Академический Документы
Профессиональный Документы
Культура Документы
≥
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
Marı́a de los Ángeles Gómez-González
Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.
앪
앝 Printed on acid-free paper which falls within the guidelines
of the ANSI to ensure permanence and durability.
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
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 ex-
pansion 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: dis-
course 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 cov-
ered 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
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 informa-
tion 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
INTERPERSONAL LEVEL
Kees Hengeveld
Mapping rules
REPRESENTATIONAL LEVEL
Expression rules
CONTEXT
COGNITION
COMMUNICATIVE
EXPRESSION LEVEL
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).
(M1: [(A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))
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 enti-
ties by using referring expressions,5 and it is the speaker who ascribes
properties to entities by applying predicates to these referring expressions.6
(Para1: [(S1: [(Cl1: [(PrP1: [(Lex1)] (PrP1)) (RP1: [(Lex2)] (RP1))] (Cl1))] (S1))] (Para1))
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 elabor-
ation in Hengeveld (1997). The correspondences may be listed as follows:
M Move M Move
E Speech Act A Discourse Act
ILL Illocution ILL Illocution
PN Speech Act Participant PN Discourse Act Participant
X Propositional Content p Propositional Content
e State of Affairs e State of Affairs
x Individual x Individual
f Property or Relation f Property or Relation
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 first-
order 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 in-
formation communicated in a discourse act from the nature of the entity
type the description of which is used to transmit that information. As a re-
sult, 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.
Mapping rules
(p1:[(e1:[(f1)(x1)](e1))](p1))
Expression rules
CONTEXT
COGNITION
COMMUNICATIVE
((Para1:[(S1:[(Cl1:[(PrP1:[(Lex1)](PrP1))(RP1:[(Lex2)](RP1))](Cl1))](S1))](Para1)))
The architecture of a FDG
GRAMMAR
9
(2) Damn!
The expressive illocution takes care of the prosodic contour of this one-
word 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 ex-
pression level again:
(3) John!
The same propositional content (p), expressed as a clause (Cl), may oc-
cur 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.
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 referen-
tial phrase:
Ascriptive acts (T) often make use of the description of a zero-order en-
tity (f) and are then expressed by means of a lexeme (Lex), as in the next
example:
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?
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 rep-
resentational 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:
(M1: [IN (A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))
Mapping rules
Kees Hengeveld
COGNITION
Expression rules
COMMUNICATIVE
((Para1:[(S1:[(Cl1:[(PrP1:[(Lex1)](PrP1))(RP1:[(Lex2)](RP1))](Cl1))](S1))](Para1)))
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 sen-
tence, unlike (11a), is an example of a so-called hedged performative
(Fraser 1975), in which the embedded clause represents the actual informa-
tion 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 ac-
tual 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) ...] (C1))] (A1))
(c2)
(CL1)
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):
Sentences like (12), with a prosodically prominent copula, serve the pur-
pose 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) ...] (C1))] (A1))
(r2)
(RP1)
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 referen-
tial phrase (RP).
(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 ref-
erence 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) ...] (C1))] (A1))
(lex1)
(Lex2)
A second case in which reference is to the code rather than to the mes-
sage 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 re-
ports 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 17
(A1: [DECL (P1)Sp (P2)Addr (C1: [... (R1) ...] (C1))] (A1))
(s1)
(S1)
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) What are the restrictions on left-right decisions within the production
process, i.e. what are the systematic restrictions on the internal con-
stitution of the interpersonal level?
(b) What are the restrictions on top-down decisions within the produc-
tion process, i.e. what do the interfaces between the three levels of
grammar look like?
(c) What is the internal structure of the cognitive component and how
does it interact with the three levels of grammar?
(d) What is the internal structure of the contextual component, particu-
larly with respect to the representation of the non-linguistic context,
and how does it interact with the three levels of grammar?
(e) 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 num-
ber 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 open-
minded 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 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 indi-
vidual 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.
2. 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).
3. 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.
4. This slot may alternatively be occupied by a performatively used speech-act
verb.
5. Cf. Lyons (1977: 177): “... the speaker ... invests the expression with refer-
ence by the act of referring”.
6. 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 indi-
viduals 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. Gat-
ward (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 fun-
cional. In: M. Teresa Caneda Cabrera and Javier Pérez Guerra (eds),
Os Estudios Ingleses no Contexto das Novas Tendencias. Vigo: Uni-
versidade 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 Phila-
delphia: 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. Cam-
bridge: 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 21
Vet, Co
1998 The multilayered structure of the utterance: about illocution, modal-
ity 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 for-
mulate 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 contri-
bution 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 work-
ing 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
into five periods. Each FGn refers not only to the period but also to its ca-
nonical FG publication. FG4, however, refers to both the period and
Hengeveld’s Functional Discourse Grammar.
(a) FG0 covers the period prior to 1978, the central work being Dik’s
dissertation (1968) on coordination;
(b) FG1 covers 1978 to 1989, beginning with the publication of Func-
tional Grammar (Dik 1978a);
(c) FG2 covers 1989 to 1997, beginning with the publication of The The-
ory of Functional Grammar (Dik 1989b);
(d) FG3 covers 1997 to September 2000, beginning with the publication
of the two volumes of The Theory of Functional Grammar (Dik
1997a; 1997b);
(e) FG4 extends from the ninth International Conference on Functional
Grammar (Madrid, Spain) onwards, where Hengeveld’s (2000)
Functional Discourse Grammar was first presented.
Total Publications
120
100
80
60
40
20
0
78
80
82
84
86
88
90
92
94
96
98
(a) PR1. The problem of structure – what place, if any, does constituent
structure have in the grammar?
(b) PR2. The problem of underlying representations – what do the un-
derlying representations (URs) actually represent and how should
they be interpreted?
(c) PR3. The problem of verbal interaction – how does FG relate to
communication as process?
(d) PR4. The problem of functional primitives – how are the primitives
defined and applied? How many are needed?
(e) PR5. The problem of discourse – how does FG account for linguistic
phenomena beyond the sentence level?
(f) PR6. The problem of psychological adequacy – what does psycho-
logical adequacy mean and does FG fulfil it?
(g) PR7. The problem of formalization – how should the notation for-
mally and explicitly represent language structure?
(2) ile(s (sdecl (SUBJ (np (npsg (DET(art(the) )+HEAD (nsg (man) ) ) ) )+PRED
(fv (fvintr (fvintr past (fvintr past 3d ps sg(came) ) ) ) ) ) ) )
prop
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 so-
cial 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 Af-
fairs 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):
In this case, ω covers term operators, ωxi defines the universe of poten-
tial 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) }
terms
nuclear predication
extended predication
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
One could ask what the four arrows represent in the picture. Is this a pro-
ductive model that mirrors a Speaker’s process, in which case the arrows
34 Matthew P. Anstey
5.3. Reviews
Having outlined the salient points of FG1, we are now in a position to con-
sider 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 ver-
bal 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): “Objec-
tions of this kind – especially the difficulty of adequately motivating
assignments [of semantic/syntactic functions] – were probably the most vi-
tal 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 solu-
tions to these problems”.
Piťha (1980: 266) concentrates on the problem of underlying represen-
tations (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”.
(a) 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 obvi-
ating the need for an abstract (VP-type) structure in the UR;
(b) semanticization – the internal dependency relations of the compo-
nents of noun phrases are isomorphically instantiated in the term
36 Matthew P. Anstey
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 Gram-
mar 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 ‘reconstruc-
tion’ is incorrect and that there is genuinely no constituent structure in
Functional Grammar. In this case, the success of Dik’s approach in de-
scribing linguistic data raises the obvious question: why does it work at all?
Why does Functional Grammar demonstrate that (a large majority of) syn-
tactic structure is apparently epiphenomenal, a mimicry of semantic
structure? And why can systematic syntactic deviations from this isomor-
phism 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
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
ent(s) of the term in question.” This suggests that neither type of UR pre-
sented 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
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 all-
encompassing 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 ab-
sent (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 con-
cerned 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 interac-
tion, or more precisely, “an integrated model of linguistic interpretation”.
Psychological adequacy reaches its most extreme (1986a: 6): “psychologi-
cal 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 haz-
ardous 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):
Between FG1 and FG2 the most important development is the introduction
of the layered structure for URs. This development was a gradual expan-
sion of the predication structure.25
Dik (1979) presents tense and aspect (τ α) distinctions as ‘predicate op-
erators’; Dik (1981b) is the first to include them in the formal notation:27
Hengeveld (1988) uses the ‘E’ variable for speech acts and the ‘e’ vari-
able for predications (Vet 1986), thereby creating the fourth layer:30
(13) E1: [π4 ILL (S) (A) (π3 X1: [proposition31] (X1))] (E1)
Not only is the clause layered horizontally, it is now also layered verti-
cally, the two layers signifying two different aspects of language: the
interpersonal and the representational. So sentence (14) is represented as
(15):32
(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 arti-
cle “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.
5.5. Summary
We see that FG1 generated much work and faced many challenges. The
occasion arose to produce a new publication to pull these sometimes-
incompatible threads together.33 It is to this publication that I now turn.
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 repre-
sentational) 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
(17) E1: [π4 ILL:σ4 (S) (A) (π3 X1: [predication] (X1): σ3 (X1))] (E1):σ5 (E1)
(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 ar-
gument: 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:
(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 dis-
course organization which are not captured here” (this volume: 5).
44 Matthew P. Anstey
Hengeveld’s 1997 model is still quite different from FG4. We need to re-
view 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 consid-
erations 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 pro-
vided 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’ vari-
able 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 it-
self. 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’ re-
spectively.
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 modu-
lar 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
A newcomer to FG4 who has only read FG2/3 would understandably strug-
gle 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 non-
transformational theory of grammar whereby linguistic expressions are
generated by expression rules operating on underlying representations sig-
nifying 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 over-
shadow 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, criti-
cisms and suggestions in which all linguistic theories find themselves
existing. FG4, which in the light of the survey above presents itself as a di-
rect 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 vari-
ous problems does not entail that its core commitments are indefensible,
only that their implementation has not been perfect.
46 Matthew P. Anstey
Focusing on the problem of structure (PR1), we may concede that the ex-
pression 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 non-
functional 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 ‘syn-
tacticon’ 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 ex-
pression category necessary to successfully transmit his communicative
intentions” (this volume: 10) Yet consider his example (2), here renum-
bered as (22):
(22) Damn!
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 expres-
sion rules to generate the correct linguistic expression. In a word, the
interpersonal layer appears to be representing the pragmatics of communi-
cation, 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 iden-
tify 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 in-
terpretations of the single-layered UR of FG1-3. To illustrate, let us return to
our three example sentences illustrating definiteness, this time with a pos-
sible FG4-like annotation:
Our next concern is PR3, i.e. the problem of verbal interaction. FG4 at-
tempts 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):
(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 as-
sume some Gricean principle of mutual trust whereby each participant is
doing a faithful job in tracking participants, identifying discourse moves, es-
tablishing 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
The obvious problem is timing: step #5 must occur before step #7. Oth-
erwise, 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 [knst] 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. Fur-
thermore, it seems that such linking of underlying representations in a
chain of verbal interaction is fraught with cognitive and notational prob-
lems (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 problem-
atic 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 ten-
dency 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 re-
definition) of the traditional Functional Grammar semantic, syntactic, and
pragmatic primitives into new locations within the model.
FG from its inception 51
principle and exists irrespective of any psychological theory. Dik notes fur-
thermore that LIPOC is explanatory with respect to linguistic data, but
what would explain LIPOC? He writes that “[w]e expect that such [ex-
planatory] 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’ proposi-
tions, 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 psychol-
ogy 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 in-
disputable 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 imitat-
ing 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 to-
tally absent from almost every current linguistic theory. Similarly, with
respect to the three-layered structure, I do not think that Hengeveld can jus-
tify 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 evi-
dence’ 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
In this case the demonstrative that would refer to ‘my poems’. This in-
dicates that the expression rules must select the same pro-form, that, to
represent both types of anaphora. Thus FG4, based on URs as meaning-
bearers, 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 trig-
ger 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
Notes
1. I would like to thank Lachlan Mackenzie, Kees Hengeveld, Dik Bakker,
Chris Butler, and the late Machtelt Bolkestein for their assistance.
2. 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.
3. The official Functional Grammar bibliography (De Groot and Olbertz 1999)
lists 1,141 references for 1978 to 1998. If one includes post-1998 publica-
tions and Dik’s works from 1966–1977, the total is at least 1,300.
4. Many of these problems have been identified by those working in computa-
tional 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 Matthew P. Anstey
12. 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”.
13. 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 se-
mantic 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 selec-
tion 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).
14. For recent critical evaluations of predicate frames in Functional Grammar
see Mairal Usón and Van Valin (2001) and García Velasco and Hengeveld
(2002).
15. 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. … Regret-
fully, I must conclude that this chapter of Dik’s book lags behind the other
ones in scientific importance” (Piťha 1980: 268).
16. This reconstruction is purely hypothetical and other Functional Grammari-
ans undoubtedly view the situation differently. The point is that it is only by
understanding if and how Functional Grammar accounts for syntactic struc-
ture that the theory can be critically evaluated.
17. ‘Syntactic’ in FG is perspectival and not related to constituent structure.
18. Thus, the symbols ‘:’, ‘,’ and and – that is, restriction, apposition, and coor-
dination – represent micro-functional semantic relations inside the term
structure, which in turn enters into macro-functional relations with other
terms and predicates.
19. 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.
20. The hierarchy is simply a succinct way of expressing the oft-quoted position
of Dik (1987b: 7): “… pragmatics is seen as the all-encompassing frame-
work within which semantics and syntax must be studied. Semantics is
regarded as instrumental with respect to pragmatics, and syntax as instru-
mental with respect to semantics”.
58 Matthew P. Anstey
21. 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”.
22. “In order to be able to generalize over all definite terms, however, we as-
sume 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).
23. 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, con-
struct an interpretation in which R2 is applied to a”.
24. 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 inter-
pretation 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.
25. 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).
26. 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.
27. 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.
28. Hengeveld joined the FG discussion group in 1985. Here he first encoun-
tered Foley and Van Valin (1984) when it was reviewed by Machtelt
Bolkestein.
29. Hengeveld (1986, 1987) calls X a predication, but it later is correctly rela-
belled 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 59
presented in Weijdema et al. (1982), but Dik introduced it first in his 1985
FG lecture notes.
30. 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:
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 ex-
plains 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 argu-
ment.
31. Hengeveld (1990a: 4) subsequently corrects his terminology to predication
in this notation since the proposition is actually π3 X1: […] (X1). Note, how-
ever, that in the same volume, Hengeveld (1990: 107) again incorrectly uses
proposition.
32. 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 grammati-
calized semantic categories. This practice possibly diminishes the
typological strength of the notation.
33. 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 re-
garding term operators; and Dik (1987) provides a typology of entities using
Bunt’s (1985) work.
34. See Keizer (1992a) for an excellent discussion of other more subtle differ-
ences, particularly Dik’s unexplained introduction of the ‘f’ variable for
zero-order entities.
35. 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 Matthew P. Anstey
36. Hengeveld (1990) suggests that all five satellites are restrictive satellites,
which accounts for the ‘:’ preceding each one. However, in the same publi-
cation, 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 re-
strictive or non-restrictive.
37. 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 predi-
cate and core predication. Similarly, the σ4 binds to ILL as there is not yet
an ‘F’ variable in this model.
38. 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 repre-
senting 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 parti-
cles and other clause combiners. Incidentally, the term ‘semanticon’
actually comes from Dik (1973b).
39. ‘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 com-
ponent of Functional Grammar.
40. 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 unproblem-
atic. 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 state-
ment/move made by the first speaker’s utterance).
41. The numbers represent a possible order of production. ‘CC’ stands for
‘communicative context’, and ‘Ph1’ for ‘phrase one’. The fifth line repre-
sents 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 anaph-
ora and discourse deixis are both referential.
42. 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 61
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 Kristof-
fersen (eds), 329–354.
1997 Cosubordination. Working Papers in Functional Grammar 63. Am-
sterdam: 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 Ja-
kobsen 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. Lach-
lan 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 informa-
tion. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214.
Bolkestein, A. Machtelt, Henk A.Combé, Simon C. Dik, Casper de Groot, Jadran-
ka 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: Cam-
bridge 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. Vis-
mans (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 Hori-
zons 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 Mil-
waukee Syntax Conference, 29–33. Indiana University Linguistics
Club.
Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gat-
ward (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 Introduc-
tion 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, Methodol-
ogy, and Philosophy of Science IV: Proceedings of the Fourth Inter-
national 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
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 disser-
tation, 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 Pa-
pers 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
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 Mathemati-
cal 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. Doc-
toral 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 Dis-
tinction 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 Uni-
versity 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 Universi-
teit Nijmegen, The Netherlands.
FG from its inception 71
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 represen-
tations. 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 en-
coding. 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.
1976); and the permanent store of our knowledge would correspond to the
second kind. A further distinction within long-term memory, i.e. that be-
tween 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 lev-
els.
Long-term knowledge makes itself very evident in vocabulary interpre-
tation, when we resort to all the encyclopaedic information stored in
individual lexical items. Descriptions of how this encyclopaedic informa-
tion 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 cog-
nitive linguist, suggests that lexical units are merely points of access, and
that therefore “concepts are simply entrenched cognitive routines”. More
precisely:
Langacker does not suggest this description out of the blue. It is based
on proposals by psychologists and computer scientists for models that in-
volve 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 re-
trieve 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 acti-
vation. The various modes of activation for concepts in a network explain
the link between words and encyclopaedic knowledge dynamically.
76 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 dif-
ferences 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 in-
terlocutors is part of the communicative settings. General knowledge about
different kinds of communicative settings (as well as of different text ‘gen-
res’ 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 long-
term information stored in the cognitive component.
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 struc-
ture 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 cer-
tain 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 com-
ponent, 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 nei-
ther see nor influence its contents.
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
Application
layer
REQUEST ACKNOWLEDGEMENT+ROUTINE
Presentation
layer
Session
layer
… … …
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 multilay-
ered 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 en-
coder 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
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 su-
perior. When he meets him, he says:
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 dif-
ferent 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 appar-
ently 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:
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 con-
cretely, 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 com-
municative context. At all stages, of course, the receiver has needed to
consult the long-term information stored in the cognitive component in or-
der 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. dis-
course 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 land-
marks 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 oth-
erwise, 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 compo-
nent. 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 dif-
ferent 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. Note that I am referring to modularity of the mind and not to modularity
within the FG model.
2. 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.
3. I give some examples in Inchaurralde (2000: 98–100).
Cognition and FDG 85
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 Linguis-
tics 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.), Lan-
guage 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
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 lan-
guages 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 repre-
sent discourse. Hence, the present paper will be concerned with a
discussion of this question.
grammars of natural languages, unlike the speech act categories of, for ex-
ample, 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 de-
fault values of the individual clauses within the stretch of language
concerned. Dealing with illocutions in terms of discourse also involves dis-
tinguishing between (i) the illocution as intended by S, written ILLS, (ii)
the illocution as coded in the expression, written ILLE, and (iii) the illocu-
tion 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 con-
nected 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 con-
cerned 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 dis-
course 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 interper-
sonal 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
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 ac-
commodate 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.
(1) ILL(DISCOURSE-EPISODE)
A: Utterance act
move
Utterance act
exchange
Kroon also, where appropriate, labels acts and/or moves as either central or
subsidiary.
Hengeveld (1997) employs a means of representing discourse which re-
flects 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 struc-
ture. 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 se-
quence.
In positing a non-linguistic part of the transaction in addition to the lin-
guistic 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 representa-
tion is illustrated in the outline example in Figure 2, which is based on a
stretch of discourse where two propositions occur in sequence, and propo-
sition 2 serves to elaborate proposition 1.
The question of discourse representation 97
RELATION elaboration
NUCLEUS proposition 1
SATELLITE proposition 2
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)
E1 E2
[L] [L]
m' 1 m' 2
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)))], Precondition
[WANT (S) (KNOW (A) (C(m1, m2)))], Precondition
(6) (M1: [(A1: [ILL (P1)S (P2)A (C1: [...(T1) (R1)...] (C1))] (A1))] (M1))
(8) (Para1: [(S1: [(Cl1: [(PrP1: [(Lex1)] (PrP1)) (RP1: [(Lex2)] (RP1))] (Cl1))]
(S1))] (Para1))
Once more, allowance is also made for higher layers, for example Sec-
tion and Chapter in books.
x y
Jay(x)
Kay(y)
x likes y
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.
k1 := x, s1, y k2 := u, s2, v
Jay(x) u=x
s1-like(x,y) s2-date(u,v)
Kay(y) 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.
(11) a. The identity of the participants, their attitudes, statuses and back-
grounds;
b. Any other sources of voices which blend into the discourse;
c. Bystanders;
d. The non-linguistic properties of the medium or media of communica-
tion employed;
e. The non-verbal communication;
f. The time and place;
g. The setting;
h. The prevailing cultural, social and institutional conventions governing
communication;
i. The referents of terms used in the discourse;
j. 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:
(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 involve-
ments
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)
lae such as ‘AS PER p171’ are employed in order to avoid having to dupli-
cate, at the interpersonal level, information that needs in any case to be
stated at the representational level.
(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))
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
5. Conclusion
In this chapter a framework for the representation of discourse has been
proposed which is intended to accommodate the main phenomena of inter-
est 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. Un-
doubtedly 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 Or-
ganisational 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 indi-
viduals 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. Gat-
ward (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 So-
cial 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 Phila-
delphia: 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: Rout-
ledge.
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
1. Introduction
My goal in this chapter is to compare two functionalist theories of indexi-
cal 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 Ulti-
mately, 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 Colum-
bia 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) as-
signment 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 as-
signments of Topic and Focus functions which I shall be discussing clearly
fall within the Interpersonal Level recognized within this format – the na-
ture of these pragmatic functions arguing strongly in favour of the top-
down relationship between the three descending levels of (respectively) In-
terpersonal, Representational and Expression, which structure the new
model.
A brief word, first of all, on the Columbia School (CS), which is rela-
tively little known as such in Europe (even though the work of several CS
118 Francis Cornish
2.1. Focus
The CS conception of Focus should not be confused with the sense of Fo-
cus 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 ‘Fo-
cus’ 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 informa-
tion2 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 knowl-
edge)’, 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 fore-
front 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 (ar-
guments) 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 – al-
though 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 IN-
FOCUS 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 under-
stander 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 subse-
quent 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
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 corres-
pond 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):
ent and the proposition; the entire utterance is thetic (presented as ‘all-new’
information to the addressee) – an event-reporting utterance in Lam-
brecht’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):
The system of Focus deals with the centering of attention on one of the par-
ticipants 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])
John bought Mary a book. The predicator is not involved in the CS FO-
CUS 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 non-
nominative clitics signal LOWER-FOCUS.
As already mentioned, where there is only one participant, in an intran-
sitive 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 particu-
lar 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 interpreta-
tion 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 refer-
ents but to establish the event”; see also on this point Hannay 1991: 146).
In fact, (4b) involves not just inversion, but also so-called there-
insertion,9 a specifically ‘existential’ construction. In this connection, Bir-
ner 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 famil-
iar/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 exis-
tence 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 confer-
ring 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 macro-
topic of the segment in which (6a), (5b) and (6b) occur is ‘the city of Anti-
bes’.
(5) 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 châ-
teau.(Extract from The Holiday Which? Guide to France by A.Ruck
1982: 112)
(6) 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 subject-
predicate 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 con-
trast. 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 Anti-
bes’), and is the reason why the prepositional phrase is preposed in this
context, connecting directly in this way with its immediate discourse con-
text. 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 be-
tween 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 adjecti-
val 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 the-
ory, 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 in-
terest’), 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. ar-
gument, 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 prever-
bal 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 information-
structure 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
(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 )
(8) a. A:....of greater interest is the Picasso museum in the old Grimaldi châ-
teau.
b. B: #That’s not true! Of no interest at all is that museum.
In this regard, Birner and Ward (1998: 156–194) argue that the function
of the preverbal constituent in inversion constructions is to link the utter-
ance 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 post-
verbal 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). More-
over, 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 conti-
nuity 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 order-
ing.
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 Gri-
maldi château in (5b), the cathedral in (6a) would receive the value IN-
FOCUS, since it is the preverbal subject of a transitive, and not intransi-
tive, 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 part-
whole relation between the referents of the two preverbal subject expres-
sions 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. Gold-
ing'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, postver-
bal 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 sub-
stance (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 con-
centration 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 expression-
FC] on the deictic scale used to refer to it.
The deictic scale which Diver (n.d.: 11) gives in this respect is formu-
lated 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 nor-
mally 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)
HIGHEST
ipse
hic
Deixis iste
ille
is
se
LOWEST
(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 dis-
course 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 reflex-
ive pronouns in languages such as Latin, French and Spanish (but not in
English, where reflexive pronouns are not clitic and so have different dis-
tributional 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.’
(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 coun-
try 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 neces-
sary 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 theo-
ries, regarding the way in which each views the discourse function
associated with the expression-types at issue. For FG (and also other cogni-
tively-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 as-
suming 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 dis-
course 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 Swa-
hili and Japanese examples which they present, discourse referents are not
simply designated ‘decrementally’ with ever-decreasing levels of DEIXIS
134 Francis Cornish
this is used in subsequent (instead of, or even as well as, initial) reference
to a given entity:
(16)
L 30
Caller: You spoke a little while ago to a man.hhh with regard to uh
law,//an’ murder.
Crandall: Yes ma’am...
L 70
Caller: ...you were talking about, or was he anyway,//that man
Crandall: 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.
Caller: Now/
Caller: Yuh know/
L 76
Caller: By the d- conversation that chu had with this man: my im-
pression 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-
Crandall: [In a- in a- in a murder case⎯
Caller: [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 news-
paper article) is given under (17):
(17) [Beginning of article “God sends a sign to the heathen in a Chevy”, The In-
dependent, 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 gi-
gantic 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 some-
where on Interstate 81 in southern Virginia at the weekend, is the latest in
my growing collection of “God” adverts...
(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)
(20) All of a sudden we saw A GIGANTIC SHARK. (Dik 1997a: 312, ex. (6))
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 utter-
ances in discourse, in addition to its purely topic-based account. It could
usefully take cognizance of the fundamental FG distinction between topi-
cality 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 At-
tention 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 ad-
dressee’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’ interpre-
tation) indicates how to retrieve various sorts of discourse referents, given
their already-existing FOCUS level; and (b) where the morpheme signal-
ling 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 con-
cerns 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 Com-
pletive Focus), thereby simplifying the system as a whole.
Notes
1. This chapter is a revised and expanded version of a paper presented at the
9th International Conference on Functional Grammar, held at the Universi-
dad 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.
2. 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 topic-
comment 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 argu-
ments (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.
3. 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 uni-
dentifiable 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.
4. Lambrecht and Polinsky (1997) present a range of distinctive properties
characterizing thetic utterances cross-linguistically.
5. 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) appli-
cation of the system to two fictional narratives.
6. I am indebted to Wallis Reid for this characterization of the CS FOCUS
system.
7. For good, fairly recent discussions of the thetic vs. categorical distinction,
see Rosengren (1997), and Lambrecht and Polinsky (1997).
Focus of attention in discourse 145
8. 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’.
9. See in particular Hannay (1985a) for an FG account of this construction.
10. In English, of course, preverbal subjects may also be focal, since (given an
appropriate context) they may receive a high pitch-accent, and thereby con-
vey new information relative to that context (A: Who ate the fish? B: JANE
ate the fish/did).
11. 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.
12. Strictly speaking, Huffman as a CS linguist would not use the traditional
terms ‘subject’ and ‘verb’ here, preferring the descriptive terms ‘(first) par-
ticipant’ and ‘event’, respectively. However, I use the more traditional
terms here for ease of comparison with the IS conception.
13. 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) FO-
CUS, 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 ‘re-
mention’ 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 sub-
ject function with respect to the verb, and in relation to the latter’s
inflectional form. In FG, predications containing a reflexive pronoun, in re-
lation 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 in-
herently pronominal verbs in French and other such languages). Hence the
reflexive (clitic in French) pronoun does not correspond to an argument se-
mantically, and cannot therefore be assigned any value pertaining to
argument expressions. See García (1977) for an excellent analysis of Span-
ish se within the CS framework.
14. The hyphen preceding given pronoun forms indicates that the clitic variant
is what is intended.
15. 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 clause-
bound level of pragmatic organization distinct from the semantic and syn-
tactic levels of clause structure”.
146 Francis Cornish
16. That is, the selection of given constituents as bearing Topic or Focus func-
tion 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.
17. See Hannay (1985b) on SubTopics.
18. Siewierska (1991: §§6.3.2–6.3.3) and Gómez-González (2001: §§5.2.1-
5.2.4) present critical surveys of the four types of Topic function within
standard FG.
19. That is, where a ‘HIGH DEIXIS’ marking might be used to introduce a ref-
erent, a ‘MID DEIXIS’ one to establish it as a topical discourse referent in
the current stretch of discourse, and subsequently a ‘LOW DEIXIS’ mark-
ing to maintain its current topicality (as long as it is not superseded by a
new referent bearing a higher claim to topic-worthiness).
20. 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.
21. 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).
22. 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.
23. 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). Gómez-
González (2001: §5.2.4) provides a survey of published criticisms of the al-
leged topical status of so-called ‘NewTops’.
24. See Cornish (2002: §4) on this particular problem, presented in terms of
French examples where subject and (intransitive) verb are inverted follow-
ing a clause-initial locative satellite, and where it is the predicate corres-
ponding 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 (ex-
ample: A l’horizon, un orage couve ‘On the horizon, a storm is brewing’).
Focus of attention in discourse 147
References
Aoyama, Takashi
1995 Deixis and value: a semantic analysis of the Japanese demonstra-
tives. 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. Am-
sterdam and Philadelphia: Benjamins.
Bolkestein, A. Machtelt
1998 What to do with Topic and Focus? Evaluating pragmatic informa-
tion. 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: Benja-
mins.
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 (con-
structive) critique. Working Paper in Functional Grammar 73.
Amsterdam: University of Amsterdam.
2001 “Modal” that as determiner and pronoun: The primacy of the cogni-
tive-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. Un-
published 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 Func-
tional Grammar of English. Journal of Pragmatics 16: 131–155.
Hannay, Mike and A. Machtelt Bolkestein (eds)
1998 Functional Grammar and Verbal Interaction. Amsterdam and Phila-
delphia: 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 Gram-
mar. Talk given at Rutgers University, 10 October 1993.
1997 Categories of Grammar: French lui and le, Amsterdam and Phila-
delphia: 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: Cam-
bridge 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: Ma-
sayoshi Shibatani and Sandra Thompson (eds), Essays in Semantics
and Pragmatics in Honor of Charles J. Fillmore, 247–278. Amster-
dam and Philadelphia: Benjamins.
Reid, Wallis
1991 Verb and Noun Number in English: A functional explanation. Lon-
don and New York: Longman.
Reinhart, Tanya
1981 Pragmatics and linguistics: An analysis of sentence Topics. Philoso-
phica 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 Lin-
guistic 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 Chi-
cago Press.
Vet, Co
1998 The multilayered structure of the utterance: about illocution, modal-
ity, and discourse moves. In: Mike Hannay and A. Machtelt Bolke-
stein (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 comprehen-
sion 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), representa-
tions of this kind, with their variables, quantifiers and open predicates, are
frankly not obvious candidates for psychological ‘reality’. Strawson, in de-
scribing the cognitive origin of ‘perspicacious grammar’, specifically
warns against introducing variables and quantifiers as opposed to individ-
ual particulars at the basis level of the logical representation of
propositions (Strawson 1974: 117). That is because it is individual particu-
lars (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 rele-
vant ‘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 re-
sponse 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 no-
tation is innocuous and does indeed serve a useful purpose, namely in
relating underlying clause structure to the lexicon, where individual predi-
cates (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 rele-
vance 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 inter-
pretable 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 em-
bedded 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 re-
flected 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 com-
plex intentions (perlocutions, etc.) find their way to expression via the
available linguistic code. The upward limit from this viewpoint would cor-
respond 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 generaliza-
tions across the grammars of many languages. The obvious direction to
move in describing a process interpretation of FG is thus top-down (com-
pare also Hengeveld this volume).
Now a cornerstone of FG is the notion that pragmatics contains seman-
tics, which in turn contains morphosyntax (as in Dik 1989: 7). So if
pragmatics really embraces all levels of grammar it should be seen to ex-
tend 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 infor-
mation, etc.).4 Note the difference from my proposal, however: Hannay
sees his message management module or component as an interface be-
tween 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
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:
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.
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 (Naka-
yama 1997: 99ff.).
(15) ča:pac-i:
canoe-make
‘He made a canoe’
(16) u-i: čapac
nice-make canoe
‘He made a nice canoe’
(17) 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 en-
clitics 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
(18) quac-’aq
person-inside
‘There is a person inside’
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 pro-
totypical ‘verbal’ morpheme) generally has to be nominalized in order to
function as a nominal (e.g. by addition of the definite article, as above).
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 ex-
tremes 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 morphosyn-
tactically 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 con-
trolled 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 discourse-
determined choice of initial predicate can be accommodated. Short of put-
ting 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 intro-
duce 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
(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))
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-plus-
restrictor (‘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 White-
headian 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 impor-
tantly, ‘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 be-
tween languages with a tighter or looser linkage of lexicon and grammar
(and between languages whose grammar reflects greater or lesser dis-
course-pragmatic vs. lexical control of grammar), one surely cannot expect
there to be a universally valid division of labour between a grammar mod-
ule 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 interper-
sonal level choices much more directly (e.g. ascriptive act decisions as to
what is to be the main predicate) than in more familiar European lan-
guages. 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 infor-
mation packets is indeed reminiscent of Nootka.
What in effect I have been arguing for in this chapter is a kind of func-
tionalist minimalism, parallel to the Minimalist Program of the generativists
168 Michael Fortescue
comes out in the wash’ in the expression rules anyway, as long as the rele-
vant 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 dif-
ferent 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 ut-
terance 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 de-
finitive answer to the question of ‘where pragmatics belongs’ in the FG
model. But perhaps we should not be so fixated upon the interrelation be-
tween grammar and discourse after all, but think rather in terms of Pattern
versus Process, a distinction that cuts across this divide, since Pattern in-
cludes 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 ac-
commodating 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-to-
right’ 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 comple-
mentarity 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 al-
ready 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 elabo-
rating 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. 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 proce-
dural 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 sen-
tences, whereas Nuyts (1992) argues that deeper cognitive representations
are needed that cover both (see Anstey this volume: 38ff).
2. Harder’s ‘process vs. product’ dichotomy is not quite the same as my ‘proc-
ess 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’ (ex-
cept in the historical sense), but rather a high-level abstraction across norms
of (communicative) behaviour.
3. 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.
4. Hannay proposes several basic types of utterance ‘mode’ at this level, which
he labels TOPIC, ALL NEW, REACTION, NEUTRAL and PRESENTA-
TIVE. 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 Fo-
cus 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 171
Compare even in English the man I saw at the station yesterday’s face, with
wide phrasal scope of genitive -s.
7. ‘TEL(IC)’ is a ubiquitous affix that, according to Jacobsen (1993: 250), sig-
nals ‘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 Michael Fortescue
which the second predicate shares any clausal mood marking of the preced-
ing one. Nakayama prefers not to use the term ‘finite’, since it smacks of
morphosyntactic finiteness of form (pers. comm.). He describes it as “mak-
ing 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.
8. But Focus in the sense of non-contrastive ‘newsworthiness’, as in the sen-
tences above, can also be assigned to (at least) object terms, in which case
they can be put in initial position, without these special relational predi-
cates, as in:
i. c’aak 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 newswor-
thy’ 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 (Engberg-
Pedersen 1991: 64).
9. Despite the general polysynthetic character of this language, Nootka word-
forms 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 preposi-
tion-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 (al-
though 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 psycho-
linguists like Slobin – and in terms of cosubordinate clause chaining within
Role and Reference Grammar). Some languages containing ‘serial verb’
Process and pattern interpretations 173
i. imiq nillir-sima-nirar-paa
water be.cold-PERF-say.that-3SG/3SG.INDIC
‘He said that the water had been cold’
12. Engberg-Pedersen points out to me that sentences such as these are com-
pletely parallel to similar utterances in Danish Sign Language (pers.
comm.).
13. 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 proc-
esses 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.
14. Deciding what is to be the main predicate given a complex communica-
tional 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 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 in-
put 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 prag-
matic 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 prehen-
sional structure at the deepest (initial) level.
15. 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.).
16. The predicate-based formalism of the original model does need to be ex-
tended, 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 pro-
posal for predicate variables (Hengeveld 1992: 31ff.).
17. 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 assign-
ment level of the traditional model.
18. 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 af-
fect as is achieved by Hannay’s discourse management mode of that name.
19. The scope of that focus is in part determined by the presence or absence of
the TEL suffix mentioned above.
20. 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 175
from language to language than we often imagine. The reason Nootka ap-
pears 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 intona-
tion 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 histori-
cal development where P1 became gradually freed up for other items than
those under Focus (but with the initial situation remaining in adult holo-
phrases), 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 ‘polysyn-
thetic’). 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 Fo-
cus often adheres to final position in the clause (except in its contrastive
focal construction).
21. Note that I am discussing a level of representation corresponding to (poten-
tially) conscious experience. This I see as essentially a matter of inferential
processes (whether on-line or automatized), both lexically-cued and gram-
mar-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 actu-
ally consists of declarative rather than procedural knowledge (it is a
situational network consisting of representations, i.e. – like the FG compo-
nent– 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 concres-
cences and describable in prehensional terms (in this case involving an
indicative prehension).
22. 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
References
Bolkestein, A. Machtelt
1998 What to do with Topic and Focus? Evaluating pragmatic informa-
tion. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214.
Chafe, Wallace
1994 Discourse, Consciousness, and Time. Chicago: University of Chi-
cago 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 Lan-
guage 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 To-
tal Communikation.
Fortescue, Michael
1992 Aspect and superaspect in Koyukon: An application of the Func-
tional Grammar model to a polysynthetic language. In: Michael
Fortescue, Peter Harder, and Lars Kristoffersen (eds), 99–141.
Process and pattern interpretations 177
1998 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. Am-
sterdam and Philadelphia: Benjamins.
Givón, Talmy
1995 Functionalism and Grammar. Amsterdam and Philadelphia: Benja-
mins.
Hannay, Mike
1991 Pragmatic function assignment and word order variation in a func-
tional grammar of English. Journal of Pragmatics 16: 131–155.
Hannay, Mike and A. Machtelt Bolkestein (eds)
1998 Functional Grammar and Verbal Interaction. Amsterdam and Phila-
delphia: Benjamins.
Harder, Peter
1992 Semantic content and structure in Functional Grammar. On the se-
mantics 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 Michael Fortescue
1998 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 Philadel-
phia: 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) mor-
phosyntax. 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 Linguis-
tics 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 linguis-
tic 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 produc-
tion. Yet the demands of psychological adequacy have encouraged
speculation on the relation between the model, in its various manifesta-
tions, 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) one can deny any relationship, insulating grammar from psycholin-
guistic findings (the traditional generativist position);
(b) one can maintain that processing mechanisms can ‘consult’ or ‘in-
voke’ 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.
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 expla-
nations, 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 infor-
mation is presented in chunks of increasing internal complexity” (Dik
1978: 212). We must beware, however, of assuming that the speaker is ac-
tually choosing to alleviate the hearer’s interpretive task: language users do
not consciously apply linguistic principles in their effort to express them-
selves. 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 anti-
functional 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.
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 commu-
nication: 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.
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
Acts come in two basic types. Simplex acts (like A1) consist only of
their Focus. (Hengeveld this volume: 11 provides the example of the ex-
clamation 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 sim-
plex 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:
Hengeveld (this volume) points out that the mapping relation between
the interactional and representational components is not one-to-one. Al-
though 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 con-
tent C, we find the representation … (T1) (R1) …, which is explained as
involving ‘the speaker who ascribes properties to entities’. If the interac-
tional 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 ele-
ment 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
(5) (M1: (express A1: SHOCK)Foc, (assert A2: (SA2:1: (T: FIRE)Foc), (SA2:2: (R:
<unspecified, nonhuman>)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 ac-
tion. 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 in-
formation 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 proc-
esses 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 con-
ceptualization, 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.
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 simulta-
neously satisfy all these constraints”.
If the representational component is seen as a set of constraints on proc-
essing, it is no longer necessary to assume that every utterance has a full
propositional form. The grammar need simply be consulted for what is nec-
essary 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 proposi-
tional 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 syn-
tactic 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 com-
bination 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 advan-
tage 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. In-
cremental processing in the production of utterances works on the
assumption that the concepts that lie behind a complex act are not all avail-
able at once. Speakers characteristically do not wait until all the concepts
are “in” before starting to speak. The various levels of activity, conceptual-
izing, formulating and articulating, as Levelt (1989) has it, work in parallel,
with conceptualization having a head start over formulating, and formulat-
ing over articulating. However, speakers are not free to express their ideas
exactly as they occur to them: the declarative grammar, with its conven-
tional 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 or-
der, 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 ad-
dressee. 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 in-
formation, and the selection of an appropriate lexicalization of that concept,
both require time. By contrast, associated information that is already acti-
vated 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 depend-
ent 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 po-
sition. 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!
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 interpre-
tive 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 se-
lected 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:
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 repre-
sentational component, into an acceptable template:
This example is a good instance of how any one utterance can contain a
FDG and language production 191
(10) If you could possibly see your way clear to providing me with some 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:
which would yield the utterance (13), a highly unlikely utterance in the cir-
cumstances at issue here:
(13) EAT?
(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:
(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 declara-
tive representational component:
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 se-
quences 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 proc-
esses 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 constrain-
ing 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 im-
plicitness 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. Gat-
ward (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 J. Lachlan Mackenzie
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 rela-
tions 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 non-
grammatical entity, the move, defined in strictly interactional terms as cor-
responding to a single communicative intention; after that stage, however,
all choices reflect coded choices. After the top interpersonal level, there-
fore, all the purely code-free communicative choices have been made.
The representational level is conceived as consisting of entities of dif-
ferent orders which are described by the clause. Propositions (third-order
entities) are the top layer; speech acts are now above the scope of the rep-
resentational 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 consist-
ing 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.
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.
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 con-
stitutes the content of a referential act made on the interpersonal level:
This illustrates the advantages of the three simultaneous levels in FDG: the
interactive choice made by the speaker simultaneously affects the morpho-
syntactic 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 tra-
ditional grammar point of view, the difference is simply a difference in
terms of mood, located in a paradigmatic slot in the Spanish verb; and tra-
ditional 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
[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 small-
est identifiable units of communicative behaviour’. A move consists of one
central act, which may be supported by one or more subsidiary acts...In or-
der 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).
A: ‘Express emotion’
B. ‘Hedged performative’:
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 rela-
tionship between grammar and discourse categories in such cases (cf.
Bolinger 1968).
The absence of that is also significant: that can be understood as an op-
erator which (like ‘declarative’) takes scope over the entire proposition (cf.
Harder 1995), reifying it into an entity (while ‘declarative’ uses it to con-
vey 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):
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 analogi-
cally 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 differ-
ent 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):
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 pre-
cisely to individuate the discourse categories that I would like to place in a
more spectacular position in the interpersonal tier; pending that, it will al-
ways 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 inter-
action. 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 be-
tween discourse elements. In cases where discourse and clausal organiza-
tion 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 be-
tween 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. Cana-
dian Journal of Linguistics 14: 3-30.
Comment clauses and FDG 209
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 ade-
quacy 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 mod-
els of linguistic competence and linguistic behaviour (Dik 1997: 13, 17).
To better suit these demands, in particular to span the perceived gap be-
tween 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 prepho-
netic strings through the operation of expression rules (Dik 1997: 58ff.;
Hengeveld 1997). Crucially, in this standard model conceptual unity is en-
tailed 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 inc-
ommensurable, and accordingly assign corresponding modules to
grammar, which is static and out-of-time, on the one hand, and to dis-
course, 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
(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 María de los Ángeles Gómez-González
(2) a. ⏐ there were \three of us in the \car ⏐ \all °rather \nervous ⏐ (LIBMSE-
CAPT04: 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.
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) topi-
cal 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 presen-
tation 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, temporar-
ily 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 impor-
tantly:
(a) 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;
(b) 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;
(c) the difficulties involved in identifying or distinguishing the four
Topic types.9
... 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 conversa-
tion, all of the others are automatically made available. (Fillmore 1982:
111)
(a) active, designating “currently lit up” information that is, as Chafe
has it, “in a person's focus of consciousness at a particular moment”;
(b) semiactive (or accessible), designating information that is in a per-
son’s peripheral consciousness or in their background awareness as a
result of: (i) deactivation from an earlier state (i.e. textual accessibil-
ity); (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).
(c) unused, designating inactive information that is “currently in a per-
son's long-term memory, neither focally nor peripherally active”.
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 infor-
mation 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 dimen-
sions 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 pack-
aging but with other linguistic dimensions as well (e.g. illocution, string-
based deaccenting) and which may also be a consequence of non-linguistic
effects. As indeed stated in FG, Focus accent is taken to represent the pro-
sodic means whereby – in combination with morphosyntactic devices such
as morphological markers, word-order variation and special lexico-
grammatical constructs – the syntactic domain of Focus (the Focus do-
main) 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 ‘presupposi-
tional’), 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 be-
tween 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 subdi-
vision within NewFoc should be also posited in order to account for the
structural and attentional domains of two construction-types that are recur-
rent across languages, namely utterance-Focus (UttFoc) and predicate-
Focus (PredFoc) (cf. Lambrecht 1994; Vallduví and Engdahl 1996; Lam-
brecht and Polinsky 1997).
In a UttFoc construction (e.g. (3) below) both the Subject and the predi-
cate 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 presup-
position attached to either the Subject or the Predicate, and resultingly the
identity of assertion and Focus (cf. Kuno’s 1972 ‘neutral description’; Al-
lerton and Cruttenden’s 1979 ‘all-new utterance’; and Kuroda’s 1972
‘thetic sentence’). UttFoc has been attributed to:
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]).
(who killed Cock Robin?, 1994: 48). All other thematic choices will be
marked – e.g. (1a), (2b) above in contrast with the other unmarked coun-
terparts.
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, un-
der 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:
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 (hous-
ing 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 orienta-
tional relationship as a result of the recency effect: discourse scaffolds
normally ground what is to be said on the previous context or the situ-
ational 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 dis-
course information within the Theme zone and comparatively less active
material in the Rheme, following the principle of Functional Sentence Per-
spective, 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 ‘clause-
external’ realization, in which case Themes either have one attentional do-
main 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 con-
ceptual ‘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 incremen-
tally 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
y
x
2 1 H 1 2
(TT) ^ (IT) ^ ET ^ (IT) ^ (TT)
narrowest scope
experiential meaning
interpersonal meaning
textual meaning
widest scope
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 ef-
fect 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 spe-
cific location.17
The presentative strategy recurs in (2b) – also a spatial Subject-verb in-
version – and in (2a), although in the latter case it is implemented
differently through the thematization of existential there. In the first incre-
ment 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 re-
cency effect to become the target of attention over the successive
attentional frames, thereby becoming the D-Topic of the subsequent dis-
course 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.)
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 corre-
sponding instructions for discourse continuity, discourse discontinuity and
both. High-low-oriented ETMs (e.g. (6) and (7) above) show how dis-
course 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 proposi-
tion 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 promi-
nence, 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 dis-
course 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, be-
cause of their informative and objective nature, readily accept logico-
conjunctive 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 ex-
position 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
Notes
1. 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’.
2. 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).
3. Carroll et al. (1971) and Goldberg, Sethuraman and Casenhiser (forthc.)
show that children begin to learn the associations between form and mean-
ing on two levels, i.e. verb-centred categories and abstract argument-
structure constructions (especially with such recurrent verbs as do, make,
get, go, etc.). Likewise, Bencini and Goldberg’s (2000) experimental re-
search suggests that individuals are likely to sort out sentences in terms of
argument-structure constructions rather than basing themselves on the lexi-
cal semantics of the matrix verb, which gives evidence for the psychological
or cognitive reality of constructions in language users’ minds.
4. 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 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.
5. 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).
6. Langacker (2001a) proposes the following list of natural paths, or cogni-
tively natural ordering of the elements of a complex structure:
8. Mackenzie and Keizer (1991: 194) and Hannay (1991: 138) consider New-
Tops as a subcategory of Focus, i.e. Presentative Focus.
9. In this connection Mackenzie and Keizer (1991: 187) argue, for example,
that not all GivTops need be contextually given, or introduced into the dis-
course 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 con-
textual 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... .
10. Note, however, that D-Topic relationships may also hold between nominal
expressions.
11. 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).
12. Worthy of mention here is Lambrecht’s (1994: 97) differentiation of activa-
tion 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 refer-
ent. 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 yester-
day. 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]).
13. 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’.
14. 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 rele-
vant 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 María de los Ángeles Gómez-González
18. 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’.
19. In LIBMSEC the rhematic area of there-constructions was found to be lexi-
cally and informatively four times heavier than the thematic part (8 vs 2).
20. 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 ac-
quisition, 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), Mean-
ing and Form: Systemic Functional Interpretations. Norwood, NJ:
Ablex. 1–64.
Bolkestein, A. Machtelt
1998 What to do with Topic and Focus? Evaluating pragmatic informa-
tion. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214.
Butler, Christopher S.
2003 Structure and Function: A Guide to Three Major Structural-
Functional 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 Han-
nay 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: Erl-
baum.
1993 English Grammar. (Vol. 2) Amsterdam and Philadelphia: Benja-
mins.
1995 Functionalism and Grammar. Amsterdam and Philadelphia: Benja-
mins.
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 239
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, Uni-
versity 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 Philadel-
phia: 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, modal-
ity and discourse moves. In Mike Hannay and A. Machtelt Bolke-
stein (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 particu-
larly concerning the optionality of layers at the representational level.
DYNAMIC, ABILITY
(1) 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
(2) They were servicing AWACs planes for the Royal Saudi Air Force and com-
pleting contract work on a controversial air defense network for Saudi
Arabia known as the Peace Shield. Boeing won't say how many of its em-
ployees are still there, but says those who stayed did so voluntarily. (CB)
DEONTIC
(3) 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
EPISTEMIC
(4) Is it possible that all human beings on earth today are descended from a
single woman? (CB)
DEONTIC
(5) 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)
EPISTEMIC
(6) Michael Jackson must have enjoyed his wedding in May, because he decid-
ed to have another one at his California ranch. (CB)
function in the clause (Halliday 1994), which take care respectively of the
conceptualization of the situation referred to in the clause and the interac-
tive 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 predi-
cation is the representational perspective on the SoA, as a description of
the situation to which the speaker refers in the clause, whereas the proposi-
tion 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 theoreti-
cally 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 be-
tween objective and inherent modality on the basis of their divergent
positions with respect to the SoA. Inherent modals serve to evaluate the ac-
tualization 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 ob-
jective modals evaluate it from an external perspective. In the FG model,
this is accommodated theoretically in terms of different positions with re-
spect to the predicational layer: the SoA-external function of objective
modals is reflected in their position as operators with scope over the predi-
cation, 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
2.4. Summary
Table 1. The FG analysis of modality
FUNCTION LAYERING CATEGORY MEMBERS CRITERIA
3.1. Conditionality
Hengeveld (1988: 236) proposes to use conditionality as an absolute crite-
rion 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 ade-
quate account of the behaviour of modal auxiliaries in conditional contexts:
subjective modals can occur in conditionals, but their interpretation is af-
fected 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 in-
stance,
In (7), the epistemic position expressed by may is crucially not the posi-
tion 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, sub-
jective 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 administra-
tion. (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 auxil-
iaries. Subjective modal auxiliaries can occur in interrogatives just like
objective ones, as shown in examples (9) and (10):
(9) 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)
(10) 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 sociol-
ogy politics? (CB)
(11) 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)
(12) 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 in-
terrogative 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, dy-
namic 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 epis-
temically (“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?”).
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 modal-
ity 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 interpreta-
tional 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 posi-
tioning, 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 func-
tion. 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 posi-
tion encoded in the subjective modal in his own turn, whereas the
interrogative allows the speaker to transfer this responsibility to the inter-
locutor 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.
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) “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)
(14) 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 ap-
proach. There is a crisis and he must act now <ICE-GB:S1B-056 #87-
89:1:F>
(15) 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)
(16) 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
In examples (13) and (14), the function of the deontic modal is to ex-
press 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 exis-
tence 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 ob-
jective 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 rela-
tive 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).
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 knowl-
edge 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 evalu-
ate the actuality of the SoA, but expresses his commitment to the
permission or obligation. It is the speaker who wants the SoA to be actual-
ized: 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 rea-
son 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 ex-
istence 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 voli-
tion 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 to-
gether 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 redistribut-
ing 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 subjec-
tive 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 in-
herent 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:
FG Alternative
Subjective (13), (14)
Objective (13), (14) (15)
Inherent (15) (16)
4.2.1. Conditionality
4.2.2. Interrogation
In Section 3.2, it was already argued that interrogation interacts with sub-
jective modality, in that the responsibility for the position encoded by the
258 Jean-Christophe Verstraete
(19) 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 dif-
ferences generally speaking between a counsellor a therapist and a
psychiatrist? (CB)
(20) 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 no-
body’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 inter-
preted 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 com-
mitment 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 prop-
erty? 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
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 re-
spect to the temporal zero-point. It might be objected that imperatives are
not tenseless but have an inherently future orientation, but this ‘future’ ori-
entation cannot be considered a tense the way future tense is in the
indicative mood. Future tense establishes a relation to the speaker’s here-
and-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 distin-
guishes indicative from imperative mood also distinguishes subjective
epistemic modality from subjective deontic modality. This becomes espe-
cially 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 in-
stance, if we compare (24) with (25):
clauses: the perfect in this case does not locate in the past, but simply indi-
cates 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 there-
fore 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.
(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 paradig-
matic equivalent of the imperative, which equally expresses the speaker’s
commitment to the desirability of the action and equally operates on tense-
less 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.
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 ex-
istence 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 be-
haves 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 mo-
dality 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 en-
coded 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 op-
erates on tensed SoAs. The availability of propositional attitude markers is
also a typically epistemic feature: propositional attitudes can only be ex-
pressed 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
Interpersonal Representational
FG model Epistemic Tense Deontic
Dynamic
Alternative Epistemic Tense Deontic
Indicative Dynamic
Deontic X
Imperative
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 proba-
bly 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 mo-
dalization 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 pro-
positional 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 ‘unlo-
calized’ 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 im-
plicit 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.
organization of the utterance. There is one module that takes care of inter-
personal 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. Top-
down orientation, on the other hand, implies that in this modular frame-
work the interpersonal module, which takes care of speaker-hearer
interaction, is primary in that it can determine choices in the representa-
tional 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 tradi-
tional layered model was the assumption that the full set of layers is
considered to be present in any type of utterance. In this perspective, asso-
ciation 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 modal-
ity – 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, be-
cause representational and interpersonal functions belong to separate
modules. The fact that epistemic and deontic modality have different do-
mains (tensed versus tenseless SoAs) is a purely representational matter
that can be dealt with in the representational component and does not inter-
fere 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
Interpersonal module
Meaning Speaker’s commitment
Echo-effect in conditionals
Subjective status of
Form
the modal
Interlocutor-transfer in interrogatives
Representational module
Notes
1. I would like to thank Bert Cornillie, Kristin Davidse, Renaat Declerck, Pat-
rick 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 Bolke-
stein (1980).
2. Examples taken from the Cobuild corpus are marked with CB, and exam-
ples taken from the ICE-GB corpus are marked with their standard ICE-GB
text code. The relevant modal is always underlined.
3. See Verstraete (2001a) for an overview of the distinction between subjec-
tive and objective modality in terms of various theories of layering.
4. As shown by Nuyts (1992, 1993), however, the difference between modal
adjectives and adverbs cannot be explained in terms of the distinction be-
tween objective and subjective function, but must be related to three
interacting parameters, viz. evidentiality, discourse functionality and per-
formativity (Nuyts 1993).
5. In fact, this was the original context where this criterion was introduced; see
Jackendoff (1972) and Bellert (1977).
6. 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 Jean-Christophe Verstraete
References
Bellert, Irene
1977 On semantic and distributional properties of sentential adverbs. Lin-
guistic 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. Lon-
don: Croom Helm.
Declerck, Renaat
1991a A Comprehensive Descriptive Grammar of English. Tokyo: Kaitaku-
sha.
1991b Tense in English. Its Structure and Use in Discourse. London: Rout-
ledge.
Declerck, Renaat and Susan Reed
2001 Conditionals. A Data-driven Analysis. Berlin and New York: Mou-
ton.
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 mo-
dality and mood in English. Foundations of Language 6: 322–361.
1975 Learning How to Mean. Explorations in the Development of Lan-
guage. 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 Predica-
tion, 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 Phila-
delphia: 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: Mi-
chael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered
Structure and Reference in a Functional Perspective, 73–97. Am-
sterdam 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 273
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 representa-
tion 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 ap-
proach 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 devi-
ates 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 con-
ceptual 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 representa-
tion 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 proce-
durally 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
clause
interpersonal (E1: [π4 ILL: (S) (A) (π3 X1: [proposition] σ3) σ4])
predication
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
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 domi-
nancy hierarchy between the qualifications involved, which is moreover
stable across various tokens of the combination. This is also true of qualifi-
cations which in the FG system are on one level, e.g. the predicational one.
7/8
The following examples are self-explanatory:
(1) a. John went skiing a few times [quant. aspect] last year [time]
b. It happened last year that John went skiing a few times
c. *It happened a few times that John went skiing last year
(2) a. John must [deontic modality] go skiing next year [time]
b. There is an obligation for John to go skiing next year
c. *It will happen next year that John must/is obliged [at the time of
speaking] to go skiing
(3) 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 incom-
plete10 – gradual conceptual hierarchy of qualifications in (4):
(4) evidentiality
> epistemic modality
> deontic modality
> time
> quantificational aspect
> qualificational aspect
> state of affairs
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 in-
volve (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 ap-
pear 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 inter-
nal 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.
tational dimension in language use (Hengeveld 1989) is not very well mo-
tivated. My objections (which I will not repeat here) partly boil down to an
intrinsic problem with the concept of separate strata in linguistic organiza-
14
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 origi-
nal 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 ap-
plies at the conceptual and not at the linguistic level, notions such as
‘speech event’ or ‘narrated event’ (which Hengeveld correlates with the in-
terpersonal 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 illus-
trate 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 desir-
able or necessary, would be representational, while an evidential category
such as inference would be interpersonal. Both just participate in the con-
15
ceptual representation of reality, no more and no less.
Thus, the layered system in FG to a considerable extent integrates inter-
actional – 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 be-
long 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 illocution-
ary 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 situa-
tion, 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
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 hi-
erarchy involves an increasing role for creative involvement of the speaker
(see Section 3.1), what causes that involvement? If attitudinal qualifica-
tions 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 intelli-
gence 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 op-
erations 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 con-
sciousness 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 discus-
sions 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 percep-
tual and behavioural systems, including the linguistic one.
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 re-
calculate 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 cru-
cial. 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 en-
counter in linguistic data in the grammar proper. As long as they deal with
structural phenomena, that is fine. But when they start dealing with seman-
tic 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 organiza-
tion in grammar – is a case in point.
Notes
1. The same is probably true of understanding linguistic expressions. But I will
mainly adopt a production perspective here.
2. The arguments and views presented in this chapter are developed in much
more detail, and are moreover underpinned by substantial amounts of ex-
perimental and corpus data, in Nuyts (2001b).
3. 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.
4. The foregoing actually does not imply that conceptualization must be imag-
istic (cf. Dik’s ‘perceptual representations’ or Jackendoff’s ‘3D
representations’). An action scene such as the ‘commercial event’ involves
Remarks on layering 293
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 diffi-
culty, 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 bio-
logical 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.
12. 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.
13. 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 be-
ing determined by semantic scope) and macro-structurally (in terms of the
general rationale behind the organization). But showing this is beyond the
present chapter.
14. Representational vs interpersonal will probably not do for that purpose
anyway: see Nuyts (1992a: 26–64, 1993b).
15. 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 concep-
tual 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 sys-
tem.
16. This is reflected in the fact that all attitudinal qualificational expressions
show a difference between performative and descriptive uses, i.e., respec-
tively, 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 295
qualifications with speaker commitment. But they are in any case fully sub-
ject 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 mo-
dality, etc.
References
Bock, Kathryn and Willem J.M. Levelt
1994 Language production: Grammatical encoding. In: Morton Ann
Gernsbacher (ed.), Handbook of psycholinguistics. San Diego: Aca-
demic 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 Linguis-
tics 1: 11–52.
1987 Linguistically motivated knowledge representation. In: Makoto Na-
gao (ed.), Language and Artificial Intelligence, 145–170. Amster-
dam: North-Holland.
1989a Towards a unified cognitive language. In: Frieda J. Heyvaert and F.
Steurs (eds), Worlds behind Words, 97–110. Leuven: Leuven Uni-
versity 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 Psychol-
ogy of Language, 155–175. Mahwah: Erlbaum.
Jackendoff, Ray
1983 Semantics and Cognition. Cambridge, MA: MIT-Press.
1987 Consciousness and the Computational Mind. Cambridge, MA: MIT-
Press.
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: Cam-
bridge 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: Mi-
chael Fortescue, Peter Harder and Lars Kristoffersen (eds), Layered
Structure and Reference in a Functional Perspective, 73–98. Am-
sterdam 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 Le-
jeune (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 Cognitive-
Pragmatic Perspective. Amsterdam and Philadelphia: Benjamins.
298 Jan Nuyts
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 inter-
nally transforming the FG apparatus itself. Within the latter line of thought,
two approaches can be distinguished: an expanding ‘upward layering’ ap-
proach (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) phe-
nomena 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 ap-
proaches 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 phe-
nomena 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, Mou-
taouakil 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 organ-
ized 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.
TEXT
(1) DISCOURSE = CLAUSE
TERM-PHRASE
WORD
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 posi-
tion 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 ‘Sen-
tence’ between text and clause, as a supra-clausal entity, which yields the
following alternative 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 ten-
dency is the postulation of a universal abstract archetypal discourse
structure whose actualization in these discourse categories takes place ac-
cording to certain parameters.
He then suggests adding to the existing term layers a fourth one which
he calls the ‘modality layer’. This yields structure (10), assumed to under-
lie both Term and Proposition :
(10)
[Mod-Op Mod-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 struc-
tural level and the relational level. The basic idea is that intra-clausal
layered structure as well as intra-clausal functional relations can be pro-
jected 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]
You know
(13) Listen , let's go to the theatre this evening.
Look
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]
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 interac-
tional operator (Inter-Op) and an interactional satellite (Inter-Sat), slots
intended to host particles like Hi!, Hey! as well as the quasi-
grammaticalized 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 follow-
ing contrast:
Discourse structure, the generalized parallelism hypothesis, and FG 307
EVENT
(16) SETTING = TYPE
STYLE
In fact, (16) does not tell the whole story about the underlying represen-
tation 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 “rhe-
torical 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
The main argument in favour of such an approach is that the three sub-
categories of SETTING codetermine, as mentioned above, the form as well
as the content of the subsequent discourse and should therefore be consid-
ered 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 dis-
course 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 actualiza-
tion of structure (17).
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 terminol-
ogy) 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 prop-
erly 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 non-
literal 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 repre-
sentational 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 be-
tween ‘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) un-
derlying 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 illocu-
tionary 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 expres-
sion rules which deliver the final syntactic form.
Discourse structure, the generalized parallelism hypothesis, and FG 311
MNLU
Grammatical
Structure
Pragmatic structure
Expression rules
Syntactic form
MNLU
PRAGMATIC MODULE SEMANTIC MODULE
Pragmatic Semantic structure
Structure
GRAMMATICAL MODULE
Expression rules
Syntactic
Syntactic form
form
The basic idea which will be developed in what follows is that Arche-
typal 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 distin-
guished 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 dis-
course 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, Mou-
taouakil 1993, and, in particular, Mackenzie 1998), cannot be said to be
less communicative. They also can be used to carry a complete piece of in-
formation, 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 de-
gree of explicitness of their communicativity.
Concerning the actualization of structure (17) in the discourse catego-
ries 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 be-
yond 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 (Mou-
taouakil 1993) that only volitional and emotional subjective modality
distinctions can occur in a modalized term (excluding evidential distinc-
tions).
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 actualiza-
tion of structure (17): on a stem standing as a nucleus, prefixes and suffixes
build hierarchically organized layers. As one might expect, the actualiza-
tion 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 repre-
sentational 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 in-
stance, the modal values are restricted, it seems to me, to pejorative
features. Third and more importantly, words do not have, as one might ex-
pect, the same hosting capacity as terms and clauses. A complex word can
hardly contain more than two prefixes. This means that it would be diffi-
cult to find words involving more than two layers at once.
A final word on the text-clause parallelism: the structural similarity be-
tween 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 direc-
tion, 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 or-
ganization from one and the same archetypal structure. It seems to me that
the parallelism phenomena at hand can be accounted for in a psychologi-
cally 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 dis-
course category which is the source of this projection (clause or text).
Interestingly enough, this permits us to avoid the undesirable ‘clause-
centricity’ of FG.
As mentioned above, the discourse categories lower than the text can be
used in two ways: they can constitute complete communicative units or oc-
cur 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 com-
plex) clauses, when coherently sequenced, form a text. In the previous sub-
section, 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, fo-
cusing 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 fol-
lowing general assumptions:
where α stands for any operator value and where X, Y, Z symbolize a se-
quence of clauses.
Assumption (c) means that the inheritance process can be partial or to-
tal. 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 deictic-
centre 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 param-
eterized actualization of ADS allows us to approach these on a new basis.
First, it becomes possible to redefine embedding as a linkage of two dis-
course 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 quali-
tative 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 em-
bedded layers.
Discourse structure, the generalized parallelism hypothesis, and FG 317
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 in-
stance, 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 rela-
tional part of ADS.
In the FG framework (cf. Dik 1997a: 7), linguistic universals are defined
within an approach which conceive of natural languages as particular solu-
tions 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 proper-
ties which fit optimally into this functional conception of linguistic
universals. Notice that, given its nature, its components and its organiza-
tion, 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 it-
self, 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 ac-
quisition 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 ex-
tending the current clause/sentence FG model and transforming it into a
more text-oriented model is built on four related assumptions that presup-
pose 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 de-
grees, 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 so-
lution 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 al-
ways 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 embed-
ded 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 differ-
ences 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 in-
teraction 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 dis-
course 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 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 satel-
lites respectively.
2. 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 em-
bedding phenomena fails to properly describe the structure of this type of
construction.
3. 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 predica-
tions, 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 informa-
tion. In: Mike Hannay and A. Machtelt Bolkestein (eds), 193–214.
Connolly, John H., Roel M. Vismans, Christopher S. Butler and Richard A. Gat-
ward (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
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. Univer-
sity 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), Func-
tional 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 collo-
que 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 prag-
matic 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 prag-
matic Interpersonal Level (IL). In addition, the interpersonal level has two
higher, discourse-related layers, viz. the Act and the Move. The informa-
tion 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
pression rules. We will take as our point of departure the proposals relating
to the expression component made in Bakker (2001) rather than the stan-
dard 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. Sugges-
tions 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 gram-
mar and its further development.
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 al-
ways 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’ ex-
pressions 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 morpho-
syntax 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 descrip-
tively adequate in the more obvious sense of generating the forms that do
occur.4 But overgeneration is not only a theoretical issue. An overgenerat-
ing 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
example, is arguably an appropriate template for the clause level.5 The ap-
propriate 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 VFIN VINFIN
TERM(SUBJECT) [IMPRF,PRES,PL,3] VERB [IMPRF, GERUND]
/are/ /growing/
DET NOUN NOMRESTR
[DEF] [PL]
LABEL
Functional aspects
CONFIGURATION
FUNCTIONAL FEATURES
Formal aspects
FORMAL FEATURES
SUBCATEGORIZATION
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 Vin-
fin. 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
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 unifi-
cation 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
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).
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 alterna-
tive versions for this node in the grammar, e.g. with a Focus, Topic or Sub-
ject 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 expan-
sion 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 com-
ponent. 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 14 15 20
2 3 4 9 12 13 16 19
5 6 7 8 17 18
/those/ /are/
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 psy-
chologist. 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 Cogni-
tion and Communicative Context in the model, there is a separate box
labelled ‘Grammar’ around the pragmatic (IL), semantic (RL) and expres-
sion (EL) levels. This suggests that, apart from the lexicon, which is
contained in Cognition, the grammar gets a separate status outside the lat-
ter, 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 inter-
preted 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 substruc-
tures. 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 Communica-
tive 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 con-
tribution 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 observa-
tions 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 informa-
tion 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 sug-
gested 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 enti-
ties 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 se-
mantic 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 in-
formation 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 prag-
matics 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 syntac-
tic 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 in-
stance. Rather, it is a broad interpretation of the meaning, the gist or
intention of what was said. Therefore, speakers will not normally remem-
ber whether they have said (8a) or (8b).
Taking a different perspective on things, one might argue that the con-
tents 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 sponta-
neous conversation completely ‘new’ things may be said that speakers do
remember later on. Also, telling ‘old’ stories may have a restructuring ef-
fect 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 re-
fined 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. Accord-
ingly, 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 stor-
age. 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 Hal-
liday’s information unit (Halliday 1994).14 Indeed, in studies of spoken
language it has been frequently observed that (spontaneous) speech gener-
ally 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
(9) Tomorrow.
To Mila.
Liam too.
Five of these, please.
The green one, if possible.
Skating.
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 as-
sume 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 con-
tinuity 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 347
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 indi-
cation 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 fol-
lowed 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 compet-
ing 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 con-
structions. Topic Continuity favours topical and other ‘old’ and
presupposed entities; Task Urgency favours focal and other ‘new’ and un-
predictable 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 deter-
mined ‘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 re-
ferring 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
Now semantics will kick in and start working on this structure. Seman-
tic 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 un-
doubtedly 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 )])
After unification with the structures in (14) and (15) we get the node in
(17).
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 underly-
ing representations, but by the expression rules themselves. Like many
authors, we see the phenomenon of Subject as the outcome of a grammati-
calization 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 con-
stituents 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, cross-
linguistically, 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 or-
der, agreement and several types of Same-Subject marking or non-
expression 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).
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.
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 ‘unneces-
sary’ 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 9
SUBJ VC SAT
2 5 7 10 12
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 )
(def x9)TEMP )])
(23) [DECL (P1)S (P2)A (C6: [(x1)TOP (x8)FOC (x9)FOC (f1)])]
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).
SENTENCE
1 4 9 14
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 addi-
tional 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’.
NODE 2’ (uninstantiated)
Lab: interj
Config: [FORM, [INTERJ], TYPE ]
FncFtrs: TYPE
In the meantime, the memory search for the value of the temporal satel-
lite 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 ele-
ments 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 struc-
ture 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 4 9 14
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 fol-
lowing RL level information:
Towards a speaker model of FG 355
(29) (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.
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 26 28
After (30) – previously (11c) – has been uttered, the monitor appears to
have been active again. The very specific version of the latest temporal sat-
ellite is relaxed somewhat in the last contribution to the current act, (11d),
now renumbered as (34).
For this speech act, which we interpret as an alternative for the com-
plete 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 sugges-
tions 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 work-
ings 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. Re-
lated to this point is the requirement that the model should provide an
answer to what may, or must be repeated after hesitations and self-
corrections, and how corrections affect the eventual representations.
A second point is the precise nature of the interaction between the re-
spective kinds of memory, and the way information is stored there, what is
retrievable, how it may be extended, and under what conditions. A sub-
problem 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. Fortu-
nately, 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
Notes
1. The authors wish to thank Matthew Anstey for a number of valuable com-
ments and interesting points of discussion with regard to an earlier version
of the text. We hope we have done justice to them.
2. 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.
3. Of course we could, if only for the sake of argument, take an extreme func-
tional 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 pre-
running the expression rules in a more or less disguised way.
4. This requirement can be made explicit by formulating the role of the ex-
pression component such that a grammar of language L is descriptively
Towards a speaker model of FG 359
11. Note that Dik (1989) proposes the FG underlying representations as candi-
dates for mental representations.
12. 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 con-
versation. Instead of short-term memory we will use the wider and more
dynamic notion of working memory. This holds the information directly re-
lated 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.
13. In this context the magical ‘7 plus or minus 2 items of information’, stem-
ming from Miller (1956), should be mentioned.
14. 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.
15. 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 expres-
sion 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.
16. For an early discussion of abstract predicates in FG meaning definitions see
Siewierska (1993), who discusses proposals in Jackendoff (1990). See Fa-
ber 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.
17. Typically, in this type of story-telling discourse, each Move will be Inform-
ing by default. The insertion of continuation markers such as And and And
then may well be the expression of this operator.
18. We will follow the convention of distinguishing abstract, language-
independent 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.
19. In order not to burden the discussion with matters irrelevant to it, we follow
the FG convention of representing adpositions via semantic functions. In
Towards a speaker model of FG 361
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 Fine-
gan
1999 Longman Grammar of Spoken and Written English. Harlow: Long-
man.
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 Lin-
guistics 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 Philadel-
phia: Benjamins.
Givón, Talmy (ed.)
1983 Topic Continuity in Discourse: Quantitative Cross-Language Stud-
ies. Amsterdam and Philadelphia: Benjamins.
Halliday, Michael A.K.
1994 An Introduction to Functional Grammar. 2nd edition. London: Ar-
nold.
Hannay, Mike
1991 Pragmatic function assignment and word order variation in a Func-
tional 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. Cog-
nitive 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 utter-
ances. 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 Philadel-
phia: 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 ver-
sions 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 struc-
ture, 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 rele-
vant 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 it-
self. The last three concern the incorporation of existing FG components
into FDG. After considering the six issues one by one, I will present a re-
vised and more detailed version of FDG in the final section of this paper.
366 Kees Hengeveld
The representation in (1) contains three types of variable from the re-
presentational 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
are encoded pronominally when they are available within the contextual
component, and nominally when they are not. The importance of the inter-
action 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 Fig-
ure, boxes represent components and levels, and circles represent
operations. The main components are the conceptual component, the con-
textual component, the acoustic component,5 and the grammatical compon-
ent. The grammatical component distinguishes an interpersonal, a
representational, and a structural level. The vertical arrows indicate the fol-
lowing: 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 en-
coded 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 indi-
cate 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
Formulation
Grammatical Component
Interpersonal Level
Representational Level
Contextual Component
Encoding
Structural Level
Acoustic Component
Articulation
Expression Level
(3) (Ω R1: [(Ω x1: [(Ω f1: LexemeN (f1): µ (f1))] (x1): µ (x1))] (R1): µ (R1))
6. Functions
As Anstey notes in his contribution to this volume, the functional hierarchy
of influence (pragmatics > semantics > syntax) is instantiated in the exis-
tence 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 ar-
guments in this volume for locating pragmatic functions at the inter-
personal 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 Siewier-
ska this volume). In this way, syntactic function assignment can be seen as
the outcome of a process in which both pragmatic factors (at the interper-
sonal level) and semantic factors (at the representational level) are taken
374 Kees Hengeveld
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 Siewier-
ska 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: acts are based on an illocutionary frame; there are certain lexemes
that have an interpersonal content only (personal names, pronouns, inter-
jections);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 pat-
terns.
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 suffi-
cient to trigger a certain prosodic pattern at the acoustic level.
Conceptual Component
Formulation
Frames,
Lexemes, Interpersonal Level
Operators
Grammatical Component
Representational Level
Contextual Component
Encoding
Templates,
Morphemes
Structural Level
Acoustic Component
Articulation
Prosodic
Patterns,
Sounds
Expression Level
Notes
1. 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 re-
interpreted, it is convenient to have a separate label to refer to this new re-
search enterprise.
2. In this volume, however, Moutaouakil presents another version of FG which
is both modular and hierarchical.
3. I have substituted the term ‘structural level’ for the term ‘expression level’
which was used in the first chapter of this volume, since expression in-
volves more than just structure. I will come back to this issue in Section 3.
4. 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.
5. Of course the acoustic component is only relevant for spoken language. For
sign language this would be a sign component, for written language an or-
thographic component.
6. Note that the structural level in fact has to be split up into a morphosyntactic
and a phonological level.
7. See e.g. Mackenzie's and Moutaouakil's contributions to this volume.
8. The difference between frames and templates is that frames are unordered
semantic configurations whereas templates are ordered syntactic configura-
tions.
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 Struc-
ture and Reference in a Functional Perspective, 387–407. Amster-
dam 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 Per-
spective, 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 Interac-
tion, 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 Pa-
pers in Functional Grammar 16.
Index of names
ability 36, 245, 247, 249, 251, 254- ascription 7, 108, 185, 198, 202,
256 212, 326, 344
aboutness 120, 131, 222 aspect 40, 54, 79, 92, 118, 126, 131,
accessibility viii, 144, 175, 180, 189, 151, 153, 156, 183, 276- 277,
219-221, 282-283, 331, 346 281-284, 290, 293, 329, 358
acoustic 368, 370, 375, 377 assertion 123, 125, 186, 223, 231,
acquisition 56, 62-63, 214, 294, 320, 235
338, 345 attention viii-ix, 11, 26, 45, 48, 76,
act 5, 7-8, 10-11, 15-16, 43-44, 47, 79, 102, 117, 119, 121, 123-125,
76, 91, 95-97, 99, 102, 108-111, 127-130, 133, 135-138, 143, 145,
135, 152-153, 167, 171, 174, 183- 175, 212-213, 217-218, 220, 222,
186, 188, 190, 192, 200-203, 205, 228-230, 233, 289, 307, 346, 353,
215, 230, 233, 235, 247, 253, 255, 363, 369
277, 286-287, 294-295, 347, 351, attitude 109, 171, 186, 206, 226,
355-356, 367, 372 261-263, 270, 277, 287, 289, 292-
activation 75, 184, 193, 220-222, 294, 372
235, 319 bottom-up vii, 2, 182, 199, 211, 338
addressee ix, 5, 15, 77, 80, 83, 90, buffer 188, 215
108, 119, 121, 125, 127- 134, bystander 370
136-140, 143-145, 186, 188-189, categorical 119-120, 122, 139, 143-
191, 202, 206, 215, 234-235, 369 144, 149
adjacency 102, 109-111, 141 category 8, 10, 17, 28, 30-31, 46, 52,
affix 157-159, 165, 169, 171, 173 56-59, 78-79, 91, 107-108, 118,
animacy 157, 161, 234 153, 158, 161, 163-164, 167, 176,
antecedent 43, 53, 129 179, 198, 207-208, 214, 218-219,
apodosis 252, 288 221, 223, 233, 244-247, 252, 254-
apposition 57, 226 259, 263-268, 270, 275, 277-279,
archetypal 300, 303, 312, 315, 320 286-287, 293-294, 300-303, 305,
architecture vii-x, 23, 25, 50, 55, 73- 307-308, 310, 312-315, 320-321,
74, 80, 89, 167, 169, 212, 243- 329-330, 333-334, 350, 353, 361,
244, 266, 268, 299, 365, 378 372
argument 16, 28, 35-36, 43, 48, 59, chain 50, 52, 75, 80, 161, 165, 198,
120-121, 124, 127, 145, 157-158, 234
161-164, 166, 174, 180, 187, 192, classifying 8
202, 204, 207, 213, 223, 231, 263, clause ix, xiii, 7, 11, 15, 25, 36, 41-
270, 277- 278, 293, 309, 349, 358 42, 44, 56, 60, 89-90, 92- 93, 95,
articulation 2, 34, 52, 55, 73, 161, 108, 118-119, 121- 122, 127, 131-
164-165, 169, 182, 187, 212- 213, 132, 136, 139-145, 152, 156-157,
326 160-166, 169, 171-172, 175, 181-
Index of subjects 385
182, 188, 197, 204, 206-209, 230- conceptual 23, 39, 60, 75, 87, 98,
234, 244-246, 251, 264-266, 275- 152, 165, 171, 175, 181, 185,
276, 284, 288, 295, 300-302, 304- 211-212, 218, 220, 226-227, 229,
305, 309-310, 313-315, 317, 320, 236, 272, 275, 277-279, 281-283,
325-326, 328- 329, 339, 348, 365- 285-290, 293-297, 345, 355, 366,
366, 372 369-370
clause combining 41 conceptualization ix, 79, 84, 187-
cleft 172, 174, 296 188, 215, 218, 227, 236, 247,
clitic 118, 122, 127-129, 143, 145, 277-278, 280-281, 285-286, 288-
158 289, 291-292, 295, 297, 307
code 16, 44, 49, 153, 199, 206, 269, conditional 16, 248-249, 251, 256-
280, 295, 329, 345 257, 269, 287-288
cognition viii-x, xiii, 3, 17, 38- 39, configuration 333, 335, 370
50, 56, 74-80, 82-84, 93, 140, connectionism 75, 79
143, 151-152, 170-172, 175, 180- connective 123, 133, 141
181, 183, 185-189, 211-215, 219- constituent 7, 26-28, 35-36, 38, 46,
221, 226-229, 232-233, 235, 241, 52, 56-57, 90, 108, 119, 124, 126,
275-276, 280, 282, 288-289, 292, 131, 138, 140, 142- 143, 145,
296, 328, 332, 348, 351, 359, 365, 154-161, 166, 169, 172, 188-189,
369 197, 207, 217, 221-222, 224, 231,
cognitive component 3, 17, 74, 76- 235-236, 285, 302, 305-307, 309,
79, 82-84, 93, 175, 212, 214, 369 316, 326, 328-329, 331-332, 335,
cognitive grammar 215, 241 347, 349, 352, 358-359, 361, 368,
cognitive-functional 275-276, 288 373, 377
coherence 126, 142, 308 constituent order (see also word or-
cohesion 161, 220, 231, 343 der) 328, 347, 349
commitment 40, 133, 246-248, 250- content ix, 3, 5-6, 8, 10-11, 15, 37,
252, 254-255, 257-259, 261-262, 46, 73, 77, 82, 90, 97-99, 101,
265, 267-268, 270, 284, 289, 291, 111, 120, 139-140, 153, 160, 168,
295 185, 188, 198, 200-206, 208, 220,
communication x, 8, 13, 26, 47, 49, 247, 251, 277, 300, 309, 320,
80-83, 94, 103, 105, 167, 175, 340-341, 344, 372, 375
181, 183-186, 217, 233, 278, 291, context 2, 12, 17, 27, 36-37, 44, 49-
301, 315, 319-321, 366-367 51, 56, 60, 76, 82-83, 90, 93, 102-
communicative competence 3, 30, 103, 108, 110, 118-121, 123, 125-
74, 326 127, 136, 139- 140, 142, 145,
communicative context 12-13, 36, 153, 155, 157, 159-161, 163, 166-
44, 49-51, 60, 76, 82-83, 140, 167, 169- 170, 173, 184, 186,
169, 184, 186, 214, 341, 344, 369 188, 190, 214-215, 220, 226, 234-
complement(ation) 41, 129, 161, 328 236, 249, 252, 256-257, 269, 286,
computational 25, 39, 55 289, 313, 341, 344, 360, 369
concept 29, 32, 52, 56, 131, 165, continuity 45, 58, 91, 126, 229, 232,
189, 234, 260, 275-276, 278, 283, 346
286, 289, 295, 300
386 Index of subjects
contrastive 145, 155, 166, 169, 172, 310, 312-321, 327, 340, 342-349,
174-175, 217, 313 355, 357, 360, 367-369
control 30, 85, 161, 167, 289, 290, discourse act 2, 5, 8, 10, 44, 47, 54,
358 93, 97, 103, 188, 201-202, 212,
controller 130 367-368
conversation 52, 77, 92, 106, 135, discourse grammar 2, 152, 154, 160,
183-184, 220, 279, 341-343, 346, 301, 343
357, 360-361 discourse particle 2, 60
coordination 24, 27, 57, 139, 226, discourse setting 132, 308
316 D(iscourse-)Topic 218-220, 222-
copula 15, 160, 162 223, 230-232, 235
coreferentiality 60, 129 domain-specific 74
current discourse space 215, 229 downward layering 7-8, 171, 198,
declarative ix, 56, 75, 90, 131, 166, 203, 372
175, 179-182, 187-188, 190-193, driving force 369
202, 204, 207, 250, 252, 270, 287, dynamic viii, x, 2, 30, 50, 90, 151,
345 182, 212, 214-215, 227, 229, 241,
decoding 73, 83, 214, 217, 259 245, 251-252, 254-256, 270, 293,
definite 37, 58, 132, 134, 137, 141, 327, 329-332, 334-335, 345, 356-
144, 149, 157, 160 357, 359- 361, 367-369, 375, 377
deixis 20, 49, 60, 127, 134, 136, dynamic modal 245, 252, 254-256,
148-149, 221, 307, 370 293
delimiter 107-108 echoic 249, 251, 256-257
demonstrative 13, 53, 60, 127- 128, embedding 15, 31, 90, 94, 152, 165,
130, 133-134, 136-137, 146, 333 167, 171, 174, 204, 207, 312,
denotation 44, 120 315-316, 321-322, 328, 342, 372
deontic modal ix, 243-246, 252-270, emic 153, 161, 164-165, 168
281-282, 284, 286, 290, 294, 296 encapsulation 74
depth-first x, 182, 212, 331 encoding 7, 18, 46, 60, 73, 81-83,
diachronic 281, 293, 295, 332 127, 129, 130, 132, 145, 187, 214,
dialogue 18, 90-92, 94, 96, 104, 106, 217, 231, 248, 251-252, 259, 261,
108, 114, 141, 232 267, 296, 326
direct speech 16, 44, 53 encyclopaedic 75-76, 141
discourse vii-ix, 2-3, 5, 8, 10, 12, 26, entities 6, 8, 11-12, 53-54, 58-59, 75,
38, 43-44, 47, 49-52, 54, 59-60, 79, 122-123, 131-132, 135, 138,
74, 83, 89-113, 119-121, 123-128, 141, 151-152, 159, 167, 170-172,
130-134, 136, 138-146, 150, 152- 185, 197-198, 212-213, 218-220,
153, 155-156, 158-161, 163-169, 223, 226, 231, 281, 287, 302-303,
171-172, 174, 181-183, 187-188, 312, 318, 326, 331-332, 338, 340,
197-199, 201-203, 206-208, 211, 342, 347, 351, 372-373
213, 215, 217-222, 224, 226-237, episode 241
249, 252, 256, 261, 269, 276, 285, epistemic modal ix, 59, 125, 244-
288, 291-292, 295, 299-303, 305- 246, 249-252, 254-256, 258-270,
Index of subjects 387
ideology 94, 107, 110 185, 206, 211, 215, 243, 263, 267,
illocution 5, 11, 18, 43, 46, 91, 99, 279-280, 285, 306, 313, 320-321,
116, 153, 164, 171, 186, 222, 287, 326, 340, 357, 366, 370-371
307, 318, 347 interface 2, 76, 93, 98, 153, 161,
imperative 90, 259-262, 270-271 182, 197-199, 204, 358
increment ix-x, 186-187, 212, 217, interjection 107, 109, 300, 353
226, 230, 338, 343-344, 355 interpersonal vii-ix, 3, 5, 7-8, 10- 11,
Incremental Discourse Cognitive 13, 15, 17, 41-45, 47-48, 51-52,
Grammar ix, 215 73, 75-77, 82-83, 91, 99, 103,
Incremental Functional Grammar 106, 109-112, 167, 169, 171, 176,
viii, 25, 212 181, 197-198, 200-204, 206-208,
indefinite 37, 136-137, 144, 190 213, 225, 227-228, 232, 243-244,
indexical 117, 127, 137-138, 146 246- 248, 251, 264, 266-270, 273,
individual vii, 2, 6, 31, 56, 75- 76, 276, 285-286, 294-295, 306, 312-
90, 92, 94, 129, 144, 151-153, 314, 318, 325, 347, 358, 367,
160-161, 173, 182, 289, 294, 331, 369-370, 372-373, 375
358, 367 interpretation 10, 27, 37, 39, 46- 47,
inference 50, 132, 136, 153, 161, 50, 58, 74-75, 83, 100, 119, 122,
167, 221, 286, 294, 346 125, 129-130, 136, 143, 151-154,
inferrable 124, 230, 235 163, 168, 170, 181, 188, 207, 219,
information viii-ix, 2-3, 5, 8, 12, 15, 248-251, 270, 282, 284, 337, 341,
29, 35-36, 40, 47-48, 51-53, 74- 366-369, 373, 375
76, 78-84, 90, 98, 102, 104, 109- interrogative 90, 156, 243-244. 248,
110, 120-122, 124-126, 131, 135- 250-252, 257-258, 261- 263, 267
137, 139-142, 144-145, 153, 167, intonation 175, 183, 190, 192, 215,
172, 174- 175, 180-181, 185-187, 219, 221, 226, 228, 232- 233,
189, 191, 201-202, 204, 207, 212- 342, 344, 346
215, 217-223, 226, 228-231, 233, inverse 157, 161, 164, 173
235-236, 247, 277-279, 282-283, knowledge 3, 37, 39-40, 44, 74-76,
285, 289-291, 294- 295, 312-313, 78-79, 84, 105, 119, 131, 175,
318, 325-326, 329-332, 334-335, 186, 212, 214-215, 221, 232, 246,
337, 340-344, 346, 348, 354, 356- 254-255, 266, 282, 284, 287-291,
360, 368, 373, 375 295-297, 308, 326, 329, 337, 339,
inherent modal 245-247, 254 344-345, 357, 359-360, 369
inheritance x, 315, 316, 322, 331 label 110, 228, 235, 290, 293, 309,
instruction 81, 90, 119, 125, 355 333, 339, 377
intention vii, 2-3, 5-6, 10-11, 34, 47, language user 39, 56, 118, 180, 191,
52, 73, 82, 153, 165, 168- 169, 233, 326, 336-337, 339, 358, 362
174, 182, 184, 187, 192, 197-198, layering vii, 40, 43-44, 47-48, 51-52,
201-202, 205, 213, 287, 300, 337, 59-60, 80-83, 91-92, 108, 111,
341, 346-347, 359 197, 203, 206-207, 246-247, 266-
interaction 25-26, 30-31, 34, 39, 44, 267, 269-270, 293, 304-308, 314,
49-50, 60, 73-74, 76, 79- 80, 90, 317-319, 321, 325, 332, 344
109-110, 115, 153, 166, 182, 184-
Index of subjects 389
left-to-right ordering x, 10, 17, 169, 331, 337, 339-348, 351, 353-355,
171, 176, 182, 186, 212- 213, 357, 360-361
329, 337, 358 message 16, 49, 73, 81, 84, 93, 98,
level viii-x, 2-3, 5-8, 10-13, 15-17, 108, 118, 132, 140, 142, 153, 166,
20, 26, 40, 43-44, 46-47, 56, 73, 175, 188, 189, 193, 200, 215, 224,
76-77, 81-83, 92-93, 99-100, 102- 226, 228, 232, 236, 320
103, 105-106, 108, 110-111, 115, message management 132, 140, 142,
122-123, 127, 129, 133, 139-143, 153, 166, 175
145- 146, 151, 153-155, 160-161, metalinguistic 13, 16, 78, 326, 340
163-165, 167-171, 174-176, 180, modality 58, 206, 243, 245, 249-258,
183, 185, 190, 197, 200-206, 213- 260-261, 264, 266-270, 273, 292-
215, 227, 233, 235, 243, 263, 265, 293, 297-298, 314, 316, 322, 372
268-269, 276, 281-283, 285-286, modal auxiliary 243, 249-251, 260,
288, 295, 300, 305, 308-309, 312- 270, 273, 293
314, 316-318, 321, 325-326, 328- modifier 33
329, 331, 337, 339, 341, 344, 347, modularity vii, 2-3, 44, 51, 74, 78,
349, 351-352, 355, 358, 366-370, 84, 92-93, 153, 197, 212, 239,
372-375, 377 243-244, 264, 266-268, 299, 309-
lexeme 12-13, 16, 54, 61, 368 310, 321, 365, 377
lexical 6, 10-11, 25, 28, 30-31, 35- module viii, 36, 44, 49, 93, 97-99,
36, 45-46, 55, 57-60, 75-76, 78, 132, 140, 153, 160, 163, 167-168,
83, 93, 102, 108, 128- 129, 137- 180, 267-269, 305, 310, 312, 321,
138, 154, 158-162, 164, 166-167, 326, 331, 339, 346, 365
173-174, 181, 184-185, 191-192, monitor 356-357, 359
199, 211, 213, 233, 278, 280-281, monologue 90, 92
288, 290, 296, 318-319, 322, 325- mood 79, 118, 172-173, 200- 201,
327, 330, 334, 337, 344, 362 206-207, 224, 259-262, 271-272
Lexical Functional Grammar 213, morphology 18, 46, 74, 131, 163-
337 164, 173, 222, 314, 318, 327,
lexical priority 334 331-332, 335, 337
lexicology 41 move 5, 10-11, 13, 43-44, 50, 60, 92,
lexicon 25, 30, 35, 46, 51, 55, 76, 98-99, 104, 108-109, 136, 153,
78, 152-154, 160-162, 164, 166, 183, 197, 201-202, 224, 246, 332,
167, 171, 174, 176, 184- 185, 347, 353
187, 190-192, 213, 288, 326, 335, multimedia 95, 105
337, 339-340, 345, 348, 353, 360 narrative 2, 99, 126, 307, 318, 321
linearization 36, 329 negation 28, 153, 156, 161, 279, 281
LIPOC 51, 180 New Focus 223
marker 107, 134, 136, 145, 157- 158, New Topic 131-132, 219
173, 251, 269-270, 316, 328, 346, newsworthiness 155, 160-161, 163-
358 164, 166, 169, 172, 235
memory 74-75, 78-79, 85, 87-88, node 75, 175, 331-335, 348-349,
212-213, 220-221, 232, 288, 291, 352-356
390 Index of subjects
noun phrase 25, 28, 35-36, 42, 56, polarity 224, 247, 277, 279
198, 285, 331 politeness 90, 106-107, 305, 307,
nucleus 35, 57-58, 92, 121, 144, 152, 319, 322
172, 181, 314, 316, 325 pragmatics 3, 10, 31-32, 35-38, 44,
object 32, 61, 142, 178, 183, 224, 47-50, 53-54, 57-58, 76, 90, 93,
236, 374 97-98, 116-117, 119, 124, 131-
objective modal 59, 243, 247-252, 132, 139-142, 145, 153, 156, 160-
256, 269, 272-273, 277, 293, 297 161, 164, 168, 170-171, 174, 181-
183, 185-187, 211-212, 219, 235,
operator 37, 40, 58, 190, 207, 247- 247, 292, 302, 309-310, 312-313,
248, 264-268, 306, 313-316, 319, 318, 321, 325-326, 328, 339- 340,
332, 347, 353, 360, 365, 372 347-348, 351, 353, 358, 359, 365,
overgeneration 328 373, 374
P1 position 5, 9-12, 14-17, 43, 46, predicate vii, 2, 7, 25, 29-32, 35- 36,
52, 82, 99, 104-108, 112, 124, 40, 43, 57, 58, 60, 76, 124, 151-
145, 166, 170, 175, 189, 200, 212, 167, 169, 171-174, 182, 185, 187-
219, 226, 231, 328-330, 333, 343, 188, 198, 223-224, 262, 276, 281,
348-352, 354- 355, 367 284, 315, 325, 332-333, 347, 349,
paragraph 7-8, 92, 133, 136, 288, 373-374, 378
326 predicate frame 2, 29-32, 35-36, 40,
parataxis 226, 237 57, 76, 152, 160-162, 164, 171,
parenthesis 285 182, 374, 378
participant 49, 108, 121-122, 124, predicate phrase 7
145, 156, 172, 229, 236, 245, predication 25, 31-33, 36-37, 40, 42-
254-255, 372 43, 57-60, 120, 122, 124, 126,
pause 344, 346, 351, 353, 355 131, 152, 154, 159-162, 164-165,
perception 19, 179, 183, 214, 282- 174, 182, 219, 247, 260, 265, 267,
283, 291, 345, 370 270, 276, 284- 285, 293, 315,
percolation x, 164, 165, 169, 332, 335, 374, 378
334-335, 350, 354 prehension 152, 165, 175
performative 15, 78, 200, 203, 205- preposition 121, 351, 354
206, 270, 279, 287-288, 290-291, presentative 123, 223, 230-231
293-294 presupposition 37, 174, 222-223,
perspective 23, 25, 46, 55, 74, 76, 231, 235-236
84, 147, 164, 167, 187, 198- 199, primacy effect 212
204, 207, 217, 229, 247, 267, process x, 2, 10, 12, 17, 26, 34, 46-
276-277, 282, 292, 299, 323, 326, 47, 53, 73, 79-83, 90, 93, 95, 151-
338, 341, 343, 358- 359, 372, 378 153, 162-164, 166-170, 174-176,
phonology 7, 31, 46, 60, 78, 103, 180, 182, 186, 194, 197, 199,
212, 222, 226, 236, 327, 330- 212-213, 215, 217-218, 221, 232-
331, 359 234, 238, 275, 277, 280, 286,
pitch 233, 236 289-290, 295, 300, 314-316, 320-
placement 189, 212, 226, 304, 329, 321, 325, 329-332, 334, 337, 339,
334
Index of subjects 391
342-343, 345, 347-350, 357- 358, referential 6-8, 10-12, 15-16, 44, 47-
366-367, 373 48, 53-54, 60, 100, 169, 171, 185,
production viii, x, 2, 7, 17, 34, 46, 188, 190, 197, 200- 201, 219,
50-52, 60, 73-75, 78, 85, 87-88, 223, 229, 307, 340, 351, 355, 367,
151, 170-172, 179-182, 185, 187- 372
188, 192, 197, 212- 213, 233, referential phrase 7, 11-12, 16, 169
241, 275-277, 288-289, 292, 295, reflexive 8, 127-129, 145, 240, 368
327, 331, 337- 338, 343, 359, 367 relativism 344
profile 220-221, 228 representation vii, 5-7, 10, 15- 17,
prominence 15, 124-125, 184, 212, 25, 28, 37, 39, 43, 45-46, 48, 50,
217-219, 222, 226-233, 235, 318 52, 77, 79, 82-83, 87, 90, 95-102,
pronoun 28, 60, 118, 127-129, 132, 109-110, 114, 119, 132, 140, 142,
134, 137, 143, 145, 147, 158 151, 164-166, 170, 175, 182, 185-
properties 6, 27-28, 31, 38, 53, 56, 186, 189-192, 198, 200, 203-204,
74, 79, 84, 103, 108, 128, 130, 213- 214, 238, 267-268, 272, 275-
141-142, 144, 158, 181, 185, 213, 278, 280, 282, 286, 290-291, 295,
217, 224, 270-271, 276, 278, 280- 300, 308-309, 324-329, 331, 333-
281, 283-284, 286, 302, 309-310, 334, 337, 339, 341, 344-345, 347-
312, 319, 338-339, 349, 366, 370, 348, 355, 359, 366-370, 372-373
372-373 request 80-83, 186, 190
proposition 58-59, 96, 108, 120, 152, restrictor 152, 165
154, 164-165, 173-174, 182-183, Resumed Topic 132, 219
207, 220, 222-223, 232, 235-236, rheme 27, 222, 231, 237
246-247, 251, 262, 265, 267, 270, rhetorical 43, 51, 92, 101-102, 109-
276-277, 304, 315 110, 270, 308-309, 312- 313
prosody 11, 46, 120, 131, 145, 172, rigid 163, 189, 214
222, 235, 375 salience 34, 126, 131, 139, 158, 163-
protasis 249, 251, 257, 288 164, 183, 221-222, 226, 228, 231,
psychological adequacy 2, 26, 28, 235-236, 307
34, 38-39, 51-52, 86, 170, 179- satellite 42, 92, 165, 172, 181, 253,
180, 182, 193, 211, 337- 338 306, 333, 350-352, 354-356
qualification 279, 284-291, 293, 296 schema 76, 183, 214, 220-221, 316
recency effect 212, 226, 230 scope ix, 44, 59-60, 84, 90, 92, 94,
recursivity 162, 329, 340 101, 158, 171, 174, 187, 197, 204,
reference ix, 3, 8, 13, 15-16, 25, 43- 207, 217, 226-228, 247, 251, 265,
44, 49, 53-55, 91, 108, 117, 133- 276, 278, 280- 281, 284-285, 288,
135, 152, 169, 185, 190, 198, 202, 294, 306-307, 319, 326
212, 219-221, 229- 230, 234, 236, selection restriction 57, 78
276, 280-281, 288, 326, 340, 372- semantics ix, 3, 6, 10-11, 26-32, 34-
373 38, 46-48, 50, 54-55, 57-59, 75-
referent 38, 100, 119-139, 143- 144, 76, 86, 88, 100, 103, 114, 118,
157, 164, 172, 175, 183, 220-221, 141-142, 145, 153, 158, 161, 166,
224, 231, 235, 347, 373 170, 177, 187-189, 200, 212-213,
222, 233, 248-249, 251, 277-281,
392 Index of subjects
283, 285, 292-295, 298, 309-310, strategy 39, 53, 128, 130, 138, 143,
312, 316, 321, 325-326, 337, 339- 155-156, 168, 183, 190, 217, 224,
342, 344-346, 348-349, 351-352, 230, 232, 270, 277, 359
355, 358-361, 365, 368, 373 structural 35-36, 38, 111, 118, 155,
semantic memory 75, 88, 337, 339- 161, 204, 221, 223-224, 227, 229,
342, 344-346, 355, 360 233, 234, 270, 277, 292, 300,
Semiotic Grammar 267, 272 302-305, 310, 312, 314, 321, 368,
sentence 2, 5, 8, 15, 17, 26-28, 30, 370, 373, 375
34, 36, 40-41, 44, 48, 51, 56, 58, style 90, 96, 98, 107, 182, 307-309,
91, 100-101, 111, 118- 121, 126, 317, 319-320
128, 132, 149, 154, 156-159, 164, subact 185-186, 188, 190, 202, 204,
168, 174, 198, 201-204, 206-207, 206, 212
223-224, 236, 248, 285, 301, 305, subcategorization 358
310, 320, 326, 334-335, 337, 343, subject 32, 142, 160-161, 174, 180,
348-350, 353 187, 223-224, 230-232, 236, 333,
serialization 156, 158-159, 161, 164 335, 349, 358, 374
short-term memory 79, 232, 288, subjective modal 242-244, 247-252,
331, 340-341, 360 255-257, 260, 264, 266-268, 270,
sign 29, 56, 135, 168, 172, 249 277, 310, 314, 317- 318
speaker vii, ix-x, 2-3, 5-6, 8, 10- 12, subsidiary 5, 96, 110, 181, 201- 202
15-16, 34, 44, 46-48, 50, 53, 60, SubTopic 126, 235
74, 77, 80, 83, 90, 94, 99, 108, syllable 183, 191-192, 375
118-119, 121-122, 127-128, 131, syntax 27, 31-32, 34-38, 47-48, 50,
133-134, 136-140, 143, 145, 155, 54-55, 57, 60, 74, 78, 100, 142,
174, 180, 183-189, 191-192, 198- 145, 158, 160-161, 164, 166, 187-
202, 213-215, 220, 224, 231-232, 189, 192, 214, 222, 226, 231, 280,
234-236, 247, 249-252, 254-259, 285, 310, 324-325, 331, 334, 337-
261-263, 265-266, 270, 277, 282, 338, 341, 347, 349, 351, 354, 365,
284-289, 291, 294, 295, 325-326, 373, 377, 378
332, 337-340, 342-343, 346, 351, Systemic Functional Grammar 94,
355, 357, 359, 361, 367, 369, 103, 213, 267, 295
372-373 Tail 206
speaker model vi, 193, 237, 325, template 153-154, 163, 169-170,
338, 340, 357 188, 190, 265, 329-330, 333, 335,
speech act 40, 43-44, 91-92, 95, 97- 343, 350
98, 102, 110, 174, 197, 247, 277, tense 40, 58, 79, 91, 106, 110, 118,
286-287, 289, 294- 295, 310, 356 173, 187, 192, 244, 247, 259,
speech situation 3, 11, 58, 336, 340, 261-263, 265-267, 270, 277, 279,
343 307, 327, 333
state of affairs 6, 30, 57, 60, 123, term 32, 35-37, 42, 57-60, 75, 90,
140, 213, 223, 245, 254, 282, 341, 110-111, 119-130, 132-133, 137,
366-367 141, 144-145, 152, 155, 162, 166,
story 57, 230, 249, 295, 308, 313, 172-173, 185, 188, 222, 266, 285,
340-341, 346, 361
Index of subjects 393
288, 300-301, 304, 313-314, 317, 214, 277, 303-305, 307-310, 312,
325, 329, 333, 335, 352, 372 316, 321, 324-329, 331, 333-334,
text x, 13, 42, 78, 89, 94, 105, 118, 337, 339, 344, 347-349, 357-360,
121, 126, 135, 144, 147, 159, 217, 365-370
220-221, 224, 229, 234, 236, 269, unification 80, 299, 334, 349, 359
300-303, 305, 309-310, 312- 316, upward layering 2-3, 8, 43, 92-93,
318, 320, 346, 358 96, 197, 299, 310, 325, 348, 372
theme 27, 249 utterance 2, 6, 27, 44, 54, 90, 94, 97-
thetic 120, 122, 139, 143-144, 149, 98, 101, 116, 118-122, 124, 126-
165, 170, 174, 223 127, 131, 138-140, 143- 144, 150,
thinking-for-speaking 199 166, 169-170, 172, 183, 185, 188,
top-down vii, x, 2, 10, 12, 17, 73, 190-192, 198- 199, 203, 211-212,
117, 153, 161, 182, 197-199, 212, 223-224, 227, 235-236, 242, 248,
243-244, 264, 266-268, 269, 325, 251, 259, 265-267, 280-281, 285-
329, 331, 337, 365 288, 290, 295, 300, 324-325, 331,
Topic viii, 77, 91, 106-108, 110-111, 337, 341-343, 359-360, 370
117, 120-122, 124, 127, 131, 133, valency 78, 162, 185, 192
136, 143-144, 158- 159, 217, 229, variable 7-8, 30, 40, 43-44, 53, 59-
303, 346, 369 60, 152, 236, 270, 344, 347, 349,
topicality 91, 120, 122, 124, 138- 355, 366
139, 143, 145-146, 186, 218- 219, verbal interaction 25-26, 31, 34, 39,
222, 308, 343, 347 44, 48, 50, 60, 153, 182, 211, 313,
transaction 82, 96, 106, 108-109, 340, 366, 369-371
111, 187 viewpoint 58, 153, 189, 199, 229,
tree 10, 96, 199, 202, 329-332, 335- 234, 240
336, 350-352, 354-356, 359 vocabulary 74-78
typology 25, 30-31, 56, 59, 162-163, voice 94, 104, 110, 180
167, 169, 211, 270, 300, 305, 312, volition 245, 254-256, 293
316-319, 321, 328, 332 West Coast Functionalism 213
undergeneration 327-328 word order (see also constituent or-
underlying structure vii, 2, 26, 28, der) 122, 155, 188, 234, 304, 375
30, 34, 36-37, 39, 45, 47, 49-51, working memory 75, 79, 337, 339-
56, 58, 77, 98, 132, 142, 151-152, 340, 342-343, 346-348, 351, 353-
160, 163, 170- 171, 174, 182, 355, 360