example (9) again, now numbered (27), and pretend B replied differently: (27)

a. If you behave well, I’ll let you read my poems. b. Is that Keats or Donne?

In this case the demonstrative that would refer to ‘my poems’. This indicates that the expression rules must select the same pro-form, that, to represent both types of anaphora. Thus FG4, based on URs as meaningbearers, models a process of referential neutralization whereby a single pro-form does many referential duties. An alternative model, based on a pragmatic account of reference, would have the Speaker select that to trigger an interpretative strategy whereby the Addressee determines the most appropriate antecedent (as suggested by Harder 1996a: 237–243). The point is clear: the solution to the problem of URs (PR2) affects how one chooses between these alternatives.

54

Matthew P. Anstey

Secondly, the FG4 account is not easily verifiable, as the UR could be interpreted as containing more referential information than is encoded in the language. As Harder writes: “If we view … 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” (1992: 320; Harder’s emphases). Not that FG is a stranger to such conundrums: witness the semantic neutralization across a single case-marked noun in Latin or the enormous multifunctionality of genitive constructions that in many languages reduces to a single morphosyntactic form. Such phenomena are unfortunately often presented as dichotomous ‘division of labour’ problems between grammar and pragmatics: is reference better explained pragmatically or syntactically? But surely the job of reference is itself shared between syntax and pragmatics. The grammar allows particular pro-forms to refer to certain types of entities (which is why *Is such a threat or a promise? is ungrammatical) and the Addressee’s interpretative strategies choose the appropriate entity from the referential “inventory” that the pro-form introduces. Lyons captures this balance well when he writes (1977: 197) that “[t]he linguist can contribute to the study of reference by describing the grammatical structures and processes which particular language-systems provide for referring to individuals and groups of individuals. It does not follow, however, that the linguist must be concerned with the actual reference of expressions in his analysis of the grammatical structure of system-sentences”. Thirdly, in defence of FG4, we must remember that models such as this one are not language-specific. Rather, they are designed to represent the superset of all possible linguistic phenomena. Thus crucial evidence against this part of FG4 would be to show that no language has a particular lexeme used for referring to previous discourse acts. But for a particular language, we simply need to determine how the language grammaticalizes referential subtleties. The issue here is to distinguish falsifying FG4 in general from demonstrating that an aspect of FG4 is irrelevant to a particular language. FG4 is an ambitious formalization of the URs of linguistic expressions. This formalization in FG4 has been aligned with the functional hierarchy of influence, allowing descriptions of linguistic data that represent more clearly than in FG1-3 the pragmatic, semantic and syntactic factors that interact in language use. Further research is necessary to determine if such superior descriptions lead to superior explanations.

FG from its inception

55

7.2. FG4 as a framework for Functional Grammarians It is always difficult to answer such questions as whether FG4 creates sufficient space for the Functional Grammar academic community to continue to share the one roof. The biggest problem is that the model is part of a larger whole that needs articulation. Theoretically, we have seen that FG4 touches upon many major areas that concern the FG community, but remains silent on substantial issues such as the primitives and the lexicon. As shown above, there has been no substantially new canonical FG monograph since 1989. If anything, this survey underlines the need for such a work, one that is perhaps more encompassing of diversity among the FG community than previous volumes. From a practical perspective, those teaching FG also deserve an up-to-date presentation of the theory, and newcomers to the theory need a more user-friendly initiation if Hengeveld’s increasingly elaborate notation becomes widely accepted. Even the question what contemporary FG theory is, is hard to answer without such a reference work. Perhaps the word ‘architecture’, therefore, is particularly apt. By analogy, I suggest FG could do with more engineers and builders. FG has been in a design frenzy for many years now, as if having the right blueprint will eventually manufacture the building all by itself. Needless to say, I suggest that addressing perennial problems concerning the theory’s foundations may offer more fruitful resources for development.

Notes 1. 2.

3.

4.

I would like to thank Lachlan Mackenzie, Kees Hengeveld, Dik Bakker, Chris Butler, and the late Machtelt Bolkestein for their assistance. The references in Figure 1 are also indicative of the important geographical centres of current Functional Grammar research: The Netherlands, Belgium, Spain, Denmark, and Morocco. The official Functional Grammar bibliography (De Groot and Olbertz 1999) lists 1,141 references for 1978 to 1998. If one includes post-1998 publications and Dik’s works from 1966–1977, the total is at least 1,300. Many of these problems have been identified by those working in computational implementations of Functional Grammar (Bakker 1994; Kwee 1994). There are other important problems, such as the problem of the lexicon – what place do observed semantic and syntactic regularities in the lexicon have in Functional Grammar? Are there language-independent lexical primitives, and more basically, where do particles, interjections, adverbs

56

5.

6.

7.

8. 9.

10.

11.

Matthew P. Anstey and prepositions belong? (Mackenzie 1992; Olbertz, Hengeveld and Sánchez García 1998; Engberg-Pedersen et al. 1996; Mairal Usón and Van Valin 2001); and the problem of sub-disciplines – how does Functional Grammar integrate with the various linguistic sub-disciplines such as diachrony, typology, language acquisition, sign language, and so forth (Bakker 1998; Boland 1999)? Compare Dik (1997a: 26): “… functional statements specify the relations of constituents to the constructions in which they occur”. Dik also notes that this idea is found in the works of Tesnière, Strang, Piťha, and Buyssens. One wonders why Dik does not refer to Whorf’s concept of ‘covert categories’ as summarized by Fillmore (1968: 3): “Many recent and not-so-recent studies have convinced us of the relevance of grammatical properties lacking obvious ‘morphemic’ realizations but having a reality that can be observed on the basis of selectional constraints and transformational possibilities”. The abbreviations are as follows: ile: independent linguistic expression, s: sentence; decl: declarative, np: noun phrase, fv: finite verb, intr: intransitive. Note that the ‘ile’ is (potentially) larger than a sentence, as in questionanswer pairs. Dik suggests that language users probably do not know how to produce such structures. Note the contrast with FG3, “…speakers are able to construe underlying clause structures and map these onto linguistic expressions” (Dik 1997a: 56). Dik writes, “… [FG0] can be viewed as a particular reconstruction of tagmemic notions” (1968: 159). They write (2000: 18) that “[i]t is then an empirical matter whether … we will have to rehash the overall organization in the sense that we assign elements and features to the UR level that would formerly be assigned to ER and the other way around”. Since the URs are already semantically complete, in the context of their paper, I understand these “elements and features” to refer to types of constituent structure. Dik quotes this from Campbell and Wales (1970). In every subsequent publication he quotes instead from Hymes (1972). Hymes’s article is based on his 1966 lectures, subsequently published in a monograph in 1971. Incidentally, it should not be assumed that Dik borrows this phrase from Hymes appropriately. Is there anything substantial in FG that would absolve it of Hymes’s pointed critique of ‘Garden of Eden’ linguistics: “The controlling image is of an abstract, isolated individual, almost an unmotivated cognitive mechanism, not, except incidentally, a person in a social world” (1972: 272)? Mackenzie (1987) is the only one to observe that the FG view of nouns as monovalent instead of avalent results directly from Dik’s appropriation of Bach’s logical-semantic formalism, which in turn derives from Carnap, Reichenbach, and others.

FG from its inception 12.

13.

14.

15.

16.

17. 18.

19.

20.

57

A further clue to their origin was supplied by Lachlan Mackenzie (p.c.), who relates this story: “I asked Simon why there were so few references to other literature in the pragmatic functions chapter in Dik (1978). He said that he had finished most of the rest of the book, but needed to add a vital chapter on pragmatic functions. To do so, he needed to shut himself off from the world for a few days, and therefore booked a hotel (in London, I think it was). He took no or a few (I can’t remember) books with him, and composed and completed the chapter in a few days. What references there are were added later, back at his desk so to speak. This must all have been around 1976/1977”. In this structure ‘type’ is the lexical category (verbal, nominal, adjectival), xn are terms, ‘s.r.’ are selection restrictions on the term, and SF is the semantic function of the term in the state of affairs (SoA) designated by the (nuclear) predication. It is hard to know where Dik derives the idea of selection restrictions from, since the only clue is an allusion to Weinrich (1966). The notion was regularly used in generative writings in the 1960s (Katz and Postal 1964: 15; Chomsky 1965: 95). For recent critical evaluations of predicate frames in Functional Grammar see Mairal Usón and Van Valin (2001) and García Velasco and Hengeveld (2002). He also dislikes Dik’s chapter on pragmatic functions: “It is deplorable that Dik’s FG only sporadically and clumsily stumbles towards the findings of the Prague school, rather than taking its starting point from there. … Regretfully, I must conclude that this chapter of Dik’s book lags behind the other ones in scientific importance” (Piťha 1980: 268). This reconstruction is purely hypothetical and other Functional Grammarians undoubtedly view the situation differently. The point is that it is only by understanding if and how Functional Grammar accounts for syntactic structure that the theory can be critically evaluated. ‘Syntactic’ in FG is perspectival and not related to constituent structure. Thus, the symbols ‘:’, ‘,’ and and – that is, restriction, apposition, and coordination – represent micro-functional semantic relations inside the term structure, which in turn enters into macro-functional relations with other terms and predicates. As evidence, consider Dik’s (1981a) article on ‘displacement’. It can also be observed that between FG1 (1978) and Bakker and Siewierska (2000) few publications discuss autonomous constituent structure. The hierarchy is simply a succinct way of expressing the oft-quoted position of Dik (1987b: 7): “… pragmatics is seen as the all-encompassing framework within which semantics and syntax must be studied. Semantics is regarded as instrumental with respect to pragmatics, and syntax as instrumental with respect to semantics”.

58 21.

22.

23.

24.

25.

26.

27.

28.

29.

Matthew P. Anstey Siewierska (1991: 228 fn. 7) notes: “This unclarity in regard to the semantic vs. semantico-syntactic nature of the underlying structures recognized in FG creates some uncertainty in regard to the type of criteria to be used in the determination of these underlying structures”. “In order to be able to generalize over all definite terms, however, we assume that even in such cases [as proper nouns and personal pronouns] the definiteness operator is present in the underlying structure of the term” (Dik 1978a: 60). Another such example is in Dik’s (1983a: 235) pragmatic interpretation of reflexivity. He sees the Addressee as inheriting the following task from the speech situation: “Given a two-place relation R2 and a single entity a, construct an interpretation in which R2 is applied to a”. The author who most clearly identifies PR2, but more from the viewpoint of instructional semantics, is Harder (1990, 1992, 1996a, 1996b; see also Mackenzie 1987 and Keizer 1992b: 148–157). He argues, for instance, that “there is strictly speaking no semantics” (1992: 305) in the accepted interpretation of FG, in the sense that “meaning” can only be attributed to an UR once it is interpreted, and no such interpretation mechanism exists in FG. In the structures given below, I have standardized the symbols to illustrate the continuity in the model’s development. They are structurally unchanged of course. The superscript ‘n’ indicates stacking. The subscript ‘β’ indicates lexical category (N, V, A, and so forth). It is in Dik (1989b: 137) that he first mentions Seuren’s work Operators and nucleus (1969b). Seuren (who reviewed Dik’s dissertation, Seuren 1969a) notes that propositional (i.e. illocutional) operators were proposed as early as 1946 by Lewis and that modal operators were proposed as early as the Middle Ages. Seuren’s complete list is as follows: the sentence qualifiers are ASSertion, QUestion, IMPerative, and SUGGestion and the qualifiers are NEGation, Pres, Fut, Perf, Past, U (for universal or gnomic tense), Poss, Nec(essity), and Perm(itted). In the light of this work, it is perhaps surprising that FG took its time to adopt a more sophisticated operator structure. The predicate and predication operators are π1 and π2 respectively in FG2 onwards, but at this stage of FG development they are conflated so I have used π1/2. Hengeveld joined the FG discussion group in 1985. Here he first encountered Foley and Van Valin (1984) when it was reviewed by Machtelt Bolkestein. Hengeveld (1986, 1987) calls X a predication, but it later is correctly relabelled as a proposition. Hengeveld (1986: 120, n. 6) notes that Van Schaaik, at the June 1985 FG symposium, was the first to distinguish between π3 and π1 operators (cf. Van Schaaik 1983). Incidentally, Hengeveld (1986) is also the first to mention in print the distinction between ILLs, ILLe, and ILLa, as

FG from its inception

30.

59

presented in Weijdema et al. (1982), but Dik introduced it first in his 1985 FG lecture notes. Vet’s (1986) paper is very important to Hengeveld, who regularly quotes from it, since he adopts Vet’s argument that predicational adverbs restrict the predication, thus justifying representations such as the following: Pierre arrived at eight o’clock PAST ei: [arriveV (d1xi: PierreN (xi))Proc] (ei): [eight o’clockN] (ei)

31.

32.

33.

34.

35.

The two (ei)’s in the structure mean that both Pierre arrives and eight o’clock apply to the discourse event (ei). This somewhat technical point explains Hengeveld’s preference for each variable in the structure to appear symmetrically on both sides of the layer the variable has scope over. See Mackenzie (1987) for an important and neglected alternative to this argument. Hengeveld (1990a: 4) subsequently corrects his terminology to predication in this notation since the proposition is actually π3 X1: […] (X1). Note, however, that in the same volume, Hengeveld (1990: 107) again incorrectly uses proposition. The reason that two ‘X’ and two ‘e’ variables occur is that epistemic and objective modality are expressed respectively by the lexical items (seemV and possibleA). Note that Hengeveld’s use of such lexeme-based forms in this example prohibits the canonical use of operators to represent grammaticalized semantic categories. This practice possibly diminishes the typological strength of the notation. There are three other contributions worth mentioning. Vester (1983) adds Change and Momentaneous to the SoA typology, based on the works of Vendler and Dowty; Brown (1985) introduces a lot more sophistication regarding term operators; and Dik (1987) provides a typology of entities using Bunt’s (1985) work. See Keizer (1992a) for an excellent discussion of other more subtle differences, particularly Dik’s unexplained introduction of the ‘f’ variable for zero-order entities. Since I only encountered Functional Grammar in 1999, I cannot speak of the broader (and clearly significant) impact of Simon’s death on the FG community in general. Machtelt Bolkestein (p.c.) kindly provided this brief comment: “After Simon’s death his PhD students were re-distributed over the other FG-minded available professors, and naturally their own works underwent both a slowing down at least for a while, and perhaps also some changes. All in all the first year after his death there was a very depressed atmosphere, which picked up after (and perhaps partly due to) the Córdoba conference on FG [1996]”.

60 36.

37.

38.

39.

40.

41.

42.

Matthew P. Anstey Hengeveld (1990) suggests that all five satellites are restrictive satellites, which accounts for the ‘:’ preceding each one. However, in the same publication, Dik, Hengeveld, Vester and Vet (1990: 63) argue that σ1 satellites are always restrictive, σ3-5 are always non-restrictive and only σ2 is either restrictive or non-restrictive. Because Hengeveld wants the σ1 to bind a variable and because in FG3, σ1 modifies the variable-less “core predication” (Dik 1997a: 87), he must bind it directly to the predicate. So the scope of σ1 is ambiguous between predicate and core predication. Similarly, the σ4 binds to ILL as there is not yet an ‘F’ variable in this model. As early as 1990, Van Valin noted that for FG “the development of such a theory [of non-relational syntactic structure] and an explicit means of representing syntactic structure is essential” (1990: 225). Mairal Usón and Van Valin (2001) reiterate this sentiment. A ‘syntacticon’ could contribute to solving this problem and provide a natural place to provide discourse particles and other clause combiners. Incidentally, the term ‘semanticon’ actually comes from Dik (1973b). ‘Mr Smith’, Mr Smith, and /mωstχ smωθ/ etc. represent conceptual, lexical, and (Australian English) phonological encoding respectively. I understand the expression layer to signify that which is input for the articulatory component of Functional Grammar. Hengeveld writes that “… the demonstrative that refers to the preceding move”, which is why ‘Am1’ is the subject of S2 in line (f). In FG, ‘A’ is used for anaphora, and coreferential uses of pronouns are understood to be anaphoric. It should be noted that Hengeveld’s statement is not unproblematic. The pronoun that may be a case of what Lyons calls impure textual deixis (quoted in Levinson 1983: 87 from Lyons 1977: 670). Such uses of pronouns seem both to refer (in this case to the state of affairs referred to be the first speaker’s utterance) and to be anaphoric (in this case to the statement/move made by the first speaker’s utterance). The numbers represent a possible order of production. ‘CC’ stands for ‘communicative context’, and ‘Ph1’ for ‘phrase one’. The fifth line represents the copying mechanism proposed in FG4 to account for verbal interaction. I am assuming that the two ascriptive acts in step #6 are want and call and that the three referential acts are you, it, and that, since anaphora and discourse deixis are both referential. The notation of steps ##6–7 is not at all straightforward. In FG1-3 one would probably have something as follows for step #7: (f1: want (xj: you)Ag [f2: call (Ø)Ag (Ae1)Go (Al1)Ref]Go, where Aei (‘it’) is coreferential with ei: concertlex1 and Al1 (‘that’) is discourse-deictic (see Levinson 1983: 86) with (just the lexeme) concertlex1, the “l” standing for ‘lexeme’. This perfectly illustrates the sort of pragmatic-semantic conflation FG4 seeks to avoid. I understand

FG from its inception

43.

44.

61

FG4 as locating reference in the interpersonal layer; thus in R3 the speaker refers to what R1 refers to and in R4 the speaker refers to the lexeme concert. This allows direct entry in the representational layer of the pronominal forms the speaker chooses to ‘represent’ the content of the referential acts R3–4. A similar problem regarding the use of Tails to clarify the main clause in response to a non-verbal clue from the Addressee was noted by Piťha (1980). Some sources of pressure include the SoA typology (Rijksbaron 1989); Subject and Object in patient-prominent languages such as Lak (Helmbrecht 1992); the problem of assigning semantic functions to terms with no relationship to a predicate (Mackenzie 1990); and the presence of linguistically questionable primitives such as ‘sim’ and ‘post’ in the FG notation (Harder 1995: 232–235).

References Auwera, Johan van der 1983 Review of Bolkestein et al. (eds), Predication and expression in Functional Grammar 1981. Journal of Pragmatics 7: 433–442. 1992 Free relatives. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 329–354. 1997 Cosubordination. Working Papers in Functional Grammar 63. Amsterdam: University of Amsterdam. Auwera, Johan van der, and Louis Goossens (eds) 1987 Ins and Outs of the Predication. Dordrecht: Foris. Bach, Emmon 1968 Nouns and noun phrases. In: Emmon Bach and Robert T. Harms (eds), Universals in Linguistic Theory, 91–122. New York: Holt, Rinehart and Winston. Bakker, Dik 1994 Some formal and computational aspects of the Functional Grammar machine model. In: Elisabeth Engberg-Pedersen, Lisbeth Falster Jakobsen and Lone Schack Rasmussen (eds), 421–440. 1998 Language change: Some implications for the theory of Functional Grammar. Working Papers in Functional Grammar 66. Amsterdam: University of Amsterdam. Bakker, Dik and Anna Siewierska 2002 Adpositions, the lexicon and expression rules. In: Ricardo Mairal Usón and Maria Jesús Pérez Quintero (eds), 125–177.

62

Matthew P. Anstey

Bauer, Winifred 1980 Review of Dik’s Functional Grammar 1978. Archivum Linguisticum 11: 50–57. Berg, Marinus E. van den 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–106. Berg, Thomas 1998 Linguistic Structure and Change: An Explanation from Language Processing. Oxford: Clarendon Press. Boland, Annerieke 1999 Functional Grammar and first language acquisition: The acquisition of adverbs in a functional approach. Unpublished Master’s thesis. Department of General Linguistics, University of Amsterdam. Bolkestein, A. Machtelt 1985 Cohesiveness and syntactic variation: Quantitative vs. qualitative grammar. In: A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds) (1985b), 1–14. 1992 Limits to layering: Locatability and other problems. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 387–407. 1998 What to do with Topic and Focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. Bolkestein, A. Machtelt, Henk A.Combé, Simon C. Dik, Casper de Groot, Jadranka Gvozdanović, Albert Rijksbaron and Co Vet. 1981 Predication and Expression in Functional Grammar. New York: Academic Press. Bolkestein, A. Machtelt, Casper de Groot and J. Lachlan Mackenzie (eds) 1985a Predicates and Terms in Functional Grammar. Dordrecht: Foris. 1985b Syntax and Pragmatics in Functional Grammar. Dordrecht: Foris. Brown, D. Richard 1985 Term operators. In: A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds) (1985b), 127–145. Bunt, Harry 1985 Mass Terms and Model-Theoretic Semantics. Cambridge: Cambridge University Press. Butler, Christopher S. 1991 Standards of adequacy in a Functional Grammar. Review of Dik’s The Theory of Functional Grammar, Part I: The Structure of the Clause 1989. Journal of Linguistics 27: 499–515. 1999 Nuevas perspectivas de la Gramática Funcional: Los estándares de adecuación de la teoría. In: Javier Martín Arista, Francisco J. Ruiz de Mendoza Ibánez, Ricardo Mairal Usón and Christopher S. Butler (eds), 219–256.

FG from its inception

63

Butler, Christopher S., John H. Connolly, Richard A. Gatward and Roel M. Vismans (eds) 1997 A Fund of Ideas: Recent Developments in Functional Grammar. Amsterdam: IFOTT. Campbell, R. and R. Wales 1970 The study of language acquisition. In: John Lyons (ed.), New Horizons in Linguistics, 242–260. Harmondsworth: Penguin. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Comrie, Bernard 1979 Review of Dik’s Functional Grammar 1978. Studies in Language 3: 267–276. 1980 Comments on Dik’s Functional Grammar. In: Michael Kac (ed.), Current Syntactic Theories: Discussion Papers from the 1979 Milwaukee Syntax Conference, 29–33. Indiana University Linguistics Club. Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds) 1997 Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter. Dahl, Östen 1971 Nouns as set constants. Gothenburg Papers in Theoretical Linguistics 3. Devriendt, Betty, Louis Goossens, and Johan van der Auwera (eds) 1996 Complex Structures: A Functionalist Perspective. Berlin and New York: Mouton de Gruyter. Dik, Simon C. 1966 Review of Robins’s General Linguistics: An Introductory Survey. 1964. Lingua 16: 399–412. 1967a Review of Vachek’s The Linguistic School of Prague: An Introduction to its Theory and Practice 1966. Lingua 18: 80–89. 1967b Some critical remarks on the treatment of morphological structure in transformational generative grammar. Lingua 18: 352–383. 1968 Coordination: Its Implications for the Theory of General Linguistics. Amsterdam: North-Holland. 1973a Crossing co-reference again. Foundations of Language 9: 306–325. 1973b Some remarks on the notion ‘universal semantics’. In: Patrick Suppes, Leon Henkin, Athanase Joja and Gr. C. Moisil (eds), Logic, Methodology, and Philosophy of Science IV: Proceedings of the Fourth International Congress for Logic, Methodology and Philosophy of Science, Bucharest, 1971, vol. 4, 833–843. (Studies in Logic and the Foundations of Mathematics 74.) Amsterdam: North-Holland.

64

Matthew P. Anstey

1975

1977 1978a 1978b 1978c

1979

1980a

1980b 1981a

1981b

1981c

1983a

1983b

1986a

The semantic representation of manner adverbials. In: Albert Kraak (ed.), Linguistics in the Netherlands 1972–1973, 96–121. Assen: Van Gorcum. Stepwise lexical decomposition. Paper presented at the Nijmegen conference on the Foundations of Semantics. Functional Grammar (North-Holland Linguistic Series 37). Amsterdam: North-Holland. Stepwise Lexical Decomposition. Lisse: Peter de Ridder Press. Word order in a Functional Grammar. In: Wolfgang U. Dressler and Wolfgang Meid (eds), Proceedings of the 12th International Congress of Linguistics, 349–351. Innsbruck: Universität Innsbruck. Seventeen sentences: Basic principles and application of Functional Grammar. Publikaties van het Instituut voor Algemene Taalwetenschap, 22. University of Amsterdam. Reply to Comrie’s comments. In: Michael Kac (ed.), Current Syntactic Theories: Discussion Papers from the 1979 Milwaukee Syntax Conference, 35–40. Bloomington: Indiana University Linguistics Club. Studies in Functional Grammar. London and New York: Academic Press. Discrepancies between predication and expression in natural languages. In: A. Machtelt Bolkestein, Henk A. Combé, Simon C. Dik, Casper de Groot, Jadranka Gvozdanović, Albert Rijksbaron and Co Vet, 19–39. Gramática Funcional. Traducción, Glosario de Términos Técnicos e Introducción a la Edición por Fernando Serrano Valverde. Trans. Leocadio Martín Mingorance. Madrid: Sociedad General Española de Librería. Predication and expression: Statement of the problem and outline of the theoretical framework. In: A. Machtelt Bolkestein, Henk A. Combé, Simon C. Dik, Casper de Groot, Jadranka Gvozdanović, Albert Rijksbaron and Co Vet, 1–17. On the status of verbal reflexives. In: Liliane Tasmowski and Dominique Willems (eds), Problems in Syntax, 321–355. (Studies in Language 2.) Ghent: Communication and Cognition. Some basic principles of Functional Grammar. In: Shiro Hattori, Kazuko Inoue, Tadao Shimomiya and Yoshio Nagashima (eds), Proceedings of the XIIIth International Congress of Linguists, August 29–September 4, 1982, Tokyo, 74–88. Tokyo: Tokyo Press. Linguistically motivated knowledge representation. Working Papers in Functional Grammar 9. Amsterdam: University of Amsterdam.

FG from its inception 1986b

65

On the notion ‘functional explanation’. Working Papers in Functional Grammar 11. Amsterdam: University of Amsterdam. 1987 A typology of entities. In: Johan van der Auwera and Louis Goossens (eds), 1–20. 1989a Towards a unified cognitive language. In: Frieda J. Heyvaert and F. Steurs (eds), Worlds behind Words; Essays in Honour of Prof.Dr. F.G. Droste on the Occasion of his Sixtieth Birthday, 97–110. Leuven: University of Leuven. 1989b The Theory of Functional Grammar. Part 1: The Structure of the Clause. Dordrecht: Foris. 1997a The Theory of Functional Grammar. Part I: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part II: Complex and Derived Structures. Berlin and New York: Mouton de Gruyter. Dik, Simon C. (ed.) 1983 Advances in Functional Grammar (Publications in Language Science 11). Dordrecht: Foris. Dik, Simon C., Kees Hengeveld, Elseline Vester and Co Vet 1990 The hierarchical structure of the clause and the typology of adverbial satellites. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 25–70. Engberg-Pedersen, Elizabeth, Lisbeth Falster Jakobsen and Lone Schack Rasmussen (eds) 1994 Function and Expression in Functional Grammar. Berlin and New York: Mouton de Gruyter. Engberg-Pedersen, Elizabeth, Michael Fortescue, Peter Harder, Lars Heltoft and Lisbeth Falster Jakobsen (eds) 1996 Content, Expression and Structure: Studies in Danish Functional Grammar (Studies in Language Companion Series 29). Amsterdam and Philadelphia: Benjamins. Faber, Pamela B. and Ricardo Mairal Usón 1999 Constructing a Lexicon of English Verbs. Berlin and New York: Mouton de Gruyter. Fillmore, Charles J. 1968 The case for case. In: Emmon Bach and Robert T. Harms (eds), Universals in Linguistic Theory, 1–88. New York. Foley, William A. and Robert D. Van Valin Jr. 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Fortescue, Michael 1985 Anaphoric agreement in Aleut. In: A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds), (1985a), 105–126.

66

Matthew P. Anstey

Fortescue, Michael, Peter Harder and Lars Kristoffersen (eds) 1992 Layered Structure and Reference in a Functional Perspective. Amsterdam and Philadelphia: Benjamins. García Velasco, Daniel and Kees Hengeveld 2002 Do we need predicate frames? In: Ricardo Mairal Usón and María J. Pérez Quintero (eds), 95–123. Gebruers, Rudi 1983 S. C. Dik’s Functional Grammar: A pilgrimage to Prague? LAUT Paper. 115: 1–20. Groot, Casper de and Hella Olbertz (compilers) 1999 Functional Grammar Bibliography. www.functionalgrammar.com. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Hannay, Mike and Elseline Vester (eds) 1990 Working with Functional Grammar: Descriptive and Computational Applications. Dordrecht: Foris. Harder, Peter 1990 Tense, semantics and layered syntax. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 139–164. 1992 Semantic content and linguistic structure in Functional Grammar. On the semantics of ‘nounhood’. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 303–327. 1996a Functional Semantics. A Theory of Meaning, Structure and Tense in English. Berlin and New York: Mouton de Gruyter. 1996b Subordinators in a semantic clause structure. In: Betty Devriendt, Louis Goossens and Johan van der Auwera (eds), 93–118. Helmbrecht, J. 1992 An exercise in Functional Grammar - a formal representation of a Lak folk-tale. Unpublished paper. Hengeveld, Kees 1986 Aspects of a Functional Grammar of Spanish [Master’s thesis]. University of Amsterdam. 1987 Clause structure and modality in Functional Grammar. In: Johan van der Auwera and Louis Goossens (eds), 53–66. 1988 Illocution, mood, and modality in a Functional Grammar of Spanish. Journal of Semantics 6: 227–269. 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. 1990a The hierarchical structure of utterances. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 1–24.

FG from its inception 1990b 1992 1997 1999

2000

Hesp, Cees 1990

67

Semantic relations in non-verbal predication. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 101–122. Parts of speech. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 29–55. Cohesion in Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 11–16. Formalizing functionally. In: Michael Darnell, Edith Moravcsik, Fritz Newmeyer, Michael Noonan and Kathleen Wheatley (eds), Functionalism and Formalism in Linguistics (Volume 2: Case Studies), 93–105. Amsterdam and Philadelphia: Benjamins. The architecture of a Functional Discourse Grammar. 9th International Functional Grammar Conference, 20–23 September, 2000. Madrid. [Revised for this volume].

A critique of FG-CNLU. Working Papers in Functional Grammar 35. Amsterdam: University of Amsterdam. Hoekstra, Teun, Harry van der Hulst and Michael Moortgat (eds) 1980 Special edition of Glot 3. 3/4. (Functional Grammar.) Hymes, Dell 1972 On communicative competence. In: J. B. Pride and J. Holmes (eds), Sociolinguistics, 269–293. Harmondsworth: Penguin. 1979 Short notice of Dik’s Functional Grammar 1978. Language in Society 8: 305–306. 1983 Short notice of Dik’s Studies in Functional Grammar 1980. Language in Society 12: 137. Jakobson, Roman 1963 Efforts towards a means-ends model of language in interwar Continental linguistics. In: Christine Mohrmann, F. Norman and Alf Sommerfeldt (eds), Trends in Modern Linguistics, 104–108. Utrecht/Antwerp: Spectrum. Johnson, David E. 1977 On relational constraints on grammar. In: Peter Cole and Jerrold M. Sadock (eds), Grammatical Relations, 151–178. New York: Academic Press. Jong, Jan R. de 1980 On the treatment of focus phenomena in Functional Grammar. Glot 3: 89–115. Katz, Jerrold J. and Paul Postal 1964 An Integrated Theory of Linguistic Descriptions. Cambridge MA: MIT Press.

68

Matthew P. Anstey

Keizer, M. Evelien 1992a Predicates as referring expressions. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 1–27. 1992b Reference, predication and (in)definiteness in Functional Grammar: A functional approach to English copular sentences. Doctoral dissertation, Free University of Amsterdam. Kroon, Caroline 1995 Discourse Particles in Latin. A Study of nam, enim, autem, vero and at. Amsterdam: Gieben. 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 17–32. Kwee Tjoe Liong 1994 Programmer’s Reflections on Functional Grammar: More Exercises in Computational Theoretical Linguistics. (Studies in Language and Language Use) Amsterdam: IFOTT. Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. Liedtke, Frank 1998 Illocution and grammar: A double level approach. In: Mike Hannay and A. Machtelt Bolkestein (eds), 107–128. Longacre, Ronald E. 1965 Some fundamental insights of tagmemics. Language 41: 65–76. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan 1987 The representation of nominal predicates in the fund. Working Papers in Functional Grammar 25. Amsterdam: University of Amsterdam. 1990 Let’s get our heads together: A reply to van der Auwera. In: Mike Hannay and E. Vester (eds), 133–144). 1992 Places and things. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 253–276. 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–296. 2000 First things first. Towards an Incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. Mackenzie, J. Lachlan and María A. Gómez-González (eds) 2003 A New Architecture for Functional Grammar. Berlin and New York: Mouton de Gruyter.

FG from its inception

69

Mairal Usón, Ricardo and Robert D. Van Valin 2001 What Role and Reference Grammar can do for Functional Grammar. Revista Canaria de Estudios Ingleses 42. 137–166. Mairal Usón, Ricardo and María Jesús Pérez Quintero (eds) 2002 New Perspectives on Argument Structure in Functional Grammar. Berlin and New York: Mouton de Gruyter. Martín Arista, Javier, Francisco J. Ruiz de Mendoza Ibánez, Ricardo Mairal Usón and Christopher S. Butler (eds) 1999 Nuevas Perspectivas en Gramática Funcional. Barcelona: Ariel. Martín Mingorance, Leocadio 1998 El Modelo Lexemático-Funcional. (Amalia Marín Rubiales, ed.). Granada: University of Granada. Matthews, Peter H. 1969 Review of Dik’s Coordination: Its Implications for the Theory of General Linguistics 1968. Lingua 23: 349–371. Moutaouakil, Ahmed 1989 Pragmatic Functions in a Functional Grammar of Arabic. Dordrecht: Foris. Nichols, Joanna 1984 Functional theories of grammar. Annual Review of Anthropology 13: 97–117. Nuyts, Jan 1983 On the methodology of a functional language theory. In: Simon C. Dik (ed.), 369–386. 1985 Some considerations concerning the notion ‘psychological reality’ in Functional Grammar. In: A. Machtelt Bolkestein, Casper de Groot and J. Lachlan Mackenzie (eds) (1985b), 91–105. 1992 Aspects of a Cognitive-Pragmatic Theory of Language. On Cognition, Functionalism and Grammar. Amsterdam and Philadelphia: Benjamins. Nuyts, Jan, A. Machtelt Bolkestein and Co Vet (eds) 1990 Layers and Levels of Representation in Language Theory: A Functional View. Amsterdam and Philadelphia: Benjamins. Nuyts, Jan and Georges de Schutter (eds) 1987 Getting One’s Words into Line. On Word Order and Functional Grammar. Dordrecht: Foris. Olbertz, Hella, Kees Hengeveld and Jesús S. Sánchez García (eds) 1998 The Structure of the Lexicon in Functional Grammar. Amsterdam and Philadelphia: Benjamins. Pérez Quintero, María Jesús (ed.) 2001 Revista Canaria de Estudios Ingleses 42. Sección Monográfica: Challenges and Developments in Functional Grammar.

70

Matthew P. Anstey

Pit’ha, Petr 1980

Review of Dik’s Functional Grammar 1978. Journal of Pragmatics 4: 265–270. 1985 Functional Grammar in Holland. The Prague Bulletin of Mathematical Linguistics 44: 63–77. Prideaux, Gary D. 1981 Review of Dik’s Functional Grammar 1978. Language 57: 717–720. Rijkhoff, Jan 1992 The noun phrase: A typological study of its form and structure. Doctoral dissertation, University of Amsterdam. Rijksbaron, Albert 1989 Aristotle, Verb Meaning and Functional Grammar; Towards a New Typology of States of Affairs. With an Appendix on Aristotle’s Distinction between Kinesis and Energeia. Amsterdam: Gieben. Rolland, Colette and Camille Ben Achour 1998 Guiding the construction of textual use case specifications. Data and Knowledge Engineering 25: 125–160. Schaaik, Gerjan van 1983 A functional analysis of aspects of Turkish grammar. Unpublished Master’s thesis. Department of General Linguistics, University of Amsterdam. Searle, John 1969 Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Seuren, Pieter A. M. 1969a Echte en onechte taalkunde. On Dik’s Coordination: Its Implications for the Theory of General Linguistics 1968. De Gids 132: 25–42. 1969b Operators and Nucleus: A Contribution to the Theory of Grammar. (Cambridge Studies in Linguistics 2.) Cambridge: Cambridge University Press. Siewierska, Anna 1991 Functional Grammar. London: Routledge. Steuten, Ans A. G. and Jan L. G. Dietz 1998 Guidelines for bridging the gap between literal sentence meaning and communicative intention. The Language/Action Perspective. 1– 14. Turennout, Miranda I. van 1997 The electrophysiology of speaking: Investigations on the time course of semantic, syntactic, and phonological processing. Doctoral dissertation. MPI Series in Psycholinguistics 1. Katholieke Universiteit Nijmegen, The Netherlands.

FG from its inception

71

Valin, Robert D. Jr. van 1990 Layered syntax in role and reference grammar. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 193–232. Verstraete, Jean-Christophe 2000 Review of Dik’s The Theory of Functional Grammar, Part 1: The Structure of the Clause (2nd edition), Part 2: Complex and Derived Constructions. Language 76: 210–212. 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. Amsterdam: University of Amsterdam. 1998 The multilayered structure of the utterance: About illocution, modality and discourse moves. In: Mike Hannay and A. Machtelt Bolkestein (eds), 1–24. Watters, John R. 1980 Review of Dik’s Functional Grammar 1978. Lingua 50: 155–171. Weigand, Hans 1992 Assessing Functional Grammar for knowledge representation. Data and Knowledge Engineering 8: 191–203. Weijdema, Willy, Simon Dik, Margo Oehlen, Clara Dubber and Akke de Blauw 1982 Strukturen in verbale interaktie. Muiderberg: Coutinho. Weinrich, Uriel 1966 Explorations in semantic theory. In: Thomas A. Sebeok (ed.), Current Trends in Linguistics, vol. 3, 395–477. The Hague: Mouton. Xinzhang Yang 2000 Review of Dik’s The theory of Functional Grammar, Part 1: The structure of the Clause (2nd edition), Part 2: Complex and derived constructions. Word 51: 308–313.

Behind the scenes: Cognition and Functional Discourse Grammar Carlos Inchaurralde

1. Introduction Psychological adequacy has always been a claim of Functional Grammar (Dik 1989: 13), but it has always been problematical in its application (cf. Butler 1991, 1999; Hesp 1990a, 1990b; Nuyts 1990, 1992: 223–236). Hengeveld (this volume) faces some aspects of the problem and describes a new architecture for a Functional Discourse Grammar (FDG), in which a top-down model is adopted and communicative decisions precede representations. This model is coherent with the findings of Levelt (1989) and other psycholinguists, according to whom the speech production process runs from intention to articulation. This is fine as far as production is concerned, but we should not forget that in linguistic interaction there is decoding as well as encoding. The production model describes the process from the communicative intention to the message. However, there is also a reception process on the hearer’s part, in which the message comes first, and from which both the content and the communicative intention are obtained. In the pages below I will try to show how both processes can be integrated into a single model that preserves the architecture proposed by Hengeveld.

2. Modularity of the mind: different kinds of storage Cognition plays an important role in the FDG model as a specific component that interacts with the interpersonal, representational, and expression levels, which have a hierarchical structure.1 The interpersonal level is linked to the representational level, and the interpersonal and representa-

74

Carlos Inchaurralde

tional levels are linked to the expression level by means of expression rules. Obviously, the influence of the cognitive component on any of the levels should have an observable result on the structures produced, and the information that can have such an influence may be of many different kinds. If we assume, as Hengeveld claims, that it “represents the (longterm) knowledge of the speaker, such as his communicative competence, his knowledge of the world, and his linguistic competence” (this volume: 3), it is important to bear in mind that we are dealing with many different types of information that may even have different psychological realizations. In modular approaches such as Fodor’s, our linguistic competence, that is, our capacity to create and interpret language, has many layers that work independently of others: our use of syntax, phonetics, morphology is automatic, with very little influence from central systems. The most important properties of modular systems, according to Fodor (1983), are information encapsulation (they perform the operations that they have been assigned independently of other systems) and domain-specificity (they use specific information that is not shared with other modular systems or central systems).2 It is the autonomous, domain-specific nature of the different ‘modules’, which are specialized in different kinds of tasks, that distinguishes this conception from the holistic one, in which all mental processes share mental resources. In a manner parallel to this modularity of the mind, the architecture of the FDG model assumes some degree of modularity for the different levels in language production, but this structure is described in such a way that there is also the possibility of having some influence from a general cognitive component. Thus, this model allows some room for the claims of the interactionist perspective, which assumes that there is interaction across all mental processes, including the linguistic ones. This is thus an intermediate point of view. But, in any case, the status of ‘linguistic competence’ (or even that of ‘communicative competence’) will need to be very different from that of ‘knowledge of the world’, which has a direct influence on the choice and the interpretation of vocabulary and on the inferences that are possible in discourse. For this kind of information, long-term knowledge could easily be equated with long-term memory.

2.1. Long-term knowledge and encyclopaedic knowledge There is considerable evidence that supports the well-known distinction between two types of memory: short-term and long-term (Loftus and Loftus

Cognition and FDG

75

1976); and the permanent store of our knowledge would correspond to the second kind. A further distinction within long-term memory, i.e. that between episodic memory (in which we recall particular episodes or events) and semantic memory (in which we integrate all our knowledge about the world; Tulving 1972) does not add much, since both kinds may be useful for interacting with the interpersonal, representational and expression levels. Long-term knowledge makes itself very evident in vocabulary interpretation, when we resort to all the encyclopaedic information stored in individual lexical items. Descriptions of how this encyclopaedic information deep in our minds can be used and linked to single words appear in the work of some cognitively oriented authors. Langacker (1987: 162), a cognitive linguist, suggests that lexical units are merely points of access, and that therefore “concepts are simply entrenched cognitive routines”. More precisely: The entity designated by a symbolic unit can (...) be thought of as a point of access to a network. The semantic value of a symbolic unit is given by the open-ended set of relations ⎯ simple and complex, direct and indirect ⎯ in which this access node participates. Each of these relations is a cognitive routine, and because they share at least one component the activation of one routine facilitates (...) the activation of another. (Langacker 1987:163)

Langacker does not suggest this description out of the blue. It is based on proposals by psychologists and computer scientists for models that involve semantic networking. The main hurdle is how to determine, for every access node, the extent of the corresponding conceptual network. Since it is possible to think of a network in terms of an unending chain of associations between nodes, there must be some mechanism, congruent with current psychological evidence, that helps determine which nodes get a place in the network, in other words which nodes are ‘relevant’. The keyword here is activation. This is a term used in connectionism to designate the degree of strength that the different nodes in a neural network may have. Models of cognition that incorporate connectionist ideas use this term profusely. Anderson (1983), for instance, in his ACT model, distinguishes declarative memory from production memory and working memory. To retrieve information from declarative memory so that we can use it in our working memory, a certain amount of energy is needed in the form of activation. The various modes of activation for concepts in a network explain the link between words and encyclopaedic knowledge dynamically.

76

Carlos Inchaurralde

Also from psychology come the now well-established assumptions of so-called ‘schema’ theories (Bartlett 1932; Rumelhart and Ortony 1977; Rumelhart 1980), according to which everything we know about the world is stored in large mental knowledge structures. Different labels have been used to refer to these structures, depending on the type of information and the way it is stored: ‘frames’, ‘scenarios’, ‘scripts’, ‘situations’, ‘idealised cognitive models’, and so on (Minsky 1975; Schank and Abelson 1977; Sanford and Garrod 1981; Fillmore 1982; Lakoff 1987, etc.) are all equally useful in understanding how people use vocabulary in context. For instance, Wierzbicka (1994, 1997) has pointed to the existence of different ‘scripts’ for interaction in different cultures and has drawn attention to how words have uses and meanings that are highly dependent on culturallyshared knowledge structures. All this seems to imply that it is necessary to leave behind us the fixed-entry approach, provided by a dictionary perspective on the lexicon, and adopt some sort of distributed, encyclopaedic view. However, at the same time, from an FG perspective, we do need fixed entries in the Fund, where all predicate frames and terms are placed at our disposal for the building of predications, and this is the place where the cognitive component can act as input. The cognitive component is the interface between our mental encyclopaedia and the lexicon that is available to us. The cognitive component provides a ‘lexicopaedia’, in the form of a fixed-entry list with points of access to a whole semantic network (cf. Inchaurralde 2000). We cannot deny the evidence that shows us how individual lexical units are a psycholinguistic reality,3 which gives psychological support to the existence of a lexicon in the Fund; but at the same time it is important that we admit that there exists semantic information of an encyclopaedic nature, and this is a natural and simple way of incorporating it in the FDG model.

2.2. Long-term knowledge and the communicative component Knowledge of the relevant parameters that can affect the coding of the interpersonal level should also be part of the long-term information held by the cognitive component, especially if we bear in mind that this is the first level in Hengeveld’s model. There are elements in the communicative context which, once they become available to the cognitive component, are relevant in further processing, and this fact is even more apparent in languages with a strong grammaticalization of pragmatic phenomena that deal with the social relationship between the interlocutors in the communicative situation. Japanese, for instance, attaches a lot of importance to the per-

Cognition and FDG

77

ceived distance between the interlocutors, which relates to two different factors that work together. One factor is politeness, which is in turn related to a strong need to keep up one’s face, as well as the other person’s, in Japanese culture. The second one is the strength of hierarchical relations in a group-oriented culture. In Japanese, apart from the neutral ‘unmarked’ register, there is a special register, which is called keigo and has three distinct sub-registers, sonkei-go (respectful language), kenjoo-go (humble language), and teinei-go (honorific language). Sonkei-go is used to show respect towards a person whom we place ‘above’ us. This person may have higher social status or may be somebody who deserves good treatment (a client, or somebody we have just met for the first time). Kenjoo-go is used by speakers who place themselves ‘below’ their interlocutors. Speakers may also show respect by downgrading their own ‘face’ before others’. Finally, teinei-go is used when formal language is required. Since this register can be used for any topic of conversation, its main function is simply to raise the level of formality involved. These registers involve special choices in the vocabulary employed, but they also require changes in the construal of the underlying situation, which in turn influences its representation. Thus, in the FDG model, a change of register would involve a different specification in the interpersonal level, especially concerning the speaker (Ps) and the addressee (Pa) and their relationship, which cannot be properly codified in the hierarchical structure as it stands. For Japanese, something like (Ps = Low) for humble language, or something like (Pa = High) for respectful language could be used, or a grading system which could show the relative social distance between both. For instance, (Pa = 10) and (Ps = 5) would show that the addressee is socially higher than the speaker. It is important to notice that theoretically it is possible to have (Ps = High) if the speaker is the Emperor, or some kind of god in a tale, but the possibility is very unusual. In English, it could be compared to a character speaking with third person verbal forms instead of using the first person. In Japanese, as there are no different verbal forms according to person, it is the special register used that helps us to know about whom something is being said. The cognitive component, according to the specification present in the interpersonal level, would then choose a suitable construal for the representation of content. Let us give an example. In sonkei-go, some verbs change. Instead of taberu ‘to eat’, or nomu ‘to drink’, Japanese speakers use the expression meshi-agaru, which literally means ‘to ascend food’. Instead of miru ‘to see, to look’, we have goran ni naru ‘to become a viewing’. In general, the expression o + verb root + ni naru can be used, so that we

78

Carlos Inchaurralde

have, for instance, the form o-kaki ni naru ‘to become a writing’ instead of kaku ‘to write’. These forms are very lexicalized, but they still involve differences in cognitive construal that require a different form at the representational level. In these cases, even if we do not accept changes at this level, it is clear that they do require different forms at the expression level, which justifies the input from both the cognitive component and the interpersonal level. There are many expressions in kei-go that differ only lexically. We have the honorific prefixes o- and go- (e.g. o-tomodachi ‘his friend’ instead of tomodachi ‘friend’), alternative forms (e.g. achira instead of asoko ‘there’), and special words for the different kinds of kei-go (e.g. iku ‘go’ and kuru ‘come’, which both change into irassharu in sonkei-go and into mairu in kenjoo-go and teinei-go). Of course, the information about the relative social distance of both interlocutors is part of the communicative settings. General knowledge about different kinds of communicative settings (as well as of different text ‘genres’ or conventional text structures), with their implications for different performative acts – cf. Austin’s (1962) mention of ‘felicity conditions’ – or for the use of different communicative functions, is also part of the longterm information stored in the cognitive component.

2.3. Linguistic competence and knowledge of the world Hengeveld refers to linguistic competence as being part of the long-term knowledge available to the cognitive component. If by linguistic competence we mean knowledge of the phonology, grammar, and vocabulary of a language, then obviously we would have to include here the knowledge of all vocabulary items and all rules of the grammar, as well as phonological rules. However, although the form and certain combinatorial characteristics of vocabulary items can be considered to be linguistic knowledge, the above-mentioned ‘lexicopaedic’ information seems to go beyond this kind of knowledge. It seems more suitable to consider lexical linguistic information to be specifically the kind of information that is available in the lexicon for predicates in terms of form, syntactic category, quantitative valency, qualitative valency, selection restrictions, meaning postulates, and meaning definitions (Dik 1989: 68ff.). This kind of information, stored in long-term memory, is useful for a metalinguistic usage of language, and for monitoring language production, but vocabulary choice seems to have some degree of automatization. The same can be said of syntactic and phonological rules. There seems to be some evidence of independent modular capacities that deal with syntax and phonology in

Cognition and FDG

79

an autonomous way (as can be seen, for instance, in the study of certain aphasias, as in Saffran 1982, Saffran et al. 1980, or Schwartz 1987); but, at the same time, there is also some evidence that interaction does take place (cf. Olson 1973, Shatz 1977). Another consideration is that we have a certain knowledge of the structure of external reality, which is based on bodily and sensory experience, as cognitive linguistics has argued (cf. Radden 1992, Lakoff 1990). This means that we have an internal representation of that structure in terms of entities that interact in time and in space, with certain properties and certain relations, and we also have an idea of the causal structure of events in the world. This long-term knowledge of an experiential nature can be mapped onto knowledge about word classes and grammar categories (nouns, verbs, categories of number, person, tense, aspect, mood, modality, etc.) and this mapping can be assumed to take place in the cognitive component, which is like a black box to the Functional Discourse Grammar model, since the three levels in this model can receive its input but can neither see nor influence its contents.

3. The computer metaphor The most interesting feature of the Functional Discourse Grammar model, however, may not lie in its structure, but rather in its functioning. I therefore want here to put forward a conceptualization, congruent with the encoding-decoding process that takes place in actual communicative interaction, of what this functioning might look like. In order to do this, I will use a model from computer science. Cognitive psychology has always been characterized, since its start in the late fifties, by its repeated use of the computer metaphor in all the models it has presented. Broadbent (1968), for instance, used flow diagrams to show the workings of our faculty of attention. Flow diagrams have also been used to show the different kinds of memories (long-term memory, short-term memory, and, later on, working memory) as storage places out of which or into which a flow of information comes and goes (Atkinson and Shiffrin 1968, Baddeley and Hitch 1974). Now that we have entered the 21st century, psychologists do not feel so enthusiastic about using flow diagrams, so that the situation has been reversed, with computer scientists now looking to the brain itself for a model in the form of connectionism. However, there is still room for the computer metaphor in the explanation of the mind’s workings, and obviously there is room too for that

80

Carlos Inchaurralde

metaphor in the explanation of linguistic activity, which clearly is cognitive in nature. Functional Grammar has always followed the computer metaphor, or rather, the computer-language metaphor, with its Prolog-like formalisms which make it so easy to implement in computer programs. Nevertheless, Functional Grammar has always had the same operational problem as Prolog: the logical relations are clearly expressed, and unification and deduction are easy; but the actual procedure, the real machinery that works behind the scenes, is hidden from us and has no specification – not even a tentative one – of any kind. What I suggest here is that we again resort to the computer metaphor when looking at the processes that are covered in Hengeveld’s model. If we take the two sides of communicative interaction, the speaker’s and the addressee’s, as the two poles of a chain of actions, then the entire encoding-decoding process would follow a process similar to that which is currently used when two operating systems are connected by means of a protocol. The layer model is customary in the architecture of present-day operating systems. In the OSI Standard (Open Systems Interconnection Reference Model),4 for instance, there are several different layers: the physical layer, the data link layer, the network layer, the transport layer, the session layer, the presentation layer, the application layer, and the user program. The different layers may generate requests for those below them (the lowest of all being the physical layer). For instance, a word processor (the user program layer) may generate a request for a certain routine that opens windows with some given colours and a given size (the presentation layer). This means that the user program layer generates a request which is passed on to the application layer and then goes on down to the presentation layer. This layer performs the routine and sends an acknowledgement primitive that goes up to the user layer again to acknowledge receipt of the request and the performance of the action requested (see Figure 1). In this respect, the Functional Grammar model proposed assumes this kind of communication moving up and down the different layers. However, the interesting point in our use of the operating systems metaphor is how we can employ the basic model for communication across local area networks as a model for natural language communication between individuals in a Functional Discourse Grammar framework. In modern information systems, the communication is arranged in fixedlength chunks of data called packets. These packets, as well as handshaking messages for regulating the communication, are transmitted between two interconnected systems as ‘frames’ created by a data-link layer and then transmitted by the physical layer. A request from the user program

Cognition and FDG

81

triggers another request from the application layer, in the form of a ‘frame’ made up of some data and a header. User program

REQUEST

ACKNOWLEDGEMENT+ROUTINE

REQUEST

ACKNOWLEDGEMENT+ROUTINE

Application layer

Presentation layer

Session layer

…

…

…

Figure 1. An example using the OSI model

This request goes down to the next layer, where it is taken as the data of a further packet with a new header; it then goes down to the next layer, and so on. In the entire process down to the physical layer, where the flow of 1s and 0s is transmitted along the communication lines, packets of a given level are included as data in packets of the next level, creating a multilayered structure that is very familiar to Functional Grammar researchers. This is the encoding process. At the other end, the receiver of these packets starts the process of unpacking the chunks of information received. At the data-link layer a packet is obtained with a header and data. If the data has some instruction that can be performed at this level, the process stops here; but, if the data is a packet of a higher level, the unpacking process goes on until it reaches the level at which the data contains instructions that can be executed. In our analogy, the sender operating system would be the encoder of the linguistic message, and the receiver operating system the decoder. Any request generated at a given level would need to be proc-

82

Carlos Inchaurralde

essed at the same level by the receiver of the message (see Figure 2).

Level of transaction

Actual communication (Physical layer)

Figure 2. The encoding-decoding process

Now, since this procedure actually works in local area networks and is a fine model of phenomena that take place in real language, we can adopt it as a procedural basis for Hengeveld’s model. Let us have a look at some examples. Let us imagine a Japanese businessman who is waiting for a superior. When he meets him, he says: (1)

Aa, Tanaka-san desu ka. Omachi shite orimashita ‘Oh, Mr. Tanaka. (I) have been waiting (for you)’

The process that has taken place is as follows. First, he has to encode his communicative intention. The cognitive component comes into action, retrieving long-term information about communicative settings and the different possibilities available and also drawing information from the communicative context. The interactants have a different relative position on the social scale, and this fact should be encoded at the interpersonal level. The ‘request’ that is sent to the first layer is formalized as follows: (2)

[A1: [DECL (P1 = Low)Sp (P2 = High)Addr (C1: [...] (C1))] (A1)]

This level sends the ‘request’ to the next layer (= level in FDG), but apparently it does not change the representation of the propositional content,

Cognition and FDG

83

which stays as it would have been, without any indication of the difference in relative social power between the interlocutors: (3)

[pi: [[PastProg ei: matsuv (x1)Ag (x2)Go] (ei)] (pi)]

This means that the request has to go down to the next layer (= level in FDG), where it produces some changes at the expression level. More concretely, the form o-machi shite is employed instead of matte ‘waiting’ (from matsu ‘to wait’); likewise, the form orimashita is used instead of imashita (‘was’ or ‘had been’ in this example). It is at this level (the equivalent to the physical layer in the computer communication model) that the form is transmitted. Then, in decoding, the process is repeated, but this time it goes upward from the most ‘physical’ level to the actual level at which the request is processed. The ‘package’ is fetched and ‘unpacked’ at the expression level. The data obtained from processing the expression then goes to the representation level. At this level, the representation can be correct from a propositional point of view, but there is still a package of data to be processed which cannot be interpreted there. The package then goes up to the interpersonal level, where it gets the interpretation it needs, which is important, because the addressee needs to evaluate the adequacy of the interpersonal settings that the speaker is using for the current communicative context. At all stages, of course, the receiver has needed to consult the long-term information stored in the cognitive component in order to be able to interpret the data packages. This approach is also compatible with discourse processing, since both macro-planning (concerned with the preparation of large stretches of speech) and micro-planning (concerned with grammatical and lexical choices; cf. Butterworth 1980: 159) can be encoded in some way in the data packages involved. There are some marks at the expression level that clearly encode information relevant to the interpersonal level (e.g. discourse markers, which show relations in discourse). This can also be applied to pauses, which are significant as indicators of processing time. As Butterworth (1980: 155) explains, “(...) the more the delays, the more cognitive operations are required by the output”. As it can be assumed that long pauses reflect more complicated cognitive processes than short pauses, they can be signals of discourse planning and show relevant landmarks of the cognitive processes involved (cf. Schilperoord 1998). As they are involuntary, it is questionable whether we should think of pauses as some kind of unit analysable at the expression level, but there is no doubt that the addressee can use them to reconstruct discourse structure.

84

Carlos Inchaurralde

4. Conclusion We have examined here, from a cognitive perspective, some questions which are relevant to the role of cognition in the Functional Discourse Grammar model proposed by Hengeveld (this volume). His model assigns a very important role to the cognitive component, since cognition provides input to all three levels proposed. However, cognition itself remains rather mysterious to the model. The view adopted here is that this cannot be otherwise, since cognition is situated outside the main core of the model. Nevertheless, some assumptions can be made about the role of cognition, particularly with regard to the various inputs from the cognitive component. The cognitive component contains not only general knowledge about the world, but also knowledge about relevant communicative parameters and about linguistic competence itself. We have argued that these are different kinds of information. Some are linked to more central cognitive systems in the mind, while some others are not if we accept the evidence supporting the modular approach proposed by Fodor (1983). A computer metaphor has also been put forward that helps conceptualize the workings of the model in a more understandable way, and cognition has its place in it. However, we must realize that while cognition, as a general capacity, cannot be ignored within the Functional Discourse Grammar model, it is wise to assume that giving a very detailed account of its internal workings would go beyond the scope of this grammatical approach and would lead us into matters of a fully psychological import. And this would take us away from our main object of study, the linguistic message.

Notes 1. 2.

3.

Note that I am referring to modularity of the mind and not to modularity within the FG model. Other properties of modular systems (cf. Fodor 1983) are that they have a compulsory operation, which is automatic and independent of voluntary processes; they work at a speed faster than other cognitive processes; their output representations are incomplete and need to be completed by central systems, which have more complete information; they are innate, since they have a pattern of ontogenetic development; they are realized by means of a fixed neural structure, that is, they tend to be clearly localized in certain fixed areas of the brain; moreover, their alterations are highly specific and selective, by contrast with what happens in central systems. I give some examples in Inchaurralde (2000: 98–100).

Cognition and FDG 4.

85

The Open Systems Interconnection Reference Model is a set of standards developed by the Open Interconnection Committee – sponsored by the International Standardization Organization (ISO) – to facilitate interconnectivity between computer systems.

References Anderson, John R. 1983 The Architecture of Cognition. Cambridge: Harvard University Press. Atkinson, Richard C. and Richard M. Shiffrin 1968 Human memory: a proposed system and its control processes. In Spence, Kenneth W. and Janet T. Spence (eds), The Psychology of Learning and Motivation (vol. II), 89–195. New York: Academic Press. Austin, John L. 1962 How to Do Things with Words. Oxford: Clarendon Press. Baddeley, Alan D. and Graham Hitch 1974 Working memory. In Gordon H. Bower (ed.), The Psychology of Learning and Motivation (vol. VIII), 47–89. New York: Academic Press. Bartlett, Frederick C. 1932 Remembering. A Study in Experimental and Social Psychology. Cambridge: Cambridge University Press. Broadbent, Donald E. 1968 Perception and Communication. London: Pergamon. Butler, Christopher 1991 Standards of adequacy in Functional Grammar. Journal of Linguistics 27: 499–515. 1999 Nuevas perspectivas de la Gramática Funcional: los estándares de adecuación de la teoría. In Christopher Butler, Ricardo Mairal Usón, Javier Martín Arista and Francisco José Ruiz de Mendoza (eds), Nuevas perspectivas en Gramática Funcional, 219–256. Barcelona: Ariel. Butterworth, Brian 1980 Evidence from pauses in speech. In Butterworth, Brian (ed.), Language production: Volume 1. Speech and talk, 155–176. London: Academic Press. Dik, Simon C. 1989 The Theory of Functional Grammar. Part I: The Structure of the Clause. Dordrecht: Foris.

86

Carlos Inchaurralde

1997a

The Theory of Functional Grammar. Part I: The Structure of the Clause. (Second, revised edition. Edited by Kees Hengeveld.) Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part II: Complex and Derived Constructions. (Edited by Kees Hengeveld.) Berlin and New York: Mouton de Gruyter. Fillmore, Charles J. 1982 Frame semantics. In Linguistic Society of Korea (ed.), Linguistics in the Morning Calm, 111–137. Seoul: Hanshin. Fodor, Jerry A. 1983 The Modularity of Mind. Cambridge MA: MIT Press. Hengeveld, Kees This vol. The architecture of a Functional Discourse Grammar. Hesp, Cees 1990a A Critique of FG-CNLU. Working Papers in Functional Grammar 35. Amsterdam: University of Amsterdam. 1990b The Functional Grammar computational natural language user and psychological adequacy. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 295–312. Inchaurralde, Carlos 2000 Lexicopedia. In Peeters, Bert (ed.), The Lexicon-Encyclopedia Interface. Amsterdam: Elsevier, 97–114. Lakoff, George 1987 Cognitive models and prototype theory. In: Neisser, Ulric (ed.), Concepts and Conceptual Development. Ecological and Intellectual Factors in Categorization, 63–100. Cambridge: Cambridge University Press. 1990 The invariance hypothesis: Is abstract reason based on imageschemas? Cognitive Linguistics 1: 39–74. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. (Vol 1. Theoretical Prerequisites.) Stanford: Stanford University Press. Levelt, Willem J. M. 1983 Monitoring and self-repairs in speech. Cognition 14: 41–104. 1989 Speaking: From Intention to Articulation. Cambridge MA.: MIT Press. Loftus, Geoffrey R. and Elizabeth F. Loftus 1976 Human Memory: The Processing of Information. Hillsdale NJ: Lawrence-Erlbaum. Minegishi Cook, Haruko 1996 The Japanese verbal suffixes as indicators of distance and proximity. In Pütz, Martin and René Dirven (eds), The Construal of Space in

Cognition and FDG

87

Language and Thought, 3–27. Berlin and New York: Mouton de Gruyter. Minsky, Marvin L. 1975 A framework for representing knowledge. In: Winston, Patrick H. (ed.), The Psychology of Computer Vision, 211–280. New York: McGraw-Hill. Nuyts, Jan 1990 Linguistic representation and conceptual knowledge representation. In: Jan A. Nuyts, A. Machtelt Bolkestein and Co Vet (eds), 263–293. 1992 Aspects of a Cognitive-Pragmatic Theory of Language. Amsterdam and Philadelphia: Benjamins. Nuyts, Jan, A. Machtelt Bolkestein and Co Vet (eds) 1990 Layers and Levels of Representation in Language Theory. Amsterdam and Philadelphia: Benjamins. Olson, Gary M. 1973 Developmental changes in memory and the acquisition of language. In: Timothy E. Moore (ed.), Cognitive Development and the Acquisition of Language, 145–157. New York: Academic Press. Radden, Günter 1992 The cognitive approach to natural language. In: Martin Pütz (ed.), Thirty Years of Linguistic Evolution, 513–541. Amsterdam and Philadelphia: Benjamins. Rumelhart, David E. 1980 Schemata. The building blocks of cognition. In: Rand J. Spiro, Bertram C. Bruce and William F. Brewer (eds), Theoretical Issues in Reading Comprehension. Perspectives from Cognitive Psychology, Linguistics, Artificial Intelligence, and Education, 33–58. Hillsdale NJ: Erlbaum. Rumelhart, David E. and Andrew Ortony 1977 The representation of knowledge in memory. In: Richard C. Anderson, Rand J. Spiro and William E. Montague (eds), Schooling and the Acquisition of Knowledge, 99–135. Hillsdale NJ: Erlbaum. Saffran, Eleanor M. 1982 Neuropsychological approaches to the study of language. British Journal of Psychology 73: 317–337. Saffran, Eleanor M., Myrna F. Schwartz and Oscar S. M. Marin 1980 Evidence from aphasia: Isolating the components of a production model. In: Brian Butterworth (ed.), Language Production. Volume 1: Speech and Talk, 221–241. London: Academic Press. Sanford, Anthony J. and Simon C. Garrod 1981 Understanding Written Language. New York: Wiley.

88

Carlos Inchaurralde

Schank, Roger C. and Robert P. Abelson 1977 Scripts, Plans, Goals and Understanding. Hillsdale NJ: Erlbaum. Schilperoord, Joost 1996 It’s about Time. Temporal Aspects of Cognitive Processes in Text Production. Amsterdam: Rodopi. Schwartz, Myrna F. 1987 Patterns of speech production deficit within and across aphasia syndromes: Application of a psycholinguistic model. In: Max Coltheart, Giuseppe Sartori and Remo Job (eds), The Cognitive Neuropsychology of Language, 163–199. Hillsdale, NJ: Erlbaum. Shatz, Marilyn J. 1977 The relationship between cognitive processes and the development of communication skills. In Charles B. Keasy (ed.), Nebraska Symposium on Motivation 1977, 1–42. Lincoln: University of Nebraska Press. Tulving, Endel 1972 Episodic and semantic memory. In: Endel Tulving and Wayne Donaldson (eds), Organization of Memory, 382–403. New York: Academic Press. Wierzbicka, Anna 1994 Cultural scripts. A new approach to the study of cross-cultural communication. In Martin Pütz (ed.), Language Contact, Language Conflict. Amsterdam and Philadelphia: Benjamins, 69–87. 1997 Understanding Cultures through their Key Words. English, Russian, Polish, German, and Japanese. Oxford: Oxford University Press.

The Question of Discourse Representation in Functional Discourse Grammar John H. Connolly

1. Introduction Ever since its inception, FG has been a pragmatically oriented theory, and a good deal of pragmatically based research on the grammar of natural languages has been carried out within its framework. However, it was not until the publication of Dik (1997a, 1997b) that the foundational text of FG came to include a specific chapter devoted to discourse, focusing upon units of language larger than the clause. In the field of language studies outside of FG, there has, of course, been a considerable amount of research into discourse from a variety of different standpoints, stretching back for many years. Some, though by no means all, of the principles established in this pre-existing work are incorporated into the final chapter of Dik (1997b). Now Hengeveld (this volume) has proposed a new architecture for FG, which he terms Functional Discourse Grammar (FDG). Significantly, this framework accommodates discourse as an integral part of the model. What has not yet evolved, however, is a consensus on the issue of how to represent discourse. Hence, the present paper will be concerned with a discussion of this question.

2. The Treatment of Discourse 2.1. Discourse in Dik’s FG Framework First, though, we need to conduct a brief overview of previous research. An appropriate place to begin is with a consideration of the scope of discourse.

90

John Connolly

In other words, what phenomena belong under the heading of discourse studies and thus merit consideration in respect of the design of frameworks for discourse representation? We may begin by noting that discourse can be regarded either as a process of communicative interaction or as the product of that process. Both these perspectives need to be accommodated, and we shall seek to do so here, using the term ‘discourse’ in whichever sense is appropriate in any particular context. The approach to discourse studies presented in Dik (1997b) may be summarized as follows. Discourse as a product is viewed as a coherent stretch of spoken or written language, which normally contains more than one clause, and which may also contain one or more extraclausal constituents (ECCs). Discourse as a process can take the form of a monologue or dialogue, and the interlocutors (or participants) are labelled S (for speaker/writer) and A (for addressee) as appropriate. Any discourse process will be embedded within a Discourse Event, which is a type of social phenomenon. There are several important aspects of discourse which apply to the event rather than the content of the language employed. These include (i) the identity of the interlocutors and their relationship to each other, (ii) the time, place and setting of the discourse, and (iii) the prevailing social conventions which govern the acceptable use of language. When embarking upon a discourse, several decisions need to be made which tend to affect the construction of the whole piece of language concerned. These include the choice of genre (such as a lecture or an interview) and the selection of style (which may be formal or informal, polite or impolite, and so forth). As the discourse proceeds, the interlocutors develop a mental picture which enables each new utterance to be interpreted in the light of what has gone before, and also in the context of the interlocutors’ pragmatic information concerning the discourse situation and, indeed, the wider universe. This evolving mental representation, which has a degree of commonality among all the interlocutors, is termed a Discourse Model (DM). Because it is constantly changing, a DM is dynamic, and because it can never include all the pragmatic information that is conceivably relevant, it constitutes only a partial model of the discourse concerned. As stated in Dik (1997a: 300–304), each individual clause has a basic illocutionary value which is to be viewed as a coded instruction whereby S seeks to modify A’s pragmatic information. The possible illocutionary values are declarative, interrogative, imperative and exclamative. These types are recognized on the basis that they tend to be coded differentially in the

The question of discourse representation

91

grammars of natural languages, unlike the speech act categories of, for example, Searle (1976). However, in Dik (1997b: 419–421) it is proposed, in addition, that whole discourses, or sections thereof, may have illocutionary values, which are duly termed Discourse Illocutions, and which set the default values of the individual clauses within the stretch of language concerned. Dealing with illocutions in terms of discourse also involves distinguishing between (i) the illocution as intended by S, written ILLS, (ii) the illocution as coded in the expression, written ILLE, and (iii) the illocution as interpreted by A, written ILLA; see Dik (1997b: 230). This entails countenancing the use of speech act categories additional to the four grammatically encoded values; see Connolly et al. (1997a) for an example of the use, within the FG tradition, of categories proposed by Searle (1976) and Hancher (1979). Certain other phenomena, too, tend to persist for whole sequences of clauses or indeed for entire discourses. Such phenomena include tense and discourse topic. Furthermore, the maintenance of topical continuity is connected with the creation of chains of anaphoric reference, which frequently extend across sentence boundaries. Within the framework of Dik (1997b), a discourse event is structured in two layers. The first is the Interpersonal Layer, which is concerned with the human relations between the interlocutors and with their mentalities in relation to the discourse and to each other. The interpersonal layer is subdivided into the Interactional, which focuses on the interrelationship and interplay between the participants, and the Attitudinal, which is concerned with their states of mind, emotions and judgments, as reflected in the discourse. The second layer is the Representational, which is related to the actual language produced by the interlocutors. This layer is subdivided into the Organisational, which deals with how the material that the discourse consists of is arranged and presented, and the Contentive, which is concerned with the facts and States of Affairs communicated. Discourse has (it is claimed) an essentially hierarchical structure, though sometimes this is complicated by phenomena such as interruptions, embeddings and recursions. The hierarchical structure can be discerned within both the interpersonal and representational layers. In the interpersonal layer the lowest-ranking unit is the speech act (which will carry a particular illocutionary value). Units of this type can combine into longer sequences of speech acts. In a dialogue, sequences of one or more speech acts are further organized into turns, and the whole sequence of turns makes up the discourse. As a complication, patterns known as Adjacency Sequences (such as question-answer or question-answer-reaction) can often

92

John Connolly

be identified as successions of speech acts over two or three turns. From the point of view of the representational layer, on the other hand, the discourse is divided into episodes, sub-episodes, sub-sub-episodes and so on (for example, chapters, sections and paragraphs in a book). Ultimately, these units are composed of individual propositions, which constitute the lowest-ranking type of discourse unit within the representational layer. Another aspect of discourse appears in the form of rhetorical relations, such as elaboration or contrast. A well-known approach to these phenomena is found in the Rhetorical Structure Theory (RST) proposed by Mann and Thompson (1987) and discussed in relation to FG by Gulla (1997). In this approach, an individual rhetorical relation associates two stretches of discourse, one of which is called the nucleus and the other the satellite. For instance, an elaboration is considered to be the satellite of what it elaborates, the latter being the nucleus. Dik (1997b) accepts that rhetorical relations of the kind recognized in RST should be accommodated into the FG-based approach to discourse.

2.2. The advent of a modular framework If Dik’s (1997b) approach to discourse is accepted as being indicative of the scope of discourse studies as far as FG is concerned, then we may safely assume FDG to be in conformity with this. Nevertheless, there still remains the question of how exactly to accommodate discourse within the descriptive framework. Let us, therefore, now summarize how this issue has been addressed within FG from 1997 onwards. One approach to discourse has been to treat it as an additional level (or set of levels) above the clause. This approach is described in Hengeveld (1997), and has become known as upward layering. The additional level was originally called the Rhetorical Layer, wherein speech acts are grouped into moves, and moves are organized into a particular type of discourse (such as a conversation). A move is realized by a paragraph within a monologue or by a turn within a dialogue. This view of discourse organisation is very reminiscent of the Sinclair and Coulthard approach to discourse, where acts are grouped into moves, moves into exchanges, exchanges into transactions, and transactions into interactions; see Coulthard (1977: 107). In fact, Steuten’s (1998a, 1998b) work on the analysis of business conversations leads her to suggest replacing Hengeveld’s original rhetorical layer with an Interactional Layer in which acts and exchanges comprise the main organisational units within dialogue.

The question of discourse representation

93

Some writers have urged a stronger separation between grammar and discourse than is entailed by upward layering, contending that discourse should be handled within a separate module. The latter is termed the Pragmatic Module (as opposed to the Grammatical Module) by Vet (1998a, 1998b), Bolkestein (1998) and Liedtke (1998), the Discourse Module (as against the Sentence Module) by Kroon (1997), and the Communicative Unit (as distinct from the Grammatical Unit) by Steuten (1998a, 1998b). Kroon argues that the lowest-ranking unit in the discourse hierarchy, which she calls the Discourse Act, is not necessarily equivalent to the clause, since ECCs, too, can function as discourse acts. For Vet and Steuten, the difference in nature between the process of discourse and the product of that process, in the form of clauses, motivates their being treated separately in the descriptive framework. In fact, the idea of some kind of modular framework seems also to be envisaged in Dik (1997b: 409). In Vet's model the grammatical module and the pragmatic module are linked by an interface. Van den Berg (1998) develops this interface into a level in its own right, calling it the Message Module. In Van den Berg's framework, the pragmatic module is the controlling unit of the discourse process, and deals with both the construction of moves and with the management of the social context within which the discourse takes place. The grammar module provides the lexical and grammatical material for the composition of the message, but the latter is actually put together by the message module in a manner determined by instructions from the pragmatic module. In Hengeveld’s FDG, three levels are recognized, namely the Interpersonal Level, the Representational Level and the Expression Level. Moreover, each of these levels is conceived of as a separate module, and each is internally layered. Rules serve to connect the modules, as appropriate. In this way, Hengeveld seeks to combine the best points of previous approaches. Another important advance is the recognition in FDG of a communicative component and a cognitive component. Both of these comprise essential elements of the context with which the three language modules interact.

2.3. Other areas of discourse studies The FG approach to discourse embraces numerous phenomena. Nevertheless, it by no means encompasses everything that is addressed in the broader field of contemporary discourse analysis. It will be appropriate,

94

John Connolly

therefore, to mention briefly a selection of other areas which at least some linguists would also consider to lie within the scope of discourse studies. Firstly, during the last twenty years or more, there has been a good deal of research into the relationship between language, power and ideology. This area of work is often termed Critical Language Study or Critical Discourse Analysis; see, for instance, Fairclough (1989) or Fairclough and Wodak (1997). Its central premise is that power is unevenly distributed in human society, and that existing inequalities of power are maintained partly with the aid of sets of supposedly common-sense assumptions, known as ideologies. For example, the prescriptive linguistic ideology that there is a correct way of speaking and writing English is still accepted without question by many people in the UK, and its maintenance, with the help of expressions like ‘bad grammar’, is something that tends to suit the purposes of the establishment. Authors in the tradition of Systemic Functional Grammar have made moves towards incorporating ideology, along with genre, into their theoretical framework; see, for instance, Martin (1992: 496). However, this has not become common practice within other approaches. Secondly, the work of Bakhtin has attracted a fair amount of interest; see, for instance, Clark and Holquist (1984). In particular, we should mention here his theory of Voice. To take an example, suppose that speaker S1 makes some utterance U1, and that this event is subsequently recounted by another speaker S2 in the form of utterance U2, such that U2 contains a piece of reported speech, structured along the lines of ‘S1 said U1’. Utterance U1, when thus embedded into utterance U2, has, in a sense, two speakers: S1 (its originator) and S2 (its reporter), and their voices intermingle into a polyphony. It is also possible to push this notion of voice further, until it is no longer necessarily identifiable with a particular speaker, for instance when someone is said to speak with the voice of a scientist. This implies the recognition of a generic as well as an individual voice as being a possible component within a polyphony. Thirdly, Kress and Van Leeuwen’s semiotically based approach to discourse deserves to be taken into account here. Kress and Van Leeuwen (1996: 39) remind us that text is in general multimodal. For example, speech is typically accompanied by non-verbal communication (NVC), such as gesture. Indeed, NVC can be used as a substitute for speech within a dialogue; for instance, from the point of view of communication, a nod is as effective as the word yes in response to a question (assuming a face-toface situation). Furthermore, written text has additional, non-linguistic but nevertheless communicative dimensions, such as the layout of a document,

The question of discourse representation

95

or the type and quality of the surface on which the text is found. In short, discourse can reasonably be viewed as a multimedia process. Given such a view, the advantage of a semiotic approach is that it has the potential to accommodate the linguistic and non-linguistic aspects of discourse within a single, coherent framework.

3. Discourse representation Even now we have by no means covered everything that might be regarded as part of discourse studies. Nevertheless, our survey certainly shows that there are many matters to consider in designing an adequate representation for discourse. Let us, therefore, now discuss the question of representation.

3.1. Notation in FG and FDG Given that not all FG authors have adopted the same framework for the description of discourse, it is not surprising that there is no consensus, either, on the notation to be employed for representing discourse or discourse models. In the chapter on discourse in Dik (1997b) we find some hints as to notation. Dik proposes that illocutions at levels higher than the clause should be represented along the following lines: (1)

ILL(DISCOURSE-EPISODE)

He also adopts the convention of showing hierarchical structure by means of indentation. A simple example devised in the spirit of this principle is the following: (2)

ENTER DISCOURSE ENTER EPISODE 1 Sequence of speech acts LEAVE EPISODE 1 ENTER EPISODE 2 Sequence of speech acts LEAVE EPISODE 2 LEAVE DISCOURSE

The ENTER-LEAVE pairs act as brackets around the stretches of language which they enclose. In the present chapter we shall call them Delimiters.

96

John Connolly

The hierarchical structure of discourse is indicated by means of horizontal trees in Kroon (1997). In the outline example in Fig 1, A and B are interlocutors in a dialogue: A:

Utterance

act move

B:

Utterance

act

Utterance

act

exchange move

Figure 1. Example of Kroon’s style of discourse representation

Kroon also, where appropriate, labels acts and/or moves as either central or subsidiary. Hengeveld (1997) employs a means of representing discourse which reflects his idea of upward layering. However, this will not be considered in detail here, as it has been overtaken by Hengeveld’s more recent work, to which we shall shortly return. Suffice to say that the method of discourse representation in Hengeveld (1997) takes the form of a vertical tree structure. Steuten's notation takes the form not of a tree but of a table; see the fifth chapter of Steuten (1998a). A simple and slightly adapted example is seen in Table 1, which is based on the assumption that each illocutionary act in the discourse has previously been identified and numbered in sequence. In positing a non-linguistic part of the transaction in addition to the linguistic discourse, Steuten particularly has in mind business conversations, where the transaction could include, for instance, the exchange of goods and money. Gulla (1997) proposes a means of representing a combination of FG and RST with the help of a bracketed structure. The design of this representation is illustrated in the outline example in Figure 2, which is based on a stretch of discourse where two propositions occur in sequence, and proposition 2 serves to elaborate proposition 1.

The question of discourse representation Table 1.

97

An illustration of Steuten's method of discourse representation COMMUNICATIVE UNIT

Level 1: Transaction

Level 2: Exchange Opening Information

Conversation Closing

GRAMMATICAL UNIT Level 4: Illocutionary Act 1 2 3 4, 5 6 7 8

Level 3: Interactional Act Greeting Greeting Question Answer Reaction Valediction Valediction

Non-linguistic Part

RELATION

elaboration

NUCLEUS

proposition 1

SATELLITE

proposition 2

Figure 2. Outline of an RST-style discourse representation

Vet (1998a), working within the framework of Modular Functional Grammar, distinguishes between the speech act and the utterance which constitutes the product of that act. The structure of an utterance is exemplified in the following, where ‘u’ stands for ‘utterance’, ‘c’ for the ‘content’ of that utterance, ‘POS’ for the truth-value ‘positive’ and ‘p’ for ‘propositional content’: (3)

a. Jay likes Kay. b. u1: [DECL (c1: [POS p1: [Pres e1: [likeV (d1x1: JayN (x1))ZeroSubj (d1x2: KayN (x2))RefObj] (e1)] (p1)] (c1))] (u1)

Such a characterization belongs to the grammatical module. In the pragmatic module, on the other hand, Vet accommodates the discourse activity, together with the mental considerations which motivate it. For instance, take the sequence:

98 (4)

John Connolly Jay likes Kay. He dates her.

The account of this in the pragmatic module would be something like that shown in Figure 3:

WANT (S) (S inform A of the romantic association between Jay and Kay) MOVE M1 : (S) WANT ((S) INFORM (A) OF (R)) E1

E2

ASSERT (S)(m1 )(A)

ASSERT (S)(m2 )(A)

LIKE (JAY, KAY)

DATE (JAY, KAY)

[L]

[L]

m' 1

m' 2

Figure 3. Example of Vet’s style of discourse representation

This can be interpreted as follows. The desire which S has to pass on to A the information about Jay and Kay’s relationship (R) leads S to make a discourse move M1, which is split into two speech acts, E1 and E2, which are intended to transmit messages m1 and m2 respectively to A. In the case of m1 the conceptual content of the message is represented as LIKE (JAY, KAY). This is expressed in the form of actual language by means of a function denoted as L, which encapsulates the choices made by S in order to put the content of the message into words. The output of L is denoted as m'1, which represents the conceptual content of m1 together with the choices just mentioned. Finally, m'1 gives rise to u1 (whose internal details were set out earlier), which is the underlying linguistic form of the utterance that expresses the content concerned. In this way, the function L manages the interface between the pragmatic and grammatical modules. This principle is encapsulated within the condensed formula given by Vet (1998a: 12), namely L(m1) = u1. Vet’s proposals are expanded and slightly modified by Connolly (1998), who elaborates the representation of the discourse move to include the be-

The question of discourse representation

99

liefs of the speaker that lie behind the move, and also the intended outcome of the move, as illustrated in the following example, in which C stands for ‘content of ’: (5)

M1: [[BELIEVE (S) (NOT (KNOW (A) C(m1, m2)))], [WANT (S) (KNOW (A) (C(m1, m2)))],

Precondition Precondition

[CONVEY (S) (m1, m2) (A)],

Nucleus

[KNOW (A) (C(m1))] [BELIEVE (S) (KNOW (A) (C(m1, m2)))] [NOT (WANT (S) (KNOW (A) (C(m1))))]] (M1)

Postcondition Postcondition Postcondition

The postconditions list the anticipated outcomes of the move. Of course, these may or may not coincide with the actual outcomes! The above are by no means the only proposals in the FG literature which relate to the representation of discourse. However, they give a reasonable feel for the variety of notations that have been put forward, prior to the emergence of FDG. In FDG the representation of discourse structure is distributed among the three modules. Within the interpersonal module the units recognized include the move, the act and the illocution. The formulation of the interpersonal structure proceeds along the following lines, where M stands for Move, A for Discourse Act, P for Participant, ILL for Illocution, C for Communicated Content, T for Ascriptive Act and R for Referential Act: (6)

(M1: [(A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))

In dialogues the layers of Turn and Exchange are also recognized. At the representational level the propositional content is addressed. The formulation of structure follows the following pattern, where p stands for Propositional Content, e for State-of-affairs, f for Property and x for Individual: (7)

(p1: [(e1: [(f1) (x1)] (e1))] (p1))

Again, higher layers are recognized if appropriate, for instance Episode and Story within the narrative genre.

100

John Connolly

The formulation of structure at the expression level is carried out along the following lines, where Para stands for Paragraph, S for Sentence, Cl for Clause, PrP for Predicate Phrase, RP for referential Phrase and Lex for Lexeme: (8)

(Para1: [(S1: [(Cl1: [(PrP1: [(Lex1)] (PrP1)) (RP1: [(Lex2)] (RP1))] (Cl1))] (S1))] (Para1))

Once more, allowance is also made for higher layers, for example Section and Chapter in books.

3.2. Discourse Representation Structures Outside of FG there have also been various attempts to develop methods of representing discourse. Perhaps the best known of these have emerged from Discourse Representation Theory (DRT); see Kamp and Reyle (1993). Espousing from the very outset a communicative (or functional) view of language, Kamp and Reyle develop a detailed theory of discourse interpretation, based upon concepts drawn from model-theoretic semantics. In their exposition they make extensive use of a type of semantic representation called a Discourse Representation Structure (DRS). A DRS, in its most basic form, contains (i) a list of symbols, such as x or y, each of which represents a discourse referent, (ii) a list of predicates, such as Jay(x) or Kay(y), indicating the respective individuals for which the aforementioned symbols stand, and (iii) an indication of the relationship among the individuals designated by the symbols (for instance, ‘x likes y’). The relationship concerned is derived from a syntactic analysis of the sentence concerned. x y Jay(x) Kay(y) x likes y Figure 4. DRS-style representation of the sentence Jay likes Kay

By way of notation, Kamp and Reyle combine the contents of a DRS into a list enclosed inside a box, as shown in Figure 4. Discourses containing more than one sentence can, of course, be accommodated, as for

The question of discourse representation

101

example in Figure 5. Kamp and Reyle go on to elaborate this notation in various ways, which are, however, beyond the scope of the present brief survey. x y u v Jay(x) Kay(y) X likes y u=x v=y U dates v Figure 5. DRS-style representation of the discourse Jay likes Kay. He dates her.

3.3. Segmented Discourse Representation Structures Asher (1993) proposes a somewhat different version of DRT from Kamp and Reyle. His version incorporates rhetorical relations, and is based around a unit called a Segmented Discourse Representation Structure (SDRS). A slightly adapted example is given in Figure 6.

k1 :=

x, s1, y Jay(x) s1-like(x,y) Kay(y)

k2 :=

u, s2, v u=x s2-date(u,v) v=y

Cause(k1,k2)

Figure 6. SDRS-style representation of the discourse Jay likes Kay. He dates her.

As will be apparent, it has been assumed here that the content of the first sentence is the cause of that of the second.

3.4. Situation Theory A brief mention should also be made here of Situation Theory (ST), and in particular Devlin's (1991) version of ST, which is discussed in relation to FG by Connolly (1998). Although not a fully-fledged theory of discourse, Devlin's version of ST does include a means of describing the situation in

102

John Connolly

which an utterance is produced. According to Devlin, a description of this nature will include the following information: (i) who is addressing whom, when and where, (ii) what is being uttered and (iii) which individuals are being referred to by means of which lexical items; see Devlin (1991: 237). Full details of the relevant ST-style notation need not be considered here, but a rough indication of what is involved in the description of an assertive speech act is the following: (9)

> and > and >.

Clearly, this is a more elaborate characterisation of a speech act than is conventional in FG, and thus merits our attention in the present context.

4. A framework for discourse representation From what has been said so far, it is clear that the representation of discourse involves coping with a large amount of information. We are also faced with two conflicting goals. On the one hand, we would wish our representation to provide an adequate coverage of discourse phenomena. On the other hand, we would want our notation to be easy to read, as well as being straightforward and quick to produce. What is presented below is a compromise between these desiderata. The problem of representation is complicated by the fact that discourse has more than two dimensions. We need to accommodate temporal (or else spatial) sequencing, hierarchical structure, different levels of analysis and also relational phenomena (in particular adjacency sequences and rhetorical relations). All these need to be projected onto the two-dimensional page. It is fairly natural to represent temporal sequencing as progressing vertically down the page, and it is convenient to organize the levels of analysis on the horizontal dimension across the page, with each level being contained in a separate column, while maintaining a horizontal alignment between entries which correspond to one another. In fact, it is proposed that we employ a series of columns, to accommodate each of the following:

The question of discourse representation (10)

a. b. c. d.

103

The contextual level The interpersonal level The representational level The expression level

As will be immediately apparent, we are here honouring the distinction between the discourse itself and the context in which it is situated. This decision is to some extent motivated by Dik's distinction between discourse proper and discourse event. It will be recalled that the phenomena which Dik includes in the analysis of the discourse event are: who the interlocutors are, how they relate to one another, where, when and in what circumstances they are communicating, and what social conventions are in force. These considerations are all matters of context, and therefore if we are to accommodate them, we need to recognize a contextual level of description. The contextual level is not actually part of language itself, but is intended to be seen as a component of the description of discourse. It may be noted that the idea of a contextual level has echoes in other frameworks besides FDG. For example, Hasan (1999: 224), writing within the framework of Systemic Functional Grammar, proposes that in describing language we should recognize four strata: context, semantics, lexicogrammar and phonology. Moreover, Stamper (1997), in his discussion of applied semiotics, puts forward a six-level model in which the social world is recognized as a stratum which encompasses much of the context in which human communication takes place; see further Connolly (2000). In fact, the contextual level turns out to be very useful, as it provides a natural home for a number of factors that need to be accommodated within a full account of discourse. These include (where relevant): (11) a. b. c. d. e. f. g. h. i. j.

The identity of the participants, their attitudes, statuses and backgrounds; Any other sources of voices which blend into the discourse; Bystanders; The non-linguistic properties of the medium or media of communication employed; The non-verbal communication; The time and place; The setting; The prevailing cultural, social and institutional conventions governing communication; The referents of terms used in the discourse; The pre- and postconditions and outcomes of particular discourse acts.

104

John Connolly

As may be apparent, the above list strongly echoes the work of Firth (1957: 203) and Hymes (1972). Let us illustrate this with the help of the following brief excerpt from a conversational dialogue: (12)

Hugh (making move M101): Jay likes Kay. He dates her.

( = P1) ( = Message m151) ( = Message m152)

Irene (making move M102): Gosh! Ella told me that he didn’t go for tall women.

( = P2) ( = Message m153) ( = Message m154) (Ella = AV1, i.e. additional voice 1)

A possible contextual description of this interactive discourse could be the following, which contains information (that will here be assumed to be true) along the lines suggested above: (13)

P1 = Hugh P2 = Irene AV1 = Ella All have equal social status All are from the USA All have conventional American social attitudes towards romantic involvements No bystanders Participants are using normal phonation Participants are smiling Time = 09.40 p.m., 4 February 2000 Place = Irene’s apartment in Kansas City Setting is an informal dinner party Prevailing communicational conventions are those of English-speaking Americans Referents of: x251 (JayN) x252 (KayN) x253 (EllaN) x254 (womanN)

The question of discourse representation

105

Preconditions(M101) = BELIEVE (P1 NOT (KNOW (P2 C(m151)))) BELIEVE (P1 NOT (KNOW (P2 C(m152)))) WANT (P1 KNOW (P2 C(m151))) WANT (P1 KNOW (P2 C(m152))) Postconditions(M101) = KNOW (P2 C(m151)) KNOW (P2 C(m152)) BELIEVE (P1 KNOW (P2 C(m151))) BELIEVE (P1 KNOW (P2 C(m152))) NOT (WANT (P1 KNOW (P2 C(m151)))) NOT (WANT (P1 KNOW (P2 C(m152)))) Outcome(M101) = Postconditions(M101) Preconditions(M102) = BELIEVE (P2 NOT (KNOW (P1 C(m153)))) BELIEVE (P2 NOT (KNOW (P1 C(m154)))) WANT (P2 KNOW (P1 C(m153))) WANT (P2 KNOW (P1 C(m154))) Postconditions(M102) = KNOW (P1 C(m153)) KNOW (P1 C(m154)) BELIEVE (P2 KNOW (P1 C(m153))) BELIEVE (P2 KNOW (P1 C(m154))) NOT (WANT (P2 KNOW (P1 C(m153)))) NOT (WANT (P2 KNOW (P1 C(m154)))) Outcome(M102) = Postconditions(M102)

Clearly, any contextual description will have to be selective. There is no possibility of including every fact about the situation. The basic principle is to include what one feels to be relevant, insofar as one’s knowledge and the evidence permits. Thus, our example is intended merely to provide an idea of how a contextual description might look. One particular point to note is how the multimedia aspects of discourse, including non-verbal communication, have been allowed for in the contextual level. Admittedly, this probably does not go far enough to satisfy the point made by Kress and Van Leeuwen about the multimodal nature of text, but it is, hopefully, a step in the right direction, as it explicitly incorporates multimedia phenomena into the FDG-related framework.

106

John Connolly

The description of the discourse proper begins at the interpersonal level. Internally the interpersonal level has a hierarchical structure, which can be represented with the help of indentation, somewhat after Dik (1997b: 423). The units in the hierarchy will be taken to be those of Sinclair and Coulthard, given that many of these have been adopted somewhere in the FG literature, and some of them are already recognized in FDG. By way of illustration, an interpersonal level description of the same dialogue is the following: (14) ENTER DISCOURSE INTERACTION D1 Genre = conversation Style = informal, unremarkably polite, concise Ideology = neutral Discourse topic = miscellaneous idle chatter Discourse tense = present ... ENTER TRANSACTION Tr5 Topic of transaction = gossip about mutual acquaintances ILL(Tr5) = representative ... ENTER EXCHANGE Ex18 Topic of exchange = Jay and Kay’s romantic association ENTER MOVE M101 PSp = P1, PAddr = P2 M101: [CONVEY (PSp) (m151, m152) (PAddr)] (M101) ENTER ACT A151 A151: [ASSERT (PSp) (m151) (PAddr)] (A151) ILLS(A151) = ILLA(A151) = representative C(m151) = REFER(x251, x252) ASCRIBE(f201) AS PER p171 LEAVE ACT A151 ENTER ACT A152 A152: [ASSERT (PSp) (m152) (PAddr)] (A152) ILLS(A152) = ILLA(A152) = representative C(m152) = REFER(x251, x252) ASCRIBE(f202) AS PER p172 A152 = Elaboration(A151) LEAVE ACT A152 LEAVE MOVE M101

The question of discourse representation

107

ENTER MOVE M102 PSp = P2, PAddr = P1 M102: [CONVEY (PSp) (m153, m154) (PAddr)] (M102) ENTER ACT A153 A153: [EXCLAIM (PSp) (m153) (PAddr)] (A153) ILLS(A153) = ILLA(A153) = expressive Attitude(A153) = surprise A153 = Acknowledgement(A151, A152) LEAVE ACT A153 ENTER ACT A154 A154: [ASSERT (PSp) (m154) (PAddr)] (A154) CONTAIN(A154, BLEND(PAddr, AV1)) ILLS(A154) = ILLA(A154) = representative C(m154) = REFER(x253, P2, x251, x254) ASCRIBE (f203) AS PER p173 ASCRIBE (f204, f205) AS PER p174 A154 = Elaboration(A153) LEAVE ACT A154 LEAVE MOVE M102 LEAVE EXCHANGE Ex18 LEAVE TRANSACTION Tr5 ... LEAVE DISCOURSE INTERACTION D1

The description is set out in a manner derived through combining the practice of several different FG authors, but organized in terms of the Sinclair and Coulthard categories. Following the opening delimiter (‘ENTER DISCOURSE INTERACTION D1’), we have a specification of various overall attributes of the discourse, such as the genre, the style, the ideology and the overall topic. For example, the style is informal (witness the inclusion of the interjection gosh), unremarkably polite (in that it is courteous but in no way obsequious) and concise (being composed of short sentences), while it is ideologically neutral in that it does not contain any overt marker of, for instance, feminism, capitalism, and so on. Next comes the sequence of transactions, of which our example happens to show the fifth, i.e. Tr5. The details of the latter are placed between its opening delimiter (‘ENTER TRANSACTION Tr5’) and its closing delimiter (‘LEAVE TRANSACTION Tr5’); we shall return to these particulars in a moment. At the end of the sequence of transactions comes the final closing delimiter

108

John Connolly

(‘LEAVE DISCOURSE INTERACTION D1’), which terminates the discourse. The opening delimiter of the transaction in our example is followed by a statement of the topic of the transaction concerned. After that we have the sequence of exchanges, of which our example shows the eighteenth, i.e. Ex18, which is enclosed between the delimiters ‘ENTER EXCHANGE Ex18’ and ‘LEAVE EXCHANGE Ex18’. The opening delimiter is followed by a statement of the topic of the exchange in question, after which are set out the series of moves of which the exchange consists. As the illustrative discourse is a dialogue, it is necessary to begin the description of each move with a specification of who is the speaker (PSp in Hengeveld’s notation) and who is the addressee (PAddr). For instance, in the case of move M101, the speaker is shown to be participant P1 and the addressee the participant P2. Next comes a statement of the essence of this move, which is for the speaker to convey to the addressee a series of two messages, i.e. m151 and m152. This is accomplished by means of acts A151 and A152. (An act is abbreviated A in Hengeveld’s notation.) The description of an act requires at least three lines. A simple example is found in act A151. Firstly, we need a specification of the essence of the act, which in this case is that the speaker asserts a particular message (m151) to the addressee. Secondly, we need a statement of the illocutionary force of the act, for which purpose in the present paper we employ Searle’s (1976) categories. In the current example, the illocutionary force of the act from the point of view of both the speaker and the addressee is ‘representative’. Thirdly, the discourse processes of reference and ascription are acknowledged. The line ‘REFER(x251, x252)’ states that the individuals x251, x252 (which are primarily constituents of the representational level, and given lexical shape as JayN and KayN at the expression level) are referring to certain phenomena in the context. These referents are enumerated in the contextual description. Thus, the interactional level serves as a bridge between language and context, and hence fulfils the function of the pointers mentioned in Connolly (1998: 184). The line ‘ASCRIBE (f201) AS PER p171’ serves to assign the property f201 (another constituent of the representational level, given lexical shape as likeV at the expression level) to the individuals x251 and x252, in the manner indicated in the representational level proposition p171. This proposition is shown in outline form in (15a), with the corresponding expression-level structure (up as far as the clausal layer, and in outline form) being given in (15b), while the outline propositions and expression-level structures for the second, third and fourth acts in the exchange are shown in (16–18). Formu-

The question of discourse representation

109

lae such as ‘AS PER p171’ are employed in order to avoid having to duplicate, at the interpersonal level, information that needs in any case to be stated at the representational level. (15)

a. (p171: [(e171: [(f201) (x251) (x252)] (e171))] (p171)) b. (Cl171: [(PrP231: [(likeV)] (PrP231)) (RP281: [(JayN)] (RP281)) (RP282: [(KayN)] (RP282))] (Cl171))

(16)

a. (p172: [(e172: [(f202) (x251) (x252)] (e172))] (p172)) b. (Cl172: [(PrP232: [(dateV)] (PrP232) (RP283: [(hePron)] (RP283)) (RP284: [(herPron)] (RP284))] (Cl172))

(17)

a. NO PROPOSITIONAL CONTENT b. Gosh

(18)

a. (p173: [(e173: [(f203) (x253) (p174: [(e174: [NOT(f204) (x251) (x254: (f205) (x254))] (e174))] (p174)) (P2)] (e173))] (p173)) b. (Cl173: [(PrP233: [(tellV)] (PrP233)) (RP285: [(EllaN)] (RP285)) (Cl174: [(PrP234: [(not goV for)] (PrP234)) (RP286: [(hePron)] (RP286)) (RP287: [(womanN: tallA)] (RP287))] (Cl174)) (RP288: [(mePron)] (RP288))] (Cl173))

Other information can also be included within the description of an act. First of all, rhetorical relations can be specified, as for instance in act A152, which is described as an elaboration of the preceding act, by means of the line ‘A152 = Elaboration(A151)’. Relationships within adjacency sequences can also be specified. For example, within the description of act A153, the final line, ‘A153 = Acknowledgement(A151, A152)’, shows the act in question to be an acknowledgement of the acts comprising the previous move. In addition, the adoption of a particular attitude can be indicated. This, too, is exemplified in act A153 (corresponding, as it does, to the interjection Gosh), which is shown to be marked by surprise, through the inclusion of the line ‘Attitude(A153) = surprise’. There are quite a number of points to be made in relation to the form of discourse representation just presented. Let us begin with some matters of notation. Firstly, it is assumed that every type of unit in the hierarchy is numbered in sequence, starting from the opening of the discourse. Hence, for instance, Move M1 would be the first move in the entire interaction, and not just within the transaction of which it forms part. Secondly, no use has been made of boxes, in contrast to the DRT notation. This is in order to make the representation quicker to produce. Thirdly, the notation as exem-

110

John Connolly

plified above is not intended to be overly prescriptive. The discourse analyst is expected to include what is relevant to his or her purposes. The notation as exemplified above is intended to be fairly explicit, and in some circumstances it would doubtless be possible to omit some of it as being obvious, although this is less likely if the notation is being used in the context of Computational FG. In any case, though, it is quite acceptable to omit information that is not relevant to the task in hand, or alternatively to add further information if necessary. For instance, if it desired to label acts as ‘central’ or ‘subsidiary’, after the fashion of Kroon (1997), then this can easily be done (by means of lines like ‘ENTER MAJOR ACT301’ or ‘ENTER SUBSIDIARY ACT302’, or by some suitable alternative). Another point of clarification concerns the units within the interaction. It is assumed that each member of the hierarchy (apart from the act) consists of a sequence of one or more units of the rank below. Thus, for instance, a move can be composed of just a single act. Let us now consider the coverage of the interpersonal level description. The latter is constructed around the discourse interaction, revealing the hierarchical structure of the latter, but it also accommodates statements about attitudes, for example ‘Attitude(A153) = surprise’. It incorporates broadscope parameters such as discourse topic, discourse tense and large-scale illocutionary force, as well as recognising ILLA, ILLS and their values. In addition, it accommodates both adjacency sequences (for instance ‘A152 = Elaboration(A151)’) and rhetorical relations (for example ‘A152 = Elaboration(A151)’). It should be noted that the overall framework accommodates certain discourse-related concepts that are not strongly associated with the FG tradition. These include ideology (alluded to in the above example by the statement ‘ideology = neutral’), as well as voice in Bakhtin’s sense of the term, reflected in the statements ‘AV1 = Ella’ in the contextual description and ‘CONTAIN(A154, BLEND(PSp, AV1))’ in the interpersonal level representation. Following the example of authors in, for example, DRT and Situation Theory, the referring elements are picked out (for instance by means of the statement ‘REFER(x251, x252)’) in the interpersonal level description, which have corresponding entries at the contextual level (in this case ‘Referents of: x251 (JayN), x252 (KayN)’). Note that the words in brackets here are intended merely as glosses for predicates which would normally be incorporated, instead, into the representational level and expression level description. All the other information in Devlin’s characterisation of a speech act, as illustrated earlier, also finds a place in one or other of the levels of description envisaged here.

The question of discourse representation

111

How do the present proposals compare with those of Hengeveld (this volume)? On the one hand, Hengeveld’s treatment of the interpersonal level of discourse is more concise, and will no doubt suffice for many circumstances in which discourse description is not the primary topic of interest. On the other hand, the present proposals are more appropriate when a detailed treatment of discourse is required, particularly since they include phenomena such as adjacency relations, which are omitted from Hengeveld’s (admittedly non-exhaustive) presentation. There is another significant difference between the two sets of proposals, which concerns the recognition or otherwise of structures such as paragraphs above the sentence layer. The stance tentatively taken here is that paragraphs can be seen as the expression-level counterpart of moves. Similarly, chapters and sections can be seen as the expression-level footprints of interpersonal-level units; thus a chapter expresses a transaction and a section expresses what we may term a sub-transaction. In this way, the hierarchical structure of discourse can be simply attributed to the interpersonal level, in which case its expression at lower levels is purely derivative, and therefore does not really need to figure in the lower levels of the linguistic description, where it adds nothing fresh. This stance accords with Verschueren's (1999: 130) view that there are no purely structural units above the sentence. It also simplifies the description, in dispensing with what Dik calls the organisational level. In these respects, then, it seems a preferable alternative to other approaches. However, to repeat, this proposal is tentative and will need to be examined more fully on another occasion. If it should prove to be untenable, then there would be no inherent difficulty in incorporating the higher-ranking structures into the expression level description. Finally, we need to consider how the four levels of description all fit together. Ideally, it would be satisfying to set the discourse description out in four parallel columns. However, limitations of space may well prevent that, in which case we may set out the contextual and interpersonal levels alongside each other as in Table 2, and then deal with the other two levels as in Table 3. The first column of Table 3 allows for a cross-reference to each act, which is dealt with in more detail in Table 2. In the second column of Table 3 we place the representational-level description of the propositional content of the act concerned, and in the third column its expression-level description.

112

John Connolly

Table 2. Illustrative excerpt from a multi-level discourse description: contextual and interpersonal levels

Contextual Description

Interpersonal Description

Preconditions(M101) = BELIEVE (P1 NOT (KNOW (P2 C(m151)))) BELIEVE (P1 NOT (KNOW (P2 C(m152)))) WANT (P1 KNOW (P2 C(m151))) WANT (P1 KNOW (P2 C(m152))) ENTER MOVE M101 PSp = P1, PAddr = P2 M101: [CONVEY (PSp) (m151, m152) (PAddr)] (M101) ENTER ACT A151 A151: [ASSERT (PSp) (m151) (PAddr)] (A151) ILLS(A151) = ILLA(A151) = representative ILLS(A151) = ILLA(A151) = representative C(m151) = REFER(x251, x252) ASCRIBE(f201) AS PER p171 LEAVE ACT A151 ENTER ACT A152 A152: [ASSERT (PSp) (m152) (PAddr)] (A152) ILLS(A152) = ILLA(A152) = representative C(m152) = REFER(x251, x252) ASCRIBE(f202) AS PER p172 A152 = Elaboration(A151) LEAVE ACT A152 LEAVE MOVE M101 Postconditions(M101) = KNOW (P2 C(m151)) KNOW (P2 C(m152)) BELIEVE (P1 KNOW (P2 C(m151))) BELIEVE( P1 KNOW (P2 C(m152))) NOT (WANT (P1 KNOW (P2 C(m151)))) NOT (WANT (P1 KNOW (P2 C(m152)))) Outcome(M101) = Postconditions(M101)

The question of discourse representation

113

Table 3. Illustrative excerpt from a multi-level discourse description: representational and expression levels

Interpersonal Act A151

Representational Description p171 e171 f201 x251 x252

Expression Description Cl171 Prp231 likeV RP281 JayN RP282 KayN

The above is offered as a feasible way of representing discourse for the purposes of FDG. As it stands, for the sake of readability it does not employ a fully formal notation, though if necessary, for instance if it were to be incorporated into a computing system, a properly formal version could be developed from it.

5. Conclusion In this chapter a framework for the representation of discourse has been proposed which is intended to accommodate the main phenomena of interest to analysts and to be compatible with FDG. An important feature of the representation is that it treats context as a separate level of description. Undoubtedly the framework stands in need of further development, but it does hopefully serve to indicate how the various and numerous phenomena of discourse can be seen to fit together into a coherent whole.

References Asher, Nicholas 1993 Reference to Abstract Objects in Discourse. Dordrecht: Kluwer. Bolkestein, A. Machtelt 1998 What to do with topic and focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. Clark, Katerina and Michael Holquist 1984 Mikhail Bakhtin. Cambridge MA: Harvard University Press.

114

John Connolly

Connolly, John H. 1998 Information, situation semantics and Functional Grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 167–189. 2000 Accommodating natural language within the organisational semiotic framework. Proceedings of the Third International Workshop in Organisational Semiotics, 1–13. Stafford: University of Staffordshire. Connolly, John H., Anthony A. Clarke, Steven W. Garner and Hilary K. Palmén 1997a A functionally oriented analysis of spoken dialogue between individuals linked by a computer network. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 33–58. Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds) 1997b Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter. Coulthard, Michael 1977 An Introduction to Discourse Analysis. London: Longman. Devlin, Keith 1991 Logic and Information. Cambridge: Cambridge University Press. Dik, Simon C. 1997a The Theory of Functional Grammar. Part I: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part II: Complex and Derived Structures. Berlin and New York: Mouton de Gruyter. Fairclough, Norman 1989 Language and Power. London: Longman. Fairclough, Norman and Ruth Wodak 1997 Critical discourse analysis. In: Teun van Dijk (ed.), Discourse as Social Interaction. London: Sage, 258–284. Firth, John R. 1957 Papers in Linguistics 1934–1951. London: Oxford University Press. Gulla, Jon Atle 1997 Combining Functional Grammar and Rhetorical Structure Theory for discourse representation. In: Connolly et al. (1997b), 75–89. Hancher, Michael 1979 The classification of co-operative illocutionary acts. Language in Society 8: 1–14. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins.

The question of discourse representation

115

Hasan, Ruqaiya 1999 Speaking with reference to context. In: Mohsen Ghadessy (ed.), Text and Context in Functional Linguistics, 219–321. Amsterdam and Philadelphia: Benjamins. Hengeveld, Kees 1997 Cohesion in Functional Grammar. In: Connolly et al. (1997b), 1–16. This vol. The architecture of a Functional Discourse Grammar, 1–21. Hymes, Dell 1972 Models of the interaction of language and social life. In: John J. Gumperz and Dell Hymes (eds), Directions in Sociolinguistics: The Ethnography of Communication, 35–17. New York: Holt, Rinehart and Winston. Kamp, Hans and Uwe Reyle 1993 From Logic to Discourse: Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and Discourse Representation Theory. Dordrecht: Kluwer. Kress, Gunther and Theo van Leeuwen 1996 Reading Images: The Grammar of Visual Design. London: Routledge. Kroon, Caroline 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 17–32. Liedtke, Frank 1998 Illocution and grammar: A double level approach. In: Mike Hannay and A. Machtelt Bolkestein (eds), 107–127. Mann, William C. and Thompson, Sandra A. 1987 Rhetorical structure theory: A framework for the analysis of texts. Papers in Pragmatics 1: 79–105. Martin, Jim R. 1992 English Text: System and Structure. Amsterdam and Philadelphia: Benjamins. Searle, John 1976 The classification of illocutionary acts. Language in Society 5: 1–24. Stamper, Ronald 1997 Organisational semiotics. In: John Mingers and Frank Stowell (eds), Information Systems: An Emerging Discipline?, 267–283. London: McGraw-Hill. Steuten, Ans A. G. 1998a A Contribution to the Linguistic Analysis of Business Conversations within the Language/Action Perspective. Delft: Steuten.

116

John Connolly

1998b

Structure and coherence in business conversations. In: Mike Hannay and A. Machtelt Bolkestein (eds), 59–75. Van den Berg, Marinus 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–103. Verschueren, Jef 1999 Understanding Pragmatics. London: Arnold. Vet, Co 1998a The multilayered structure of the utterance: about illocution, modality and discourse moves. In: Mike Hannay and A. Machtelt Bolkestein (eds), 1–23. 1998b Epistemic possibility in the layered structure of the utterance. In: Mike Hannay and A. Machtelt Bolkestein (eds), 149–165.

Focus of attention in discourse Francis Cornish

1. Introduction My goal in this chapter is to compare two functionalist theories of indexical reference in discourse in terms of topichood, seeking to critically evaluate each both in its own terms and in terms one of the other.1 Ultimately, I aim to show where each theory may usefully illuminate the other, resulting in a more adequate overall account of this complex and still not fully understood area of language use. I will begin by characterizing in Section 2 the North-American Columbia School of Linguistics’ account of what it calls the Focus and Deixis systems of attention concentration in discourse, and will then re-examine in Section 3 the four Topic subtypes recognized in standard FG (Dik 1997a: Ch. 13; also Hannay 1985a, 1985b, 1991; Bolkestein 1998, 2000; Mackenzie and Keizer 1991; Siewierska 1991: Ch. 6; Gómez-González 2001) in the light of the earlier characterization of these systems. Finally, Section 4 briefly discusses how the conception of topic (and focus) assignment which emerges from this comparison might be integrated into the new Functional Discourse Grammar (FDG) model put forward by Kees Hengeveld (this volume). In terms of this new Functional Discourse Grammar framework, the assignments of Topic and Focus functions which I shall be discussing clearly fall within the Interpersonal Level recognized within this format – the nature of these pragmatic functions arguing strongly in favour of the topdown relationship between the three descending levels of (respectively) Interpersonal, Representational and Expression, which structure the new model. A brief word, first of all, on the Columbia School (CS), which is relatively little known as such in Europe (even though the work of several CS

118

Francis Cornish

linguists is individually quite well known: the work of Robert Kirsner on the Dutch demonstratives, of Erica García on the Spanish clitic pronoun system, or of Ricardo Otheguy on the Spanish articles, to take but three examples). The Columbia School, founded by the late William Diver, is a functionalist sign-based theory which rests on the twin pillars of Saussurean structuralism and the primacy of the communicative function of language. It is an avowedly inductive theory, suspicious of a priori categories, notions and constructs. All such theoretical devices must be demonstrated to be necessary on the basis of evidence of patterning in actual texts (there are shades of FG here – though CS adheres much more strictly to this principle). CS linguists seek out the existence of microsystems such as those of tense, aspect, mood, number, case, the demonstratives, and so on, in terms of Saussurean constellations of values, or socalled signal-meaning hypotheses: the existence of systematic oppositions of forms correlates with systematic oppositions in meaning, or sense (signified in Saussurean terms) within the micro-system. From the user’s point of view, an actual message is fleshed out on the basis of a given set of values expressed via certain signals belonging to the systems at issue, in conjunction with relevant features of the context (cotext and context-of-utterance – including the speaker/writer’s hypothesized intentions in producing the utterance in question). That is, as in Sperber and Wilson’s Relevance Theory of understanding in context, as also in standard FG and the new FDG model, the formal structure of an utterance (CS does not recognize the traditional category sentence) always underdetermines the message(s) which it can serve to convey, given particular contexts. The gap between (structural) meaning (or signified), based on given networks of formal and semantic oppositions, and message, has to be bridged via inferences on the part of the language user. CS linguists operate very much at the frontier between clausal structure and textual product, in the sense that they make heavy use of text counts of given features and their combinations, leading to statistically significant tendencies or skewings across texts from different genres. These skewings, once apparent, are then used as empirical validations of the hypotheses set up to describe and account for the behaviour of given forms in context. The introduction by Ellen Contini-Morava to Contini-Morava and Goldberg (1995) gives a very good overview of CS theory.

Focus of attention in discourse

119

2. The systems of Focus and Deixis in CS

2.1. Focus The CS conception of Focus should not be confused with the sense of Focus in FG terms in the context of so-called Information Structure (henceforth IS). See Gundel (1999) and Cornish (1999: §5.3.2, 2000a: §2.2) for clarification of the relationship between these two senses of ‘Focus’ in the literature. The reader will need to be alert to this in what follows, and to keep his or her eye firmly on the terminological ball. To keep the two intended senses of ‘focus’ distinct, I will capitalize this term throughout when referring to the CS conception, and will use only initial capitalization for the FG one. FOCUS in CS terminology does not mean ‘the most important information2 conveyed by a clause, representing the difference between P(A)S (the speaker’s representation of the addressee’s current state of pragmatic knowledge) and (PS) (the speaker’s own current state of pragmatic knowledge)’, as in the standard FG account, but rather ‘what the speaker is assuming the hearer is already concentrating on, what s/he has at the forefront of his/her consciousness at the time of utterance’. This is therefore more akin to the FG notion of Topic than of Focus. Only participants (arguments) with respect to the Event represented in a clause fall within the CS FOCUS systems. The CS value NOT IN-FOCUS is then (confusingly) equivalent by default to the information-structural sense of Focus – although there is no prediction within CS theory that constituents bearing this value will in fact have Focus status (in the Information Structure sense). Wallis Reid (pers. comm.) confirms my understanding that INFOCUS status is in fact equivalent to that of Topic in the IS sense, and that NOT IN-FOCUS equates with non-topical. He goes on to point out that the value NOT IN-FOCUS only applies to participants that could have been IN-FOCUS (i.e. clausal topics). The instruction conveyed to the understander via a constituent bearing the value [IN-FOCUS] (the case of the term Jane in (1a) below, under its topic-comment interpretation) is ‘pay more attention to the referent at issue, as it will be important in the subsequent discourse’, and via one carrying the value [NOT IN-FOCUS] (a tarantula in (1a), and, presumably, the entire sentence in (1b)), ‘pay somewhat less attention to the referent involved’.

120

Francis Cornish

(1)

a.

Jane [IN-FOCUS] has just spotted a tarantula [NOT IN-FOCUS] (categorical utterance: ‘predicate-focus’ in Lambrecht’s 1994 terminology). b. A bomb exploded yesterday morning in Armington Valley high street. (thetic utterance: ‘sentence focus’ pace Lambrecht 1994)

I am assuming here that in the immediate context of (1a), the discourse has been dealing with Jane, that is, that this referent is topical in Dik's (1997a) sense of the term. Thus, (1a) should be understood as a reply to the question What did Jane do?, or What about Jane?, where the IS Focus would correspond to the content of the VP has just spotted a tarantula. Alternatively, (1a) could be a response to the question What has Jane just seen/spotted?, where the IS Focus would correspond to the second argument a tarantula alone (‘Jane has just seen/spotted something’ being presupposed, following Erteschik-Shir 1997: 13). (1a) may be analysed as an instance of the classic ‘topic-comment’ (categorical) information structure, since a relation of ‘aboutness’ may easily be attributed to the referent of the subject expression and to the denotation of the predicator-cum-second argument (given the type of context indicated above), as expressed in (2a): (2)

a.

As for/About Jane, she has just spotted a tarantula.

b.

#As for/About a tarantula, it has just been spotted by Jane.

(2b), which is very unnatural as a paraphrase of the relevant part of (1a), clearly shows that the second argument a tarantula does not bear the topic (i.e. [IN-FOCUS]) relation with regard to the remainder of the utterance.3 Two well-known problems with assessing written, decontextualized sentences of the type in (1a) are first, precisely the lack of a context which would enable the assignment of a given information-structural analysis on the basis of that context; and second, the lack of an explicit prosodic structure which would motivate that assignment. This latter point makes itself felt particularly acutely in English. (1b) as a thetic utterance does not presuppose any particular prior context; as such, it could occur discourse-initially. CS would not assign INFOCUS status to the subject term a bomb here, since the predication is intransitive. As such, it would be assigned the value ‘MORE FOCUS (needed)’ (see below), in comparison to inverted constructions where the subject term is postverbal. The fact that the value ‘MORE FOCUS (needed)’ is assigned to this referent correctly predicts that it is not (yet) topical: for here there is no ‘aboutness’ relation involved between its refer-

Focus of attention in discourse

121

ent and the proposition; the entire utterance is thetic (presented as ‘all-new’ information to the addressee) – an event-reporting utterance in Lambrecht’s (1994) terms.4 That the argument expression a bomb, in subject position, does not bear the topic, or an [IN-FOCUS] relation with respect to the remainder of the utterance, is shown by (3): (3)

#As for/About a bomb, it exploded yesterday morning in Armington Valley high street.

However, were this utterance to correspond to a transitive sentence (as in A bomb destroyed the Armington Valley city centre yesterday morning), the value IN-FOCUS would then be assigned to the referent of the subject term, thereby incorrectly predicting it to be the topic of that utterance. Now, there is an ambivalence in the conception of FOCUS within the CS framework: on the one hand, an item signalled as IN-FOCUS represents information which the speaker is assuming the addressee is already concentrating on; and on the other, the addressee is being tacitly instructed to maintain the concentration differential between this IN-FOCUS item and other NOT IN-FOCUS items, for the immediately ongoing discourse. The following quotation from Huffman (forthc.: 31) reflects this ambivalence: The system of Focus deals with the centering of attention on one of the participants in an event. The opposition here is a binary one: a participant is either IN FOCUS or NOT IN FOCUS. A participant IN FOCUS is thereby stated to be the center of attention with respect to the particular event, and when a participant is signalled to be NOT IN FOCUS, the speaker is asserting that this one should not have the hearer’s attention focused on it. (Huffman forthc.: 31 [emphasis mine])

Although for convenience, illustrations of the FOCUS system are often given in the context of a single clause, CS linguists point out that it can only really be detected within a complete text, since it is the relation of participants to the wider discourse which the participant FOCUS values instantiate.5 The Columbia School system of FOCUS has two basic values, then, restricted to encoded differential attention contrasts solely with respect to the arguments of finite verbs:6 for canonical transitive predications (in FG terminology, A1 subject - predicator - A2 object), IN-FOCUS for the A1 subject, and NOT IN-FOCUS for the A2 object (Reid 1991:178). The latter value (NOT IN-FOCUS) would also be attributed to a third participant (i.e. an argument which is nuclear, that is, not preceded by a preposition), as in

122

Francis Cornish

John bought Mary a book. The predicator is not involved in the CS FOCUS system, only argument expressions (i.e. ones denoting participants) are. As suggested above, this distinction would appear to hold only in the case of so-called categorical utterances, where the utterance as a whole may be construed as conveying information about the entity denoted by the topic (here subject) expression – although CS linguists do not appear to draw the distinction between categorical and thetic utterances.7 All entities coded as non-participants (i.e. non-nuclear arguments, as well as satellites, in FG terminology) – e.g. via a prepositional phrase – are ipso facto presented as not entering the FOCUS system at all. In languages possessing case-marked clitic pronouns, nominative clitics signal HIGHER-FOCUS, while nonnominative clitics signal LOWER-FOCUS. As already mentioned, where there is only one participant, in an intransitive clause, for example, the two values are said to be MORE FOCUS for the participant in preverbal subject position, and LESS FOCUS when it is postposed. The contrast here is then between two construction types, not one between participants (arguments) within the same predication. That is, these word order signals form an oppositional micro-system; see in particular Huffman (1993) and Reid (1991: 180–181, ex. (29)), for justification and practical illustration based on attested texts. (4a,b) from Reid (1991) illustrate: (4)

a. A fly (MORE FOCUS) was swimming in my soup. (Reid 1991: ex. (26)) b. There was a fly (LESS FOCUS) in my soup. (Reid 1991: ex. (27))

Note that both (4a) and (4b) may be analysed as thetic utterances, not as categorical ones. Reid’s annotation here goes some way towards implying this in (4a), since the assignment of ‘MORE FOCUS (needed)’ (as in the case of the term the bomb in example (1b) above) entails that the referent at issue is not already IN-FOCUS (i.e. topical). As such, neither occurrence of the NP a fly in (4a) and (4b) would be construed as a topic expression, both of them necessarily carrying a high level of pitch-accent.8 Now, while the analysis ‘LESS FOCUS (needed)’ may be plausible as an interpretation of a fly in (4b), that of ‘MORE FOCUS (needed)’ is not evidently so in the case of the token functioning as subject in (4a), since the thetic interpretation of the containing utterance here treats the referent of this expression, not as an entity having a separate existence, but as an integral part of the situation conveyed as a whole (see for example Siewierska 1991: 161–162, who says of her comparable example (18b) A tiger chased a tourist: “...in

Focus of attention in discourse

123

using (18b) the primary purpose of the speaker is not to present the referents but to establish the event”; see also on this point Hannay 1991: 146). In fact, (4b) involves not just inversion, but also so-called thereinsertion,9 a specifically ‘existential’ construction. In this connection, Birner and Ward (1998: 102–106) argue that there is a discourse-pragmatic constraint affecting the postverbal NP in this construction, to the effect that its referent must not be ‘hearer-old’ (i.e. assumed to be already familiar/known to the hearer); no such constraint is claimed to regulate the canonical word-order variant, as seen in (4a), however. This restriction may be seen to stem from what Hannay (1985a: 101) argues is the raison d'être of presentative constructions of this kind, namely to “assert the existence of an entity or state of affairs in the discourse world”. Siewierska (1991: 161–162) distinguishes between the two construction types also within an FG framework in suggesting that sentences of type (4a) serve to report a bare event, while existential ones, as in (4b), function to introduce or present and highlight a specific entity (or alternatively, a State of Affairs as a whole). Huffman (1993) presents the three construction types as a three-way decreasing scale of FOCUS marking: subject-verb order conferring a high level of attention focus on the referent of the subject term, that of verb-subject a lower or mid level, and the there + subject + extension (in Hannay's 1985a analysis), the lowest level of the three. (5b) is a naturally-occurring example, contrasted with (5a), which is adapted from it. The immediate left-hand and right-hand co-text of (5b) is given under (6a) and (6b), respectively. Moreover, the immediate macrotopic of the segment in which (6a), (5b) and (6b) occur is ‘the city of Antibes’. (5)

(6)

a. ...the Picasso museum in the old Grimaldi château is of greater interest. b. ...of greater interest is the Picasso museum in the old Grimaldi château.(Extract from The Holiday Which? Guide to France by A.Ruck 1982: 112) a. ...In the old town [of Antibes] the cathedral, like so many others, has an ill-lit Bréa; (4b). b. ...(6a) + (5b) The rest of Antibes is a modern and bustling resort...

In comparing (5a) and (5b) as occurring in the left- and right-hand contexts specified in (6a) and (6b), respectively, it would seem that for the subjectpredicate ordering as in (5a) to be fully coherent and natural in this context, the adversative connective but would be expected preceding it. The force of this connective is in fact implicit in the original predicate-subject order-

124

Francis Cornish

ing as in (5b), the attested version, but the surface presence of this item is not essential in such a context. I would suggest that the reason for the choice of predicate-subject ordering in (5b) is entirely due to the need to relate (5b) to (6b), its immediate context, via a relation of comparative contrast. The very fact of mentioning a feature of a town in a travel guide such as this (here the fact that Antibes cathedral has an “ill-lit Bréa”) implies that it is potentially ‘of interest’ to the prospective visitor. This is then the topic of utterance (5b) (‘features of potential interest to the visitor to Antibes’), and is the reason why the prepositional phrase is preposed in this context, connecting directly in this way with its immediate discourse context. In FG terms, it would be placed in P1 position – an intra-clausal position reserved (among other features) for constituents bearing special pragmatic values. This, then, is the common thematic feature holding between the two sentences, in terms of which the propositions they express are connected. See Hannay (1991a: 145) for a presentation and analysis of a very similar example (his (19a)). Hannay notes that the preposed adjectival phrase Particularly interesting... in his example is “topical inferrable information” (on the basis of the immediate context which he provides), and that it is a “staging device” rather than a Topic per se. As with CS theory, so with FG only nominal expressions (terms) may be assigned one of the Topic functions. Now, one may agree with the CS analysis here that the referent of the term the Picasso museum in the old Grimaldi château in (5a) is the (new) topic of this utterance (‘As for the Picasso museum, this/it is of greater interest’), and therefore bears the value ‘[MORE FOCUS (needed)]’, since it is the subject of an intransitive predication; and that the predicational prepositional phrase of greater interest is non-topical, not participating in the FOCUS system (since it does not correspond to a participant – i.e. argument, in FG terms). But surely this degree of [FOCUS] (i.e. concentration of attention) on the preverbal term in (5a) is not greater than that which the inversion construction illustrated in (5b) manifests? Of course, the difficulty here is that we are in grave danger of confusing the CS sense of Focus (and what it is applied to) and the information-structural one. CS would say that more attention-focus is concentrated on the preverbal term in the non-inverted construction in (5a) than on the postverbal one in (5b). At the same time, it is apparent that in their spoken realizations, the comparative adjective greater in (5a) (being part of the informationstructure focus) would have a more prominent pitch-accent on it than it would in (5b). The placing of the subject term the Picasso museum in the old Grimaldi château in postverbal predicate position in (5b) seems indeed

Focus of attention in discourse

125

to “direct the attention of the hearer(s) to its intension, by being uttered” (Erteschik-Shir 1997: 11), which is how this linguist defines the IS sense of Focus; but this is not the case in (5a), where it is in preverbal, canonical subject position.10 Erteschik-Shir’s (1997: 14) ‘lie-test’ confirms this difference. The lie-test consists in contradicting various parts of an utterance, and seeing whether the connection between the two utterances is coherent. It is essentially a test of what is asserted in the utterance at issue (Erteschik-Shir 1997: 15), and so reflects the potential IS Focus domain(s). (7)

a.

A: ....of greater interest is the Picasso museum in the old Grimaldi château. b. B: That’s not true! The ill-lit Bréa is much more interesting...(adapted from the context of (4b) in The Holiday Which? Guide to France, 1982: 112 )

Compare the naturalness of (7b) as a contradiction of the assertion in (7a), with the more bizarre character of (8b) as a contradiction of the information conveyed by the preverbal PP in (7a) (here (8a)): (8)

a.

A:....of greater interest is the Picasso museum in the old Grimaldi château. b. B: #That’s not true! Of no interest at all is that museum.

Clearly then, by the ‘lie-test’ the most prominent information in (5b) is that associated with the postverbal term, not the preverbal one. And the inversion construction seems equally clearly to place even greater informational prominence on the postverbal term than is the case in the unmarked topic-comment construction illustrated in (5a). The CS claim, then, that inversion constructions such as (4b) and (5b) involve the signalling of less attention on the postverbal referent than on the preverbal one in non-inverted constructions seems to be valid only under the epistemic interpretation of the CS FOCUS systems – where the addressee’s attention is simply assumed not already to be concentrated on the referent at issue. Otherwise there would appear to be a clear conflict in discourse values: for CS theory, the postverbal subject acts as an instruction to the addressee to pay LESS attention to the referent of this term (than if it had occurred in preverbal subject position); and for IS linguists (Erteschik-Shir, Lambrecht, and others – including Dik and other FG linguists), the referent of this term is focal information, to which the addressee’s attention is specifically being directed.11

126

Francis Cornish

In this regard, Birner and Ward (1998: 156–194) argue that the function of the preverbal constituent in inversion constructions is to link the utterance to the prior discourse context (see the PP of greater interest in the attested example (5b)), and that, for the inversion to be felicitous, the postverbal constituent must not represent information which is more ‘discourse-old’ than that of the preverbal one. This is clearly the case with the term the Picasso museum in the old Grimaldi château in (5b). Moreover, the verb in inversion constructions is a ‘light’ one: intransitive, copular, or denoting existence or emergence and so forth. For Huffman (1993), verb-subject12 ordering (where it is possible) is a means of unobtrusively introducing a new referent, with maximum continuity with respect to the preceding co-text. Subject-verb ordering, on the other hand, always assumes that the subject term’s referent is important and salient in relation to the immediate concerns of the current discourse; thus, when a new referent is introduced in this position, there is an effect of discontinuity which is not apparent in the case of verb-subject ordering. In the text from which (5b) was taken, the subsequent co-text does not go on to mention the referent concerned (the Picasso museum in the old Grimaldi château) – but neither does it do so in the case of the cathedral in (6b), whose referent is in fact introduced via subject-verb ordering in this sentence. And unlike the term the Picasso museum in the old Grimaldi château in (5b), the cathedral in (6a) would receive the value INFOCUS, since it is the preverbal subject of a transitive, and not intransitive, predication. In the case of the two sentences (6a) and (6b), the subject expressions the cathedral (IN-FOCUS), and the rest of Antibes (MORE FOCUS (needed)) would both correspond to the FG Topic type known as ‘SubTopic’, since they each pick out an aspect of the town of Antibes, the immediate macro-topic of the text unit in which they occur. It is this more macro-discourse relation which, in my view, makes for the coherence of the three sentences (in the order in which they actually occur in the text). The writer’s assumption that the reader can easily infer the intended partwhole relation between the referents of the two preverbal subject expressions and the immediate macro-topic lessens the tension created by the placing of (relatively) new material in this position. It is possible that the feature which Huffman ascribes to texts generally is in fact genre-specific, occurring no doubt in narrative texts – W. Golding's Lord of the Flies and W. Cather's O Pioneers!, used as a testing ground by Huffman – but not in travel guides such as the one illustrated by the extracts in (6a), (5b) and (6b).

Focus of attention in discourse

127

One last general point about CS FOCUS systems: given that, as Hannay (1991: 140), Reinhart (1981: 56), Lambrecht (1994), Erteschik-Shir (1997) and others point out, a given clause may sustain several different topic and focus assignments, depending on its context at the utterance level, postverbal argument terms (direct and indirect objects) may fulfil the topic function. CS theory, as presently constituted, would seem to have no means of predicting these possibilities.

2.2. Deixis DEIXIS in CS terms is said to constitute a grammatical system whose substance (in Saussurean terms) is “degree of insistent pointing toward the intended referent” (cf. Diver n.d., 1995: 34–37 on the Latin demonstrative and anaphoric pronouns; Leonard 1995 on the Swahili demonstratives, and Aoyama 1995 on the Japanese ones). The degree of DEIXIS (required concentration of the addressee’s attention) is claimed (Diver, n.d.: 12) to reflect the following principle (known as Principle G): The more obvious the referent, the lower the item [indexical expressionFC] on the deictic scale used to refer to it.

The deictic scale which Diver (n.d.: 11) gives in this respect is formulated in terms of the Latin demonstratives (see also Dik’s 1997b: 223 similar table (21) structured in terms of preferred choices of Latin indexical forms for realizing NewTops, Focus, and GivTops, as well as Bolkestein 2000, for FG). Bolkestein examines the discourse conditions under which each of these forms (with the exception of ipse) occur, including of course the zero subject pronoun, which would be placed at the extreme low end of the scale in (9) below, as well as the subject relative pronoun qui. Ipse, the emphatic demonstrative, is at the highest end of Diver’s scale, and the clitic reflexive se is at the lowest end. The referent of se is normally what is in focus with respect to the associated finite verb. This is by virtue of the reflexive relation with the nominative expression, if present, where any nominative term and the verb inflection already encode HIGH FOCUS. According to Principle G, the less obvious the speaker assumes the intended referent in context to be for the addressee, the more a form from the upper part of the scale will tend to be used; and correspondingly, the more obvious, the more one from the lower part will tend to be chosen.

128

Francis Cornish

(9)

Deixis

HIGHEST ipse hic iste ille is se LOWEST

Hence, to use my own English example, in (10), the distal demonstrative determiner is used initially, since the intended referent is new to the discourse, the addressee being assumed not already to have noticed it: (10)

[It is dusk, and John and Mary are returning from a shopping trip. As John is parking the car, Mary exclaims:] Good God! Look at that incredibly bright light! [Mary gestures towards a point in the evening sky] What on earth do you think it could be? (Cornish 1999: 26, ex. (2.6a))

Here, the indexically ‘stronger’ determiner that, together with the lexical component of the term and the associated gesture, are used initially, since the addressee John is assumed not already to be attending to the bright light in question; while the indexically ‘weaker’ ordinary pronoun it is used in the second sentence to refer back to something assumed by this point to be within his attention focus. But it is not just the ‘degree of obviousness’ of a given referent in a discourse which is the criterial factor in the organization of the field of DEIXIS, in the CS sense. It is also, and mainly, that referent’s degree of importance, as judged by the speaker, within the immediate discourse at the point of use, which determines which form will be chosen from within the relevant scale. This is the so-called ‘direct’ strategy for referent establishment, the strategy in terms of the assumed degree of obviousness of the referent being the ‘indirect’ one. DEIXIS, in the CS conception, is claimed to be encoded by (among other expression-types) clitic pronouns of different varieties. Clitic reflexive pronouns in languages such as Latin, French and Spanish (but not in English, where reflexive pronouns are not clitic and so have different distributional properties) signal the lowest degree of DEIXIS (in this sense of

Focus of attention in discourse

129

the term), namely that very little effort need be expended by the addressee in order to retrieve the intended referent – why? Presumably because it is IN-FOCUS;13 and non-nominative, non-reflexive clitics signal a high(er) level of DEIXIS – since here, the intended referent is no doubt (assumed to be) NOT IN-FOCUS. Let us take the ambiguous example (11) from French as a basis for illustration: (11)

Elle [LOW DEIXIS] sait bien que la décision lui [HIGH/LOW DEIXIS] incombe. ‘Shei well knows that the decision is up to himk/heri/j.’

One point which (11) immediately illustrates is that, contrary to CS predictions, under the coreferential interpretation of the subject pronoun elle and the dative clitic pronoun -lui,14 both pronoun forms must bear identical (LOW) DEIXIS status, the coreferential interpretation pulling the dative clitic within the dominion of the antecedent subject one (see Van Hoek 1997 for this type of approach to pronominal anaphora, within Langacker’s Cognitive Grammar framework). Under the non-coreferential interpretation, it is possible that the DEIXIS status of the two referents concerned differs. (11) shows the difficulty encountered by accounts such as the CS one (also Gundel et al.’s 1993 ‘Givenness Hierarchy’), which frame their scales in terms of individual lexical items, rather than in terms of given uses of forms.

2.3. The CS FOCUS and DEIXIS systems compared However, this distinction between the subsystems of FOCUS and DEIXIS is not entirely convincing, since it fails to capture what is common to both (which is a good deal, in fact) in a revealing way. In addition, there is the ever-present danger of redundancy, depending on which definition of the FOCUS system is retained: that is, ‘attention is already concentrated on a given referent’, or ‘attention needs to be so concentrated’. It would be tempting to suggest (as would seem logical, after all) that (a) the very same morphemes which encode the value LOW DEIXIS are signals of HIGH/CENTRAL FOCUS, and that (b) those morphemes signalling HIGH DEIXIS simultaneously encode the value LOW/PERIPHERAL FOCUS. There is evidence for the (b) relationship in the shape of the French clitic complement pronouns lui-, leur-, le-, la-, and les-, which are said to encode the values PERIPHERAL (= LOW) FOCUS and HIGH DEIXIS (see Huffman’s 1997: 211 Figure 5.3). In the case of (a), however, the French reflexive se is claimed to encode the values LOW DEIXIS, but also PE-

130

Francis Cornish

RIPHERAL (LOW) FOCUS (Huffman 1997: 211). Yet it is at least plausible that an alternative analysis in terms of CENTRAL FOCUS might be made of French se, owing to its obligatory coreference with the nominative term (‘subject’) which acts as its controller (see also footnote 13 above). For García (1977: 152), Spanish se is outside the FOCUS system altogether. Furthermore, in the case of Latin, as Wallis Reid (pers. comm.) points out, the demonstratives ipse and hic are at the top of the Deictic Scale (see (9) above), and so clearly encode the value HIGH DEIXIS; and yet since both pronouns may be fully declined, each may take nominative case (giving them HIGH FOCUS status) or alternatively an oblique case (where they would signal LOW or PERIPHERAL FOCUS: in the latter case, this combination would be as predicted by (b) above). Similarly, the demonstrative is, which occurs near the bottom of the Deictic Scale in (9), may take an oblique case, and so encode the value LOW FOCUS. Now, given that these combinations hold, how can we make sense of the apparently contradictory instructions which the ‘non-conforming’ combinations signal? That is, HIGH FOCUS + HIGH DEIXIS: ‘the intended referent is assumed already to be in the addressee’s attention focus, but will need a high degree of effort in order to locate it’; and LOW FOCUS + LOW DEIXIS: ‘the intended referent is not assumed already to be in the addressee's attention focus, but at the same time it is not necessary to expend a great deal of effort in order to locate it’. In my view, the only sensible interpretation of these apparently contradictory combinations of instructions would be as follows: ‘the intended referent is/is not already in the addressee's attention focus (FOCUS system), but the referent at issue is/is not important for the current discourse purpose (DEIXIS system)’. Thus, it is the direct, not the indirect strategy associated with the DEIXIS system which necessarily comes to the fore in such cases. More thought clearly needs to be given to these kinds of relationships between the various values encoded by given forms from within the two systems at issue here, in order to try to better integrate the two complementary systems of DEIXIS and FOCUS in CS theory.

3. The four Topic statuses recognized in standard FG Functional Grammar uses both discourse-pragmatic and formal-coding criteria to distinguish Topics in relation to Foci. Two basic properties of discourse relevant to Topic and Focus status are those of topicality and fo-

Focus of attention in discourse

131

cality, considered as functions of the speaker's assumptions regarding his/her addressee's current pragmatic knowledge state at any given point in the discourse. Topicality has to do with the set of referents “‘about’ which information is to be provided or requested in the discourse” (Dik 1997a: 312). This notion is close to Brown and Yule’s (1983) concept topic framework. Focality, on the other hand, “attaches to those pieces of information which are the most important or salient with respect to the modifications which S [the Speaker] wishes to effect in PA [the addressee's current state of pragmatic knowledge], and with respect to the further development of the discourse” (Dik 1997a: 312). Importantly, topicality and focality are not mutually exclusive: areas where they overlap include constructions involving Contrastive or Parallel Focus. The ‘aboutness’ test which we used in Section 2.1 is employed to determine the Topic relation between an entity and a predication, as is the Question-Answer test (which also determines the potential Focus domain(s) of the utterance concerned). The latter test establishes the degree to which a given (declarative) utterance may be construed as a natural or coherent response to a question about a specified referent, providing the information required. This type of test is criticized by Gómez-González (2001), though it would seem that no very convincing evidence is marshalled against it. Now, Lambrecht’s (1994: §4.1.2) important distinction between topicreferent and topic-expression is reflected in the FG approach to topicality and Topic assignment, in that the tests just mentioned serve to isolate the referent bearing the Topic function in relation to a given Focus domain within an utterance; however, the various formal criteria needing to be satisfied in order for a constituent to be analysed as Topic obviously relate to the topic expression, in Lambrecht’s terms. That is, in order for a discourse entity to be recognized as a Topic of some kind, its linguistic exponent must be singled out in a systematic way by the language system involved: via morphological form, a particle, a specific type of construction (e.g. the left-detached construction), a specific word-order pattern, or a particular prosodic form. As we shall see, this form-oriented aspect of Topic recognition in FG has taken priority over the discourse-cognitive dimension15 (as developed in Reinhart 1981, or Lambrecht 1994, for example), in that it has motivated the recognition of a Topic type known as New Topic. However, according to the discourse-cognitive criteria, such referents are not topics at all within the clause in which they are expressed: rather they are clearly focal (see the references listed in footnote 23 on this issue). Moreover, it has led one pair of authors within FG (Mackenzie and Keizer 1991) to conclude that no Topic function exists at all in English, since according

132

Francis Cornish

to them, there is no systematic coding in that language which would single out expressions bearing the Topic function in relation to other, non-topical expressions (but see Hannay 1991: 142, fn. 9 for an opposing view). I believe that this over-reliance on systematic, overt coding as the main condition determining the existence of particular pragmatic functions in a language is a reflection of the fact that the assignment of such functions is standardly assumed to take place within the grammar, as the final specification before the complete underlying clause structure representation is transferred to the expression component for its formal realization. A number of linguists within FG have argued against this locating of pragmatic function assignment16 within the grammar, for a variety of reasons (e.g. Vet 1998, Bolkestein 1998, Van den Berg 1998). More recently the type of model presented by Hengeveld (this volume) is conducive to representing the looser kind of relationship between a pragmatic module and a grammatical one which these authors argue for. See also Hannay’s (1991) five proposed message management modes (to be discussed in Section 4 below) and their relevance for grammatical coding. Now, as Siewierska (1991: 148) points out, the subdivision of the Topic function within FG was designed to relate the specification of clause structure to the wider discourse setting in which the clause under analysis may occur. The four Topic statuses recognized in FG (Given, Sub, Resumed and New) would appear to subsume the distinction in CS between FOCUS and DEIXIS. Given, Sub and Resumed Topics would realize CS DEIXIS, and New Topics, FOCUS (i.e. they would correspond to the CS value NOT IN-FOCUS). FG GivTops, SubTops and ResTops clearly form a scale of deicticity, in the CS sense of the term, GivTops retrieving obvious referents, where no competition is involved between referents of the same type: examples used so far in this chapter are it in the second sentence of example (10), and elle in (11) (where this pronoun is unaccented). SubTops17 require the drawing of a minimal inference from a contextually given set of entities to the existence of a particular member of that set, or from a whole to one of its parts, and so on – thus, encoding a slightly higher degree of DEIXIS than that associated with GivTops: examples would be the preverbal subject terms the cathedral and the rest of Antibes in examples (6a) and (6b) above. Both these definite terms manifest a partwhole relation within the immediate macro-topic of their containing discourse segment, namely the city of Antibes. Finally, ResTops encode the highest degree of deicticity of the three subtypes, the referent being assumed to exist within the current discourse model, but no longer active or present in the addressee’s current con-

Focus of attention in discourse

133

sciousness. Thus the intended referent is assumed to be the least ‘obvious’ of the three at the point of use. An attested example is given under (12): (12)

... Elliot Morley, the Countryside minister, said he thought it was possible, with the new funds available, to aim at not just halting, but reversing the declines [of UK farmland bird species] in the medium term. “I'm in no doubt about the scale of this task, but I believe it is realistic to aim to do this by 2020”, he said. (Last paragraph of a 9 paragraph newspaper article ‘Ministers pledge £1bn to save farmland birds’⎯The Independent, 7 August 2000).

The demonstrative term this task in the fourth line of this extract refers back over the two preceding discourse units to the initial (macro-discourse) topic, introduced in the first, namely, the UK government’s commitment to halt and reverse the massive decline in farmland bird species in that country by the year 2020. Neither of Dik’s (1997a: Ch. 13) other two formal criteria supposedly accompanying terms with the ResTop function appear here, however (a connective indicating the start of a new discourse unit, and a specific reference back to the discourse unit to which the ResTop expression effects a return pop). It is the fact that the term itself occurs in a unit which is discourse-final, marking the conclusion of the discourse as a whole, I believe, which obviates the need for these two supposedly necessary conditions here. I will deal with the fourth subcategory of Topics recognized in FG (‘NewTops’) later in this section.18 Now, an important difference in emphasis arises between the two theories, regarding the way in which each views the discourse function associated with the expression-types at issue. For FG (and also other cognitively-oriented approaches like the ‘Givenness Hierarchy’ of Gundel, Hedberg and Zacharski, 1993), the context-bound Topic functions are predicated solely on the discourse-cognitive status which the speaker is assuming the referent to enjoy in the addressee’s discourse model at the point of use. For CS, on the other hand, the signals of DEIXIS are said to serve as instructions to the addressee to concentrate a given level of attention on the referent involved: HIGH – the referent is important to the current discourse concerns and so must be foregrounded relative to other referents; MID – the referent is of medium importance with regard both to HIGH DEIXIS and LOW DEIXIS referents; and LOW – the referent is of only background, ‘scene-setting’ (for example) importance. As Leonard (1995) and Aoyama (1995), respectively, point out in relation to the attested Swahili and Japanese examples which they present, discourse referents are not simply designated ‘decrementally’ with ever-decreasing levels of DEIXIS

134

Francis Cornish

marking as they are introduced, established, and maintained anaphorically throughout a discourse.19 Leonard’s (1995) DEIXIS scale as realized by the Swahili demonstratives is given in (13): (13)

h h-o le

HIGH DEIXIS MID DEIXIS LOW DEIXIS

The Swahili DEIXIS system (Leonard 1995: 273, item (1))

The LOW DEIXIS Swahili demonstrative -le may indeed function anaphorically to maintain reference to a recently-introduced entity, by virtue of its differential deictic meaning ‘no particular effort of concentration is needed to retrieve this referent’, as illustrated in Leonard’s (1995: 272) example (2a), here presented as (14); note that -le here would be glossed in English by the definite article the: (14)

Alimwona kijana. Kijana yu-le alisimama mlangoni. ‘He saw a youth; le-youth (LOW-DEIXIS) was standing at the door’

Clearly also, as expected, where a new referent is introduced into the discourse, and where that referent is also of importance to the speaker and to the current discourse purpose, a HIGH DEIXIS marker is used: (15)

[An oathgiver explains an oathing procedure to a man. He says:] “First I’ll speak, you’ll listen, then when I’m finished ⎯ nataka wewe useme h-ivi “Mikale mikale...” ‘I want you to speak in h- (HIGH-DEIXIS) manner [i.e. say the following]: “Mikale mikale...” (Leonard 1995: ex. (3))

The HIGH DEIXIS marker h- here is clearly equivalent to the English proximal demonstrative this, which Strauss (1993) argues is the HIGH DEIXIS member of the triplet this/that/it in that language, whereby the speaker is conveying a high degree of personal, subjective involvement with the referent, instructing the addressee to accord a high degree of attention-concentration on the referent, and to assume it is going to be important for the ensuing discourse. The distal demonstrative that is said to realize the value MID DEIXIS/FOCUS,20 and the ordinary inanimate pronoun it, the value LOW DEIXIS/FOCUS. As with Leonard’s (1995) attested examples of Swahili texts, Strauss gives examples where English HIGH DEIXIS

Focus of attention in discourse

135

this is used in subsequent (instead of, or even as well as, initial) reference to a given entity: (16) L 30 Caller: Crandall: L 70 Caller: Crandall:

Caller: Caller: L 76 Caller:

Crandall: Caller:

You spoke a little while ago to a man.hhh with regard to uh law,//an’ murder. Yes ma’am... ...you were talking about, or was he anyway,//that man well, no:: Now you see this the- again we come back to this: Murder is the act of killing someone, by, an- a specific dee:duh, described in the law of the sovereign state in which you live, as being unjustified. Now/ Yuh know/ By the d- conversation that chu had with this man: my impression was that there’s something drastic’ly wrong with us as a nation. Because I- uh when you say, that the Federal government cannot interfere,.hh in u-uh:: condoned murder. Like you[In a- in a- in a murder case⎯ [Like you mentioned in Mississippi (Strauss 1993: ex. (12))

Strauss points out (1993: 410) that the use of the HIGH DEIXIS/FOCUS expression this man in the first line of L 76, as fourth mention of the referent involved, is not evoking ‘new information’, but is rather motivated by the caller’s attempt to regain her turn, lost to Crandall at line 71. Another attested example (from my own corpus this time, here a newspaper article) is given under (17): (17)

[Beginning of article “God sends a sign to the heathen in a Chevy”, The Independent, 11 August 1999, p. 13] You are driving along in the relaxed American way, just getting from A to B and minding your own business, when your attention is grabbed by a gigantic billboard, “That ‘Love Thy Neighbour’ Thing”, reads the neat, white text, “I meant it – God”.

136

Francis Cornish This particular hoarding, which thrust itself into my line of sight somewhere on Interstate 81 in southern Virginia at the weekend, is the latest in my growing collection of “God” adverts...

From a purely ‘referent-tracking’ point of view, there is clearly no motivation for the use of a HIGH DEIXIS marker at the start of the second paragraph of this extract. The particular billboard at issue has been introduced by the indefinite nominal a gigantic billboard in the first paragraph, and it is already in the reader’s focus of attention by the end of that paragraph. If the adjective particular were omitted from the second-mention demonstrative NP, then that hoarding or the hoarding, or even (if the appositive relative clause following it were absent too), simply it, would have been appropriate. The HIGH DEIXIS proximal demonstrative phrase is clearly chosen here in order to underline the writer’s empathy or strong subjective involvement with the referent at the point recounted in her car journey when she first noticed it – and also to signal the fact that the referent is being envisaged in a slightly different context in this second paragraph, as simply one of many such billboards proliferating along American highways (the generic topic of that and the subsequent paragraphs, in fact). Nevertheless, the CS account of DEIXIS in terms of three degrees of effort needed by the addressee to concentrate his/her attention on the intended referent is not general enough. I have argued (in Cornish 2001) that these progressively greater degrees, as we move up the scale, of ‘urging’ or ‘insistent pointing’ on the speaker’s part that the addressee attend to the referent may be predicted from what can be seen to be more basic cognitive-interactional values encoded by the members of a given ‘deictic scale’. If we take as a model the three English indexicals this, that and it, for example,21 then the proximal (marked) variant has the effect of placing the referent within the speaker’s discourse sphere, signalling his/her strong subjective involvement with that referent and allowing the inference that this referent does not constitute ‘negotiated’, mutually-validated information. From this and the ‘egocentricity’ principle which clearly regulates discourse interpretation generally, it can be deduced that the speaker considers the referent at issue to be highly important for the ensuing discourse, and that the addressee should pay special attention to it. Particular uses of this in English such as colloquial ‘indefinite’ this and anticipatory or cataphoric this (both instances of discourse deixis in my view), as in (18a, b), are natural extensions of this value:

Focus of attention in discourse (18)

137

a. There was this terrible boy at the circus, who.... b. Now listen to this: a man went into a butcher’s shop and bought a whole pig...

The distal (unmarked) variant, that, would have the discourse-interactional value whereby its referent is placed outside the speaker’s subjective sphere and into a shared, ‘negotiated’ space, in which the speaker’s and addressee’s stances are in co-alignment. As such, the referent is not (necessarily) envisaged as representing new information for the addressee, and thus the somewhat lower degree of insistent urging that s/he attend to it is predicted (see also Cheshire 1996 and Glover 2000 for similar analyses of English this and that). The ‘recognitional’ or ‘reminder’ use of that provides a good illustration (see also the demonstrative term that ‘Love Thy Neighbour’ thing in example (17) above): (19)

Tony Blair was said to be on the phone last week to his German oppo, Chancellor Schröder, trying to talk through an upbeat final instalment of that dismal industrial soap opera called Longbridge. (The Guardian, 3 May 2000, p. 12)

Finally, the non-demonstrative, inanimate (unaccented) pronoun it signals not only that its intended referent constitutes shared information, but also that that information is currently in the addressee’s (and the speaker’s) ongoing attention focus. Again, this value implies that the addressee need expend no particular effort at all in locating and retrieving the intended referent, since it is (assumed to be) ‘right there’, in his/her current consciousness. Unlike both the CS account of DEIXIS and the Gundel et al. (1993) Givenness Hierarchy, which are formulated in terms of particular lexical items, the four FG Topic types are characterized in terms of purely discourse-functional roles: although there are claimed to be certain prototypical indexical markers which realize such roles (zeros and unaccented 3rd person pronouns for GivTops, definite lexically-headed terms for SubTops, expanded definite or demonstrative terms for ResTops, and indefinite lexically-headed terms for NewTops), it is also clear that, in given contexts, other expression types may also be used to realize each of these functions (for example, the demonstratives, whether as determiners or pronouns, may in fact realize all of them – except ResTops, where demonstrative NPs rather than pronouns are required: see Cornish 1998 for some discussion, based on an English-language newspaper article).

138

Francis Cornish

This would seem to me to be desirable, since as we have seen, other, strategic-discourse as well as expressive and viewpoint-signalling factors come into play in addition to purely referent-tracking ones in the use of indexical expressions. However, no allowance seems to be made within FG for proactive, strategic uses, based on interactional or discourse-constructional motivations such as the ones which we briefly saw in examples (16) – (17). But the FG account has the advantage over the CS and the Givenness Hierarchy (GH) ones, in that specific lexical items are not ‘tied’ to particular positions on a hierarchy of forms in terms of relative indexical ‘strength’.22 In any case, we have seen that the CS scales in terms of degrees of insistent pointing towards the intended referent may be predicted on the basis of more fundamental discourse-interactional values; and it is likely that the GH may similarly be subsumable under a more precise discourse-interactional characterization of each of the form types represented on the Hierarchy. FG NewTops, on the other hand (the most problematic of the four subtypes recognized in the theory), clearly fall within the FOCUS system distinguished in CS theory. In fact, where such NewTops are introduced in postverbal position (as is most often the case in Dik’s 1997a presentation: see (20) below for an illustration), they would be viewed in CS theory, as we saw earlier, as being NOT IN-FOCUS – that is, as not worthy of the addressee’s attention at all! So here we have what at first sight is a diametrically opposite analysis of a given form type by each theory. Actually, as we saw in the case of the comparison between non-inverted constructions such as (5a) and inverted ones such as (5b), the contrast is a paradox rather than a true contradiction, since CS is saying that Dik’s NewTops are nontopical (which is correct, as we shall see below), while in FG they realize one type of Focus (in the Information Structure sense); and clearly, both statuses can simultaneously characterize the same given constituent. In fact, even within FG as they are defined in Dik (1997a: Ch.13), they can be analysed as simply constituting in the unmarked case one type of Completive Focus, since we are dealing here with the introduction of a referent which is anticipated to play a role in the subsequent discourse. Dik (1997a: 312) in fact specifically mentions NewTops in illustrating the possibility of overlap between topical and focal elements in a discourse. He characterizes NewTops as the introduction of new participants or entities about which the speaker intends to say something later on, and by virtue of that introduction, as constituting the main point of the containing utterance. His example of this is (20):

Focus of attention in discourse (20)

139

All of a sudden we saw A GIGANTIC SHARK. (Dik 1997a: 312, ex. (6))

However, it is debatable whether in fact the referent of the NP A GIGANTIC SHARK in this context (in fact, a thetic utterance) is topical at all – that is, whether the utterance of (20) may be construed as a categorical rather than thetic one, and ‘about’ a gigantic shark: witness the incoherence of #As for a gigantic shark, all of a sudden we saw one/it. Applying Erteschik-Shir’s (1997) ‘lie-test’ clearly shows that the content of a gigantic shark in (21) is part of the focal information conveyed via this utterance: (21)

A: All of a sudden we saw A GIGANTIC SHARK. B: That’s not true! It was a BABY WHALE.

The notion of NewTop is in fact an incoherent one itself, if it is conceived (as I believe it must be) in discourse-contextual terms. After all, discourse topics do not emerge fully-fledged just through being introduced into a given discourse; they have to be established, and then maintained if they are not to fade from salience or be superseded by subsequent topics (see Cornish 1998 for some discussion). New topics may be established on the basis of the introduction of a new referent into a discourse, as in the case of a gigantic shark in (20); but I think that it is a mistake to believe that the referents of such expressions ipso facto have the status of topics.23 Topic introduction and establishment is now widely conceived within the literature as being a joint, cooperative undertaking, involving the coordination of both speaker and addressee, and is not the prerogative of the speaker alone. In any case, all the examples of NewTops presented in Dik’s (1997a: 315-318) subsection on NewTops (§ 13.3.1) in fact fall within thetic, not categorical utterances, where Dik specifically makes the point that such referents tend to be introduced via expressions occurring towards the end rather than at the beginning of the clause.

4. Pragmatic functions within the new FDG model Quite clearly, the assignment of pragmatic functions within the new Functional Discourse Grammar model (Hengeveld, this volume) must be part of the Interpersonal Level specified in the model, a level which subsumes and ‘controls’ the specifications made at the two lower levels in the system (those of Representation and Expression). The marking of information-

140

Francis Cornish

structural relations at the level of the clause needs to be sensitive both to features of the communicative context, and to the speaker’s evaluation of the addressee’s current cognitive state. This can be done once the Cognition and Communicative Context components of the FDG model which feed and are in turn fed by the Interpersonal Level (as indicated in Hengeveld, this volume: Fig. 1) are adequately fleshed out. Adopting Hannay’s (1991: 148) five modes of message management (All-New information, Topic, Reaction, Neutral and Presentative), it is clear that this initial planning stage must be situated at the Interpersonal Level recognized in the FDG model, where communicative intentions are formed, expressed minimally by the Move, which is realized by one or more constituent Discourse Acts. See also for a more specific elaboration of this component Van den Berg’s (1998) ‘pragmatic’ and ‘message’ modules within his ‘pragmatic functional grammar’; here, pragmatic functions are handled by the ‘message module’ (1998: 100). I assume his ‘pragmatic’ module would correspond to the Interpersonal Level in Hengeveld’s FDG model, and that his message module would correlate with the FDG Representational Level. Examples illustrating Hannay’s message management modes presented so far in this chapter are (1b) and (4a) for All-New, (1a) for Topic, and (4b) and (5b) for Presentative. The Neutral mode is said not to involve an initial Topic or Focus expression and comprises a ‘dummy’ subject (it or there). An example would be It was surprising that the Greens won the election. As for the Reactive mode, it involves the preposing of immediate, expressive-subjective material bearing the Focus function, as in Hannay’s (1991: 143) example (2b): (2) Q: Did you get wet?… (b) Wet? Bloody soaking I was. Now, clearly, each message management mode, with the exception of the first (the All-New mode), presupposes a specific kind of context. This must be represented in the Communicative Context component, whose task is to provide a representation of the discourse developed prior to the utterance about to be produced, as well as of the prevailing situational context. A very important part of this discourse representation must be (a) an indication of the current (local and global) macro-topics and (b) the structure of this discourse (i.e. the level or ‘unit’ of the discourse under development which the discourse has reached). The Presentative, Neutral and All-New modes may serve to initiate a new discourse unit, since their essential function is to introduce new referents or states of affairs. In the case of the Presentative mode, as we saw in example (5b), there is often a preposed thematic locative, temporal (or more generally, circumstantial) phrase introducing the constituent manifesting it; in which case, the content of the

Focus of attention in discourse

141

clause as a whole is presented as tightly connected with the preceding cotext. Thus, such clauses may serve either as transitions between one discourse unit and the next, or as links between utterances within the same unit. Constructions realizing the All-New mode, however, introduce wholly new situations or events, and re-set the space, time and thematic coordinates within a given discourse. In the case of Reaction mode clause realizations, these are mainly restricted to face-to-face spoken language of a dialogic kind (as in the illustrative example given above); as its very name suggests, the Reaction mode presupposes a ‘first pair-part’ to which it constitutes a ‘reaction’. Thus, its clausal exponents are restricted to occurring within minimal spoken discourse units (the adjacency pair in conversational discourse). The Topic mode too – that is, where the Topic type at issue is a GivTop – is by definition restricted to intra-unit occurrences. In terms of the intra-clause Topic functions, most clearly, ResTops (manifesting the highest level of CS deicticity of the three discourse-bound FG Topic functions) would refer to an entity evoked within an earlier, distinct discourse unit in relation to the one in which their exponents occur. But (as Dik 1997a: 325 himself implies), they could not open a new discourse unit on their own: Dik claims that a specific connective signalling the start of a new discourse unit is needed, presumably for this purpose. SubTops, too, may signal the start of a new discourse unit, but one which is clearly subordinate to the immediately preceding one – or which is at least at the same level of subordination as this unit. Whereas the Communicative Context component within the FDG model must clearly keep track of the preceding current discourse, as we have seen, and must feed this information into the Interpersonal Level in order for it to be able to manage the organization of future messages, the Cognitive component must contain long-term representations of both linguistic and non-linguistic kinds: encyclopaedic information concerning real-world properties and relationships, as well as personal information assumed by the current speech participants to be mutually shared, and which is relevant to and evoked via the current discourse. It must also contain information regarding semantic relations between lexemes (such as hyponymy, meronymy, synonymy and antonymy), thus motivating the choice at the Expression Level of appropriate first restrictors within definite term structures manifesting GivTop or SubTop functions within the wider discourse (see Hannay, 1985b: 55–56, exx. (26) – (35) for the range of semantic or pragmatic relations in terms of which SubTops may be realized).

142

Francis Cornish

With this subsuming, controlling relationship between the higher-level Interpersonal Level and the lower, intermediate-level Representational one, which forms a message on the basis of the pragmatic structuring specified at the Interpersonal Level, Focus assignment or the lack of such assignment may be shown to have an effect on the semantic nature of the predicator selected at the lower Representational level, something which was impossible or very difficult to show under the standard model of FG (Dik 1997a: Ch. 13).24 Bolkestein (1998: 204–206) likewise gives Latin examples where the assignment of Focus function to given constituents in a clause may block Subject or Object assignment to certain other constituents. Thus, in her words (p. 205), “the location of main Focus must be part of the input for the rule that takes care of syntactic function assignment”. All these phenomena show clearly that pragmatic function assignment must take place before, and not after, the stages at which semantic and syntactic properties are established within the clause. In any case, as the various tests, such as the Question-Answer test for Topichood and potential Focus domains, clearly indicate, information from the wider discourse context in which a given clause is set is crucial for the assigning of pragmatic functions to relevant constituents within it. The structure of the discourse corresponding to a particular Move, then, is established at the Interpersonal Level in the new model, in terms of a coherence relation with respect to the last such Move, and to the wider discourse and thematic structure as specified by the Communicative component. The mode of message management adopted at this level will in part determine at the lower Representational Level a particular underlying clause structure, with Topic and Focus assignments marked (as appropriate) as a function of the message structure indicated at the Interpersonal Level. Finally, the Expression rule component will convert these specifications into an actual object-language expression (but now, there is no requirement of a one-to-one relation between a type of systematic coding in the language in question and a given Topic or Focus function: Topic referents, in particular, are established as such independently of a given coding type, and indeed are already marked as such at a more languageexclusive level of representation – the Interpersonal Level). So topics (referents) may be such without there needing to be dedicated languagespecific coding devices for expressing them, the Interpersonal and Representation Levels now being distinct and containing non-isomorphic types of units.

Focus of attention in discourse

143

5. Conclusions Both theories (CS and FG, but CS more so than FG) need to take account more centrally of the thetic/categorical distinction in predicting possible Topic-Focus relations in discourse – and in doing so, CS theory needs to take on board the information-structural dimension of the flow of utterances in discourse, in addition to its purely topic-based account. It could usefully take cognizance of the fundamental FG distinction between topicality and focality in discourse, as distinct from the assignment of particular Topic or Non-Topic (i.e. IN-FOCUS or NOT IN-FOCUS) roles to given constituents within the clause. As it is, its notion of Focus of Attention leaves the focality dimension totally out of account – even though its presence is implicit in CS analyses. Under the conception where FOCUS in CS is defined in terms of the addressee’s being instructed to concentrate or otherwise his/her attention on a given referent, it is basically no different from the system of DEIXIS (at least under the indirect strategy interpretation), whose signals likewise enjoin the addressee to accord different levels of attention concentration on a particular referent, as a function of its assumed degree of obviousness for the addressee or reader – but where three degrees of ‘insistent urging’ rather than two, as for FOCUS, are involved. Viewed in this light, there is no essential difference between the two systems (apart, as we have seen, from the partially complementary set of clitic pronoun signals involved in coding each system). There would appear to be two interpretations where the FOCUS and DEIXIS signalling systems have differing functions: (a) where CS FOCUS has the cognitive/AI interpretation (where the referent concerned is simply assumed already to be at the forefront of the addressee’s consciousness at the point of use): here, the FOCUS system signals what is considered to be topical or laying claim to the addressee’s conscious awareness, and the DEIXIS system (under its ‘indirect’ interpretation) indicates how to retrieve various sorts of discourse referents, given their already-existing FOCUS level; and (b) where the morpheme signalling DEIXIS has the direct status of indicating the level of discourse importance of a given referent, from the speaker/writer’s point of view. In fact, with the systems of DEIXIS and FOCUS, CS theory only concerns itself with topic/non-topic signalling in discourse – i.e. with what the speaker is or is not centrally talking about in a given utterance; his or her marking of what s/he wants the addressee to do with (or to) this referent (via the various types of Focus function, in the IS sense) appears to be left out of account.

144

Francis Cornish

FG, for its part, might usefully pay more heed to establishing criteria for determining the various attention-focus or accessibility statuses of given referents in the flow of discourse, thus motivating the successful prediction of the use of the various subtypes of Topic expression in accessing them (this point is also made by Gómez-González 2001: 165). Finally, one of the four subtypes of FG Topics, NewTops, may, I think, be safely abandoned (since its provision is already catered for, in the unmarked case, by Completive Focus), thereby simplifying the system as a whole.

Notes 1.

2.

3.

4. 5.

6. 7.

This chapter is a revised and expanded version of a paper presented at the 9th International Conference on Functional Grammar, held at the Universidad Nacional de Educación a Distancia, Madrid, from 20 to 23 September 2000. My thanks in particular to Wallis Reid, Kees Hengeveld and Knud Lambrecht as well as the editors of the present volume for their very helpful comments on earlier versions. This term should be understood here in the sense which Lambrecht (1994) adopts, namely where discourse information is that which is derived by an addressee from the combination of a topic with a focus (within a topiccomment utterance), or in more general terms, from that of items one with another in the flow of text. In CS theory, FOCUS (‘IN-’, ‘NOT IN-’, ‘MORE’ or ‘LESS’) is assigned by the speaker-writer to individual arguments (participants) both in relation to the particular predicator chosen (what it calls ‘EVENT’) within the clause, and to the wider discourse. I am indebted to Knud Lambrecht for this clarification. As Knud Lambrecht points out (pers. comm.), this is partly due to the fact that the term in question is indefinite (thereby marking its referent as unidentifiable by the addressee), and not definite (and hence potentially identifiable). It is also due, of course, to its position to the right of the verb, and to the fact that, in its spoken realization, it would receive a nuclear pitch-accent. Lambrecht and Polinsky (1997) present a range of distinctive properties characterizing thetic utterances cross-linguistically. In this connection, see Reid’s (1991: 181–182) presentation and analysis of a newspaper report of a marathon race, as well as Huffman’s (1993) application of the system to two fictional narratives. I am indebted to Wallis Reid for this characterization of the CS FOCUS system. For good, fairly recent discussions of the thetic vs. categorical distinction, see Rosengren (1997), and Lambrecht and Polinsky (1997).

Focus of attention in discourse 8.

9. 10.

11.

12.

13.

14. 15.

145

See Birner and Ward (1998: 159), as well as Gundel (1999: 299–300) on the prosodic difference between ‘semantic focus’ (i.e. unmarked focus in a topic-comment structure), and ‘contrastive focus’. See in particular Hannay (1985a) for an FG account of this construction. In English, of course, preverbal subjects may also be focal, since (given an appropriate context) they may receive a high pitch-accent, and thereby convey new information relative to that context (A: Who ate the fish? B: JANE ate the fish/did). See also the point made by Reinhart (1981: 57) that the description ‘focus of speaker’s [and addressee’s] attention’ may characterize topics and foci alike. Strictly speaking, Huffman as a CS linguist would not use the traditional terms ‘subject’ and ‘verb’ here, preferring the descriptive terms ‘(first) participant’ and ‘event’, respectively. However, I use the more traditional terms here for ease of comparison with the IS conception. This is not always the case, however: for example, in Huffman's (1997: 209–210) analysis of French clitic pronouns, the 3rd person clitic reflexive se is claimed to blend CENTRAL and PERIPHERAL (as he calls it) FOCUS, since on the one hand it automatically links up with the IN-FOCUS ‘subject’ term (nominative in form, if a clitic pronoun: il/ils, though not elle/elles, since these forms may occur in other relations to the EVENT (verb) than that of P1 (initial argument)); and on the other, it involves a ‘remention’ of the same participant. Nonetheless, I think that a case could be made for French clitic se to encode the value CENTRAL FOCUS, by virtue of its automatic binding by the HIGH FOCUS nominative term bearing subject function with respect to the verb, and in relation to the latter’s inflectional form. In FG, predications containing a reflexive pronoun, in relation to ones with a full term phrase (i.e. transitive clauses), are considered to have undergone ‘argument reduction’, the reflexive pronoun being the surface marker of this relation. Such clauses are clearly intransitive (like inherently pronominal verbs in French and other such languages). Hence the reflexive (clitic in French) pronoun does not correspond to an argument semantically, and cannot therefore be assigned any value pertaining to argument expressions. See García (1977) for an excellent analysis of Spanish se within the CS framework. The hyphen preceding given pronoun forms indicates that the clitic variant is what is intended. As Siewierska (1991: 148) points out, “though a constituent may be topical or focal on discourse grounds, unless its topical or focal status is coded in the structure of the clause, there is no basis for recognizing a special clausebound level of pragmatic organization distinct from the semantic and syntactic levels of clause structure”.

146

Francis Cornish

16.

That is, the selection of given constituents as bearing Topic or Focus function in terms of the wider discourse context and the assignment of these functions to expressions conforming to the formal coding requirements at the same level. See Hannay (1985b) on SubTopics. Siewierska (1991: §§6.3.2–6.3.3) and Gómez-González (2001: §§5.2.15.2.4) present critical surveys of the four types of Topic function within standard FG. That is, where a ‘HIGH DEIXIS’ marking might be used to introduce a referent, a ‘MID DEIXIS’ one to establish it as a topical discourse referent in the current stretch of discourse, and subsequently a ‘LOW DEIXIS’ marking to maintain its current topicality (as long as it is not superseded by a new referent bearing a higher claim to topic-worthiness). Strauss (1993: 403–404), though recognizing that the three-member system of indexical pronouns she is dealing with would normally fall within the CS DEIXIS system, nevertheless terms it ‘FOCUS’. This terminological shift is highly significant in the light of the discussion in Section 2.3 above. See also the studies of the Finnish demonstratives by Östman (1995) and Laury (1997), or of the Dutch demonstrative determiners by Maes (1996: ch. 4). Admittedly, the GH is an ‘implicational’ scale, so that particular forms may move up or down the scale according to the drawing of certain implicatures based on one or the other of Grice’s two Quantity maxims. See also Mackenzie and Keizer (1991: 193–194), Hannay (1991: 138), and Siewierska (1991: 162) for a similar view within standard FG. Outside FG, see also Lambrecht (1994: 129–130, 168). Mackenzie and Keizer (1991: 194) suggest that Dik's (1997a) ‘NewTops’ should be analysed as a type of ‘Presentative Focus’ (a subtype of New or Completive Focus). GómezGonzález (2001: §5.2.4) provides a survey of published criticisms of the alleged topical status of so-called ‘NewTops’. See Cornish (2002: §4) on this particular problem, presented in terms of French examples where subject and (intransitive) verb are inverted following a clause-initial locative satellite, and where it is the predicate corresponding to the head noun of the subject term expression which is assigned the Focus function, thereby acting as predicator (example: A l’horizon couve un orage ‘On the horizon (there) brews/is brewing a storm’); or where canonical subject–verb order obtains, the verb then receiving Focus function and assuming its full lexical value as predicator for the clause (example: A l’horizon, un orage couve ‘On the horizon, a storm is brewing’).

17. 18.

19.

20.

21.

22.

23.

24.

Focus of attention in discourse

147

References Aoyama, Takashi 1995 Deixis and value: a semantic analysis of the Japanese demonstratives. In: Ellen Contini-Morava and Bonnie S. Goldberg (eds), 289– 320. Birner, Betty J. and Gregory Ward 1998 Information Status and Noncanonical Word Order in English. Amsterdam and Philadelphia: Benjamins. Bolkestein, A. Machtelt 1998 What to do with Topic and Focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. 2000 Discourse organisation and anaphora in Latin. In: Susan Herring, Pieter van Reenen and Lene Schoesler (eds), Textual Parameters in Older Languages, 107–137. Amsterdam and Philadelphia: Benjamins. Brown, Gillian and George Yule 1983 Discourse Analysis. Cambridge: Cambridge University Press. Cheshire, Jenny 1996 That jacksprat: an interactional perspective on English that. Journal of Pragmatics 25: 369–393. Contini-Morava, Ellen and Barbara S. Goldberg (eds) 1995 Meaning as Explanation: Advances in Linguistic Sign Theory. Berlin and New York: Mouton de Gruyter. Cornish, Francis 1998 Les ‘chaînes topicales’: leur rôle dans la gestion et la structuration du discours. Cahiers de Grammaire 23: 19–40. 1999 Anaphora, Discourse, and Understanding. Evidence from English and French. Oxford: Clarendon Press. 2000a L’accessibilité cognitive des référents, le centrage d’attention et la structuration du discours: une vue d’ensemble. Verbum 22: 7–30. 2000b The Functional Grammar conception of discourse anaphora: A (constructive) critique. Working Paper in Functional Grammar 73. Amsterdam: University of Amsterdam. 2001 “Modal” that as determiner and pronoun: The primacy of the cognitive-interactional parameter. English Language and Linguistics 5: 297–315. 2002 'Downstream' effects on the predicate in Functional Grammar clause derivations. Journal of Linguistics 38: 247–278. Dik, Simon C. 1997a The Theory of Functional Grammar. Part 1: The structure of the clause. Berlin and New York: Mouton de Gruyter.

148

Francis Cornish

Dik, Simon C. 1997b The Theory of Functional Grammar. Part 2: Complex and derived constructions. Berlin and New York: Mouton de Gruyter. Diver, William (n.d.) Avoidance of the obvious: the pronoun as a minimax solution. Unpublished course text for Linguistics G 6801. 1995 Predictions from the hypotheses about the characteristics of the texts (§3.4.4.1), and Measuring the strength of the correlation (§3.4.4.2), extracts from “Theory”. In: Ellen Contini-Morava and Bonnie S.Goldberg (eds), 34–35 and 35–37. Erteschik-Shir, Nomi 1997 The Dynamics of Focus Structure. Cambridge: Cambridge University Press. García, Erica 1977 On the practical consequences of theoretical principles. Lingua 43: 129–170. Glover, Karen 2000 Proximal and distal deixis in negotiation talk. Journal of Pragmatics 32: 915–926. Gómez-González, María de los Ángeles 2001 The Theme-Topic Interface. Evidence from English. Amsterdam and Philadelphia: Benjamins. Gundel, Jeanette K. 1999 On different kinds of Focus. In: Peter Bosch and Rob van der Sandt (eds), Focus. Linguistic, Cognitive, and Computational Perspectives, 293–305. Cambridge: Cambridge University Press. Gundel, Jeanette K., Nancy Hedberg and Ron Zacharski 1993 Cognitive status and the form of referring expressions in discourse. Language 69: 274–307. Hannay, Mike 1985a English Existentials in Functional Grammar. Dordrecht: Foris. 1985b Inferrability, discourse boundness and subtopics. In: A. Machtelt Bolkestein, Casper de Groot, and J. Lachlan Mackenzie (eds), Syntax and Pragmatics in Functional Grammar, 49–63. Dordrecht: Foris. 1991 Pragmatic function assignment and word order variations in a Functional Grammar of English. Journal of Pragmatics 16: 131–155. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Hengeveld, Kees This vol. The architecture of a Functional Discourse Grammar.

Focus of attention in discourse

149

Huffman, Alan 1993 Cognitive and semiotic modes of explanation in Functional Grammar. Talk given at Rutgers University, 10 October 1993. 1997 Categories of Grammar: French lui and le, Amsterdam and Philadelphia: Benjamins. forthc. The Control-Focus interlock. In: Modern English: A Columbia School Grammar. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Lambrecht, Knud and Maria Polinsky 1997 Typological variation in sentence-focus constructions. In: Singer, Kora, Randall Eggert and Gregory Anderson (eds), Papers from the Panels, 33rd Regional meeting of the Chicago Linguistic Society. 189–206. Laury, Ritva 1997 Demonstratives in Interaction. The emergence of a definite article in Finnish. Amsterdam and Philadelphia: Benjamins. Leonard, Robert A. 1995 Deixis in Swahili: attention meanings and pragmatic function. In: Ellen Contini-Morava and Bonnie S. Goldberg (eds), 271–287. Mackenzie, J. Lachlan and M. E. Keizer 1991 On assigning pragmatic functions in English. Pragmatics 1: 169– 215. Maes, Alfons 1996 Nominal Anaphors, Markedness, and the Coherence of Discourse. Leuven: Peeters. Östman, Jan-Ola 1995 Recasting the deictic foundation, using Physics and Finnish. In: Masayoshi Shibatani and Sandra Thompson (eds), Essays in Semantics and Pragmatics in Honor of Charles J. Fillmore, 247–278. Amsterdam and Philadelphia: Benjamins. Reid, Wallis 1991 Verb and Noun Number in English: A functional explanation. London and New York: Longman. Reinhart, Tanya 1981 Pragmatics and linguistics: An analysis of sentence Topics. Philosophica 27: 53–94.

150

Francis Cornish

Rosengren, Ingrid 1997 The thetic/categorical distinction revisited once more. Linguistics 35: 439–479. Siewierska, Anna 1991 Functional Grammar. London and New York: Routledge. Strauss, Susan 1993 Why this and that are not complete without it. In: Katie Beals et al. (eds), Proceedings of the 29th Regional Meeting of the Chicago Linguistic Society, 403–417. Chicago Linguistic Society. Van den Berg, Marinus 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–103. Van Hoek, Karen 1997 Anaphora and Conceptual Structure. Chicago: University of Chicago Press. Vet, Co 1998 The multilayered structure of the utterance: about illocution, modality, and discourse moves. In: Mike Hannay and A. Machtelt Bolkestein (eds), 1–23.

The complementarity of the process and pattern interpretations of Functional Grammar Michael Fortescue

1. Introduction From its inception, the Functional Grammar formalism has in principle been interpretable in two ways, either as a static system of functional choices, or as a dynamic process model (more specifically, a production model). Interest in the second kind of interpretation has increased with the rapprochement between cognitive and functional linguistics in recent years, and not everyone within the FG sphere has been as sanguine as Dik himself in considering the formalism directly implementable as a computer model of cognition.1 What I want to explore in this chapter is the question of what the consequences of interpreting FG as a process model really are. If one wants the FG model to represent production and/or comprehension directly, one should, I would suggest, base it on underlying cognitive processes which have an intuitive feel of plausibility about them (not, for instance, on hypothetical neural circuits at some dehumanized ontological level). As has been argued by Harder (1992: 313ff.), one aspect of the model that simply does not fit a process/procedural interpretation is the first-order predicate formalism of the lowest level of the model.2 Taken straight from the abstract world of logicians like Quine (1960), representations of this kind, with their variables, quantifiers and open predicates, are frankly not obvious candidates for psychological ‘reality’. Strawson, in describing the cognitive origin of ‘perspicacious grammar’, specifically warns against introducing variables and quantifiers as opposed to individual particulars at the basis level of the logical representation of propositions (Strawson 1974: 117). That is because it is individual particulars (and their more abstract, higher-order entities derived by ‘sub-

152

Michael Fortescue

stantiation’) that enter into the act of reference, not variables. The basic cognitive processes as I see them are rather Whiteheadian ‘prehensions’, the grasping and assimilation of both perceptual and conceptual ‘data’, specifically ‘propositional prehensions’, which link the prehension of a predicate to the prehension of a set of logical subjects (the totality of relevant ‘entities’ bound by the predicate – cf. Fortescue 2001: §1.2.5). At this level of cognition we are not yet talking of linear chains of linguistic signs: the processual ‘concrescences’ whereby multiple data-to-be-expressed are integrated on the way towards a determinate linguistic output should be taken as having their own internal complexity (unless purely automatic response is involved). Now Dik’s term variable ‘x’ in underlying predications simply stands for a first-order entity (as opposed to a proposition ‘X’, for example). In a nuclear predication variables and their first restrictors are replaced by terms. ‘Firstness’ is all that determines which predicate restrictor is to be treated as term head by binding with a preceding variable. This kind of notation is innocuous and does indeed serve a useful purpose, namely in relating underlying clause structure to the lexicon, where individual predicates (including those defining terms) can be represented as a predicate frame relating directly to a particular kind of State of Affairs. However, this is a matter of pattern, not of process. There is no guarantee that the straightforward linkage between predicates and SoAs has any direct relevance for cognitive processes as such. I would like now to suggest that we may need two versions of the FG model, a Process and a Pattern version, rather than just one (Dik’s original one, say) serving both purposes – or, if there is agreement on a monolithic framework again, then it should at least be of such a nature as to be interpretable in two clearly distinct ways and not remain indeterminately ‘hybrid’. This has been the subject of much debate in recent years within FG circles, in the form of discussions of the upward limit of layering of the theory. Dik himself proposed that his model should be envisaged as embedded in – or side by side with – a distinct discourse grammar (Dik 1997: 409ff.), and this idea has been followed up in two principal ways (as reflected in the various articles in Bolkestein and Hannay 1998),3 either by advocating the extension of the original model to include such a ‘discourse grammar’ or by calling for their treatment as two separate, interacting modules. I would suggest that neither of these options is really what we need if the original functional model is to be seen in cognitive terms. For you cannot simply embed pattern within process (or vice versa). The model as it is (as pattern) needs a process interpretation – where by ‘process’ I re-

Process and pattern interpretations

153

fer to all procedurally interpreted levels of the model, from the choice of basic predicate (and its relationship to the lexicon) up to the highest level of illocutionary intention (and its relationship with still broader discourse context). My own position as regards the upper limit of the model as such is thus twofold: from the pattern point of view (the actual ‘emic’ coding choices of grammar) it is the illocution, just as Dik proposed. From the process view, illocutions are simply an intermediate stage of organization whereby complex intentions (perlocutions, etc.) find their way to expression via the available linguistic code. The upward limit from this viewpoint would correspond to the Whiteheadian ‘subjective aim’ defining the individual communicative act. Such an intentional stance is, along with propositional content, an integral part of the Whiteheadian judgment (which, unlike those of standard propositional logic, is not limited to the assignment of positive and negative truth-values). The way in which the intention to utter such a judgment is worked out with the help of the ‘template’ of grammar is one of the constraints on pure expression. A grammar as such is simply a set of abstract ‘forms of definiteness’ (Whiteheadian ‘eternal objects’), and the FG model (as pattern) is best regarded as a higher-level set of generalizations across the grammars of many languages. The obvious direction to move in describing a process interpretation of FG is thus top-down (compare also Hengeveld this volume). Now a cornerstone of FG is the notion that pragmatics contains semantics, which in turn contains morphosyntax (as in Dik 1989: 7). So if pragmatics really embraces all levels of grammar it should be seen to extend down to the very lowest level of the process model it embraces, namely the organization of predications. This will be my starting point in what follows. Also for Hannay (1991: 146ff.) the ‘place to start’ is with pragmatic/message organizing (choices to do with new versus given information, etc.).4 Note the difference from my proposal, however: Hannay sees his message management module or component as an interface between a product-oriented model and a process one (of verbal interaction). In fact he indicates choice of ‘mode’ as taking place at the highest level of the extended FG model, i.e. as part of the ‘pattern’ grammar, e.g. TOPIC (mode) DECL E1: [X1: ...]. My own approach is not really modular at all: both discourse and grammar have their pattern and their process aspects, only the pattern aspect of discourse is far more heuristic, relying heavily on inference and background in ways that are not necessarily coded ‘emically’ in any grammar, and its patterns are across situations of use and means of achieving ends, not across coding categories. The consequence of taking

154

Michael Fortescue

the overarching nature of pragmatics seriously must, it seems to me, result in a process interpretation of language that does not set discourse and grammar side by side on the same ontological level, but rather treats grammar as abstract pattern or template at the service of communicative purposes. Let us avoid formalizing the patterns of pragmatics as if they were merely an extension of the patterning of grammar. This amounts to having one’s cake and eating it as regards the choice between incorporating a discourse ‘level’ into the existing FG model (in the manner of Hengeveld 1997) or interfacing it with a separate ‘discourse grammar’ (in the manner of Hannay 1991). It is in fact compatible with Hengeveld’s more recent proposal (this volume) for combining the two approaches to discourse within the FG model.5 In this chapter I shall be extending the notion of letting the highest (discourse-pragmatic) level of a processual interpretation of the FG model ‘dip down’ to the lowest, predicate level, as proposed for predicate formation rules in Koyukon and West Greenlandic (Fortescue 1992: 131). Languages vary on this dimension, some allowing whole propositions as input to certain derivational rules, others strictly limiting input to stem predicates (as English, by and large, does).6 In the following section I shall specifically look at relevant phenomena in Nootka. My representations will focus on the level of the (extended) predication (= Whiteheadian proposition, as ingredient in the ‘judgment’, which corresponds more or less to Dik’s ‘proposition’), since that is where I claim the link to psychological reality must be made. The ‘missing’ relationship between this level and the lexicon (with its reflection of SoAs) will be returned to in the final section.

2. Choice of predicate in Nootka In Nootka, as in the other Wakashan and neighbouring Salishan languages of western Canada, the distinction between lexical nouns and verbs is notoriously slippery. Moreover, the choice of constituent that is treated as (main) sentence predicate often goes contrary to European expectations. What corresponds to an adverbial or a quantifier, for instance, might be chosen as predicate, followed by a bare (structurally indeterminate) verbal or nominal expression whose relation to the predicate – e.g. as subject – is very loose (and must be inferred). Thus the following sentences appear in Mithun (1999: 188–189, citing Nakayama 1997):

Process and pattern interpretations (1)

(2)

(3)

(4)

hu:ak-a-qu:s mamu:k early-TEL7-CONDIT.1SG work ‘I would work early’ a:n-u:a-qu:-č hauk only-do.together.with-TEL-CONDIT-INFER eat ’He used to eat alone with his mother’ mu:-a-qu: maayi hi: four-TEL-CONDIT.3 family there.in.house ’There used to be four families living in this house’ sačica-am-it-iš-a wai:q unceasingly-CAUS-PAST-IND.3-PL homesick ‘They constantly got homesick’

155

umi:qs-ak mother-POSS

Swadesh (1936: 84) gives a further example with a locative predicate (from a dialect with a different indicative paradigm): (5)

hi-ap-’a-’at-ma there-CAUS-TEL-SHIFT-INDIC ’He is allowed to sleep there then’

waič sleep

It is important to realize that although the constituent that is chosen as predicate in such sentences is somehow focused or emphasized, one should not assume that we are dealing here with Focus assignment at the pragmatic-function level of the traditional FG sort (although the etic functions concerned do of course overlap). The ‘focused’ element is rather what the speaker in a general way considers to be ‘newsworthy’ in the discourse context (this reflects the strategy of ‘most newsworthy first’ found in many North American languages with pragmatically controlled word order – cf. Mithun 1987). Note that the language does have a separate contrastive focus construction when focus is on a term, using predicates uħ (for subject) or u:kwi (for object), as in:8 (6)

a. qaħsa:p uħ John muwac kill do.it John deer ‘(It was) John (who) killed the deer’ b. qaħsa:p-’a kwatya:t u:kwi kill-TEL Kwatyaat do.it.to ‘Kwatyaat killed the wolf’

qwayac’i:k wolf

156

Michael Fortescue

The phenomenon in (1) to (5) is part of a structural phenomenon which Nakayama calls ‘serialization’, which combines multiple clauses into units expressing a single State of Affairs (actually some of the example sentences strictly speaking represent other parallel clausal relationships such as ‘modification’, which do not allow word-order variation according to pragmatic factors in the way serialization as Nakayama defines it does). This is an event-encoding strategy for combining predicates into informationally dense predications – dense in terms of participant structure or event specification – which cannot be expressed in a single word (Nakayama 1997: 119ff.).9 The shared inflectional suffixes (none of them actually obligatory) are generally gathered on the first constituent, whatever it is. The non-initial element can take a more limited array of suffixes (mainly for aspect, as in (7) below). Serialization proper serves a number of discourse purposes besides that of the kind of construction illustrated above. Sentences such as (2) do not really bear any special focus according to Nakayama (pers. comm.), since there is no flexibility of ordering choice with the particular predicate involved. Similarly, the order is obligatory with negative and interrogative words such as the following, where there is no special focus – or rather only ‘default’ focus on the initial constituents: (7)

(8)

(9)

wik’-a hauk-ši not-TEL eat-MOM ‘He did not eat’ aqish-w’it’as-a-it-ħ suw’a inm’i:-či why-about.to-TEL-PAST-INTERR you snail-INCHO ‘Why would you become a snail?’ wa:-yaq-pi-mit-ħ awa: town which-most-PAST-3SG.INTER near town ‘Which was the closest town?’

In (9) note that even the modifying suffixes that belong semantically with the adjectival predicate awa: has been positioned on the interrogative constituent appearing as predicate of the whole sentence. In the following the adverbial predicate glossed ‘very’ can be said to be well-suited as focus owing to its inherent meaning: (10)

i:ħ

kwi:s-ħi very strange-DUR ‘It is very strange’

Process and pattern interpretations

157

Note that it can still remain as main sentence predicate when converted semantically to a transitive verbal expression by addition of the ‘inverse’ (‘shifter’) suffix -’at. This indicates action on an undergoer higher on the animacy scale than the subject (here 3SG outranks 3PL).10 A rough gloss of (11) would be ‘they did it very much to him, being strange’, with being strange less central to the construction than doing very much. (11)

i:ħ-’at

kwi:s-ħi very-SHIFT strange-DUR ‘People treated him very strangely’

A parallel phenomenon is found within the nominal phrase, as in the following, where the only inflection present (the definite article) is attached to the modifier constituent (which must come first), even though it goes semantically with the head: (12)

camaħta-i qwayac’i:k real-DEF wolf ‘the real wolves’

There is clearly no marked focus on the modifying constituent here. To return to the verbal sentence examples, the next question is: if the initial element of serialized construction is the (main) predicate, how is the logical subject of these sentences expressed? There are several possibilities. It may be an overt argument later in the sentence, or some referent inferred from context, or it may be a whole verbal expression understood as ‘nominalized’ (as in the first sentence). As can be seen, there is no direct marking of what the subject of the sentence is – with the exception of the few examples of pronominal inflection. Compare also the following examples of the clausal possessor-incorporating construction, whereby the possession relationship is marked by a special affix on the main sentence predicate immediately preceding an incorporated pronominal marker of the possessor: (13)

(14)

m’aw’a:-a-’at-ukw-ina hiyiqtup deliver-TEL-SHIFT-POSS-1PL things ‘He brought our things’ (or ‘We had our things brought by someone’) siy’aq y’ap’ic-uk-kw-is blue-DUR-POSS-1SG I (SUBJ) ‘Mine is blue’

158

Michael Fortescue

Note that the independent pronoun in (14) is in its subject form, not its possessor form: the first-person possessor affixed to the first word has subject properties, as Nakayama puts it, and the possession marker itself produces a verbal expression. In such constructions the possessor is topic, i.e. salient in the discourse, and treated as a grammatical argument of the predicate expression. Nakayama argues elsewhere that the category of syntactic ‘subject’ (and ‘object’), as opposed to semantic ‘agent’ (and ‘undergoer’), is in general difficult to maintain for this language (Nakayama 1997: 99ff.).

3. Nootka phrasal affixes and zero-verbalization Now it could be argued that the ‘suffixes’ attached obligatorily to the initial constituent in the serialization type of construction are enclitic and that that constituent is not ‘really’ the predicate of the sentence at all. But Nakayama argues that this is not the case, for whether the suffixes concerned are inflectional or derivational they have properties that are un-clitic-like. Thus in Nakayama (1994: 266) he compares the following sentences, all containing ‘phrasal affix’ –i:  ‘make’: (15)

(16)

(17)

ča:pac-i: canoe-make ‘He made a canoe’ u-i: čapac nice-make canoe ‘He made a nice canoe’ mu:kw-i: u čapac four-make nice canoe ‘He made four nice canoes’

As he argues, the ‘verbalizing’ suffix here takes a whole phrase as its base (though it must morphosyntactically stand on the first word of the phrase, whatever that is). Its scope properties are directly comparable to those of such West Greenlandic affixes as -liur- ‘make’, which can also be added to an NP leaving the modifying element stranded (in the instrumental case).11 Whether one can jump from this to the claim that these suffixes are not enclitics at all is perhaps a matter of definition: they are at least not prototypical enclitics, being lexically much weightier than clitic elements usually are.

Process and pattern interpretations

159

There is another kind of lexical suffix in Nootka than the illustrated ‘governing’ type, namely the ‘restrictive’ type as in (18), which has no parallel in West Greenlandic. There are a great many suffixes of this type for expressing location and instrument, etc., in Nootka, as in most of the languages of the surrounding Northwest American linguistic area: (18)

quac-’aq person-inside ‘There is a person inside’

Here the topic is person (tracked as such in following discourse), as opposed to canoe in (15) for example. The verbal/nominal/adverbial status of the suffix is lexically quite indeterminate – it is only in the context of a predication, such as this sentence, that it is clear that its function is to modify (for location) the left-most constituent understood as a zero-verbalized nominal, i.e. in an existential context. It only differs from other predicates in being an obligatorily bound form. This leads us back to the question of the status of nominal versus verbal expressions in Nootka in general, for as Nakayama puts it, although it is not easy to differentiate between isolated verbs and nouns, it is generally clear what element is functioning as a verbal or a nominal, either because a verbal morpheme functioning as a nominal or a nominal one functioning as a (sentence) predicate is marked as such by affixation or because the discourse context renders it apparent (when, for instance, serialization is involved). In the following sentences (from Swadesh 1936, phonemicized according to Nakayama) the verbal affix on the first word and the nominal one on the second make this quite clear, whichever of them is chosen as sentence predicate:12 (19)

(20)

mamu:k-ma qu:as-i work-INDIC man-DEF ‘The man is working’ qu:as-ma mamu:k-i man-INDIC work-DEF ‘The working one is a man’

The situation is somewhat obscured by the fact that sentential affixes (inflections) are not obligatory, unlike, say, the inflections of West Greenlandic. Nevertheless, a lexical item referring to an entity (a prototypical ‘nominal’) cannot function as a verb except in copular sentences of

160

Michael Fortescue

identification and class-assignment or the like (as above), i.e. where one alternatively could talk of a zero-marked derivational copula (something actually quite widespread in the languages of the world; see Fortescue 1998: 62ff.). Likewise, a lexical item referring to an action or state (a prototypical ‘verbal’ morpheme) generally has to be nominalized in order to function as a nominal (e.g. by addition of the definite article, as above).

4. Word classes and the relationship between predication and lexicon in FG The question I want to turn to now is this: how are the facts about Nootka sketched above to be fitted into the FG model? First, the choice of the ‘most important/newsworthy’ element that is positioned first in the typical clause (as in sentences 1–5) could in theory be made at the level of pragmatic function assignment (wherever that may be). On the other hand, Nootka ‘phrasal’ suffixes (as in 15–17), like semantically similar West Greenlandic ones, can be treated by individual predicate formation rules, and its ‘zero copula’ construction (as in 20) by term-predicate formation rules in the manner of Dik (1989: 172) or, preferably, by non-verbal predicate frames in the manner of Hengeveld (1992: 32). The lack of evidence for the assignment of syntactic Subject status to particular arguments (as mentioned in connection with sentences 13–14) would seem to be accounted for by aligning Nootka with those numerous languages where FG recognizes that Subject assignment simply does not occur. Furthermore, Dik’s fundamental principle of treating both nominals and verbals as underlying predicates would seem to be amply justified for this language, where all lexical content morphemes are indeed predicates in the sense that they can function – given an appropriate morphosyntactic ‘slot’ – as verbals or nominals or modifiers of either. However, with the standard version of FG, designed principally for non-polysynthetic languages, there are problems with all of these suggestions.13 Specifically as regards clause-initial ‘focused’ constituents, it appears that the choice of first constituent is not simply a matter of Focus in Dik’s original pragmatic function sense. The constituent chosen is either one displaying general ‘newsworthiness’ (in the discourse context) or one required by the lexical items and/or construction involved (as in 7–9). There has been much discussion within FG in recent times as to whether Focus (and other discourse-related functions) really belong in the model here at all or should be part of a separate discourse grammar or module (as discussed

Process and pattern interpretations

161

briefly in Section 1 above). Nootka is a language in which complex discourse-pragmatic factors affect the basic syntactic choices that modulate the linear succession of constituents, and not all of this complexity can be fitted into the simple binary choices of FG pragmatic function assignment. In part this is a matter of the interface between lexicon and basic predication. ‘Newsworthiness’ is a complex matter of weighting various contextual and lexical factors, not one of simple ‘digital’ choices. In other words, it is hard to argue here for a discrete pragmatic function assignment level of the classical FG kind.14 The problem with the lack of Subject assignment, on the other hand, is that Nootka does have a morphologically marked inverse construction (extending also to impersonal subject constructions, as in 10), which is in part a matter of syntactic function level ‘perspectivization’ but is in part also determined (and quite automatically so) by semantic animacy – similar to the situation with the inverse construction in Athabaskan languages. The ‘emic’ (coded) grammatical categories of Nootka appear to be exceptionally context- (and semantics-) dependent. As regards parts of speech, Nakayama distinguishes, as we have seen, only verbal and nominal expressions (apart from uninflected particles), both of which are possible as arguments in serialized/complement syntactic configurations or as modifiers. Basic clause structure is, as Nakayama puts it, imposed top-down (Nakayama 1997: 78ff.). It is the basic propositional articulation of the speech chain that holds it all together structurally. Here is a language with maximal pragmatic control of syntax – much inference from context is needed to interpret its syntactic cohesion, though its pragmatic articulation is transparent. It is a language in which ‘highlevel’ illocutionary or pragmatic factors may reach down all the way to affect the choice between predicate and argument status. There are very few structural clues to signal this: no inflectional categories are obligatory, and if these are overtly expressed they are typically attached to the first words of a phrase, whatever its function, with a multiplicity of functions and predicate frames being associated with individual lexical morphemes. In sentences in which constituents expressing time, negation, or other ‘peripheral’ adverbial elements are chosen as the main clause predicate, as in (2), context-dependent emphasis of some sort is generally involved – although this is not always strictly a matter of Focus. It may rather be a matter of default, ‘unmarked’ focusing (e.g. with the negative ‘verb’, which must come first). The first element chosen in the serialization construction, where there is a choice, is simply the ‘most newsworthy’ in the broader discourse context.

162

Michael Fortescue

As regards ‘phrasal suffixes’, the situation is much the same as I have described for West Greenlandic (Fortescue 1992: 131), namely that if this is a matter of derivation in the Fund (a perfectly reasonable assumption), then such a derivational process must be able to ‘dip down’ (or rather ‘up’, depending on the kind of diagramming of the FG model one has in mind) into layers of clause structure where not just extended predications but even whole propositions have been elaborated as input for the derivational process concerned. The bound derivational suffixes of Nootka and West Greenlandic need in any case to be treated independently of the predicate/term formation processes into which they might enter and belong in the lexicon with their own predicate frames. As the examples from Japanese and English in note 5 show, this cross-level recursivity of derivation is not limited to polysynthetic languages. Finally, as regards the ‘zero copula’ construction, if one chooses to follow Dik’s approach to copular constructions, a conversion of a lexical ‘term’ to a predicate expression is assumed, but if all lexical items are ambiguous as to term or verbal status in Nootka, there is no need for such a conversion at all – unless one first produces a term by term formation and then converts it back to a predicate by term-predicate formation, a seemingly unmotivated extra complication for this type of language. If, on the other hand, one follows Hengeveld’s approach mentioned above, one avoids having to postulate ad hoc predicate formation rules, and can treat non-verbal predicates in terms of predicate variables (f1), just like verbal ones. However, there is very little evidence in Nootka for distinct predicate frames (with predetermined argument sets) being associated with particular non-verbal predicate construction types any more than with verbal predicates, so it is hard to tell if this is an advantage or not. I shall return to this matter below. In sum, the nub of the problem is the relationship between the lexicon – with its predicate frames corresponding to discrete SoAs – and the rest of the model (in particular the choice of elements in the basic predication). A given lexical morpheme (or “predicate”) in Nootka may enter into various constructional “frames”, with the valency even of a specific verbal frame being rather vague compared to the situation in more familiar language types. In Nootka, lexical morphemes are not anchored in the lexicon in quite the same way as in more familiar languages: it would appear to be a typological fact one must live with (and FG claims typological adequacy) that some languages have a much closer correspondence between States of Affairs (SoAs) and lexical items than do others (compare Fortescue 1992 for another polysynthetic language – Koyukon – which does, unlike

Process and pattern interpretations

163

Nootka, have a very close correspondence between the two). Perhaps the mistake is to take the SoAs of the original model to be truly universal. An attempt to incorporate a typology of languages with varying degrees of rigidity of word-class distinctions within FG has been made by Hengeveld (1992: 45ff.). At first sight Nootka would appear to go with languages like Tongan, representing the highly ‘flexible’ type, in which predicates can be used in practically any function. This contrasts with highly ‘rigid’ languages like Cayuga, which may have only one part of speech, the predicate, but where this must be morphologically derived to function as other parts of speech than the verb. In fact Hengeveld’s extremes meet, since in ‘rigid’ languages any one morpheme may be extremely flexible as regards morphosyntactic context, and the real contrast is with ‘specialized’ languages like English with a full array of morphosyntactically distinct parts of speech. The problem with fitting Nootka into this typology is that the distinct morphological marking of predicate words in the one or the other sentential or phrasal function is not obligatory, so that it partakes of both the Tongan and the Cayugan type depending on whether there is suffixation present or not (and, as we have seen, suffixes tend to be shifted to the first word of constructions, whatever it is). Hengeveld (pers. comm.) suggests that Nootka might represent a ‘rigid’ language in which the relevant morphology has relatively recently lost its obligatory status. It seems then that some languages, like Nootka, are simply more controlled by discourse pragmatics than others which are more lexically controlled and in which there is a closer link to SoAs.15 Nootka will, for instance, split what in other languages correspond to unitary predications into two clauses (of varying degrees of tightness) to emphasize the salience of the undergoer argument. If this is so, we must look for another place in the model (or an extension of it) where such matters as discoursedetermined choice of initial predicate can be accommodated. Short of putting virtually all of the morphosyntax of Nootka within a ‘discourse module’ (on the nature and categories of which there is no consensus within FG today), the only obvious possibility would seem to be to introduce discourse factors determining the choice of main clause predicate at the lowest level of clause structure (e.g. by a feature ‘most newsworthy’), obviously a severe, and undesirable, disruption of the model. This can be overcome, I would suggest, by clearly distinguishing between a ‘process’ and a ‘pattern’ interpretation of the FG model. It is the former that must be reassessed in the light of the kind of evidence I have presented. Let me elaborate the point.

164

Michael Fortescue

Nootka understood as ‘pattern’ (or grammatical ‘template’) presents ample support for Dik’s treatment of both nominals and verbals as underlying predicates.16 On the other hand, the question of which element to treat as main clausal predicate and which as argument (and of how many clauses to split up one proposition into) is hardly relevant to the pattern perspective at all if, as I have shown, this is determined by higher discourse-pragmatic factors, which must ‘reach down’ from the topmost level. It would only be relevant if there were a more or less direct link between the State of Affairs to be expressed and a particular predicate frame corresponding to it in the lexicon (built up usually around a verbal core), but this link is precisely what is extremely tenuous in Nootka, where the main predicate could just as well be an adverbial or a quantifying expression, for example. It is difficult to see how the pattern perspective could accommodate ‘percolation’ down from the level of illocution (or beyond) in the way that Hannay (1991) has suggested, since this would have to extend right down into the lexicon itself – besides which, ‘percolation’ is surely a process and not part of a pattern description at all. The nature of the link between clause/predication structure and discourse articulation is also much less direct in Nootka than in English, where the clause can accommodate several arguments and satellites of single events. By contrast, a unitary predication may in Nootka be broken into several clauses, e.g. with one salient referent in each. The Nootka clause is structurally very simple and less flexible than in English as regards meeting more elaborate discourse needs than just indicating ‘newsworthiness’. It makes sense to describe the pattern grammar of Nootka as being without syntactic – and indeed with very limited pragmatic – function assignment. (This is something the FG model already allows for.) Thus Focus in the general sense of ‘newsworthiness’ could in theory be left right out of the grammar as an etic, not an emic category, although in the serialization construction, where it triggers an actual ordering choice, it had better be retained.17 The morphological ‘inverse’ can in turn be treated as a matter of predicate formation in the Fund rather than of syntactic function assignment, just like causative (and other morphological) valency-shifting derivation processes. In other words, all that has to go from the traditional pattern model is the direct link between lexical items and specific SoAs. From the ‘process’ perspective, on the other hand, one can indeed envisage higher-level pragmatic choices (at the level of the illocution or above) percolating all the way down to the basic predicate level and affecting the choice of (main) predicate. Compare the following two representations of sentence (1) above (I would work early, with early cho-

Process and pattern interpretations

165

sen as predicate); (21a) is formulated within standard static FG grammar (with a traditional Foc introduced at the pragmatic function assignment level); (21b) is a representation of actual processes leading to the expression concerned on the basis of a unitary (though complex) Whiteheadian prehension: (21)

a. DECL E1: [X1: [HABIT e1: mamu:k (d1x1 :1p(x1))Ag (hu:ak)TimeFoc (e1)] (X1 )](E1 ) b. DECL ((hu:ak)Foc? (HABIT mamu:k (1SG)Act))

The representation in (21b) should be understood as being embedded in a specific discourse-pragmatic background which articulates the embedded proposition into a focal part (in the broadest sense) and a presupposed part (i.e. as the source of the focal marking on hu:ak ‘early’). This corresponds to an outer Whiteheadian prehension of the concept ‘early’ applied to (i.e. qualifying) an inner proposition consisting in turn of an ‘indicative’ prehension of the subject I and a ‘conceptual’ prehension of the predicate work. The articulation of the predicative structure reflects the higher (or rather outer) level choice: the two form an indivisible unitary actual occasion of language use and cannot be separated in either time or space. The sum, when coupled to an appropriate illocutionary intention (here simplified to ‘DECL’), corresponds to a Whiteheadian judgment. Note that the features of the discourse-level trigger which cumulatively assign Focus are not ‘emic’ elements of any finite system of contrasts but simply a collection of relevant presuppositions and communicative intentions, whose weighted effect is to incite the choice of early as main clausal predicate (the question mark on the feature is meant to leave open the question whether it is the same Foc as in 21a). I assume that also the ‘telic’ affix -a is introduced by percolation from the outer discourse level, functioning as it does to indicate that the predication as expressed does indeed represent one unitary clause (the speech chain is thus modulated to reflect the propositional structure). This affix would be difficult to introduce at all in the static FG model (as in 21a), since its packaging function is purely a matter of discourse management. In contrast to (21a), representation (21b) reflects the selection of early as the basic predicate right from the start. In (21a) this choice would have to be suspended until the operation of higher levels of the model closer to the word-ordering expression rules – up to that point work would have to be regarded as the basic predicate selected to correspond to the given SoA

166

Michael Fortescue

(with the adverbial satellite, or predicate restrictor if one follows Hengeveld’s approach, being focused later). This can hardly be called psychologically plausible. Recall what was suggested above as regards two-tiered thetic propositions: these too need to be indicated as such at the initial phase of a process representation.18 Note also that term variables have been eradicated from (21b). Moreover, the semantic role Actor has replaced Ag, since these roles are, according to Nakayama (1997: 90ff.), restricted to Actor and Undergoer in Nootka. The role Actor is required by the predicative construction (an intransitive declarative clause) rather than being closely tied to the lexical predicate itself. The Actor is no doubt the default argument associated with mamu:k (the default function of which is in turn verbal), but other arguments, e.g. of place, could optionally have been indicated instead of the Actor (nominal arguments are not usually marked morphologically for semantic or syntactic role). As to how the Nootka sentences we have looked at relate to Hannay’s specific ‘message management’ types, one could say that (1–5) and (19– 20) at least would correspond to his REACTION mode (presupposing a discourse context that requires Focus on the first non-term element of the utterance). (6a) and (6b) would presumably also be in that mode, but with contrastive Focus on a core term (one of the few constructions in Nootka where Focus does not lie on the initial constituent). (7–9) could be seen as reflecting the same mode, but perhaps NEUTRAL is more fitting, since the initial position of the element concerned is lexically determined, there being no choice as to what is regarded as the sentential predicate; note that Focus here has to be indicated in the lexicon as adhering to these words, since they automatically trigger initial positioning. (10–11) and (13–14) probably do reflect a choice and thus go with (1–5), but (15–17) are again probably NEUTRAL, since the positioning of modifiers before heads is obligatory. (12) and (18), finally, are not relevant here. Is this very helpful? Certainly not as useful as for English, since Focus in the broad sense of ‘newsworthiness’ is so strongly associated with initial P1 position in Nootka that it virtually always adheres to that position (except in the Contrastive Focus construction), and there is virtually nothing to choose between – only what particular item in the message is to fill it.19 In general, it looks as if the structure of Nootka is so strongly determined by discourse-pragmatic factors that there is little to be found that parallels the complex interaction between such ‘triggers’ and the often conflicting requirements of sentence-internal syntax displayed by English.

Process and pattern interpretations

167

5. A minimalistic functionalism? The question remaining, then, is whether the two perspectives compared in the previous section are compatible. I would claim that they are actually not, either as a unitary model embracing both, or as reflecting two separate modules on the same ‘ontological’ plane. The relationship is – and must remain – one of complementarity. This is in part because the variable-plusrestrictor (‘predicate calculus’) format of the pattern grammar has no direct process correlate – at least not one involving experientially interpretable meaning. I have proposed equating the processes involved with Whiteheadian prehensions. These are situated within real contexts, in a world of entities rather than variables and of complex embedded communicational purposes rather than discrete digital choices of ‘function’. More importantly, ‘emics’ cannot be matched directly to ‘etics’. Purposes are the domain of the latter; the former define the various means at hand to express them. Given the typological distinction that this chapter has illustrated between languages with a tighter or looser linkage of lexicon and grammar (and between languages whose grammar reflects greater or lesser discourse-pragmatic vs. lexical control of grammar), one surely cannot expect there to be a universally valid division of labour between a grammar module and a discourse module. What we need is not so much a module or a grammar of discourse, but a set of (rational) principles allowing us to link specific communicational intentions to the means provided by the abstract categories of grammar (cf. Itkonen 1983: 177). Perhaps Hengeveld’s proposed reformulation of the basic architecture of FG can be understood in just these terms. Presumably the idiosyncracy of Nootka, from this perspective, has to do with the relative transparency or fluidity of the expression level in this language, which matches interpersonal level choices much more directly (e.g. ascriptive act decisions as to what is to be the main predicate) than in more familiar European languages. However the FG model is extended, reformulated or reinterpreted in the future, we can at least require of it that it should be able to account for languages of the Nootkan type in which morphosyntactic indication of function is rather minimal and where inference from context is constantly required.20 If FG is to accommodate signed languages within its overall framework the same could be said – the pragmatically determined fluidity with which these articulate the flow of communication into discrete information packets is indeed reminiscent of Nootka. What in effect I have been arguing for in this chapter is a kind of functionalist minimalism, parallel to the Minimalist Program of the generativists

168

Michael Fortescue

(Chomsky 1993). Having shaken off the ‘superfluous’ baggage of transformations and the deep-versus-surface structure distinction and thus returned towards the central nature of the linguistic sign as binding form and content, minimalists continue their search for ever more economic generalizations as to the universals of language patterning. Which is fine, as long as neither they nor we persist in the mistake of thinking this has anything much to do with real psychological processes. We, as functionalists, however, should not be complacent in our belief that we, by default, as it were, must be the guardians of the ‘true’ key to Process. If we are to do so, we must first strip our process view of language of all its hidden pattern trimmings, just as some generativists (though perhaps not Chomsky himself) have been willing to abandon the process, or strategy, interpretation of their evolving models. As regards the process interpretation of the FG model in particular, I propose that we limit our representations of sentence meaning to simple layered structures that take experiential ‘prehensions’ as the baseline.21 As regards the specific question I started out with, how to handle Focus in Nootka and where in general it should be positioned in the FG model (on some higher pragmatic/illocutionary level or in a separate interfacing discourse module), I can only propose that Focus should be part of both the pattern and the process representations of utterances in Nootka, but that this does not necessarily hold for all languages. (Nor, as mentioned above, is Focus in the process sense of cumulative etic triggers necessarily equivalent to the corresponding pattern sense of emic coding choice.) In some languages, like Yukagir, Focus is highly grammaticalized within the patterning of its morphosyntax, and in others it is not grammaticalized at all, focality being borne solely by the ‘uncoded’ or analogue ‘etics’ of its prosody (Nootka is somewhere in between in this respect). As to where in the model Focus should be introduced, it looks as if it needs to be admitted on process representations at the very outset, i.e. at the outset of ‘concrescences’ that produce determinate utterances from undifferentiated, holistic ‘intentions’ and work towards greater and greater differentiation en route towards their final ‘satisfaction’ (= matching of intention and output). Using the standard formalism (that of Dik 1989), this actually means locating Focus at the lowest predicational level (albeit percolated down from the highest, pragmatic level), since choices there may have an immediate influence on choice of structure at ensuing levels. It should be borne in mind that process representations are by necessity language-specific. As regards pattern interpretations or representations (i.e. in the traditional FG model), it really does not matter very much, since ‘everything

Process and pattern interpretations

169

comes out in the wash’ in the expression rules anyway, as long as the relevant functions are marked somewhere. In Hengeveld’s revised model one would presumably start with Focus choices made at the interpersonal level (where the factors affecting its assignment are obviously ‘at home’, with direct access to the communicative context), but it is also possible that different kinds of Focus are relevant to distinct levels. It may turn out that both contrastive constituent focus and the kind of ‘newsworthiness’ focus leading to choice of first constituent in Nootka are determined from the outset at the ‘interpersonal level’, whereas the actual articulation of the utterance into successive predicate and referential phrases realizing these choices must be accomplished (as process proceeds generally ‘from left to right’) by reference to pattern possibilities belonging to the other two levels (for example activating the contrastive focus template or summoning and positioning the ‘telic’ affix -a). This may be a disappointing conclusion for those who would like a definitive answer to the question of ‘where pragmatics belongs’ in the FG model. But perhaps we should not be so fixated upon the interrelation between grammar and discourse after all, but think rather in terms of Pattern versus Process, a distinction that cuts across this divide, since Pattern includes the more ritualized/grammaticalized aspects of pragmatics as well as core clausal grammar. It is not yet clear how exactly Hengeveld’s new ‘architecture’ is to be understood in this respect. One possibility for accommodating both perspectives in one model (which appears to be Hengeveld’s intention) would be, as hinted at above, to interpret it as pure Pattern on the vertical axis, but as Process on the horizontal axis (‘left-toright’ as he puts it). This at least avoids the problem of psychologically suspect ‘percolation’ processes – as long as all relevant triggering factors are potentially there ‘in the top left-hand corner’ from the start.22 Whatever the way forward, I suggest that the goal must be to bring the Pattern and Process perspectives on FG into line in a relation of complementarity which does not blur over the essential distinction between them. There is still much mileage to be derived from applying standard FG to the abstraction of generalizations across linguistic pattern at all levels of its layered template (including grammaticalized higher-level ones), thus achieving still greater typological adequacy. Elaborating a specifically Process variety of FG does not, it seems to me, require further theoretical or formal sophistication – on the contrary, we need to simplify what we already have in the Pattern model. Meanwhile, let us at least avoid ontologically suspect hybrid models. When talking about Pattern let us do just that, but when talking about Process let us make it quite clear that we

170

Michael Fortescue

are talking about something else. This does not preclude eventually elaborating a single model that can be interpreted both ways. What we do need is clear principles for (and constraints on) how to interpret such a model in the one or the other way.

Notes 1.

2.

3.

4.

Cf. Dik (1989: 13) for the possibility of interpreting FG as a production model that achieves psychological adequacy and which is said to “lay out recipes for construing linguistic expressions from their basic building blocks”. Grammar he sees from this point of view as tripartite (production and comprehension plus a common store of elements and principles). In this context he approved (1989: 13, fn. 9) of Nuyts’s proposal to give a procedural interpretation to FG, although his own approach was different, being closer to the predicational representations of standard FG. Thus Dik (1988) suggested that the representations used by a hypothetical logic component could be the same as those of the underlying propositional structure of sentences, whereas Nuyts (1992) argues that deeper cognitive representations are needed that cover both (see Anstey this volume: 38ff). Harder’s ‘process vs. product’ dichotomy is not quite the same as my ‘process vs. pattern’ one. For example, his representation of the phrase the old elephant as: (1 (prop: old (ent: elephant))) is to be understood as a set of nested instructions to produce the phrase in question, which would then be the static ‘product’ of the procedural representation. By ‘pattern’ I mean the grammar as such, the relational template of coding choices which constrains the processes of expression in a given language. This is not a ‘product’ (except in the historical sense), but rather a high-level abstraction across norms of (communicative) behaviour. See in particular Bolkestein (1998) for discussion of the question of the level of the revised FG model at which the pragmatic function of Focus should be assigned. Hannay proposes several basic types of utterance ‘mode’ at this level, which he labels TOPIC, ALL NEW, REACTION, NEUTRAL and PRESENTATIVE. The last-mentioned, for example, corresponds to ‘thetic’ as opposed to ‘categorial’ judgments typically expressed by special constructions for introducing a NewTop, whereas REACTION mode in English triggers Focus in P1, and TOPIC mode Topic in P1. In general, the choice of mode determines what will fill the P1 slot (in English). So for Hannay the initial choice in a processual interpretation of FG concerns which entity – or other semantic element – is to go in initial P1 position in the accruing utterance. As Mackenzie interprets this (Mackenzie 1996: 144) these modes may be

Process and pattern interpretations

5.

6.

represented as operators on the level of the illocution, whose effect percolates through the hierarchy of levels, constraining the assignment of pragmatic functions to elements of the utterance. Nuyts (1992: 281), too, assumes that the first thing that has to be decided when mapping an underlying cognitive ‘Singular State of Affairs’ onto an FG clause structure is the choice of the basic predicate (and its terms). The factors he sees behind this choice include higher-level matters of illocution, politeness and attitude, etc. (as well as the availability in the lexicon of the language of predicate frames corresponding to that SoA). However, he does not elaborate on how, within his own cognitive interpretation of FG, such complex choices are actually to be implemented. More specifically, Hengeveld’s reformulated ‘interpersonal level’ can easily be interpreted in terms of Whiteheadian complex subjective aims (e.g. illocutions embedded within discourse moves), and propositional prehensions, which combine indicative prehensions (Hengeveld’s ‘referential acts’) and predicative conceptual prehensions (Hengeveld’s ‘ascriptive acts’). The downward layering at this level is what is essentially new in Hengeveld’s reformulation. This allows for the alignment of an ascriptive act at the interpersonal level with an entity of any order at the representational level below it, not just a zero-order predicate. In the Whiteheadian framework of ‘eternal objects’ (here ‘predicates’) and ‘nexūs’ (here read ‘entity types’) this is to be welcomed, since the same eternal objects may be involved in nexūs of various degrees of abstraction. The left-to-right restrictions on decisions in production at the interpersonal level that Hengeveld calls for are in fact largely provided for by the Whiteheadian theory of the successive stages of the ‘concrescence’, as applied to linguistic utterances (cf. Fortescue 2001: Appendix 2). The phenomenon is by no means limited to polysynthetic languages like Eskimo, and can be illustrated for Japanese, for example, as in the following structure from Shibatani (1990: 248), where the affixed nominal tyuu ‘middle of’ is attached to an entire clause: i.

7.

171

[Yamada-san ga tyuukosya o hanbai]-tyuu Mr. Yamada NOM used.car ACC sell ′in the middle of Mr. Yamada’s selling used cars’

Compare even in English the man I saw at the station yesterday’s face, with wide phrasal scope of genitive -s. ‘TEL(IC)’ is a ubiquitous affix that, according to Jacobsen (1993: 250), signals ‘finitehood’ of a predicate, i.e. its relative independence as a clause, not just the second part of a serializing construction, which in Jacobsen’s Role and Reference terms is a ‘nuclear cosubordination’ construction in

172

8.

Michael Fortescue which the second predicate shares any clausal mood marking of the preceding one. Nakayama prefers not to use the term ‘finite’, since it smacks of morphosyntactic finiteness of form (pers. comm.). He describes it as “making the event feel more immediate and punctuated” and points out that it plays an important role in discourse organization. Mithun glosses it as ‘momentaneous’, and Swadesh as ‘now, at that time’. In general, I have used Nakayama’s glosses in the examples. Jacobsen’s characterization of the affix’s principal function seems to me convincing, as distinguishing clausal from nuclear cosubordination in the case of ambiguous ‘absolutive’ (uninflected) forms of predicates with no marking of illocutionary mood (or subordination or modality) following initial predicates that may well show such marking. Note that the CONDIT(IONAL) mood is used here to mark habitual action. But Focus in the sense of non-contrastive ‘newsworthiness’, as in the sentences above, can also be assigned to (at least) object terms, in which case they can be put in initial position, without these special relational predicates, as in:

i.

9.

c’aak t’a:ps-at-a-qu: river dive-sinking.into(water)-TEL-3.COND.INFER ‘He would dive into the river every once in a while’

A similar distinction between contrastive stress and merely ‘most newsworthy’ emphasis on a given entity referred to can be observed in most languages. In English, for example, one would use respectively either a cleft construction (or simply strong stress on a constituent) and (other) prosodic means of emphasis. In Danish Sign Language the former is also expressed either by emphasis (an emphatic production of a sign) or by something like a cleft construction, as opposed to raised eyebrows marking a referent (or any other segment of an utterance) as the most newsworthy part (EngbergPedersen 1991: 64). Despite the general polysynthetic character of this language, Nootka wordforms can only accommodate a single verbal stem, typically marking no more than one participant, so that transitive clauses are often split up into two parts, one indicating the subject’s action, another involving the preposition-like transitive focusing predicate ‘do with respect to’ mentioned above to indicate the object. The study of cross-linguistic variation in the density of packaging of information in single clause units is still in its infancy (although it is being broached in various ways, e.g. in terms of alternative event structure construals – by cognitive linguists like Langacker – and of the distinction between ‘verb and satellite framed’ languages – by psycholinguists like Slobin – and in terms of cosubordinate clause chaining within Role and Reference Grammar). Some languages containing ‘serial verb’

Process and pattern interpretations

10.

11.

constructions (like Chinese and Japanese) seem to fall somewhere between English and Nootka as regards constraints on how much information can be packaged in a single clause. The issue is complicated by the further questions of how (and how much) new information is distributed in successive clauses (for example, one such unit per clause, as Chafe 1994: 109ff. suggests) and whether semantically obligatory serial verb constructions in Papuan languages, for instance, really represent more than one predicate per clause (as opposed to a single complex one). Typically of Nootka, neither 3rd person subject nor object are overtly marked when context can supply them, as here. Nakayama glosses the ‘inverse’ marker as ‘SHIFT’, since, as he points out, its function is broader than that suggested by the former term. As in kusanartu-mik qaja-liur-puq (handsome-INSTR kayak-make3SG.INDIC) ‘he made a handsome kayak’. There is no way the affix here could be construed as ‘enclitic’, by the way – it must follow the ‘incorporated’ head. In West Greenlandic there are even a few affixes, for example -nirar- ‘say that’, which can take as base an entire proposition with its own tense or modality marker (stripped of only person/mood inflection), as in (i); though again its position is fixed i.

12.

13.

14.

173

imiq nillir-sima-nirar-paa water be.cold-PERF-say.that-3SG/3SG.INDIC ‘He said that the water had been cold’

Engberg-Pedersen points out to me that sentences such as these are completely parallel to similar utterances in Danish Sign Language (pers. comm.). It is not that traditional FG has any particular problem with languages that have only two parts of speech (as can be argued is the case for Nootka), namely nominals and verbals, which are distinguished almost entirely by morphological criteria, i.e. by the type of inflectional and derivational processes to which they are subject respectively (Dik 1989: 163 specifically mentions neighbouring Salishan languages in this context). What is more problematical is the association of such a wide array of lexical stems with both verbal and nominal (and other) uses. Term predicate formation in Dik’s sense (also corresponding derivations from semantically adjectival and adverbial elements) would on the one hand have to apply across a high percentage of all sentences and, on the other, a wide array of potential SoAs would have to be associated with individual lexical entries. Deciding what is to be the main predicate given a complex communicational intent can also be problematical with languages which do not display the same ambiguity of major word classes as Nootka. Thus consider the ini-

174

15.

16.

17.

18.

19. 20.

Michael Fortescue tial choices in Japanese watashi wa kono hoo ga ii ‘I like this one’, in which watashi, which corresponds to the English subject I, is marked as Topic (or Theme), the inner grammatical Subject is kono hoo (‘this one’) and the predicate is ii ‘good’. The standard FG treatment must start with basic predication this one is good and somehow attach (as for) me at a later stage, say as Theme at the pragmatic function assignment level, even though it is clearly central to the basic communicational intention (the propositional input to be expressed). In the following French sentence, it seems even more difficult to decide on the basic predication: j’ai ma voiture qui est en panne ‘my car has broken down’, where the Topic is actually me and the domain the whole predication (a kind of possessive cleft). Should the possessive predicate be included in the underlying structure or introduced at the pragmatic level? It is evidently a matter of focus, involving the presupposition that the speaker has a car but with the focus actually on the whole sentence as the essential new information (Lambrecht 1994: 14). Note that thetic, presentational propositions like this are often expressed with the help of clefts or dummy locative subject constructions, effectively splitting the proposition into two – in Whiteheadian terms this corresponds to expressing the indicative and the propositional prehensions behind a judgment as if they were separate propositions (they are indeed separate ‘speech acts’), and it would be natural for a process model to reflect this two-tiered prehensional structure at the deepest (initial) level. Note that Nakayama’s ‘predication’ corresponds roughly to Dik’s State of Affairs, but this is for him not a matter of the lexicon at all (Nakayama 1997: 94ff.). The predicate-based formalism of the original model does need to be extended, however, to allow for predicate-level restrictors (for the ‘restrictive’ lexical suffixes of Nootka at least) of the kind I argued for in connection with Koyukon and in line with Hengeveld’s independently motivated proposal for predicate variables (Hengeveld 1992: 31ff.). The distinct focus construction illustrated in (6a) and (6b) above (which highlights an argument for essentially contrastive purposes) does at all events need to be marked – presumably on the pragmatic-function assignment level of the traditional model. The case is parallel: THETIC or PRESENTATIVE would be part of the outermost bracketing – reflecting the act of embedding the inner proposition with its NewTop within a broader contextual nexus – i.e. much the same affect as is achieved by Hannay’s discourse management mode of that name. The scope of that focus is in part determined by the presence or absence of the TEL suffix mentioned above. Dik (1989: 8ff.) himself stressed that much of communication is implicit and not overtly coded, and this evidently is something that can vary more

Process and pattern interpretations

21.

22.

175

from language to language than we often imagine. The reason Nootka appears to be close to the extreme here is the high degree to which purely discourse-pragmatic factors directly determine coding choices (by contrast, English employs to a greater degree subtle, analogue differences of intonation to articulate such matters). Mackenzie (1998) follows Dik’s cue in his recent exploration of the ontogenetic expansion of the ‘holophrase’, taken as the essential item under Focus in an information ‘chunk’ that initially fills the P1 position. This meshes nicely with Hannay’s emphasis on P1 as the locus of initial message management choices. If English itself could then be seen as in some sense also the phylogenetic result of a long historical development where P1 became gradually freed up for other items than those under Focus (but with the initial situation remaining in adult holophrases), one might want to say that Nootka represents a language that is still much closer to its holophrastic roots (not for nothing is it also ‘polysynthetic’). P1 is still reserved for Focus (in some sense) in Nootka in virtually all its utterances types – here is positioned the one item of an information unit that cannot just be dropped. This contrasts, by the way, with equally polysynthetic West Greenlandic (which is basically SOV rather than VSO like Nootka), where there is no P1 with that function – in that language Focus often adheres to final position in the clause (except in its contrastive focal construction). Note that I am discussing a level of representation corresponding to (potentially) conscious experience. This I see as essentially a matter of inferential processes (whether on-line or automatized), both lexically-cued and grammar-cued (see Givón 1995: 364ff.). These may well in turn involve basic operations such as ‘search and retrieve’, ‘establish new node’, ‘connect node’, etc., of the kind Givón proposes – or indeed Nuyts’s (1992: 272ff.) sentencing (i.e. the breaking up of information into clause-sized units) and the like – but this is surely at a level of assumed cognitive activity below what is accessible to conscious experience. Moreover, they are not relevant to minimalizing the FG model as such. Nuyts’s cognitive component actually consists of declarative rather than procedural knowledge (it is a situational network consisting of representations, i.e. – like the FG component– represents pattern), the claim being that such knowledge is relevant to both logic (process) and to conceptual knowledge (pattern). The processes meant to implement it are only vaguely adumbrated, however. The kind of basic processes I am concerned with are more like Givón’s referent pointing (a matter of attracting attention), easily understandable as complex concrescences and describable in prehensional terms (in this case involving an indicative prehension). In general, interpreting the vertical axis of Hengeveld’s revised model as universal process does not seem warranted. As regards Nootka, for exam-

176

Michael Fortescue ple, a left-to-right elaboration of an interpersonal level structure would require very little reference to expression-level categories (parts of speech and syntactic phrase units) at all. All the relevant triggers for expression are arguably there at the interpersonal level (interacting with the lexicon). The only involvement of the representational level in production would in turn be in decisions to construe the propositional content in terms of first-order or higher-order terms and/or bare predicates – in other words as nominal or verbal expressions (this, as we have seen, being determinate and overt if inflectional suffixes are present). Languages can no doubt be typologized as regards the degree of relevance of the two lower levels for production/comprehension decisions. These would, at all events, appear to be more language-specific than those at the interpersonal level.

References Bolkestein, A. Machtelt 1998 What to do with Topic and Focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. Chafe, Wallace 1994 Discourse, Consciousness, and Time. Chicago: University of Chicago Press. Chomsky, Noam 1993 A minimalist program for linguistic theory. In: Ken Hale and Samuel Jay Keyser (eds), The View from Building 20, 1–52. Cambridge MA and London: MIT Press. Dik, Simon C. 1988 Concerning the logical component of a natural language generator. In: Michael Zock and Gerard Sabah (eds), Advances in Natural Language Generation, vol.1, 73–91. London: Pinter. 1989 The Theory of Functional Grammar, Part 1: The Structure of the Clause. Dordrecht: Foris. 1997 The Theory of Functional Grammar, Part 2: Complex and Derived Constructions. Berlin and New York: Mouton de Gruyter. Engberg-Pedersen, Elisabeth 1991 Lærebog i tegnsprogs grammatik. København: Døves Center for Total Communikation. Fortescue, Michael 1992 Aspect and superaspect in Koyukon: An application of the Functional Grammar model to a polysynthetic language. In: Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds), 99–141.

Process and pattern interpretations 1998

177

Language Relations Across the Bering Strait. London and New York: Cassell. 2001 Pattern and Process: A Whiteheadian Perspective on Linguistics. Amsterdam and Philadelphia: Benjamins. Fortescue, Michael, Peter Harder and Lars Kristoffersen (eds) 1992 Layered Structure and Reference in a Functional Perspective. Amsterdam and Philadelphia: Benjamins. Givón, Talmy 1995 Functionalism and Grammar. Amsterdam and Philadelphia: Benjamins. Hannay, Mike 1991 Pragmatic function assignment and word order variation in a functional grammar of English. Journal of Pragmatics 16: 131–155. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Harder, Peter 1992 Semantic content and structure in Functional Grammar. On the semantics of “Nounhood”. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), 303–327. Hengeveld, Kees 1992 Parts of speech. In: Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds), 29–55. 1997 Cohesion in Functional Grammar. In: John Connolly, Roel Vismans, Christopher Butler, and Richard Gatward (eds), Discourse and Pragmatics in Functional Grammar, 1–16. Berlin and New York: Mouton de Gruyter. Itkonen, Esa 1983 Causality in Linguistic Theory. London and Canberra: Croom Helm. Jacobsen, William 1993 Subordination and cosubordination in Nootka: Clause-combining in a polysynthetic verb-initial language. In: Robert Van Valin, Jr. (ed.), Advances in Role and Reference Grammar, 235–274. Amsterdam and Philadelphia: Benjamins. Lambrecht, Knud 1994 Information Structure and Sentence Form. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan 1996 Functional Grammar and the analysis of English. In: Jürgen Klein and Dirk Vanderbeke (eds), Anglistentag 1995 Greifswald, 135–148. Tübingen: Niemeyer.

178 1998

Michael Fortescue

The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–295. Mithun, Marianne 1987 Is basic word order universal? In: Russell Tomlin (ed.), Coherence and Grounding in Discourse, 281–328. Amsterdam and Philadelphia: Benjamins. 1999 The Native Languages of North America. Cambridge: Cambridge University Press. Nakayama, Toshihide 1994 Phrasal suffixation in Nootka. In: Osahito Miyaoka (ed.), Languages of the North Pacific Rim, 263–272. Sapporo: Hokkaido University Publications in Linguistics No. 7. 1997 Discourse-pragmatic dynamism in Nuu-chah-nulth (Nootka) morphosyntax. Ph.D. dissertation, University of California at Santa Barbara. Nuyts, Jan 1992 Aspects of a Cognitive-Pragmatic Theory of Language. Amsterdam and Philadelphia: Benjamins. Quine, Willard V.O. 1960 Word and Object. Cambridge MA.: MIT Press. Shibatani, Masayoshi 1990 The Languages of Japan. Cambridge: Cambridge University Press. Strawson, Peter F. 1974 Subject and Predicate in Logic and Grammar. London: Methuen and Co. Swadesh, Morris 1936 Nootka internal syntax. International Journal of American Linguistics 9 (2–4), 77–102.

Functional Discourse Grammar and language production J. Lachlan Mackenzie

1. Introduction One of the central requirements placed on a Functional Grammar is that “… such a grammar must also aim at psychological adequacy, in the sense that it must relate as closely as possible to psychological models of linguistic competence and linguistic behaviour’ (Dik 1997a: 13). The FG model is presented in a quasi-productive mode (Dik 1997a: 57), but at the same time sees itself as essentially generative – indeed the primary purpose of the computer implementation of FG (Dik 1992) was to test and enhance its generativity. Recognizing that to confuse generation with production is a category error, FG has insisted that it is not a model of language production. Yet the demands of psychological adequacy have encouraged speculation on the relation between the model, in its various manifestations, and the ever more sophisticated models of language production currently available. The purpose of this chapter is to examine the potential of Functional Discourse Grammar for a further rapprochement between FG and a psycholinguistic consensus on language production. Jackendoff (1997: 7–8) has argued that there are three possible positions on the relation between grammar and the processes of speech production and perception: (a) (b)

one can deny any relationship, insulating grammar from psycholinguistic findings (the traditional generativist position); one can maintain that processing mechanisms can ‘consult’ or ‘invoke’ a declarative grammar;

180

J. Lachlan Mackenzie

(c)

or one can claim that the processor embodies the grammar, i.e. that grammar is itself procedural.

I believe that there are certain developments in current FG that suggest that greater psychological adequacy is achievable through a combination of Jackendoff’s positions 2 and 3. More specifically, I will explore the possibility that Hengeveld’s (this volume) interactional and expression levels can be modelled as procedural, with the intermediate (and in some cases by-passed) representational level functioning as a declarative module that is consulted by the processing mechanisms.

2. FG as a declarative grammar A central tenet of the functionalist stance is that linguistic form results from a complex of choices. For example, there is in certain languages a choice whether or not to apply the Subject function to a first argument, which is reflected in English in the active vs. passive voice (Dik 1997a: 248–250). The choice would appear to be the speaker’s. However, this raises an apparent problem for FG, and other functionalist approaches that stress ‘choice’: psycholinguistic findings strongly suggest that most of these choices are simply not accessible to the language-user’s awareness. As Levelt (1989: 21) puts it, “A speaker doesn’t have to ponder the issue of whether to make the recipient of GIVE an indirect object (as in John gave Mary the book) or an oblique object (as in John gave the book to Mary)”. This is known as ‘cognitive impenetrability’: many aspects of language processing are automatized, and not accessible to choice. The conclusion must be that, when FG invokes psycholinguistic explanations, it is trying to make understandable why the automatic processes of language are as they are. This applies as much to the hearer’s processing of utterances as to the speaker’s production of them. To take an example, the explanation offered for LIPOC is that “[i]t is easier [sc. for the hearer, JLM] to perceive, process, and store complex information when this information is presented in chunks of increasing internal complexity” (Dik 1978: 212). We must beware, however, of assuming that the speaker is actually choosing to alleviate the hearer’s interpretive task: language users do not consciously apply linguistic principles in their effort to express themselves. Surely the explanation offered by Dik must be understood as historical and selectionist: of all possible orderings of information, the one that has sedimented into automatic processing is that which, everything

FDG and language production

181

else being equal, conforms with LIPOC. Note that there is nothing antifunctional about this understanding of ‘choice’. On the contrary, automatic processing can itself be seen as functional, since it certainly contributes to efficiency, freeing the speaker’s and hearer’s minds to concentrate on the subject-matter of the discourse. Applying these insights to the construction of a model of FG, we will need to regard the various generalizations made within FG as constraints upon the cognitively impenetrable processes of language production. What would then distinguish a functional grammar from a non-functional one would be that the constraints are not arbitrary. Rather, they are understandable a posteriori in terms of efficient interpersonal communication. The position that suggests itself, then, is one that is compatible with Jackendoff’s second option: an FG is declarative and takes the form of a set of constraints on processing. Across billions of communicative acts, these constraints have proved their worth. It is the functionalist’s task to identify them and to provide explanations for their existence.

3. FG as a model of language production As mentioned in Section 1, some recent work has been moving FG in the direction of Jackendoff’s third position. The main exception to the dominant interpretation of FG as simply a declarative grammar is to be found in the work of Jan Nuyts, who has forcefully presented his Functional Procedural Grammar (Nuyts 1992). He has argued that a number of matters traditionally regarded in FG as being part of grammar proper should be assigned to non-grammatical cognition. Questions such as the assignment of pragmatic functions should in his view be handled in conceptual and/or textual components, and in any case fall outside the grammar. Another important consequence of Nuyts’s work concerns matters such as the distinction between operators and satellites, or the distribution of information over main and subordinate clauses. For both phenomena, their treatment in the grammar is entirely dependent upon their ultimate formal properties: operators correlate with grammatical, and satellites with lexical expression; similarly, there is no simple or direct correlation between cognitively nuclear information and main clauses, or between cognitively subsidiary information and satellite clauses. Nuyts argues that these distinctions are also more plausibly understood as resulting from cognitive (pre-linguistic) operations. He takes Jackendoff’s third position in propos-

182

J. Lachlan Mackenzie

ing a model of clause production. He nevertheless refers to it as a grammar, not least because the processes described culminate in a modified, slimmed-down version of FG, including a rather orthodox expression rule component. He has thus clarified the stages that take us from the first communicative intention through to the representation of the predication. Bakker and Siewierska (this volume) have now provided an existence proof of the possibility of regarding the expression rules, too, as following production. They assume a fully specified underlying representation of the clause which is static and declarative in style. But the expression rules they propose can and should be seen as dynamic, working top-down, depth-first, and - crucially, as in actual language production, left-to-right. The trend towards a re-interpretation of (at least part of) FG as a production model is reflected, too, in Hengeveld’s (this volume) call for a conversion of FG from a bottom-up grammar taking the predicate frame as its starting point and adding layers in an upward fashion to a top-down grammar, “starting with the communicative intention and ending with the articulation of the linguistic expression”. This call cannot be divorced from the increasing interest in reconciling Functional Grammar and discourse analysis. This urge to include discourse considerations in the overall model arises from a desire to achieve ‘pragmatic adequacy’ rather than ‘psychological adequacy’. Nevertheless, we are aware that psychologists, too, have been moving away from the examination of individual utterances towards considering ‘arenas of language use’, as Clark (1992) has it, and the entire range of activities that occur in those arenas. There may be less distinction between ‘psychological’ and ‘pragmatic’ adequacy than in the early days of FG. The relationship between discourse and grammar can, I believe, be seen in three different ways, all of which are being tried out in current work. Firstly, we may seek to model discourse and grammar in separate compartments of our theory; we will then try to link them through an interface (this is essentially the position taken in work by Kroon 1997 and by Vet 1998). Secondly, we may examine the proposition that discourse is structured in analogy to clausal grammar – this is how I read Chapter 18 of Dik (1997) and, in different ways, the earlier and recent work of Hengeveld (1997, this volume) and Moutaouakil (this volume). Thirdly, and this is the position that seems most consistent with a production-mimicking approach, we can see discourse production as a dynamic process occurring in real time and the expression of the clause as a similarly real-time process. Clark (1996: 285) has pointed out that “Utterances are often viewed as the prerogative of speakers – products that speakers formulate and produce

FDG and language production

183

on their own. Nothing could be further from the truth”. Dik (1997a: 8) similarly stresses that we are concerned with verbal interaction. At this level conscious decisions are involved: not the choice whether to assign Object function to a Recipient or not, but a split-second determination of what move to make at a particular juncture in the joint activity of communication: to contest what the other has said, to crack a joke to defuse a tricky situation, etc. The choice of move is a complex matter involving, as discourse analysts have shown us, a number of goals at once, not least that of keeping the conversation going in a mutually co-operative manner.

4. Acts and subacts Let us, with Hengeveld (this volume), take the speaker’s strategic move as the point of departure for analysis. This ‘minimal free unit of discourse’ consists of a number (n ≥ 1) of acts. Under normal circumstances, their sequence in cognition will be reflected by their sequence in expression. In this view, the schema Theme, Predication, Tail familiar from Dik’s early work (1978: 130) will be reinterpreted as reflecting a single move consisting of three acts: the Theme corresponds to an act that introduces a discourse referent, the Predication to a cognitively subsequent proposition involving that referent in some way, and the Tail to a final act, correcting or amplifying the preceding act, possibly as a result of self-monitoring. Each Act corresponds in principle to a single intonation unit, although many extraneous factors may upset the biuniqueness. In a language like English the intonation unit generally contains one particularly accented and pitch-changing (tonic) syllable, which helps to mark a possible transition point for the other conversation partners. This accented syllable is in most cases also a guide to the communicatively most salient aspect of the act. In FG this corresponds to the assignment of the pragmatic function Focus. The Focus is thus the essential component of each act; i.e. there is no act, no matter how brief, without a Focus. Let us consider an utterance such as (1): (1)

Oh my God, it’s on fire, my hair.

This utterance will, let us assume, reflect one move by the speaker. The utterance will most likely manifest three intonation contours, each with a tonic syllable (God, fire, hair respectively). The move can correspondingly be seen as containing three sequenced acts which rather naturally reflect

184

J. Lachlan Mackenzie

the succession of cognitive events: shock; perception; identification. Using, but adapting, Hengeveld’s proposals, we may represent the outline of this move as (2): (2) (M1: [(A1), (A2), (A3)])

where the numbering of the acts corresponds to their sequence in cognition. Acts come in two basic types. Simplex acts (like A1) consist only of their Focus. (Hengeveld this volume: 11 provides the example of the exclamation Damn!.) Simplex acts do not involve the grammar in any way, but are passed directly from the interaction component to the expression component. Hengeveld assumes that such acts involve a consultation of the lexicon. In contrast, I should wish to claim that the words uttered in a simplex act do not come from the lexicon, but are furnished directly by the expression component. The lexicon contains predicates, and is therefore an essential support for the representational component. But where the latter is not involved, the lexicon is also not in play. Simplex acts, which possibly represent a perpetuation of more primitive forms of communication, i.e. animal ‘calls’, instruct the expression rules to activate formulae appropriate to the communicative function of the act. Speakers may of course differ within a speech community in which formulae they prefer. Consider how various acts of exclamation offer a range of ready-made expressions, each with its own social implications: (3) agreement disagreement regret

Yes, Absolutely, You said it, Too right, … No, No way, Absolutely not, … Sorry, Pardon me, Excuse me, …

We also find acts of solidarity, covering various greetings and valedictions, but also echoes and vocatives. Echoes (where the speaker, for one of a variety of reasons, repeats some fragment of the interlocutor’s previous speech) involve the intention to (re-)express an element of the communicative context; the lexicon need not be consulted. And vocatives similarly draw upon a very specific part of that context, the identity of the interlocutor, with the expression rules applying, where that is relevant for the language being spoken, the appropriate vocative marking. Another prominent type of simplex act is the act of organization, which is oriented to keeping the conversation going: the forms used for this purpose are generally highly formulaic, ranging from interjections like ah and oh to more

FDG and language production

185

complex forms that mimic the complexity of lexical formulations (technically speaking, to be more specific, …). Complex acts differ from simplex acts in consisting of one or more subacts: in (2) above, A2 (expressed as it’s on fire) and A3 (expressed as my hair) are complex acts. One of these subacts carries the pragmatic function Focus that characterizes the act as a whole. The subacts in question are subacts of reference (R) or of ascription (T): both reference and ascription are thus seen as activities carried out by the speaker in interaction. If a complex act contains only one subact, this subact then necessarily carries the Focus function (as in A3). Where a complex act contains more than one subact (as in A2), one of these carries the Focus function. All the acts and subacts discussed here are initiated and organized at a pre-linguistic, conceptual level. Whereas simplex acts are sent directly to the expression component, complex acts, which I take to be definitional of the more sophisticated communication systems of the human being, have a more complex history. All complex acts involve the representational component, and may also involve the lexicon, although not necessarily so. Thus both (4a) and (4b) result from complex acts, each involving a subact of reference and a subact of ascription, but only (4a) invokes lexical material: (4) a. b.

The president smokes. He does.

Hengeveld (this volume) points out that the mapping relation between the interactional and representational components is not one-to-one. Although there are default correlations between an ascriptive subact and a predicate (T ↔ f), and between a referential subact and a term (R ↔ p, e, x), other possibilities are available: ‘predicate nominals’ thus reflect the mapping of an ascriptive subact to a term (T ↔ x). A striking characteristic of Hengeveld’s proposal is that the valency structure rightly posited for the representational component, i.e. (f1) (x1), is also present in the interactional component: within the communicated content C, we find the representation … (T1) (R1) …, which is explained as involving ‘the speaker who ascribes properties to entities’. If the interactional component is to be viewed as reflecting production processes carried out in real time, however, an alternative view suggests itself. Let us assume that the Focus of each act corresponds to the temporally first cognitive element activated in the preparation of an utterance. After all, it is the very communication of that Focus that justifies saying anything at all. Indeed, in

186

J. Lachlan Mackenzie

appropriate circumstances, determined by such matters as urgency, or familiarity between the interlocutors, the Focus is all that need be communicated. In the holophrastic speech of infants, too, it is focal information that is expressed; it is only as they develop a ‘theory of mind’, i.e. a social competence, that they come also to express non-focal (typically topical) information (cf. Levelt 1999: 85). In this view, then, the cognitive assembly of an act commences with the identification of a Focus: this is the first (and possibly only) subact. Associated with any complex act is a fundamental stance: either a giving stance, where the speaker is offering information or goods/services to the addressee, or a demanding stance, in which the speaker is eliciting information or goods/services from the addressee (cf. Halliday 1994: 68–69). The stance will be determined by the speaker’s initial motivation for linguistic communication, and it is this stance towards the Focus that determines the basic ‘illocutionary’ status of the act as a whole. Thus a giving stance correlates with an assertion, a demanding-information stance with a question and a demanding-goods/services stance with an order or request. Rather than associating the illocution with the entire act, as Hengeveld does, I would therefore prefer to indicate the speaker’s stance on the subact in Focus. This subact may be supported by further subacts, in line with the speaker’s awareness of her participation in the joint project of communication with the addressee. Here she draws upon the sources of knowledge available in the communicative context (Dik’s ‘pragmatic information’, 1997a: 10) to establish a framework in which her Focus will affect the addressee in the desired way. The more purely cognitive relations between the supportive subacts and the Focus subact are familiar from existing work in FG: the supportive subacts may be Topics of various kinds (Given, Resumed, etc.) or possibly Settings; other subacts may be inspired by the desire to reflect the (power) relationship between the interlocutors, such as those that result in indirect formulations, attitudinal adverbials, etc. The representation that suggests itself will thus differ in various ways from that proposed by Hengeveld (this volume) for the interactional component. The representation will be incremental, i.e. the left-to-right ordering of acts and subacts will be designed to reflect actual sequence in cognition. An outline suggestion for the representation of (1) above would appear as follows: (5) (M1: (express A1: SHOCK)Foc, (assert A2: (SA2:1: (T: FIRE)Foc), (SA2:2: (R: )Top), (assert A3: SA3:1: (R: HAIR)Foc)

FDG and language production

187

The various elements in capital letters stand for activated emotions and concepts; the operators in bold print indicate the type of act. As soon as the process represented in (5) begins, the expression component is set into action. This is in line with “a basic consensus in the language production literature” (Levelt 1999: 88), namely the notion of incremental production. This says “that the next processing component in the general flow of information can start working on the still incomplete output of the current processor. A processing component will be triggered into action by any fragment [Levelt’s emphasis] of its characteristic input. As a consequence the various processing components are normally simultaneously active, overlapping their processes as the tiles of a roof”. Whereas simplex acts can be sent directly to the expression component (as outlined above), complex acts call upon the declarative grammar, which may be modelled much as is proposed in Hengeveld’s representational component. There, the units corresponding to the various subacts, Focus and non-Focus, will generally, as with A2 in (5), be grouped around a predicate into the familiar propositional form. The predicate is selected from the Fund, i.e. the lexicon expanded by the various derivational processes permitted by the language. The selection of the predicate, i.e. of the State of Affairs, may well reflect the speaker’s perspective upon her conceptualization, as with the selection of buy versus sell for a commercial transaction; the perspective may also manifest itself in syntactic function assignment. As Levelt (1999: 91–92) points out, the perspective may be influenced by pragmatic factors attributable to the ongoing discourse, but the speaker does have some freedom in this regard.

5. The representational and expression components The entire layering system of the FG representational component now comes into play, setting down the ground rules for the linguistic formulation of the cognitive processes that are to be expressed. They form a complex of constraints upon the translation of “intention into articulation” (Hengeveld this volume), constraints that are partially universal and partially language-specific. The universal constraints cover such matters as argument structure, the ordering of semantic functions, the relative scope of the layers; the language-specific constraints relate to the distribution of information over operators and satellites, the presence or absence of tense, of Subject assignment, of evidentiality, etc. Further constraints derive from the Fund: certain predicates strongly impose a certain argument structure,

188

J. Lachlan Mackenzie

such as English put, which in its literal sense calls for specification of Agent, Goal and Location. In Levelt’s (1999: 95) formulation, “[i]n a way grammatical encoding is like solving a set of simultaneous equations. Each lemma [in FG, ‘predicate’ JLM] requires particular syntactic constraints from its environment and the emerging syntactic structure should simultaneously satisfy all these constraints”. If the representational component is seen as a set of constraints on processing, it is no longer necessary to assume that every utterance has a full propositional form. The grammar need simply be consulted for what is necessary to the task at hand, or as Hengeveld (this volume) has it, “it is no longer necessary to assume that every discourse act contains a propositional content”. Thus an act such as A3 in (x) above, containing only one referential subact, undergoes only those constraints that relate to term structure, including, in those languages that require this, a semantic or syntactic function that determines its case form. In this way, it is possible to treat so-called fragmentary utterances without recourse to rules of deletion. The complete clause then emerges as a system for supporting the Focus, providing a context within which the addressee can understand it as well as possibly satisfying other communicative goals of the speaker’s. The interpretation of Hengeveld’s model that I am proposing is a combination of Jackendoff’s second and third options: the overall grammar is procedural (position 3), but in the construction of complex acts, it needs to consult, indeed conform to, a declarative grammar (position 2). The advantage of a procedural approach is that FG will accord better with the “psychological models of linguistic competence and linguistic behaviour” which Dik (1997a: 13) wishes the practitioners of FG to bear in mind. Incremental processing in the production of utterances works on the assumption that the concepts that lie behind a complex act are not all available at once. Speakers characteristically do not wait until all the concepts are “in” before starting to speak. The various levels of activity, conceptualizing, formulating and articulating, as Levelt (1989) has it, work in parallel, with conceptualization having a head start over formulating, and formulating over articulating. However, speakers are not free to express their ideas exactly as they occur to them: the declarative grammar, with its conventional principles, will often mean that particular parts of a message are put into a buffer before they can be expressed. Clausal structure appears to have exactly this buffering function. At least, this seems to be plausible for English, a language with rather fixed word order, by which I mean that the syntactic template plays an important role in expressing the relations imposed by the declarative grammar. But what of a

FDG and language production

189

language with a so-called ‘free word order’? By this we understand that the order of constituents is much more strongly influenced by their cognitive status as New, Given, Inferred, etc. For such languages the temporal order in which the concepts become available may be much more directly reflected in word order, with the declarative grammar serving to ensure above all that the semantic relations between the constituents remain transparent to the addressee. In such languages we might expect the Focus to be expressed first, in line with its cognitive priority; an example of this situation is discussed in depth by Fortescue (this volume). Yet even in language with relatively ‘free’ word order, there is often one syntactic position, frequently not P1, which is reserved for the Focused constituent. In Turkish (Van Schaaik 2001: 45) this is the immediately preverbal position; in Hungarian (De Groot 1989: 105) it is the position between Topic and Verb. The non-initial positioning of Focus in actual utterances, as against its cognitive priority, may also be understandable from a processing viewpoint. The cognitive identification of a Focus concept, which is typically new information, and the selection of an appropriate lexicalization of that concept, both require time. By contrast, associated information that is already activated and thus highly accessible (i.e. Topics and Settings) is immediately available for expression. The speaker who expresses accessible information first allows herself more time for the identification and lexicalization of the Focus. Hannay (1991) has shown that even for a syntactically rather rigid language like English, the relative positioning of Topic and Focus is dependent upon what he calls the ‘message mode’: in an urgent situation, the Reaction mode encourages initial placement of the Focus, with optional back-up from a following Topic; where there is a highly accessible Topic, the Topic mode induces a postponement of the Focus to a later syntactic position. It may indeed be advisable to add a specification of the message mode to the representation of utterances in the interactional component. One consequence to be drawn from the preceding discussion is that the expression component is no longer, as it is in orthodox FG, the only place where the order of constituents is determined. Its function now emerges as that of balancing the competing demands of the interactional component, which ‘wishes’ to see each component of its emerging message expressed as soon as possible, of the representational component (as in traditional FG), which requires an unequivocal reflection of its demands, and of the expression component itself, which has its own language-specific rules and regulations. These are the ‘simultaneous equations’ to which Levelt (1999: 95) refers.

190

J. Lachlan Mackenzie

6. Some examples Let us conclude by sketching out a couple of examples of how things might work. If I feel hunger, I can conceive of grapes as something I want. Under certain circumstances, for example if I am surrounded by willing slaves, it will be sufficient for me to utter a complex act with a single and inevitably Focused subact of reference: (6) Grapes!

to be analysed at the interactional level as (7) (7) (M1: (request A1: (SA1:1: (R: GRAPES)Foc)))

Here only that part of the lexicogrammar is involved that deals with subacts of reference. The lexicon offers the countable noun GRAPE and the declarative grammar (of English) requires that the referential subact be specified for at least definiteness and number; the expression rules react appropriately to the operators indefinite and plural. The operator request triggers the appropriate intonation contour. The rest lies with the interpretive abilities of my slaves. Under less utopian circumstances, I will have to devise a strategy to achieve my purpose. The focal reference to ‘grapes’ now appears in the context of this overall strategy. Let us assume that the strategy I have selected is one of moderate politeness, and is associated with the formula Can I have …. This kind of unit is reminiscent of the expression of simplex acts in being a ready-made formula, probably learned and stored as such (cf. Nattinger and Decarrico 1992). The following representation suggests itself, where the Move takes the operator pol for politeness: (8) (pol M1: (request A1: (SA1:1: (R: GRAPES)Foc)))

This is passed directly to the expression rules, which fit the two elements Can I have and grapes, the latter (only) having passed through the representational component, into an acceptable template: (9) Can I have grapes?

This example is a good instance of how any one utterance can contain a

FDG and language production

191

mixture of elements that come directly from the interactional component and elements that have passed through the representational component. In this way we are able to account for the fact that (9) functions in the same way as (7), but differs only in degree of politeness. At the same time, (9) remains parsable: if so desired, a language user can refer it back to the declarative grammar, as appears to happen when a speaker, in an effort to be hyperpolite, creates a baroque variant of the socially recognized formula: (10) If you could possibly see your way clear to providing me with some grapes.

An example in which the representational component is more thoroughly implicated is (11), where capitalization indicates the tonic syllable: (11) Did you EAT the grapes?

Here, let us assume, the speaker is aware of the addressee and the grapes: both are given information, and the Focus is on eat. The initial structure in the interactional component could therefore be: (12) (M1: (inquire A1: (SA1:1: (T: EAT)Foc)))

which would yield the utterance (13), a highly unlikely utterance in the circumstances at issue here: (13) EAT?

The representational component, however, will not be ‘satisfied’ with (13), since the lexical entry for eat contains an Agent and a Patient. The lexicon also contains the information that eat without a Patient has a different, here irrelevant sense ‘have one’s meal’. The representational component therefore demands specification of the Patient, yielding (14), which would be a possible utterance in the given situation: (14) EAT the grapes?

In other words, in creating the representation in the interactional component, the speaker heeds the declarative rules of the language, giving: (15) (M1: (inquire A1: (SA1.1: (T: EAT)Foc), (SA1.2: (R: GRAPES))))

192

J. Lachlan Mackenzie

As was mentioned, the lexicon also recognizes the Agent of the eating as an argument, which forms another, slightly less binding, constraint on the speaker’s communicative intention as represented in the interactional component: (16) You EAT the grapes? (17) (M1: (inquire A1: (SA1.1: (T: EAT)Foc), (SA1.2: (R: GRAPES)), (SA1.3 (R: ADDRESSEE))))

Finally, the representational component of English has rules regarding the finiteness of propositions and the tense of predications which can also be imposed upon the translation of the intention into expression. It is these that are involved in the production of (11). More precisely, both the structure (17) and a representational structure (18) in which the relevant operators are specified will form the input to the expression rules: (18) (finite p1: (past e1: (f1: eat [V] (d1x1: [-S, +A])AgSubj (d1x2: f2: grape [N])Go)))

The final structure of (11) results from the expression rules’ reacting to a collaboration between the on-line interactional component and the declarative representational component: (a) (b) (c)

(d)

the speaker decides to produce an act of inquiry; the speaker focuses this act on EAT, as shown in (12) above; this activates the lexical entry eat and implicates the representational component, which (as ever) imposes finiteness, tense, valency and syntactic function assignment; the speaker now starts to formulate the utterance, gradually enriching the interactional representation to take account of the requirements of the representational component: did in utterance initial position helps to signal the act of ini. quiry (along with the intonation contour) and satisfies the representational component’s demand for finiteness and tense; you satisfies the representational component’s demand for ii. valency and subject assignment; iii. EAT, with tonic syllable, expresses the Focus of the utterance; iv. the grapes satisfies the representational component’s demand for valency.

All three possible formulations impart the Focus (diii); (14) arises from

FDG and language production

193

also heeding (div), (16) from also heeding (dii) and (div), and (11) from heeding all the demands of the representational component.

7. Conclusion I have taken as my point of departure Hengeveld’s (this volume) proposal for the architecture of a Functional Discourse Grammar. With a view to enhancing the psychological adequacy of FG and realizing the ambition of bringing FG closer to a production model, I have suggested that the interaction and expression levels can be reformulated as operating incrementally in real time. At the interactional level, moves consist of sequences of acts, which in turn consist of sequences of subacts; at the expression level, too, there is a sequence of constituents. Speakers typically activate the expression level before completing the underlying act at the interactional level. The representational level is regarded as a declarative component that can be called upon to constrain these two real-time processes where lexical items from the fund are involved. A major contribution of this level is to provide supportive material against which the Focus of the Speaker’s message can be understood, encouraging and often constraining the speaker to provide more information than that Focus alone. It is hoped that this approach can offer something to the understanding of the ontogenesis of multi-word speech in infants and to the study of the implicitness of much conversational speech.

References Bakker, Dik and Anna Siewierska This vol. Towards a speaker model of Functional Grammar. Clark, Herbert H. 1992 Arenas of Language Use. Chicago: University of Chicago Press. 1996 Using Language. Cambridge, New York and Melbourne: Cambridge University Press Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds) 1997 Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter. Dik, Simon C. 1978 Functional Grammar. Amsterdam: North-Holland.

194 1992

J. Lachlan Mackenzie

Functional Grammar in Prolog: An Integrated Implementation for English, French, and Dutch. Berlin and New York: Mouton de Gruyter. 1997a The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part 2: Complex and Derived Constructions. Berlin and New York: Mouton de Gruyter. Fortescue, Michael This vol. The complementarity of the process and pattern interpretations of Functional Grammar. Groot, Casper de 1989 Predicate Structure in a Functional Grammar of Hungarian. Dordrecht: Foris. Halliday, Michael A.K. 1994 An Introduction to Functional Grammar. 2nd edition. London: Arnold. Hannay, Mike 1991 Pragmatic function assignment and word order variation in a Functional Grammar of English. Journal of Pragmatics 16: 131–155. Hengeveld, Kees 1997 Cohesion in Functional Grammar. In: John Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 1–16. This vol. The architecture of a Functional Discourse Grammar. Jackendoff, Ray S. 1997 The Architecture of the Language Faculty. Cambridge MA and London: MIT Press. Kroon, Caroline 1997 Discourse markers, discourse structure and Functional Grammar. In: John Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 17–32. Levelt, Willem J.M. 1989 Speaking: From Intention to Articulation. Cambridge MA and London: MIT Press. 1999 Producing spoken language: A blueprint of the speaker. In: Colin M. Brown and Peter Hagoort (eds). The Neurocognition of Language, 83–122. Oxford: Oxford University Press. Moutaouakil, Ahmed This vol. Discourse structure, the generalized parallelism hypothesis and the architecture of Functional Grammar. Nattinger, James R. and Jeanette S. Decarrico 1992 Lexical Phrases and Language Teaching. Oxford: Oxford University Press.

FDG and language production

195

Schaaik, Gerjan van 2001 The Bosphorus Papers: Studies in Turkish Grammar 1996–1999. Istanbul: Boğaziçi University Press.

Comment clauses, Functional Discourse Grammar and the grammar-discourse interface Peter Harder

1. Introduction Hengeveld (this volume) introduces a new format for the incorporation of discourse into the linguistic model of Functional Grammar, in which the ‘upward layering’ and the ‘modular’ approaches are integrated (FDG). In this chapter I would like to take up the basic idea of a three-level integrated model and use one of Hengeveld’s examples to discuss both the advantages of his proposal and the reasons why a moderate change of the model in the ‘modular’ direction might improve the model’s capacity to bring out relations between discourse and grammar. Hengeveld’s most radical break with previous FG models is perhaps in the top-down derivational process, and in the motivation for that choice in the form of an explicit adoption of a theory of speech production matching the linguistic model point by point. At the interpersonal level, the initial choice is an explicitly nongrammatical entity, the move, defined in strictly interactional terms as corresponding to a single communicative intention; after that stage, however, all choices reflect coded choices. After the top interpersonal level, therefore, all the purely code-free communicative choices have been made. The representational level is conceived as consisting of entities of different orders which are described by the clause. Propositions (third-order entities) are the top layer; speech acts are now above the scope of the representational component. The main element of coding choice here is, therefore, how to represent the referential and ascriptive choices made in the interpersonal component. The expression level is described as consisting of constituent structure, from paragraphs to words; I shall assume,

198

Peter Harder

however, that linear templates are part of the model, but have merely been left out for purposes of simplification.

2. Interface rather than underlyingness A valuable feature of FDG is that it represents a movement in the direction of looking at complex constructions in an ‘interface’ perspective rather in terms of a ‘depth’ perspective where actual linguistic expressions end up as the ‘surface’. In FDG, we have three different analyses of the same linguistic chain, representing three different sets of facts about it. Not all levels are involved in every case; but in standard cases they are. Moreover, under specifiable conditions there may be correspondences or discrepancies that create different types of relationships between the three levels. The linguist using the model to describe an utterance will have his task defined in terms which make good sense from both a formal and a functional point of view: he has to look at interactive organization, representational organization, and organization of the expression side, and see how they interlock. An example of how the interface perspective is brought out in FDG is the element of ‘downward layering’, in which the interpersonal level is expanded to cover communicated content, reference and ascription instead of leaving that to the representational component. The functions of reference and ascription are standardly but not inevitably carried out by means of noun phrases and verbs; and the separation between the interpersonal acts and their linguistic representation makes possible a much more adequate account of the complex relationship between sentence organization and entity categories than previous models. Combined with the top-down perspective, the simultaneous operation of all three levels makes it possible to see a descriptive procedure as a set of choices involving all levels, with an anchor outside the grammar itself in the form of the speaker's communicative intention. From a discourse point of view, this is clearly an advance in adequacy over a model that begins the execution of a communicative intention by choosing a basic predicate.

Comment clauses and FDG

199

3. The grammar-coding perspective vs the intention-coding perspective Although the layout is well motivated in terms of adequacy from the discourse viewpoint, it raises a fundamental question of the natural ontological relations between grammatical choices and communicative choices. Should we expect the speaker’s ‘decision tree’ of communicative choices in making an utterance to be identical to the decision tree of making grammatical choices? It is noteworthy that over the past decades, models of grammar, whether generative or more functional, have become increasingly lexically based in their descriptive procedure, with the verb as the central source of those dependencies that determine grammatical options. This trend suggests that if we want to describe how grammatical choices are structured, there are factors that favour beginning with lexical elements – rather than placing them at the end of a long hierarchical procedure. On the other hand, it is natural to assume that there is a form of isomorphism in the sense that unless the speaker can adapt his choice process to the choices offered by the code, it is difficult to see how he could produce adequate utterances. Slobin and Berman’s ‘thinking-for-speaking’ research (cf. Slobin 1987; Berman and Slobin 1994), although focusing on other aspects of the issue, also suggests that we need to provide descriptions where the two sides match up. Until we know more, however, a number of different solutions are possible: apart from point-by-point matching of linguistic and communicative choices, one might imagine a first phase of communicative choices gradually blending into a later phase of coding choices, or a process in which the interface between communicative aims and coding options is present from the beginning until the end. For reasons described elsewhere, I think the linguistic hierarchy has an inherent order which reflects a stage of ‘compilation’ rather than ‘execution’, or ‘recipe’ rather than ‘cooking’ (cf. Harder 1996: 214ff.), where you have to relate operators to operands before you can execute the program, in which case there would always be a simultaneous bottom-up and top-down logic; cf. also Fortescue (1992: 122) on the bottom-up ‘centrifugal principle’ that seems to apply to expression rules. Whatever the exact truth may be, I believe that an explicit grammardiscourse interface is an advantage in a framework that seeks the integration of discourse into a functional linguistic theory; a point-by point matching process would then not be excluded, but it would emerge as a special case. I think such a model would be in keeping with the central part

200

Peter Harder

of the aims stated in Hengeveld (this volume), even if it differs in the way it conceives of the three strata. Taking my point of departure in an example used by Hengeveld, I am going to illustrate the issue with an example from English grammar, that of so-called comment clauses.

4. The hedged performative and the report of emotional state in FDG To demonstrate the kind of complex phenomena that are brought into the purview of the model by FDG, Hengeveld (this volume: 15) provides an analysis of the ambiguity between the Spanish sentences (1) and (2): (1) (2)

(Me) to.me Me to.me

temo que I.am.afraid that temo que I.am afraid that

Juan Juan Juan Juan

esté is.SUB está is.IND

enfermo ill enfermo ill

Both translate into ‘I am afraid that Juan is ill’, but in the indicative version the subclause contains the main message, which is reflected in its status on the representational and the interpersonal levels: the content of the subclause constitutes the content of a referential act made on the interpersonal level: (A1:[DECL(P1)Sp(P2)Addr(C1: [...

(R1) (c2) (CL1)

...] (C1))] (A1))

This illustrates the advantages of the three simultaneous levels in FDG: the interactive choice made by the speaker simultaneously affects the morphosyntactic expression and the linguistic representation of the coded content, by inserting a different content (c2) as constituting the main representational element. In this respect the model represents an extension of the advantages of earlier FG versions over a traditional grammatical description. From a traditional grammar point of view, the difference is simply a difference in terms of mood, located in a paradigmatic slot in the Spanish verb; and traditional discussions about indicative and subjunctive mood were conducted in terms of notional differences such as that between ‘real’ and ‘unreal’, which were very imprecise and often difficult to apply to concrete cases without a great deal of ‘semantics’ in the wrong sense of the word.

Comment clauses and FDG

201

As demonstrated in depth in a number of publications by Hengeveld in relation to modality, the layered model is useful precisely as a way to do two things at once: to relate a grammatical choice like the indicativesubjunctive contrast systematically to those aspects of sentence meaning that it affects, and to show what the difference is in different types of cases. This particular case requires access to more than pre-discourse layering. Unless there were three simultaneous levels, with relations explicitly indicated, and with the possibility of interpersonal factors moving ‘downward’, this could not be captured. But the question is how best to configure and distribute the discourse information and the grammatical information over the model. Consider now the English translation of the two Spanish clauses, which can receive either of the two readings: (3)

I am afraid that John is ill.

To describe this sentence adequately and to capture its ambiguity, we need to be able to assign either of the FDG descriptions to it, rather than one for each mood, as in Spanish. Because of the orientation of FDG towards matching the speaker’s choices, with the point of departure in his communicative intention, it is logical to expect that the differences in communicative intention will be clearly expressed at the interpersonal level; and in terms of the basic features of Hengeveld’s model as laid out in the introduction, this is also the aim. Quoting selectively from Hengeveld’s description (this volume: 5) of what goes on at the interpersonal level, we see that [a]t the interpersonal level a central unit of analysis is the move (M), ... the vehicle for the expression of a single communicative intention of the speaker. In order to achieve his communicative intention, the speaker executes one or more discourse acts (A), defined in Kroon (1995: 65) as ‘the smallest identifiable units of communicative behaviour’. A move consists of one central act, which may be supported by one or more subsidiary acts...In order to build up the communicated content the speaker may have to execute one or more ascriptive acts (T) and one or more referential acts (R).

Hengeveld, moreover, leaves open the possibility of including larger structures such as turns and exchanges where necessary. The model as described in the passages quoted above seems to me well suited to bring out the difference between the two readings. The question I raised above

202

Peter Harder

comes in when we look at what information is provided by the actual descriptive format. In Hengeveld’s schematic version of the ‘interpersonal level’, the information reflects the familiar grammar-oriented structuring principle, covering participants in the discourse seen as ‘arguments’ of the illocutionary frame. The ‘discourse act’ information quoted above is not immediately visible; instead we see a format that is tailored to cover the generic features of declarative sentence types. Because of that, the interpersonal level suggested for the Spanish indicative version, cf. (2) above, would seem to cover the subjunctive version as well: an act coded by a declarative sentence with a speaker, an addressee and a content which involves an act of reference performed by means of the subclause (which serves as the second argument of the verb temo). The difference, it appears, is only reflected in the substitution of ‘communicated content’ on the representational level and its relation to the expression level. This appears to me less than ideal as an illustration of the aim of integrating discourse structures explicitly into the model. Ultimately, it may be a matter of notational variants, but it seems to me that if the difference between the two readings reflects an interpersonal choice, i.e. the choice between ‘saying what one feels’ and ‘conveying a mitigated piece of bad news’, the interpersonal level should also be the central point at which the difference is visible in the model. ‘Hedging’ would appear to be a good example of a subsidiary discourse act attached to a central act with face-threatening potential (with the two together constituting a move, according to the description given by Hengeveld). As such, hedging should be represented in the decision tree for ‘realization of intentions’ (cf., e.g. Brown and Levinson 1987: 60, showing the decision tree of someone who is considering a face-threatening act). Therefore the interpersonal level should specify two different choices, perhaps describable as follows (S = speaker, H = hearer, C = communicated content, T = ascription, R = reference): (a)

(b)

Communicative intention: Relieve one’s mind by entering into ‘troubles talk’, realized by Move: ‘report of emotional situation’, divided into a main act (A1): ‘S tell H of C = emotional state’ (T = fear, R1 = S) + subact (a1): ‘specify object of emotion’ (= R2) Communicative intention: Do one’s duty towards H, realized by Move: conveying relevant ‘bad news’, divided into a main act (A2): S tell H of C (T = ill, R = J) + subact (a2): mitigate by showing sympathy ‘S tell H of C (T = fear, R1 = S)’

Comment clauses and FDG

203

` I would like to proceed to a revision also in the representational level, which can be seen as the reverse of Hengeveld’s ‘downward layering’ innovation of the interpersonal level: just as the interpersonal level in Hengeveld’s model is extended downwards so that C, T and R overlap with items on the representational layer, I think representation should cover all linguistically represented meaning (including that which codes interpersonal choices). One example of why this would make sense is the analysis of the sentence I apologize as on the one hand a performative utterance or on the other hand as the answer to a question such as What do you do when you realize you’ve hurt someone? A linguistic model that aims to specify relations between grammar and discourse should provide a level where it is specified that the sentence codes the same represented content in both cases – as well as indicate that under appropriate contextual conditions, the sentence acquires the status of a direct coding of an interpersonal act rather than of informative content. An illustration might be: The performative reading: Interpersonal Level

Act: apology: S apologize towards H (No C) Layer of linguistically represented A-coding: content Decl (pres (apologize (S))

The informative reading: Interpersonal level

Act: Informative: S tell H C C= T, R Layer of linguistically represented C-coding: content Decl (Pres (apologize (S)) Figure 1. Performative and informative readings of I apologize.

The details of this representation are very preliminary, of course. The point is to bring out even more clearly than in Hengeveld's version what I see as a major advantage of the three-tier system, namely the possibility of matching structures at different levels of analysis in order to stress the similarities and differences (a perspective that has recently been in the

204

Peter Harder

focus of interest of the Danish Functional Grammar group under the heading ‘iconicity and structure’). Let us look at the analysis of the ambiguous English sentence (3) above. In terms of linguistically represented content, there is presumably no ambiguity in English: under both readings, it is a declarative clause with two argument positions, the latter being filled by an embedded clause. In the FDG analysis there is no level at which this identity of coded content is apparent, which is why (cf. above) I would like to extend the representational level to accommodate the linguistically represented content of I am afraid in its interactive function. If we do this, the interesting facts about the ambiguity can be brought out by interfacing the two different interpersonal readings with the same linguistically represented content. It is generally assumed (cf. Hannay and Mackenzie 1996: 114) that the main information is standardly put in main clauses, yielding an unmarked ‘iconic’ matching relation between the interpersonal level and the level of linguistic representation. If we see the information in the subclause as a subact in its own right (it may be somewhat implausible, but not, I think, out of the question; I propose it for purposes of illustration rather than as a hypothesis), this yields the ‘interface’ analysis of the main features of the ‘emotion report’ reading shown in Figure 2 below. In English, there is no linguistic trigger of the discrepant second reading; it is simply a fact of usage. In Spanish, where the same structural discrepancy is found, the indicative está provides a coding explanation, because it overrules the motivation for a subjunctive that emerges from the matrix verb, declaring everything with its scope to be ‘real’ – which is discrepant with the standard influence of a mental matrix verb.

Comment clauses and FDG

205

A: ‘Express emotion’ Interpersonal level

Intention: Relieve one’s mind by entering into ’troubles talk’

Level of linguistically represented content

Main act: Move: S tell H of C Express emotional state (C = state of being afraid T = afraid R1 = S) Decl (pres (fear,(S, arg2)))

Level of expression

Subact: Refer to object of emotion (R2 = J’s illness)

Arg2: That (pres(ill, J))

Main Cl (…)

Sub-Cl (…)

Move:

Main act:

Subact:

Convey relevant bad news

S Inform H that C (C= R) R = J’s illness T = ill R´ = J

Mitigate by sympathy S express C towards H (C = own feelings, T = afraid, R´´ = H)

B. ‘Hedged performative’: Interpersonal level

Level of linguistically represented content

Expression level

Communicative intention: Explain to H why J will not be available

Decl (pres (afraid (S, arg2))) Arg2: That (pres (ill, J)) Sub-Cl (…)

Figure 2. Analysis of I am afraid that John is ill.

Main Cl (…)

206

Peter Harder

5. Comment clauses and the coding of hedges in English Even if English does not have a mood distinction for this purpose, it provides its own way to code the ‘hedged performative’ reading unambiguously, i.e. by means of a so-called ‘comment clause’ construction (cf. Quirk et al. 1985: 1112). This construction involves removing the matrix clause from its fronted position, putting it at the end or in mid-sentence position: (4) (5)

John is ill, I am afraid. John, I am afraid, is ill.

Both possible positions exemplify the interaction of discourse principles and coding principles. FG provides slots for extraclausal elements both before and after the clausal positions, and the ‘tail’ position is generally recognized as a home for what may, iconically, be called ‘afterthoughts’ in relation to clause content. The clause-internal mid-sentence position is also used by ‘sentence adverbials’ of various kinds, including modal and attitudinal adverbials. In languages like German or Danish the mid-sentence position is also the home of particles that code attitudinal status. Such particles are absent in English, and the similarity between the content of certain particles in German and Danish and certain types of comment clauses in English reveals something about the coding pattern of English. When comment clauses occupy mid-sentence position they are standing at the beginning of a potential grammaticalization path that has not been followed in English: certain Danish particles have historically developed out of comment-clause-like constructions (cf. Davidsen-Nielsen 1996: 298). Thus the removal of the matrix clause ‘chunk’ to these positions is one way of bringing about a more motivated relationship between its interpersonal status and the way it is linguistically represented. The representational level needs to contain the matrix clause in order to capture this relationship. First of all, if we remove I am afraid from the representational layer, we lose the possibility of explaining what qualifies represented content of that kind to code a hedge, namely that it is an indication of sympathy with the addressee’s feelings. Moreover, it follows from the coding principles above that the content of I am afraid ... achieves a slightly different status in each case, while its relation to the subact of ‘hedging’ on the interpersonal level is constant. In one case it aligns itself (roughly speaking) with attitudinal elements in the clause, in another it aligns itself with extraclausal ‘dislocated’ elements. In spite of the differ-

Comment clauses and FDG

207

ence in structure, the English mechanisms share with the Spanish mood distinction the element of undermining the power of the matrix clause over the subclause – something which is important in understanding the relationship between grammar and discourse categories in such cases (cf. Bolinger 1968). The absence of that is also significant: that can be understood as an operator which (like ‘declarative’) takes scope over the entire proposition (cf. Harder 1995), reifying it into an entity (while ‘declarative’ uses it to convey a piece of information). Therefore the absence of that when the matrix clause is in the comment clause position is different from cases where that is optionally absent. By being preposed without the complementizer, the subclause is effectively promoted into a full declarative clause. The realignment of matrix and subclause does not bring about a full parallelism between discourse status and grammatical status. Distributional tests would still class it with matrix clauses. This is not so obvious with I am afraid, however, so for ease of illustration let us use the formal version I fear instead. In sentence (6): (6)

John is ill, I fear.

the verb fear is prima facie without a second argument, and the only way of bringing about a canonical dependency situation is by having a grammatical interpretation under which John is ill remains embedded in relation to the verb fear, while at the same time I fear is treated analogically with subclauses in terms of discourse status. An investigation of this field, I suspect, will lead to the need for many subdivisions of the representational layer of organization with many different links to both the expression and the interpersonal levels. The work of pursuing this angle will also require a reconsideration of the status of the sentence from a discourse perspective. Hansen (2001: 128) takes up this issue, quoting an example from German from Franck (1985: 234): (7)

Das war / also im Jahre 1907 / that was / then in:the year 1907 / ‘approx. So that was in 1907 I was born.’

bin ich was I

geboren. born

Here we see a constituent with what Franck called ‘double-bind’ status, serving both in a sentence begun in the previous instalment and a sentence completed in the following instalment. Hansen’s conclusion is that the sentence should perhaps be replaced as the relevant unit of analysis in favour

208

Peter Harder

of something smaller than sentences. As suggested in Section 1 above, the problem might also be that from the point of view of online production, there is no guarantee that the relevant units will match grammatical units point by point.

6. Final remarks As pointed out in Hengeveld (this volume), there are many problems that need to be solved in order to have a fully operational Functional Discourse Grammar. The suggestions I make raise the additional problem of how precisely to individuate the discourse categories that I would like to place in a more spectacular position in the interpersonal tier; pending that, it will always be unclear how much empirical substance there is in the interfaces of the model. However, problems of that kind need to be faced by anyone who would like to relate the categories of the code with categories of interaction. In spite of the problems, I see the three-tier model as a clear step forward in the quest to integrate discourse and grammar from a functional point of view. The main point I have defended is that if the interface potential of the model is to be fully realized, it would be an advantage to have a level at which linguistically represented content is fully specified, to be interfaced with an interpersonal level focusing more explicitly on interrelations between discourse elements. In cases where discourse and clausal organization go hand in hand, this will mean that information is systematically duplicated at the two levels – but rather than being a disadvantage, this is an intuitively obvious way of marking iconicity between discourse and grammar when it occurs. And in cases where there are discrepancies between structuring of linguistic content and structuring of discourse, these can be read straight off the model.

References: Berman, Ruth A. and Slobin, Dan I. 1994 Relating Events in Narrative: A Cross-linguistic Developmental Study. Hillsdale NJ: Erlbaum. Bolinger, Dwight 1968 Postposed phrases: An English rule for Romance subjunctive. Canadian Journal of Linguistics 14: 3-30.

Comment clauses and FDG

209

Brown, Penelope and Stephen C. Levinson 1987 Politeness. Some Universals in Language Usage. (Studies in International Sociolinguistics 4.) Cambridge: Cambridge University Press. Davidsen-Nielsen, Niels 1996 Discourse particles in Danish. In: Elisabeth Engberg-Pedersen, Michael Fortescue, Peter Harder, Lars Heltoft and Lisbeth Falster Jakobsen (eds), Content, Expression and Structure: Studies in Danish Functional Grammar, 283–314. Amsterdam and Philadelphia: Benjamins. Fortescue, Michael 1992 Aspect and superaspect in Koyukon: An application of the Functional Grammar model to a polysynthetic language. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered Structure and Reference in a Functional Perspective, 99–141. Amsterdam and Philadelphia: Benjamins. Franck, Dorothea 1985 Sentences in conversational turns. A case of syntactic “double bind”. In: Mario Dascal (ed.), Dialogue, 233–245. Amsterdam and Philadelphia: Benjamins. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Hannay, Mike and J. Lachlan Mackenzie 1996 Effective Writing in English. A Resource Guide. Groningen: Martinus Nijhoff. Hansen, Maj-Britt Mosegaard 2001 L’importance de l’analyse des interactions pour l’étude grammaticale de la langue. Revue Romane 36. 115–131. Harder, Peter 1995 Subordinators in a semantic clause structure. In: Betty Devriendt, Louis Goossens and Johan van der Auwera (eds), Complex Structures. A Functionalist Perspective, 93–118. Berlin and New York: Mouton de Gruyter. 1996 Functional Semantics. A Theory of Meaning, Structure and Tense in English. Berlin and New York: Mouton de Gruyter. Hengeveld, Kees 1987 Clause structure and modality in Functional Grammar. In: Johan van der Auwera and Louis Goossens (eds), Ins and Outs of the Predication. 53–66. Dordrecht: Foris. 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. This vol. The architecture of a Functional Discourse Grammar.

210

Peter Harder

Kroon, Caroline 1995 Discourse particles in Latin. A study of nam, enim, autem, vero and at. Amsterdam: Gieben. 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), Discourse and Pragmatics in Functional Grammar. 17–32. Berlin and New York: Mouton de Gruyter. Quirk, Randolph, Sidney Greenbaum, Geoffrey N. Leech and Jan Svartvik 1985 A Comprehensive Grammar of English. London and New York: Longman. Slobin, Dan 1987 Thinking for speaking. In: Berkeley Linguistics Society: Proceedings of the Thirteenth Annual Meeting. 435–445. Berkeley CA: Berkeley Linguistics Society.

Functional Grammar and the dynamics of discourse María de los Ángeles Gómez-González 1. Introduction From its inception to the present day (see Anstey, this volume) Functional Grammar (FG) has striven to satisfy the three criteria of functional adequacy it set for itself, i.e. pragmatic adequacy, psychological adequacy and typological adequacy.1 Accordingly, a prime concern for FG co-workers has been to provide linguistic descriptions with a strongly universalist bent so as to not only relate as closely as possible to the rules and principles governing verbal interaction, but also to dovetail with psychological models of linguistic competence and linguistic behaviour (Dik 1997: 13, 17). To better suit these demands, in particular to span the perceived gap between discourse/cognition and grammar (Mackenzie and Keizer 1991; Gómez-González 1998b, 2001), a number of FG representatives have felt it necessary to work towards a discourse- and cognitively oriented expansion of the standard model, which is, broadly, a lexico-semantically based, three-staged bottom-up layered framework converting lexical elements and terms into pragmatico-semantic constructs, converted in turn into prephonetic strings through the operation of expression rules (Dik 1997: 58ff.; Hengeveld 1997). Crucially, in this standard model conceptual unity is entailed between discourse and grammatical analysis by treating each utterance as a mini-discourse (cf. also Moutaouakil, this volume). By contrast, Kroon (1997), Bolkestein (1998), Steuten (1998a, b), Vet (1998), Van den Berg (1998), Connolly (1998, this volume) and Fortescue (this volume) assume that grammar and discourse are in essence incommensurable, and accordingly assign corresponding modules to grammar, which is static and out-of-time, on the one hand, and to discourse, which is dynamic and ongoing in time, on the other. A different position is taken by Nuyts’s (1992, this volume) Functional Procedural

212

María de los Ángeles Gómez-González

Grammar (FPG) and Mackenzie’s (1998, 2000, this volume) Incremental Functional Grammar (IFG), which draw closer attention to work in language production and comprehension. These represent two alternative, though compatible top-down attempts to accommodate grammatical description to the dynamic nature of discourse organization, conceived in both programmes as an interactive, flexible process that unfolds in real time in ‘chunks’ or in mentally digestible units. FPG claims that linguistic structures emerge from abstract cognitive factors or conceptualizations which are fundamentally different from linguistic (lexico-semantic) organization, yet are still strongly language-bound (based on predicateargument constructs). The whole process of language production is claimed to be a matter of gradually singling out packets of information which get coded in single utterances (rarely full clauses, especially in speech), using mid-to-short-term memory to adapt long-term conceptual knowledge to the actual communicative situation and to the process of language production. And it is only the last part of this process, the single utterance, that grammar actually deals with (Jackendoff 1987, 1990). IFG, in turn, linking up with research into starting points (MacWhinney 1977), posits an incremental or gradual left-to-right build-up of chunks of information (utterances or discourse acts), from minimally a P1 in Focus position, the first increment or subact (the holophrase) which exploits the primacy effect or prominence of first mention, to rightward expansions or linear sequences of subacts, with late, possibly last placement of the Focus in conformity with the recency effect, according to which information offered last is best retained (cf. Gómez-González 2001). Concomitant intuitions are found in Bakker and Siewierska’s (this volume) rendering of expression rules as dynamic, and working top-down, depth-first, and left-to-right as in actual language production. Hengeveld’s Functional Discourse Grammar (FDG; this volume) seems to accommodate the insights gleaned from the aforementioned broad-brush accounts. With a view to constructing a new architecture for FG, FDG reveals itself as a top-down hierarchical modular approach, in which entities of a purely pragmatic Interpersonal Level (IL) at the top (acts/moves of reference and ascription) and purely semantic entities of varying orders at a Representational Level (RL) next down are mapped by expression rules (ER) onto the Expression Level (EL) at the bottom, i.e. the phonological string that eventually triggers articulation. Speakers – Hengeveld claims – draw on the cognitive component, covering (long-term) knowledge including communicative and linguistic competence, at each of the three levels, while the communicative component, containing (short-term) linguistic co(n)textual

FG and the dynamics of discourse

213

information, is fed by the interpersonal and expression levels and simultaneously feeds the representational level in order to enable later reference to earlier acts and expressions. This discourse- and cognitively oriented expansion of FG (see also Harder, Connolly, Verstraete, this volume) parallels the trend of other functionalist approaches, notably Systemic Functional Grammar (SFG) and West Coast Functionalism (WCF; e.g. Cognitive Grammar, Construction Grammar, Emergent Grammar, Conversational Analysis, etc.),2 as well as the more formally-oriented proposals of e.g. Lexical Functional Grammar (LFG), Generalized Phrase Structure Grammar (GPSG) or Head-Driven Phrase Structure Grammar (HDPSG). Some common ground thus seems to be emerging, seeking speaker-oriented models that treat the speech production process as running from intention to articulation (Levelt 1989), thereby assigning discourse and cognition corresponding levels of representation. In these emergent approaches linguistic expressions (and by implication universals) seem to arise out of the interplay of languagespecific properties (e.g. lexico-grammatical semantics, constructional meaning) and factors that are not specifically linguistic, such as constraints on attention and memory, the dynamics of discourse, co(n)textual characteristics, channel restrictions etc., as processed by the human cognitive system, despite the limitation that at present there is no full-blown answer as to how this system works or what it looks like. Possibly to some extent sprouting from this limitation, there are aspects of FDG – as Hengeveld concedes – that deserve further elaboration in order to become a fully-fledged framework. Indeed, in line with Bakker and Siewierska (this volume), one misses a more detailed account of the role of the lexicon in the model, of the processes whereby semantic representations come into existence, and of the reasons why left-to-right expansions are only assumed for the interpersonal level when we have every reason to suspect that these restrictions equally affect the entire production process. Furthermore, FDG claims that in the coding of his/her communicative intention, the speaker relies on entities of the third, second, first or zero order (i.e. propositional contents (p), states of affairs (e), entities or individuals (x), and properties (f)), but at the same time the possibility is admitted that all entity types may also be expressed in a non-hierarchical way (i.e. through lexical items). One could find in this last procedure an argument to support the view that, instead of a threefold hierarchical ordering of IL, RL and EL, only the latter is required (see Nuyts and Fortescue, this volume, for concomitant suggestions to minimalize the model), with the lexicon and the grammar forming a continuous spectrum of symbolic assemblies or

214

María de los Ángeles Gómez-González

constructions, apparently in conformity with evidence reported from language processing, language acquisition, aphasia and categorization (Bates and Goodman 1997).3 As a corollary, there would be no need to posit underlying structures or derivations, nor (expression) rules to link the different components. Different substantiations of this kind of non-hierarchical non-modular framework are subscribed to by the exponents of WCF. Roughly, these approaches, which are, as has been remarked, in many respects analogous to FDG, (a) defend a materialist model of perception with an empiricist functionalist and universalist orientation, and (b) postulate real-time dynamic programmes that discard the division between a core grammar and a periphery grammar, under the assumption that all constructions, whether relatively recurrent or highly idiosyncratic, stand on an equal footing for description, resorting to, among other things, (c) non-discrete categories or prototype theories, (d) monostratal or ‘surface level’ syntactic descriptions, and (e) making no sanctioned distinction between grammatical knowledge, knowledge of language use and other sorts of knowledge under the assumption that “knowledge of language is knowledge” (Goldberg 1995; Langacker 1987, 1991, 2001b; Noonan 1999, Kay and Fillmore forthc.). This latter stance (e) apparently flies in the face of Hengeveld’s outline of FDG (this volume), in which Cognition and the Communicative Context are presented at the sides of the schema as two separate, though connected modules. This representation could lead to the assumption – counterintuively in my view – that Cognition (i.e. (long-term) knowledge) and communicative context (i.e. (short-term) linguistic co(n)textual information) can be dissociated or can operate independently of each other. Instead, we have every reason to suspect that in encoding and decoding strategies the communicative context is filtered through and interpreted by the speaker’s cognitive system: the latter encompasses the former as it were, both with a certain amount of independence but nonetheless coordinated, the underlying rationale being that processing proceeds simultaneously on multiple time scales, and with respect to numerous parameters. Moreover, it seems that oneway arrows running from and towards Cognition, the Communicative Context and the three levels of grammar in different guises infuse the model with a rigid one-way directionality that impoverishes the supposedly interactionist (bi-directional or multidimensional) dynamicity that the model strives to confer to the cognition-discourse-grammar relationship. In sum, a modification of this design is here called for to better symbolize Hengeveld’s claim that “the three levels interact with a cognitive component and with a communicative component” (this volume: 3).

FG and the dynamics of discourse

215

Also in this volume, Cornish, Inchaurralde and Hengeveld himself stress the need for fleshing out the Cognition and Communicative Context components of FDG, but without moving away from our main object of study: the linguistic message. Of interest in this context is Mackenzie’s (this volume) suggestion that the interaction and expression levels should be viewed as incrementally operative in real time, while the representational level would act as a buffer between these two real-time processes, serving to organize the material into linguistic form. Convergent with this re-interpretative spirit and with the ongoing discussion on the discourseand cognitively oriented expansion of FG (and other approaches), I would argue for an Incremental Discourse Cognitive Grammar (IDCG), as outlined in Figure 1 below. Gathering inspiration from Hengeveld’s FDG, Nuyts’s FPG, Mackenzie’s IFG, and cognitive grammar (CG; Langacker 1987, 1988, 2000, 2001b), IDCG pictures the connection between discourse expressions and cognition as a dynamic real-time process in which information is incrementally presented in coherent ‘packets’ of digestible size. Broadly identifiable with intonation units,4 information packets are represented as framed (or co(n)textually delimited) bi-dimensional events, comprising both conceptualization (embracing IL and SL) and expression (EL), each with various co-ordinated planes. EL includes discourse expressions of all sorts, segmental, suprasegmental, and paralinguistic or kinesic. Conceptualization incorporates the interlocutors’ apprehension of the ground (including the speech event itself, the speaker (S) and addressee (A), and their immediate spatio-temporal circumstances), on the one hand, and the current discourse space (CDS), on the other, that is, the mental space comprising the elements and relations construed as being shared by S and A at a given moment in the flow of discourse. Therefore, in this model CDS embodies speakers’ shared knowledge, which in turn includes the context of speech involving physical, mental and socio-cultural variables. At the centre of the speech context is the ground, where S and A are engaged in a co-operative viewing of some facet of any world (real or imagined).

María de los Ángeles Gómez-González

Figure 1. An outline for an IDCG

216

FG and the dynamics of discourse

217

In this process both S and A, whether playing an active or reactive role, have to deal with gradual or incremental conceptualizations and expressions concerning their focus of attention (symbolized as →). In other words, speakers’ decoding and encoding strategies involve timeconstrained conceptualizations and expressions of lexico-grammatical constructs that are contextually relevant, that is, which respond to text-building and speech management strategies (i.e. information structure, textual progression, genre characteristics, conversational routines, etc.) A finergrained elaboration of IDCG lies well beyond the scope of this chapter. Space constraints oblige us to concentrate on the implications that this model may bring to the dynamics of discourse. Section 2 revises what in an IDCG are felt to be the most relevant analytical tools to explore this issue, namely the notions of choice, perspective and prominence as manifested in three important attention-related dimensions: Topic, Focus, and the Theme zone. The chapter closes with a summary (Section 3).

2. The dynamics of discourse Effective communication is an essentially strategic encoding and decoding process, in which both S and A, whether playing an active or reactive role, are continuously monitoring choices as to what is being said, what they want to say, and how they want to say it, generally following two basic cooperative principles: ‘be orderly’ and ‘be consistent with the existing topic framework’ (Grice’s 1975 maxims of relevance and quantity). Inherent to the notion of choice is therefore the fact that all languages offer (minimally) distinct available options at different levels – such as those exemplified in (1) and (2) – which derive their distinctiveness fundamentally from their syntagmatic properties or the associative relationships holding among their constituents, on the one hand, and from their paradigmatic potential on the other, that is, from the oppositive or contrastive relationships established with other units. It is these two different kinds of relationships, paradigmatic and syntagmatic, that make one option more or less appropriate as a motivated choice for a particular discourse co(n)text. (1)

a. and ∨there ⏐ on the /table ⏐ where my ∨newspaper had been ⏐ was my packet of \biscuits ⏐ (LIBMSECGPT03: 046-47)5 b. and my packet of biscuits was there, on the table, where my newspaper had been.

218 (2)

María de los Ángeles Gómez-González a. ⏐ there were \three of us in the \car ⏐ \all °rather \nervous ⏐ (LIBMSECAPT04: 018) b. in the car (there) were three of us, all rather nervous. c. three of us were in the car, all rather nervous. d. three of us were all rather nervous, in the car.

In what follows it will be argued that, even though truth-conditionally equivalent or describing precisely the same objective situation, different discourse expressions like those in (1) and (2) above correspond to different conceptual organizations that imply different directions of mental scanning, i.e. different paths of mental access, in the conceptualizers’ minds (both S and A) when constructing a full conception of a given SoA.6 In other words, different discourse expressions establish different viewpoints (or vantage points) and determine different perspectives (global or local) on the experience being constructed. For one thing, discourse involves a succession of attentional frames, each of which represents the scene being ‘viewed’ and acted on by S and A at a given instant, and it is this viewing process as perspectivized by S and A that determines the ‘camera angle’ or ‘camera movement’ of discourse, structuring it into separate attentional units or episodes (see also note 3). One particular framing may be unmarked, or prototypical, in contrast with an array of comparatively marked alternatives, but all may be familiar and conventionally sanctioned in specific situations.7 A major dimension of mental accessing is the degree of discourse prominence accorded to various elements within a conceptualization, most importantly topical, focal and thematic prominence. We will here not go into the nitty-gritty of the categories of Topic, Focus, Theme – and other related concepts – as a detailed analysis has been given elsewhere (GómezGonzález 2001). Rather, in the following subsections we shall offer a sketchy presentation of these notions – for the most part coinciding with FG accounts, although in some occasions departing from or complementing them, especially in the case of Theme – so as to demonstrate how these different kinds of prominence affect the Focus of attention of discourse in keeping with the newly posited IDCG.

2.1. Topic Dik’s view (1997: 312) will be endorsed that topicality concerns the status of those entities ‘about’ which information is to be provided or requested in the discourse. However, an expansion of the FG notion of Discourse Topic (D-Topic) is suggested here which will entail a relationship rather than an

FG and the dynamics of discourse

219

entity or a thing. As the reader will remember, Dik (1978: 19) explains that Topic is “the entity ‘about’ which the predication predicates something in the given setting”. A threefold implication follows: (a) only one Topic per predication is allowed; (b) this function can only be assigned to terms (i.e. discourse referents/participants, mainly Subjects or Objects), and (c) topical elements are endowed with different degrees of referential accessibility in terms of being (i) Given Topics (GivTop), (ii) SubTopics (SubTop), (iii) Resumed Topics (ResTop), or (iv) New Topics (NewTop): the first presentation of a D-Topic is a NewTop,8 and an entity so introduced becomes a GivTop if re-introduced; likewise, an entity which is mentioned, temporarily neglected and later revived is a ResTop, while an entity inferred from other entities is labelled SubTop. This account of D-Topics in terms of entities endowed with different degrees of referential accessibility raised a number of problems explained in detail elsewhere (Gómez-González 1998b, 2001: 156–168), most importantly: (a)

(b)

(c)

the assumption that only entities qualify for D-topical status, while a person’s pragmatic information also includes the SoAs in which those entities play a role; the apparent identification of Topic with predication-internal entities, as opposed to Theme and Tail (both predication-external), for it seems that P1/P2 assignment neither is a discrete distinction nor can be upheld as the only means of distinguishing these categories; the difficulties involved in identifying or distinguishing the four Topic types.9

In an attempt to surmount at least some of the aforementioned complications, I will here resort to a more comprehensive, more cognitivelyoriented notion of D-Topic, from Kemmer (1995: 58): a prominent conceptualisation which acts as a kind of cognitive anchoring point; other [...] conceptualisations are brought into the discourse by virtue of their perceived ties to the reference point.

Crucially D-Topics of this kind, whether local or global, entail relationships rather than things or entities.10 This interpretation reflects the fact that connected discourse comprises a succession of attentional frames – a series of intonation units mapped onto clauses, or parts of clauses – each profiling a relationship (Langacker 2000, 2001a, b; Gómez-González 2001). At the

220

María de los Ángeles Gómez-González

basis of this relationship, an expression evokes a certain body of conceptual content, its ground or base. This directs attention to some particular substructure – the profile – construed as the relationship (subsuming the entity or entities) that the expression designates or refers to. By way of illustration, (1a) and (1b) above profile an abstract presentational relationship, (1b), (2c), and (2d) one of spatial attribution, and (2a) one of ‘existence’ (see Sections 2.3 and 2.4 for further reference to these alternate relationships). A D-Topic therefore refers schematically to the relationship (or thing) profiled. It has a prospective, or forward-looking potential by virtue of signalling that the profiled relationship or entity will function as a conceptual reference point for purposes of interpreting a subsequent proposition or establishing mental contact with another reference point (its target) within – as is most likely – or outside its cognitive schema, or dominion, that is to say: ... any system of concepts related in such a way that to understand any of them you have to understand the whole structure in which it fits; when one of the things in such a structure is introduced into a text, or into a conversation, all of the others are automatically made available. (Fillmore 1982: 111)

In addition, D-Topics have a retrospective or backward-looking dimension, in the sense that they normally refer to information that is already accessible in the discourse co(n)text. This is implied by the very notion of building a coherent or cohesive structure for, when structure is added, it is usually anchored to what has already been built, by virtue of the conceptual overlap. It follows that – as implied in FG – D-Topics tend to be endowed with the features of identifiability and activation. Identifiability accounts for the difference between referents for which the speaker assumes a file has already been opened in the discourse register and those for which such a file does not yet exist (Chafe 1976). What counts for the linguistic expression of this cognitive distinction is whether or not the speaker is able to pick out a referent from among all those which can be designated by a particular linguistic expression and identify it as the one which the speaker has in mind, provided that, given the appropriate discourse context, a referent is more or less permanently stored in the long-term memory of S and A, and can be retrieved without difficulty at any particular time. A referent may be identifiable either because it has been mentioned in a discourse (i.e. through anaphoric reference) or because it is either visible or otherwise sa-

FG and the dynamics of discourse

221

lient in the speech setting or because it is ‘inalienably possessed’ or otherwise anchored in the individuality of one of the interlocutors (i.e. through deictic reference, e.g. those ugly pictures or the woman in the green hat over there, your left leg, or my sister's second ex-husband). All instances of identifiability and their expression by a single grammatical category assume the existence of a cognitive schema, or frame. Activation, on the other hand, evokes Chafe’s (1987: 22) idea “that our minds contain very large amounts of knowledge or information, and that only a very small amount of this information can be focused on, or be ‘active’ at any one time”. Accordingly, three following degrees of activation are posited: (a) (b)

(c)

active, designating “currently lit up” information that is, as Chafe has it, “in a person's focus of consciousness at a particular moment”; semiactive (or accessible), designating information that is in a person’s peripheral consciousness or in their background awareness as a result of: (i) deactivation from an earlier state (i.e. textual accessibility); (ii) inference from a cognitive schema or frame, that is, from some other active or accessible element in the universe of discourse (i.e. inferential accessibility); and (iii) presence in the text-external world referent (i.e. situational accessibility). unused, designating inactive information that is “currently in a person's long-term memory, neither focally nor peripherally active”.

Note that, although D-Topics normally entail some degree of identifiability and/or accessibility, not all identifiable or accessible items become D-Topics, for the latter must also show a prospective dimension (see note 13). The problem arises how to identify these cognitively oriented DTopics. For this sketchy presentation it suffices to assume that D-Topics generally profile relationships normally anchored in the situational and linguistic co(n)text, in conformity with natural paths of mental accessing (see notes 5 and 14). This grounding process occurs in successive attentional frames (intonation units with corresponding Focus spans), whose arrangement is basically linear and affected by the order of presentation (in principle what is presented first will profile what follows, see Section 2.3), and which are superimposed on structural elements, whose arrangement is both linear and hierarchical. There is a natural tendency for attentional frames to coincide with grammatical constituents, but it is only a tendency, not an inviolable principle. Conceptual organization need not be directly mirrored in grammatical constituency, for grammar is a tool for building up

222

María de los Ángeles Gómez-González

to – and symbolizing – complex conceptualizations, but it is not to be identified with those conceptualizations. Further, this conception of D-Topic does not always point to one particular element in a proposition, or to one particular constituent. For one thing, D-Topics also allow for degrees of topical relevance and degrees of activation and identifiability that are incrementally built up in the scaffolding of discourse, not to forget the elusiveness and fuzziness implicit in the criteria upheld to assign D-topical status, i.e. ‘aboutness’, ‘activation’, or ‘identifiability’.

2.2. Focus Here we shall depart somewhat from FG practice, which attaches Focus status to those pieces of information which represent the “relatively” most important or salient information (Dik 1978: 19; 1997). To my mind, this description is better captured under the title ‘Focus of attention’ (Section 2.4) – equivalent to the FG account of focality11 – because discourse information may become prominent (‘important’ or ‘salient’) not only as a result of its focal status, but also, among other things, for its topical quality and its thematic (vs. rhematic or final) arrangement, each of these dimensions involving corresponding lexico-grammatical, phonological and paralinguistic expressions (cf. Cornish this volume). In this chapter Focus represents a purely phonological term, equivalent to Focus accent. It designates intonational prominence in an informational unit (see note 3), which may be associated not only with information packaging but with other linguistic dimensions as well (e.g. illocution, stringbased deaccenting) and which may also be a consequence of non-linguistic effects. As indeed stated in FG, Focus accent is taken to represent the prosodic means whereby – in combination with morphosyntactic devices such as morphological markers, word-order variation and special lexicogrammatical constructs – the syntactic domain of Focus (the Focus domain) is expressed. Accordingly, a semantic element belonging to the Focus component of a pragmatically structured proposition is said to be in Focus or focal (as opposed to being ‘in the presupposition’ or ‘presuppositional’), regardless of whether the constituent coding it carries an accent or not. For example, in I went to the movies as a reply to What did you do last night?, both the (relatively) unaccented constituents went and to together with the accented constituent movies are in Focus (Lambrecht 1994: 209) [emphasis mine].12 And, again as remarked in FG, Focus accent may fall either upon New (inactive) information or upon Given (active) information.

FG and the dynamics of discourse

223

Examples given by Siewierska (1991: 174) are Sonia and Joyce came to help me. Sonia worked like mad, but Joyce was horribly slow; or I heard that Peter got married. Peter’s married. How amazing! I don’t believe it! [emphasis mine]. In principle, I see no objection to Dik’s (1997: 315ff.) distinction between New Focus (NewFoc) and Contrastive Focus (ContrFoc), broadly evoking the widely agreed differentiation between broad and narrow Focus (Selkirk 1984; Lambrecht 1994; Ladd 1996).13 However, I would contend that besides assigning ContrFoc a number of functionalities (Parallel Focus and Counter-presuppositional Focus, which comprises Replacing Focus, Expanding Focus, Restricting Focus and Selecting Focus), a further subdivision within NewFoc should be also posited in order to account for the structural and attentional domains of two construction-types that are recurrent across languages, namely utterance-Focus (UttFoc) and predicateFocus (PredFoc) (cf. Lambrecht 1994; Vallduví and Engdahl 1996; Lambrecht and Polinsky 1997). In a UttFoc construction (e.g. (3) below) both the Subject and the predicate are in Focus, i.e. the Focus extends over the entire proposition. The purpose of the assertion is to express a proposition which is linked neither to an already established D-Topic nor to a presupposed open proposition.14 In other words, crucial in UttFoc constructions is the absence of a presupposition attached to either the Subject or the Predicate, and resultingly the identity of assertion and Focus (cf. Kuno’s 1972 ‘neutral description’; Allerton and Cruttenden’s 1979 ‘all-new utterance’; and Kuroda’s 1972 ‘thetic sentence’). UttFoc has been attributed to: (a)

(b) (c)

impersonal expressions, that is, events, states, situations or facts in which no referential entity is present and therefore nothing can be said about it; presentative constructions, which introduce entities, but fail to report an event about them; any state of affairs presented as a compact whole representing nothing but new information.

It is the absence of a presupposition and the identity of assertion and Focus that distinguishes UttFoc in (3) both from the PredFoc construction in (4) and the ContrFoc in (5) below. In the PredFoc category the Predicate is in Focus and the Subject (and possibly some other argument or adjunct) is within the presupposition: the purpose of the assertion is to pragmatically predicate some property of an already established discourse referent;

224

María de los Ángeles Gómez-González

whereas in the ContrFoc category, the Subject (or some other argument) is in Focus and the predicate is within the presupposition (examples are taken from Lambrecht and Polinsky 1997 [emphasis in original]). (3)

(4)

(5)

UTTERANCE FOCUS (UttFoc): Context sentence: Why didn’t she come work today? Sentence: Her SON is sick. PREDICATE FOCUS (PredFoc): Context sentence: Why didn’t she come to work today? Sentence: Her son had an ACCIDENT. CONTRASTIVE FOCUS (ContrFoc): Context sentence: Why didn’t she didn’t come to work today? Sentence: Her SON is responsible.

Of importance here is that this threefold distinction of Focus types in terms of UttFoc, PredFoc and ContrFoc seems to fit in well with FG work, that is, Hannay’s (1991, 1998) three message modes, i.e. Grounding, Neutral and Reaction, which also attach corresponding attentional and structural properties to these three utterance types (see Section 2.4).

2.3. The Theme zone For compelling evidence and reasons explained in detail elsewhere (Gómez-González 1998a, 2001), behind which is the premise that the speaker’s choice of point of departure, or Theme, represents a strategic move in natural language (Mackenzie 1998, 2000, this volume) giving orientation or anchoring for what is to follow in the background of a co(n)text, the Theme zone (to be underlined in the following examples) will be taken to extend as far as an initial ‘experiential’ constituent, the experiential Theme (ET), i.e. (a) Subject, (b) Process/Verb, (c) Complement/Object or (d) Circumstance – all with realizations ranging from words, phrases to (parts of) clauses – including any (non-restrictive) postmodification.15 Cross-cutting illocutionary moods and Theme choices, a thematic arrangement will be unmarked when ETs correspond to the first element of each mood type, and it will otherwise be marked. Accordingly, following Halliday (1994), we shall assume that in English the unmarked Theme of declaratives is the Subject (e.g. Little Bo-peep has lost her sheep, 1994: 43),16 the WH-element in exclamatives (e.g. how cheerfully he seems to grin, 1994: 47), and in interrogatives, either an auxiliary (carrying the expression of polarity) plus the Subject, in yes/no questions (e.g. can you find me an acre of land?, 1994: 48), or the WH-element in WH-questions

FG and the dynamics of discourse

225

(who killed Cock Robin?, 1994: 48). All other thematic choices will be marked – e.g. (1a), (2b) above in contrast with the other unmarked counterparts. In addition, also relevant for the ongoing discussion is the fact that the Theme zone may house different functional loads. I dub ‘extended multiple Themes’ (EMTs) those cases in which ETs are preceded and/or followed by elements with a textual or an interpersonal orientation, which I shall call ‘textual Themes’ (TTs) and ‘interpersonal Themes’ (ITs) respectively, under the assumption that the latter mark a boundary between the Theme zone and the Rest (or Rheme). (6) – (9) below illustrate different kinds of EMTs: (6)

(7)

(8)

(9)

a. ⏐ a —journalist ∨convoy ⏐ of about \fifteen ∨cars ⏐ was a\ssembled ⏐ and — lights ∨flashing ⏐ it was ∨dusk ⏐ we —tore ‘down the \road ⏐ ‘holding —white ∨flags ⏐ —out of our \windows ⏐ (LIBMSECA04: 080) (ET) (TT^ET^ET) b. ⏐ yet —she for her /treasure will en—dure ⏐ and ∨tremble ⏐ and —so find — peace ⏐ that \passeth our under\standing ⏐ (LIBMSECH04: 033) (TT^ET^ET) c. ⏐ /well ⏐ _em ⏐ \those that \come from agricultural \families \will ‘work ⏐ ∨on the ‘land⏐ (LSECJPT06: 279–80) (TT^TT^ET) a. ⏐ and of \course they ‘lived in ∧ ‘segregated ∨quarters ⏐ which was \maybe O‘K in the be∨ginning ⏐ but ∧ \later there was a ∨doorman ⏐ on the ∧ ∨women's building ⏐ ‘and they ∧ just —weren't a‘llowed to ∧ ‘visit at \all ⏐ (LIBMSECJ06: 016– 017) (TT^IT^ET) b. ⏐ because ‘obviously Su∨dan is ⏐ ‘by and large a ∨Muslim country ⏐ (LIBMSECC06: 297) (TT^IT^ET) a. ⏐ West∨morland for e°xample ⏐ became par\ticularly ∨passionate ⏐ when \ talking about the °influence ⎯television re∨porting from Viet°nam ⏐ had \had on the ∨White °House ⏐ in the °late \sixties ⏐ (LSECAPT03: 030) (ET^TT) b. ⏐ —one figure inci‘dentally ⏐ you \may find \useful for ⏐ ∨tax ‘purposes ⏐ (LIBMSECF02: 064) (ET^IT) a. ↑and in ⎯Mara\dona ⏐ without \doubt ⏐ they °had the \star of the ∧ °compe°tition ⏐ (LIBMSECJPT01: 065) ) (TT^ET^IT) b. ⏐ the \Shi'ites in∨side ⏐ are \not the ‘hard cases of the _terrorist \world ⏐ in‘deed ∨ most of them ⏐ ‘probably did ‘little /more ⏐ than ‘be in the —wrong /place ⏐ at the _wrong \time ⏐ (LIBMSECA07: 028) (TT^ET^IT) c. ⏐ thus one /third for e‘xample ⏐ was ∨written as ⏐ —three _vertical \strokes ⏐ ↓with the ∨mouth symbol placed a\bove them ⏐ (LIBMSECD03: 040) (TT^ET^TT ⎯the two TTs being of different kinds⎯)

226

María de los Ángeles Gómez-González

The excerpts above reveal a number of most important things. The first is the vast richness entailed by the Theme zone, in terms of function (housing discourse markers, discourse connectors, vocatives, attitudinal markers, circumstantial specifications, and/or entities of various orders), syntactic realization (with the potential of having complex realizations, expanded through relationships of subordination, coordination or apposition) and volumewise (having the possibility of being packed within one intonation unit or extending over more than one unit). Incidentally, the same potential applies to the Rest (or Rheme). For this reason, here both the Theme and Rheme zones are viewed not as a discrete locations, with fixed boundaries, but rather as incremental and flexible zones entailing different kinds of wave-like or pulse-like points of prominence. Theme profiles an orientational relationship as a result of the recency effect: discourse scaffolds normally ground what is to be said on the previous context or the situational context. By contrast, Rheme exploits the recency effect, i.e. the saliency of what is said last, in conformity with the principles of End Focus (the late placement of focal accent) and End Weight (the last mention of weighty units). Both kinds of prominence conspire to place active discourse information within the Theme zone and comparatively less active material in the Rheme, following the principle of Functional Sentence Perspective, i.e. the Given-before-New array of information. But, again, these are only tendencies, not inviolable principles. Now, to go back to the Theme zone and to summarize, its flexibility admits: (a) recursion (paratactic, hypotactic or appositive) within each of its three orientational subfields, ET, IT, TT – as shown in (6) above – and as a corollary, (b) either a P1 realization within a larger intonation frame (e.g. (2a), (6b), (7a), (7b)), or a P2 realization, that is, a broadly ‘clauseexternal’ realization, in which case Themes either have one attentional domain of their own (e.g. (7a), (7b), (8a), (8b), (9c)) or otherwise extend over more than one frame (e.g. (1a), (6a), (6c), (9a), (9b)). The presence/absence of intonational integration at the beginning and/or end of an utterance implies the presence/absence of corresponding Focus spans or message peaks (Bolkestein 1998; Hannay 1994). Hence, Themes that are not marked off intonationally result in their phonological and conceptual ‘compression’, thereby profiling the subsequent discourse relationship as a single, complex characterization: more has to be squeezed into a single, limited span of processing time, resulting in a somewhat less articulated realization. By contrast, detached Theme zones are incrementally coded as separate attentional gestures, thereby enhancing their cognitive salience and that of the elements within their scope, if only by

FG and the dynamics of discourse

227

being accorded intonational prominence, more processing time and thus a fuller realization. The two implementations are quite distinct both experientially and at the level of motor activity. And, most importantly, they show that – as noted above – different structural configurations involve some differences in conceptual organization. In keeping with this, another issue that we will consider before closing this section is the cognitive repercussions of the different arrangements of the orientational subfields, i.e. ET, IT, TT, within the Theme zone. Here we shall assume that the order of expression always induces a conceptual ordering of the notions symbolized, as one facet of our dynamic apprehension of linguistic meaning. But, needless to say, not every facet of dynamic conceptualization is necessarily correlated with expression order; neither does the sequence of presentation in discourse necessarily coincide with the chronological ordering implied. The data handled in Gómez-González (1998a, 2001) reveal that EMTs tend to display the organization TT^IT^ET^IT^TT – as shown in excerpts (6) – (9) above and Figure 2 below – with the constraint that, if TTs and/or ITs occur in the pre-ET subfield, then items of the same kind do not recur in the post-ET subfield (for a classification of the types of TTs and ITs see Gómez-González 2001: 110, 180–185). Further, an utterance may minimally consist of one kind of Theme only, be it TT, IT, or ET, which will therefore be in Focus (e.g. Tomorrow?, Fine and Great) as possible reactions to Maria is coming for a visit).

2 (TT)

^

1 (IT)

PRE-ORIENTATION

^

y x H ET

^

INTERNAL ORIENTATION

1 (IT)

2 (TT)

POST-ORIENTATION

narrowest scope experiential meaning interpersonal meaning textual meaning widest scope Figure 2. The Theme zone’s maximal expansion

^

228

María de los Ángeles Gómez-González

The unmarked array of EMTs illustrated in Figure 2 represents y (TT), x (IT) and H (ET) as being arranged in relation to their scope of influence, suggesting that ET, IT (1) and ET (2) have increasing scopes of influence, each within the span of the next. Thus ETs profile a backward-looking and forward-looking relationship pointing to the message itself, but with ITs, this profiling involves in addition the interlocutors (S and A), while with TTs the profiling expands to the global logical organization of discourse. All other arrays are claimed to be marked, or less frequent, or otherwise determined by item-specific mobility conditions. Note, for example, the flexibility of however, a textual item, which may appear (a) in the thematic pre-orientational field (e.g. However, the man didn´t come), (b) in the thematic post-orientational field (The man, however, didn´t come), (c) crammed within an EMT (The man, however, the one we saw yesterday, didn´t come), or (d) outside the Theme zone (The man didn´t come, however). Consistent with the unmarked array of the Theme zone, three further orientations will be attributed to EMTs: the high-low orientation, the lowhigh orientation and the combined type (cf. Smits 2002). High-low orientations are those profiled when TTs and/or IT occur in the pre-orientational field, normally in decreasing scopal spans, in order to provide a textual and interpersonal ground according to which the subsequent discourse is to be interpreted. Illustrations can be found in the EMTs in (6) and (7) above. By contrast, low-high orientations obtain when TTs and/or ITs occur in the post-orientational zone, as shown in (8) above. Normally arrayed in increasing scopal spans, and in corresponding intonation units, postorientational ITs and/or TTs separate ETs from what we have dubbed the Rest, thereby distributing the focus of attention between both chunks of information, the ET and the Rest, which are zoomed in on and accordingly receive cognitive salience (more processing time and thus a fuller realization) and focal prominence. In mixed orientations – as the label suggests – the functional capacities of both high-low orientations and low-high orientations are combined (as in (9) above). In the following section we shall expand the view that the different kinds of orientation profiled by the Theme zone, whether representing marked and unmarked choices, entail different instructions for the camera angle of discourse.

FG and the dynamics of discourse

229

2.4. Focus of attention Resulting from the merging of attentional framing and structural constraints, linguistic representations are here viewed as instructions to modify the S’s and A’s Focus of attention or current discourse space in particular ways (cf. Hannay 1991, 1998; Harder 1996; Vallduví and Engdahl 1996; Gómez-González 2001). These instructions – and naturally the corresponding expressions – are here regarded as interpersonally-oriented textual choices that are cognitively rooted, on the basis of their broadly bidimensional functionality, i.e. their perspective-creating capability, and their action as discourse markers, packaging information into digestible units, both aspects influencing the way reality is monitored (encoded or decoded) by the human (S’s and A’s) cognitive system. In unmarked cases, the instructions given by discourse expressions tend to seek for sameness and usually reflect a static viewing: a constant perspective that entails instructions of continuity of discourse participants, of referential predictability and of foregrounded or discourse-focused material. By contrast, as a deviation from the norm, marked discourse expressions are natural symptoms of a subjective viewpoint and signal an invitation for a dynamic viewing, entailing, among other things: a topic change, a change in focus of attention to formerly backgrounded (or defocused) material, and digressions or turns within the structure of the text (Fox 1985; Givón 1985b, 1987, 1988; García 1994: 337). By way of illustration, we would argue that (1a) above, a Subject-verb inversion, in contrast with its non-inverted counterpart (1b), and the thereexistential construction in (2a), with a non-existential marked (inverted) version (2b) and two unmarked alternatives between two different kinds of expansions or tails (2c, 2d), entail different conceptual organizations or directions of mental scanning. In each case the Theme zone, whether or not it carries intonational prominence (Focus), provides an orientational frame with reference to which rightward expansions or increments of discourse are to be processed, thereby instructing both S and A to focus their attention accordingly. In (1a) the Theme zone entails a locational path of access, nested from more generic spatial reference (i.e. focal There) to more detailed location, on the table, a focal spatial modifier that houses an equally focalized participant, my newspaper, in the relative elaboration, thereby empathizing with or providing a profiled relationship to the narrator via possession, the possessor, my, and the possessed, newspaper. The stage being set through this thematic spatial and empathic grounding, it then specifies what is

230

María de los Ángeles Gómez-González

found in that location. Thus, the camera angle of the scene shifts to my packet of biscuits, which is also focal, thereby establishing the postverbal Subject as a prominent participant that is introduced in discourse in order to, in this particular case, close a story about biscuits (the global D-Topic: a stranger eats the narrator’s biscuits), exploiting the recency effect, or prominence of last mention. Alternatively, (1b) lacks the presentative effect and simply profiles a spatial relationship attributing a location to my biscuits (a local D-Topic), again by means of a zooming-in strategy whereby the focus of attention shifts from a more generic to a more specific location.17 The presentative strategy recurs in (2b) – also a spatial Subject-verb inversion – and in (2a), although in the latter case it is implemented differently through the thematization of existential there. In the first increment of (2a) the Theme zone profiles an abstract setting, a presentational frame through which the Subject (the three of us), speaker-empathic through personal reference, receives focal prominence and exploits the recency effect to become the target of attention over the successive attentional frames, thereby becoming the D-Topic of the subsequent discourse spans. First it is assigned a location (in the car), then a state of mind (all rather nervous), and finally attention zooms in on one of its members (the third reporter), through reference chains coding Given information, as shown in (10) below, an expanded version of (2a). (2c) and (2d) lack this presentative potential – as was the case with (1b): a reference point (the three of us) (the local D-Topic) is first assigned a spatial grounding and then a mental grounding in (2c), and vice versa in (2d). (10) ⏐ there were \three of us in the \car ⏐ \all °rather \nervous ⏐ the ∨third re‘porter ⏐ was with the —Washington \Post ⏐ a /war corre‘spondent ⏐ for \ twenty ∨years ⏐ who'd \covered ‘Viet\nam ⏐ the —Washington ∨Post ‘man ‘said ⏐ he ∨hoped ⏐ that ∧ at an —army ∨checkpoint ⏐ just \before the ‘final \ \ ⏐ \stretch ⏐ to ‘Suchi toto ⏐ they would stop us from ‘going \through (LIBMSECAPT04: 018ff.)

To summarize, it can be argued that in contrast to the canonical varieties, inversions, like preposings or left detachments, display a leftward movement within the clause which has a substantial effect, whereas a different effect arises from the rightward movement entailed by existentialthere constructions, or right detachments and extrapositions. Inversions normally act as instructions to place active, particularly inferrable, elements clause-initially, which in English are exempt from Subject status and

FG and the dynamics of discourse

231

are thereby unexpected in their discourse role as point of departure, creating an effect of emotivity, typically over major discourse shifts. This effect is also contributed to by the marked position of the Subject, which is thereby signalled to be unexpected as well: in principle some other potential referent would have been more likely or expected. The Subject referent does not need to be ‘brand new’, but merely less salient/familiar than the fronted element. The presentative nature of inversions resides in their foregrounding of the speaker’s syntactic point of departure, which (a) usually receives intonational prominence (ContrFoc),18 (b) precedes a clausal presupposition, and (c) serves a variety of communicative purposes (viz. to solve problems of encoding, to act as a spatio-temporal framework-setting device for the introduction of new participants in discourse, to establish a cohesive tie, or to create an effect of emphasis). These wide-ranging discourse functions make inversions particularly appropriate for literary styles or for relatively vivid accounts of an event, as well as for expository and argumentative discourse, in which case their role is normally to contribute to the scaffolding of the S’s explanation or argumentation. Alternatively, existential-there constructions lack a presupposition of existence and, in the unmarked cases, place in P1 a Stager, i.e. a scenesetting spatial or temporal adverbial, which acts as a presentative device which, in combination with BE or another existential verb, creates an existential/presentational framework, bringing to the rhematic zone a (lexically and informatively) ‘heavier’19 Existent that receives focal prominence (UttFoc), as dictated by the principles of End Focus and End Weight. This array of information contributes to facilitating the processing of thereconstructions: placed last, heavy rhematic constituents do not disturb the processing of the thematic part, which at the same time provides an existential-presentational framework for grounding the most informative material. Generally, the point of introducing this informative material (the Existent) by means of a there-construction is to develop it as local D-Topic over the ensuing discourse span(s). Because of their functional features, there-constructions are particularly suited for informative and descriptive discourse seeking spatio-temporal locations for focal or discourse-prominent information (entities and situations). The side-effect is often to silence the agency of a State of Affairs, imbuing these constructions with a depersonalized quality, and thereby permitting speakers to avoid claims about the responsibility for the assertion in question, a precondition of objective discourse. In keeping with this, existential-there constructions are, for processing reasons, scarce in active dialogic discourse. The notional Subject is delayed until the end of the con-

232

María de los Ángeles Gómez-González

struct, which puts the A’s short-term memory under strain and decelerates the process of comprehension, an undesirable effect for dialogic discourse, which normally works on an immediate-response basis. Lastly, to round off this section, unmarked high-low, low-high and combined orientations within EMTs can also be associated with corresponding instructions for discourse continuity, discourse discontinuity and both. High-low-oriented ETMs (e.g. (6) and (7) above) show how discourse continuity is ensured by firstly providing a logico-conjunctive peg on which speaker’s attitudes and the core message – normally in this order – is to be hung, whether viewed through one attentional window or more. An alternative discontinuous low-high strategy – as in (8) – assigns focal and cognitive prominence to the profiled relationship, normally encoded in the ET. Detached from the Rest (or Rheme) by means of IT and/or TT, the ET specifies a domain of knowledge, a ‘mental address’, and the proposition by the following clause is ‘delivered’ to that address, i.e. integrated into the domain of knowledge centred on the (local) D-Topic. In most cases, not only ETs, but also post-field ITs and/or TTs, are framed within their own intonation units, which are accorded focal prominence, more processing time and a fuller realization. Both strategies and their corresponding discourse effects are present in the combined type (shown in (9)).20 Owing to their different orientations, EMTs tend to be used either in constructive discourse, which often requires the help of logico-conjunctive signposts to develop their specialized contents, or in dialogic discourse, where speakers can freely express their points of view and have to gain or maintain the discourse floor. Besides, both constructive and dialogic discourse tend to involve the type of State of Affairs preferred by EMTs, i.e. material processes in the sense of Halliday (1994: 106–107), which, because of their informative and objective nature, readily accept logicoconjunctive connectors and interpersonal hedging.

3. Conclusion In this chapter we have attempted to offer a cognitively-oriented extension of Hengeveld’s (this volume) FDG, dubbed IDCG in line with the recent trends observed within and outside FG - mostly Mackenzie’s IFG, Nuyts’s FPG and Langacker’s CG. Space constraints have forced us to give but a cursory exposition of what the model as a whole would look like, to zoom in, in subsequent sections, on different aspects of the dynamics of discourse.

FG and the dynamics of discourse

233

The process of discourse production and comprehension has been portrayed as involving attentional framings superimposed on structural domains, following the shifts of the S’s and A’s focus of attention. The progression of this viewing movement is realized incrementally, or in chunks, through cognitive-structural directions or instructions to focus attention on what is felt to be prominent in discourse. We have seen that very high on the scale of discourse prominence are the notions of Topic, Focus and Theme. It is to be hoped that, its conciseness notwithstanding, this piece of work – part of current research – may at least act as a catalyst for further discussion.

Notes 1.

2.

3.

4.

I am grateful to Lachlan Mackenzie, Knud Lambrecht, Ronald Langacker and Francisco Gonzálvez-García for contributing data, comments, and/or suggestions to this chapter. The research reported here was supported by the Xunta de Galicia, grant number pgidt00pxi20402PR, research project ‘Análise do discurso na lingua inglesa: Aspectos sincrónicos e contrastivos con referencia ó galego o mailo castelán’. WCF can be taken to align with the functional paradigm insofar as (a) it admits that the primary function of the language is to serve as a vehicle of communication and (b) it strives to account for those aspects of grammar that are non-arbitrary in terms of functional principles, that is, relating them to language use (for details see Butler 2003). Carroll et al. (1971) and Goldberg, Sethuraman and Casenhiser (forthc.) show that children begin to learn the associations between form and meaning on two levels, i.e. verb-centred categories and abstract argumentstructure constructions (especially with such recurrent verbs as do, make, get, go, etc.). Likewise, Bencini and Goldberg’s (2000) experimental research suggests that individuals are likely to sort out sentences in terms of argument-structure constructions rather than basing themselves on the lexical semantics of the matrix verb, which gives evidence for the psychological or cognitive reality of constructions in language users’ minds. An intonation unit can be characterized as being typically demarcated by pauses or breaks in timing, by acceleration and deceleration, by changes in pitch level and terminal pitch contours, and as being constrained to a length of four words in English, oftentimes coinciding with clauses, though many others are parts of clauses (Chafe 1994: 69). This characterization suggests a cognitive constraint on how much information can be fully active in the mind at one time. Givón (1995: 358) calls this the ‘one new or informative chunk-per-clause constraint’, and he argues that in many cases it is often sufficient to mention only this chunk (Givón 1989: 209; cf. Du Bois 1987:

234

5.

6.

María de los Ángeles Gómez-González 819). It is no less true, however, that we often cram more than one clause or concept into intonational frames, or stretch out a single clause into more than one frame. Unless otherwise stated, data are taken from the Lancaster IBM Spoken English corpus (LIBMSEC). For further details on this corpus see Taylor and Knowles (1988) and Gómez-González (2001: 192–206). Langacker (2001a) proposes the following list of natural paths, or cognitively natural ordering of the elements of a complex structure: -

7.

speech time (word order, expression order in general) locational path of access (nested or chained) temporal sequence of events successive points along any scale chain of reference point of relationships empathy (speaker > hearer > other; human > animate > inanimate; etc.) chain of cause-effect relationships clausal organization (main clause > subordinate clause) viewpoint chain partonomy (hierarchy of whole-part relations) grammatical relations (subject > object > other)

He contends that, ceteris paribus, there is a tendency for natural paths to coalign, and for the sequences in which this applies to be more basic and easier to process (i.e. unmarked), although there are also cases in which this tendency is overridden by other factors. Here ‘markedness’ refers to probabilities of choice within a linguistic system. Linguistic choices normally skew, in that one or some of their terms are marked or less probable, in contrast with unmarked, or more prototypical or recurrent items. Further, with Kies (1988: 74), we shall take the view that marked options have: (a) comparatively lower frequency of occurrence, (b) comparatively higher structural complexity; and (c) more restricted distribution in definable environments (Givón’s 1993: 178 condition of ‘discourse distribution’). However, it should be borne in mind that marked patterns are not always less frequent – less expected from the addressee’s point of view – than their unmarked counterparts, for they are affected by such factors as text type (Siewierska 1988: 12), or by matters of position and context (Mithun 1987: 313). A case in point is discourse-initial sentences, which belong to highly marked situations and therefore give rise to the so-called markedness reversal, i.e. to a local affinity with marked constructions (Fox 1987; Pu and Prideaux 1994).

FG and the dynamics of discourse 8. 9.

10. 11.

12.

13.

14.

235

Mackenzie and Keizer (1991: 194) and Hannay (1991: 138) consider NewTops as a subcategory of Focus, i.e. Presentative Focus. In this connection Mackenzie and Keizer (1991: 187) argue, for example, that not all GivTops need be contextually given, or introduced into the discourse by means of a NewTop, but can also be situationally given (Situationally GivTops) or generally given (Generally GivTops): e.g. Watch out! The ceiling is caving in!). Similarly, it seems that not all inferrable elements are inferred from a GivTop or fulfil a SubTopic function, for many inferrable elements may also act as NewTops. These may be new at the contextual level but given or inferrable with regard to the addressee`s general or situational pragmatic information: e.g., What did you see in the circus? Well, there was an elephant that amazed us with his tricks... . Note, however, that D-Topic relationships may also hold between nominal expressions. In FG ‘focality’ encompasses “those pieces of information which are the most important or salient with respect to the modifications which S [...] wishes to effect in P [....], and with respect to the further development of the discourse” (Dik 1997: 312). Worthy of mention here is Lambrecht’s (1994: 97) differentiation of activation accent and Focus accent in that a point of prosodic prominence is not necessarily an indicator of either a Focus relation or inactiveness of a referent. It may be one or the other, or both at the same time. An activation accent expresses temporary cognitive states of discourse referents and may fall on a constituent expressing a presupposed proposition (even if it is not in Focus): e.g. the underlined constituents in I saw Mary and John yesterday. She says hello, but he’s still angry at you. vs. I saw Mary yesterday. She says hello (John was very busy that morning) [emphasis mine]). If only that the label ‘NewFoc’ does not seem to be very felicitous, since non-contrastive Focus need not fall on New information. Besides, both kinds of Focus, NewFoc and ContrFoc, entail a certain degree of newness or newsworthiness, albeit of different kinds. Hence both kinds could deserve the label ‘NewFoc’. We shall endorse Lambrecht’s (1988: 1) definition of ‘presupposition’ and ‘assertion’: […] [presupposition entails] the proposition or set of propositions which the speaker assumes the hearer considers true (believes, knows) and is aware of at the time of utterance and which is relevant in the context of utterance. […] [By contrast, assertion is what is] added to or superimposed on the pragmatic presupposition by an utterance [the proposition which the hearer is expected to know as a result of hearing a sentence].

236

15. 16.

17.

María de los Ángeles Gómez-González Of importance here is to bear in mind that ‘givenness’ and ‘presupposition’ evoke two different concepts. Broadly speaking, givenness is a discourse notion referring to the informational status of the constituents of a message as determined by the speaker’s view of the situational and linguistic co(n)text and indicated by attenuated morphosyntactic and phonological forms (e.g. pronominalization, definiteness, lower pitch, weaker stress, etc.). Conversely, presupposition is a logico-semantic notion realized by sentence form which involves a ‘proposition’ (i.e. a potential sentence having the capacity of being true or false), whose assumability is required for the success of the message (Jackendoff 1972: 276–278). Accordingly, a proposition may convey Given information, but need not be presupposed (e.g. It can't be true, as an answer to I saw the man, where it stands as given information, but the proposition is not presupposed). Or, vice versa, informationally new items may occur within a presupposition (e.g. What the duke gave to my aunt was that teapot, where it is presupposed that the duke gave something to someone (my aunt), but the identity of that something is presented as news). An ‘open proposition’, in turn, refers here to a kind of proposition “which contains one or more variables, and it represents what is assumed by the speaker to be salient [in the presupposition] in the discourse [...]. The variable of the open proposition is instantiated with the focus of the utterance” (Ward 1985: 5). For alternative demarcations of the Theme-Rheme partitions see e.g. Halliday (1994), Berry (1996), and Dik (1997). Subject has special status as a point of attachment to what has already been constructed and a point of access to what is currently being constructed (Chafe 1994), hence its tendency to be given rather than new in discourse. This is consonant with Langacker’s (1998, 1999) view that Subject and Object are the first and second conceptual reference points accessed in building up to the full conception of a profiled relationship: Subject is a normally initial participant invoked for purposes of arriving at the conceptualization of a profiled relationship, while an Object is a target evoked in the context of conceptualizing a relationship in which the Subject participates. Langacker concludes that, since the Subject anchors the conception of a relationship encompassing the Object, both tend to coalign with Given vs. New information respectively. Observe the awkwardness of the ‘zooming out’ strategies below ⎯ note that there here is not to be confused with existential there: i. ii.

and my packet of biscuits was on the table, where my newspaper had been, there. and on the table, where my newspaper had been, there was my packet of biscuits.

FG and the dynamics of discourse 18.

19. 20.

237

The marked ‘heavy’ Themes move an informational Focus to the front (ContrFoc, in LIBMSEC normally uttered on a high level), and at the same time, the Subject, which would not be in focus in CWO counterparts, also acquires focal status (in LIBMSEC usually with Falling tones) and is thereby put ‘on stage’. In LIBMSEC the rhematic area of there-constructions was found to be lexically and informatively four times heavier than the thematic part (8 vs 2). According to the results reported for LIBMSEC, most EMTs are obtained by the presence of a prefield TT, thereby having a high-low or combined orientation, that is, structural or conjunctive Themes (e.g. and, but), linking hypotactic or paratactic clause complexes, or continuatives (e.g. em, well, so), punctuating an exchange and staging discourse turns.

References Anstey, Matthew This vol. Functional Grammar from its inception. Bakker, Dik and Anna Siewierska This vol. Towards a speaker model of Functional Grammar. Bates, Elizabeth and Judith C. Goodman 1997 On the inseparability of grammar and the lexicon: evidence from acquisition, aphasia and real-time processing. Language and Cognitive Processes 12: 507–584. Bencini, Giulia M.L. and Adele E. Goldberg 2000 The contribution of argument structure constructions to sentence meaning. Journal of Memory and Language 43: 640–651. Berry, Margaret 1996 What is Theme? A(nother) Personal View. In: Margaret Berry, Christopher Butler, Robin Fawcett and Guowen Huang (eds), Meaning and Form: Systemic Functional Interpretations. Norwood, NJ: Ablex. 1–64. Bolkestein, A. Machtelt 1998 What to do with Topic and Focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. Butler, Christopher S. 2003 Structure and Function: A Guide to Three Major StructuralFunctional Theories. Amsterdam and Philadelphia: Benjamins. Carroll, John B., Peter Davies, and Barry Richman 1971 Word Frequency Book. New York: Houghton Mifflin.

238

María de los Ángeles Gómez-González

Chafe, Wallace L. 1976 Givenness, contrastiveness, definiteness, Subjects, Topics and point of view. In: Charles Li (ed.), Subject and Topic, 26–56. New York: Academic Press. 1987 Cognitive constraints on information flow. In: Russell S. Tomlin (ed.), Coherence and Grounding in Discourse, 21–51. Amsterdam and Philadelphia: Benjamins. 1994 Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago: University of Chicago Press. Connolly, John H. 1998 Information, situation semantics and Functional Grammar. In Hannay and Bolkestein (eds), 167–189. This vol. The question of discourse representation in Functional Discourse Grammar. Connolly, John H., Roel Vismans, Christopher Butler, and Richard Gatward (eds) 1997 Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter. Cornish, Francis This vol. Focus of attention in discourse. Dik, Simon C. 1978 Functional Grammar. Amsterdam: North Holland. 1997 The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. Du Bois, John W. 1987 The discourse basis of ergativity. Language 63: 805–855. Fortescue, Michael This vol. The complementarity of the process and pattern interpretations of Functional Grammar. Fox, Barbara A. 1987 Morphosyntactic markedness and discourse structure. Journal of Pragmatics 11: 359–373. Givón, Talmy 1989 Mind, Code and Context: Essays in Pragmatics. Hillsdale NJ: Erlbaum. 1993 English Grammar. (Vol. 2) Amsterdam and Philadelphia: Benjamins. 1995 Functionalism and Grammar. Amsterdam and Philadelphia: Benjamins. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago and London: The University of Chicago Press.

FG and the dynamics of discourse forthc.

239

Argument realization: The role of constructions, lexical semantics and discourse factors. In: Mirjam Fried and Jan-Ola Östman (eds), Construction Grammar(s): Cognitive and Cross-Language Dimensions. Amsterdam and Philadelphia: Benjamins. Goldberg, Adele E., Nitya Sethuraman and Devin Casenhiser forthc. Learning argument structure generalizations. Ms. University of Illinois and University of California at San Diego. Gómez-González, María de los Ángeles 1998a A corpus-based analysis of extended multiple Themes in Present-day English. International Journal of Corpus Linguistics 3: 81–114. 1998b Aspects of Topic and topicality in Functional Grammar. Working Papers in Functional Grammar 65. Amsterdam: University of Amsterdam. 2001 The Theme-Topic Interface: Evidence from English. (Pragmatics and Beyond New Series 71.) Amsterdam and Philadelphia: Benjamins. Grice, H. Paul 1975 Logic and conversation. In: Peter Cole and Jerrold L. Morgan (eds), Studies in Syntax 3: Speech Acts. 41–58. New York: Seminar Press. Halliday, Michael A.K. 1994 An Introduction to Functional Grammar. 2nd edition. London: Arnold. Hannay, Mike 1991 Pragmatic function assignment and word order variations in a Functional Grammar of English. Journal of Pragmatics 16: 131–155. 1994 The Theme zone. In: Ronny Boogaart and Jan Noordegraaf (eds), Nauwe Betrekkingen. 107–117. Amsterdam: Stichting Neerlandistiek VU and Münster: Nodus. 1998 Message modes in a modular FG. Paper delivered to the 8th International Conference on Functional Grammar. Amsterdam: Vrije Universiteit. 6–11 July 1998. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Harder, Peter 1996 Functional Semantics. A Theory of Meaning, Structure and Tense in English. (Trends in Linguistics. Studies and Monographs 87). Berlin and New York: Mouton de Gruyter. This vol. Comment clauses, Functional Discourse Grammar and the grammardiscourse interface. Hengeveld, Kees 1997 Cohesion in Functional Grammar. In John H. Connolly Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 1–16.

240

María de los Ángeles Gómez-González

This vol. The architecture of a Functional Discourse Grammar. Inchaurralde, Carlos This vol. Behind the scenes: Cognition and Functional Discourse Grammar. Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. Cambridge MA and London: MIT Press. 1987 Consciousness and the Computational Mind. Cambridge MA: MIT Press. 1990 Semantic Structures. Cambridge MA: MIT Press. Kay, Paul and Charles J. Fillmore forthc. Construction Grammar. Stanford: CSLI Publications. Kemmer, Suzanne 1995 Emphatic and reflexive –self: Expectations, viewpoint and subjectivity. In Dieter Stein and Susan Wright (eds), Subjectivity and Subjectivisation, 55–82. Cambridge: Cambridge University Press. Kies, Daniel 1988 Marked Themes with and without pronominal reinforcement, their meaning and distribution in discourse. In: Erich M. Steiner and Robert Veltman (eds), Pragmatics, Discourse and Text, 47–75. London: Pinter Publishers. Kroon, Caroline 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward, 17–32. Ladd, D. Robert 1996 Intonational Phonology. Cambridge: Cambridge University Press. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Lambrecht, Knud and Maria Polinsky 1997 Typological variation in sentence-Focus constructions. Chicago Linguistic Society 33. 1–20. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Stanford: Stanford University Press. 1998 Topic, Subject, and possessor. Linguistic Notes from La Jolla 19: 1–28. 1999 Assessing the cognitive linguistic enterprise. In: Theo Janssen and Gisela Redeker (eds), Cognitive Linguistics: Foundations, Scope, and Methodology, 13–59. (Cognitive Linguistics Research 15.) Berlin and New York: Mouton de Gruyter.

FG and the dynamics of discourse 2000

241

A dynamic usage-based model. In: Michael Barlow and Suzanne Kemmer (eds), Usage-Based Models of Language, 1–63. Stanford: CSLI Publications. 2001a Dynamicity in grammar. Axiomathes 12. 7–33. 2001b Discourse in cognitive grammar. Cognitive Linguistics 12. 143–188. Levelt, Willem J. M. 1989 Speaking: From Intention to Articulation. Cambridge MA: MIT Press. Mackenzie, J. Lachlan 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–296. 2000 First things first: towards an Incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. This vol. Functional Discourse Grammar and language production. Mackenzie, J. Lachlan and M. E. Keizer 1991 On assigning pragmatic functions in English. Pragmatics 1: 169– 215. MacWhinney, Brian 1977 Starting points. Language 53. 152–168. Mithun, Marianne 1987 Is basic word order universal? In: Russell Tomlin (ed.), 281–328. Moutaouakil, Ahmed This vol. Discourse structure, the Generalized Parallelism Hypothesis and the architecture of Functional Grammar. Noonan, Michael 1999 Non-structuralist syntax. In: Michael Darnell, Edith Moravcsik, Fritz Newmeyer, Michael Noonan and Kathleen Wheatley (eds), Functionalism and Formalism in Linguistics, 11–31. Amsterdam and Philadelphia: Benjamins. Nuyts, Jan 1992 Aspects of a Cognitive-Pragmatic Theory of Language. Amsterdam and Philadelphia: Benjamins. This vol. Remarks on layering in a cognitive-functional language production model. Pu, Ming-Ming, and Prideaux, Gary D. 1994 Coding episode boundaries with marked structures: A crosslinguistic study. Canadian Journal of Linguistics 39: 283–296. Selkirk, Elisabeth O. 1984 Phonology and Syntax: The Relationship between Sound and Structure. Cambridge, MA: MIT Press.

242

María de los Ángeles Gómez-González

Siewierska, Anna 1988 Word Order Rules. London: Croom Helm. 1991 Functional Grammar. London: Routledge. Smits, Aletta 2002 How writers begin their sentences: Complex beginnings in native and learner English. Doctoral dissertation, University of Amsterdam. Steuten, Ans A. G. 1998a A Contribution to the Linguistic Analysis of Business Conversations within the Language/Action Perspective. Doctoral dissertation, University of Delft. 1998b Structure and coherence in business conversations. In: Mike Hannay and A. Machtelt Bolkestein (eds), 59–75. Taylor, Lita J., and Gerald Knowles 1988 Manual of Information to Accompany the LIBMSEC Corpus: The Machine-Readable Corpus of Spoken English. Unit for Computer Research on the English Language Bowland College, University of Lancaster. Bailrigg: Lancaster. Tomlin, Russell (ed.) 1987 Coherence and Grounding in Discourse. Amsterdam and Philadelphia: Benjamins. Vallduví, Enric, and Elizabeth Engdahl 1996 The linguistic realization of information packaging. Linguistics 34: 459–519. Van den Berg, Marinus 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–103. Van der Auwera, Johan, and Louis Goossens (eds) 1987 Ins and Outs of the Predication. Dordrecht: Foris. Verstraete, Jean-Christophe This vol. The problem of subjective modality in the Functional Grammar model. Vet, Co 1998 The multilayered structure of the utterance: About illocution, modality and discourse moves. In Mike Hannay and A. Machtelt Bolkestein (eds), 1–24. Ward, Gregory Louis 1985 The Semantics and Pragmatics of Preposing. Michigan: University Microfilms International.

The problem of subjective modality in the Functional Grammar model Jean-Christophe Verstraete

1. Introduction The purpose of this chapter is twofold.1 I will first propose a number of modifications to the analysis of subjective modality in the Functional Grammar model (as developed in Hengeveld 1987, 1988, 1989, Dik 1997), using the English modal auxiliaries as a test case. I will then confront the alternative view of subjective modality resulting from these modifications with the architecture for FG proposed in Hengeveld (this volume). I will show how the proposed modifications can be dealt with more easily in the modular, top-down architecture than in the traditional FG model, but I will also argue that some further change is required in the model, more particularly concerning the optionality of layers at the representational level.

1.1. The subjective-objective distinction The first part of this chapter will be devoted to a critical examination of the distinction between subjective and objective modality in the traditional FG model. I will first propose a number of modifications to the criteria used to support the subjective-objective distinction. It is often argued that objective modality can be questioned and hypothesized, whereas subjective modality cannot. I will show that this is not entirely correct: conditionality and interrogation are not excluded for subjective modality, but interact with it in a special way. I will argue that this interaction can be explained as a consequence of the fundamentally interpersonal nature of subjective modality. On the basis of these modifications, I will then re-assess the position of deontic modality with respect to the subjective-objective distinction. In the FG model, as in many other frameworks, deontic modality is excluded from the subjective category. On the basis of the functions of deontic

244

Jean-Christophe Verstraete

modals and their behaviour in reaction to the criteria of conditionality and interrogation, I will show that this exclusion is not justified, and that deontic modality can be subjective just like epistemic modality. In the final section, I will argue that what sets deontic modals apart from their epistemic counterparts is not their potential for subjective status, but rather the type of domain over which they operate: subjective epistemic modality deals with the truth of propositions, whereas subjective deontic modality deals with the desirability of actions. This functional distinction is also grammatically reflected in the terms of the feature of tense: subjective epistemic modality operates over tensed SoAs, whereas subjective deontic modality operates over tenseless SoAs.

1.2. Implications for the FG model(s) In the second part of this chapter, I will confront this alternative analysis of subjective modality both with the traditional FG model in Hengeveld (1989) and Dik (1997) and with the alternative architecture for FG proposed by Hengeveld (this volume), showing how the alternative analysis of subjective modality proposed in this chapter can be accommodated more easily in that alternative architecture. The modular separation between the interpersonal and the representational components allows for a subjective analysis of deontic modality in parallel with epistemic modality, independently of the fact that subjective epistemic and subjective deontic modality are associated with different types of domains. This difference in domain between epistemic and deontic modality, on the other hand, can be dealt with in terms of the top-down organization of the model, as an example of how choices within the interpersonal component have repercussions on choices within the representational component. To round off, I will also discuss one problem that is not dealt with in Hengeveld’s (this volume) proposed model. In spite of the modular separation between the interpersonal and the representational components, the model still implicitly assumes that the full set of representational layers is relevant for every main clause. In order to adequately model subjective deontic modality, however, it is necessary to recognize optionality of layers in the representational component.

The problem of subjective modality in the FG model

245

2. The analysis of modality in FG I will first briefly outline the most important features of the analysis of modality in FG, as it has been developed in the work of Hengeveld (1987, 1988, 1989) and Dik (1997).

2.1. Three categories The basis of the FG analysis of modality is a distinction between three different categories, which is itself based on a distinction between different functional layers in the clause with which the modal categories can be associated. More particularly, Hengeveld (1988, 1989) and Dik (1997) distinguish between the categories of subjective, objective and inherent modality. Inherent modality includes “all those linguistic means through which S can characterize the relation between a participant in an SoA and the potential actualization of that SoA” (Hengeveld 1988: 233). In terms of the traditional tripartition of epistemic, deontic, and dynamic modality (see, for instance, Palmer 1990), FG’s inherent category includes the dynamic modals of ability and volition – the traditional participant-oriented modals – as well as those uses of deontic modality “in which it is reported that some participant in a state of affairs is under the obligation or has received permission to perform in that state of affairs” (Hengeveld 1988: 234). These types are exemplified respectively in examples (1), (2) and (3), in which the modal expressions in question have been underlined: (1)

(2)

(3)

DYNAMIC, ABILITY Well, certainly marriage has changed him. It mellowed him to the extent that he lost his drive and his motivation. In the midst of all that, he thought 2 to himself, “Hey, wait a minute. I can still play this game.” (CB) DYNAMIC, VOLITION They were servicing AWACs planes for the Royal Saudi Air Force and completing contract work on a controversial air defense network for Saudi Arabia known as the Peace Shield. Boeing won't say how many of its employees are still there, but says those who stayed did so voluntarily. (CB) DEONTIC But Ramadan means more than just physical deprivation. It has spiritual and moral obligations, too. Followers must refrain from bad thoughts, words and actions, perform special acts of charity and spend even more time than usual in worship. (CB)

246

Jean-Christophe Verstraete

Objective modality includes “all those linguistic means through which S can evaluate the actuality of an SoA in terms of his knowledge of possible SoAs” (Hengeveld 1988: 233). The relevant knowledge here can be either epistemic “knowledge of possible situations obtaining in S’s conception of reality or of a hypothesized situation” (Hengeveld 1988: 234) or deontic “knowledge of possible situations relative to some system of moral, legal or social conventions” (Hengeveld 1988: 234). Accordingly, the objective category includes both epistemic and deontic modality, exemplified in (4) and (5): (4)

(5)

EPISTEMIC Is it possible that all human beings on earth today are descended from a single woman? (CB) DEONTIC The bandages they wrapped around his face were loose enough to allow him to breathe, yet tight enough not to slip. “You must not move a muscle,” Ram Das warned him. “That would cause a riot of terror. We’re going to cover you with saffron dust to show we’re carrying a corpse to the holy river.” (CB)

Subjective modality, finally, involves “all those linguistic means through which S can express his commitment with regard to the truth of a proposition” (Hengeveld 1988: 233). This applies only to epistemic modality, exemplified in (6): (6)

EPISTEMIC Michael Jackson must have enjoyed his wedding in May, because he decided to have another one at his California ranch. (CB)

2.2. Theoretical motivation Within the FG framework, the distinction between subjective, objective and inherent modality is theoretically motivated in terms of an association with different functional layers in the structure of the clause (Hengeveld 1989, compare Foley and Van Valin 1984, partly also Lyons 1977).3 The starting point for the FG account of the different categories of modality is the distinction between a propositional and a predicational layer in the structure of the clause. According to Hengeveld (1989), this is grounded in the distinction between a representational and an interpersonal

The problem of subjective modality in the FG model

247

function in the clause (Halliday 1994), which take care respectively of the conceptualization of the situation referred to in the clause and the interactive intentions of the speaker towards the interlocutor. In this sense, the predicational and propositional layers represent two different functional perspectives on the traditional notion of State of Affairs (SoA): the predication is the representational perspective on the SoA, as a description of the situation to which the speaker refers in the clause, whereas the proposition is the interpersonal perspective on the SoA, as the propositional content for a particular communicative act performed by the speaker in the clause. In terms of this distinction between interpersonal and representational functions of the SoA, a first distinction in the domain of modality is made between subjective modality on the one hand and objective and inherent modality on the other hand. Subjective modality serves to express the speaker’s commitment, and is therefore an interpersonal category, taking care of the speaker’s interactive positioning with respect to the SoA as the object of his speech act. In the FG model, this is accommodated theoretically by analyzing subjective modality as an operator of the propositional layer. Objective and inherent modality, on the other hand, do not have any interpersonal function, and are therefore associated with the predicational layer. Within the non-interpersonal domain, a further distinction is made between objective and inherent modality on the basis of their divergent positions with respect to the SoA. Inherent modals serve to evaluate the actualization of the SoA from a perspective internal to the SoA, more particularly by indicating a particular type of relation (for instance ability or willingness) between one of the participants and the SoA, whereas objective modals evaluate it from an external perspective. In the FG model, this is accommodated theoretically in terms of different positions with respect to the predicational layer: the SoA-external function of objective modals is reflected in their position as operators with scope over the predication, whereas the SoA-internal function of inherent modals is reflected in their position as categories internal to the predication.

2.3. Criteria Hengeveld (1988) argues that the functional distinction between subjective and objective modality is also reflected in their grammatical behaviour, and adduces a number of grammatical criteria to which subjective and objective types of modality react differently. In addition to tense, polarity and prag-

248

Jean-Christophe Verstraete

matic appropriateness of questioning the source of information, which will not be dealt with in this chapter, Hengeveld also mentions conditionality and interrogation: “Objectively modalized predications can be questioned, subjectively modalized ones cannot”; “Objectively modalized predications can be hypothesized in a conditional sentence, subjectively modalized ones cannot” (Hengeveld 1988: 236).

2.4. Summary Table 1. The FG analysis of modality FUNCTION

LAYERING

CATEGORY

MEMBERS CRITERIA

Interpersonal

Prop. operator

Subjective

Epistemic

-{interr, cond}

Representational Pred. operator

Objective

Epistemic, Deontic

+{interr, cond}

Representational Pred.-internal

Inherent

Deontic, Dynamic

+{interr, cond}

3. Criteria: refinement and explanation In order to demonstrate the need for an alternative analysis of the subjective-objective distinction, I will first show that the criteria used to distinguish between subjective and objective modality, especially conditionality and interrogation, are not entirely accurate as they are formulated in Hengeveld (1988). Subjective modals are not excluded from conditional and interrogative constructions: rather, using subjective modals in these constructions leads to modifications in their interpretation that do not occur with objective modals. These descriptive refinements, however, do not imply a rejection of conditionality and interrogation as criteria for the subjective-objective distinction. The semantic modifications observed for subjective modals in these constructions provide an important key to their fundamentally interpersonal status: the semantic effects of interrogation and conditionality can be explained in terms of the interpersonal function of subjective modality described in Hengeveld (1988: 233), viz. the fact that they serve to encode

The problem of subjective modality in the FG model

249

positions of commitment with respect to the SoA described in the utterance.

3.1. Conditionality Hengeveld (1988: 236) proposes to use conditionality as an absolute criterion for the subjective-objective distinction, and argues that objective modality can occur in the protasis of conditional constructions, whereas subjective modality cannot. This absolute interpretation of the conditionality criterion may work for the distinction between modal adjectives and adverbs,4 but it is not an adequate account of the behaviour of modal auxiliaries in conditional contexts: subjective modals can occur in conditionals, but their interpretation is affected by the conditional context. Subjective modals in the protasis of conditional constructions invariably become echoic (see also Palmer 1990: 182): they do not express the current speaker's opinion, but echo an opinion that has been voiced in or is implied by the preceding discourse. For instance, (7)

In distilling a statement of theme from a rich and complicated story, we have, of course, no more encompassed the whole story than a paleontologist taking a plaster mold of a petrified footprint has captured a living brontosaurus. A writer (other than a fabulist) does not usually set out with theme in hand, determined to make every detail in the story work to demonstrate it. Well then, the skeptical reader may ask, if only some stories have themes, if those themes may be hard to sum up, and if readers will probably disagree in their summations, why bother to state themes? (CB)

In (7), the epistemic position expressed by may is crucially not the position of the ‘sceptical reader’ who uses the conditional construction, but an opinion voiced by some other speaker which the ‘sceptical reader’ echoes in his conditional construction without committing himself to it. Thus, subjective modals are not excluded in conditional contexts, but their interpretation is affected by the conditional. This semantic effect does not occur with non-subjective modals like the modal of ability in (8): the interpretation of can’t in the conditional in (8) is not different from its interpretation in non-conditional contexts. (8)

Phillips: Oh, I think it’s just the–the same thing as the sign that they had during the campaign, that everything else was secondary to the economy. If Bill Clinton can’t deal with the economy in the next year or year and a half,

250

Jean-Christophe Verstraete he’s really going to be in trouble in–in the whole breadth of his administration. (CB)

3.2. Interrogation Hengeveld (1988) argues that objective modality can be questioned, whereas subjective modality cannot. Again, I agree that this may be the case when subjective and objective modality are realized as modal adverbs and adjectives respectively,5 but it is certainly not the case for modal auxiliaries. Subjective modal auxiliaries can occur in interrogatives just like objective ones, as shown in examples (9) and (10): (9)

(10)

THOMPSON: If the trials are successful, might this be used as the first line of treatment of cancer? THORPE: Well that’s exactly where we are going. We are hoping that this will replace conventional chemotherapy. (CB) I mean maybe we’ve got no choice if we want to do it but I mean erm do you think it’s possible to do anything worthwhile in history lessons or can this work really only take place in somewhere like civics social studies sociology politics? (CB)

As with conditionality, the difference between subjective and objective modals is not a matter of acceptability in interrogative contexts, but of the way in which the interpretation of the modal is affected by the interrogative construction. With subjective modality, the commitment encoded by the modal undergoes a shift of orientation under the influence of interrogation. In declarative contexts, the responsibility for the position is assigned to the speaker, but in interrogative contexts like (9) or (10), the responsibility for the position encoded by subjective modals like epistemic might and can is no longer taken by the speaker, but is assigned to the interlocutor in the next turn: “Are you committed to this position?” rather than “I am committed to this position”. For non-subjective modals, on the other hand, interrogation does not affect the interpretation of the modal. Since the interpretation of nonsubjective modality does not require any assignment of responsibility, it does not undergo any shift of orientation under the influence of interrogation. For instance, if we look at examples like (11) or (12): (11) (12)

They’ve already worried the charts and won many admirers with the single Fun for Me. Can they do it live? It’s time to find out. (CB) First-time buyers should ask: Who is responsible for a title search and abstract? Who will provide title insurance? Must the termite inspection be

The problem of subjective modality in the FG model

251

paid before closing (in which case you will need a receipt), or can it be paid at settlement? (CB)

Unlike the subjective modals in (9) and (10), the non-subjective uses of can and must in (11) and (12) do not undergo any shift of orientation in interrogative contexts. They do not imply any responsibility or a position of commitment, and therefore they are not affected by interrogation. Rather than taking part in the speaker-interlocutor transfer of responsibility, dynamic can and deontic must in (11) and (12) belong to the content with respect to which positions of responsibility are taken or transferred. With must in (12), for instance, the obligation or necessity denoted by must is part of the content about which speaker and interlocutor negotiate epistemically (“Do you think it is obligatory or necessary to pay the termite inspection before closing?”; see further in Section 5.2.). The same goes for can in (11): the ability denoted by the modal is part of the content that is subject to epistemic negotiation (“Do you think they are able to do the same thing live?”).

3.3. Explanation of the behavioural distinctions The descriptive modifications proposed in the previous two sections do not invalidate the basic idea that conditionality and interrogation can be used as criteria for the subjective-objective distinction. The behaviour of subjective and objective modality in reaction to conditionality and interrogation is indeed different, only not in terms of unacceptability but in terms of the effect on the interpretation of the modal. In this section, I will show how these semantic effects do not merely serve as criteria to differentiate between subjective and objective uses of modal auxiliaries, but that the difference in behaviour can actually be linked back to the functional basis of the subjective-objective distinction proposed in Hengeveld (1989). The effects of echo interpretation for conditionals and shift of orientation for interrogatives can be explained as a consequence of the fact that subjective modals take part in the interpersonal organization of the clause: they serve to encode positions of commitment with respect to the SoA described in the utterance. The fact that subjective modality becomes echoic in the protasis of conditional constructions is a consequence of the clash between the function of the subjective modal, which is to encode a position of commitment by the speaker, and the function of the conditional marker, which is to suspend the speaker’s commitment with respect to the proposition in its scope. Condi-

252

Jean-Christophe Verstraete

tionals are special contexts in that they operate with suppositions rather than assertions (Declerck and Reed forthcoming): the typical function6 of the conditional marker if is to suspend the commitment of the speaker with respect to the proposition in its scope (Dancygier 1998) and thus to treat it as a mere supposition in the context of the apodosis. For objective modals, this suspension of commitment is unproblematic, because objective modality does not encode any type of speaker commitment. For subjective modals, however, the suspension of speaker commitment in conditionals clashes with the very function of subjective modality, which is to encode the speaker’s commitment with respect to the proposition. This clash leads to an echoic reinterpretation, which could be regarded as the interpretational compromise between the two conflicting functions in the modal and the conditional: the position encoded by the modal is no longer a position of the current speaker, because the conditional context suspends such positioning, but resumes a position taken in the preceding discourse, usually by another speaker. That is the only way a modal can still encode a position in a context which does not allow such positioning by the current speaker. The fact that subjective modality undergoes a shift in orientation in an interrogative context is again a consequence of its position-encoding function. A position of commitment necessarily implies responsibility for that position, and that is what the declarative-interrogative contrast operates on. The declarative allows the speaker to take the responsibility for the position encoded in the subjective modal in his own turn, whereas the interrogative allows the speaker to transfer this responsibility to the interlocutor in the next turn (Davies 1979). Unlike subjective modals, objective modals do not serve to encode any positions of commitment, and therefore do not undergo any shift of orientation under the influence of interrogation. Interrogation can only interact with modality if the function of the modal implies assignment of responsibility in discourse.

4. The position of deontic modality On the basis of the modified criteria, the position of deontic modality in the overall model for modality in FG can now be reconsidered. In Hengeveld (1988), deontic modality is included in the objective and inherent categories, but crucially not in the subjective category. FG is not the only framework to treat deontic modality differently from epistemic modality: deontic modality has often been set apart in the analysis of the subjective-objective distinction. Foley and Van Valin (1984), for instance,

The problem of subjective modality in the FG model

253

do not recognize any subjective-objective ambiguity for deontic modality at all, and treat the deontic modals as ‘core’ operators (more or less the equivalent of FG’s predicational operators) together with the dynamic modals. Even Lyons (1977), who seems to recognize both subjective and objective types for deontic modality, still differentiates his analysis of subjectivity by associating subjectivity with different functional layers (in terms of Hare’s 1970 tropic-neustic-phrastic framework) depending on the epistemic or deontic nature of the modal. Halliday (1970) is in fact the only one who recognizes a subjective function for deontic modality that is fully parallel with the subjective function of epistemic modality. In this section, I will try to defend this position both on the basis of the function of deontic modals (Section 4.1) and on the basis of their behaviour in reaction to some of the criteria for subjectivity (Section 4.2). In the next section, I will show that the difference between epistemic and deontic modality does not lie in their potential for subjective status, but in the domain over which they operate.

4.1. The functions of deontic modals: inherent, objective and subjective 4.1.1. Types of deontic modality In terms of function, there are at least three types of deontic modality that can be distinguished in English, illustrated in (13) – (16) below: (13)

(14)

(15)

(16)

“I need to see Izzy,” I said. “I told you, she's sound asleep. Deeply asleep.” “May I see for myself?” “You may not, you shit-sucking liar! You cheat. You coward. You sit!” I pulled back the heavy dining room chair and sat before the typewriter. (CB) What we want is for the right honourable gentleman to use the full weight of his office. We are getting tired of a cosmetic approach an oversanguine approach. There is a crisis and he must act now But Ramadan means more than just physical deprivation. It has spiritual and moral obligations, too. Followers must refrain from bad thoughts, words and actions, perform special acts of charity and spend even more time than usual in worship. (CB) But to reach orbit an object must accelerate to a speed of about 17,500 miles per hour (28,000 kilometres per hour, called satellite speed or orbital velocity) in a horizontal direction; and it must reach an altitude of more

254

Jean-Christophe Verstraete than 100 miles (160 kilometres), in order to be clear of the atmosphere.

In examples (13) and (14), the function of the deontic modal is to express the speaker’s commitment to the permission or obligation encoded in the modal: may and must in these examples can be paraphrased as “I do not allow you to …” and “I oblige him to …”. An example like (15), on the other hand, does not express the speaker’s commitment to the obligation signalled by the modal (“I want the followers to …”), but rather reports on the existence of a particular obligation without necessarily committing the speaker to it. Unlike in (13) and (14), the speaker is not the deontic source of the obligation or permission, but merely the one who describes the existence of such an obligation, which may itself originate from another source. An example like (16), finally, is different from both (13) and (14) and from (15) in that no deontic source is involved at all. Must in (16) does not express an obligation originating from a deontic source (irrespective of the question whether this source is identical with the speaker or not), but denotes a necessity that is inherent in the situation:7 the state of things is such that if the object in question is to reach orbit, a speed of 17,500 mph is required. The FG analysis of modality also recognizes functional ambiguity for deontic modals, and accounts for it in terms of the distinction between objective and inherent modality. Examples like (13) and (14) would be classified as objective, i.e. expressing an evaluation of the actuality of the SoA on the basis of the speaker’s “knowledge of possible situations relative to some system of moral, legal or social conventions” (Hengeveld 1988: 234). Examples like (15), on the other hand, would be classified as inherent, i.e. “report[ing] that some participant in a state of affairs is under the obligation or has received permission to perform in that state of affairs” (Hengeveld 1988: 234).

4.1.2. An alternative to the FG categorization of the types I do not believe that the objective-inherent dichotomy is an adequate way to account for the functional distinction between (13) and (14) on the one hand and (15) on the other hand. In this section, I will argue that in terms of function it is more plausible to analyse (13) and (14) as subjective in parallel with epistemic modality rather than as objective. Examples like (15), on the other hand, should be analysed as the objective counterparts of

The problem of subjective modality in the FG model

255

such subjective uses rather than grouping them together with the dynamic modals of ability and volition in the inherent category. Examples like (16), finally, which are not explicitly dealt with in the FG framework,8 are the most plausible ‘deontic’ candidates for the inherent category, because of their functional similarity with the dynamic modals of ability and volition. The FG analysis of examples like (13) and (14) as representatives of the objective category implies that the function of the modal here is to provide an evaluation of the actuality of the SoA in terms of the speaker’s knowledge of conventions or morals. I do not think that this is an adequate description of the function of may and must in (13) and (14), as also argued by Goossens (1996: 49–50): in these examples, the speaker does not evaluate the actuality of the SoA, but expresses his commitment to the permission or obligation. It is the speaker who wants the SoA to be actualized: the speaker is not evaluating descriptively but acting interpersonally. Functionally, therefore, this use comes closer to the category of subjective modality, which “expresses [the speaker’s] commitment” (Hengeveld 1988: 233). Given the function of examples like (13) and (14), I see no reason to exclude deontic modality from the subjective category: deontic modals can be subjective just like epistemic ones, the only difference being that the commitment in question does not concern the truth of propositions, but the desirability of actions (see further in Section 5). For examples like (15), the functional characterization proposed by FG is entirely adequate. Examples like these do not involve the speaker’s commitment to the obligation: the speaker is the one who reports on the existence of an obligation for some participant to act but does not coincide with the deontic source of this obligation. In spite of the adequacy of the functional characterization for such examples, however, I do not think that they should be grouped together with dynamic modals of ability and volition as members of the inherent category, as is the case in the FG analysis. There are two important arguments for distinguishing examples like (15) from the modals of ability and volition. On the one hand, the obligation expressed by must in (15) still implies some external source, whereas the dynamic modals of ability and volition never involve any external source but are entirely internal to the SoA. In this sense, (15) naturally groups together with (13) and (14) in contrast with the dynamic modals of ability and volition: (13), (14) and (15) always imply some external source (which may or may not coincide with the speaker), whereas the dynamic modals of ability and volition do not involve any source at all. The second reason for keeping must in (15) apart from the dynamic modals of ability and volition is that within the deontic system there is a another use of must which does

256

Jean-Christophe Verstraete

naturally group together with the modals of ability and volition in that it does not involve any external source. Must in (16) is similar to the dynamic modals and different from (13), (14) and (15) because it is not related to any deontic source: must in (16) does not express an obligation coming from an external source but simply expresses the existence of a necessity internal to the SoA. I believe that we can do justice to both these arguments by redistributing examples like (15) and (16) over the objective and inherent categories: (15) should be regarded as objective rather than as inherent, and (16) should be included in the inherent category. Including examples like (15) in the objective category does justice to their natural affinity with subjective modality: in both cases, the modal is related to a deontic source, and the difference between the two relates to the question whether the deontic source coincides with the speaker (subjective) or not (objective). On the other hand, the ‘gap’ that is left by taking examples like (15) out of the inherent category is more adequately filled by examples like (16): grouping these examples together with the dynamic modals of ability and volition does justice to their shared SoA-internal function.9 Thus, a consideration of the various functions of deontic modality leads to the following alternative proposal for the subjective-objective-inherent matrix: Table 2. The functions of deontic modality FG Subjective Objective Inherent

(13), (14) (15)

Alternative (13), (14) (15) (16)

4.2. Criteria applied to deontic modality A subjective analysis of deontic modals like (13) and (14) in parallel with epistemic modality is further justified by their behaviour in reaction to the various criteria that can be used to distinguish between subjective and objective modality. Subjective deontic modals are just as much speakerhearer-related as their subjective epistemic counterparts, and therefore show the same behaviour with respect to conditionality and interrogation.

The problem of subjective modality in the FG model

257

4.2.1. Conditionality As already argued in Section 3.1, subjective epistemic modality becomes echoic in conditional constructions: it does not express a position taken by the speaker who uses the conditional, but echoes one voiced by another speaker in the preceding discourse context. With subjective deontic modality, conditionality can have the same effect – although because of the almost systematic ambiguity with non-subjective functions this is not automatic as it is for epistemic modals. For instance, if we compare (17) with (18): (17)

(18)

The key stumbling block remained Republican insistence on a Medicare premium increase. Mr Clinton argued that Medicare increases were not necessary to meet demands for a balanced Budget. If America must close down access to quality education, a clean environment and affordable health care for our seniors in order to keep the government open, then that price is too high,” Mr Clinton said in vetoing the temporary spending Bill. (CB) I suffer from acne but also burn very easily - can I use sun oils? It is always best to tan gradually, but especially so if you have a sensitive skin. If you must be exposed to sun that may burn you, it may be wise to use a sunscreen preparation. (CB)

In (18), the non-subjective deontic modal is unproblematic in the conditional: it does not echo any position taken by another speaker, but simply expresses that if a situation occurs that requires exposure to sun, it is wise to use a sun-screen preparation. In (17), on the other hand, the subjective deontic modal is echoic: must in the protasis does not simply express the existence of necessity or obligation, but echoes another speaker’s commitment to such an obligation. This is evident from the fact that the conditional in (17) is rhetorically directed against those who are committed to Medicare premium increases. The speaker in (17) takes up the Republicans’ position in his conditional protasis without committing himself to it: the must in (17) is clearly a must that is attributed to the Republicans, and not a position to which Clinton himself subscribes.

4.2.2. Interrogation In Section 3.2, it was already argued that interrogation interacts with subjective modality, in that the responsibility for the position encoded by the

258

Jean-Christophe Verstraete

modal is transferred to the interlocutor’s next turn in an interrogative context. The same applies to subjective deontic modality, as illustrated in the following examples: (19)

(20)

Dr Gwyn Adshead thank you very much for joining us. And please may I call you Gwyn? Certainly. Thank you. That's a good start. What are the differences generally speaking between a counsellor a therapist and a psychiatrist? (CB) You’ve got to be there by nine o’clock in the morning at the latest. You’ll be crossing the main refugee routes. Shouldn’t be too bad.” Must I leave my platoon, sir? At this moment?” Stop arguing and get down there. It’s nobody’s fault but yours that you speak fluent German. You know perfectly well every linguist’s name is listed.” (CB)

Both in (19) and in (20), the deontic modals express some position of commitment to the permission and the obligation to carry out the action in question, but the responsibility for this commitment is not the speaker’s. Rather, the speaker transfers this responsibility to the interlocutor’s next turn, i.e. asks whether the interlocutor is committed to it: “Will you allow me to call you Gwyn?” and “Are you obliging me to leave my platoon?”. For non-subjective deontic modality, on the other hand, the modals do not take part in the speaker-interlocutor exchange: if a modal does not express any position of commitment, there is no responsibility to be assigned. An example like (21) below cannot be paraphrased as “Do you want the money to be repaid to A?” like (19) or (20), but should rather be interpreted as “Do you think it is obligatory/necessary to repay the money?”. What is exchanged between speaker and interlocutor in this example is not a position of deontic commitment, but rather a position of epistemic commitment to the existence of necessity or obligation (see further in Section 5.2). (21)

[From a letter asking advice about legal matters] Can I use a portion of the proceeds from the first sale effect improvements to the second property? If so, must it be repaid to A later? If a contract of sale has been signed by all parties and my aunt was to die, will that contract still be binding? (CB)

4.3. Conclusion If we look at the function of deontic modals in examples like (13) or (14), and their behaviour in conditionals and interrogative contexts, there is no

The problem of subjective modality in the FG model

259

reason to exclude deontic modality from the subjective category. Deontic modals can be subjective just like epistemic modals, the only difference being the domain of the modality. Whereas epistemic modals express positions concerning the truth of propositions, deontic modals express positions concerning the desirability of actions. Within the traditional FG conception of layering, it is probably this distinction of domains that has been the motivation for the exclusion of deontic modality from the subjective category, as I will argue in Section 6.

5. Subjective deontic versus subjective epistemic modality: different domains As I tried to show in the previous section, deontic modality can be subjective just like epistemic modality, both in terms of the function of encoding positions of commitment, and in terms of the behaviour in reaction to the different criteria for subjectivity. In this section, I will argue that the difference between the two categories lies in the domain over which they operate rather than in their subjective status. Subjective epistemic modality encodes positions of commitment to the truth of propositions, whereas subjective deontic modality encodes positions of commitment to the desirability of actions. I will show that this functional difference is also reflected in the grammatical behaviour of the categories, more particularly in terms of the category of tense: subjective epistemic modality operates over tensed domains, whereas subjective deontic modality operates over tenseless domains.

5.1. Tensed versus tenseless domains The distinction between tensed and tenseless domains has traditionally been associated with the distinction between the basic mood types of indicative and imperative (Bolinger 1967, 1977; Lyons 1977: 746–747; Bolkestein 1980: 41). (22) (23)

John is aware of the problems. Give me the money!

In indicative clauses like (22), the SoA represented is always located with respect to the temporal zero-point (the time of encoding and/or decoding by speaker and/or hearer, Declerck 1991b): the category of tense locates the actualization of the SoA in the past, present or future relative to

260

Jean-Christophe Verstraete

the temporal zero-point. In imperative clauses like (23), on the other hand, there is no tense marking and the SoA is consequently not located with respect to the temporal zero-point. It might be objected that imperatives are not tenseless but have an inherently future orientation, but this ‘future’ orientation cannot be considered a tense the way future tense is in the indicative mood. Future tense establishes a relation to the speaker’s hereand-now by locating the actualization of an SoA at a point in time which is later than the time of utterance, but the so-called ‘future’ orientation in imperatives cannot reasonably be said to locate the actualization of the SoA at a particular point in time later than the moment of utterance. As an object of the speaker’s desire or will, the SoA in an imperative is not ‘located’ but rather a purely ‘virtual’ concept (Bolinger 1968, 1977). Virtuality is very different from location in the future: location in the future still entails some relation with the world of the speaker’s here-and-now, which can be evaluated in terms of truth and falsity, whereas virtuality implies no relation at all with the world of the here-and-now and is therefore outside the realm of truth and falsity. The same contrast between tensed and tenseless domains that distinguishes indicative from imperative mood also distinguishes subjective epistemic modality from subjective deontic modality. This becomes especially clear if we look at the present-perfect contrast for the main verb following the subjective modal auxiliary, which has a different functional value depending on the epistemic or deontic nature of the modal. For instance, if we compare (24) with (25): (24) (25)

John must surely be aware of the problems. John must surely have been aware of the problems. Jack must give me the money, or I'll kill him. Jack must have given me the money (by ten), or I'll kill him.

In subjective-epistemic (24), the function of the contrast between present and perfect main verb is to locate the actualization of the SoA differently with respect to the temporal zero-point: perfect locates it in the past relative to the zero-point, whereas present locates it in the non-past. Thus, subjective epistemic modality operates over a tensed domain, and is therefore paradigmatically equivalent to the bare indicative mood. In subjectivedeontic (25), on the other hand, the contrast between present and perfect does not serve to locate the actualization of the SoA differently relative to the temporal zero-point. Both with present and with perfect main verbs, the SoA is a desired and therefore still virtual SoA just as in imperative

The problem of subjective modality in the FG model

261

clauses: the perfect in this case does not locate in the past, but simply indicates that actualization of the virtual SoA denoted by the predication is desired by a specific point in time (here lexically realized in by ten). Thus, subjective deontic modality operates over tenseless domains and is therefore paradigmatically equivalent to the imperative mood: its SoA is not located in time with respect to the temporal zero-point, but is simply an unlocalized, virtual SoA.

5.2. Subjective versus non-subjective deontic modality The paradigmatic equivalence between the imperative mood and subjective deontic modality has sometimes been challenged on the basis of the fact that deontic modals can have features which are normally excluded for imperatives, like tense marking and propositional attitude markers (Bolkestein 1980: 36–47). In this section, I will show that these arguments apply only to nonsubjective (objective or inherent) uses of deontic modality, but crucially not to subjective uses. More particularly, I will argue that what is interpersonally at issue in structures with nonsubjective deontic modals is the subjective epistemic modalization realized in the indicative mood: this is what explains the presence of typically epistemic features such as tense and propositional attitude marking in such structures. (26)

You must ask him if he fancies me and love him and ask him why he says he’d phone me that often, ask him that, you must say, right yeah but don’t tell him that I told you to ask him, yeah (CB) (27) The 16-strong Scottish squad is free to travel south and whoop it up. Johnston’s ex-Rangers pals - six are named in the party - are likely to take advantage, but the proviso is they must be back at Motherwell by 2.30 pm on Sunday to work off their excesses.

The deontic modal must in (26) above is subjective: it serves to encode the speaker’s commitment to the desirability of the action in question, and the SoA with respect to which the speaker expresses this commitment is not located relative to the temporal zero-point but still ‘virtual’. This subjective use of deontic modality can unproblematically be regarded as a paradigmatic equivalent of the imperative, which equally expresses the speaker’s commitment to the desirability of the action and equally operates on tenseless SoAs. In this sense, it is not a coincidence that subjective deontic must in (26) naturally alternates with imperative structures in one and the same stretch of discourse.

262

Jean-Christophe Verstraete

The deontic modal must in (27) is objective: it does not express the speaker’s commitment to the desirability of the action, but merely reports on the existence of an obligation originating from another deontic source. Structures with this type of deontic modality behave very differently from structures with subjective deontic modality. (28) (29) (30)

The Scottish squad had to be / will have to be back at Motherwell by 2.30 pm on Sunday to work off their excesses. Fortunately, the Scottish squad must be back at Motherwell by 2.30 pm on Sunday to work off their excesses. Must the Scottish squad be back at Motherwell by 2.30 pm on Sunday?

First of all, they are not tenseless but tensed: the issue in these structures is no longer the speaker’s commitment to the obligation, but rather the existence of the obligation in question. This existence can be located in the past, present or future just like any other SoA, as shown in (28). Secondly, structures like (27) also allow the expression of propositional attitude markers like probably, possibly, fortunately or sadly (Bolkestein 1980: 40– 42), as shown in (29). Again, the object with respect to which the attitude is expressed is the existence of the obligation rather than any speaker’s commitment to such obligation. Finally, objective deontic modality behaves differently from its subjective counterpart in reaction to interrogation. The responsibility transferred to the interlocutor’s next turn in (30) is not deontic (“Do you want the players to …?”) as would be the case for subjective deontic uses like (26) but epistemic (“Is it the case that the players are obliged to …?”). What these features show is that structures with objective deontic modality are subjectively modalized in epistemic terms, by the indicative mood. What is interpersonally at issue in these structures is not deontic commitment to desirability of actions but rather epistemic commitment encoded by the indicative mood: the speaker’s or interlocutor’s commitment to the truth of a proposition about the existence of an obligation (“Is it or is it not the case that this obligation exists?”). The presence of subjective epistemic modalization in such structures can explain the availability of tense: as shown in the previous section, subjective epistemic modality operates on tensed SoAs. The availability of propositional attitude markers is also a typically epistemic feature: propositional attitudes can only be expressed in epistemic utterances, where the truth of a proposition is at stake, but not in subjective deontic utterances where the desirability of an action is at stake (Verstraete 2000, see also Moutaouakil 1996: 212–213). The be-

The problem of subjective modality in the FG model

263

haviour of interrogation, finally, can also be linked up with the subjective epistemic modalization encoded in the indicative. What the interrogative transfers to the interlocutor in (30) is the responsibility for an epistemic rather than a deontic position, typically paraphrased with an epistemic predicate (“Do you think it is the case that the players are obliged to …?” rather than “Do you want the players to …?”). In this sense, the objections raised in the literature against the paradigmatic equivalence between deontic modality and the imperative mood are valid only for nonsubjective deontic modality: structures with nonsubjective deontic modals are epistemically modalized by the indicative mood, and can therefore not be regarded as paradigmatically equivalent to imperatives. But this argument crucially applies only to nonsubjective deontic modality. Subjective deontic modality does not show any of these features: it does operate on tenseless SoAs, does not allow expression of propositional attitude markers10 and the speaker-interlocutor transfer effected by the interrogative is deontic rather than epistemic. In this sense, the arguments against paradigmatic equivalence with the imperative cannot be used for the deontic category as a whole, but should also take into account the functional diversity within the deontic category.

5.3. Summary of the model Table 3 below summarizes the most important differences between the traditional FG treatment of modality and the alternative proposal put forward in this chapter. The central point of divergence between the two proposals is the possibility of subjective status for deontic modality, but this also has repercussions for the overall organization of the model. At the level of speaker-hearer interaction, this implies that there is a functional dichotomy between utterances where speaker and interlocutor negotiate about the desirability of actions, and those where speaker and interlocutor negotiate about the truth of propositions. This bifurcation is also continued at the representational level, in the sense that the material negotiated in the interaction is different depending on the subjective-deontic or subjectiveepistemic nature of the negotiation. The relevant parameter here is the category of tense, with its function of locating the SoA with respect to the speaker’s temporal zero-point: subjective-deontic modality operates on bare SoAs, which are not located with respect to the temporal zero-point, whereas subjective-epistemic modality operates on tensed SoAs, because

264

Jean-Christophe Verstraete

location with respect to the temporal zero-point is a necessary prerequisite for arguing about truth or falsity (see Halliday 1994: 75). Table 3. The standard FG model and the proposed alternative

FG model Alternative

Interpersonal Epistemic

Representational Tense

Epistemic Indicative

Tense

Deontic Imperative

X

Deontic Dynamic Deontic Dynamic

6. Implications for the FG model(s) In the final section of this chapter, I will confront my analysis of subjective modality both with the traditional layered model of the clause as it is presented in Hengeveld (1989) and Dik (1997), and with the alternative model proposed in Hengeveld (this volume). I will argue that the exclusion of deontic modality from the subjective-modal category (Hengeveld 1988, 1989) is a natural consequence of the way the theory of layering has been formulated so far: given the assumption that the full set of layers is present in every (main) clause, the difference in domain between epistemic and deontic modality naturally excludes the possibility of subjective deontic modality. I will also show, however, that the modular top-down organization proposed in Hengeveld (this volume) is better suited to dealing with subjective deontic modality. On the one hand, the modular separation between the interpersonal and representational components allows one to dissociate the question of the subjective status of epistemic and deontic modality from the question of the type of domain over which the modals operate: the domains of the modal operators are a representational issue, whereas the subjective status of the operator is an interpersonal issue that can be dealt with in its own right. On the other hand, the connection between the interpersonal distinction [subjective epistemic vs subjective deontic] and the representational distinction [tensed SoAs vs tenseless SoAs] can still be incorporated in the model as an example of top-down organization, where representational distinctions are steered from the interpersonal component.

The problem of subjective modality in the FG model

265

6.1. Theory-internal problems with subjective-deontic modality The exclusion of deontic modality from the subjective category in earlier FG proposals can probably be attributed to the understanding of layering that is implicit in the model presented in Hengeveld (1989) and Dik (1997). Given the difference in domain between epistemic modality (the truth of propositions) and deontic modality (the desirability of actions), it is a logical step within the traditional FG version of layering to exclude deontic modality from the subjective category, as I will try to show in this section. The basic difference between subjective epistemic modality and (what I claim to be) subjective deontic modality is a difference in domain. As shown in the previous section, subjective epistemic modality expresses commitment to the truth of propositions, whereas subjective-deontic modality expresses commitment to the desirability of actions. In grammatical terms, this difference in domain is reflected in the presence or absence of tense in the non-modalized part of the utterance: subjective epistemic modality operates over tensed SoAs, located with respect to the speaker’s temporal zero-point, whereas subjective deontic modality operates over tenseless, virtual SoAs. This is a descriptively relevant distinction that will probably be recognized in most analyses, but the question is how to incorporate it in the model. The decision within FG to locate the ‘highest’ types of epistemic and deontic modality at different sides of the tense operator can be regarded as one way to do justice to the distinction in domains – epistemic modality at the proposition level, with scope over the tense operator, and deontic modality at the predication level, without scope over the tense operator. I fully agree that the distinction between subjective epistemic and deontic modality in terms of tense should somehow be included in the model, but the question is whether association of the modals with different layers is the right way to do this. In the FG conception of layering, association with different layers necessarily also implies different status for operators, and this makes it impossible to analyse both epistemic and deontic modality as subjective. If the highest type of deontic modality is associated with the predication, and the highest type of epistemic modality is associated with the proposition, it is impossible to do justice to the functional and behavioural similarities between the two categories that point towards a shared subjective status. The reason is that the full set of layers is considered to be relevant for every type of (main) clause, more or less in the form of a template: in this sense, utterances with subjective deontic modality as an operator over the predicational layer still have to contain a propositional

266

Jean-Christophe Verstraete

layer with an associated slot for the subjective epistemic operator,11 so that the deontic modal cannot be analysed as subjective in the same way as epistemic modals. Thus, the assumption that every main clause contains a propositional layer is the most important obstacle to the inclusion of subjective deontic modality in the model. The presence of a propositional layer need not be regarded as a universal characteristic of clauses, however: this idea probably derives from the assumption implicit in many frameworks that every utterance is fundamentally about knowledge. Following Halliday (1975, 1994) and McGregor (1997), I believe that the subjective-deontic domain of action and the subjective-epistemic domain of knowledge constitute a fundamental dichotomy in the system, such that in terms of subjective modalization any utterance belongs either to the deontic domain or to the epistemic domain, i.e. that any utterance argues either about action or about knowledge. This also implies that only epistemic utterances contain a propositional layer. Subjective deontic structures lack all propositional characteristics (McGregor 1997: 216–217), both functionally (because they argue about virtual action) and grammatically (because they are tenseless): what subjective deontic modality operates on is not propositions but ‘unlocalized’ SoAs, tenseless ‘virtual’ SoAs which are considered by the speaker in terms of desirability. Thus, one alternative to the traditional FG analysis is to give up the implicit assumption that every main clause structure has a propositional layer.12 In this way the distinction in tense between subjective epistemic and subjective deontic modality can be incorporated in the model (in terms of the presence and absence of a propositional layer in the structure of the clause) in a way that does not a priori exclude the parallel analysis of the two modal categories as subjective. As I will show in the following section, however, implementing this alternative proposal requires two types of modification in the traditional formulation of the FG model: a modular separation between the interpersonal and the representational structures of the clause, and optionality of layering in the representational structure.

6.2. The usefulness of modularity and top-down organization The two most important features in Hengeveld’s (this volume) alternative architecture for the FG model are modularity and top-down orientation. Modularity means that the continuous layering from terms to illocutionary frames as seen in the traditional model is abandoned in favour of a distinction between different modules dealing with different aspects of the

The problem of subjective modality in the FG model

267

organization of the utterance. There is one module that takes care of interpersonal organization, and another that takes care of matters of representation, much as in frameworks like Halliday’s (1994) Systemic Functional Grammar and McGregor’s (1997) Semiotic Grammar. Topdown orientation, on the other hand, implies that in this modular framework the interpersonal module, which takes care of speaker-hearer interaction, is primary in that it can determine choices in the representational module. In this section, I will show how both of these features are directly reflected in the behaviour of subjective modality as it has been analysed in this chapter. As explained in the previous section, the main problem with the traditional layered model was the assumption that the full set of layers is considered to be present in any type of utterance. In this perspective, association with different layers for the highest types of epistemic and deontic modality necessarily implies different status. If the highest type of deontic modality is associated with the predication, this implies that it cannot have the same subjective status as the highest type of epistemic modality, since every utterance – including those with the ‘highest’ type of deontic modality – always contains a propositional layer with a slot for precisely this subjective-epistemic modal operator. As argued in the previous section, I fully agree with the descriptive motivation – the divergent behaviour of tense – for associating the highest types of epistemic and deontic modality with the proposition and the predication, respectively, but I do not agree that this should imply a different status for the two categories of modality. In a modular system, like that of Halliday (1994), McGregor (1997) or Hengeveld (this volume), the representational association of the modal categories no longer has any influence on their interpersonal status, because representational and interpersonal functions belong to separate modules. The fact that epistemic and deontic modality have different domains (tensed versus tenseless SoAs) is a purely representational matter that can be dealt with in the representational component and does not interfere with the question of subjective status. The subjective status of epistemic and deontic modality, on the other hand, is an interpersonal issue in its own right that can be determined on the basis of function (encoding positions of commitment) and behaviour (in reaction to conditionality and interrogation) and is not related to representational questions at all. The connections that do exist between the interpersonal and representational components, finally, can be captured in the top-down orientation of the model. The representational choice between tensed and tenseless SoAs is steered from the interpersonal component by the choice between subjective

268

Jean-Christophe Verstraete

epistemic and subjective deontic modality. Table 4 below provides a schematic representation of how the analysis of subjective modality proposed in this study can be dealt with in terms of the modular architecture proposed by Hengeveld (this volume). In sum, we can say that the descriptive problems discussed in the first part of this chapter can generally be accommodated more easily in a modular architecture with top-down orientation between interpersonal and representational components than in the traditional FG model. Because the modular approach separates the interpersonal status of an operator from possibly divergent representational associations, it also removes the most important obstacle to the inclusion of deontic modality in the subjective category: association with different layers at the representational level does not necessarily have implications for the interpersonal status of epistemic versus deontic modals. On the other hand, the top-down orientation still allows one to incorporate the link between the epistemic-deontic distinction at the interpersonal level and the tensed-tenseless distinction at the representational level, in the sense that this representational choice is steered by a choice at the interpersonal level.

Table 4. Modularity and top-down organization for subjective modality in English Interpersonal module Meaning Subjective status of the modal

Speaker’s commitment Echo-effect in conditionals

Form Interlocutor-transfer in interrogatives Epistemic (truth) and deontic (desirability)

Representational module Meaning

‘Located’ propositions

‘Virtual’ actions

Form

+ Tense

- Tense

Domain Of the modal

The problem of subjective modality in the FG model

269

There is one problem, however, which still remains unaddressed in the model proposed by Hengeveld (this volume), and that is the problem of optionality of layering at the representational level. The top-down orientation of the model can easily deal with the fact that the interpersonal choice between epistemic and deontic modality leads to a representational choice between tensed SoAs and tenseless SoAs, but this still requires a way for the representational module to deal with the fact that some structures are tensed and others are not. Assuming that the distinction between tensed and tenseless SoAs corresponds to the distinction between propositions and predications in the representational module,13 it is necessary to explicitly include optionality in the representational layering. Optionality of the propositional layer is a necessary prerequisite for adequate modelling of structures with subjective deontic modality.

Notes 1.

2.

3. 4.

5. 6.

I would like to thank Bert Cornillie, Kristin Davidse, Renaat Declerck, Patrick Goethals, Kees Hengeveld, Peter Lauwers, Bill McGregor, Jan Nuyts, William Van Belle, Dieter Vermandere, and an anonymous referee for more general discussions about modality and/or comments on a previous draft of this chapter. A preliminary version of this chapter was presented at the 9th International Conference on Functional Grammar in Madrid. I would like to thank the members of the audience for their remarks and criticisms. Thanks also to the late Machtelt Bolkestein for making available a copy of Bolkestein (1980). Examples taken from the Cobuild corpus are marked with CB, and examples taken from the ICE-GB corpus are marked with their standard ICE-GB text code. The relevant modal is always underlined. See Verstraete (2001a) for an overview of the distinction between subjective and objective modality in terms of various theories of layering. As shown by Nuyts (1992, 1993), however, the difference between modal adjectives and adverbs cannot be explained in terms of the distinction between objective and subjective function, but must be related to three interacting parameters, viz. evidentiality, discourse functionality and performativity (Nuyts 1993). In fact, this was the original context where this criterion was introduced; see Jackendoff (1972) and Bellert (1977). It is important to note that it is this commitment-suspending function of the conditional marker rather than the presence of the conditional marker as such that leads to an echo effect. As shown in Declerck and Reed (2001),

270

7.

8. 9. 10.

11.

12.

13.

Jean-Christophe Verstraete not all clauses introduced by if have the function of suspending the speaker's commitment: there are for instance what they call performative conditionals, where the if-clause contains a performatively used performative verb. Accordingly, such conditionals will not trigger any echo effect for subjective modals, since it is only the commitment-suspending function of if that leads to echo interpretation for subjective modals. In fact, this use of deontic modality is more typically realized by periphrastic modals like have to rather than modal auxiliaries like must (Palmer 1990: 113–116, Declerck 1991a: 376–377). See, however, the more descriptively oriented typology in Hengeveld (forthcoming). In this respect it is not surprising that Palmer (1990: 113) lists such examples as dynamic necessity. That is, of course, without change of interpretation. Adding a propositional attitude marker to a subjective deontic expression is always possible, but necessarily shifts it from the subjective to the objective category. It is in fact a common rhetorical strategy to add propositional attitude markers to orders or prohibitions, in order to present the obligation as somehow existing independently of the speaker, as in Unfortunately, you cannot enter this building. As I have shown in Section 5.2, such configurations do occur, but crucially only in structures with nonsubjective deontic modality, which are interpersonally epistemic utterances about the existence of necessity or obligation. There have been a number of proposals in the same direction within the FG tradition, for instance in Moutaouakil’s (1996) argument that only declarative clauses contain a propositional layer and in Hengeveld’s (1990) argument that imperative clauses lack a propositional layer. If the propositional layer is tied to the grammatical feature of tense, however, neither of these proposals is entirely adequate: epistemic interrogatives are tensed just like their declarative counterparts, and subjective-deontic declaratives are tenseless just like their imperative counterparts. In the traditional model, the distinction between proposition and predication was motivated in terms of the interpersonal-representational distinction (Hengeveld 1989: 127–131). In Hengeveld’s new proposal (this volume), what used to be the propositional layer now belongs both to the interpersonal component (as the ‘C’ variable) and to the representational component (as the ‘p’ variable). This implies that the proposition-predication distinction is no longer motivated by the interpersonal-representational distinction but requires a new motivation. The structural property of presence vs absence of tense is probably a good candidate for this purpose: location with respect to the temporal zero-point is a necessary prerequisite for epistemic negotiation about truth or falsity, as argued by Halliday (1994: 75).

The problem of subjective modality in the FG model

271

References Bellert, Irene 1977 On semantic and distributional properties of sentential adverbs. Linguistic Inquiry 8. 337–350. Bolinger, Dwight 1967 The imperative in English. In: To Honor Roman Jakobson, 335–362. The Hague: Mouton. 1977 Is the imperative an infinitive? In: Dwight Bolinger, Meaning and Form, 152–182. London: Longman. Bolkestein, A. Machtelt 1980 Problems in the Description of Modal Verbs. Assen: Van Gorcum. Dancygier, Barbara 1998 Conditionals and Prediction: Time, Knowledge and Causation in the Grammar of English if-then Constructions. Cambridge: Cambridge University Press. Davies, Eirian 1979 On the Semantics of Syntax. Mood and Condition in English. London: Croom Helm. Declerck, Renaat 1991a A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. 1991b Tense in English. Its Structure and Use in Discourse. London: Routledge. Declerck, Renaat and Susan Reed 2001 Conditionals. A Data-driven Analysis. Berlin and New York: Mouton. Dik, Simon 1997 The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. Foley, William and Robert Van Valin 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Goossens, Louis 1996 English Modals and Functional Models: A Confrontation. (Antwerp Papers in Linguistics 86.) Antwerp: University of Antwerp. Halliday, Michael A. K. 1970 Functional diversity in language as seen from a consideration of modality and mood in English. Foundations of Language 6: 322–361. 1975 Learning How to Mean. Explorations in the Development of Language. London: Longman. 1994 An Introduction to Functional Grammar. London: Arnold.

272

Jean-Christophe Verstraete

Hare, Richard 1970 Meaning and speech acts. Philosophical Review 79: 3–24. Hengeveld, Kees 1987 Clause structure and modality in Functional Grammar. In: Johan van der Auwera and Louis Goossens (eds), Ins and Outs of the Predication, 53–66. Dordrecht: Foris. 1988 Illocution, mood and modality in a functional grammar of Spanish. Journal of Semantics 6: 227–269. 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. 1990 The hierarchical structure of utterances. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), Layers and Levels of Representation in Language Theory. A Functional View, 1–23. Amsterdam and Philadelphia: Benjamins. This vol. The architecture of a Functional Discourse Grammar. forthc. Mood and modality. In: Geert Booij, Christian Lehmann, and Joachim Mugdan (eds), Morphology: A Handbook on Inflection and Word Formation. Berlin and New York: Mouton. Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Lyons, John 1977 Semantics. Vol. 2. Cambridge: Cambridge University Press. McGregor, William 1997 Semiotic Grammar. Oxford: Oxford University Press. Moutaouakil, Ahmed 1996 On the layering of the underlying clause structure in Functional Grammar. In: Betty Devriendt, Louis Goossens and Johan van der Auwera (eds), Complex Structures. A Functionalist Perspective, 201–227. Berlin and New York: Mouton. Nuyts, Jan 1992 Subjective vs. objective modality: What is the difference? In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered Structure and Reference in a Functional Perspective, 73–97. Amsterdam and Philadelphia: Benjamins. 1993 Modality and the layered representation of conceptual and linguistic structure. Linguistics 31, 933–969. Palmer, Frank 1990 Modality and the English Modals. London: Longman. Verstraete, Jean-Christophe 2000 Attitudinal disjuncts and illocutionary force in clause combining – a response to Bill McGregor. Functions of Language 7: 117–131.

The problem of subjective modality in the FG model 2001

273

Subjective and objective modality: interpersonal and ideational functions in the English modal auxiliary system. Journal of Pragmatics 33: 1505–1528.

Remarks on layering in a cognitive-functional language production model Jan Nuyts

1. Introduction In this chapter I will take another look at the phenomenon of the layering of qualificational categories, in language and beyond. The layered representation of the clause is a core concept in the framework of Functional Grammar (henceforth FG) as developed by Dik (1989b, 1997) and Hengeveld (1989, this volume). In this chapter I will adopt a basic approach to linguistic theorizing which is in a few respects substantially different from that advocated in ‘standard FG’ (as represented in Dik 1997), viz. what I have called a cognitive-pragmatic or cognitive-functional approach (Nuyts 1992a, 2001a). This approach shares with FG a strongly functionalist orientation in the analysis of linguistic structure. But it deviates from mainstream FG in adopting a much more radically cognitive orientation, and, to the extent that there are cognitive assumptions relating to (dimensions of) FG (cf. Dik 1987a, 1989), in adopting a different view on the relationship between linguistic structures in the grammar and conceptual structures figuring in human thought. Correspondingly, a grammatical framework based on these principles – what I have called a Functional Procedural Grammar – has to assume more levels of representation and processing, some of which (notably the conceptual ones) involve degrees of abstractness which go well beyond what would commonly be admitted in FG. Moreover, the present framework is much more procedurally oriented, assuming that producing linguistic expressions is a strongly interactive and flexible process.1 As a consequence, the present framework leads to a view of the layering of qualifications which is substantially different from the standard FG con-

276

Jan Nuyts

ception, both in terms of its position in a model of language production and in terms of its precise format. This chapter, then, will present a sketch of the basic elements of layering in the cognitive-functional perspective adopted here, and I will discuss the how and why of some of the most obvious contrasts with the standard FG view of layered representation.2 My repeated reference to the ‘standard conception’ of FG is important. In the last few years some FG scholars have proposed modified conceptions of FG, which at least in certain respects constitute steps towards a more cognitive and procedural approach to grammar. Probably the first such proposal was Mackenzie’s (1998, 2000). More recent, but completely different from Mackenzie’s, is what is proposed in Hengeveld (this volume). Whether Hengeveld’s proposal has consequences for the issues to be discussed here is not obvious to me as yet. Hengeveld’s concerns are primarily towards an integration of the discourse level into grammar, and this clearly does cause some changes in the role and format of the layered system in FG, inter alia in terms of its applying at or spreading out over multiple levels in the model. But, at least at face value, the fundamental issues which I will be discussing below remain unaffected. Whether this first impression is accurate is to be decided on the basis of a future elaboration of the new trend(s), however, and is therefore beyond the scope of the present chapter. I will assume basic acquaintance with the principles of the (classical) FG layered proposal. I will use Hengeveld’s (1989) version as my reference, but the points to be made are equally applicable to Dik’s (1997) (slightly different) version. For ease of reference, Figure 1 offers a summary overview of the essence of the system. clause interpersonal representational

(E1: [π4 ILL: (S) (A) (π3 X1: [proposition] σ3) σ4]) π2 e1: [π1 pred (x1) ... (xn) σ1] σ2 predication

Figure 1. Basic elements of layered clause structure in Functional Grammar Predicate operators (π1) and satellites (σ1): additional properties of the SoA π1: qualificational aspect ((im)perfective, inchoative, progressive, etc.), predicate negation σ1: additional participants, manner, spatial orientation

Remarks on layering

277

Predication operators (π2) and satellites (σ2): setting of the SoA π2: tense, quantificational aspect (iterative, semelfactive, etc.), objective modality (epistemic and deontic), polarity σ2: spatial and temporal setting, setting relative to other SoAs (cause, cooccurrence, condition, reason, purpose) Proposition operators (π3) and satellites (σ3): validity of the propositional content π3: subjective modality (epistemic, boulomaic), evidentiality (inferential, quotative, experiential) σ3: propositional attitude, validity of proposition (source, evidence, motivation, condition) Illocution operators (π4) and satellites (σ4): communicative strategy of the speaker π4: mitigation or reinforcement of illocutionary force σ4: manner of speech act, communicative setting of speech act

2. A basic assumption: Linguistic and conceptual structure are substantially different I should start by briefly repeating a basic assumption which is critical for understanding the present cognitive-pragmatic view of layered representation, concerning the nature of the human conceptual apparatus and its relation to the linguistic systems.3 Pace Dik’s (1987, 1989a) view that FG predications might serve as the coding device for conceptual representation, there is good evidence for assuming that conceptualization is nonlinguistic. Probably the strongest argument for this assumption derives from a combination of empirical observations about the structural and functional nature of sets of semantically related expression types (henceforth called ‘semantic paradigms’), and basic functionalist views (also accepted in FG) on what a grammar can and cannot do. Firstly, among (the numerous examples of) such semantic paradigms are the alternative expressions of epistemic modality (Nuyts 1993a, 2001a), or the alternative expressions of a basic action scene such as the ‘commercial event’ (Fillmore 1977, 1985, Kay 1996). The variants in such paradigms can be shown to be due to various functional factors which are essentially independent of the basic semantic categories underlying the paradigm, including (quite prominently) matters of perspectivization and information structure. From a language production perspective, this means that all those variants can be considered to originate in one common conceptual notion or cluster, the differences between them emerging in the course of the production process due to, inter alia, contextualization procedures adjusting the information to the local

278

Jan Nuyts

communicative situation. Secondly, the most important basic functionalist principle (also a standard assumption in language psychology, cf. Bock et al. 1992, Bock and Levelt 1994) is the implausibility of transformational operations in a grammar. In this vein, the linguistic variants in semantic paradigms are so fundamentally different that they cannot be derived from each other in the grammar. Consequently, none of the variants can be used to represent the basic semantic category. This unavoidably leads to the conclusion that the semantic category must have a format different from its linguistic realizations. The implications of all this go further than just the claim that conceptualization is non-lexical (i.e. that it does not use lexical elements occurring in natural languages). It also extends to the claim that its principles of organization are fundamentally different from linguistico-semantic organization. Thus, even a model such as Jackendoff’s (1983, 1987, 1990) Conceptual Semantics – which postulates conceptual structures which are nonlexical yet still strongly language-bound, i.e. based on the principle of predicate-argument organization – faces problems with semantic paradigms such as those mentioned above (Nuyts 2001a: 296ff.). None of this should come as a surprise to functionalist researchers: The functional requirements for a central information processing and storage system, which conceptualization is, are completely different from the functional requirements of a communication system such as language. So it is only natural that they should have a completely different shape and organization.4

3. Layering is a conceptual, but not a linguistic phenomenon If conceptualization is non-linguistic, what does it look like? There is no full-blown answer at present, but the evidence mentioned in Section 2 does uncover at least one major property of conceptualization: it must be layered. In fact, the argument in Section 2 simultaneously forces one to conclude that notions such as epistemic modality are not specifically linguistic, but basically conceptual (Nuyts 2001a). And this reasoning applies in equal measure to other qualificational categories (most of which also involve ‘semantic paradigms’), and by extension to the principle that there is relative semantic scope between all these qualificational categories, which is at the heart of the concept of layered representation in FG. Perfectly in line with this conclusion is the observation (Nuyts 1998) that there is a ‘mismatch’ (albeit one clearly motivated by functional prin-

Remarks on layering

279

ciples) between qualificational categories and linguistic forms, in the sense that there is no one-to-one relation between them. Space prevents me from elaborating this here (see for details Nuyts 2001a: 305ff.), but in strongly generalized terms the mismatch comes in two types. One type is involved in the phenomenon of semantic paradigms introduced above: one qualificational category is expressed in a range of different form types. There are two subtypes of this. One qualification as a whole can be expressed in alternative forms: for example, epistemic modality codes into adverbs, adjectives, auxiliaries, and mental state and 5 similar predicates. Or different dimensions of one qualification can on occasion be expressed in two forms, as for instance when a negative polar epistemic evaluation is simultaneously expressed by one form denoting the epistemic scaling and one denoting the negative polarity (cf. I doubt vs I don’t think, or it is doubtful vs it is improbable vs it is not probable). Such one-meaning-to-multiple-forms mappings are to a large extent due to the multi-functionality of language use, i.e. to the interaction of the semantic category with other functional factors pertaining to its linguistic expression (such as, very prominently, information structure), and the different demands those functional dimensions impose on structure (Nuyts 2000). The other type of mismatch works the opposite way: one linguistic form is used by two or more qualificational categories. This, too, comes in two subtypes. Two qualifications may be expressed simultaneously in one form: I think, for example, expresses epistemic likelihood and subjectivity (and as Perkins 1983 shows, such combinations are rather common). Alternatively, two or more qualifications may on different occasions use one and the same form: past tense, for instance, normally marks time, but can sometimes express epistemic modality, viz. when it is used as a weakener on other epistemic expressions, as in the ‘past’ modals in English (e.g. may vs. might) or in ‘performative past’ mental state predicates (i.e. certain uses of I thought). These kinds of mismatch are clearly not accidental or unsystematic. The former subtype is no doubt due to the frequent co-occurrence of the two qualifications in certain usage conditions (for example, subjectivity and epistemic uncertainty frequently go hand in hand in spontaneous conversation). And the latter subtype is a by-product of the principle of expressibility, viz. the long-term striving for qualificational categories to find expression, whereby expression forms are not created out of the blue, but are initially ‘borrowed’ from other meanings on the basis of the principle 6 of (most probably) metonymic transfer. The mismatch between meanings and forms not only underscores the need to separate a conceptual system of qualifications from a linguistic sys-

280

Jan Nuyts

tem of qualificational expressions. It obviously also implies that there is a discrepancy between the two. One may take this to mean that there must be two layered systems, which differ in more or less substantial ways: one in grammar and one in conceptualization. However, the observations rather force one to rethink the role of layered representation in grammar in more fundamental ways. As argued in Nuyts (1998, 2001a), there is no need for a layered system in the cognitive procedures dealing with linguistic structures at all. Layering as such is a purely semantic issue, viz. a matter of semantic scope. But grammar deals with forms, not with meanings. And where forms respond to semantic properties, as is often the case in the behaviour of qualificational expressions, this should be handled in terms of the procedural interaction between linguistic structure and the layered system in conceptualization, rather than by duplicating the latter in grammar. I am not arguing that layering plays no role in grammar: quite on the contrary, its effects on grammatical/syntactic structure and processing are tremendous. But this is precisely the right way to state the issue: we are dealing with effects, not with an inherent feature. To put it metaphorically: layering is an ‘emergent’ property of linguistic structure (more or less in Hopper’s 1998 sense). Some effects of the semantics of layering certainly do become inherent in grammar, i.e. when they are structuralized or grammaticalized in linguistic form. But this process leads to different results for different qualificational expression types, so that it cannot be stated in generalized terms anyway (contrary to what is implied by the postulation of one global layered system in grammar). Thus, if we stick to our example of epistemic modality, in languages such as English, Dutch or German there are considerable differences between the grammatical (auxiliary), lexical (adverbial) and predicational (adjectival, verbal) expressions. This is true in terms of which distinctions they code: the adverbs and adjectives allow a fairly subtle rendering of all positions on the epistemic scale (even more so when combined with modifiers), but the auxiliaries and verbal predicates only allow a vague reference to a few areas on the scale (and they hardly allow additional modifiers). And this is equally true of the behaviour of the forms in these expression types, for example in the range of word-ordering possibilities. Moreover, all these features can be stated in terms of the traditional dimensions of grammatical description, i.e. which forms are available, how can they be combined, and how can they be ordered in an utterance, in view of the basic properties of the ‘parts of speech’ involved. There is no need for any additional constructs such as a layered system. Once again, there is no doubt that the linguistic properties of qualificational forms are

Remarks on layering

281

motivated by the conceptual properties of the qualifications they express. But that is a matter of diachronic effects of meaning on form, and of synchronic procedural links between forms and the basic conceptual dimensions they express.

4. Basic properties of the layered system in conceptualization But then, if we try to model the layered system in conceptualization, can the layered system postulated in FG not serve as a source of inspiration? The foregoing already implies that the format of ‘entities’ in the FG system cannot be taken over (i.e. lexical structures such as the predicate and arguments, or the distinction between operators and satellites). Yet maybe the semantic notions involved in the system and the principles of their organization could be maintained? In some respects the answer may be positive, but in some other respects it appears to be negative. In the following I will try to sketch what layering in conceptualization may involve, with reference to some of the major features of the layered system in FG.

4.1. The gradual nature of the layered system, and the rationale behind it The layered system in conceptualization is (unlike the FG system) most probably a gradual system. This is apparent in observations regarding the semantic scope effects that arise when combining different qualificational expressions in an utterance: one can always determine a semantic dominancy hierarchy between the qualifications involved, which is moreover stable across various tokens of the combination. This is also true of qualifications which in the FG system are on one level, e.g. the predicational one. 7/8 The following examples are self-explanatory: (1)

(2)

(3)

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

John went skiing a few times [quant. aspect] last year [time] It happened last year that John went skiing a few times *It happened a few times that John went skiing last year John must [deontic modality] go skiing next year [time] There is an obligation for John to go skiing next year *It will happen next year that John must/is obliged [at the time of speaking] to go skiing a. John may [epistemic modality]9 go skiing next year [time] b. There is some kind of likelihood that John will go skiing next year c. *It will happen next year that John may go skiing

282

Jan Nuyts

In terms of this kind of reasoning, then, one arrives at the – very incomplete10 – gradual conceptual hierarchy of qualifications in (4): (4)

evidentiality > epistemic modality > deontic modality > time > quantificational aspect > qualificational aspect > state of affairs

This hierarchy (and hence the linguistic-semantic observations on which it is based) is clearly not accidental or arbitrary: there is a straightforward deeper rationale behind it, which ties in with very basic aspects of human cognitive functioning. (This rationale is implicitly or subliminally contained in, or closely approximated by, some aspects of the motivation for the layered system in FG, but it got blurred in the FG account due to the interference of other dimensions which are irrelevant at a conceptual level; see Section 4.4 below.) Climbing up the cline in (4) correlates with a gradual widening of the perspective on the State of Affairs (henceforth SoA). The cline develops from, at the lower end, qualifications which further specify internal features of the SoA, to, at the higher end, qualifications which provide a global assessment of the status of the SoA. Correspondingly, qualifications low in the system require no or hardly any information other than knowledge of the SoA itself, while qualifications high in the system are predominantly or exclusively based on information external to the SoA. In even more basic terms, the system relates to the tension between perception and interpretation in cognitive functioning. Climbing up the system in a way involves a decreasing role for direct perception of the SoA, and an increasing role for interpretation and creative involvement on the part of the speaker. In principle, the basic level of the conceptual representation of the SoA is closest to, i.e. has been acquired through or should be accessible for, direct perception of the world. Any instance of perception is obviously necessarily local, restricted to what is within the direct reach of the perceptive system (whichever is involved) at one point in time. The higher one climbs up in the hierarchy, the more the qualifications appear to be concerned with specifying aspects of this SoA which are beyond the locality of immediate perception, and are thus dependent on, or aim at, abstraction and generalization, and/or on perceptions or information outside the SoA proper.

Remarks on layering

283

For example, qualificational aspect requires the perception of an SoA and its actual state of development, no more and no less. Quantificational aspect already requires repeated perceptions of SoAs and a generalization over them, e.g. in terms of the uniqueness of one perceived SoA relative to others, or the similarity of different perceived SoAs such that this ‘type’ of SoA can be characterized as having a certain frequency, or as being generic, etc. ‘Situation in time’ requires linking the perception of an SoA to the (in itself very complex) perception of natural cycles of the sun and the moon and/or the even more abstract estimation of numbers of such cycles using conventional, hence completely abstract, definitions of time (e.g. the calendar). (Situation in space is similar: determining that one is ‘in Paris’, for example, goes far beyond mere immediate perception of one’s whereabouts.) And epistemic modality is not at all a matter of direct perception of an SoA any more, but of abstract deductive reasoning from perceptions of other SoAs, or generalizations over them, to the tentative postulation of a possible, hypothetical SoA (which should be accessible to perception if the hypothesis is to be verified, of course). Obviously, the fact that the system of layering of qualifications ties in with such very basic features of human perception and information processing lends strong further support to our assumption that we are dealing here with a basic conceptual phenomenon, not with a specifically linguistic 11 one.

4.2. Stacks in the system The foregoing, however, does not mean that the FG concept of ‘stacks’ in the layered system is completely inaccurate, even at the conceptual level. In fact it is possible to distinguish, superimposed upon the gradual hierarchy, groups of qualifications which share certain semantic properties (although the groups do not fully correspond to those in the FG system). One distinction corresponds perfectly to that between the predicate-level and predication-level qualifications in FG: in the system in (4), this line should be drawn between quantificational and qualificational aspect. Qualifications below it further specify the internal structure of the SoA: qualificational aspect, for instance, concerns the state of development of the SoA (and manner of action – cf. note 10 – could be characterized in a similar way). Above this line, qualifications no longer affect the internal structure of the SoA, but situate the SoA as a whole in the world: this applies to quantificational aspect, which concerns the frequency of the SoA, and it applies even more clearly to temporal (and spatial) situation. Another

284

Jan Nuyts

semantic division in (4) – one not present in the FG system, however – lies in between time and deontic modality. Deontic and epistemic modality and evidentiality (and emotional attitude, if it is included in the system) all involve (different) kinds of speaker attitudes, i.e. explicit statements of degrees to which the speaker is committed (in different ways) to the SoA. Qualifications from time downward cannot be called ‘attitudinal’, however (cf. Nuyts 2001a: 344ff.). These stacks do not bring any new ‘logic’ into the system: as will appear from their definition above, they fit perfectly into the general rationale provided in Section 4.1. This also means that the borders between them do not constitute radical breaks (for the rationale implies graduality). Consider the middle stack. Quantificational aspect does not change the properties of the SoA as such, but nevertheless it still draws fairly closely on those internal properties, since it requires a comparative check whether there is identity between instances of a potentially recurrent SoA. Temporal and spatial qualification, however, involve a specification of or situation in the external dimensions of time and space of an SoA, for which purpose the internal structure of the SoA matters much less. If we jump to the upper stack, deontic modality brings in the issue of speaker commitment, which naturally arises due to the increased role of interpretation that emerges from the clash between the SoA and other, external knowledge. But in a way, deontic modality also still situates the SoA, viz. in the social world 12 and its moral values, without putting its reality status as such at issue. Epistemic modality, however, does not situate the SoA any more, but rather broaches the question whether it needs to be situated at all. Clearly, then, the breaks in the system are no more than somewhat more drastic qualitative jumps culminating from an accumulating number of smaller jumps underneath them, which all fit into the same general logic.

4.3. Explaining scope extensions in terms of the rationale behind the system The rationale behind the layered system discussed in Section 4.1 also naturally explains the differences in the scope of qualifications, not only relative to each other, but also over domains of the qualified SoA. That is, in the present account scope is a natural by-product of the logic inherent in the system. FG deals with the scope of qualifications over domains of the SoA by relating the qualifications to different hierarchical levels in clause structure (the predicate, the predication, etc.). This is an essential dimension of the FG account which I have neglected so far. However, this kind

Remarks on layering

285

of construction is probably as superfluous in syntax as is the semantic hierarchy of qualifications itself. For it follows directly from the layered system at the conceptual level. Obviously, if low-level qualifications further specify internal aspects of the SoA, they will naturally affect the relation between the participants in the SoA. Since this relation usually gets expressed as the predicate in a linguistic expression, they appear to be predicate modifiers. If medium-level qualifications situate the SoA as a whole, they affect not only the relation between the participants, but also the participants themselves. So in linguistic expression they will usually appear to affect the predicate plus the core participants, plus the expressions of the low-level qualifications, of course. Finally, high-level committing qualifications affect whatever there is about the SoA, including its external situation, so it is only normal that their expressions will usually also affect the situating operators and satellites. Note that I am relativizing these statements with the labels ‘usually’ or ‘normally’. In fact, the situation need not be as described. Corpus data (Nuyts 2001a: Section 2.4, 3.4 and 4.4) show that a qualification such as epistemic modality, for example, sometimes only affects a subpart of a linguistic expression, such as a single constituent or a parenthetical insertion. Moreover, it has been observed that most or all qualificational dimensions which can operate at the sentence level can also occur within the noun 13 phrase (the term operators in FG – cf. Rijkhoff 1990). Our present approach allows for a straightforward explanation. Depending on how the speaker, in view of the communicative circumstances, decides to organize the chunk of information he aims to express, a relation in conceptualization that is qualified somehow (e.g. epistemically) can be linguistically coded, not in the main pattern of an utterance, but in some subdomain (a noun, a constituent, a parenthetical, etc.). In such cases, then, the qualificational expression can hardly be considered to be attached to – in the epistemic case – the predication in utterance syntax. This observation offers another good reason to consider the extension of scope over domains of the SoA to be a matter of the layered system in conceptualization, and not in syntax.

4.4. Qualifications and dimensions of interaction and discourse In spite of certain correspondences between the FG account of layering and the present approach, many aspects of the organization in the FG system are entirely absent here. I have argued elsewhere (cf. Nuyts 1992a: 196f., 1992b) why the distinction in FG between an interpersonal and a represen-

286

Jan Nuyts

tational dimension in language use (Hengeveld 1989) is not very well motivated. My objections (which I will not repeat here) partly boil down to an intrinsic problem with the concept of separate strata in linguistic organiza14 tion correlating with separate language functions (whatever these are). This clearly involves a departure from Dik’s (1986) view that there is a multiple mapping from functions to forms, such that there is no one-to-one relation between the two dimensions. I wholeheartedly endorse Dik’s original view (cf. Section 3 above), but to go into this matter here would lead us astray. More relevant in the present context is that, if the layered system applies at the conceptual and not at the linguistic level, notions such as ‘speech event’ or ‘narrated event’ (which Hengeveld correlates with the interpersonal and representational dimensions, respectively) are not at stake anyway. A conceptual structure is neither a narrated nor a speech event, it is just a coding of derivations from percepts (SoAs), including a marking of their status (qualifications). That is, a conceptual system, including the layered system (in a way – see Section 5 below), just represents. To illustrate this directly in terms of qualifications: it is hard to see why deontic modality, i.e. whether a speaker considers something good or bad, or desirable or necessary, would be representational, while an evidential category such as inference would be interpersonal. Both just participate in the con15 ceptual representation of reality, no more and no less. Thus, the layered system in FG to a considerable extent integrates interactional – and correlated with this, discursive – elements, not only in the rationale offered for the organization of the system, but also in terms of concepts and notions actually figuring in the system. But none of these belong in a conceptual system of layering, even if some of them do relate to dimensions of the layered system in conceptualization. Let me explain these cryptic statements in some more detail. Firstly, the FG layered system fully integrates the matter of illocutionary force. This is justified in an account of the behaviour of qualificational forms in the grammar, since speech act markers clearly interact with and have properties comparable to expression devices for qualifications. But illocutionary force as such is not a conceptual category. It is certainly not a qualification of an SoA: it is a matter of the speaker’s plans and intentions with an utterance vis-à-vis the hearer in the actual communicative situation, in relation to his/her deeper (non-linguistic) intentions with respect to the world (as relevant at the conceptual level). Thus, unlike qualifications which assess the position of an SoA in the world, illocutionary force is only relevant in connection with the process of producing linguistic expres-

Remarks on layering

287

sions in an actual communicative situation. It is, in fact, a major element driving this process. Certainly, illocution is related in very profound ways to the conceptual system, and even to certain qualificational dimensions there, i.e., probably not accidentally, the high-level ones, belonging to the ‘attitudinal stack’ (see Section 4.2). For one thing, illocution shares with these qualifications the matter of performativity. In fact, quite like illocution, speaker attitudes 16 are bound to the hic et nunc of the speaker. Of course, what is performed in qualification and in illocution are quite different, i.e. a conceptual operation of determining an attitude towards an SoA vs a linguistic act towards an interlocutor. Nevertheless, the ‘performing entity’ is no doubt the same in both (cf. Section 5 below). For another thing, the sources of illocutionary force are obviously also conceptual, and in this respect too, there are links with (the same) qualificational categories. Information questions ensue from gaps or unclarities in conceptual knowledge, i.e. essentially from epistemic uncertainty, and the intention to fill or resolve them. And orders or requests for action, or promises, probably result from an epistemic assessment of the (non-)existence and a deontic assessment of the (un)desirability and/or (non-)necessity of an SoA, and the intention to do something about that. This may explain why illocutionary markers and deontic and epistemic markers are often closely related. It also explains why performative episte17 mic expressions do not occur in questions. Thus, a speaker can deal with an informational gap in two ways, depending on what (s)he aims to achieve. Either (s)he can communicate his/her own view, with an epistemic assessment, of how it might be filled, leading to a declarative with an epistemic expression. Or (s)he can ask the hearer to fill it, leading to a question thematizing the gap or unclarity as such. But then, there is no use for an epistemic expression. In fact, if there is such an expression in other than a speech act-modifying function, then this is automatically part of the thematized gap, hence descriptive (cf. note 16). A second and even more complex issue is how to deal with conditional, temporal or causal relations, relations of purpose, reason, motivation, etc. in linguistic expressions. FG fully integrates these in the layered system, as satellites at the different levels. But conceptually, these dimensions are obviously quite different from ordinary qualifications of SoAs: they concern relations between SoAs. Thus, even as satellites they usually take the form of full subordinate clauses, although the same links exist between syntactically independent clauses. So these relations extend beyond the domain of a single utterance or SoA. How such relations get expressed, then, is a dis-

288

Jan Nuyts

cursive or interactional issue, relating to how the speaker organizes his/her discourse in view of his/her communicative purposes. Still, it is quite obvious that such clause connections do relate to the hierarchy of qualifications, though probably in very complex ways. For example, temporal connections are unavoidably based on the temporal situation of each of the related SoAs. Or consider conditionals. Performative epistemic expressions do not occur in a conditional protasis (Nuyts 2001a: Sections 2.3, 3.3 and 4.3). The explanation may be similar to that for the absence of performative epistemic forms in questions. In the protasis, a speaker brings up a hypothesis about how some conceptual gap might be filled. But (s)he is not conjecturing how (s)he expects the gap to be filled; in that case an epistemic assessment would be to the point. Rather, (s)he is stating one possible way the gap might be filled, with the aim of pointing out (in the apodosis) what consequences that might have, irrespective of its probability. Hence, there is no use for an epistemic expression in the protasis. This analysis does show that conditionals have an intimate link with the domain of epistemic qualification. Nevertheless, this and other dimensions of SoA linking are not part of the layered system as 18 such. How combining SoAs in relation to the layered system in conceptualization does work in a language production system is a matter that goes beyond the scope of the present chapter. With reference to the current concerns in FG with discourse structure, the foregoing obviously implies that in the present cognitive-functional framework, discourse organization as such does not belong in grammar either, i.e. in the cognitive system for processing linguistic structures (I am using the term ‘grammar’ in this chapter in this narrow sense, obviously). I am not certain where or how in Hengeveld’s (this volume) new proposal the lexicon comes in. But assuming that at least the ‘expression level’ is fully lexical, since a complete unit there is a paragraph, I take it that discourse organization is ultimately considered to be entirely part of linguistic structure and processing. In the current cognitive-functional framework, however, grammar is a device solely for utterance processing. Surely, grammatical processing is critically determined by discourse factors, but that is a matter of processing the effects of discourse on the internal constitution of a single utterance. Discourse organization as such is handled in several steps in Functional Procedural Grammar. But central to it is a system (the ‘situational network’) which performs a first adaptation of longterm conceptual knowledge to the actual communicative situation. This system still works in terms of conceptual structures, unlike grammar proper. But unlike the central conceptual system, it uses mid- to short-term

Remarks on layering

289

memory, and in some respects it already adapts conceptual structures to the 19 process of language production. Unfortunately, going further into this debate would lead me too far astray from my present concern with layering 20 (but see De Schutter and Nuyts 1983, Nuyts 2001a: 273ff).

5. A few more thoughts about conceptual layering Let me briefly make a few more final observations pertaining to the status of qualifications – and especially the high-level, attitudinal ones – in conceptualization, which reveal complexities of the issue of layering which clearly extend far beyond what belongs in a linguistic grammar.

5.1. Control If attitudinal qualifications (like speech acts) are ‘performed’ (see Section 4.4), what is it that ‘performs’ them? If climbing up the qualificational hierarchy involves an increasing role for creative involvement of the speaker (see Section 3.1), what causes that involvement? If attitudinal qualifications all involve an explicit statement about types of speaker commitment to the SoA, and can be expressed with or without speaker commitment to the qualification itself (the performativity vs descriptivity issue – see note 16), where does that commitment come from? All these questions strongly point in the direction of a concept which has a long tradition in AI but which apparently must be assumed in cognitive theories of natural intelligence as well: they all suggest the existence of a ‘control unit’ which steers, coordinates and supervises at least some of the operations of the cognitive system. And this no doubt includes in a quite direct way the operations leading to attitudinal qualifications of information about the world (SoAs): they result from the control system’s evaluative comparison of the chunk of knowledge (the SoA) that is to be qualified to other knowledge stored somewhere in the conceptual system. This is not the context in which to elaborate much on this notion of control and its implications (for discussion of its links to other highly interesting notions such as consciousness and attention see Nuyts 1992a, 2001a: 357ff.). But it is well known that this element has pervasive effects on all kinds of perceptive and behavioural systems in human cognition, witness the continuing discussions about automaticity vs control in many areas of cognitive processing. As such, it stands beyond all those individual cognitive systems, including the linguistic. But if attitudinal qualifications are so closely linked to the

290

Jan Nuyts

operations of this unit, then those qualifications (and the system they are part of) can hardly be less independent of the various task-specific perceptual and behavioural systems, including the linguistic one.

5.2. An SoA is specified for all qualifications, conceptually, but not linguistically Speakers rarely mention more than one or a few qualifications of an SoA in an utterance. But that does not mean they have no notion of the status of that SoA in terms of all other qualificational dimensions. What is mentioned in a specific utterance is a matter of what is relevant for the hearer in the very specific, local communicative situation. But knowing about an SoA is not so local, of course: it means anchoring it in one’s long-term knowledge about the world. In line with Chafe’s (1994: 129) adage that “consciousness cannot function without being oriented in space, time, society, and ongoing background events”, then, it is very plausible to assume that for each SoA in conceptual knowledge all qualificational dimensions in the layered system must be present and set for a certain value. Knowledge about events and objects is probably not fully anchored in one’s conceptual system until all dimensions of its status are clear, and if some of those dimensions are not clear, the control unit will strive to resolve the gaps as soon as possible. ‘Anchoring’ an SoA thus means knowing how it relates to other conceptual information, its frequency (aspect), its spatial and temporal situation, its social value (deontic modality), its reality status (epistemic modality), and how one got to know about it (evidentiality). Of course, there is no reason to burden the linguistic system with this: grammar only needs to deal with what actually appears in the linguistic expression.

5.3. Qualifications as representations and as operations In grammar, a qualificational expression is unavoidably always a (lexical) label, a representation. To be sure, it is processed in grammar, but even in that process it remains a label: grammar processes lexical elements, and those are representational things. Conceptually, however, giving qualification the status of a label or a representation will not do: as is implied in Section 5.1, qualifying something is an operation, a process. The matter is complicated, however. Let me lift a corner of the veil. A few times, I have mentioned the distinction between performative and descriptive uses of expressions of high-level qualifications such as epistemic modality. In conceptual terms, the difference can be stated as

Remarks on layering

291

follows. Performative expressions linguistically render the conceptual output of operations of the central control unit directly. But descriptive, commitment-less expressions do not result from an operation of the control system, but render qualifications which have been represented and stored 21 as part of the speaker’s knowledge of the world. This does not mean, however, that the distinction can simply be grasped in terms of an opposition between an operation and a stored representation. For, obviously, an operation over an SoA still results in a value being assigned to the SoA; and, as was implied in Section 5.2, this value must even be placed in longterm storage. What really differentiates qualifications with speaker commitment from those without, then, is this: the former, after having been executed, remain ‘validated’, i.e. they remain coded within the layered system, and this is probably the case as long as the speaker is not forced to reassess them. This does not apply to the latter. Descriptive qualifications involving the speaker’s own former views originally involved validated qualifications in the layered system, which have become invalidated by information acquired later. But for some reason they have not simply been erased from memory, but have been stored as knowledge about one’s former ideas. Descriptive qualifications involving another speaker’s views, on the other hand, are acquired purely through perception, either because one has been informed about them, or because one has inferred them from other observations. So they immediately enter the conceptual system as knowledge about facts in the world, i.e. as a representation of an SoA. When a speaker brings up a performative qualification in communication, however, the ‘operational’ status of the qualification, probably unavoidably, becomes acute again. Let us take the example of epistemic modality again. A speaker will probably only bring up his/her degree of epistemic commitment to an SoA in a discourse if the status of the SoA in this regard is not obvious, to him/herself, or because the hearer turns out to hold a different view, or because there is otherwise new information relevant for the speaker’s view. Thus, one may reasonably assume that an SoA in long-term storage in conceptualization is normally certain knowledge. That is, the epistemic slot in the layered system over it is set for certainty. Consequently, its epistemic status is not an issue and is not brought up, at least not by the speaker, when talking about the SoA. When knowledge is epistemically less than certain to the speaker, however, it is inherently unstable, and (s)he will keep trying to resolve the uncertainty. This means that (s)he keeps processing information potentially relevant to the SoA, thereby epistemically reassessing it. (One may actually wonder whether uncertain information is normally stored for more than a short to interme-

292

Jan Nuyts

diary period of time at all. Similarly, one may wonder whether clearly counterfactual information, i.e. information marked for ‘certainty that not’, makes it into the long-term store of world knowledge, except for rare, highly significant cases, such as ‘that job of a lifetime which one did not get’.) Of course, even if the speaker’s knowledge about an SoA is certain, if a hearer brings in an alternative view, the speaker is again forced to recalculate his/her view in function of what the hearer adds in terms of background information for the SoA. Whatever the details, it is obvious that it is precisely when talking about one’s commitment to an SoA – i.e. precisely when epistemic or other attitudinal elements enter the grammar — that the operational character of these high-level qualifications is crucial. Yet, the grammar is hardly the place to handle those high-level operations as such – in fact, it does not have the infrastructure for them.

6. Conclusion The point of this chapter can be summarized as follows. Linguistic theories of grammar have always shown a tendency to deal with whatever they encounter in linguistic data in the grammar proper. As long as they deal with structural phenomena, that is fine. But when they start dealing with semantic and pragmatic phenomena, they quickly run into the problem of overburdening the grammar with constructs which do not belong there, but belong in other areas of cognition, such as conceptualization. That is, they create grammars with an ‘overcapacity’. I suspect that the layered system in FG – as well as the current tendency in FG to render discourse organization in grammar – is a case in point.

Notes 1. 2.

3. 4.

The same is probably true of understanding linguistic expressions. But I will mainly adopt a production perspective here. The arguments and views presented in this chapter are developed in much more detail, and are moreover underpinned by substantial amounts of experimental and corpus data, in Nuyts (2001b). I have elaborated the arguments for this assumption at length elsewhere (Nuyts 1990, 2001a), so for reasons of space I will not repeat them here. The foregoing actually does not imply that conceptualization must be imagistic (cf. Dik’s ‘perceptual representations’ or Jackendoff’s ‘3D representations’). An action scene such as the ‘commercial event’ involves

Remarks on layering

5.

6.

7. 8.

9.

10.

11.

293

many dimensions which are entirely non-visual (e.g. possession, trading conventions, etc.), and epistemic modal evaluations are by definition entirely non-perceptual. What all of this means for the nature and (no doubt highly complex) organization of the human conceptual systems is beyond the present chapter (but see Nuyts 2001b for discussion). The FG distinction between subjective and objective epistemic (and deontic) modality is conceptually problematic, as I have argued in Nuyts (1992b, 2001a, 2001). So I will not draw this distinction in the remainder of this chapter. Even if one does draw it, however, the arguments made in this chapter remain intact, since each subtype will still attract at least two alternative expression forms. On the pervasiveness of this principle of ‘borrowing’ and its links to diachronic processes such as subjectification and grammaticalization, see Nuyts (1998, 2001a). An asterisk in these examples obviously refers to a semantic anomaly, not (necessarily) to a grammatical one. It is more difficult, if not impossible, to give an example combining a performative epistemic and deontic qualification. But that has little to do with the hierarchical relation between them. This combination – quite like a combination of performative epistemic and evidential forms or of performative deontic and evidential forms, for that matter – is hard to get for reasons that have to do with the basic nature of these qualifications as ‘attitudinal’ or ‘committing’ categories (see below) and the limits this imposes on conceptual processing. See Nuyts (2001a) for an explanation. As mentioned before, I do not adopt the distinction between subjective and objective modality here. But for those who do accept this distinction in FG, example (3) is based on the assumption that in current FG modal auxiliaries would classify as objective (i.e. belonging at the predication layer). The argument actually works equally well with epistemic predicative adjectives, which are no doubt considered objective. Obviously, there are quite a few qualificational categories missing from this list, at least including (inter)subjectivity and mirativity, boulomaic or emotional attitude, volition, and kinds of spatial situation and direction. The reason for omitting them is that their precise position in the system or their exact status relative to other categories is not clear to me at present. Notions such as dynamic modality and manner of action probably fit under the label of one of the types of aspect. The very fundamental nature of the layered system obviously also explains why this system has such a pervasive effect on linguistic development. It is well known that the ontogenesis of modal meanings (dynamic, deontic, epistemic) largely follows one, cross-linguistically recurrent path, which corresponds to climbing up the qualificational hierarchy in (4) (see, for in-

294

12.

13.

14. 15.

16.

Jan Nuyts stance, Stephany 1993 and Hickmann et al. 1993). Similarly, the diachronic evolution of qualificational expressions in general is known to follow the cline in the layered system (cf., for example, Traugott’s many illustrations of the principle which she dubs ‘subjectification – Traugott 1989, 1995, 1997, among others – and which clearly correlates with ‘gaining height’ in the layered system in (4) – cf. Nuyts 2001a). For acquisition, the reason for the correlation is obvious. Increasing reliance on information external to the SoA proper and increasing abstractness and generalization obviously lead to increasing complexity involved in determining the status of the SoA. And, presumably, higher complexity correlates with greater acquisitional difficulty, hence later acquisition. For diachronic evolution this explanation is less evident: why should complexity in an individual mind correlate with order of long-term development in a linguistic community? Yet in the biological world evolution also appears to lead to increasingly complex organisms. So it is tempting to assume that even for diachrony there is something to the matter of the complexity of the qualificational categories, even if it is not immediately apparent in which way. The SoA can be either real or non-real: cf. It is a good thing that you did that vs. You should do that soon. However, this reality status as such is not at issue, but is a ‘precondition’ for the deontic evaluation. Probably the parallelism between the two levels goes very far, in the sense that qualificational expressions at both levels are at least roughly subject to the same organizational principles, microstructurally (i.e. their ordering being determined by semantic scope) and macro-structurally (in terms of the general rationale behind the organization). But showing this is beyond the present chapter. Representational vs interpersonal will probably not do for that purpose anyway: see Nuyts (1992a: 26–64, 1993b). If anything, deontic modality would then appear to have more to do with an interpersonal or social dimension than inference, although that is exactly the opposite of what the FG account implies. But in a way, everything conceptual is social or interpersonal. This is clearly true of ‘good’ or ‘bad’, but the same applies, for example, to what one considers fast or slow, or large or small (i.e. kinds of aspectual qualifications). Mind is a thoroughly social phenomenon, but that observation helps little to structure the layered system. This is reflected in the fact that all attitudinal qualificational expressions show a difference between performative and descriptive uses, i.e., respectively, between uses in which the speaker expresses his/her own current attitude towards the SoA and uses in which a speaker reports on someone else’s, or his/her own but former, attitude without thereby committing him/herself to the SoA at the time of speaking. Compare I think he is the

Remarks on layering

17.

18.

19.

20.

21.

295

murderer and John thinks he is the murderer or At that moment I thought he was the murderer. A similar distinction exists in the use of speech act verbs. But it does not exist in expressions of qualifications from time downwards. It is actually important to stress that the layered system in (4), and the discussion relating to it, only concerns the performative ‘version’ of deontic and epistemic modality and evidentiality. The status of descriptive uses is a completely separate story. I will briefly return to this matter in Section 5 below. If they do, they lose their epistemic status and become speech act modifiers: Dutch misschien ‘maybe’, for instance, can figure in a question, but then it does not express epistemic uncertainty, but acts to turn the question into a tendentious one. See Nuyts (2001a). This relation between qualifications and clause linking immediately explains why there are so many diachronic and synchronic connections between qualificational and discourse markers, as observed, for example, by Traugott and König (1991), and why diachronic developments even of discourse markers do fit into the pattern of continuing subjectification, as Traugott calls it, i.e. of climbing up the ladder of the qualificational hierarchy: they do so by virtue of the relation of discourse markers to qualifications in the layered system. The difference between Hengeveld’s handling of discourse and the account in the present framework could be summarized as follows. Hengeveld seems to adopt an ‘accumulation’ concept of discourse production: the steps taken at the interpersonal and representational levels accumulate at the expression level. Functional Procedural Grammar adopts a ‘tear apart and code into language’ concept: large clusters of information going into what will ultimately emerge in the hearer’s mind as a coherent discourse are present in the speaker’s conceptualization and early pre-linguistic stages of communicative processing only. The whole process of language production is a matter of gradually singling out small chunks of information which get coded in single utterances (albeit in discursively adequate ways). Grammar actually only deals with the last part of that process, the single utterance. It may be relevant to refer to the fact that Systemic Functional Grammar (e.g. Halliday 1994), too, deals with discourse organization in the semantic stratum, and not in the lexico-grammar, for good reasons, which have been elaborated at length in the Systemic literature. This does not necessarily mean that, as part of one’s conceptual representation of the world, descriptive qualifications are not distinguished from the facts about the world which they qualify. Maybe they do maintain a special status, for knowing about their qualificational status is obviously necessarily part of one’s knowledge about them. So they may form a sort of intermediate level between the representation of the SoA and the hierarchy of

296

Jan Nuyts qualifications with speaker commitment. But they are in any case fully subject to the latter, as is obvious from the fact that in linguistic expression they can be affected by any qualification in the layered system with speaker commitment. A descriptive epistemic qualification, for instance, can be qualified by the speaker for aspect, time, deontic modality, epistemic modality, etc.

References Bock, Kathryn and Willem J.M. Levelt 1994 Language production: Grammatical encoding. In: Morton Ann Gernsbacher (ed.), Handbook of psycholinguistics. San Diego: Academic Press, 945–984. Bock, Kathryn, Helga Loebell and Randal Morey 1992 From conceptual roles to structural relations: Bridging the syntactic cleft. Psychological Review 99: 150–171. Dik, Simon C. 1986 On the notion “functional explanation”. Belgian Journal of Linguistics 1: 11–52. 1987 Linguistically motivated knowledge representation. In: Makoto Nagao (ed.), Language and Artificial Intelligence, 145–170. Amsterdam: North-Holland. 1989a Towards a unified cognitive language. In: Frieda J. Heyvaert and F. Steurs (eds), Worlds behind Words, 97–110. Leuven: Leuven University Press. 1989b The Theory of Functional Grammar, Part 1. Dordrecht: Foris. 1997 The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. Fillmore, Charles J. 1977 Topics in lexical semantics. In: Roger W. Cole (ed.), Current Issues in Linguistic Theory, 76–138. Bloomington: Indiana University Press. 1985 Frames and the semantics of understanding. Quaderni di Semantica 6: 222–254. Halliday, Michael A.K. 1994 An Introduction to Functional Grammar. London: Arnold. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. Hickmann, Maya, Christian Champaud and Dominique Bassano 1993 Pragmatics and metapragmatics in the development of epistemic modality. First Language 13: 359–389.

Remarks on layering

297

Hopper, Paul J. 1998 Emergent grammar. In: Michael Tomasello (ed.), The New Psychology of Language, 155–175. Mahwah: Erlbaum. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, MA: MIT-Press. 1987 Consciousness and the Computational Mind. Cambridge, MA: MITPress. 1990 Semantic Structures. Cambridge, MA: MIT-Press. Kay, Paul 1996 Intraspeaker relativity. In: John J. Gumperz and Stephen Levinson (eds), Rethinking Linguistic Relativity, 97–114. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan 1998 The basis of syntax in the holophrase. In: Mike Hannay and Machtelt Bolkestein (eds), Functional Grammar and Verbal Interaction, 267– 296. Amsterdam and Philadelphia: Benjamins. 2000 First things first: Towards an incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. Nuyts, Jan 1990 Linguistic representation and conceptual knowledge representation. In: Jan Nuyts, Machtelt Bolkestein and Co Vet (eds), Layers and Levels of Representation in Language Theory, 263–293. Amsterdam and Philadelphia: Benjamins. 1992a Aspects of a Cognitive-Pragmatic Theory of Language. Amsterdam and Philadelphia: Benjamins. 1992b Subjective vs. objective modality: What is the difference? In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered Structure and Reference in a Functional Perspective, 73–98. Amsterdam and Philadelphia: Benjamins. 1993a From language to conceptualization: The case of epistemic modality. Chicago Linguistic Society 29, vol. 2: 271–286. 1993b On determining the functions of language. Semiotica 94: 201–232. 1998 Layered models of qualifications of states of affairs: Cognition vs. typology? In: Johan van der Auwera, Frank Durieux and Ludo Lejeune (eds), English as a Human Language, 274–284. München: Lincom Europa. 2000 Tensions between discourse structure and conceptual semantics: The syntax of epistemic modal expressions. Studies in Language 24: 103–135. 2001a Epistemic Modality, Language and Conceptualization: A CognitivePragmatic Perspective. Amsterdam and Philadelphia: Benjamins.

298

Jan Nuyts

2001b

Subjectivity as an evidential dimension in epistemic modal expressions. Journal of Pragmatics, 33: 383-400. Perkins, Michael R. 1983 Modal Expressions in English. London: Pinter. Rijkhoff, Jan 1990 Toward a unified analysis of terms and predications. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), Layers and Levels of Representation in Language Theory, 165–191. Amsterdam and Philadelphia: Benjamins. Schutter, Georges de and Jan Nuyts 1983 Toward an integrated model of a functional grammar. In: Simon C. Dik (ed.), Advances in Functional Grammar, 387–404. Dordrecht: Foris. Stephany, Ursula 1993 Modality in first language acquisition: The state of the art. In: Norbert Dittmar and Astrid Reich (eds), Modality in Language Acquisition, 133–144. Berlin and New York: De Gruyter. Traugott, Elizabeth C. 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. 1995 Subjectification in grammaticalisation. In: Dieter Stein and Susan Wright (eds), Subjectivity and Subjectivisation, 31–54. Cambridge: Cambridge University Press. 1997 Subjectification and the development of epistemic meaning: The case of promise and threaten. In: Toril Swan and Olaf J. Westvik (eds), Modality in Germanic Languages: Historical and Comparative Perspectives, 185–210. Berlin and New York: De Gruyter. Traugott, Elizabeth C. and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In: Elizabeth C. Traugott and Bernd Heine (eds), Approaches to Grammaticalization, 189–218. Amsterdam and Philadelphia: Benjamins.

Discourse Structure, the Generalized Parallelism Hypothesis and the Architecture of Functional Grammar Ahmed Moutaouakil 1. Introduction One of the most extensively discussed issues in the recent FG literature is the way in which the current sentence-oriented model can be extended to handle supra-sentential phenomena. In order to come to grips with this problem, two kinds of solutions have been proposed: an ‘extrinsic’ solution consisting in combining FG with other theories (e.g. Gulla 1997) and an ‘intrinsic’ solution intended to achieve the extension in question by internally transforming the FG apparatus itself. Within the latter line of thought, two approaches can be distinguished: an expanding ‘upward layering’ approach (Hengeveld 1997; Moutaouakil 1998, among others) involving a continuum of successively larger units, and a modular approach (Kroon 1997) according to which sentential and supra-sentential (discourse) phenomena are to be dealt with in two separate modules. Hengeveld (this volume) proposes a model of Functional Discourse Grammar (FDG) which integrates these two approaches. My aim in this study is to further argue for Hengeveld’s claim and show that the two approaches in question should indeed not be considered incompatible. In the light of this assumption and in the perspective of a new architecture for FG, I will suggest a general frame for describing and explaining discourse phenomena which is both hierarchical and modular. This frame differs in certain substantial as well as organizational aspects from Hengeveld’s model of FDG but it goes in the same direction and strives to achieve the same theoretical goal, simplification and unification of the theory of FG. Elaborating on the idea put forward in Dik (1997b) that the hierarchical structure and the functional relations postulated for the clause can be pro-

300

Ahmed Moutaouakil

jected onto the discourse level, I will push a step forward the structural parallelism hypothesis advocated in earlier works (Rijkhoff 1992, Moutaouakil 1993) by assuming that the different discourse categories (from word to text) can be said to be underlain, although with various degrees of surface explicitness, by one and the same archetypal, hierarchically organized structure whose (quantitative and/or qualitative) actual realization is regulated by structural and typological parameters. It will be hypothesized, in line with Moutaouakil (1999), that the representation of this structure can be perfectly ‘transmodular’ in the sense that its different parts (i.e. levels and layers) can be represented in separate but interacting modules.

2. Discourse and Discourse Categories The claim that will be advocated here presupposes a re-examination of the notion ‘discourse’ as well as the notion ‘discourse categories’.

2.1. On defining discourse As is well known, discourse is one of the notions that are used rather polysemically in the linguistic (and the paralinguistic) literature. I do not intend, here, to explore the different meanings that the term ‘discourse’ covers in traditional and modern linguistic theories. Rather I will restrict myself to the use of this term within the FG community. According to Mackenzie and Keizer (1991), ‘discourse’ is understood in two main ways: it may be seen as (a) the product of text-creating activity, or (b) the ongoing text-creating process itself. They rightly point out that FG takes the former view, and in current FG literature ‘discourse’ is indeed being used primarily to refer to supra-sentential (textual) stretches (cf. Dik 1997b: 379). In the remainder of this chapter, I adopt the ‘product of text-creating activity’ view but with the difference that I suggest that we call ‘discourse’ any ‘complete communicative unit’, i.e. any utterance fulfilling a communicative purpose in a given setting or, in other words, any utterance with both a content and a communicative intention. According to this conception, any utterance (be it a single word or a mere interjection) can be taken as a discourse if and only if it serves to achieve a certain communicative goal, whatever its length, as we will see below.

Discourse structure, the generalized parallelism hypothesis, and FG

301

Worthy of notice is that, in this view, the concept of discourse covers not only the linguistic expression proper but also its contextual correlates. In this sense, we can say that FG, being a pragmatically oriented grammar, has always been a discourse grammar although it has been, in general, actually restricted to sentential discourses. This means that the problem now being extensively discussed in the FG literature should be reformulated: the required extension should take place not from a sentence FG to a discourse FG but rather from a sentential discourse FG to a textual discourse FG.

2.2. Discourse categorization If we adopt the view advocated above, we can distinguish between four discourse categories: text, clause, term-phrase and word, yielding: (1)

DISCOURSE =

TEXT CLAUSE TERM-PHRASE WORD

The discourse categorization visualized in (1) calls for a number of preliminary remarks. Firstly, the term ‘text’ is taken here to designate any coherent combination of (sequences of) simple or complex clauses. In this sense, it replaces the traditional term ‘discourse’, which will henceforth be used as a generic term referring, as mentioned above, to any complete communicative unit (including a text). Secondly, as rightly pointed out by Dik (1997b: 379), Natural Language Users (hereafter NLUs) typically communicate with each other by means of full texts. In these cases, hierarchical relationships arise between the four discourse categories, as visualized in the following hierarchy, which we may call the Discourse Categories Hierarchy (DCH): (2)

Discourse Categories Hierarchy (DCH): Text > Clause > Term-phrase > Word

However, NLUs can also communicate by using single (simple or complex) clauses: (3)

a. It's cold in here. b. Can you pass the salt, please?

302

Ahmed Moutaouakil

And perfect communication can take place with one-constituent utterances such as the following: (4)

a. Some tea! b. Cheers!

As shown in Moutaouakil (1993, 1996) and extensively demonstrated in Mackenzie (1998), ‘holophrastic’ constructions like (4a–b), contrary to what is commonly asserted, are complete stretches of discourse rather than reduced sentences. In sum, Clause, Term-phrase and Word may display two different uses: ‘integrated’ and ‘free’ uses. In the former case, they organizationally function as shown in hierarchy (2); in the latter case, they behave as autonomous complete discourse units. It will be shown in Section 3 that the difference between the two uses is reflected in the internal structure of these discourse categories. A third preliminary remark concerns a particular type of constructions referred to as ‘Extraclausal constituent + Clause’ constructions in FG. Here are some illustrative examples : (5)

a. b. c. d.

Mary, can you help me? Well, we can now continue the lecture. As for John, he will come tomorrow. She is a nice girl, your neighbour.

The problem constructions like (5a–d) pose relates to their status. In this respect, two approaches can, it seems to me, be suggested. First, one can conceive of the construction at hand as a distinct full-fledged discourse category which one may call ‘Expression’ (Cuvalay 1997) or ‘Sentence’ (Moutaouakil 1988, Dik 1997b). This is indeed, as far as I know, the position commonly taken so far in the FG community (cf. Dik 1978, 1989, 1997b; Moutaouakil 1988, 1989, 1998; Cuvalay 1997, among others). If one adopts this approach, one can position this discourse category ‘Sentence’ between text and clause, as a supra-clausal entity, which yields the following alternative DCH: (6)

Discourse Categories Hierarchy (DCH) Text > Sentence > Clause > Term-phrase > Word

Discourse structure, the generalized parallelism hypothesis, and FG

303

Second, a closer look at the pragmatic and structural properties of extraclausal constituents, as discussed in Dik (1997b), reveals that they are not necessarily associated with clauses. Rather, they may co-occur with any discourse category, i.e. with clauses or whole texts, but also with termphrases or words. Moreover, some extraclausal constituents seem to cooccur only with discourse categories other than clauses. To take an example, ‘topic shifters’, ‘push and pop markers’ and ‘finalizers’ typically function as textual (more particularly conversational) discourse markers. Within an approach that takes these facts into account, it becomes difficult to speak of ‘Sentence’ as a distinct discourse category. Leaving the question open, I will assume in the rest of this chapter that both approaches are tenable and that hierarchies (2) and (6) are thus equally relevant. For the sake of simplicity, however, I will refer by DCH to hierarchy (2) rather than to hierarchy (6). A fourth and final remark is that a text can be divided into sub-parts that are commonly called, depending on the text type, ‘episodes’, ‘passages’, ‘moves, etc. Since these result from a strictly internal sub-categorization of text and otherwise have no existence, these entities cannot be taken as fullfledged discourse categories. That is why they are not mentioned in DCH.

3. Discourse Structure One of the recent tendencies in FG has been to try and establish a structural parallelism between the different discourse categories. In the remainder of this chapter, I will show that one of the logical endpoints of such a tendency is the postulation of a universal abstract archetypal discourse structure whose actualization in these discourse categories takes place according to certain parameters.

3.1. Generalized Parallelism Hypothesis By the ‘Generalized Parallelism Hypothesis’ (GPH), I refer to the assumption that there is a certain structural parallelism between all the different discourse categories discussed above. In the evolution of this idea, two types of structural parallelism may be distinguished which are perceptible in the last decade of FG work: a surface parallelism and an underlying parallelism.

304

Ahmed Moutaouakil

3.1.1. Surface structural parallelism The source of the assumption that certain discourse categories display a surface formal similarity can, in my opinion, be traced back to Dik's (1989) formulation of ordering principles (especially the ‘Principle of CrossDomain Harmony’) which he conceives of as governing word order within both the clause and the term-phrase. In a similar vein, De Groot (1990) argues for an approach based on the idea that word structure (mainly the position of affixes with respect to the stem) can be described in the same way as clause structure, i.e. in terms of placement rules, ordering principles and patterns.

3.1.2. Underlying structural parallelism As for the assumption of underlying structural parallelism, we may say that it has been developed and extended in three steps. In his insightful work on the Noun Phrase, Rijkhoff (1992) puts forward the idea that Predication and Term display a structural parallelism to the extent that they both consist of three layers: a quality layer, a quantity layer and a locality layer. According to Rijkhoff's analysis, the common structure of Predication and Term can be represented as follows: (7)

[Loc-Op [Quant-Op [Qual-Op [Nucleus] Qual-Sat] Quant-Sat] Loc-Sat]

In Moutaouakil (1993), the parallelism assumption is pushed a step further. Starting from constructions like (8a–b) and (9) whose peculiarity is to exhibit modalized terms (i.e. terms with subjective, exclamative or volitional modality), Moutaouakil observes that a parallelism can also be established between the term and the layer at which this type of modality is supposedly located, i.e. the proposition layer: (8)

a. French

Marie travaille avec quel enthousiasme! ‘With what enthusiasm Mary works!’ b. What a nice girl I saw yesterday!

(9)

a. Standard Moroccan Arabic kana l-marhumu kariman was the blessed-nom generous-acc ‘The hopefully to-be-blessed was generous.’

Discourse structure, the generalized parallelism hypothesis, and FG

305

He then suggests adding to the existing term layers a fourth one which he calls the ‘modality layer’. This yields structure (10), assumed to underlie both Term and Proposition : (10) [Mod-Op

Mod-Sat]

[Loc-Op [Quant-Op [Qual-Op [Nucleus] Qual-Sat] Quant-Sat] Loc-Sat]

The next and final step in the development of the parallelism hypothesis comes with Dik’s (1997b) proposal that Clause and Text (Discourse in his terminology) can be said to display a strong similarity, both on the structural level and the relational level. The basic idea is that intra-clausal layered structure as well as intra-clausal functional relations can be projected onto Text. According to this view, the structure which can be taken as common to both Clause and Text may be represented as follows: (11) [Ill-Op [Mod-Op

Mod-Sat] Ill-Sat]

[Loc-Op [Quant-Op [Qual-Op [Nucleus] Qual-Sat] Quant-Sat] Loc-Sat]

It should be noted, finally, that the works (Hengeveld 1997, Cuvalay 1997 and Moutaouakil 1998 among others) in which the suggestion is made to deal with supra-clausal phenomena in terms of an expanding upward-layering approach (where Text is conceived of as a supra-clausal, or indeed a supra-sentential, layer rather than constituting a separate module) can somehow also be regarded as contributions to the development and the extension of GPH in the sense that they all postulate, though to different degrees, a certain structural similarity between clause, sentence and text.

3.2. Archetypal Discourse Structure Before it can be taken as underlying the various discourse categories discussed in the previous section, structure (11), requires, it seems to me, some further refinements and enrichments on both the constituency and the relational levels. Here are some suggestions. According to the typology proposed in Dik (1997b: 384–386), greetings

306

Ahmed Moutaouakil

and leave-takings, summonses, addresses and certain polite expressions such as those exemplified in (12a–d) form a subclass of extraclausal constituents whose function is to manage interaction, to create and maintain “the interactional conditions which must be fulfilled for a discourse event to be implemented”: (12)

a. b. c. d.

Good morning, sir, what time is it? Hey there, where are you going? John, can you give me your pencil? Excuse me, could I have a cigarette?

We can also consider as pertaining to this subclass expressions like You know, listen, look etc., which function, in their grammaticalized use (as in (13) for instance), as mere interactional devices: (13)

You know Listen Look

, let's go to the theatre this evening.

It is clear that, given the status and the function of these expressions, structure (11), as it stands, cannot handle them: they can be located in none of the five available layers. In order to do justice to this kind of expression, I would suggest enriching structure (11) with a third interpersonal layer which we may call the ‘Interactional layer’. Such an enrichment results in the following structure: (14) [Inter-Op [Ill-Op [Mod-Op

Mod-Sat]Ill-Sat]Inter-Sat]

[Loc-Op [Quant-Op [Qual-Op [Nucleus] Qual-Sat] Quant-Sat] Loc-Sat]

Some remarks are in order here. First, the added constituent is a real full layer in the sense that it involves, as do the other layers, both an interactional operator (Inter-Op) and an interactional satellite (Inter-Sat), slots intended to host particles like Hi!, Hey! as well as the quasigrammaticalized expressions exemplified in (12). Second, in structure (14), the interactional layer is the outermost and highest one, taking all the other layers in its scope. This is indeed what the data suggest, witness the following contrast:

Discourse structure, the generalized parallelism hypothesis, and FG (15)

307

a. Hey there, frankly, I don't like your hat! b. *Frankly, hey there, I don't like your hat!

It becomes clear from (15a–b) that interactional expressions preferably occupy the absolute initial position preceding all the other constituents including illocutionary satellites (such as frankly), which corroborates the assumption that they pertain to a layer higher than the illocutionary one. In the discussion of the extension of FG, the problem of the number and the nature of the relevant units (or layers) has received more attention than the relations holding between these units. Yet the notion ‘structure’ implies, as is well known, constituency as well as relationships. In order to give a complete and precise picture of discourse structure, we must also take into account the various relations it involves. In FG, relational structure subsumes two main types of relations: functional and referential relations. Since the FG definition and characterization of these two types of relations are largely well known, I will not go into this matter in any more detail. Let us only keep in mind the following three salient facts. First, the two types of relations differ from each other in the sense that functional relations hold within the discourse structure itself, i.e. between its constituents, whereas referential relations obtain between the structure as a whole and what it refers to. We can speak, therefore, of strictly ‘internal’ relations as opposed to ‘external’ relations. Second, as repeatedly pointed out and emphasized in Dik (1997a and 1997b), the world within which the discourse refers is a mental ‘Discourse Model’ which represents either S's conceptualization of the outside world or a purely fictitious world. Third, as we will see below, not all functional relations are involved in all the discourse categories distinguished so far. Still on constituency, Dik (1997b: 415–422) draws attention to the importance of “the global discourse decisions” that S takes in building up a discourse, such as choosing a discourse type, a discourse style, a discourse world, etc. The importance of such decisions, as Dik convincingly shows, lies in the fact that they co-determine the internal structure of (the sub-part of) the discourse they take in their scope. This obviously means that these decisions must be represented in the underlying discourse structure. As regards the way in which this can be done, I think that the following general reflections may be of some relevance. The exhaustive list of the discourse decisions in question, as established by Dik, includes those which relate to the choice of discourse event (in particular the deictic centre), discourse genre (conversational, narrative, argumentative, etc.), discourse style (formal/informal, polite/familiar, etc.),

308

Ahmed Moutaouakil

discourse world, discourse illocution, discourse tense and discourse topics. In the section he devotes to discourse coherence, Dik (1997b: 433–435) mentions ‘Frame’, which he describes as the “organized knowledge concerning what can be done and said within a given institutional setting”. Given this definition, it is possible, it seems to me, to consider Frame as a special, highly institutionalized discourse type. This reinterpretation is supported by the fact that Dik (1997b: 435) exemplifies this notion with a sonnet, which he mentions as a discourse genre. In the same vein, given the close interdependence between discourse genre and discourse world (cf. Dik 1997b: 418) and in order to reduce the number of layers to a minimum, these two notions can be collapsed into the generic notion ‘discourse type’ and taken as forming together one and the same layer. This collapsing is supported by the fact that Dik (1997b: 422) does not mention discourse world when he speaks about “the highest brackets”, which he restricts to discourse event, discourse genre and discourse style. In Dik's view, all discourse decisions are subsumed by a single generic category of ‘discourse settings’. As regards underlying representation, it is worthy of notice that some of the decisions in question can be coded in the already existing layers of structure (14), as is the case for illocutionary, temporal and topical decisions. Consequently, only the remaining features, i.e. discourse event, discourse type and discourse style, can be included in the category settings. The values this category can take are, thus, the following: EVENT

(16)

SETTING

=

TYPE STYLE

In fact, (16) does not tell the whole story about the underlying representation of the category at hand, for two major problems remain to be solved. They can be formulated as follows: first, where must SETTING be located in the underlying discourse structure; second, what is its exact status there? Concerning the location problem, it is clear that SETTING should stand at the opening of this structure. This is indeed Dik's view. In (1997b: 422) he conceives of the features represented in (16) as opening up “the highest brackets” involving the discourse as a whole. As for the problem of the status of SETTING, I think that one of the most reasonable ways to deal with it is by adding a third level – which we may call by default the “rhetorical level” – to the existing ones, i.e. the Interpersonal and the Representational levels. At this level, three – at least – hierarchically or-

Discourse structure, the generalized parallelism hypothesis, and FG

309

ganized layers may be distinguished, in descending order for the discourse event, the discourse type and the discourse style. With the addition of this level the full underlying discourse structure becomes something like (17): (17)

Archetypal Discourse Structure (ADS)

Rhetorical level

[Ev-Op [Typ-Op [Sty-Op

Sty-Sat] Typ-Sat] Ev-Sat]

Interpersonal level

[Inter-Op [Ill-Op [Mod-Op

Mod-Sat]Ill-Sat]Inter-Sat]

Representational level

[Loc-Op[Quant-Op[Qual-Op

[Nucleus]

Qual-Sat]Quant-Sat]Loc-Sat]

The main argument in favour of such an approach is that the three subcategories of SETTING codetermine, as mentioned above, the form as well as the content of the subsequent discourse and should therefore be considered to function as operators just like the operators of the other, well established layers. In the same vein, although it seems to be relatively less easy to find linguistic expressions which function as discourse type or discourse style satellites, we cannot conclude that satellites of this kind do not exist.1 In any case, satellites are optional constituents and their existence can therefore not be taken as a crucial defining feature of layerhood. Dik (1997b) points out that the discourse decisions under discussion here at hand are ‘global’ in the sense that they affect not a single clause but rather a whole text. This means that the added rhetorical level – or at least some of its layers – is typically a textual level. We will return to this issue in Section 3, where we will discuss the parameters regulating the actualization of structure (17).

3.3. Unified representation or modular representation As regards the underlying representation of the discourse structure discussed above, we can opt for one of two procedures, which I propose to label the ‘unified procedure’ and the ‘modular procedure’. According to the procedure adopted so far in the FG tradition, all the (semantic and pragmatic) properties are coded in the same underlying structure, which is represented in the Grammatical Module of the Model of Natural Language User (MNLU). It is this structure that Expression Rules

310

Ahmed Moutaouakil

operate on. On the other hand, according to a recent tendency (cf. Kroon 1997, Bolkestein 1998, Vet 1998, Van den Berg 1998, Liedtke 1998), a separate module should be added to the MNLU. In fact, I think that there is some ambiguity on the nature and the function of this additional module, probably attributable to the fact that ‘discourse’ and ‘pragmatics’ are not always systematically differentiated. In Kroon’s view, it is conceived of, as far as I can judge, as a ‘discourse module’ (a text module in our terminology) intended to handle supra-sentential phenomena. In this sense, it stands in contrast with a ‘sentence module’ whose task it is to take care of properly intra-sentential phenomena. In Vet’s view, the added module is quite different. It is a pragmatic module intended to deal with the contextually determined properties of linguistic expressions – whatever their length, and even extending to whole texts) – and in particular with speech acts (or nonliteral illocutions). In Moutaouakil (1999), a proposal along the lines of Vet’s view is made which can be further developed as follows. We can say that the two representational procedures described above are theoretically equivalent in the sense that they both fit in with the principles and the organization of FG. It is clear that the definition and the categorization of discourse opted for here, in particular when viewed in the light of GPH, leave no room for the distinction between a ‘discourse (or text) module’ and a ‘sentence module’. All the discourse categories distinguished in hierarchy (2) (or hierarchy (6)) are dealt with in the same way, according to either the unified or the modular procedure. In the latter case, once the added module is understood as a pragmatic (and not exclusively a textual) module, the opposition between ‘upward layering’ and modularity is neutralized and the two approaches can thus go perfectly hand in hand. Two kinds of organization of the MNLU are possible. In one of these, two modules contribute together to taking care of the relevant underlying properties of any complete communicative unit whatever its category (text, clause, term-phrase etc.). In the grammatical module, a (grammatical) underlying structure represents the semantic and structural features, whereas in the pragmatic module (in the sense of Vet) a (pragmatic) underlying structure encodes the contextual speaker-oriented features such as illocutionary force, subjective modality and pragmatic functions. These two underlying structures are taken together as inputs, possibly with underlying structures from other (epistemic, logical, social etc.) modules, to expression rules which deliver the final syntactic form.

Discourse structure, the generalized parallelism hypothesis, and FG

MNLU PRAGMATIC MODULE

GRAMMATICAL MODULE Grammatical Structure

Pragmatic structure

Expression rules

Syntactic form

Figure 1. One organization of the MNLU

MNLU PRAGMATIC MODULE

SEMANTIC MODULE Semantic structure

Pragmatic Structure

GRAMMATICAL MODULE Expression rules

Syntactic Syntactic form form

Figure 2. An alternative organization of the MNLU

311

312

Ahmed Moutaouakil

Alternatively, pure semantic properties (semantic and perspectivizing functions, operators and satellites pertaining to the representational level, etc.) can be represented in a separate semantic module. In this case, the expression rules constituting an autonomous grammatical module operate on the information coded in the two underlying structures represented in the pragmatic and the semantic modules. These two types of organization are visualized in Figures (1) and (2) respectively. Notice that in Hengeveld’s proposal (this volume), the representational and the interpersonal entities are regarded as hierarchically organized modules whereas here they are assumed to be, together with the rhetorical entity, merely layered levels of the same structure which are located in separate but related modules.

4. Archetypal Discourse Structure Actualization Structure (17) can be said to be ‘archetypal’ in the sense that it is postulated as a generic abstract structure whose actualization in stretches of discourse is regulated, as we will assume, by – at least – a structural and a typological parameter.

4.1. The structural parameter As was pointed out above, discourse categories can have two kinds of use, ‘free’ and ‘embedded’. It will be shown here that the actualization of structure (17) in these discourse categories depends on the way in which they are used.

4.1.1. ADS actualization in free use The basic idea which will be developed in what follows is that Archetypal Discourse Structure (ADS) is fully actualized in texts, whereas its explicit actualization in the lower discourse categories is dependent on the (decreasing) hospitality of these categories. Examination of the internal structure of the discourse categories distinguished so far shows that, in their free (non-embedded) occurrences, they differ from each other in their capacity to host ADS. As one might expect, hosting capacity diminishes as we run from the top to the bottom of the DCH. The text is thus the most hospitable category and the word is the least

Discourse structure, the generalized parallelism hypothesis, and FG

313

hospitable. In fact, one may say that ADS is potentially present in any normal communicative event, whatever the discourse category through which it is mediated. The difference between the discourse categories in question resides, therefore, rather in how they house ADS. In general, the full, explicit actualization of ADS takes place in the text; in the other discourse categories, only a part of ADS is expressed, the other part being implicitly taken care of by the context/situation. As may be deduced from the characterization of verbal interaction given in Dik (1997b: 409), the discourse category through which a complete communicative event takes place in the most explicit way is the text. This is why NLUs speak in texts rather than in isolated sentences. We may thus expect the full actualization of structure (17) to obtain optimally in a whole text. This indeed emerges clearly from the contrastive examination of three typologically quite different texts: Brunhoff's fairy-tale The story of Babar the little elephant (already analysed from the point of view of Topic-Focus assignment by Mackenzie and Keizer 1991), Najib Mahfouz's novel Han Al-Halili ('Al-Halili Quarter') and Diderot's Jacques le fataliste. It should be noticed here that the other discourse categories (clause, term-phrase and word), as pointed out in many works (Dik 1997a, Moutaouakil 1993, and, in particular, Mackenzie 1998), cannot be said to be less communicative. They also can be used to carry a complete piece of information, with the difference, however, that part of this information remains implicit, i.e. is transmitted by contextual and/or situational means. In other words, the difference between the discourse categories at hand does not reside in their communicative capacity proper but rather in the degree of explicitness of their communicativity. Concerning the actualization of structure (17) in the discourse categories other than text (clause, term-phrase and word), data suggest that it becomes more restricted as we run from the top to the bottom of DCH. As regards this restriction, it will become clear through the following examination of the internal structure of the clause, the term and the word (where they stand alone as performing autonomous and complete speech events) that (a) this is, in fact, a consequence of the decreasing hosting capacity of these discourse categories, (b) it may affect the layering, the operator values and/or the relations and (c) it operates according to a certain directionality, i.e. from higher to lower levels and layers. When used as a complete discourse unit, a clause can generally display all the layers of the interpersonal and representational levels as well as, though to a lesser extent, as (a) and (c) predict, those of the rhetorical level (which typically occur in larger stretches of discourse). As regards func-

314

Ahmed Moutaouakil

tional relations, pragmatic functions are clearly text-based notions (cf. Dik 1997a: 309–338), although they may be expressed within the clause. As shown above, term structure can be said to run parallel to clause structure. However, the layering of term structure typically does not go beyond the lower layer in the interpersonal level, i.e. the modal layer. It is hard to speak of a ‘term illocutionary layer’, at least as far as explicit term structure is concerned. The values of the modal operator are more restricted in the term than they are in the clause: I have shown elsewhere (Moutaouakil 1993) that only volitional and emotional subjective modality distinctions can occur in a modalized term (excluding evidential distinctions). In languages with a concatenative morphology, word formation is achieved, as is well known, through an affixation (prefixation, suffixation, infixation or circumfixation) rule which applies to a stem. Approached in terms of the GPH, word structure can be conceived of as a partial actualization of structure (17): on a stem standing as a nucleus, prefixes and suffixes build hierarchically organized layers. As one might expect, the actualization of structure (17) in the word is more restricted than it is in the other discourse categories. First, affixes can express all the features of the representational level through a quality layer (Republican, Chinese, panelling, anti-social etc.), a quantity layer (hypercritical, supernatural, overdressed, subhuman etc.), and a locality layer (subway, transatlantic, pre-marital, post-classical, unfair, etc.). Word layering can attain the interpersonal level, as in the case of words with pejorative affixes such as: (misleading, malodorous etc.). However, it cannot go beyond the modal layer: I am aware of no affixes which can be said to express illocutionary features. Second, not all the operator values may obtain at the word level. For instance, the modal values are restricted, it seems to me, to pejorative features. Third and more importantly, words do not have, as one might expect, the same hosting capacity as terms and clauses. A complex word can hardly contain more than two prefixes. This means that it would be difficult to find words involving more than two layers at once. A final word on the text-clause parallelism: the structural similarity between the clause and the text is commonly described in terms of the projection of the structure of the former onto the latter. The facts examined above clearly show that the projection process – if we can speak of such a process at all – should be conceived of as taking place in the reverse direction, i.e. from the text to the clause. With the approach proposed here we no longer need the notion of projection: the similarity between the different discourse categories is not due to a structural projection from one category

Discourse structure, the generalized parallelism hypothesis, and FG

315

to another; it rather resides in the fact that they all borrow their internal organization from one and the same archetypal structure. It seems to me that the parallelism phenomena at hand can be accounted for in a psychologically more adequate way in the light of the assumption that NLUs organize their discourse according to a common archetypal communication structure than if we consider that they do so by projecting the structure of a given discourse category, especially since there is no agreement about the discourse category which is the source of this projection (clause or text). Interestingly enough, this permits us to avoid the undesirable ‘clausecentricity’ of FG.

4.1.2. ADS actualization in integrated use As mentioned above, the discourse categories lower than the text can be used in two ways: they can constitute complete communicative units or occur as (embedded) parts of larger categories. In the latter case, words can constitute parts of terms which stand as parts of a clause; a clause can be embedded in another clause yielding a complex clause; (simple and complex) clauses, when coherently sequenced, form a text. In the previous subsection, we were concerned with the free use of the discourse categories; our aim in what follows is to examine the way in which ADS is actualized in these discourse categories when they are integrated into each other, focusing particularly on the embedding of clauses into clauses and clauses into texts. In the FG framework, several works have been devoted to embedding phenomena (cf. Bolkestein 1990, Moutaouakil 1987, Hengeveld 1996 and Dik 1997b, among others). Once reinterpreted in terms of the actualization of ADS, the basic ideas advocated in these works can be reduced to the following general assumptions: (a) (b) (c)

The actualization of ADS is more restricted in integrated (embedded) parts than it is in integrating (embedding) parts; Integrated parts ‘inherit’ features (layers or operator values) from integrating parts; The inheritance process is a matter of degree.

As regards assumption (a), it is now well established in FG circles that the subordinate part of a complex clause need not be a complete clause: as argued in Hengeveld (1996), it can be a clause or a proposition or even a

316

Ahmed Moutaouakil

predication, depending on the nature of the superordinate predicate or on the semantic function it bears. Assumption (b) is in fact a generalization of the claim advanced in Dik (1997b: 421) that “the Tense operator specifications on successive clauses are inherited from the temporal value which has been fixed for the whole discourse unit”. In fact, successive clauses forming the same (phase of a) text inherit not only the temporal value but all the operator values (illocutionary, modal, aspectual etc.) globally fixed for this (phase of the) text. So, the schema which represents the temporal specifications inheritance in Dik (1997b: 421) becomes, once it has been generalized, something like (18) : (18)

α((X). (Y). (Z)...)

where α stands for any operator value and where X, Y, Z symbolize a sequence of clauses. Assumption (c) means that the inheritance process can be partial or total. In other words, the embedded parts can inherit either some operator values only or all of them. In the former case, the embedded parts (i.e. the clauses of a text for example) can be ultimately reduced, at the underlying structure level, to nothing but their nuclear predications. So, the nucleus of (a phase of) a highly homogeneous text, i.e. a text with neither deicticcentre change nor discourse-type shift, can consist of merely the set of the nuclear predications of its constituent clauses. The re-examination of embedding phenomena in the light of the parameterized actualization of ADS allows us to approach these on a new basis. First, it becomes possible to redefine embedding as a linkage of two discourse units U and U' where a part of the ADS as actualized in U' is specified in U. To put it another way, U' is embedded in U if some values of the ADS actualized in U' are inherited from U. This definition permits us also to handle those cases where embedding is not expressed by any formal marker (i.e. without a subordinator).2 It furthermore makes it possible to characterize coordination as a linkage of discourse units with actualization of the same (part of) ADS. Second, it may be possible to establish a more precise typology of embedded constructions on both quantitative and qualitative grounds: it becomes feasible to classify these constructions with respect to the degree of embedding (which is calculated on the basis of the number of the layers specified in the matrix) and to the values of the embedded layers.

Discourse structure, the generalized parallelism hypothesis, and FG

317

4.2. The typological parameter The third parameter is ‘typological’ in the sense that it implies that the actualization of ADS takes place according to three factors: language type, discourse type and discourse style.

4.2.1. ADS and language types The examination of data from several typologically different languages (namely English, French, Standard Modern Arabic as well as Moroccan, Egyptian and Syrian colloquial Arabics) allows us to make the following general observations. All the languages investigated display the three levels distinguished in ADS as well as the layers these levels subsume. They differ from each other, however, with respect to the values of these layers. Here are some examples illustrating the point. Firstly, Modern Standard Arabic as well as the various Colloquial Arabics examined seems to be more interpersonally oriented than the other languages. The data suggest that Arabic displays a relatively greater number of vocative particles and expressions (cf. Moutaouakil 1989). Moreover, it is a well known fact that “Egyptians always speak by implicature” (and I think that this feature can be generalized to all Arabic languages); if this observation is correct, we can say that Egyptian is a language with a rich illocutionary layer – at least in comparison to the other languages examined. We can say in the same vein that Arabic is one of the more highly modalizing languages in the sense that it displays a rather rich set of subjective modality values both on the clause and the term levels (cf. Moutaouakil 1993, 1999). Secondly, at the representational level, we often find in the Arabist literature the claim that Arabic, in comparison with Indo-European languages, is an aspectual rather than a temporal language. If this claim turns out to be tenable, we may hypothesize that there are languages with rich aspectual quality and quantity layers in contrast with languages with a rich temporal locality layer. And as a last example, Dik (1997a: 181–182) points out that languages differ in the number of demonstratives (from a minimum of two to a maximum of well over twenty) as well as in the types of distinctions these elements express. This again can be taken as a criterion with which to classify languages with respect to the locality layer in the term.

318

Ahmed Moutaouakil

In addition, languages differ not only in the number and the kinds of the layer values but also in the means by which these values are expressed. Some languages use grammatical means; others use lexical ones. For instance, in Arabic, illocutionary and subjective modality values are mediated by morphological means (particles, special morphemes, etc.) rather than lexically. Keeping in mind that grammatical and lexical means are underlyingly represented by operators and satellites respectively and generalizing this observation to all the layers involved in ADS, we may speak of ‘operator-prominent languages’ in contrast to ‘satellite-prominent languages’. A large part of the FG literature has been devoted to the selection and the assignment, in different languages, of perspectivizing functions as well as to the study of Topic and Focus types and sub-types. The results of this work can be reinterpreted in terms of typological actualizations of the relational part of ADS.

4.2.2. ADS and discourse types The realization of ADS in an actual discourse is codetermined by the type of that discourse. Let us concentrate on narrative discourse, whose main distinguishing features (extensively discussed in Moutaouakil 1998) are as follows. In narrative texts, the interpersonal level, as one might expect, is much less activated than it is in other text types. Consider, for example, that in a ‘pure’ narrative text with no deictic-centre change, the fixed global illocution is what Dik (1997b: 419) calls a “default Declarative Illocution”. What is more, in this genre, implicature phenomena hardly occur. Likewise, highly neutral narrative texts (‘récits’ in Benveniste’s 1966 terminology) have the prominent feature that they rarely involve subjective modality distinctions: the narrator is, in this type of text, reduced to a mere ‘paper’ entity (to use Barthes’s 1976 metaphor). At the representational level, on the other hand, the more frequently activated temporal and aspectual values are Past and Perfective respectively (unless the so-called ‘Historical Present’ is used). And as regards pragmatic functions, given that the information conveyed in a narrative text is typically new (unshared), the type of Focus selected is, as one might expect, Completive Focus.

Discourse structure, the generalized parallelism hypothesis, and FG

319

4.2.3. ADS and discourse style The notion ‘style’ is defined in Dik (1997b: 417) as “a sequence of choices ... of means of expression which have consistent values along such polar dimensions as Formality-Informality, Politeness-Familiarity and Conciseness-Redundancy”. Dik points out that the choice of a given style is strongly determined by the choice of the discourse type in the sense that not all kinds of styles can fit in with all discourse types. As for the impact of the style on the discourse, the definition given above seems to restrict it to formal means of expression. In fact, if we represent the style as a layer operator, it becomes clear that it plays, like any operator, a twofold role: on the one hand, it combines with other operators to trigger the rules responsible for the formal realization of the discourse; on the other hand, it codetermines the values of the operators it has in its scope, i.e. the operators of lower layers. Let us take as an example the polar dimension Politeness-Familiarity. The actualization and the formal realization of ADS in polite discourse involve special operations such as the selection of particular interactional devices, the extensive activation of the indirect illocutionary values, the choice of specific lexical items as well as of certain forms of predicates and terms. Notice that the three typological factors under discussion can interact with each other in an interesting way in the sense that certain languages handle certain discourse types and styles better than others do. The most well known example is the case of so-called ‘polite languages’ such as, for example, Japanese and, to a less extent, colloquial Egyptian.

4.2.4. ADS as a communicative universal In the FG framework (cf. Dik 1997a: 7), linguistic universals are defined within an approach which conceive of natural languages as particular solutions to the communication problem, i.e. to the problem of establishing “high-level communicative relationships between human beings”. It is clear that ADS can be considered as one of the common linguistic properties which fit optimally into this functional conception of linguistic universals. Notice that, given its nature, its components and its organization, ADS is compatible with functional theories in which universals are conceived of as Formal-Functional correlations. The following arguments can be taken as militating in favour of the universality of ADS. Firstly, structure (17) reflects, as pointed out above,

320

Ahmed Moutaouakil

the main components and the organization of the communication event itself, which typically involves three operations: (a) representing some State of Affairs occurring in some possible world and constituting the message that S intends to communicate; (b) establishing relationships between S and A on the one hand and between S and the content of this message on the other; (c) choosing the discourse type and the discourse style in which (a) and (b) are to be delivered. Secondly, from the FG point of view, the acquisition of language is approached, as is pointed out in Dik (1997a: 7), in terms of “its development in communicative interaction between the maturing child and its environment”. Here again, the notion ‘communicative interaction’ is central. This would enable us to think of ADS as representing a relatively advanced stage of the child’s progressive mastery of linguistic communicative abilities. And thirdly, it becomes clear from the facts discussed in the previous sub-sections that ADS can be assumed to underlie, with quantitative and qualitative parametrical variance, natural discourses of various categories, types and styles in different types of languages (and probably all language types). Moreover, if we are able to assume that any communicative process involves the three levels of ADS, it becomes not unreasonable to hypothesize that a structure with similar components and organization is also at work in non-verbal (pictorial, musical,3 etc.) communication systems. In sum, ADS can be viewed as a universal structure in which language types make quantitative and/or qualitative choices. In another, stronger formulation, language types partially result from different choices in ADS. Within each language type, the actualization of this structure is regulated by further factors relating to discourse type, discourse style and discourse category.

5. Conclusions One of the possible solutions that can be proposed to the problem of extending the current clause/sentence FG model and transforming it into a more text-oriented model is built on four related assumptions that presuppose each other: (a) that the notion ‘discourse’ covers all kinds of utterances expressing a complete communicative event, (b) that discourse can be formally mediated through four formal categories: text, clause, term-phrase and word (or perhaps five if we take sentence as a full-fledged discourse category), (c) that these categories are underlain, at different degrees, by one and the same archetypal structure, and (d) that this

Discourse structure, the generalized parallelism hypothesis, and FG

321

hierarchically organized structure fits perfectly into a modular FG. This solution can be said to have the following major advantages. First of all, it permits us to unify the FG approach to the different discourse categories. Second, it makes it possible to avoid two theoretically undesirable options: (a) an unnecessarily recourse to other theories and (b) a costly and not always very convincingly justified multiplication of the modules within MNLU on the other hand. Third, it allows us to redefine and extend the central notion of embedding and to further refine the typology of embedded clauses. Fourth, the perhaps most important gain is that this solution provides the universal part of FG with a tool permitting us to describe and explain in a more principled and unified way the similarities and the differences between natural language types as well as those between types of natural discourse. Moreover, Dik’s (1997b: 415) insightful idea that a text “can in many ways be likened to a piece of music” leads to the assumption – hopefully to be verified by further research – that, if the transposability of ADS turns out to be feasible, the solution proposed here can be viewed as paving the way for a General Functional Theory (GFT) whose main task would be to describe the general structure of the communicative process and to account for its actualization in verbal and non-verbal modes of interaction as well as providing particular theories for the various communicative systems. Within such a general theory, FG would stand as a member of a subset of linguistic theories to be compared to and evaluated in relation with the theories of the non-linguistic subset. In this view, ADS would be taken as the actualization in natural language of a more abstract archetypal communication structure constituting one of the primitives of this all-encompassing functional theory. In this study, we have mainly been concerned with the underlying (pragmatic and semantic) side of the structural parallelism between the discourse categories. I think that it would be of great interest to verify, in future research, the extent to which a similar parallelism can be said to also hold (within the grammatical module) at the surface (morphosyntactic) level as well as the extent to which the latter parallelism can be taken as resulting from a projection of the former.

Notes 1.

In arguing for the relevance of the discourse-type layer, Hengeveld (1997) reports that in some languages (Turkish, Tauya and Krongo) certain verb forms only occur in narrative texts. As for the discourse-style layer, its rele-

322

2.

3.

Ahmed Moutaouakil vance is evidenced, for example, by the fact that, in ‘polite’ languages such as Japanese, politeness is coded by both grammatical and lexical means which are to be underlyingly represented as Politeness operators and satellites respectively. By ‘non-formal embedding’, I refer to the kind of linkage that is found in constructions like Paul said: “John is an excellent linguist.” where John is an excellent linguist is an embedded clause functioning as a (Goal) second argument of the matrix predicate said. More generally, this is the type of embedding involved in (parts of) texts where successive clauses inherit some of their (illocutionary, modal, temporal, etc.) features from globally fixed values. It becomes clear that any formally oriented approach to embedding phenomena fails to properly describe the structure of this type of construction. For the structural parallelism between a text and a piece of music, see Dik (1997b: 415). In the same vein, J. Amjad (p.c.) has pointed out to me that the structure of a musical composition is quite similar to the ADS advocated here.

References Barthes, Roland 1976 S/Z. Paris: Points. Benveniste, Emile 1966 Problèmes de linguistique générale. Paris: Seuil. Berg, Marinus van den 1998 An outline of a pragmatic functional grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), 77–106. Bolkestein, A. Machtelt 1990 Sentential complements in Functional Grammar: embedded predications, propositions, utterances in Latin. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds), Layers and Levels of Representation in Language Theory: A Functional View, 71–100. Amsterdam and Philadelphia: Benjamins. 1998 What to do with Topic and Focus? Evaluating pragmatic information. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214. Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds) 1997 Discourse and Pragmatics in Functional Grammar. Berlin and New York: Mouton de Gruyter. Cuvalay, Martine 1997 The Arabic Verb. A Functional Approach to Verbal Expressions in

Discourse structure, the generalized parallelism hypothesis, and FG

323

Classical and Modern Arabic. Berlin and New York: Mouton de Gruyter. Dik, Simon C. 1978 Functional Grammar. Amsterdam: North-Holland. 1989 The Theory of Functional Grammar. Part 1: The Structure of the Clause. Dordrecht: Foris. 1997a The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. 1997b The Theory of Functional Grammar. Part 2: Complex and Derived Structures. Berlin and New York: Mouton de Gruyter. Groot, Casper de 1990 Morphology and the Typology of Expression Rules. In: Mike Hannay and Elseline Vester (eds), Working with Functional Grammar: Descriptive and Computational Applications, 187–201. Dordrecht: Foris. Gulla, J. Atle 1997 Combining Functional Grammar and Rhetorical Structure Theory for Discourse Representation. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 75–89. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. Amsterdam and Philadelphia: Benjamins. Hengeveld, Kees 1996 The internal structure of adverbial clauses. In: Betty Devriendt, Louis Goossens and Johan van der Auwera (eds), Complex structures: A functionalist perspective, 119–147. Berlin and New York: Mouton de Gruyter. 1997 Cohesion in Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 11– 16. This vol. The architecture of a Functional Discourse Grammar. Kroon, Caroline 1997 Discourse markers, discourse structure and Functional Grammar. In: John H. Connolly, Roel M. Vismans, Christopher S. Butler and Richard A. Gatward (eds), 17–32. Liedtke, Frantz 1998 Illocution and grammar: A double level approach. In: Mike Hannay and Machtelt Bolkestein (eds), 107–128. Mackenzie, J. Lachlan and Keizer, M. Evelien 1991 On assigning pragmatic functions in English. Pragmatics 1: 169– 215.

324

Ahmed Moutaouakil

Mackenzie, J. Lachlan 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–296. Moutaouakil, Ahmed 1987 Term-predication in Arabic: A functional approach. Paper. University Mohamed V, Rabat. 1988 Essais en Grammaire Fonctionnelle. Rabat: SMER. 1989 Pragmatic Functions in a Functional Grammar of Arabic. Dordrecht: Foris. 1993 Reflections on the layered underlying representation in Functional Grammar. Paper, University Mohamed V, Rabat. 1996 On the layering of the underlying structure in FG. In: B. Devriendt, L. Goossens and J. van der Auwera (eds), 201–227. Berlin and New York: Mouton de Gruyter. 1998 Benveniste’s récit and discours as Discourse operators in Functional Grammar. In: Mike Hannay and A. Machtelt Bolkestein (eds), Functional Grammar and Verbal Interaction, 25–42. Amsterdam and Philadelphia: Benjamins. 1999 Exclamation in Functional Grammar: Sentence type, illocution or modality? Working Papers in Functional Grammar 69. Amsterdam: University of Amsterdam. forthc. Préliminaires à une théorie fonctionnelle de discours. Premier colloque sur la Grammaire Fonctionnelle. Mohammedia. Rijkhoff, Jan 2002 The Noun Phrase. Oxford: Oxford University Press. Vet, Co 1998 The multilayered structure of utterance: About illocution, modality and discourse moves. In: Mike Hannay and A. Machtelt Bolkestein (eds), 1–24.

Towards a speaker model of Functional Grammar Dik Bakker and Anna Siewierska

1. Introduction According to the standard model of an FG grammar as presented in Dik (1997: 58ff.) a linguistic utterance is constructed in three stages.1 First of all the relevant lexical elements and terms are selected from the Fund, if necessary after predicate and/or term formation (Stage I). Then from these basic elements, an underlying representation is constructed in an inside-out fashion, starting with the layer of the nuclear predicate and ending with the pragmatic (clausal) layer. In the process, the corresponding operators and semantic, syntactic and pragmatic functions are assigned to each layer (Stage II). And finally, once this pragmatico-semantic construct has been completed, this is converted into a prephonetic string by the expression rules (Stage III). The Functional Discourse Grammar model (FDG) as presented in Hengeveld (this volume) proposes an alternative for the organization of Stages I and II above, and indeed an extension of the latter. An important distinction between FDG and the standard model is the separation of pragmatic and semantic information into two separate levels. The pragmatic top layer of the underlying representation in the standard model has been separated from the now purely semantic Representational Level (RL) and added as the inner layer to an organizationally higher and purely pragmatic Interpersonal Level (IL). In addition, the interpersonal level has two higher, discourse-related layers, viz. the Act and the Move. The information contained in the IL and RL is constructed in a top-down, outside-in fashion rather than the inside-out direction (upward layering) followed throughout in the standard model. The result is that pragmatics goes before

326

Dik Bakker and Anna Siewierska

semantics, and semantics before expression, which is an improvement from the perspective of the principle of functional explanation (Dik 1986). Furthermore, the scope between the layers is respected in the sense that choices are determined on the basis of scope relations.2 The relevant part of the information on the interpersonal level, mainly reference and ascription, is mapped onto the representational level, and is then expressed from there. Purely pragmatic information is expressed directly from IL, without intermediate semantic mapping. Although Hengeveld’s model does not discuss the role of the lexicon in any detail, we assume that both the interpersonal and the representational levels have access to it via the Cognition module. As opposed to the pragmatic and semantic levels, the FDG model does not contain any new proposals for the organization of the expression rules (ER) in a direct sense; they simply map IL and RL entities onto the Expression Level (EL), i.e. the morphophonological string that will eventually trigger articulation. However, several aspects of the model suggest that the overall structure of the standard expression component should be reconsidered in order to fit the rest of the model. Firstly, the expression rules now take their input from two levels rather than just from the traditional fully specified underlying representation, as is the case in the standard model. This considerably complicates the interaction between the respective components of the grammar. Secondly, the expression level communicates with both the Cognition module and the Communicative Context. The Cognition module contains the linguistic competence, i.e. the lexical and grammatical knowledge of the language user, which is also represented in the standard model, although there the grammar is mainly presupposed. However, this module also represents communicative competence and all other information needed during the formulation and encoding of the speaker’s intentions. The interaction with the Communicative Context, not represented in the standard model at all, should be considered vis-à-vis expression in more detail as well. Hengeveld’s model suggests that the Communicative Context module is one outlet for expressions. Thus, after having been uttered, expressions also go there in their raw form, i.e. in the shape of the output of the expression rules. This is to take care of metalinguistic operations, where anaphoric reference is made to forms rather than meanings. Finally, the model seems to assume a hierarchical organization for the (pre-phonetic) output of the expression rules, going from paragraph to sentence to clause to the different types of constituents and further down. The traditional expression rules and their output lack any form of constituency at all. In this chapter, we wish to consider in some detail the implications of the grammar model proposed by Hengeveld, henceforth FDG, for the ex-

Towards a speaker model of FG

327

pression rules. We will take as our point of departure the proposals relating to the expression component made in Bakker (2001) rather than the standard expression rules as presented in Dik (1997). The major aspects of this dynamic model of expression will be presented in Section 2. Section 3 will briefly discuss linguistic modelling in general. Furthermore, we give a critical discussion of some aspects of the FDG model in this light. Suggestions are given to clarify some of our points. In Section 4 we will look at the implications of FDG for our revised expression rules. We will integrate the two models, and simulate the production of a short stretch of spoken discourse on the basis of this integral dynamic model. Finally, in Section 5 we draw some conclusions for the overall organization of an F(D)G grammar and its further development.

2. Overview of the dynamic model of expression In their standard form, the FG expression rules come in three phases. First, starting out from a fully specified underlying representation, which contains the predicates and the relevant functions and operators, the grammatical elements are generated. This gives us both the independent grammatical morphemes such as determiners, auxiliaries and prepositions, and the respective kinds of bound morphemes, such as tense markers and agreement suffixes. In the second phase, all lexical and morphological material is ordered into a linear string. In the third phase this string is given its final phonological shape. In Bakker (1994; 1999) it is shown that there are two major shortcomings to the way the expression rules are traditionally organized. Firstly the expression component as it stands undergenerates, i.e. it cannot produce certain forms that actually occur in languages. This endangers the descriptive adequacy of an FG grammar in a rather obvious sense. A clear, and ubiquitous example of undergeneration is provided by constructions in which form and order interact. Examples (1) from Breton (Borsley and Stephens 1989) and (2) from Dutch are a case in point. (1)

a. Ar vugale ne lenn-ont ket levrioù the children PCL read-3PL not books b. Ne lenn ket ar vugale levrioù PCL read not the children books ‘The children do not read books.’

328

Dik Bakker and Anna Siewierska

(2)

a. Jij speel-t goed vandaag! you play-2SG well today ‘You are playing well today!’ b. Vandaag speel je goed! today play-ø you well ‘Today you are playing well!’ c. Ik zie dat jij vandaag goed speel-t! I see that you today well play-2SG ‘I see that you are playing well today!’

In the (a) examples, with SV order, we find a 2nd person singular subject agreement marker on the verb. In the (b) examples, with VS order, there is no such marker; the verb appears as a bare stem. In the complement clause of (2c), with its more or less fixed SOV order, the agreement marker is always expressed. If the expression rules first generate the grammatical forms and only then establish linear order, as in the standard model, there is no way to determine whether the suffix should be generated or not. The only solution would be to look at the factors that determine the eventual constituent order, in this case pragmatic aspects of the subject constituent, the availability of another potential P1 filler, and the level of embedding. But this would imply ‘prerunning’ the ordering rules in some way or other.3 A second problem that we encountered with the expression rules in their current form is that they overgenerate, i.e. they will produce all kinds of forms that actually do not occur in any existing language. At first sight, overgeneration seems to be a less serious problem than undergeneration: we may simply assume that the theory will prevent such ‘impossible’ expressions from arising at all because they express ‘impossible’ underlying representations. However, under the assumption that there is a niche in the grammar which allows for a certain amount of autonomy within morphosyntax there might indeed exist certain (universal or typological) constraints on what can be a well-formed expression in languages which are not directly reducible to functional criteria (cf. Croft 1995). A linguistic theory should make these constraints explicit, much as it should be descriptively adequate in the more obvious sense of generating the forms that do occur.4 But overgeneration is not only a theoretical issue. An overgenerating expression component puts into question the learnability of a FG grammar, and as such endangers the cognitive adequacy of the theory. In Bakker (2001) a model is proposed for the expression component that tries to repair these shortcomings. In correspondence with the standard

Towards a speaker model of FG

329

model, the point of departure is a fully specified underlying representation (UR), with all the predicates and operators present. The morphosyntactic templates and the corresponding placement rules are also retained, albeit in another guise. A major contrast with the standard approach is that our model is dynamic rather than static in the sense that the top-down, left-toright order in which elements of the corresponding URs are expressed plays a decisive role in determining the final shape of the expression. It is shown that at least some of the problems that the original model runs into can be solved by this dynamicity. It provides both a set of constraints on what may and may not be found in expressions and a way of getting the right information at the right place during expression. An almost logical implication of this dynamicity is the integration of phase I (form generation) and phase II (linearization) into one process, interleaving form and order operations. Another aspect that distinguishes the dynamic model from the standard approach is that the templates necessary for the expression of underlying representations are combined into a tree-like structure as in Figure 1 below. Although such a hierarchical organization of templates has always been possible in theory – and may even have been implied; see also Hengeveld (this volume: Figure 6) – it has never to our knowledge been explicitly introduced as such, nor has any other form of constituent structure been posited within FG. Although Figure 1 may be seen as the traditional (static) representation of the constituent structure underlying the corresponding expression, in our case it attempts to code the history of the process leading to that expression. Thus, at any level in the tree, the corresponding expression pattern (‘template’) is selected on the basis of the actual information contained in some specific part of the UR. The respective elements of that part of the UR are then distributed over the slots of the pattern on the basis of slot specific selection rules (‘placement rules’). For each (filled) slot in the pattern this process is repeated recursively until no further distribution is possible. This recursive process of slot filling and template selection is represented statically in Figure 1 for the (simplified) underlying representation in (3). (3)

[Decl E1: [X1: [Pres e1: [Progr [expand [V] (Def Pl x1: tree [N]: (Prox Sg x2: theory [N])POSS)ZERO,SUBJECT ] ] ] ] ] ‘The trees of this theory are expanding.’

In Figure 1 below, templates are found as a row of categories at the respective levels of expansion. P1 VFIN VINFIN on the second row, for

330

Dik Bakker and Anna Siewierska

example, is arguably an appropriate template for the clause level.5 The appropriate template for the actual filler of the P1 position, the subject term in this case, turns out to be DET NOUN NOMRESTR. MAIN CLAUSE

P1 TERM(SUBJECT)

VFIN [IMPRF,PRES,PL,3]

VINFIN VERB [IMPRF, GERUND]

/are/ DET [DEF]

NOUN [PL]

/the/

/trees/

/growing/

NOMRESTR

TERM

ADPOS [POSS]

[PROX,SG]

/of/

/this/

NOUN

/theory/

Figure 1. Expression tree with nested templates

This process of alternating template selection and slot filling provides us with a tree in which functional categories (the ones in bold print) and formal categories alternate. The process bottoms out in a row of nodes with lexical and grammatical categories, such as NOUN and ADPOS, or a cluster of grammatical elements, such as [PROX, SG]. By organizing the distribution of information in this way, we attribute a dynamic character to the process of tree construction, in which the sequence of slot filling, and therefore time, plays a crucial role. This makes it possible to postpone the determination of the final form of both predicates and grammatical elements until the moment at which they are eventually expressed. Elements that have influence on the form of elements that will be expressed before them, e.g. because they assign case to them, have to be available in time only in as far as their assigning (functional) aspects are concerned, not nec-

Towards a speaker model of FG

331

essarily in their final form. Although it is not clear to us at what stage of the expression process phonological details of the respective lexical and grammatical entities are available, we will nevertheless give our terminal elements a (pseudo)phonological shape indicated by forward slashes, but will furthermore remain agnostic about the phonology.6 In underlying clauses we will follow the practice of assuming that a more abstract representation of predicates is present, indicated by the uses of single quotation marks. We will organize this process of dynamic tree construction by applying the following principles. We would like to stress that the model almost completely ignores the phonological level. Far from assuming that the phonological module may simply be stuck on as a separate component to the model presented here, we nevertheless think that the phonology might be interleaved with the morphosyntax in the same fashion in which we interleave syntax and morphology below.7 Principle 1: Constituent structures are developed top-down from the material in the underlying representation. Thus, first the higher syntactic constituents, such as noun phrases, are developed, then the individual free morphemes, next the bound morphemes, and finally their phonological form. In this way, and in contrast to the standard model, the explanatory hierarchy of FG as discussed in Dik (1986) is respected, which runs from syntax to morphology to phonology. Principle 2: Development takes place from left to right. This is the ‘natural’ order in which linguistic forms are uttered in the first place. Leftto-right ordering may be expected to shape and have shaped language over time, at least to some extent, and more than any other order. Principle 3: Development works depth-first. This means that of any two contiguous elements to be expressed, at whatever syntactic or morphological level, the leftmost one will be completely expanded up to its terminal forms before the rightmost one is considered, and this applies recursively. An implication of this is that only a fraction of the complete information which is necessary for the production of the whole utterance will be available at any one time. This considerably reduces the problem for short-term memory which is inherent to breadth-first development. In combination with Principle 2, this may solve a number of problems of the type illustrated in example (1) above. Principle 4: For any node N in the tree, all features found on a direct path from that node to the top node are available for inheritance by N, at least in principle. Trivially, this implies that overt primary operators are available for every node in the tree. However, a distinction will be made

332

Dik Bakker and Anna Siewierska

between ‘raw’ UR material (functions, π and ω operators, and inherent features of predicates such as Animacy and Gender) on the one hand, and µ operators on the other hand. In order to be accessible for inheritance, it will be assumed that any UR feature has to be transformed into a µ operator, either in a one-to-one or a more-to-one fashion, as in the case of portmanteau operators. An implication of this is that only those features of URs qualify as such if they are turned into a µ operator at some stage during expression. Apart from this there may be universal or language-(type)specific downward barriers that make features inaccessible to lower nodes. Barriers to inheritance may be either functional or formal in nature, i.e. either determined by a layer or substructure in the UR or determined by some syntactic or morphological boundary, i.e. a specific node in the tree. Principle 5: Operators may percolate, i.e. move upwards to higher nodes. For percolation there may be universal and language-(type)-specific upward barriers that make them inaccessible to higher nodes in the tree. Barriers to percolation are only formal in nature. These five principles should contribute to the (cognitive and typological) adequacy of the dynamic model. They have two interesting implications for the global working of the model. On the one hand, linguistic forms are produced precisely in the order in which they are uttered by the speaker in a live setting, thus giving the model a real-time procedural flavour. On the other hand, several constraints follow from the principles, which restrict the formal power of the expression component. Indeed, it may now be determined what information has to be available at what stage in the tree expansion process. In this way, it may be determined more precisely what information necessarily has to come in from the UR as opposed to arising in the course of expression, and thus, which elements have predicational status in the language under consideration and which ones do not. Furthermore, these constraints provide a means of determining the order in which grammatical material has to be available in the first place. For example, a case-marking postposition is needed on some node in the tree before the elements on which the corresponding case is expressed are created. Since grammatical forms may come into existence at several stages in the expansion process, we also get an instrument to distinguish between partially and fully grammaticalized elements. The former will appear relatively early (‘high’) and the latter relatively late (‘low’) in the constituent structure. Arguably, this distribution may reflect the length of their diachronic path through the language, typically leading from a full predicate to a fully grammatical entity, and their relative role in the syntax and morphology.

Towards a speaker model of FG

333

The constraints implied by principles 1–5 above determine the information flow through the expression tree that is dynamically constructed by them. It is crucial what functional and formal information may be available at which node in the tree during tree expansion. In order to accommodate this information, we think that nodes should have the following internal structure, depicted in Figure 2.

LABEL Functional aspects CONFIGURATION

FUNCTIONAL FEATURES

Formal aspects FORMAL FEATURES

SUBCATEGORIZATION

Figure 2. Structure of a node

A node has the following fields associated with it. The label (Lab) represents the category of a slot in a template, such as P1, Subject or Vinfin. The configuration (Config) is a specification of some part of the underlying representation that will be expressed by this node. It cannot be just any part of the UR; it should be part of the configuration of the mother node of the one under consideration, for example the specification of the subject term, a temporal satellite, the head of a term, or the main verbal predicate. Typically, such descriptions will contain one or more predicates;

334

Dik Bakker and Anna Siewierska

in those cases we will call them functional configurations. A configuration may also be the description of a grammatical element, such as an auxiliary, a demonstrative, or a tense suffix. In that case we will call them formal configurations. The functional features (FncFtrs) are the primary µ operators relevant for the slot under consideration. They may be derived from operators, functions and lexical elements of the Config, or they are directly inherited from the FncFtrs field of the mother node. Examples of functional features are Number, Tense and Animacy. Formal features (FrmFtrs) are the auxiliary µ operators relevant for this slot. They may be derived from the Config and the FncFtrs, be inherited from the mother node, or percolate upwards from a daughter node. Examples of formal features are Case and Finiteness. Finally, the Subcategorization (SubCat) is a template consisting of nodes for the functional categories into which Config is to be split up. The appropriate template is selected from the set of templates of the language on the basis of the concrete value of the functional part of the node, via more or less general rules. A template may also be inserted on the basis of lexical information, typically from the head of the structure in Config, for cases where the default node specifications would lead to the selection of the wrong syntactic environment. This is an instance of lexical priority. All these aspects together represent what in the standard model of the expression rules are phase I and phase II. The templates and placement rules of phase II are replaced by the subcategorizations and the functional configurations of our model, respectively. Furthermore, formal configurations and both feature sets replace the formation rules of phase I. To round off this short sketch of the dynamic model of the expression component, we give the complete derivation of a very simple sentence. It is the expression of the underlying representation in (4). (4)

[decl E1:[X1:[pres e1:[smart [A] (prox pl x1:girl [N])ZERO,SUBJECT ]]]] ‘Those girls are smart.’

The first step in the expression process is the selection of a node that will express the complete UR of (4). In order to cater for (4), this node needs to have a more or less maximum representation of an UR for its Config field, with variables for all relevant functions, operators and predicates and all layers present. After its selection, the Config field will be unified with the UR of (4), and the respective variables will get the corresponding values. Example (5) below gives a simplified version of this top node before unification with the UR and example (6) after unification. Elements in upper case such ‘TENSE’ indicate free variables; lower case elements such as

Towards a speaker model of FG

335

‘pres’ represent the values for these variables after they are bound with the corresponding values from the UR.8 (5)

(6)

NODE 1 (uninstantiated) Lab: sentence Config: [ILLOC [[TENSE [ PRED CAT ARG1 ]]]] FncFtrs: ILLOC, TENSE, CAT, NUMBER.SUBJECT, PERSON.SUBJECT NODE 1 (partially instantiated) Lab: sentence Config: [decl E1:[X1:[pres e1:[smart [A] (prox pl x1: girl [N])ZERO,SUBJECT ]]]] FncFtrs: decl, pres, [A], pl, [-sp,-hr]

The last step for this node is the selection of the right template for its SubCat field. The information in the Config field is instrumental in this. The result is given in (7). (7)

NODE 1 (fully instantiated) Lab: sentence Config: [decl E1:[X1:[pres e1:[ smart [A] (prox pl x1: girl [N])ZERO,SUBJECT ]]]] FncFtrs: decl, pres, [A], pl, [-sp,-hr] SubCat: p1, subject, vfin, mainpred

Node 1 ends up being fully instantiated because there are no variables left to be bound. If this were to be the case, then the only way left for these variables to be bound by a value would be via percolation from lower nodes. In a fully expanded tree there may be no unbound variables left. The next step in the expression of the UR in (3) is the development of the p1 position in its subcategory. Typically, there will be several alternative versions for this node in the grammar, e.g. with a Focus, Topic or Subject term for its configuration, but also with adverbials. In the concrete case of (3), the UR element chosen will be the Subject. Filling the p1 slot with the Subject constituent from the Config field will trigger a term template for the filler of the SubCat of Node 2. The next step will then be the expansion of the Determiner node of this template, assuming that this is the first slot. Figure 3 below gives the complete flow of the expansion of Node 1, up to the expression of the slot for the main predication, mainpred. Note that here, the plural morpheme s is introduced by node 6, i.e. via a rule. Our model is neutral towards the type of morphology, i.e. it could just

336

Dik Bakker and Anna Siewierska

as well handle a word-based approach which has the inflected forms in the lexicon. This should serve for a brief overview of the dynamic expression component. In Section 4 we will see to what extent it has to be adapted to fit the multi-level FDG model. 1: SENTENCE 1

10 11

2:P1

2

3

14

15

7:VFIN

4

9

12

4:NOMHEA

3:DET 5

6

7

13

8:VERBFORM

8

/girl/

9:MAINPRED

16

19

10:ADJHEAD 17

/those/

5:NOUN

20

18

/are/

6:NUMSFX

/s/

11:ADJECTIVE

/smart/

Figure 3. Full tree expansion

3. Linguistic modelling and FDG When discussing a model in linguistics, the central – indeed the first – question is what is it that we are modelling. In general, a model seeks to give an operational, explicit and non-ambiguous version of a theory. It is through a model that a theory may be subjected to experimentation and testing. For the present discussion, we will distinguish between two types of linguistic models: the grammar model and the model of the language

Towards a speaker model of FG

337

user. A grammar model represents the respective components a theory distinguishes in the grammar of a language and the relations between these components. Furthermore, it provides formal representations for the respective rules and structures. It abstracts away from any speech situation and perceives the language and its grammar as an abstract, formal construct. The grammar model shows in what way the data structures (the static elements of the grammar, such as the lexical items and morphological and syntactic rules) lead to underlying semantic and syntactic representations and these to actual linguistic expressions. Usually, these models are restricted to the sentence as the highest level of representation and production. There may be a fixed order in the flow of information through the respective grammar components. Often, the ‘later’ components not only expand the structures generated by the ‘earlier’ components but also operate as filters on ill-formed structures. In principle, such models are bidirectional in the sense that they relate the lexicon and underlying representations to expressions and vice versa via ‘left-to-right’ or ‘right-to-left’ applications of the rules of the grammar. However, the traditional interpretation of such models is from underlying representations to expressions; for this reason they are often called generative. The grammar model is usually thought to represent the linguistic competence of the ‘ideal speaker-hearer’. It does not deal with typical performance aspects such as speech errors and hesitations. This, indeed rather abstract model is the most common one found in the linguistic literature. The model of FG presented in Dik (1997: 58) is an example of this type. Sells (1985) presents such models for Government-Binding Theory (1985: 24), Generalized Phrase Structure Grammar (1985: 79) and Lexical Functional Grammar (1985: 137), respectively. Although, as a model of the theory, such models should have descriptive adequacy, they are not required to be psycholinguistically adequate. In Dik (1997: 1) the ultimate goal of a linguistic theory is formulated as the construction of a model of the language user, a model that can approximate the communicative performance of human beings when using their linguistic competence. It might therefore be assessed with regard to its psychological adequacy. Such a model should not only contain linguistic knowledge. It should also deal with all kinds of other knowledge, with beliefs, prejudices, etc. Furthermore, it crucially takes into consideration the various types of memory: working memory, semantic memory and episodic memory. While a grammar model is neutral, a model of the language user should distinguish between speaking and hearing, given that these two modes of linguistic behaviour differ in many respects. A model of the

338

Dik Bakker and Anna Siewierska

speaker should give a precise representation of the process of language production, running from the initial, prelinguistic intention to the eventual utterance, while specifying all relevant intermediate steps in the right order. This information flow is mainly top-down, as in the generative grammar model. However, psycholinguistic research strongly suggests that the production of an utterance is not a purely linear process but an incremental one, managed by several more or less independent modules which work in parallel, at least partially, rather than consecutively. It is above all speech error data which are suggestive of this parallelism, as well as the occurrence of hesitations, pauses, repetitions, etc. A model of the speaker should cater for the extremely complex data flow during such a process as well as for the ‘performance’ phenomena related to it. Finally, this kind of model should include a representation of the speaker’s knowledge about the conversational partner(s). Hearers mainly use the same knowledge resources as speakers do. However, the procedural aspects of a model of the hearer are rather different from those of speaking. The hearing mode interleaves bottom-up processes with top-down strategies, such as the construction of cohorts of phonologically similar strings, or the priming of semantically related entities during recognition. Another example of this is the choice of the most promising syntactic structure at an early stage of speech analysis via probabilistic devices. In the (rare) case of erroneous intermediate choices, backtracking mechanisms are employed. One of the differences between speaker and hearer that the respective models should make clear is the often observed gap between a language user’s passive knowledge and active use of a language. Eventually, both models should make clear how first and second language acquisition take place.9 Hengeveld contrasts the FDG model sharply with the grammar model in Dik (1997). The latter is seen – mistakenly, we think – as a speech production model which lacks psychological adequacy, and therefore runs counter to the adequacy requirements of the theory (cf. Dik 1997: 13f.). Dik (1989) makes it clear, however that an FG model of the speaker necessarily has other properties than the standard model of the theory and then sketches such a Model of the Natural Language User.10 Hengeveld (this volume) observes that FDG is an improvement on the FG grammar model precisely with respect to psychological adequacy. This, and the fact that the model could be described in terms of the decisions a speaker makes during language production, is highly indicative of the fact that FDG should be seen as a model of the speaker rather than of the grammar. However, when we take a closer look, it seems to have aspects of both models. We will briefly mention these, and then assume for the rest of the discussion that a speaker

Towards a speaker model of FG

339

model is the ultimate goal of the enterprise. We will not assume that this should be a full psychological model of the speaker; this might better be constructed from the perspective of psychological modelling, and by a psychologist. The FG version would be a linguistic model of the speaker which is as much as possible in harmony with psycholinguistics. Just like other generative models of grammar, FDG (Hengeveld this volume) shows how complete (‘fully specified’) underlying representations are first constructed and then fed to the expression rules which generate the corresponding forms. Layered formal representations are introduced for the respective intermediate structures, which are an extension of the traditional underlying clause formalism first introduced into FG in Hengeveld (1989). A novelty vis-à-vis the standard model is the introduction of hierarchical tree-like structures suggested for the output of the expression rules. Most of the structures at the respective levels are static and ready made, i.e. nothing is said about how they come into existence. No explicit claims are made as to their psychological status, i.e. whether they should be seen as mental representations or just as formalisms. The latter possibility therefore seems to be the more likely.11 Only for the representation at the pragmatic level is an outside-in history claimed: Move goes before Act and this goes before Communicated Content. Interestingly, though there are modules for Cognition and Communicative Context in the model, there is a separate box labelled ‘Grammar’ around the pragmatic (IL), semantic (RL) and expression (EL) levels. This suggests that, apart from the lexicon, which is contained in Cognition, the grammar gets a separate status outside the latter, with the inclusion of the projection rules and the expression rules which link IL, RL and EL. Linguistic competence, however, is located within Cognition. So it may be safe to assume that the three levels are just levels of representation, while the rules that create them and link them are stored in Cognition. Probably the label ‘Grammar’ should then be interpreted as ‘created by the grammar’. On the other hand, FDG has several properties that are suggestive of a model of the language user, more specifically of the speaker. Firstly, in contrast to the standard grammar model, which takes the lexicon as a point of departure for the generation of sentences via the respective sets of rules, FDG introduces Cognition. This is the complete knowledge store of the speaker, of which the lexicon and the rest of the grammar are just substructures. Indeed, Cognition controls the generation process at all stages. It introduces long-term memory to the model, or rather semantic memory.12 Short-term memory (or rather working memory) is introduced via the Communicative Context module. This stores both the formal representa-

340

Dik Bakker and Anna Siewierska

tions of the Moves and (certain aspects of?) their expression. Interestingly, in FDG the Content of the respective Acts is not stored in the Communicative Context, but only influenced by it. Furthermore, there is a link from this short-term device to the long-term Cognitive module. This feedback loop accounts for the speaker’s monitoring function. A second point that gives FDG the flavour of a speaker model is that it takes the Move as its highest entity of processing, which brings discourse notions into the arena. The feedback loop via working memory adds to this discourse orientation by making aspects of previous utterances available to current or following ones. This caters for problems such as long-distance anaphoric and metalinguistic reference. Thirdly, in FDG a distinction is made between the pragmatic and semantic modules. The output of the former partially feeds into the latter, but both modules have their own, in principle parallel contribution to make to the final output via independent access to the expression rules. This parallelism introduces the possibility of testing the model with respect to its performance characteristics. Assuming, then, that FDG should be perceived as a model of the speaker rather than of the grammar, we may make some further observations which in our view should be taken into consideration for future developments of FDG. Firstly, there is the role of the Communicative Context. In Hengeveld (this volume: 3) this is supposed to be “the (short-term) linguistic information derivable from the preceding discourse and the non-linguistic, perceptual information derivable from the speech situation”. The arrows in Hengeveld's Figure 6 make it clear that the information currently at the speaker’s IL and EL levels is included in this information. If we are right in assuming that ‘short-term’ in the quote above should be read as ‘contained in short-term memory’, then the point is that this information can only be very restricted, i.e. correspond to a few seconds of the verbal interaction at most.13 It is not likely that this will always be the complete intended Move at hand, including its Central Act and possible subordinate Acts as suggested in Figure 6. The simple reason is that full recursion must be assumed for Acts since they will get semantic content via referential and ascriptive acts which in their turn are expressed by recursive linguistic entities such as NPs, PPs and Ss. Since there seems to be no upper limit to the length and complexity of a Move, the information contained at the IL and EL levels, which are thought to feed into the Communicative Context, will regularly exceed the maximum capacity of short-term memory. We think that it may in fact be more realistic to assume that what is contained in the IL, RL and EL levels is actually in working memory, at

Towards a speaker model of FG

341

least partially, with the inclusion of the rules that operate on them. The complete set of materials necessary for the construction of these entities is stored more permanently in semantic memory. Alongside the grammar and the lexicon, this is both the ‘full story’ that is being told and the scripts, scenarios or mental models that are relevant for it (cf. Schank and Abelson 1977; Johnson-Laird 1983). A central planning mechanism sends coherent bits of this material to working memory for linguistic expression, where it is operated upon in three stages, leading to the representations at the IL, RL and EL levels. In its EL form, it is finally send to the articulator, which runs it as a set of instructions for the speech apparatus. The feedback loop from Communicative Context to Cognition would then mean that whatever information is held in working memory at any one time is fed back to semantic memory for more permanent storage, or at least certain aspects of it. In this respect it is interesting that there is no feeding arrow from the semantics of the RL level to the communicative context. Rather, RL gets part of its input from the Communicative Context, probably anaphoric information above all. Since Communicative Context is related to short-term memory only, this must be very local information. This would imply that no (direct) information about the precise semantics of an utterance finds its way back to semantic (sic!) memory, but only information about the pragmatics and the form. This is not a very probable state of affairs. The extensive literature on recall experiments shows that hearers and speakers retain the content of what they hear, read or say rather than the form. See Brandsford and Franks (1971), Schweller et al. (1976) and Johnson-Laird et al. (1974) for classical experiments that refute the recollection of syntactic structure, precise words and parts of speech, respectively. But even though it is the meaning of an utterance rather than anything else that is remembered, it is not necessarily the concrete linguistic meaning in terms of the meaning representation as given by an Underlying Clause, for instance. Rather, it is a broad interpretation of the meaning, the gist or intention of what was said. Therefore, speakers will not normally remember whether they have said (8a) or (8b). (8)

a. The window was open. b. The window was not closed.

Taking a different perspective on things, one might argue that the contents of what is being said are stored in semantic memory even before speaking starts, and that there is no need for them to be fed back. This may be true to some extent for a coherent and structured story or event in the

342

Dik Bakker and Anna Siewierska

past that has been remembered and that is being told. However, in spontaneous conversation completely ‘new’ things may be said that speakers do remember later on. Also, telling ‘old’ stories may have a restructuring effect on the way they are remembered and told on a later occasion. Thus, it seems to be necessary to have feedback from the semantics of an utterance to semantic memory, too. In the light of the above we would suggest that the FDG model be refined in such a way that what resides in working memory and what in permanent storage is determined more precisely. A possible scenario is the following. Assuming that all basic material – both building blocks and rules – are located in semantic memory, an utterance is then constructed on three more or less independent levels: IL, RL and EL. For all three levels construction starts out from semantic memory, which sends (parts of) the contributions of pragmatics, semantics and morphosyntax to working memory at certain intervals. These intervals and the amount sent are related to the capacity of working memory and to what is already present in it. This process follows the hierarchy pragmatics > semantics > form. Accordingly, some bits of the pragmatics are first sent to working memory, say some aspects of the Move and the Act. When this has arrived, semantics starts filling in the relevant parts, while pragmatics moves further on in the process, e.g. to the first subordinate Act. When semantics is finished with the first chunk and is ready to start working on the second chunk it gets from pragmatics, the expression rules kick in and start working on the first chunk of RL material. Apart from the fact that a great deal of the material to be uttered by the speaker is in semantic memory in some form or other before any utterance is made at all, we assume that it is IL material and above all RL material that is fed back to semantic memory. There it is used by an autonomous process that constructs a version of the total discourse, to the extent that what is being said is a complete paraphrase or a more or less new version of what was already residing there in some form or other anyway. Only very little of the EL material is fed back for permanent storage. Only this, and what is left in working memory, is available for monitoring processes. As to the shape and size of the contents of working memory at any one time, we should probably best think of entities such as Chafe’s intonation unit (Chafe 1987; and see Crystal 1975 for an earlier version of it) or Halliday’s information unit (Halliday 1994).14 Indeed, in studies of spoken language it has been frequently observed that (spontaneous) speech generally proceeds via relatively short contributions, which may contain one or more given elements and rarely introduce more than one element that is

Towards a speaker model of FG

343

new to the conversation. Such units typically take 1 to 2 seconds at most, and are embedded between intermediate pauses (cf. Givón 1975: 202f.). The most common unit is of the Topic-Comment type, prototypically expressed in Subject-Predicate form. Quite often however, utterance units do not consist of the complete intended Move or Act, and may also be ‘incomplete’ from the perspective of GrammarN.15 This incompleteness may be based on processing constraints. It may also be caused by the fact that part of the topical information is thought by the speaker to be redundant. In the latter case, this ties in neatly with the discussion in Mackenzie (1998), which is a plea for adapting FG such that it also produces holophrases like those in (9) below. (9)

Tomorrow. To Mila. Liam too. Five of these, please. The green one, if possible. Skating.

Indeed, many utterances that appear to be ‘incomplete’ from the perspective of a sentence grammar turn out to be complete from the perspective of a discourse grammar. Speakers, it is assumed, only produce ‘complete’ sentences in order to create more cohesion in their discourse when the speech situation calls for this. In such cases they opt for what Hannay (1991) calls the Topic Mode. However, in informal discourse speakers will try to limit themselves as often as possible to new, focal information, thus producing chunks that need only the P1 (clause-initial) position of the traditional FG template for their expression. The latter is above all the case in the Reaction Mode, where speakers are frequently taking turns in the conversation. Such chunks may fit into working memory in their entirety, which makes for ease of processing in the highly interactive and competitive set-up of a face-to-face conversation. On the other hand, it is precisely the size of working memory which puts constraints on what can be maximally dealt with during speech production in the first place. So, in the case of weightier Moves and Acts, spoken language typically comes in Sub-Acts and Sub-Clauses, or even parts of these, i.e. mere fragments of GrammarN complete clauses. Quite frequently, in spontaneous speech consecutive Sub-Clauses do not even combine to form GrammarN-complete and well-formed ones. We think this is also a strong indication of ‘incompleteness’ at the higher levels of linguistic planning, of the role of feedback

344

Dik Bakker and Anna Siewierska

mechanisms, and of the parallel, incremental nature of the process of speech production in general. Example (10), from Chafe (1987: 23), is a case in point. In (10), a hyphen indicates lengthening of the preceding segment; two dots indicate a brief temporal break; three dots indicate a pause; the acute accent shows an intonation peak. Each line would be an intonation unit in Chafe's terms. (10)

a. b. c. d. e.

… A-nd .. he would come into cláss … a-t .. uh- you know three or f … precísely one minute after the hóur, or something like thát … a-nd he- .. wou-ld .. immédiately open his … nótes up

The proposal in Mackenzie (2000) for an incremental model of FG (IFG) should be seen in the light of the above remarks. A last point we would like to make here is related to the nature of the representations of the information given at the respective levels in Hengeveld (this volume). For the RL level the usual Underlying Clause (UC) structure is assumed, cut off at the propositional layer. For EL we propose the structures introduced in Section 2; the link with the other levels of representation will be worked out further in the next section. However, some aspects of the IL representation remain unclear to us. The outer layers – Move, Act and Illocution – are formally of the Underlying Clause (UC) type. Their contents are variables, operators, and references to speech-act participants, i.e. they are formal in nature. The inner layer, under variable C, has referring acts (with the variable R) and ascription acts (with the variable T) as its contents. These acts get their semantic content only after projection on the RL level, in the form of predicates of the language, plus the corresponding variables, operators and functions. The point now is how this predicational content should be determined. This problem relates directly to the way in which mental representations are made in semantic memory in the first place, both of the knowledge to be reported in general and more specifically of the communicative context. If linguistic relativism is assumed, as is suggested by the stepwise lexical decomposition philosophy proposed within FG ever since Dik (1978), then the predicates of the language concerned make a considerable contribution to the ‘stuff which thoughts are made of’. In addition, Dik (1989) proposes to take UCs as the representational formalism for those types of knowledge which are non-pictorial in nature. In other words, FG promotes a linguistically motivated type of knowledge representation. If this were indeed valid,

Towards a speaker model of FG

345

then projection from IL to RL would in fact be trivial: the R and T variables would simply be linking devices between UC-like mental representations and newly made underlying clauses, via discourse variables of the right type. However, such an extreme relativistic position seems to be untenable. There are a number of arguments, both logical and experimental in nature, which point towards a non-linguistic, abstract representation of knowledge. We will mention but a few. If language were to be the code in which (conceptual) knowledge was stored, it would not be clear how knowledge about language itself (or rather, a specific language) could be acquired by the pre-linguistic brain. There must be some built-in mode of operation, necessarily prelinguistic, with which to bootstrap the brain to thinking and knowledge acquisition. There must therefore be some kind of thinking before language, and indeed without language altogether, since the congenitally deaf do think, and not in a dramatically different way from the non-deaf (cf. Furth 1966). Furthermore, there is no evidence for differences in the way conceptual knowledge on the one hand and knowledge acquired through perception on the other are stored and processed in the mind. The same goes for declarative knowledge versus procedural knowledge. For most types of knowledge that we possess, we do not remember how it was acquired, nor can we determine by ourselves how it is internally represented. Furthermore, we seem to reason and to process it all in much the same way. This all points in the direction of a unified form of representation, both for the knowledge that is linguistically expressible, like most declarative knowledge, and knowledge that is not, or only partially so, like procedural and perceptual knowledge. For further arguments we refer to Clark and Clark (1977) and Fodor (1983) and references therein. In what follows, our assumption will be that the planning of utterances is largely pre- and indeed non-linguistic, i.e. it proceeds in terms of abstract, language-independent concepts. This means that the mapping from IL to RL involves translation from such abstract concepts to predicates of the language concerned. The crucial link for this translation process lies in the meaning definitions of the predicates in the lexicon. We will assume that these are constructed on the basis of such abstract concepts, possibly mixed with language-specific elements. Arguably, the ongoing discourse – the complete Communicative Context – as represented in semantic memory is coded in the same, more or less language-independent fashion.16 This concludes our critical discussion of the more general aspects of the FDG model. In the next section we will see how FDG, with the provisions made above, may be linked to our dynamic model for the expression rules.

346

Dik Bakker and Anna Siewierska

4. The FDG model and expression We will now try and fit the FDG model and our expression module together. We will do this on the basis of example (9), repeated here in a slightly simplified form under (11): (11)

a… A-nd .. he would come into cláss b.… a-t .. uh- three or f c. … precísely one minute after the hóur, d…. or something like thát

These four intonation units are part of a longer fragment, which comprises 40 units in all. In its turn, this is part of a tape-recorded conversation during an informal dinner party. The participants are telling each other stories about the university teachers they had. In the 8 units that precede (11a) the speaker has introduced a certain (male) teacher she particularly remembers. We will assume that the information in that introduction is represented in the speaker’s semantic memory in an activated, readily accessible form. In that stretch of text, the teacher has been introduced as a topic; he has been referred to several times pronominally. Furthermore a typical ‘school’ script has probably been activated, from which all kinds of default aspects can be derived, such as the article-less use of ‘class’ in (11a). Finally, the speaker will assume that her audience is in the same ‘frame of mind’. Let us now try to trace the possible flow of information within the FDG model as sketched above, with special attention to the expression of the respective utterances. The whole story of 40 intonational units, which is not interrupted by anything other than assertive noises and laughter, consists of a series of Moves. All of them express the communicative intention of the speaker to inform the audience about this special teacher. We shall interpret stretch (11a) – (11d) as one Move, which we will call M6. It consists of a central Act, A6. M6 is about the fact that the teacher in question always manages to arrive shortly after the moment at which the class should start. We will assume that this specific piece of recollection is represented in the semantic memory of the speaker in an abstract, non-linguistic way, as an inference made over a set of observations over time concerning the behaviour of this teacher. M6 starts with a pause, prolonged by the lengthened discourse continuity marker And and another short pause. During this pause the contents of M6 are planned, and part of its expression form is calculated in working memory. This will give rise to the first intonation unit, which is part of A6.

Towards a speaker model of FG (12)

347

… A-nd .. he would come into cláss. [previously (11a)]

According to FDG, after the mental material for M6 has been gathered, the first step in this process will be the decision that this is going to be an Informing move. Although Hengeveld (this volume) does not give an indication in this direction, we will assume that there will be an M operator INFRM for this.17 Then the Act is constructed. The illocution chosen is DECL(arative), the default choice under the INFRM operator. It is followed by the variables for the discourse participants PS and PA and the actual Content C6. Together this is a paraphrase of the intention ‘I tell you that C6’. C6 will be filled in by a number of referring and ascriptive acts Ri and Tj. The order in which this happens is determined by various competing pragmatic principles. Good candidates are Topic Continuity and Task Urgency (Givón (ed.) 1983; Givón 1987, 1988). Both are scalar variables which determine the formal expression of referring and property-assigning entities in terms of constituent order and other devices, such as special constructions. Topic Continuity favours topical and other ‘old’ and presupposed entities; Task Urgency favours focal and other ‘new’ and unpredictable entities. For our discussion, our point of departure is that these forces work on the interpersonal level, above all in determining the order in which referring and ascribing acts take place. This pragmatically determined ‘underlying’ order is reflected only indirectly in the final syntactic constituent order. It is reflected in the availability at the RL level and the constituent-order rules of the language concerned. More concretely, Topic Continuity, Task Urgency and possibly other forces determine the order in which the respective Ris and Tjs will enter C6. In the case of (12) the unmarked surface form is an indication that three elements have been entered into C6 initially, and probably in the (default) order Topical > Focal: the topical referent for the teacher (R1); the focal referent for the notion CLASS (R2); and the third element for the ascriptive notion GO INTO (T1).18 In the case of R1, instead of an abstract predicate, the existing discourse variable for the teacher, say x1, will represent the referring act. In the case of referents which are new to the discourse, new variables are introduced. Something like the representation under (13) may then be in working memory. Note that here and in later representations we leave out bits of the FG formalisms that seem to be less relevant for our discussion.

348

Dik Bakker and Anna Siewierska

(13)

(M6: [INFRM [(A6: [DECL (P1)S (P2)A (C6: [R1(x1)TOP R2(x8:CLASS)FOC T1(f1: GO INTO)])])]])

Now semantics will kick in and start working on this structure. Semantic representations will be created for the referents of R2 and T1 by searching the lexicon and by adding the operators which are relevant given the grammar and the different types of background information such as the ongoing discourse and the script. We will not go into the details of this undoubtedly very complicated process of translating cognitive structures into semantic ones, precisely because we lack a representation for the former, and a theory about the translation process. The result in working memory will be as in (14).19 (14)

(p1: [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT )])

This will be passed on to the expression component. In parallel, parts of (13) have been handed to the expression component as well. This might be something like the information in (15). (15)

[DECL (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (f1)])]

Here, unlike in the version introduced in Section 1 where we used the traditional multi-layered underlying clause, we are dealing with two structures for expression, (14) and (15). Several options are now open. One is to merge these two structures into one structure, which will give us something like the traditional underlying clause. The variables shared by the two structures should make this possible. However, this would mean reintroducing the upward layering procedure. Also, in order to be combinable, the two structures should be complete in the sense of the traditional underlying clause. It seems to be more interesting to see whether we could also maintain the parallelism in the model during expression by keeping the two structures apart. This can be accomplished by extending the Config field of the top node – and other nodes for that matter – such that it can hold both the pragmatic structure of type (15) and the semantic one of type (14). The top node for the expression will then look like (16).

Towards a speaker model of FG (16)

349

NODE 1 (uninstantiated) Lab: sentence Config: [ILLOC PS PA CONTENT], TENSE [PRED CAT ARG1] FncFtrs: ILLOC, TENSE, ASPECT, CAT, NUMBER.SUBJECT, PERSON.SUBJECT FrmFtrs: VARIABLE.SUBJECT

After unification with the structures in (14) and (15) we get the node in (17). (17)

NODE 1 (partially instantiated) Lab: sentence Config: [decl (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (f1)])], [ [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT ) ] ) FncFtrs: decl, past, hab, [V], NUMBER.SUBJECT, PERSON.SUBJECT FrmFtrs: VARIABLE.SUBJECT

Note that the values for number, person and the variable of the Subject have not been bound. This was not possible because no Subject has yet been assigned. In contrast to the position taken in the standard theory of FG, we think that grammatical functions should not be assigned to underlying representations, but by the expression rules themselves. Like many authors, we see the phenomenon of Subject as the outcome of a grammaticalization process, arguably involving left-dislocated topics. The result of this process is a set of formal relations in morphosyntax, a subset of which is obligatory for those languages which have Subject. Therefore, syntax seems to be the right place for assigning this function to one of the constituents that qualify for it. The choice is typically restricted to arguments of the main predicate, and motivated on the basis of a language-dependent set of aspects of constituents, such as their semantic function and a number of other properties, e.g. animateness or person. However, crosslinguistically, and maybe not surprisingly, it seems to be the case that in the great majority of utterances in spoken discourse, Subject is assigned to the constituent which is also the Topic.20 Subject assignment tends to have effects at the highest level of the sentence: among these are constituent order, agreement and several types of Same-Subject marking or nonexpression in coordinated and subordinated clauses. Therefore, the relevant choices are necessarily made at the highest level of the syntactic structure, i.e. at the top node of the expression. For the English example in (17), in fact only one constituent is available for the assignment of Subject func-

350

Dik Bakker and Anna Siewierska

tion: the Ag argument, which happens also to have Topic function. This leads to the binding of the Subject-related variables in (17), which gives us (18). (18)

NODE 1 (partially instantiated) Lab: sentence Config: [decl (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (f1)])], [(past e1: [hab f1: come [V] (def x1)AG,SUBJECT] (gen sg x8: class [N])ILLAT ) ] ) FncFtrs: decl, past, hab, [V], sg, 3 FrmFtrs: x1

The next step is the selection of the right template, for which the actual contents of Config are scanned. This has been done in (19), where three slots are introduced. Note that there is a position for the subject rather than a general P1 position, since it is ‘known’ that the subject will land here in this case. The other positions are for the verbal cluster and the satellite. (19)

NODE 1 (fully instantiated) Lab: sentence Config: [decl (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (f1)])], [(past e1: [hab f1: come [V] (def x1)AG,SUBJECT] (gen sg x8: class [N])ILLAT ) ] ) FncFtrs: decl, past, hab, [V], sg, 3 FrmFtrs: x1 SubCat: subj, vc, sat

This will now be expressed further via the process sketched in Section 1. This will lead to the expression of (12), on the basis of the partial tree depicted in Figure 4 below (the numbers next to the arrows give the order of creation; we have left out the upward-pointing percolation arrows). Note that the tree below is in fact completely ‘flat’, i.e. there are no ‘unnecessary’ nodes and branches. We think that structure is represented in the first place on the RL and IL levels and is not necessarily reflected iconically and without further arguments in the structure of morphosyntax. Further elaboration of the theory of expression should make clear whether more formal structure and more formal categories are necessary to represent the linguistic facts beyond some (relative or absolute) minimum.21

Towards a speaker model of FG

351

SENTENCE 1

4

SUBJ

9

VC

2

5

PRO 3

/he/

SAT 7

10

AUX 6

PRED 8

PREP 11

/would/

/come/

/into/

12

N 13

/class/

Figure 4. Partial expression tree; stage 1

After (12) – previously (11a) – has been articulated, working memory will contain this tree, or at least such a substructure of it as to support the rest of the syntactic entity under construction. The same will be assumed for the IL and RL structures of (14) and (15). There is a short pause now, in which new pragmatic and semantic material is being processed, corresponding to (20). (20)

… a-t .. uh- you know three or f

This unit is complicated and particularly interesting. It starts with a preposition, which is an indicator of a temporal satellite, as we know on hindsight. This preposition is articulated, but in a hesitant, prolonged way, apparently because the speaker is not sure about the temporal details of the satellite. Therefore, extensions will be made at the IL and RL levels. We will represent these as follows. First, an update will take place on the structure in (13). An extra referential act R3 will be added at the IL level, being the cognitive version of the temporal satellite under creation. This leads to (21); the new part is printed in bold. (21)

(M6: [INFRM [(A6: [DECL (P1)S (P2)A

(C6: [R1(x1)TOP R2(x8:CLASS)FOC T1(f1: GO INTO) R3(x9:TIME)FOC])])]])

352

Dik Bakker and Anna Siewierska

After completion of this, the IL level will be updated with the semantic version of the temporal satellite. This gives us (22); again, the new bit is in bold. Separately, the relevant part of the IL extension will be sent to the expression component. This will give us (23). In both cases we will effect these changes not in an independent structure, but directly in the Config field of the relevant node of the expression, in this case Node 1. So we will have the following updated structures there. (22)

(p1: [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT )

(23)

[DECL (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (x9)FOC (f1)])]

(def x9)TEMP )])

Expression continues from Node 1 with the new element that was added to the RL structure. After this update, a new node is created to express the new satellite. After this Node 15 has been fully specified it will look like (24). (24)

NODE 15 (fully instantiated) Lab: sat Config: [(x9)FOC ], [ (def x9)TEMP] SubCat: prep

Thus, the tree structure of Figure 5 is incrementally extended with a new satellite constituent SAT for its rightmost constituent. Only the PREP position will be expressed since there is no more material in this term. SENTENCE 1

4

SUBJ 2

9

VC 5

PRO 3

6

AUX

/he/

/would/

14 SAT 10

7

PRED 8

/come/

Figure 5. Partial expression tree; stage 2

SAT 15

12

PREP 11

/into/

N 13

/class/

PREP 16

/a-t/

Towards a speaker model of FG

353

It turns out now that the hesitation was not long enough, and has not led to a solution of the search problem. It is therefore followed by three additional pauses of different types: silence; the prolonged uh- and the meaningless interjection you know. The first two belong structurally to move M6. We think, however that the you know bit should be seen as a new Move, M7, pushed on top of whatever is in working memory. Arguably, whatever was there is ‘saved’ in the background, for later elaboration. The new move M7 is a very simple one. The contents of its Act, A7, are a sole illocutionary operator. Its function is to recapture the attention of the hearers which may be flagging because of the long pause. The structure of M7 is as in (25) below. It will be sent directly to the expression component, after which a lexicon search will provide its form. (25)

(M7: [INFRM [(A7: [ATT])]])

For this side step a new top node is created, which gets the pragmatic structure of (25) as the only filler of its Config field. The ATT operator will get expression via a special interj(ection) node, which searches the lexicon for a node of the grammatical category ‘interjection’. (26)

NODE 1’ (fully instantiated) Lab: sentence Config: [att] FncFtrs: att SubCat: interj NODE 2’ (uninstantiated) Lab: interj Config: [FORM, [INTERJ], TYPE ] FncFtrs: TYPE NODE 2’ (partially instantiated) Lab: interj Config: [FORM, [INTERJ], att ] FncFtrs: att NODE 2’ (fully instantiated) Lab: interj Config: [ynow , [INTERJ], att ] FncFtrs: att SubCat: ynow

354

Dik Bakker and Anna Siewierska

In the meantime, the memory search for the value of the temporal satellite has been completed. Copies of the saved IL and RL structures are popped back to working memory again. Our (tentative) analysis in terms of FDG is that IL is updated as in (27). (27)

(M6: [INFRM [(A6: [DECL (P1)S (P2)A [R1(x1)TOP R2(x8:CLASS)FOC T1(f1: GO INTO) R3(x9: 3+ MINUTE)FOC])])]])

Recall that expression had stopped at the SAT node as given in Figure 5 above, of which the preposition has been expressed. Control returns to this Node 15. Its Config field will be extended as follows. Again, the new elements are represented in bold characters. (28)

NODE 15 (fully instantiated) Lab: sat Config: [(x9)FOC ], [(def (3 OR 4) x9: minute [N])TEMP] SubCat: prep

Two things will happen after this. On the one hand, the syntactic structure of Figure 5 will be further expanded with the new elements on node 15. This gives us the expression tree in Figure 6 below. SENTENCE 1 SUBJ 2

PRO 3

/he/

5

AUX 6

/would/

4

9

VC 7

10

PRED 8

/come/

14 SAT

PREP 11

12

15

N PREP NUM 13 16 18

/into/ /class/

/a-t/

/three/

SAT 21

OP NUM 20 22

/or/

/f/

Figure 6. Partial expression tree; stage 3

At the same time, the new information on Node 15 will percolate upwards to the top node. As a result, the Config field of Node 1 will contain the following RL level information:

Towards a speaker model of FG (29)

355

(p1: [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT ) (def (3 OR 4) x9: minute [N])TEMP] ))])

When the rest of the satellite is finally expressed the speaker interrupts herself because she wants to revise her initial estimate. This is an example of self-monitoring, which is represented in the model. Information from the working memory is fed back to semantic memory and added to the overall discourse structure after evaluation. It is this evaluation which leads to the instruction to stop expression, and reconsider. A new planning period starts with a new pause. It is followed by a corrected version of the contents of referential act R3. (30)

… precísely one minute after the hóur,

This is the result of an incremental update of the three levels. We first give the updated IL and RL levels, the former without an attempt at a conceptual representation. (31) contains the new referential act R4. Note that, although the act is new, the variable that it refers to is that of R3, i.e. x9. This variable stands for the mental notion under discussion. (31)

(M6: [INFRM [(A6: [DECL (P1)S (P2)A (C6: [R1(x1)TOP R2(x8:CLASS)FOC T1(f1: GO INTO) R3(x9: 3+ MINUTE)FOC R4(x9: … )FOC])])]])

Correspondingly, the Config field of Node 15 will again be updated with the contents of the time satellite. This gives us the following. Note that again only the material represented in bold will be expressed. (32)

NODE 15 (fully instantiated) Lab: sat Config: [(x9)FOC ], (def (1: precise [A]) x9: minute [N]: (def sg x10: hour [N])POST)TEMP] SubCat: prep num n x

Accordingly, the tree of Figure 6 will be further expanded. Expression will continue at the rightmost satellite node again. Figure 7 shows the results; only a partial tree is given this time. Again, the new information percolates upwards to Node 1. This time,

356

Dik Bakker and Anna Siewierska

however, it is not added to the Config structure, but replaces part of it. The new contents of the top node will be as in (33).22 (33)

(p1: [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT ) (def (1: precise [A]) x9: minute [N] : (def sg x11: hour [N])POST))TEMP]))])

SAT 15 23

PREP 16

NUM 18

/a-t/

/three/

OP 20

/or/

/precisely/

26

NUM 22

NUM

/4/

25

24

/one/

/minute/

28

N

27 29

/of/

SAT

30

31

/the/ /hour/

Figure 7. Partial expression tree; stage 4

After (30) – previously (11c) – has been uttered, the monitor appears to have been active again. The very specific version of the latest temporal satellite is relaxed somewhat in the last contribution to the current act, (11d), now renumbered as (34). (34)

or something like thát

For this speech act, which we interpret as an alternative for the complete time satellite we give only the RL update. Syntactic expansion proceeds again from the top node, via a third SAT node. The final result in the Config of Node 1 is as in (35). (35)

(p1: [(past e1: [hab f1: come [V] (def x1)AG ] (gen sg x8: class [N])ILLAT ) (def (1: precisely) x9: minute [N]: (def sg x10: hour [N])POST)TEMP] ) OR (indef sg x11: [-S,-A,-M,-F]: (x9)COMPAR ) TEMP )])

This terminates the Act, the Move and our discussion of the integration of the FDG model and our dynamic model of the expression rules.

Towards a speaker model of FG

357

4. Conclusions In our contribution to this volume we have done three things. First, we gave a short introduction to our dynamic model for the FG expression rules. Then we looked critically at the Functional Discourse Grammar model as proposed by Hengeveld (this volume). We made some suggestions for the further elaboration of FDG. Finally, we tried to amalgamate our slightly adapted version of FDG and the expression component into one dynamic model of a discourse-oriented FG grammar of the Speaker. A short example from spoken language served as a rough sketch of the workings of this model. Most of the problems that we ran into have been solved, at least to some extent. Many aspects should be worked out in much greater detail, and in the light of more empirical material, preferably recorded and annotated versions of informal spoken conversation. We mention just a few of the major problematic issues that spring to mind. There is the – old – point of keeping track of the material that has been expressed, and the related question of which underlying elements of the IL and RL levels may be expressed more than once. This is of much greater importance in a dynamic model such as the one given above than in the traditional grammar model which deals with the ideal speaker-hearer. Related to this point is the requirement that the model should provide an answer to what may, or must be repeated after hesitations and selfcorrections, and how corrections affect the eventual representations. A second point is the precise nature of the interaction between the respective kinds of memory, and the way information is stored there, what is retrievable, how it may be extended, and under what conditions. A subproblem is the way discourse is represented in the speaker’s memory. This, in turn, calls for a deeper insight into the way speakers monitor their own speech process. Underlying this is the fundamental, and extremely complex problem of finding an acceptable way of representing abstract concepts. Undoubtedly, there are many problems that have been overlooked by us altogether. These can not be solved without state-of-the-art knowledge of the intricacies of human mental models of information processing. Fortunately, within the Cognitive Linguistics (CL) enterprise there are many hundreds of linguists, psychologists and computer scientists working in this area. CL seems to be one of the obvious scientific arenas FG should turn to for discussion, suggestions and solutions about these and other matters. A dynamic speaker model is not only a research goal in its own right. We are confident that the combined work on such a construct will put the

358

Dik Bakker and Anna Siewierska

FG researcher in a better position to test all kinds of hypotheses about spoken language. A rather concrete aspect of the integrated model presented above, one which could introduce an interesting line of research to the theory is the selection of the right subcategorization for the pragmatic and semantic material offered to the expression component, i.e. the semanticssyntax interface. The constraints the model presents may help us see better how cognitively or semantically coherent stretches of input to the expression rules control and shape morphosyntax, with the obvious connections to grammaticalization. We might also find out how certain grammaticalized patterns might in their turn ‘pull’ on semantics via the feedback mechanisms, thus to some extent shaping the rest of the planning process. Eventually, this should lead to a better understanding of the behaviour of the language user, and to a better explanation of why languages are the way they are.

Notes 1.

2.

3.

4.

The authors wish to thank Matthew Anstey for a number of valuable comments and interesting points of discussion with regard to an earlier version of the text. We hope we have done justice to them. Actually, outside-in expansion of layers (or left-to-right as it is called by Hengeveld) is only assumed for the interpersonal level. Nothing is said about the coming into existence of semantic representations. Although we think that here, too, the hierarchy among the respective layers has some predictive value in the outside-in sense (cf. Bakker 1994 for more on this), there can be no direct correspondence between the universal order in which underlying clauses are structured and language-specific surface orders of constituents. Of course we could, if only for the sake of argument, take an extreme functional perspective, denying all autonomy to morphosyntax. It might then turn out to be possible in the grammar of individual languages to determine the combination of all underlying factors which will eventually lead to certain order patterns. These factors could be made available as µ operators to the Phase I expression rule that determines the form of the finite verb. Unless we see the form of the agreement marker as the expression of this conglomerate of factors, i.e. not only of person/number aspects of the Subject, but also as information about its pragmatic status, this would indeed boil down to prerunning the expression rules in a more or less disguised way. This requirement can be made explicit by formulating the role of the expression component such that a grammar of language L is descriptively

Towards a speaker model of FG

5.

6.

7.

8.

9.

10.

359

adequate if it generates precisely the set of well-formed expressions of L. And a theory of grammar is descriptively adequate if it allows for all and only the forms and structures that may occur in human language as such. This is just an example of a possible tree expansion. We have no intention here to develop an optimal constituency theory for English or FG for that matter. That will be a separate and very elaborate exercise, which should be done on the basis of an in-depth formal morphosyntactic analysis of a typologically well-founded sample of languages. Much can be gained from the insights of other functionally based theories which have developed a morphosyntax, such as Role and Reference Grammar (Van Valin and LaPolla 1999). Certain types of speech errors make it clear that phonological information must be available at a relatively early stage of the production of an utterance, such as the frequently observed reversal of initial consonant clusters between two words, either accidental or on purpose (‘spoonerisms’). It would be a challenge to implement intriguing examples such as (1) and (2) below (from Nespor and Vogel 1986: 238), where the presence or absence of flapping in American English depends on the strength of the semantic or pragmatic link between the two sentences: Turn up the heat. I am freezing (… hea[ſ]) vs. Turn up the heat. I am Frances (*… hea[ſ]). Technically speaking, ours is by far not the only way the strategy envisaged here might be implemented. However, in this form we follow models in the psycholinguistic literature, such as Kempen and Hoenkamp (1987) and Levelt (1989), thereby adding to the cognitive adequacy of the FG model. For a fully worked out implementation of the dynamic model, a featurevalue notation will be assumed including a set of logical operators and a unification mechanism. Such a formal system is necessary to make the representations explicit and unambiguous, and to define the operations on them in a straightforward way. A proposal for such a system may be found in Bakker (1989; 1994: 263–296). Since these are mainly technicalities which may obscure the current discussion we will stick to a more informal version here. In the representation of nodes we will leave out the fields that have no value. Typically, the models for speaker and hearer are sketched more or less independently, although they use the same linguistic and non-linguistic knowledge resources. However, since speakers obviously hear and monitor their own output a realistic model of the speaker should at least contain a feedback loop to the hearer model. Hesp (1990) is a critical discussion of the MNLU proposal from the perspective of the psychology of language. The author shows that the MNLU model as introduced in Dik (1989) cannot work, and is for a number of reasons not in harmony with the models and experiments found in psychological literature. See our discussion below and Nuyts (this volume).

360

Dik Bakker and Anna Siewierska

11.

Note that Dik (1989) proposes the FG underlying representations as candidates for mental representations. Long-term memory is assumed to be split up between semantic memory for relative static bodies of knowledge such as the grammar and lexicon, and episodic memory for the storage of events, e.g. what happens during a conversation. Instead of short-term memory we will use the wider and more dynamic notion of working memory. This holds the information directly related to the utterance, typically for a very short time, but it also includes – and runs – the processes which operate on that information. See Carroll (1994: 49f.) for a more detailed discussion of the several types of memory relevant for language processing. In this context the magical ‘7 plus or minus 2 items of information’, stemming from Miller (1956), should be mentioned. In working memory, data and processes are in competition for space and time. In this connection it is tempting to think of grammaticalization as a method to economize on the resources of working memory. We will use the notion GrammarN for the normative standard grammar of the language, as it is used in writing and in formal speech. Good examples of short and incomplete utterances in informal discourse may be found in the contributions to Chafe (1980), in which versions of the renowned Pear Story are discussed. DuBois (1987), also taking the Pear Film as a point of departure, contains interesting observations about the utterances typically used during recall sessions of this short film. More particularly, he focuses on the distributions of nominal and pronominal (or zero) NPs in the expression of grammatical relations. A recent survey of spoken English may be found in Biber et al. (1999), where it is observed that more than a third of the utterances in the spoken English section of the British National Corpus are non-clausal, with an average length of just under 2 words. For an early discussion of abstract predicates in FG meaning definitions see Siewierska (1993), who discusses proposals in Jackendoff (1990). See Faber and Mairal Usón (1999) for a recent discussion of meaning and the lexicon in FG which also takes abstract concepts in meaning definitions as a point of departure. Typically, in this type of story-telling discourse, each Move will be Informing by default. The insertion of continuation markers such as And and And then may well be the expression of this operator. We will follow the convention of distinguishing abstract, languageindependent concepts from the related English words by using capital letters for the former. For the rest, no attempt will be made to say anything about the way mental concepts might be represented. In order not to burden the discussion with matters irrelevant to it, we follow the FG convention of representing adpositions via semantic functions. In

12.

13. 14.

15.

16.

17.

18.

19.

Towards a speaker model of FG

20. 21.

22.

361

Bakker and Siewierska (2002), and following Mackenzie (1992), we defend the position that some adpositions should be seen as predicates rather than grammatical elements. Givón (1983), basing himself on text counts, gives an estimate of between 80% and 90% of the cases in running text. This has, of course, a direct bearing on the (sub)theory of syntactic categories. To date, FG has not developed any theory of the kind; notions such as Noun Phrase and Prepositional Phrase are used in a more or less theoryindependent fashion (cf. Rijkhoff 1992). Our point of departure will be that a syntactic category is in the first place the prototypical formal correlate of some underlying element: e.g. an NP is the prototypical expression form of a term. To the extent that (formal) generalizations are possible over such categories, they may be assigned a language-independent, or universal status. It is unclear to us at the moment to what extent syntactic categories sui generis will turn out to be necessary. Of course, there are different ways to represent corrected underlying elements. For example, we could merge instead of replace, and maintain the old, rejected material. Then we could have something like (i): i.

[(def (3 OR 4) (1: precise [A]) x9: minute [N]: (def sg x10: hour [N])POST)TEMP]

We think, however, that our replacement solution is closer to the intuition that the speaker will most probably remember the latest version as part of the discourse structure. The fact that he made a correction is possibly remembered as such in the episodic memory, which also stores a number of details about the conversation, but not as a part of the story as it is told.

References Bakker, Dik 1989 A formalism for Functional Grammar expression rules. In: John H. Connolly and Simon C. Dik (eds). 45–64. 1994 Formal and Computational Aspects of Functional Grammar and Language Typology. Amsterdam: IFOTT. 1999 FG Expression Rules: From templates to constituent structure’. Working Papers in Functional Grammar 67. 2001 The FG expression rules: A dynamic model. Revista Canaria de Estudios Ingleses 42. 15–53. Bakker, Dik and Anna Siewierska 2002 Adpositions, the lexicon and expression rules. In: Ricardo Mairal Usón and Maria Jesús Pérez Quintero (eds). 125–177.

362

Dik Bakker and Anna Siewierska

Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 Longman Grammar of Spoken and Written English. Harlow: Longman. Borsley, Robert D. and Janig Stephens 1989 Agreement and the position of subjects in Breton. Natural Language and Linguistic Theory 7: 407–428. Brandsford, John D. and Jeffrey J. Franks 1971 The abstraction of linguistic ideas. Cognitive Psychology 2: 331– 350. Carroll, David W. 1994 Psychology of Language. Pacific Grove: Brooks/Cole Publishing Company. Chafe, Wallace L. 1987 Cognitive constraints on information flow. In: Russell Tomlin (ed.). 21–51. Clark, Herbert H. and Eve V. Clark 1977 Psychology and Language. New York: Harcourt Brace Jovanovich. Connolly, John H. and Simon C. Dik (eds) 1989 Functional Grammar and the Computer. Dordrecht: Foris. Croft, William 1995 Autonomy and functionalist linguistics. Language 71: 490–532. Crystal, David 1975 The English Tone of Voice. London: St. Martin. Dik, Simon C. 1978 Stepwise lexical decomposition. Lisse: Peter de Ridder. 1986 On the notion “functional explanation”. Belgian Journal of Linguistics 1: 11–52. 1989 FG*C*MNLU: Functional Grammar computational model of the natural language user. In: John H. Connolly and Simon C. Dik (eds). 1–28. 1997 The Theory of Functional Grammar, Part I. Berlin and New York: Mouton de Gruyter. Du Bois, John 1980 The discourse basis of ergativity. Language 63: 805–855. Faber, Pamela and Ricardo Mairal Usón 1999 Constructing a Lexicon of English Verbs. Berlin and New York: Mouton de Gruyter. Fodor, Jerrold A. 1983 The Modularity of the Mind. Cambridge MA: MIT Press. Furth, Hans 1966 Thinking without Language. New York: Free Press.

Towards a speaker model of FG

363

Givón, Talmy 1975 Focus and the scope of assertion: Some Bantu evidence. Studies in African Linguistics 6: 185–205. 1987 Beyond foreground and background. In: Russell Tomlin (ed.), 175– 188. 1988 Pragmatics of word order: Predictability, importance and attention. In: Michael Hammond, Edith Moravcsik and Jessica Wirth (eds), Studies in Syntactic Typology, 243–284. Amsterdam and Philadelphia: Benjamins. Givón, Talmy (ed.) 1983 Topic Continuity in Discourse: Quantitative Cross-Language Studies. Amsterdam and Philadelphia: Benjamins. Halliday, Michael A.K. 1994 An Introduction to Functional Grammar. 2nd edition. London: Arnold. Hannay, Mike 1991 Pragmatic function assignment and word order variation in a Functional Grammar of English. Journal of Pragmatics 16: 131–155. Hannay, Mike and A. Machtelt Bolkestein (eds) 1998 Functional Grammar and Verbal Interaction. (Studies in Language Companion Series 44). Amsterdam and Philadelphia: Benjamins. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127–157. This vol. The architecture of a Functional Discourse Grammar. Hesp, Cees 1990 A critique of FG-CNLU. Working Papers in Functional Grammar 35. Amsterdam: University of Amsterdam. Jackendoff, Ray 1990 Semantic Structures. Cambridge MA: MIT Press. Johnson-Laird, Philip N. 1983 Mental Models. Cambridge: Cambridge University Press. Johnson-Laird, Philip H., Clive J. Robins and L. Velicogna 1974 Memory for words. Nature 251: 704–705. Kempen, Gerard and Eduard Hoenkamp 1987 An incremental procedural grammar for sentence formulation. Cognitive Science 2: 201–258. Levelt, Willem J. M. 1989 Speaking. From Intention to Articulation. Cambridge MA: MIT Press.

364

Dik Bakker and Anna Siewierska

Mackenzie, J. Lachlan 1992 English spatial prepositions in Functional Grammar. Working Papers in Functional Grammar 46. Amsterdam: University of Amsterdam. 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), 267–295. 2000 First things first: towards an incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. Mairal Usón, Ricardo and María Jesús Pérez Quintero (eds) 2002 New Perspectives on Argument Structure in Functional Grammar. Berlin and New York: Mouton de Gruyter. Miller, George A. 1956 The magic number seven, plus or minus two. Psychological Review 63: 81–97. Nespor, Marina and Irene Vogel 1986 Prosodic Phonology. Dordrecht: Foris. Rijkhoff, Jan 1992 The noun phrase. Doctoral dissertation: University of Amsterdam. Schank, Roger C. and Robert P. Abelson 1977 Scripts, Plans, Goals and Understanding. Hillsdale: Erlbaum. Schweller, Ken G., William F. Brewer and D.A. Dahl 1976 Memory for illocutionary forces and perlocutionary effects of utterances. Journal of Linguistic and Verbal Behavior 15: 325–337. Sells, Peter 1985 Lectures on Contemporary Syntactic Theories. Stanford: CSLI. Siewierska, Anna 1993 Semantic functions and theta-roles. Working Papers in Functional Grammar 55. Amsterdam: University of Amsterdam. Tomlin, Russell (ed.) 1987 Coherence and Grounding in Discourse. Amsterdam and Philadelphia: Benjamins. Van Valin, Robert D. and Randy LaPolla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press.

Epilogue Kees Hengeveld 1. Introduction In the first chapter of this volume I presented a basic outline of Functional Discourse Grammar (FDG),1 pulling together various strands of research in FG over the last decade. Basic features of FDG not shared by earlier versions of Functional Grammar are (i) its top-down organization, (ii) an architecture which is both modular and hierarchical,2 (iii) the presence of a structural3 module as a separate component of the underlying clause structure, and (iv) the location of the cognitive and contextual components. My initial presentation of FDG focused on just these characteristics, ignoring many other issues. The subsequent articles raise quite a number of these issues, and in this chapter I will try to address some of them.4 Obviously, many topics will remain untouched: FDG is a research programme rather than a fully-fledged theory. The main questions that I will touch upon here are the following: (i) What is FDG a model of; (ii) How does FDG operate; (iii) What do the cognitive and communicative components look like; (iv) What are the relevant operator and modifier slots in FDG; (v) Where are semantic, syntactic, and pragmatic functions located in FDG; (vi) What is the position of the fund in FDG? The first three questions concern the architecture of FDG itself. The last three concern the incorporation of existing FG components into FDG. After considering the six issues one by one, I will present a revised and more detailed version of FDG in the final section of this paper.

366

Kees Hengeveld

2. A model of encoded intentions and conceptualizations Since the introduction of the layered structure of the clause in FG (Hengeveld 1988, 1989; Dik 1997), the question has been raised (e.g. by Bolkestein 1992 and Harder 1996) whether this structure 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. These two interpretations have become known as the process and pattern views on underlying representations. In this volume Fortescue raises the same issue explicitly in relation to FDG, and concludes that it is a process model. Anstey, Bakker and Siewierska, Harder, and Nuyts all arrive at the same conclusion. It is important to note that this is not how I intend FDG to be taken. I conceive of FDG as a pattern model, i.e. as a model that represents linguistic facts. The process interpretation given to it by the aforementioned authors is probably due to two facts. First of all, FDG is presented in Hengeveld (this volume) as the grammatical component of a wider theory of verbal interaction (cf. Dik 1997) and as such interacts with a conceptual and a contextual component. These latter components, however, are not part of the grammar as such, even though many grammatical phenomena can be studied much more fruitfully if such components are assumed to exist. Secondly, in FDG, as in FG, the patterns of language are described as reflecting the process of communication. This, however, does not mean to say that FDG is a model of that process. The latter point can be clarified starting from a familiar and generally accepted underlying representation in FG: (1)

(e1: [(f1: collapseV (f1)) (x1: houseN (x1)Ø)Pat] (e1)) ‘the collapsing of a house’

The representation in (1) contains three types of variable from the representational level: (e) for states-of-affairs, (f) for relations and properties, and (x) for spatial objects. These variables are defined in terms of entity types in the external, extra-linguistic world. Yet nobody would claim that the formula in (1) represents part of the external world, as it would if it were an ontological representation. Rather, (1) represents a linguistic unit in terms of its function of designating part of the external world. This fact has been obscured by informal terminological usage. It is common to say that (1) represents a state of affairs, whereas it would be more appropriate to say that (1) represents a linguistic unit in terms of its ideational function.

Epilogue

367

Thus, (1) is a linguistic representation based on non-linguistic factors. Similar considerations apply to the representation of discourse acts. Consider the following underlying representation: (2)

(A1: [DECL (P1)Sp (P2)Addr (C1: [(T1: collapsed (T1)) (R1: house (R1))] (C1))] (A1)) ‘The house collapsed.’

This representation contains five types of variables from the interpersonal level: (A) for discourse acts, (P) for discourse participants, (C) for communicated contents, (T) for ascriptive acts, and (R) for referential acts. These variables are defined in terms of the units relevant to the unfolding speech event. Now, in exact parallel to what was said in relation to the representation of states of affairs, this representation should not be interpreted as a direct representation of the actual communicative situation, but rather as representing a linguistic unit in terms of its function in communication. Thus, (2) does not represent a discourse act, as it would if it were an action theory representation, but should be interpreted as the formalization of a linguistic unit in terms of its interpersonal function. This is precisely why FDG can be called a functional model of language: it captures the structure of linguistic units in terms of the world they describe and the communicative intentions with which they are produced, i.e. in terms of their representational and interpersonal functions.

3. The dynamic construction of linguistic expressions A pattern model of language is not necessarily a static one. In this volume the contributions by Bakker and Siewierska, Harder, and Mackenzie stress the importance of a dynamic implementation of the model. The same issue is raised in Bakker (1999; 2001) and Mackenzie (1998; 2000). For a model of grammar a dynamic interpretation entails an implementation that mirrors the language production process in individual speakers. Again, this does not mean that the grammatical model is a model of the speaker. Rather, the model is assumed to be more effective, the more closely it resembles this language production process. For the implementation of FDG this means that the various levels operate simultaneously, though with a slight delay applying from higher to next lower levels. That is, as soon a certain decision at the interpersonal level allows the selection of elements at the representational level, the latter level becomes active.

368

Kees Hengeveld

And structural choices are made as soon as sufficient information is available from the interpersonal and representational levels. One could visualize this provisionally as in Figure 1, anticipating a more elaborate picture in Section 8. Interpersonal Level activated Representational Level activated Structural Level activated →→ time axis →→ Figure 1. The dynamic interpretation of FDG

The dynamic approach faces an important problem that touches upon the basic principles of FG. The dynamically produced order of constituents at the structural level is linear. But the underlying semantic representations in FG are generally assumed to be unordered. As a result, in a dynamic implementation of the model either the semantic representations have to be specified before expression starts, or the semantic representations have to be linearly ordered too. In this volume, Bakker and Siewierska opt for the latter solution. This is a rather drastic departure from current FG practice, and seems to suggest that differences in the structural organization of languages reflect differences in their semantic organization. Phenomena which contradict such an approach are the existence of cataphora, and of reflexive pronouns preceding their antecedents. I will not attempt to formulate a complete alternative solution for this problem here. A promising possibility involves a strict separation between lexemes on the one hand, and the frames in which they occur on the other, as proposed in García Velasco and Hengeveld (2002). Once these two elements are dissociated from each other, one might argue that expression starts at the moment that frames have been selected, and the first lexeme is inserted into the appropriate slot. Expression then evolves parallel to the insertion of further lexemes into the remaining slots. A full account of dynamic expression should not stop at the structural level. A structural representation is part of the underlying representation of discourse acts, but is not the same as the acoustic, orthographic, or signed output. This output requires a separate component within the overall model. The place of the output in relation to the FDG model will be presented at the end of the next section.

Epilogue

369

4. The conceptual and contextual components In my initial presentation of FDG (Hengeveld this volume) the cognitive component was defined rather sketchily, and far too restrictively, as one which defines the communicative and linguistic competence of a speaker and his/her knowledge of the world. Several authors in this volume, but particularly Nuyts, have drawn attention to the necessity of a further elaboration of this component. A dynamic implementation of FDG, as discussed in the previous section, begs for a ‘driving force’ to trigger the creation of linguistic expressions. This driving force is embodied by a conceptual component, within which communicative intentions develop and combine with suitable conceptualizations. The conceptual component is not part of the grammatical model, but in actual language use serves as a trigger for the grammar to operate. Anstey (2002) presents a formal representation of the conceptual level which basically mirrors the underlying linguistic representation, but abstracts away from purely linguistic distinctions, and thus generalizes, for instance, across operators and satellites. Within the conceptual component there is an important distinction between communicative intentions on the one hand, and conceptualizations on the other. This is reflected at the grammatical level in the presence of interpersonal and representational levels. In his contribution to this volume, Connolly makes exactly the same distinction for the contextual component, incorporating insights from Discourse Representation Theory. He proposes a detailed and convincing format for the description of the contextual component, distinguishing between the contextual description and the interpersonal description of the communicative context. It is important to note that in Connolly’s view, which I share, the contextual component is interpreted as a discourse domain (cf. Vet 1986) and thus contains a description of the knowledge shared by the interlocutors. This interpretation of the contextual component is different from the one advocated in Anstey’s contribution to this volume. He assumes different discourse models for speaker and addressee. Together, the proposals by Anstey for the conceptual component and Connolly for the contextual component allow for a systematic treatment of certain linguistic phenomena that can only be understood in terms of a wider theory of verbal interaction. Chief among these is the question of pronominalization, which is the topic of Cornish’s contribution to this volume. At the conceptual level, referents do not change: coding differences do not reflect differences in perception. To simplify matters a bit, referents

370

Kees Hengeveld

are encoded pronominally when they are available within the contextual component, and nominally when they are not. The importance of the interaction between the conceptual and contextual modules is also stressed by Gómez-González in her contribution to this volume. On the basis of the discussion in the previous section and in this one we can give a more detailed presentation of FDG, shown in Figure 2, as the grammatical component of a wider theory of verbal interaction. In this Figure, boxes represent components and levels, and circles represent operations. The main components are the conceptual component, the contextual component, the acoustic component,5 and the grammatical component. The grammatical component distinguishes an interpersonal, a representational, and a structural level. The vertical arrows indicate the following: the conceptual component drives the grammatical component, both as regards the formulation of the interpersonal and of the representational level; the underlying representation resulting from this operation is encoded at the structural level;6 and interpersonal choices together with the structural configuration determine the phonetic properties of the utterance. The other way round, the result of these operations feeds the conceptual component through the contextual component. The horizontal arrows indicate that all grammatical and articulatory levels feed this contextual component, creating possible antecedents at the representational level. But the interpersonal level draws on the contextual component too, for instance with respect to bystander deixis.

Epilogue

371

Conceptual Component

Interpersonal Level

Representational Level

Encoding

Acoustic Component

Structural Level

Contextual Component

Grammatical Component

Formulation

Articulation

Expression Level

Figure 2. Main components of FDG within a wider model of verbal interaction

372

Kees Hengeveld

5. Operators and modifiers The FDG-model introduces a number of additional layers and as a result creates a number of additional potential slots for operators and modifiers. There are two areas where the additional slots seem to be particularly helpful: (i) the distinction between communicated (C) and propositional (p) contents which results from upward layering at the representational level offers new opportunities to account for speaker-bound and non-speakerbound modalities; (ii) the introduction of referential acts (R) through downward layering at the interpersonal level provides a new perspective on the layered structure of term phrases. With respect to the first issue, one should note that in FDG propositional contents (as opposed to communicated contents) are situated at the representational level, rather than at the interpersonal level. This calls for a redefinition of modal categories: the distinction between communicated and propositional contents creates an additional level of analysis. In fact, this is a welcome addition since it helps to distinguish between modal, and in particular evidential, categories which are restricted to main clause (i.e. speaker-bound) uses, and modal, and in particular epistemic, categories which may be used in embedded clauses and then express the propositional attitude of a main-clause participant, rather than the propositional attitude of the speaker. In his contribution to this volume, Verstraete does indeed locate speaker-bound modalities at the interpersonal level, i.e. as operators at the level of the communicated content. With respect to the second issue, consider the following full representation of a referential act (R) in which reference is made to a first order entity and in which an ‘Ω’ indicates an operator slot and a ‘µ’ a modifier slot. (3)

(Ω R1: [(Ω x1: [(Ω f1: LexemeN (f1): µ (f1))] (x1): µ (x1))] (R1): µ (R1))

In this representation of referential acts in FDG the same number of layers, operators and modifiers can be distinguished as in Rijkhoff (1992), but now they can also be provided with a systematic underpinning in terms of the functions of the various layers. Thus, f-operators concern the nature of the property (f) itself, x-operators concern properties of the set of designated entities, and R-operators situate the referential act in the actual communicative situation. For modifiers within term phrases similar things may be said. Consider the following series of examples:

Epilogue (4)

a b c

373

a poor driver a poor man Poor guy!

In (4a) the property of being a driver is modified by the adjective, which does not state anything about the driver entity itself. One might well talk about a poor driver who is very rich or even about a rich poor driver. Bolinger (1967) calls the type of modification in (4a) ‘reference modification’. In (4b) the adjective selects a subclass of the set of entities defined by the head noun. This is the type of modifier typically treated in the FG literature. In Bolinger’s terms this is a case of referent modification. In (4c), finally, the adjective embodies the subjective evaluation by the speaker of the entity referred to. The functional differences between these three types of modifiers are reflected in structural differences in various languages, as has for instance been demonstrated for Spanish in Berniell (1995). The characterizations given to the various uses of the adjective poor in (4) furthermore tie in nicely with their interpretation as f-modifier, x-modifier, and R-modifier respectively.

6. Functions As Anstey notes in his contribution to this volume, the functional hierarchy of influence (pragmatics > semantics > syntax) is instantiated in the existence and ordering of the interpersonal, representational and structural levels respectively. Similarly, these levels provide straightforward slots for pragmatic, semantic and syntactic functions. With respect to pragmatic functions Cornish and Mackenzie provide arguments in this volume for locating pragmatic functions at the interpersonal level. An important reason to situate pragmatic functions at the interpersonal level is that the selection of a predicate at the next level down, the representational level, is sensitive to the information status of constituents (see e.g. Bolkestein and Risselada 1985; 1987). Semantic functions, as in earlier versions of FG, are situated at the representational level. They are part of the frames which are used to build up a semantic representation. Finally, FDG offers a new location for syntactic functions: I propose to situate these at the structural level (cf. also Bakker and Siewierska this volume). In this way, syntactic function assignment can be seen as the outcome of a process in which both pragmatic factors (at the interpersonal level) and semantic factors (at the representational level) are taken

374

Kees Hengeveld

into consideration. Expression rules sensitive to these factors determine which constituents come out as Subject or Object, where this is relevant. This proposal has consequences for the conception of syntactic functions in FG: here they are considered to be purely grammatical rather than semantic notions. The perspectival flavour of syntactic functions is then the product of more basic pragmatic and semantic notions which trigger their appearance, rather than being part of their meaning.

7. The fund In the initial presentation of FDG I did not say anything about the location of the fund in FDG, as noted by Gómez-González and Bakker and Siewierska in their contributions to this volume. Here I can only present a first sketch of this part of the model. The basic idea is that for every level within the model the fund contains the set of basic units which are used to build up that level. Basic units are language-specific inventories. The fund of a given language contains at least the sets of basic units given in Table 1: LEVEL Interpersonal

Representational

Structural Acoustic

BASIC UNITS Illocutionary Frames Lexemes Operators Predication Frames Lexemes Operators Templates Morphemes Prosodic Patterns Sounds

Table 1. Basic units stored in the fund

García Velasco and Hengeveld (2002) propose a new organization of the fund in which predicate frames are split into lexemes on the one hand and predication frames on the other. They give arguments to show that from the perspectives of psychological, pragmatic and descriptive adequacy such an approach is called for. Predication frames and lexemes, together with the relevant operators, are the building blocks of the representational level. A similar approach may be applied at the interpersonal

Epilogue

375

level: acts are based on an illocutionary frame; there are certain lexemes that have an interpersonal content only (personal names, pronouns, interjections);7 and the interpersonal level has its own set of operators. The structural level draws on a set of (word order) templates8 and on a set of grammatical morphemes. This is what Anstey calls a ‘syntacticon’ in his contribution to this volume. Finally, the acoustic level contains at least prosodic patterns and sounds, but conceivably also syllabic and word patterns. Note that this organization of the fund into components corresponding to the various levels is crucial if one strives for a dynamic interpretation of FDG. It allows for expression to start as soon as sufficient information has come in. This information may come from as high up as the interpersonal level and go as deep down as the acoustic level. For instance, the selection of a particular illocutionary frame at the interpersonal level may be sufficient to trigger a certain prosodic pattern at the acoustic level.

8. Summary: expanding the model By way of summary, Figure 3 gives an expanded version of the FDG model which incorporates the various adaptations that have been discussed in the previous sections, and which were triggered by the many interesting issues raised in the various contributions to this volume.

376

Kees Hengeveld Conceptual Component

Frames, Lexemes, Operators

Interpersonal Level

Representational Level

Encoding

Templates, Morphemes

Acoustic Component

Structural Level

Articulation

Prosodic Patterns, Sounds

Expression Level

Figure 3. An expanded version of FDG

Contextual Component

Grammatical Component

Formulation

Epilogue

377

Notes 1.

2. 3.

4.

5.

6. 7. 8.

The use of the label FDG does not mean to suggest a radical departure from the basic principles of FG. However, since the top-down organization of FDG does imply that all existing components of FG have to be reinterpreted, it is convenient to have a separate label to refer to this new research enterprise. In this volume, however, Moutaouakil presents another version of FG which is both modular and hierarchical. I have substituted the term ‘structural level’ for the term ‘expression level’ which was used in the first chapter of this volume, since expression involves more than just structure. I will come back to this issue in Section 3. I am grateful to Matthew Anstey, Annerieke Boland, Lachlan Mackenzie, and Gerry Wanders for discussion of several of the topics dealt with in this chapter. Of course the acoustic component is only relevant for spoken language. For sign language this would be a sign component, for written language an orthographic component. Note that the structural level in fact has to be split up into a morphosyntactic and a phonological level. See e.g. Mackenzie's and Moutaouakil's contributions to this volume. The difference between frames and templates is that frames are unordered semantic configurations whereas templates are ordered syntactic configurations.

References Anstey, Matthew 2002 Layers and operators revisited. Working Papers in Functional Grammar 77. Bakker, Dik 1999 Functional Grammar expression rules: From templates to constituent structure. Working Papers in Functional Grammar 67. 2001 The FG expression rules: A dynamic model. Revista Canaria de Estudios Ingleses 42: 15–53. Berniell, Silvia 1995 La posición del adjetivo en Español: Un análisis en el marco de la Gramática Funcional. MA thesis. University of Amsterdam. Bolinger, Dwight 1967 Adjectives in English: Attribution and predication. Lingua 18: 1–34.

378

Kees Hengeveld

Bolkestein, A. Machtelt 1992 Limits to layering: Locatability and other problems. In: Michael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered Structure and Reference in a Functional Perspective, 387–407. Amsterdam and Philadelphia: Benjamins. Bolkestein, A. Machtelt and Rodie Risselada 1985 De tekstuele funktie van valentie: Latijnse drieplaatsige predikaten in kontekst. Tijdschrift voor Taal- en Tekstwetenschap. 5 (2): 161–176. 1987 The pragmatic motivation of syntactic and semantic perspective. In: J. Verschueren and M. Bertuccelli-Papi (eds), The Pragmatic Perspective, 497–512. Amsterdam and Philadelphia: Benjamins. Dik, Simon C. 1997 The Theory of Functional Grammar. Part I: The Structure of the Clause. Kees Hengeveld (ed.), 2d rev. ed. Berlin and New York: Mouton de Gruyter. García Velasco, Daniel and Kees Hengeveld 2002 Do we need predicate frames? In: Ricardo Mairal Usón and María J. Pérez Quintero (eds), New Perspectives on Argument Structure in Functional Grammar, 95–123. Berlin and New York: Mouton de Gruyter. Harder, Peter 1996 Functional Semantics. A Theory of Meaning, Structure and Tense in English. Berlin and New York: Mouton de Gruyter. Hengeveld, Kees 1988 Layers and operators. Working Papers in Functional Grammar 27. 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25 (1): 127–157. this vol. The architecture of a Functional Discourse Grammar. Mackenzie, J. Lachlan 1998 The basis of syntax in the holophrase. In: Mike Hannay and A. Machtelt Bolkestein (eds), Functional Grammar and Verbal Interaction, 267–296. Amsterdam and Philadelphia: Benjamins. 2000 First things first. Towards an Incremental Functional Grammar. Acta Linguistica Hafniensia 32: 23–44. Rijkhoff, Jan 1992 The noun phrase: a typological study of its form and structure. Ph.D. dissertation. University of Amsterdam. Vet, Co 1986 Pragmatic approach to tense in Functional Grammar. Working Papers in Functional Grammar 16.

Index of names

Abelson 76, 341 Amjad 322 Anderson 75, 149 Anstey vii, x, 23, 170, 211, 358, 366, 369, 373, 375, 377 Aoyama 127, 133 Asher 101, 113 Atkinson 79 Austin 78 Bach 31, 57 Baddeley 79 Bakhtin 94, 110 Bakker x, 29, 34, 55, 57, 182, 212213, 325, 327-328, 358-359, 366368, 373-374 Barthes 318 Bartlett 76 Bates 214 Bauer 34 Bellert 269 Ben Achour 40 Bencini 233 Benveniste 20, 318 Berg 52 Berman 199 Berniell 375 Berry 236 Biber 360 Birner 123, 145 Bock 278 Boland 52, 56, 377 Bolinger 207, 259-260, 375 Bolkestein 2, 17, 25, 27, 44, 55, 5859, 93, 117, 127, 132, 142, 152, 170, 211, 226, 259, 261- 262, 269, 310, 315, 366, 373 Borsley 327 Brandsford 341 Broadbent 79

Brown, G. 131 Brown, P. 202 Brown, R. 59 Brunhoff 313 Bunt 59 Butler 19-20, 25, 31, 55, 73, 233 Butterworth 83 Buyssens 56 Campbell 56 Carnap 57 Carroll 233, 360 Casenhiser 233 Chafe 173, 220-221, 233, 236, 290, 342-343, 360 Cheshire 137 Chomsky 30, 39, 57, 168 Clark, E. 345 Clark, H. 16, 182, 345 Clark, K. 94 Comrie 34 Connolly viii, 25, 91, 98, 101, 103, 108, 211, 213, 369 Contini-Morava 118 Cornillie 269 Cornish viii, 117, 119, 128, 136-137, 139, 215, 222, 369, 373 Coulthard 5, 92, 106-107 Crevels 18 Croft 328 Cruttenden 223 Crystal 342 Cuvalay 302, 305 Dahl 32 Dancygier 252 Davidse 269 Davidsen-Nielsen 206 Davies 252 De Groot 55, 189, 304 De Jong 40

380

Index of names

De Schutter 25, 289 Decarrico 190 Declerck 15, 251, 259, 269-270 Devlin 101, 110 Devriendt 25 Dietz 40 Dik vii, x, 2, 5, 24-43, 45, 50- 51, 55-60, 73, 78, 89-93, 95, 103, 106, 111, 117, 120, 125, 127, 131, 133, 138-139, 141- 142, 151-154, 160, 162-163, 168, 170, 173-175, 179-180, 182-183, 186, 188, 211, 218, 222-223, 235-236, 243-245, 264-265, 275-277, 286, 292, 300303, 305, 307-309, 313, 315-321, 325-327, 331, 337- 338, 344, 359360, 366 Diver 118, 127 Dowty 59 Du Bois 234 Engberg-Pedersen 25, 56, 172, 173 Engdahl 223, 229 Erteschik-Shir 120, 125-126, 139 Faber 25, 30, 41, 360 Fairclough 94 Fillmore 32, 56, 76, 214, 220, 277 Firth 104 Fodor 74, 84, 345 Foley 40, 58, 246, 252 Fortescue viii-ix, 25, 34, 151- 152, 154, 160-162, 171, 189, 199, 211, 213, 366 Fox 229, 235 Franck 207 Franks 341 Fraser 15 Fried 239 Furth 345 García, E 118, 130, 145, 229 García Velasco 57, 368, 374 Garrod 76 Gebruers 32 Gerrig 16 Givón 175, 229, 234, 342, 347

Glover 137 Goethals 269 Goldberg 118, 214, 233 Gómez-González ix, xi, 25, 117, 131, 144, 211-212, 218- 220, 224, 227, 229, 234, 370 Gómez Soliño 18 Gonzálvez-García 233 Goodman 214 Goossens 25, 255 Grice 49, 146, 217 Gulla 92, 96, 299 Gundel 119, 129, 133, 137, 145 Halliday 29, 186, 224, 232, 236, 246, 253, 263, 266-267, 270, 295, 342 Hancher 91 Hannay 2, 17, 25, 117, 123-124, 126, 132, 140-141, 145, 152- 153, 164, 166, 170, 174-175, 189, 204, 224, 226, 229, 235, 343 Hansen 207 Harder ix, 44, 53-54, 58, 151, 170, 197, 199, 207, 213, 229, 366-367 Hare 253 Hasan 103 Hedberg 133 Helmbrecht 61 Hengeveld vii-x, 2, 5, 7-8, 15, 2324, 26, 34, 40-53, 55-60, 73- 74, 76, 78, 80, 82, 84, 89, 92- 93, 96, 108, 111, 117, 132, 139-140, 144, 153-154, 160, 162-163, 165, 167, 169, 171, 174, 180, 182-188, 197, 199-203, 211-215, 232, 243-252, 254-255, 264-267, 269-270, 275276, 285-286, 288, 295, 299, 305, 312, 315, 321, 325-327, 329, 338340, 344, 347, 357-358, 365-369, 374 Hesp 39, 73, 359 Hickmann 293 Hitch 79 Hoekstra 25

Index of names Hoenkamp 359 Holquist 94 Hopper 280 Huffman 121-123, 126, 129, 144145 Hymes 34, 56, 104 Inchaurralde viii, 73, 76, 85, 215 Itkonen 167 Jackendoff 179-181, 188, 212, 236, 269, 278, 292, 360 Jacobsen 171-172 Jadir 18 Jakobson 27, 30 Johnson 32 Johnson-Laird 341 Kamp 100-101 Katz 57 Kay 214, 277 Keizer 15, 25, 58-59, 117, 131, 211, 235, 300, 313 Kemmer 219 Kempen 359 Kies 234 Kirsner 118 Knowles 234 König 295 Kress 94, 105 Kroon 5, 17, 25, 44, 93, 95-96, 110, 182, 201, 211, 299, 309 Kuno 223 Kuroda 223 Kwee 55 Ladd 223 Lakoff 76, 79 Lambrecht 119-121, 125-126, 131, 144, 174, 222-224, 233, 235 Langacker 75, 129, 172, 214-215, 220, 232-234, 236 LaPolla 359 Laury 146 Lauwers 269 Leonard 127, 133-134 Levelt 2, 7, 73, 180, 186-189, 213, 278, 359

381

Levinson 60-61, 202 Lewis 58 Liedtke 18, 46, 93, 310 Loftus, E. 75 Loftus, G. 74 Longacre 27 Lucy 8 Lyons 18, 40, 54, 60, 246, 253, 259 Mackenzie viii-xi, 2, 17, 25, 44, 51, 55-59, 117, 131, 171, 175, 179, 204, 211-212, 215, 224, 232-233, 235, 276, 300, 302, 313, 343-344, 367, 373, 377 MacWhinney 212 Maes 146 Mahfouz 313 Mairal Usón 30, 41, 56-57, 60, 360 Mann 92 Marin 87 Martin 94 Martín Arista 25 Martín Mingorance 25, 41 Mathesius 32 Matthews 29 Miller 360 Minegishi Cook 86 Minsky 76 Mithun 154-155, 172, 234 Moutaouakil ix, 2, 25, 44, 182, 211, 262, 270, 299-300, 302, 304-305, 310, 313-315, 317-318, 377 Nakayama 154, 156, 158-159, 161, 166, 172-174 Nattinger 190 Nespor 359 Nichols 31 Noonan 214 Nuyts ix, 25, 38, 73, 170-171, 175, 181, 212-213, 215, 232, 269, 275, 277-280, 284-285, 288-289, 292295, 297, 360, 366, 369 Olbertz 25, 55-56 Olson 79 Ortony 76, 87

382

Index of names

Östman 146 Otheguy 118 Palmer 245, 249, 270 Pérez Quintero 25, 41 Perkins 279 Piťha 32, 34, 56-57 Polinsky 144, 223 Postal 57 Prideaux 34, 235 Pu 235 Quine 151, 178 Quirk 206, 210 Radden 79, 87 Reed 251, 269, 271 Reichling 32 Reid 119, 121-122, 130, 144 Reinhart 126, 131, 145 Reyle 100-101, 115 Rijkhoff 25, 42, 285, 300, 304, 372 Rijksbaron 61-62, 64 Risselada 17, 375 Rolland 40 Rosengren 144 Ruck 123 Rumelhart 76, 87 Saffran 79, 87 Sánchez García 56 Sanford 76, 87 Schank 76, 341 Schilperoord 83, 88 Schwartz 79, 87, 88 Schweller 341, 365 Searle 40, 91, 108, 115 Selkirk 223, 242 Sells 337 Sethuraman 233, 239 Seuren 58, 70 Shatz 79 Shibatani 171 Shiffrin 79 Siewierska x, 25, 29, 57-58, 117, 122-123, 132, 145, 182, 212-213, 223, 234, 325, 360, 365, 366-368, 373-374

Sinclair 5, 92, 106-107 Slobin 173, 199 Smits 228 Sperber 118 Stamper 103 Stephany 293 Stephens 327 Steuten 18, 40, 92-93, 96-97, 211 Strang 56 Strauss 134-135 Strawson 151 Swadesh 155, 159, 172 Sweetser 8 Taylor 234 Tesnière 56 Thompson 92 Traugott 294-295 Tulving 75 Vallduví 223, 229, 242 Van Belle 269 Van Buuren 46 Van den Berg 2, 18, 44, 93, 132, 140, 211, 310 Van der Auwera 34, 44 Van Hoek 129, 149 Van Leeuwen 94, 105 Van Schaaik 58, 189 Van Turennout 52 Van Valin 40, 56-58, 60, 246, 252, 359 Vendler 59 Vermandere 269 Verschueren 111 Verstraete ix, 42, 213, 243, 262, 269, 374 Vester 25, 59-60 Vet 2, 17, 18, 40, 44, 59-60, 93, 9798, 132, 182, 211, 310, 369 Vogel 359 Wales 56 Wanders 18, 377 Ward 123, 125, 145, 236 Watters 35, 71 Weigand 40, 71

Index of names Weijdema 59, 71 Weinrich 29, 57, 71 Whorf 56 Wierzbicka 76, 88 Wilson 118

Wodak 94, 114 Xinzhang Yang 42, 71 Yule 131 Zacharski 133

383

Index of subjects

ability 36, 245, 247, 249, 251, 254256 aboutness 120, 131, 222 accessibility viii, 144, 175, 180, 189, 219-221, 282-283, 331, 346 acoustic 368, 370, 375, 377 acquisition 56, 62-63, 214, 294, 320, 338, 345 act 5, 7-8, 10-11, 15-16, 43-44, 47, 76, 91, 95-97, 99, 102, 108-111, 135, 152-153, 167, 171, 174, 183186, 188, 190, 192, 200-203, 205, 215, 230, 233, 235, 247, 253, 255, 277, 286-287, 294-295, 347, 351, 355-356, 367, 372 activation 75, 184, 193, 220-222, 235, 319 addressee ix, 5, 15, 77, 80, 83, 90, 108, 119, 121, 125, 127- 134, 136-140, 143-145, 186, 188-189, 191, 202, 206, 215, 234-235, 369 adjacency 102, 109-111, 141 affix 157-159, 165, 169, 171, 173 animacy 157, 161, 234 antecedent 43, 53, 129 apodosis 252, 288 apposition 57, 226 archetypal 300, 303, 312, 315, 320 architecture vii-x, 23, 25, 50, 55, 7374, 80, 89, 167, 169, 212, 243244, 266, 268, 299, 365, 378 argument 16, 28, 35-36, 43, 48, 59, 120-121, 124, 127, 145, 157-158, 161-164, 166, 174, 180, 187, 192, 202, 204, 207, 213, 223, 231, 263, 270, 277- 278, 293, 309, 349, 358 articulation 2, 34, 52, 55, 73, 161, 164-165, 169, 182, 187, 212- 213, 326

ascription 7, 108, 185, 198, 202, 212, 326, 344 aspect 40, 54, 79, 92, 118, 126, 131, 151, 153, 156, 183, 276- 277, 281-284, 290, 293, 329, 358 assertion 123, 125, 186, 223, 231, 235 attention viii-ix, 11, 26, 45, 48, 76, 79, 102, 117, 119, 121, 123-125, 127-130, 133, 135-138, 143, 145, 175, 212-213, 217-218, 220, 222, 228-230, 233, 289, 307, 346, 353, 363, 369 attitude 109, 171, 186, 206, 226, 261-263, 270, 277, 287, 289, 292294, 372 bottom-up vii, 2, 182, 199, 211, 338 buffer 188, 215 bystander 370 categorical 119-120, 122, 139, 143144, 149 category 8, 10, 17, 28, 30-31, 46, 52, 56-59, 78-79, 91, 107-108, 118, 153, 158, 161, 163-164, 167, 176, 179, 198, 207-208, 214, 218-219, 221, 223, 233, 244-247, 252, 254259, 263-268, 270, 275, 277-279, 286-287, 293-294, 300-303, 305, 307-308, 310, 312-315, 320-321, 329-330, 333-334, 350, 353, 361, 372 chain 50, 52, 75, 80, 161, 165, 198, 234 classifying 8 clause ix, xiii, 7, 11, 15, 25, 36, 4142, 44, 56, 60, 89-90, 92- 93, 95, 108, 118-119, 121- 122, 127, 131132, 136, 139-145, 152, 156-157, 160-166, 169, 171-172, 175, 181-

Index of subjects 182, 188, 197, 204, 206-209, 230234, 244-246, 251, 264-266, 275276, 284, 288, 295, 300-302, 304305, 309-310, 313-315, 317, 320, 325-326, 328- 329, 339, 348, 365366, 372 clause combining 41 cleft 172, 174, 296 clitic 118, 122, 127-129, 143, 145, 158 code 16, 44, 49, 153, 199, 206, 269, 280, 295, 329, 345 cognition viii-x, xiii, 3, 17, 38- 39, 50, 56, 74-80, 82-84, 93, 140, 143, 151-152, 170-172, 175, 180181, 183, 185-189, 211-215, 219221, 226-229, 232-233, 235, 241, 275-276, 280, 282, 288-289, 292, 296, 328, 332, 348, 351, 359, 365, 369 cognitive component 3, 17, 74, 7679, 82-84, 93, 175, 212, 214, 369 cognitive grammar 215, 241 cognitive-functional 275-276, 288 coherence 126, 142, 308 cohesion 161, 220, 231, 343 commitment 40, 133, 246-248, 250252, 254-255, 257-259, 261-262, 265, 267-268, 270, 284, 289, 291, 295 communication x, 8, 13, 26, 47, 49, 80-83, 94, 103, 105, 167, 175, 181, 183-186, 217, 233, 278, 291, 301, 315, 319-321, 366-367 communicative competence 3, 30, 74, 326 communicative context 12-13, 36, 44, 49-51, 60, 76, 82-83, 140, 169, 184, 186, 214, 341, 344, 369 complement(ation) 41, 129, 161, 328 computational 25, 39, 55 concept 29, 32, 52, 56, 131, 165, 189, 234, 260, 275-276, 278, 283, 286, 289, 295, 300

385

conceptual 23, 39, 60, 75, 87, 98, 152, 165, 171, 175, 181, 185, 211-212, 218, 220, 226-227, 229, 236, 272, 275, 277-279, 281-283, 285-290, 293-297, 345, 355, 366, 369-370 conceptualization ix, 79, 84, 187188, 215, 218, 227, 236, 247, 277-278, 280-281, 285-286, 288289, 291-292, 295, 297, 307 conditional 16, 248-249, 251, 256257, 269, 287-288 configuration 333, 335, 370 connectionism 75, 79 connective 123, 133, 141 constituent 7, 26-28, 35-36, 38, 46, 52, 56-57, 90, 108, 119, 124, 126, 131, 138, 140, 142- 143, 145, 154-161, 166, 169, 172, 188-189, 197, 207, 217, 221-222, 224, 231, 235-236, 285, 302, 305-307, 309, 316, 326, 328-329, 331-332, 335, 347, 349, 352, 358-359, 361, 368, 373, 377 constituent order (see also word order) 328, 347, 349 content ix, 3, 5-6, 8, 10-11, 15, 37, 46, 73, 77, 82, 90, 97-99, 101, 111, 120, 139-140, 153, 160, 168, 185, 188, 198, 200-206, 208, 220, 247, 251, 277, 300, 309, 320, 340-341, 344, 372, 375 context 2, 12, 17, 27, 36-37, 44, 4951, 56, 60, 76, 82-83, 90, 93, 102103, 108, 110, 118-121, 123, 125127, 136, 139- 140, 142, 145, 153, 155, 157, 159-161, 163, 166167, 169- 170, 173, 184, 186, 188, 190, 214-215, 220, 226, 234236, 249, 252, 256-257, 269, 286, 289, 313, 341, 344, 360, 369 continuity 45, 58, 91, 126, 229, 232, 346

386

Index of subjects

contrastive 145, 155, 166, 169, 172, 174-175, 217, 313 control 30, 85, 161, 167, 289, 290, 358 controller 130 conversation 52, 77, 92, 106, 135, 183-184, 220, 279, 341-343, 346, 357, 360-361 coordination 24, 27, 57, 139, 226, 316 copula 15, 160, 162 coreferentiality 60, 129 current discourse space 215, 229 declarative ix, 56, 75, 90, 131, 166, 175, 179-182, 187-188, 190-193, 202, 204, 207, 250, 252, 270, 287, 345 decoding 73, 83, 214, 217, 259 definite 37, 58, 132, 134, 137, 141, 144, 149, 157, 160 deixis 20, 49, 60, 127, 134, 136, 148-149, 221, 307, 370 delimiter 107-108 demonstrative 13, 53, 60, 127- 128, 130, 133-134, 136-137, 146, 333 denotation 44, 120 deontic modal ix, 243-246, 252-270, 281-282, 284, 286, 290, 294, 296 depth-first x, 182, 212, 331 diachronic 281, 293, 295, 332 dialogue 18, 90-92, 94, 96, 104, 106, 108, 114, 141, 232 direct speech 16, 44, 53 discourse vii-ix, 2-3, 5, 8, 10, 12, 26, 38, 43-44, 47, 49-52, 54, 59-60, 74, 83, 89-113, 119-121, 123-128, 130-134, 136, 138-146, 150, 152153, 155-156, 158-161, 163-169, 171-172, 174, 181-183, 187-188, 197-199, 201-203, 206-208, 211, 213, 215, 217-222, 224, 226-237, 249, 252, 256, 261, 269, 276, 285, 288, 291-292, 295, 299-303, 305-

310, 312-321, 327, 340, 342-349, 355, 357, 360, 367-369 discourse act 2, 5, 8, 10, 44, 47, 54, 93, 97, 103, 188, 201-202, 212, 367-368 discourse grammar 2, 152, 154, 160, 301, 343 discourse particle 2, 60 discourse setting 132, 308 D(iscourse-)Topic 218-220, 222223, 230-232, 235 domain-specific 74 downward layering 7-8, 171, 198, 203, 372 driving force 369 dynamic viii, x, 2, 30, 50, 90, 151, 182, 212, 214-215, 227, 229, 241, 245, 251-252, 254-256, 270, 293, 327, 329-332, 334-335, 345, 356357, 359- 361, 367-369, 375, 377 dynamic modal 245, 252, 254-256, 293 echoic 249, 251, 256-257 embedding 15, 31, 90, 94, 152, 165, 167, 171, 174, 204, 207, 312, 315-316, 321-322, 328, 342, 372 emic 153, 161, 164-165, 168 encapsulation 74 encoding 7, 18, 46, 60, 73, 81-83, 127, 129, 130, 132, 145, 187, 214, 217, 231, 248, 251-252, 259, 261, 267, 296, 326 encyclopaedic 75-76, 141 entities 6, 8, 11-12, 53-54, 58-59, 75, 79, 122-123, 131-132, 135, 138, 141, 151-152, 159, 167, 170-172, 185, 197-198, 212-213, 218-220, 223, 226, 231, 281, 287, 302-303, 312, 318, 326, 331-332, 338, 340, 342, 347, 351, 372-373 episode 241 epistemic modal ix, 59, 125, 244246, 249-252, 254-256, 258-270,

Index of subjects 277-288, 290-293, 295-298, 310, 372 etic 155, 164, 168 event 59, 90-91, 94, 103, 121, 123, 145, 156, 172, 215, 223, 231, 277, 286, 292, 306-308, 313, 320, 341, 367 evidential 187, 269, 277, 282, 284, 286, 290, 293, 295, 298, 314, 372 exchange 5, 96, 106, 108, 237, 258 exclamative 90, 304 existential 123, 159, 230-231, 236 expression vii-viii, x, 2-3, 5-8, 1011, 13, 15-17, 18, 25, 28- 29, 31, 37, 44-49, 51, 53, 56, 60, 73, 75, 77, 83, 91, 100, 103, 108, 110111, 120-122, 127, 131, 133, 135, 137, 140, 142, 144, 153-154, 157158, 162, 164-165, 167, 169-170, 180-187, 189-190, 192-193, 197202, 205, 207, 211-212, 214-215, 220, 224, 227, 234, 262-263, 270, 277, 279-280, 285-288, 290, 293, 295, 301, 306, 310, 312, 319, 325-330, 332, 334-335, 339, 341343, 345-362, 368, 375, 377 expression rule 2-3, 29, 37, 45-47, 53, 62, 74, 165, 169, 182, 184, 190, 192, 199, 211-212, 310, 312, 325-328, 334, 339- 340, 342, 345, 349, 356-358, 361-362, 377 extraclausal xiii, 90, 206, 303, 305 finite 56, 121, 127, 165, 172, 192, 358 focality 125, 131, 138-139, 145, 165, 175, 185, 190, 218, 222, 226, 228-232, 237, 343, 347 Focus viii, 117, 119-121, 124, 127, 130-131, 138, 140, 142-144, 155, 160-161, 164-166, 168-170, 172, 175, 183-189, 191-193, 212, 217218, 221-224, 226-227, 229, 231, 233, 235, 237-238, 240, 318, 322, 335, 363

387

formal feature 28, 334 formalization 26, 29, 31, 39-40, 49, 53-54, 367 formula 98, 190-191, 366 fragment 184, 187, 346 Functional Discourse Grammar (FDG) vii, ix, 2-4, 17, 24, 73-74, 76-77, 79-80, 82-84, 89, 92-93, 95, 99, 103. 106, 117-118, 139141, 148, 179, 197-198, 200-201, 204, 208-209, 212-215, 232, 299, 325-327, 335-336, 338-340, 342, 345-347, 354, 356-357, 365-377 functional feature 231, 333 Functional Lexematic Model 41 functional minimalism ix, 167 Functional Procedural Grammar ix, 181, 212, 275, 288, 295 Fund x, 63, 76, 162, 164, 187, 325 Generalized Parallelism Hypothesis x, 241, 299, 303 Generalized Phrase Structure Grammar 213, 337 generative 27, 31, 37-38, 57, 179, 199, 337, 339 genre 90, 94, 99, 107, 217, 307, 318 Given Topic 219 grammaticalization 76, 206, 293, 298, 349, 358, 360 head 33 Head-Driven Phrase Structure Grammar 213 hedge ix, 206 hierarchy ix, 2-3, 5-6, 36, 38, 43, 48, 73, 77, 91, 93, 95, 102, 106, 109111, 138, 171, 199, 212-213, 221, 234, 272, 281- 285, 288-289, 293, 295, 299-303, 310, 326, 329, 331, 339, 342, 358, 365, 373, 377 holophrase 20, 175, 185, 212, 302 hypotaxis 226, 237 iconicity 52, 204 identifiability 5, 94, 144, 201, 215, 220-222

388

Index of subjects

ideology 94, 107, 110 illocution 5, 11, 18, 43, 46, 91, 99, 116, 153, 164, 171, 186, 222, 287, 307, 318, 347 imperative 90, 259-262, 270-271 increment ix-x, 186-187, 212, 217, 226, 230, 338, 343-344, 355 Incremental Discourse Cognitive Grammar ix, 215 Incremental Functional Grammar viii, 25, 212 indefinite 37, 136-137, 144, 190 indexical 117, 127, 137-138, 146 individual vii, 2, 6, 31, 56, 75- 76, 90, 92, 94, 129, 144, 151-153, 160-161, 173, 182, 289, 294, 331, 358, 367 inference 50, 132, 136, 153, 161, 167, 221, 286, 294, 346 inferrable 124, 230, 235 information viii-ix, 2-3, 5, 8, 12, 15, 29, 35-36, 40, 47-48, 51-53, 7476, 78-84, 90, 98, 102, 104, 109110, 120-122, 124-126, 131, 135137, 139-142, 144-145, 153, 167, 172, 174- 175, 180-181, 185-187, 189, 191, 201-202, 204, 207, 212215, 217-223, 226, 228-231, 233, 235-236, 247, 277-279, 282-283, 285, 289-291, 294- 295, 312-313, 318, 325-326, 329-332, 334-335, 337, 340-344, 346, 348, 354, 356360, 368, 373, 375 inherent modal 245-247, 254 inheritance x, 315, 316, 322, 331 instruction 81, 90, 119, 125, 355 intention vii, 2-3, 5-6, 10-11, 34, 47, 52, 73, 82, 153, 165, 168- 169, 174, 182, 184, 187, 192, 197-198, 201-202, 205, 213, 287, 300, 337, 341, 346-347, 359 interaction 25-26, 30-31, 34, 39, 44, 49-50, 60, 73-74, 76, 79- 80, 90, 109-110, 115, 153, 166, 182, 184-

185, 206, 211, 215, 243, 263, 267, 279-280, 285, 306, 313, 320-321, 326, 340, 357, 366, 370-371 interface 2, 76, 93, 98, 153, 161, 182, 197-199, 204, 358 interjection 107, 109, 300, 353 interpersonal vii-ix, 3, 5, 7-8, 10- 11, 13, 15, 17, 41-45, 47-48, 51-52, 73, 75-77, 82-83, 91, 99, 103, 106, 109-112, 167, 169, 171, 176, 181, 197-198, 200-204, 206-208, 213, 225, 227-228, 232, 243-244, 246- 248, 251, 264, 266-270, 273, 276, 285-286, 294-295, 306, 312314, 318, 325, 347, 358, 367, 369-370, 372-373, 375 interpretation 10, 27, 37, 39, 46- 47, 50, 58, 74-75, 83, 100, 119, 122, 125, 129-130, 136, 143, 151-154, 163, 168, 170, 181, 188, 207, 219, 248-251, 270, 282, 284, 337, 341, 366-369, 373, 375 interrogative 90, 156, 243-244. 248, 250-252, 257-258, 261- 263, 267 intonation 175, 183, 190, 192, 215, 219, 221, 226, 228, 232- 233, 342, 344, 346 inverse 157, 161, 164, 173 knowledge 3, 37, 39-40, 44, 74-76, 78-79, 84, 105, 119, 131, 175, 186, 212, 214-215, 221, 232, 246, 254-255, 266, 282, 284, 287-291, 295-297, 308, 326, 329, 337, 339, 344-345, 357, 359-360, 369 label 110, 228, 235, 290, 293, 309, 333, 339, 377 language user 39, 56, 118, 180, 191, 233, 326, 336-337, 339, 358, 362 layering vii, 40, 43-44, 47-48, 51-52, 59-60, 80-83, 91-92, 108, 111, 197, 203, 206-207, 246-247, 266267, 269-270, 293, 304-308, 314, 317-319, 321, 325, 332, 344

Index of subjects left-to-right ordering x, 10, 17, 169, 171, 176, 182, 186, 212- 213, 329, 337, 358 level viii-x, 2-3, 5-8, 10-13, 15-17, 20, 26, 40, 43-44, 46-47, 56, 73, 76-77, 81-83, 92-93, 99-100, 102103, 105-106, 108, 110-111, 115, 122-123, 127, 129, 133, 139-143, 145- 146, 151, 153-155, 160-161, 163-165, 167-171, 174-176, 180, 183, 185, 190, 197, 200-206, 213215, 227, 233, 235, 243, 263, 265, 268-269, 276, 281-283, 285-286, 288, 295, 300, 305, 308-309, 312314, 316-318, 321, 325-326, 328329, 331, 337, 339, 341, 344, 347, 349, 351-352, 355, 358, 366-370, 372-375, 377 lexeme 12-13, 16, 54, 61, 368 lexical 6, 10-11, 25, 28, 30-31, 3536, 45-46, 55, 57-60, 75-76, 78, 83, 93, 102, 108, 128- 129, 137138, 154, 158-162, 164, 166-167, 173-174, 181, 184-185, 191-192, 199, 211, 213, 233, 278, 280-281, 288, 290, 296, 318-319, 322, 325327, 330, 334, 337, 344, 362 Lexical Functional Grammar 213, 337 lexical priority 334 lexicology 41 lexicon 25, 30, 35, 46, 51, 55, 76, 78, 152-154, 160-162, 164, 166, 167, 171, 174, 176, 184- 185, 187, 190-192, 213, 288, 326, 335, 337, 339-340, 345, 348, 353, 360 linearization 36, 329 LIPOC 51, 180 marker 107, 134, 136, 145, 157- 158, 173, 251, 269-270, 316, 328, 346, 358 memory 74-75, 78-79, 85, 87-88, 212-213, 220-221, 232, 288, 291,

389

331, 337, 339-348, 351, 353-355, 357, 360-361 message 16, 49, 73, 81, 84, 93, 98, 108, 118, 132, 140, 142, 153, 166, 175, 188, 189, 193, 200, 215, 224, 226, 228, 232, 236, 320 message management 132, 140, 142, 153, 166, 175 metalinguistic 13, 16, 78, 326, 340 modality 58, 206, 243, 245, 249-258, 260-261, 264, 266-270, 273, 292293, 297-298, 314, 316, 322, 372 modal auxiliary 243, 249-251, 260, 270, 273, 293 modifier 33 modularity vii, 2-3, 44, 51, 74, 78, 84, 92-93, 153, 197, 212, 239, 243-244, 264, 266-268, 299, 309310, 321, 365, 377 module viii, 36, 44, 49, 93, 97-99, 132, 140, 153, 160, 163, 167-168, 180, 267-269, 305, 310, 312, 321, 326, 331, 339, 346, 365 monitor 356-357, 359 monologue 90, 92 mood 79, 118, 172-173, 200- 201, 206-207, 224, 259-262, 271-272 morphology 18, 46, 74, 131, 163164, 173, 222, 314, 318, 327, 331-332, 335, 337 move 5, 10-11, 13, 43-44, 50, 60, 92, 98-99, 104, 108-109, 136, 153, 183, 197, 201-202, 224, 246, 332, 347, 353 multimedia 95, 105 narrative 2, 99, 126, 307, 318, 321 negation 28, 153, 156, 161, 279, 281 New Focus 223 New Topic 131-132, 219 newsworthiness 155, 160-161, 163164, 166, 169, 172, 235 node 75, 175, 331-335, 348-349, 352-356

390

Index of subjects

noun phrase 25, 28, 35-36, 42, 56, 198, 285, 331 nucleus 35, 57-58, 92, 121, 144, 152, 172, 181, 314, 316, 325 object 32, 61, 142, 178, 183, 224, 236, 374 objective modal 59, 243, 247-252, 256, 269, 272-273, 277, 293, 297 operator 37, 40, 58, 190, 207, 247248, 264-268, 306, 313-316, 319, 332, 347, 353, 360, 365, 372 overgeneration 328 P1 position 5, 9-12, 14-17, 43, 46, 52, 82, 99, 104-108, 112, 124, 145, 166, 170, 175, 189, 200, 212, 219, 226, 231, 328-330, 333, 343, 348-352, 354- 355, 367 paragraph 7-8, 92, 133, 136, 288, 326 parataxis 226, 237 parenthesis 285 participant 49, 108, 121-122, 124, 145, 156, 172, 229, 236, 245, 254-255, 372 pause 344, 346, 351, 353, 355 perception 19, 179, 183, 214, 282283, 291, 345, 370 percolation x, 164, 165, 169, 332, 334-335, 350, 354 performative 15, 78, 200, 203, 205206, 270, 279, 287-288, 290-291, 293-294 perspective 23, 25, 46, 55, 74, 76, 84, 147, 164, 167, 187, 198- 199, 204, 207, 217, 229, 247, 267, 276-277, 282, 292, 299, 323, 326, 338, 341, 343, 358- 359, 372, 378 phonology 7, 31, 46, 60, 78, 103, 212, 222, 226, 236, 327, 330331, 359 pitch 233, 236 placement 189, 212, 226, 304, 329, 334

polarity 224, 247, 277, 279 politeness 90, 106-107, 305, 307, 319, 322 pragmatics 3, 10, 31-32, 35-38, 44, 47-50, 53-54, 57-58, 76, 90, 93, 97-98, 116-117, 119, 124, 131132, 139-142, 145, 153, 156, 160161, 164, 168, 170-171, 174, 181183, 185-187, 211-212, 219, 235, 247, 292, 302, 309-310, 312-313, 318, 321, 325-326, 328, 339- 340, 347-348, 351, 353, 358, 359, 365, 373, 374 predicate vii, 2, 7, 25, 29-32, 35- 36, 40, 43, 57, 58, 60, 76, 124, 151167, 169, 171-174, 182, 185, 187188, 198, 223-224, 262, 276, 281, 284, 315, 325, 332-333, 347, 349, 373-374, 378 predicate frame 2, 29-32, 35-36, 40, 57, 76, 152, 160-162, 164, 171, 182, 374, 378 predicate phrase 7 predication 25, 31-33, 36-37, 40, 4243, 57-60, 120, 122, 124, 126, 131, 152, 154, 159-162, 164-165, 174, 182, 219, 247, 260, 265, 267, 270, 276, 284- 285, 293, 315, 335, 374, 378 prehension 152, 165, 175 preposition 121, 351, 354 presentative 123, 223, 230-231 presupposition 37, 174, 222-223, 231, 235-236 primacy effect 212 process x, 2, 10, 12, 17, 26, 34, 4647, 53, 73, 79-83, 90, 93, 95, 151153, 162-164, 166-170, 174-176, 180, 182, 186, 194, 197, 199, 212-213, 215, 217-218, 221, 232234, 238, 275, 277, 280, 286, 289-290, 295, 300, 314-316, 320321, 325, 329-332, 334, 337, 339,

Index of subjects 342-343, 345, 347-350, 357- 358, 366-367, 373 production viii, x, 2, 7, 17, 34, 46, 50-52, 60, 73-75, 78, 85, 87-88, 151, 170-172, 179-182, 185, 187188, 192, 197, 212- 213, 233, 241, 275-277, 288-289, 292, 295, 327, 331, 337- 338, 343, 359, 367 profile 220-221, 228 prominence 15, 124-125, 184, 212, 217-219, 222, 226-233, 235, 318 pronoun 28, 60, 118, 127-129, 132, 134, 137, 143, 145, 147, 158 properties 6, 27-28, 31, 38, 53, 56, 74, 79, 84, 103, 108, 128, 130, 141-142, 144, 158, 181, 185, 213, 217, 224, 270-271, 276, 278, 280281, 283-284, 286, 302, 309-310, 312, 319, 338-339, 349, 366, 370, 372-373 proposition 58-59, 96, 108, 120, 152, 154, 164-165, 173-174, 182-183, 207, 220, 222-223, 232, 235-236, 246-247, 251, 262, 265, 267, 270, 276-277, 304, 315 prosody 11, 46, 120, 131, 145, 172, 222, 235, 375 protasis 249, 251, 257, 288 psychological adequacy 2, 26, 28, 34, 38-39, 51-52, 86, 170, 179180, 182, 193, 211, 337- 338 qualification 279, 284-291, 293, 296 recency effect 212, 226, 230 recursivity 162, 329, 340 reference ix, 3, 8, 13, 15-16, 25, 4344, 49, 53-55, 91, 108, 117, 133135, 152, 169, 185, 190, 198, 202, 212, 219-221, 229- 230, 234, 236, 276, 280-281, 288, 326, 340, 372373 referent 38, 100, 119-139, 143- 144, 157, 164, 172, 175, 183, 220-221, 224, 231, 235, 347, 373

391

referential 6-8, 10-12, 15-16, 44, 4748, 53-54, 60, 100, 169, 171, 185, 188, 190, 197, 200- 201, 219, 223, 229, 307, 340, 351, 355, 367, 372 referential phrase 7, 11-12, 16, 169 reflexive 8, 127-129, 145, 240, 368 relativism 344 representation vii, 5-7, 10, 15- 17, 25, 28, 37, 39, 43, 45-46, 48, 50, 52, 77, 79, 82-83, 87, 90, 95-102, 109-110, 114, 119, 132, 140, 142, 151, 164-166, 170, 175, 182, 185186, 189-192, 198, 200, 203-204, 213- 214, 238, 267-268, 272, 275278, 280, 282, 286, 290-291, 295, 300, 308-309, 324-329, 331, 333334, 337, 339, 341, 344-345, 347348, 355, 359, 366-370, 372-373 request 80-83, 186, 190 restrictor 152, 165 Resumed Topic 132, 219 rheme 27, 222, 231, 237 rhetorical 43, 51, 92, 101-102, 109110, 270, 308-309, 312- 313 rigid 163, 189, 214 salience 34, 126, 131, 139, 158, 163164, 183, 221-222, 226, 228, 231, 235-236, 307 satellite 42, 92, 165, 172, 181, 253, 306, 333, 350-352, 354-356 schema 76, 183, 214, 220-221, 316 scope ix, 44, 59-60, 84, 90, 92, 94, 101, 158, 171, 174, 187, 197, 204, 207, 217, 226-228, 247, 251, 265, 276, 278, 280- 281, 284-285, 288, 294, 306-307, 319, 326 selection restriction 57, 78 semantics ix, 3, 6, 10-11, 26-32, 3438, 46-48, 50, 54-55, 57-59, 7576, 86, 88, 100, 103, 114, 118, 141-142, 145, 153, 158, 161, 166, 170, 177, 187-189, 200, 212-213, 222, 233, 248-249, 251, 277-281,

392

Index of subjects

283, 285, 292-295, 298, 309-310, 312, 316, 321, 325-326, 337, 339342, 344-346, 348-349, 351-352, 355, 358-361, 365, 368, 373 semantic memory 75, 88, 337, 339342, 344-346, 355, 360 Semiotic Grammar 267, 272 sentence 2, 5, 8, 15, 17, 26-28, 30, 34, 36, 40-41, 44, 48, 51, 56, 58, 91, 100-101, 111, 118- 121, 126, 128, 132, 149, 154, 156-159, 164, 168, 174, 198, 201-204, 206-207, 223-224, 236, 248, 285, 301, 305, 310, 320, 326, 334-335, 337, 343, 348-350, 353 serialization 156, 158-159, 161, 164 short-term memory 79, 232, 288, 331, 340-341, 360 sign 29, 56, 135, 168, 172, 249 speaker vii, ix-x, 2-3, 5-6, 8, 10- 12, 15-16, 34, 44, 46-48, 50, 53, 60, 74, 77, 80, 83, 90, 94, 99, 108, 118-119, 121-122, 127-128, 131, 133-134, 136-140, 143, 145, 155, 174, 180, 183-189, 191-192, 198202, 213-215, 220, 224, 231-232, 234-236, 247, 249-252, 254-259, 261-263, 265-266, 270, 277, 282, 284-289, 291, 294, 295, 325-326, 332, 337-340, 342-343, 346, 351, 355, 357, 359, 361, 367, 369, 372-373 speaker model vi, 193, 237, 325, 338, 340, 357 speech act 40, 43-44, 91-92, 95, 9798, 102, 110, 174, 197, 247, 277, 286-287, 289, 294- 295, 310, 356 speech situation 3, 11, 58, 336, 340, 343 state of affairs 6, 30, 57, 60, 123, 140, 213, 223, 245, 254, 282, 341, 366-367 story 57, 230, 249, 295, 308, 313, 340-341, 346, 361

strategy 39, 53, 128, 130, 138, 143, 155-156, 168, 183, 190, 217, 224, 230, 232, 270, 277, 359 structural 35-36, 38, 111, 118, 155, 161, 204, 221, 223-224, 227, 229, 233, 234, 270, 277, 292, 300, 302-305, 310, 312, 314, 321, 368, 370, 373, 375 style 90, 96, 98, 107, 182, 307-309, 317, 319-320 subact 185-186, 188, 190, 202, 204, 206, 212 subcategorization 358 subject 32, 142, 160-161, 174, 180, 187, 223-224, 230-232, 236, 333, 335, 349, 358, 374 subjective modal 242-244, 247-252, 255-257, 260, 264, 266-268, 270, 277, 310, 314, 317- 318 subsidiary 5, 96, 110, 181, 201- 202 SubTopic 126, 235 syllable 183, 191-192, 375 syntax 27, 31-32, 34-38, 47-48, 50, 54-55, 57, 60, 74, 78, 100, 142, 145, 158, 160-161, 164, 166, 187189, 192, 214, 222, 226, 231, 280, 285, 310, 324-325, 331, 334, 337338, 341, 347, 349, 351, 354, 365, 373, 377, 378 Systemic Functional Grammar 94, 103, 213, 267, 295 Tail 206 template 153-154, 163, 169-170, 188, 190, 265, 329-330, 333, 335, 343, 350 tense 40, 58, 79, 91, 106, 110, 118, 173, 187, 192, 244, 247, 259, 261-263, 265-267, 270, 277, 279, 307, 327, 333 term 32, 35-37, 42, 57-60, 75, 90, 110-111, 119-130, 132-133, 137, 141, 144-145, 152, 155, 162, 166, 172-173, 185, 188, 222, 266, 285,

Index of subjects 288, 300-301, 304, 313-314, 317, 325, 329, 333, 335, 352, 372 text x, 13, 42, 78, 89, 94, 105, 118, 121, 126, 135, 144, 147, 159, 217, 220-221, 224, 229, 234, 236, 269, 300-303, 305, 309-310, 312- 316, 318, 320, 346, 358 theme 27, 249 thetic 120, 122, 139, 143-144, 149, 165, 170, 174, 223 thinking-for-speaking 199 top-down vii, x, 2, 10, 12, 17, 73, 117, 153, 161, 182, 197-199, 212, 243-244, 264, 266-268, 269, 325, 329, 331, 337, 365 Topic viii, 77, 91, 106-108, 110-111, 117, 120-122, 124, 127, 131, 133, 136, 143-144, 158- 159, 217, 229, 303, 346, 369 topicality 91, 120, 122, 124, 138139, 143, 145-146, 186, 218- 219, 222, 308, 343, 347 transaction 82, 96, 106, 108-109, 111, 187 tree 10, 96, 199, 202, 329-332, 335336, 350-352, 354-356, 359 typology 25, 30-31, 56, 59, 162-163, 167, 169, 211, 270, 300, 305, 312, 316-319, 321, 328, 332 undergeneration 327-328 underlying structure vii, 2, 26, 28, 30, 34, 36-37, 39, 45, 47, 49-51, 56, 58, 77, 98, 132, 142, 151-152, 160, 163, 170- 171, 174, 182,

393

214, 277, 303-305, 307-310, 312, 316, 321, 324-329, 331, 333-334, 337, 339, 344, 347-349, 357-360, 365-370 unification 80, 299, 334, 349, 359 upward layering 2-3, 8, 43, 92-93, 96, 197, 299, 310, 325, 348, 372 utterance 2, 6, 27, 44, 54, 90, 94, 9798, 101, 116, 118-122, 124, 126127, 131, 138-140, 143- 144, 150, 166, 169-170, 172, 183, 185, 188, 190-192, 198- 199, 203, 211-212, 223-224, 227, 235-236, 242, 248, 251, 259, 265-267, 280-281, 285288, 290, 295, 300, 324-325, 331, 337, 341-343, 359-360, 370 valency 78, 162, 185, 192 variable 7-8, 30, 40, 43-44, 53, 5960, 152, 236, 270, 344, 347, 349, 355, 366 verbal interaction 25-26, 31, 34, 39, 44, 48, 50, 60, 153, 182, 211, 313, 340, 366, 369-371 viewpoint 58, 153, 189, 199, 229, 234, 240 vocabulary 74-78 voice 94, 104, 110, 180 volition 245, 254-256, 293 West Coast Functionalism 213 word order (see also constituent order) 122, 155, 188, 234, 304, 375 working memory 75, 79, 337, 339340, 342-343, 346-348, 351, 353355, 360