Kees Hengeveld & J. Lachlan Mackenzie. Functional Discourse Grammar. A typologically-
based theory of language structure. Oxford: Oxford University Press 2008, xxiii+503. (ISBN
Reviewed by Johan van der Auwera and Frank Brisard (University of Antwerp)
The term ‘Functional Grammar’ was never a good proper name or perhaps it was too good and
thus suffered from its own success. Depending on one’s background one will think of the
theories of André Martinet (1979), Alexander Bondarko (1991), Michael Halliday (Halliday &
Matthiessen 2004) or Simon Dik (Dik 1978, 1997). These theories indeed show family
resemblances and also resemble theories that do not contain the adjective ‘functional’ (especially
Role and Reference Grammar), but they are rather different. The one that is at issue here,
associated with Simon Dik, comes in three versions (with elements of a protoversion in Dik’s
pre-1978 work). For the last version, the leading designers, Kees Hengeveld and Lachlan
Mackenzie, have chosen to change the name somewhat, from ‘Functional Grammar’ (FG) to
‘Functional Discourse Grammar’ (FDG). Strictly speaking, there was no need to do this.
Hengeveld and Mackenzie could have retained the simple term ‘Functional Grammar’, as Dik
(1997) did when he overhauled his own first version (Dik 1978). The changes between 1997
Functional Grammar and 2008 Functional Discourse Grammar are certainly substantial, but so
were the ones between 1997 Functional Grammar and 1978 Functional Grammar. The name
change does have two advantages, though. First, it profiles the theory in a more succinct way:
just as ‘Lexical-Functional Grammar’ is guaranteed not to be confused with e.g. the work of
Halliday, the addition of the word discourse will give this third-generation ‘Functional
Grammar’ the better and more unique reference. Second, it highlights that the major difference
between the 1997 and the 2008 versions indeed concerns the treatment of discourse or, if one
prefers, pragmatics. Another effect of the name change is that it marks the end of the Dik era. It
seems to be difficult to position a linguistic theory without one or perhaps two or three linguists
becoming the primi inter pares. The text-book act of Hengeveld & Mackenzie (2008) confirms
the linguists with those names as these primi.
The FDG book has a substantial introductory chapter and then four additional chapters,
one for each of the four levels of the FDG model. The introduction describes the basic features of
the model and it contextualizes the theory relative to functionalism, typology and language
processing. As to functionalism, the term ‘refers to an approach to linguistic analysis that is
based on the belief that the properties of linguistic utterances are adapted to those communicative
aims which the language user, in interaction with other language users, seeks to achieve by using
those utterances’ (p. 26). In this respect FDG is like its FG predecessors, even though in the
latter’s incubation period (Dik 1968: 160) the term ‘function’ rather referred to the primacy of
the syntactic functions of subject and object as irreducible to phrase structure (contrary to what is
still being claimed in generative grammar) or to semantic/pragmatic functions. Next, FDG is
‘typologically neutral’ (p. xi; i.e., applicable to all languages) in that it ‘is capable of providing a
framework for the enunciation and comparison of universals and of offering lines of explanation’
(p. 32). This, too, is fully in line with the FG predecessors, even from its incubation: the
functional theory (functional, because it gave pride of place to subjects and objects) that the early
Simon Dik liked most was tagmemic theory, for it was applicable to a wide variety of languages
(Dik & Kooij 1970: 151). Furthermore, as in FG the primitives and levels are taken to be
universal, at least in the sense that they are available for every language. And just like FG, at
least since it appeared in Dik’s 1968 book, FDG is a quasi-productive theory: it reflects the
process of language production in its basic theory. It does so because, if one accepts the tenet
that functional pressures may shape grammar, then language processing mechanisms figure
among the prime candidates for consideration (Butler 2003: 12). The nature of the vocal(-
auditory) channel and that of the associated psychological processing mechanisms or, in other
words, ‘the means by which languages are implemented’ (Dik 1986: 21) are bound to be
responsible not just for many of the phonological characteristics of languages (ease and economy
of articulation, the perceptibility of sounds), but also for certain syntactic principles, such as that
of end-weight (cp. Dik’s Language-Independent Preferred Order of Constituents). On the whole,
for readers that merely need an introduction to the theory, chapter 1 is excellent, as is Hengeveld
& Mackenzie (to appear).
Chapters 2 and 3 concern the so-called ‘Interpersonal’ and ‘Representational’ levels of
the theory, good for 81 and 153 pages and, as stated in the book’s preface, whose discussion
represents a significant advance on FG. The Morphosyntactic and Phonological levels, which are
said to have their own principles of organization and are thus not seen as merely ‘expressive’, go
to Chapters 4 and 5 and take up 138 and 41 pages. The size of the chapters is indicative of the
importance of each level. Thus FDG is not the average phonologist’s first choice and it is clear
that FDG is a strongly meaning-based account of grammar, for this is what Chapters 2 and 3 are
all about. In separating the ‘Interpersonal’ and the ‘Representational’ level, FDG is notably in a
position, or so it claims, to investigate the hitherto largely underemphasized complexities of the
former, and the grammatically relevant interactions between these two levels. All chapters have a
parallel structure, as is also the case for the levels themselves. The levels that are most parallel
are the first two as they both concern meaning in a wide sense, but whereas the ‘Interpersonal’
level concerns pragmatic aspects, the ‘Representational’ one concerns semantic ones, with
‘semantic’ in a more restricted sense of the term. The distinction between semantics and
pragmatics is, of course, theory-dependent and for FDG pragmatics involves all aspects of
speech acts, interjections, specifications of Speaker and Addressee (even their number and
gender), topic, focus, contrast, reported speech, the difference between referring and so-called
‘ascribing’ (roughly equivalent to ‘predicating’), (in)definiteness and (non-)specificness, and
also attitudinal uses of adjectives such as old in (1).
I am sorry for old Bill.
Semantics involves the meanings of simple and complex units ‘in isolation from the ways these
are used in communication’ (p. 129). Distinctions include the ones between properties,
individuals and states of affairs, certain dimensions of modality and evidentiality, tense and
aspect, polarity, participant roles such as Actor and Undergoer, and lexeme classes such as Noun
On all four levels entities are claimed to involve variables - the only obligatory
component -, heads, modifiers, operators, and functions. They are represented in a quasi-logical
way, familiar from Functional Grammar.
variable: head (variable): modifier(s) (variable)
To illustrate this with a Representational-level meaning, a nominal like this green snake involves
some entity (a variable) that is a snake (head), but not just any snake, i.e., a green one (modifier).
The meaning carried by the demonstrative this is represented like a quantifier in predicate logic,
and a function, like Actor or Undergoer (as in This green snake bit me versus He killed this green
snake), is added as a subscript to the whole entity.
(operator(s) variable: head (variable): modifier(s) (variable))function
(3) is the maximal formula, in the sense that not every component needs to be present. And what
is most crucial is that every phrase will be represented on four levels. Next to the
Representational-level information sketched in the above for the green snake, the Interpersonal-
level description will specify whether or not the green snake has a referential function, for
instance, or whether the green snake is identifiable for the hearer and whether it functions as a
topic or focus. The Morphosyntactic level will make clear that English has green snake rather
than snake green, and the Phonological level will specify the phonological features. This
quadrilevel strategy is new compared with Functional Grammar. Interestingly, the four FDG
levels partially correspond to two FG 1997 levels or, better, components. A good part of the
morphosyntactic analysis and all of the phonological analysis were the province of the so-called
‘Expression rules’ (testifying to the more derivational orientation of original FG) and the input to
these rules was just one representation, containing all semantic as well as pragmatic distinctions
and even some syntactic ones, viz. the specification of the syntactic functions of subject and
object. Despite their names the latter were then considered imbued with semantics of their own,
viz. that of the perspective on the state of affairs (Dik 1997 vol. 1: 247 ff), a hypothesis that has
very nearly been given up in FDG (except for what could be taken as a ‘relic’ statement on pages
35-36). As to the difference between what FDG calls ‘interpersonal’ and ‘representational’
aspects of meaning, FG 1997 was able to represent these too, but of course these aspects figured
on just one level, and there are two now. At least for many phenomena the difference between
one more complex representation and two more simple ones probably does not matter. The basic
reason for the strong divide between pragmatics and semantics is that (i) the former relates in a
more direct way to intentions of the Speaker, (ii) FDG is organized as a quasi-productive model
that starts from the Speaker’s intention, and therefore (iii) ‘FDG takes the functional approach to
language to its logical extreme: within the top-down organization of the grammar, pragmatics
governs semantics […]’ (p. 13).
As a matter of fact, the model presents itself as the grammatical, ‘constructive’ part of a
wider theory of verbal interaction. The question then rises: how extreme, in fact, is this position,
which might arguably be called ‘usage-based’ (since it starts from the communicative use of bits
of language or Discourse Acts)? In FDG terms, this question directly translates into the problem
of the relationship between the grammatical component of the theory and the other components,
primarily the contextual and the output ones (though the relation with the conceptual component
is relevant to related discussions of linguistic relativity). Now, in contrast with such truly usage-
based theories as Cognitive Grammar, various Construction Grammars, and other models within
interactional linguistics, FDG explicitly maintains the difference between the lexical and the
grammatical (p. 9), which it needs to motivate the distinction between modifiers and operators. It
also proposes a neat contrast between ‘systematic’ (grammatical) and ‘statistical’ aspects of
context (p. 11), while various contemporary grammars would claim exactly the opposite, viz.,
that grammar actually bears the imprint of the statistical properties of its genesis or ‘emergence’.
This illustrates the theory’s ‘conservative stance’ (Butler 2008) on what to call grammatical facts
of language, and it raises questions about its self-proclaimed extreme functionalism. In fact,
despite the quote above, Hengeveld & Mackenzie position FDG between what they call ‘radical
functionalism’ and ‘radical formalism’. It shares with formalist approaches the exclusive focus
on ‘criterial’ (p. 28) properties of language, defined as those conventional aspects of language
use that introduce systematic oppositions within the grammar. This equation between grammar
and system, in itself quite unproblematic, is all the more formally oriented in FDG, because it
considers such a system as by necessity ‘covert’, ‘governed by rules that predict the form taken
by idealized linguistic units’, and thus ultimately not recoverable ‘directly from data’ (p. 26). All
of this effectively turns the model into a ‘form-oriented function-to-form model’ (ibid.). What
remains functionalist in this approach is the promise that the linguistic structures revealed (the
Speaker’s knowledge of conventional units and their combinations) will be explained by relating
them to the uses they are put to (the Speaker’s communicative aims), just like a bicycle is best
described, not only in relation to its structure, but also in relation to human purposes of
But if functionalism is somehow also to be equated with the quest for a meaning-based
grammar, then surely the most extreme position would be to reconsider the foundational status of
the autonomy of grammar itself. Indeed, it seems reasonable to assume that every phonological
structure - every part of the bike, including non-criterial ones such as lights or a bell - is
motivated, i.e. associated with a semantic structure (a communicative aim). However, this
possibility is considered by only few, and not even by most of those who situate themselves in
the usage-based camp (cf. Langacker 2005). It is tempting, apparently, to retain some notion of a
‘constructional’ meaning that is irreducible to semantics at the very core of one’s conception of
grammar, regardless of one’s ultimate position on the functionalist cline (more, or less, formal).
FDG, like so many others then, seems to think that to acknowledge the fully symbolic nature of
grammar (which it does not) would be tantamount to eliminating grammar or claiming that it
does not exist (‘denying the cognitive reality of linguistic structure’, p. 26), which in our view is
evidently not true: a grammar still has to be learned by children and described by linguists as
such, because it does not emerge or fall out automatically, so to speak (i.e., it cannot be fully
predicted from meaning and other independent factors). Thus, every theory, no matter how
radically functionalist, will always have the need for some specifically morphosyntactic level of
description and analysis, without contradicting its own assumptions about the non-autonomous
nature of grammar. Again, the only question is: is this morphosyntax itself necessarily imbued
with meaning, i.e. symbolic? We shall see in its treatment of ‘tense copying’ that sometimes in
FDG it is not.
In FDG, to uphold a radically reductionist position one would have to assume that all
grammatical rules should have some relevance at the Representational level and above. A case in
point is the phenomenon of tense copying or sequence of tenses (Leufkens 2009), which FDG
describes in terms of ‘operator agreement’ at the Morphosyntactic level. In (4a), the operator
‘Past’ in the main clause is copied to the embedded clause, which takes over its tense
specification; there is no corresponding operation at the Representational level. If the Speaker
chooses not to follow the tense-copying rule (4b), the operator at the Morphosyntactic level is
interpreted as an absolute tense operator at the Representational level.
a) Ursula said that Manfred was gay.
ep1: (e1: (f1: [ (f2: say) (x1)A (ep2: (e2: (f3: [ ((f4: gay) (x1)U ] )))U ] ))
(Vp1: (Vw1: say : <Past>)) … (Vp2: (Vw2: is : <Past>))
b) Ursula said that Manfred is gay.
ep1: (e1: (f1: [ (f2: say) (x1)A (Pres ep2: (e2: (f3: [ ((f4: gay) (x1)U ] )))U ] ))
(Vp1: (Vw1: say : <Past>)) … (Vp2: (Vw2: is : <Pres>))
First of all, it is strange to see that tense copying, a putative grammatical rule, can be violated on
the basis of a semantic condition, viz., that the situation described in the subordinate clause still
continues in the present. Furthermore, the analysis of (4a) might work for readings whereby a
past-tense form in the embedded clause sounds like an absolute past tense (i.e., Manfred is no
longer gay). In this case, the hearer does interpret the tense form at the Representational level.
But for strict cases of tense copying, FDG simply provides no corresponding unit in RL, which is
also strange, given that one might expect a hearer always to look for a meaning attached to a
tense form (psychological plausibility). Assuming, however, that a form like was in (4) is really a
relative tense, indicating simultaneity (and remaining undecided on the matter of continuing
applicability), it would have to operate at the Representational, rather than the Morphosyntactic,
level. Within the Representational level, relative tense is said in FDG to work on a different layer
than absolute or deictic tense: the former belongs to the layer of the State-of-Affairs (e), while
the latter is associated with Episodes (ep), or ‘one or more States-of-Affairs that are thematically
coherent’ (p. 157) in time, location, and participants. The relative-tense analysis of Ursula said
that Manfred was gay requires, then, that there should be an operator for e2 at the
Representational level, as in (5).
(5) RL Past ep1: (e1: (f1: [ (f2: say) (x1)A (ep2: (Sim e2: (f3: [ ((f4: gay) (x1)U ] )))U ] ))
The operator in the embedded clause is of a different nature now than that in the main clause,
and hence cannot be the consequence of mere operator agreement.
Perhaps the authors do not accept the idea that English also has relative tense markers
(proposed by, among others, Declerck & Tanaka 1996), but even then the fact remains that tense
copying, as worked out in FDG, is a typical blind grammatical rule in the formalist sense, for
which no ultimate explanation is offered and which in turn offers no explanation for the apparent
gaps it exhibits. It accepts the possibility of certain tense forms having no meaning at all, which
is hard to maintain, especially among more cognitively oriented linguists. One could easily
imagine semantic accounts of tense behavior, including in indirect speech, as always depending
on choices the Speaker makes, so that there are no blind spots to be explained as such. Tense
models that talk about ‘intensional domains’ (Declerck 1991) or ‘mental spaces’ (Fauconnier
1997) do just that. This would also lead to a possibly more unified treatment of both deictically
used and discursively used tenses, and it would generally force grammarians, in the end, to
consider (more) carefully the links and analogies between the non-linguistic situation of speech
(context) and the linguistic discourse (co-text) in which an utterance is simultaneously produced
The term ‘discourse’ in the proper name ‘Functional Discourse Grammar’ is not quite
obvious. Basically, it refers again to the decision to organize the grammar from the top-down
perspective, which has pragmatics at the top. The basic unit is essentially the speech act, which
in FDG is called ‘Discourse Act’. Crucially, FDG is not a discourse or text grammar. It is not
designed to unravel the macro text structure. However, if elements of macro-structure are
reflected on what is traditionally called the ‘sentence’ or ‘clause’ level, FDG will have to account
for them. FDG is thus on the whole still a ‘grammar’ in much the traditional sense, and even a
sentence grammar, as indeed much of 20th-century grammar has been, at least since the Prague
Circle functionalist Vilém Mathesius held that ‘the basic element of the communicative process
[…] is the sentence’ (quoted from Mathesius 1975: 79; see also Sornicola to appear).
Interestingly, the authors plead against the concept of the sentence (pp. 17, 38), and they indeed
succeed in avoiding the term ‘sentence’ altogether, but it is not obvious why notions such as
‘discourse act’ are fundamentally any better, except perhaps in recalling notions of ‘pragmatic
adequacy’ (relating to the use of linguistic expressions) and ‘semantic equivalence’ which Dik
used to discuss, e.g., constructional variants in the same pragmatic setting (say, question-answer
sequences). It is also interesting that FDG has very little to say about the lexicon and the rules
that are operative there. Here FDG differs from FG, but it may be a matter of time before the
FDG light will shine on the lexicon too. At the moment, it seems as if FDG is primarily trying to
get away from the projection-based orientation it ascribes to Dikkian FG: rather than having the
construction of linguistic expressions be constrained by the syntactic requirements of lexical
items (predicate frames), it proposes to separate Frames and Lexemes, thereby removing the
need for Frames to be associated with any descriptive content (i.e. to have meaning; García
Velasco & Hengeveld 2002). Instead, Frames are said to act as indicators triggering contextually
appropriate interpretations for a given Lexeme. To explain cases of coinage and coercion (e.g.
inserting a proper noun into a verbal Frame, as in He Ben Johnsoned his promising swimming
career), FDG has opted to rely much more heavily on general Gricean principles of
interpretation, underlining ‘the idea that meaning is created cooperatively by speakers on the
basis of common knowledge, assessment of each others’ beliefs and contextual clues’ (García
Velasco 2008: 20). The relevant material (clues and beliefs) will be found in the model’s
contextual and conceptual components, thus both demonstrating the need for these components
and reflecting ‘the choice the Speaker often has in describing one and the same entity through a
variety of lexemes with different connotations and/or denotations’ (p. 19). This, too, goes against
much of what is assumed in current constructional approaches.
The typological orientation of FDG is at least as important as the functionalist one, as
befits a book with the subtitle ‘A typologically-based theory of language structure’. FDG is
explicitly offered as a model that is applicable to all languages (e.g., p. 36), and indeed probably
each entity, construction, level or rule is illustrated from a wide variety of non-English and more
generally non-European languages. The authors also successfully realize part of the further
desire to use FDG descriptions of various languages as input for comparison, which then allows
them to make generalizations about what is possible and impossible and frequent and infrequent.
These generalizations often take the form of implicational hierarchies. (6) is an example: it deals
with Subject assignment and says that any language allowing Subject assignment to a category
on the right of the hierarchy will allow it to a category on the left.
Generalizations of this type are also the bread and butter of typology itself, and thus many
grammarians of the FDG denomination are also typologists. Yet the relevance of FDG work to
typology has a paradoxical flavor. Most if not all of the implicational generalizations one will
find in FDG work can be understood by non-FDG linguists with minimal understanding of FDG.
FDG work thus achieves a cross-theoretical significance and this is excellent. But isn’t it also
somewhat not good? At least for typology there might not be any strong need for choosing FDG
rather than another model, such as simple FG, or indeed just ‘basic linguistic theory’ (p. 36) and
if that should be too restrained then ‘framework-free grammatical theory’ (Haspelmath to
This book impresses the reader by its coherence, clarity, amplitude and typological
relevance. Theoretical positions are often plausible and always interesting. Therefore the book
will be welcomed by many more than the members of the ‘FDG community’
(www.functionalgrammar.com). It reads easily. The book is beautifully produced. There are
three adequate indexes (language, name, and subject). Features that are bound to be controversial
are the following. First, despite the enormous amount of work done since Lehmann (1982), the
realization that the relation between lexical and grammatical categories (and also between
various grammatical categories) is one of gradience, and the fact that Hengeveld & Mackenzie
(2008) are fully aware of this, they nevertheless assume a strict distinction between lexicon and
grammar (p. 9). In this respect FDG is no different from its FG predecessors, which had received
criticism for this reason too (e.g., van der Auwera 2003: 117). Second comes the somewhat
unclear relation of the model to psycholinguistic production models. On the one hand, FDG
reflects real production in the sense that it describes linguistic construction in terms of a flow that
starts from prelinguistic Speaker’s intentions and ends with the articulatory output. On the other
hand, though, FDG is not really a model of the Speaker (the old ‘Natural language User’) in the
sense of, e.g., Levelt (1989), with concrete predictions at the level of online speech production
and processing/interpretation. So FDG is something in between, and in any case definitely more
oriented towards production issues, so that this particular aspiration of the model still remains
largely theoretical, as in the past. Third, because of the comprehensiveness of the model, the
formalizations can be very complex. These formalizations are intended as ‘a means to the end of
the insightful analysis of linguistic phenomena’ (p. 42). There will be readers who find the
complex succession of brackets, variables and subscripts to be a hindrance rather than a help.
van der Auwera, Johan. 2003. La grammaire de la modalité. In Mohammed Jadir, (ed.)
Développements récents en Grammaire Fonctionnelle, 109-120. Mohammedia: Université
Bondarko, Alexander V. 1991. Functional Grammar: A field approach. Amsterdam: Benjamins.
Butler, Christopher S. 2003. Structure and function: A guide to three major structural-functional
theories. Amsterdam: Benjamins.
Butler, Christopher S. 2008. Interpersonal meaning in the noun phrase. In Jan Rijkhoff & Daniel
García Velasco (eds.) The noun phrase in Functional Discourse Grammar, 221-261. Berlin:
Mouton de Gruyter.
Declerck, Renaat. 1991. Tense in English: Its structure and use in discourse. London: Routledge.
Declerck, Renaat & Kazuhiko Tanaka. 1996. Constraints on tense choice in reported speech.
Studia Linguistica 50. 283-301.
Dik, Simon C. 1968. Coordination: Its implications for the theory of general linguistics.
Dik, Simon C. 1978. Functional Grammar. Amsterdam: North-Holland.
Dik, Simon C. 1986. On the notion “functional explanation”. Belgian Journal of Linguistics 1.
Dik, Simon C. 1997. The theory of Functional Grammar. 2 volumes. New York: Mouton de
Dik, Simon C. & Jan G. Kooij. 1970. Beginselen van de algemene taalwetenschap. Utrecht:
Fauconnier, Gilles. 1997. Mappings in thought and language. Cambridge: Cambridge University
García Velasco, Daniel. 2008. Innovative coinage: Its place in the grammar. In Christopher S.
Butler & Javier Martín Arista (eds.) Deconstructing constructions, 3-23. Amsterdam: John
García Velasco, Daniel & Kees Hengeveld. 2002. Do we need predicate frames? In Ricardo
Mairal Usón & María J. Pérez Quintero (eds.) New perspectives on argument structure, 95-123.
Berlin: Mouton de Gruyter
Halliday, Michael A. K. & Christian M. I. M. Matthiessen. 2004. An introduction to Functional
Grammar. London: Hodder Arnold.
Haspelmath, Martin. To appear. Framework-free grammatical theory. In Bernd Heine & Heiko
Narrog (eds.) The Oxford handbook of linguistic analysis. Oxford: Oxford University Press.
Hengeveld, Kees & J. Lachlan Mackenzie. To appear. Functional Discourse Grammar. In Bernd
Heine & Heiko Narrog (eds.) The Oxford handbook of linguistic analysis. Oxford: Oxford
Langacker, Ronald W. 2005. Construction Grammars: Cognitive, radical, and less so. In
Francisco J. Ruiz de Mendoza Ibáñez & M. Sandra Peña Cervel (eds.) Cognitive Linguistics:
Internal dynamics and interdisciplinary interaction, 101-159. Berlin: Mouton de Gruyter.
Lehmann, Christian. 1982. Thoughts on grammaticalization: A programmatic sketch. Köln:
Institut für Sprachwissenshaft der Universität Köln. [1995, revised edition. München: Lincom
Leufkens, Sterre. 2009. Tense copying in Functional Discourse Grammar. Research Master
paper, University of Amsterdam.
Levelt, Willem J.M. 1989. Speaking: From intention to articulation. Cambridge, MA: MIT
Martinet, André. 1979. Grammaire fonctionnelle du français. Paris: Crédif.
Mathesius,Vilém. 1975. A functional analysis of present day English on a general linguistic
basis. Ed. by Josef Vachek. The Hague: Mouton.
Sornicola, Rosanna. To appear. European functionalism. In: Bernd Kortmann & Johan van der
Auwera (eds.) The field of linguistics: Europe. Berlin: Mouton de Gruyter.