Jump to ContentJump to Main Navigation
Functional Discourse GrammarA Typologically-Based Theory of Language Structure$

Kees Hengeveld and J. Lachlan Mackenzie

Print publication date: 2008

Print ISBN-13: 9780199278107

Published to Oxford Scholarship Online: September 2008

DOI: 10.1093/acprof:oso/9780199278107.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 18 October 2019



(p.1) 1 Introduction
Functional Discourse Grammar

Kees Hengeveld (Contributor Webpage)

J. Lachlan Mackenzie (Contributor Webpage)

Oxford University Press

Abstract and Keywords

The chapter presents Functional Discourse Grammar (FDG) as part of a wider theory of verbal interaction, specifying its distinguishing features and detailing its architecture and notational conventions. It is explained how the grammar can be implemented in linguistic analysis and how it relates to linguistic functionalism and to language typology.

Keywords:   top-down organization, Discourse Act, Conceptual Component, Contextual Component, Output Component, formulation, encoding, levels of representation, layering

1.1 Functional Discourse Grammar

This introduction provides a general overview of Functional Discourse Grammar (FDG) as part of a wider theory of verbal interaction. It starts out by describing various distinguishing features of the FDG model in Section 1.2. Section 1.3 goes on to present the architecture of FDG, introducing notions that will be expanded and justified in the remaining chapters of the book and explaining in general terms how the grammar can be implemented in linguistic analysis. The following Section (1.4) discusses the relation of FDG to linguistic functionalism, the relevance of FDG for language typology and various methodological prerequisites. The penultimate section, 1.5, sets out various notational conventions to be observed in the following chapters, which are briefly previewed in Section 1.6.

1.2 Basic properties

1.2.1 Introduction

There are a number of distinguishing features that set off Functional Discourse Grammar from other structural‐functional theories of language (Butler 2003). These features, which are discussed in the following sections, are the following: FDG has a top‐down organization (1.2.2); FDG takes the Discourse Act as the basic unit of analysis (1.2.3); FDG includes morphosyntactic and phonological representations as part of its underlying structure, alongside representations of the pragmatic and semantic properties of Discourse Acts (1.2.4); and, as the Grammatical Component of the theory of verbal interaction, FDG systematically links up with a Conceptual, a Contextual, and an Output Component (1.2.5).

1.2.2 Top‐down organization

FDG starts with the speaker's intention and then works down to articulation. This is motivated by the assumption that a model of grammar will be more effective the more its organization resembles language processing in (p.2) the individual. Psycholinguistic studies (e.g. Levelt 1989) clearly show that language production is a top‐down process, which starts with intentions and ends with the articulation of the actual linguistic expression. The implementation of FDG reflects this process and is accordingly organized in a top‐down fashion. This does not mean that FDG is a model of the speaker: FDG is a theory about grammar, but one that tries to reflect psycholinguistic evidence in its basic architecture (cf. 1.2.5 below).

Two major operations have to be distinguished in the top‐down construction of utterances: FORMULATION and ENCODING. Formulation concerns the rules that determine what constitute valid underlying pragmatic and semantic representations in a language. Encoding concerns the rules that convert these pragmatic and semantic representations into morphosyntactic and phonological ones. The operation of Formulation involves three interlinked processes: the selection of appropriate frames for the Interpersonal and Representational Levels; the insertion of appropriate lexemes into these frames; and the application of operators symbolizing the grammatical distinctions required in the language under analysis. Encoding also involves three processes: the selection of appropriate templates for the Morphosyntactic and Phonological Levels; the insertion of free and bound grammatical morphemes; and the application of operators that play a role in the process of articulating the output of the grammar. Details will emerge from the relevant chapters.

Our presentation, in progressing from formulation to encoding and within encoding from morphosyntax to phonology, clearly mimics the sequence found in production. Despite this seductive analogy between the architecture of FDG and the processes of speech production, it is important to emphasize, as pointed out by Hengeveld (2004b: 366–7), that FDG is a ‘model of encoded intentions and conceptualizations’ rather than, as is Levelt's ‘blueprint for the speaker’ (1989: 8 ff.), a model of language production. FDG aims to understand how linguistic units are structured in terms of the world they describe and the communicative intentions with which they are produced, and models this in a dynamic implementation (Bakker and Siewierska 2004) of the grammar, i.e. the sequence of steps that the analyst must take in understanding and laying bare the nature of a particular phenomenon. This is how our discourse in this book is to be understood, for example where we remark that some operation precedes another one, or that two units are available simultaneously.

Note that, although the presentation of the FDG model will focus on the generation of utterances, the model could in principle be turned on its head to account for the parsing of utterances. It is clear that listeners analyse phonetic input into phonological representations, which are subsequently grouped into morphosyntactic constituents, from which meaningful representations are then constructed.

(p.3) The top‐down organization of the model is a precondition for a grammatical theory that aims at describing discourse units rather than clauses. In a discourse‐oriented model the clause is just one of the options that the speaker can use to contribute to the ongoing discourse, for which reason formulation has to precede encoding. This is the topic of the next section.

1.2.3 Discourse grammar

There are many grammatical phenomena that can only be interpreted in terms of units larger than the individual clause. Examples of these are narrative constructions, the use of discourse particles, anaphorical chains, and tail‐head linkage. By way of example, consider the following instance of tail‐head linkage in Tidore (van Staden 2000: 275):

(1) …turus jafa cahi     saloi ena=ge turus

…then Jafa carry.on.the.back basket 3.NH=there then

ena=ge paka ine.    Ine   una oka koi…

3.NH=there ascend go.upwards go.upwards 3.SG.M pick banana

‘… then Jafa carried the basket upwards and picked the bananas…’

“… then Jafa carried the basket and went upwards. Went upwards he picked the bananas …”

In many Indo‐Pacific languages there are several grammatical phenomena that are a faithful and direct reflection of discourse organization. In Foley's (1986: 176) words: ‘A text is a coherent linking of clauses and sentences, and this coherence is achieved by rules of the language which state how clauses and sentences can be joined’. Example (1) illustrates one of these linking devices. Episodes within stories are in Tidore often realized as single linguistic expressions containing strings of clauses. The linguistic expressions are linked to each other by means of tail‐head linkage: the last verb of the one linguistic expression is repeated as the first verb of the next linguistic expression, as illustrated in (1).

The crucial point here is that, as stated in the quotation from Foley (1986), phenomena such as tail‐head linkage are governed by rules of the language and thus form part of the grammatical system as it applies to narratives. Grammatical phenomena like these thus clearly show the need for a grammatical model that allows for the treatment of units larger than the individual clause and of the relations that obtain between and within these units.

As argued in Mackenzie (1998b), the need for a discourse‐oriented grammar also becomes apparent when units smaller than a clause are considered. The following examples illustrate what he treats as holophrases of various types: (p.4)

  1. (2) (What are you eating?) A donut.

  2. (3) Congratulations!

  3. (4) Oh John!

The answer in (2), the exclamation in (3), and the vocative expression in (4) all take a non‐clausal form. Yet in the appropriate circumstances they all count as full and complete contributions to the discourse. In fact, any further elaboration of (2), for example, would lead to a relatively less natural exchange. These utterances are accordingly not interpreted as reduced clauses, but as being non‐clausal right from the start. The model should thus find a way of dealing with non‐clausal utterances which recognizes the fact that they constitute fully grammatical discourse units.

The conclusion that FDG draws from the facts discussed in the preceding sections is that the basic unit of discourse is not the clause but the Discourse Act. Discourse Acts combine into larger discourse structures, such as Moves. These larger structures account for the units larger than the individual clause discussed above. On the other hand, Discourse Acts may be manifested in language as clauses, but also as fully grammatical clause fragments, phrases or words. The latter point is a crucial one: it requires the grammatical model to be capable of mapping the unit of Discourse Act onto morphosyntactic units of various kinds. This mapping procedure in turn requires a top‐down approach.

Moves and Discourse Acts are notoriously difficult to define. Anticipating a more extensive discussion in Chapter 2, we here use the definitions offered in Kroon (1995: 65–6; see also Hannay and Kroon 2005), who following Sinclair and Coulthard (1975) defines a Move as ‘the minimal free unit of discourse that is able to enter into an exchange structure’ and a Discourse Act as ‘the smallest identifiable unit of communicative behaviour#x2019;. Note that a Move consists of a single central Discourse Act, which may be supported by one or more Subsidiary Discourse Acts.

1.2.4 Levels of representation

The organization of Moves and Discourse Acts is dealt with at one level of the grammar, the Interpersonal Level. This is one of four levels of organization distinguished in FDG: two levels for formulation (the Interpersonal and Representational Levels, for pragmatic and semantic analysis respectively) and two for encoding (the Morphosyntactic and Phonological Levels). One of the reasons for having these four levels of linguistic organization is that anaphoric reference is possible to any of them. This means that these levels should be (p.5) available as potential antecedents in underlying representations. Consider the following examples:

Interpersonal Level

  • (5) A Get out of here!

  • B Don't talk to me like that!

Representational Level

  • (6) A There are lots of traffic lights in this town.

  • B I didn't notice that.

Morphosyntactic Level

  • (7) A I had chuletas de cordero last night.

  • B Is that how you say ‘lamb chops’ in Spanish?

Phonological Level

  • (8) A I had /tʃu ‘letasdekor’dero/ last night.

  • B  Shouldn't that be ‘/tʃu ‘letasdeヴor’ dero /’?

In (5B) the anaphoric element that refers back to the communicative strategy chosen by A, which is indicative of the presence of an Interpersonal Level in the underlying representation of (5A). In (6B) that refers back to the situation in the external world that is described within (6A). This purely semantic reference shows that the underlying structure of (6A) contains a Representational Level of organization.

The anaphoric references in (7B) and (8B) are different since they are metalinguistic in nature. They are instances of ‘reflexive language’ (Lucy 1993) or ‘messages about the code’ (Jakobson 1971). In (7B) that does not refer to the entity described by chuletas de cordero but to the phrase ‘chuletas de cordero’ as such. This phrase is a morphosyntactic unit, hence the conclusion must be that this phrase is present in underlying structure and can therefore function as an antecedent for anaphoric reference. A similar line of reasoning can be set up for the anaphoric reference in (8B), the only difference being that here the antecedent is a phonological rather than a morphosyntactic unit.

From these facts it may be concluded that the underlying representation of an utterance contains four levels of organization: an Interpersonal Level (pragmatics), a Representational Level (semantics), a Morphosyntactic Level (morphosyntax), and a Phonological Level (phonology). Note that all these levels are purely linguistic in nature. This holds for the Interpersonal Level and the Representational Level too: these levels describe language in terms of its functions and meanings, but only in so far as these functions and meanings are encoded in the grammar of a language. Thus the Interpersonal Level (p.6)


Figure 1. FDG as part of a wider theory of verbal interaction

represents a linguistic unit in terms of its communicative function, and the Representational Level in terms of its semantic category.

1.2.5 Conceptual Component, Contextual Component, and Output Component

FDG as the Grammatical Component in a wider theory of verbal interaction is linked to a Conceptual Component, an Output Component, and a Contextual Component within an overall model of verbal interaction. These non‐grammatical components interact in various ways with the Grammatical Component. The Conceptual Component ( is responsible for the development of both a communicative intention relevant for the current speech event and the associated conceptualizations with respect to relevant extra‐linguistic events. The Output Component ( generates acoustic, signed, or orthographic expressions on the basis of information provided by the Grammatical Component. The Contextual Component ( contains a description of the content and form of preceding discourse and of the actual perceivable setting in which the speech event takes place and of the social relationships between Participants. The relationships among the Components are sketched in Figure 1.

This general design of a wider theory of verbal interaction is again inspired by the extensive research into the processes of speech production embodied in Levelt (1989). His model distinguishes three fundamental modules: the Conceptualizer, the Formulator, and the Articulator. Very roughly, these correspond to our Conceptual Component, Grammatical Component, and Output (p.7) Component respectively. The distinction within the Grammatical Component between formulating and encoding also owes much to Levelt's own use of these terms, although for him encoding is an aspect of formulation (1989: 11–12). The Conceptual Component

The Conceptual Component is the driving force behind the Grammatical Component as a whole. It is here that is represented the ideational and interactive material presupposed by each piece of discourse under analysis and the various communicative Moves and Discourse Acts that it contains. The Conceptual Component does not include every aspect of cognition that is potentially relevant for linguistic analysis, but only those that affect the immediate communicative intention. Harder (2004: 202) gives various pertinent examples, for example that given in (9):

(9) Speaker does his/her duty towards Addressee by conveying relevant bad news (‘John is ill’), mitigated by showing sympathy.

This will be expressed in Spanish as (10), an example drawn from Hengeveld (2004a) and also discussed by Harder (2004):

(10) Me tem‐o  que Juan está   enfermo.

1.SG fear‐1.SG.PRS COMP Juan cop.3.SG.PRS.IND ill

‘I am afraid that Juan is ill.’

There are two vital linguistic facts about (10) that must be captured in the Grammatical Component. Firstly, there is the presence of the indicative mood in the embedded clause, as opposed to the subjunctive mood in (11), which expresses a quite different communicative intention, namely the Speaker's expression of his fear that Juan may be ill:

(11) (Me) tem‐o  que Juan esté   enfermo.

1.SG fear‐1.SG.PRS COMP Juan cop.3.SG.PRS.SBJV ill

‘I fear that Juan may be ill.’

Secondly, we note the obligatory status of the reflexive pronoun me in (10) as against its optionality in (11). Without entering here into the actual analysis (but see Hengeveld 2004a: 15), we may observe that the communicative intention behind (10) is represented rather informally in (9) in language and not in abstract conceptual structures, which we will not go into in this book. See, on the many rivalling proposals for conceptual representation, Pederson and Nuyts (1997) and, for the necessity of distinguishing between semantic (‐pragmatic) and conceptual representations, Levinson (1997).

Slobin (1996) stresses how thinking for speaking is language‐specific and involves ‘picking those characteristics of objects and events that (i) fit some (p.8) conceptualization of the event, and (ii) are readily encodable in the language’ (1996: 76). Examples he gives (1996: 72) of ‘picking characteristics’ are the witnessed/non‐witnessed opposition in Turkish or the perfective/imperfective distinction in Spanish. For FDG, however, thinking for speaking is not part of the Conceptual Component. Rather, the selection of the language‐specific distinctions of the type discussed by Slobin is a task of the Grammatical Component, specifically the operation of Formulation, which has the task of translating conceptual configurations into the semantic and pragmatic distinctions available within a specific language.

In the informal representation of the language user's intention shown in (9), the material in normal print corresponds to the pragmatic, interpersonal side of the interaction, while the material in bold print lines up with its semantic, representational side. This distinction corresponds well with Butler's (2008b: 10) proposal that the Conceptual Component should distinguish a ‘conceptual component proper’ and an ‘affective/interactional component’, an opposition which he tentatively links to neurophysiological notions and the chemistry of brain processes. In turn, this distinction correlates nicely with the two aspects of formulation to be distinguished within the Grammatical Component: the formulation of the Interpersonal Level and that of the Representational Level, dealt with in Chapters 2 and 3 respectively. The Output Component

Let us now turn briefly to the Output Component, which—again to adopt the language of dynamic implementation—converts the final structures of the Grammatical Component into output. This output will in the case of speech (the kind of discourse that will primarily be considered in this book) be acoustic in nature and consist of articulatory gestures of the respiratory, laryngeal, and supralaryngeal structures of the human anatomy. With signed languages, which have been shown to have all the grammatical levels required for the description of spoken languages (including a phonological level, cf. Uyechi 1996), the output will consist of manual and other bodily gestures; and with written languages, the Output Component will oversee the motor control required for the production of orthographic expressions. Its function in speech may be seen as translating the digital (i.e. categorical, opposition‐based) information in the grammar into analogue (i.e. continuously variable) form: thus an utterance boundary in the grammar will yield inter alia a pause of so many milliseconds in the Output Component; or a syllable with a ‘falling’ operator will effect a decline in the fundamental frequency of the corresponding segment of the output. The Output Component will accordingly also be the location for long‐term settings, such as the tempo at which an individual's speech, signing, or writing is carried out: allegro forms (p.9) attributable to fast speech, or less accurate signing due to high tempo, or indeed ‘sloppier’ handwriting due to rapid use of the pen or keyboard are the kind of phenomenon to be treated here.

The distinction between the analogue nature of the Output Component and the digital nature of the grammar gives us an opportunity to emphasize an important characteristic of FDG. The analysis of linguistic data does not always lead to clear‐cut results. Criteria used to distinguish between word classes, for example, do not always give unequivocal classifications when applied to the forms found in a particular language; and the data drawn from corpus analysis will often show statistical (>0% and <100%) rather than categorical (0% or 100%) distributions. This has led a number of current grammatical approaches to promote the notion of gradience, the position that boundaries between categories are fluid and that categorization should be based upon prototypes rather than on inviolable criteria (for discussion, cf. Aarts 2007); gradience would then be taken to apply within grammar.

In particular, this notion of gradience has been extended to the distinction between lexical and grammatical phenomena. From a diachronic viewpoint, it is undeniable that grammatical phenomena derive overwhelmingly and uni‐directionally from lexical units, an observation that has been developed and deepened in the substantial literature on grammaticalization. As a corollary of this process, individual phenomena may find themselves somewhere on a scale between the initiation and the completion of a historical change and thus sharing properties of both the initial and final stages thereof. From a synchronic viewpoint, however, FDG postulates a sharp distinction between the lexical and the grammatical, a distinction that is integral to the way in which items will be represented in our analyses (but see Anstey 2006: 61–70 for a critical examination of this standpoint). The lexical‐grammatical distinction will return extensively in Chapters 2 and 3, where it correlates strongly with the opposition between modifiers and operators. The Contextual Component

Functional Discourse Grammar is so called because it seeks to understand the structure of utterances in their discourse context, though it is in no sense a discourse‐analytical model. The intention developed by the speaker does not arise in a vacuum, but in a multifaceted communicative context. For some FDG‐related suggestions as to the many aspects of the sociocultural situatedness of verbal interaction, see Connolly (2004). With the last of the nongrammatical components to be introduced in this chapter, the Contextual Component, FDG as presented here makes no effort to offer anything like a complete description of the overall discourse context. Rather, this Component contains two types of information, both of them limited in scope. Firstly, it houses the immediate information received from the Grammatical (p.10) Component concerning a particular utterance which is relevant to the form that subsequent utterances may take. Secondly, it contains longer‐term information about the ongoing interaction that is relevant to the distinctions that are required in the language being used, and which influence formulation and encoding in that language. The influence on formulation and encoding of both kinds of information, immediate and longer‐term, is symbolized by the arrows from the Contextual Component to the Grammatical Component in Figure 1. Just as with the Conceptual Component, we will not go into the internal constitution of the Contextual Component in this book.

As examples of long‐term settings within the Contextual Component, we may consider the sex of the speech‐act participants as well as the social relation between them. These are both relevant for Spanish, as shown in example (12):

(12) ¡Qué pálid‐a  est‐ás!

what pale‐F.SG  COP‐IND.PRS.2.SG.FAM

‘How pale you look!’

Here the choice of the forms pálida (rather than pálido ‘PALE‐M.SG’) and estás (rather than está ‘COP‐IND.PRS.2.SG.POL’) reflects specifications in the Contextual Component, i.e. the sex of the Addressee and the formality of the relation between Speaker and Addressee respectively. For an account of the grammatical properties of the corresponding utterance in English, as in the translation of (12), no such specification is required.

FDG adopts what Butler (2008a) refers to as a ‘conservative stance’ on the Contextual Component. Many of the matters that he himself includes in such a Component, like the factors that would induce selection of the informal lexeme kid rather than child in English to designate a child, would not find their way into an FDG Contextual Component. There are so many aspects of the context of interaction that could be argued to have an incidental impact upon a speaker's linguistic choices that modelling them within our theory would deprive it of much of its power. In an informal context, after all, a child may indeed be evoked by means of kid, but nothing prevents the choice of child. For this reason, factors relating to matters of genre, register, style, etc. will be included only where these can be shown to have a systematic effect upon grammatical choices in formulation (as in example (12) above); on the difficulties inherent in any attempt to include such factors in grammatical description, see Falster Jakobsen (2005).

Further examples of the type of phenomena which call upon the Contextual Component are reflexives, anaphora, and instances of narrative chaining, all of which we will deal with at the respective stage of the presentation. In languages with logophoric pronouns, for example, the Contextual Component will have to keep track of the status of (typically human) entities as belonging to a (p.11) particular embedded discourse domain or not. In such languages a systematic formal opposition is made between the two readings of He said that he was ill, according as the second instance of he identifies the creator of the embedded domain (i.e. the referent of the first instance of he; this is indicated by the logophoric form) or some other male individual, indicated by the non‐logophoric form (see for discussion). Similarly, according as a language permits reflexive pronouns to apply across larger or smaller stretches of discourse, the Contextual Component will be adjusted to make particular possible antecedents available.

Note that the short‐term information in the Contextual Component must be continually kept up to date. Anaphoric chains depend upon the availability in the Contextual Component of valid antecedents. As the discourse progresses, so some of these cease to be available while others arise as potential antecedents. The Contextual Component will be responsive to the requirements of the particular language in this respect. This also applies to narrative chaining, where the positioning of a State‐of‐Affairs within an Episode must be specified with regard to previous or later States‐of‐Affairs. Where the anaphora or narrative chaining works forwards in time (cataphora), the Contextual Component will create an empty position constraining the formulator to supply the awaited information.

As seen in Figure 1, the input to the Contextual Component does not only come from the result of formulation but also from the result of encoding, in other words the Morphosyntactic and Phonological Levels within the Grammatical Component. This is because, as we saw in 1.2.4, anaphoric reference is possible not only to pragmatic and semantic constructs but also to sections of the actual morphosyntactic structure of clauses and phonological structure of utterances. In the following chapters, we will detail various ways in which there is interaction between the Contextual Component and the various Levels of the Grammatical Component.

In 1.2.5 we classified the Conceptual, Output, and Contextual Components as non‐grammatical. Our discussion of the three non‐grammatical Components has shown, however, that they are certainly not non‐linguistic. Indeed, all three will differ from language to language, according to the impact that each has on linguistic form. The decision whether to include a particular phenomenon in the grammar or in one of the flanking Components will be taken language‐specifically and will be determined by considerations of systematicity. If, for example, every single utterance in a language ends in a lengthened syllable, this should be shown as a systematic aspect of the grammar; if there is a statistical tendency to utterance‐final syllable‐lengthening, this is something to put into the Output Component. If a language expresses all commands as a question about ability (Can you open the window? etc.), (p.12) then this is a grammatical fact about that language. If the Speaker may express commands either directly by means of an Imperative Illocution or indirectly as a question about ability, the circumstances determining that choice are a matter for the Conceptual Component while the alternative formulations are a matter for the grammar.

We have now seen in general terms how FDG operates as a top‐down grammar of the Discourse Act, recognizes four Levels of description and interacts with Conceptual, Output, and Contextual Components. In the following section, we will consider the architecture of FDG in greater depth.

1.3 The architecture of FDG

1.3.1 Overall organization

The general architecture of FDG and the Components that flank the Grammatical Component may now be represented as in Figure 2, in which the Grammatical Component is presented in the centre, the Conceptual Component at the top, the Output Component at the bottom, and the Contextual Component to the right. Note that this figure fleshes out Figure 1.

Within the various Components, circles contain operations, boxes contain the primitives used in operations, and rectangles contain the levels of representation produced by operations. In line with the top‐down organization of FDG, we start our discussion of Figure 2 at the top.

As mentioned in, in the prelinguistic Conceptual Component a communicative intention (e.g. issuing a warning) and the corresponding mental representations (e.g. of the event causing danger) are relevant. Through the operation of Formulation these conceptual representations are translated into pragmatic and semantic representations at the Interpersonal and the Representational Levels, respectively.

The rules used in Formulation are language‐specific, i.e. FDG does not presuppose the existence of universal pragmatic and semantic notions. As a result, similar conceptual representations may receive different pragmatic and semantic representations in different languages. To give just one example: warnings are in some languages encoded as a distinct type of speech act, whereas in others they receive the same treatment as orders. This type of crosslinguistic variation may be expected to be governed by typological hierarchies, just like morphosyntactic and phonological variation.

Formulation rules make use of a set of primitives that contains frames, lexemes, and operators (see The configurations at the Interpersonal and the Representational Levels are translated into a morphosyntactic structure at the Morphosyntactic Level through the operation of Morphosyntactic Encoding. The Morphosyntactic Encoding rules draw on a set of primitives (p.13)


Figure 2. General layout of FDG

containing Morphosyntactic Templates, Grammatical Morphemes, and Morphosyntactic Operators (see Similarly, the structures at the Interpersonal, Representational, and Morphosyntactic Levels are translated into a phonological structure at the Phonological Level. The phonological encoding rules draw on a set of primitives containing Phonological Templates, Suppletive forms, and Phonological Operators (see

By organizing the Grammatical Component in this way, FDG takes the functional approach to language to its logical extreme: within the top‐down organization of the grammar, pragmatics governs semantics, pragmatics and semantics govern morphosyntax, and pragmatics, semantics, and morphosyntax govern phonology.

(p.14) The Phonological Level of representation is the input to the operation of Articulation, which, in the case of an acoustic (as opposed to written or signed) Output Component, contains the phonetic rules necessary for arriving at an adequate utterance. Articulation takes place outside the grammar proper.

The various levels of representation within the grammar feed into the Contextual Component, thus enabling subsequent reference to the various kinds of entity relevant at each of these levels once they are introduced into the discourse. The Contextual Component feeds into the operations of formulation and encoding, so that the availability of antecedents, visible referents, and speech‐act participants (and possibly bystanders; cf. Rijkhoff 1995) may influence the composition of (subsequent) Discourse Acts. Note that the representation of these feeding relations in Figure 2 is a simplification when looked at from the perspective of the language user. In order to create a contextual specification, the Addressee has to reconstruct all the levels of representation within the grammar on the basis of the actual output of that grammar, i.e. the phonetic utterance. Since in this book we restrict ourselves to the perspective of language production and concentrate on the Grammatical Component, we abstract away from this complication by provisionally assuming direct feeding relationships between the Grammatical Component and the Contextual Component.

1.3.2 Levels and Layers Introduction

Each of the levels of representation distinguished within the Grammatical Component in Figure 2 is structured in its own way. What all the levels have in common is that they have a hierarchically ordered layered organization and are displayed as a layered structure. In its maximal form the general structure of layers within levels is as follows:

(13) (πv1: [head (v1)ϕ]: [σ (v1)ϕ])ϕ;

Here v1 represents the variable of the relevant layer, which is restricted by a (possibly complex) head that takes the variable as its argument, and may be further restricted by a modifier σ that takes the variable as its argument. The layer may be specified by an operator π and carry a function Φ. Heads and modifiers represent lexical strategies, while operators and functions represent grammatical strategies. The difference between operators and functions is that the latter are relational, holding between the entire unit and other units at the same layer, while the former are not, applying only to the unit itself.

Of course, not all relations between units are hierarchical. In those cases in which units together form a non‐hierarchical (equipollent) configuration, (p.15)


Figure 3. The Interpersonal Level

they are enclosed between square brackets, as exemplified in (13), where the relationship between a head and its argument and a modifier and its argument is indicated by square brackets.

The levels differ in the sense that at each level a linguistic expression is analysed in terms of the distinctions relevant to that level. It should be stressed again that the representations at all levels are purely linguistic in nature, so that only those distinctions are provided that are actually reflected in the grammar of the language involved. The Interpersonal Level

At the Interpersonal Level the hierarchical structure given in Figure 3 applies.

As indicated in 1.2.3, we recognize as a unit of analysis at the Interpersonal Level the Move (M), which may contain one or more (N) Discourse Acts (A). Each Discourse Act contains an Illocution (F), which specifies a relation between speech‐act Participants (P, the Speaker S, and the Addressee A) and (except in the case of Expressives, the Communicated Content (C). The Communicated Content contains a varying number of Ascriptive (T) and Referential (R) Subacts. Note that the latter two units are operative at the same layer, i.e. there is no hierarchical relation between them; in such cases of equipollence, square brackets are applied. In general, then, at the Interpersonal Level units are analysed in terms of their communicative function. The Representational Level

At the Representational Level the relevant layers are those presented in Figure 4.

At this level of analysis linguistic units are described in terms of the semantic category they designate (see Hengeveld 1989, 2004a; Mackenzie 2004c). These categories are of different types, such as Propositional Contents (p), which may contain one or more (n) episodes (ep) (see Gómez Soliño 1995), which may contain one or more descriptions of States‐of‐Affairs (e); the latter, in turn, are characterized by one or more Properties (f1), which may contain descriptions of Individuals (x1) and further Properties (f2). Further classes of semantic category are presented in Chapter 3. Note that, as is indicated by the


Figure 4. The Representational Level

(p.16) square brackets, the Individuals and further Properties in Figure 4 belong to the same layer, i.e. there is no hierarchical relation between them.

The nature of a semantic category is not indicative of the way the linguistic unit describing that category is used within a Discourse Act. Semantic categories, as the name indicates, are categories, not functions. The functional analysis is given at the Interpersonal Level. Thus, the same Property (f) may be either ascribed (T) or referred to (R). The following examples illustrate this point (note that the formal difference between tall and tallness is accounted for at the Morphosyntactic Level as arising from coercion, cf. 4.6.1):


  1. a. The teacher is tall.

    (Ascription of Property: T/f)

  2. b. Tallness impresses the teacher.

    (Reference to Property: R/f)

Similarly, an Individual may be ascribed or referred to:


  1. a. Sheila is my best friend.

    (Ascription of Individual entity: T/x)

  2. b. My best friend visited me last night.

    (Reference to Individual: R/x)

A more elaborate representation of (14a–b) is given in (16a–b):


  1. a. (CI:   [(TI)

    (pi: (epi: (ei: [(fi: [(fj: tall (fi))

    (RI)   ]    (CI))

    (xi: teacher (xi))ϕ ] (fi)) (ei)ϕ]) (epi)) (pi))

  2. b. (CI:   [(TI)   (RI)

    (pi: (epi: (ei: [(fi: [(fj: impress (fj)) (fk: tallness (fk))ϕ

    (RI)   ]    (CI))

    (xi: teacherN (xi))ϕ ] (fi)) (ei)ϕ]) (epi) (pi))

Examples like these show that, though there are regular correspondences between the Interpersonal Level and the Representational Level, the two are basically independent of each other, allowing for a wide variety of interactions between them. The Morphosyntactic Level

Figure 5 shows the general maximal template for morphosyntactic frames at the layer of the linguistic expression, where each unit may occur more than once.

At this level a linguistic unit is analysed in terms of its syntactic constituents, with, from the highest to the lowest layers: Linguistic Expressions (Le), Clauses (p.17)


Figure 5. The Morphosyntactic Level

(Cl), Phrases of several types (Xp), and Words of several types (Xw). Within Words we furthermore distinguish Morphemes of several types (Xm), not shown in Figure 5. As is to be explained in 1.4.4, the notion ‘sentence’ is not applied in FDG.

There is no necessary one‐to‐one mapping between semantic and pragmatic units on the one hand and morphosyntactic units on the other. As argued earlier, Discourse Acts may be expressed as Clauses, Phrases, or Words. To give another example: semantic predications consisting of a unit designating a Property and two units designating Individuals may be realized in one language as a Clause with three constituents and in others as a single Word. Consider the following examples, from English and Southern Tiwa (Allen et al. 1984: 293; the gloss 1.SG.SBJ>PL.OBJ in (18) should be read as ‘first person singular subject acting on plural object’).

(17) I made shirts.

(18) Te‐shut‐pe‐ban


‘I made (the) shirts.’

The English Clause in (17) can be subdivided into three syntactic constituents corresponding to the three semantic units mentioned earlier: a unit designating a Property (made) and two units designating Individuals (I, shirts). The same semantic configuration is expressed in Southern Tiwa as a single syntactic constituent, as shown in (18). The Actor argument is expressed by means of a prefix on the Verb and does not have to be expressed independently. The Undergoer argument is incorporated into the Verb. The fact that the Undergoer is cross‐referenced on the verb shows that it is really an argument of that Verb. Assuming a similar underlying semantic representation for (17)– (18), these examples thus clearly demonstrate that there are many possible mappings between the Representational and the Morphosyntactic Levels. The Phonological Level

The Phonological Level is equally language‐specific, and contains both the segmental and the suprasegmental phonological representation of an Utterance, which is the largest phonological unit considered in FDG. Figure 6 shows the phonological template for an Utterance, with a number of simplifications for purposes of exposition. Again, every unit may occur more than once.



Figure 6. The Phonological Level

At this level the linguistic expression is analysed in terms of the phonological units it contains, such as the Utterance (U), the Intonational Phrase (IP), the Phonological Phrase (PP), and the Phonological Word (PW).

Again, there is no necessary one‐to‐one mapping between pragmatic, semantic, and morphosyntactic units on the one hand, and phonological units on the other. Thus, in some languages subordinate clauses are set off from the main clause by means of a break between two Intonational Phrases, whereas in others they form a single Intonational Phrase with the main clause. To give another and perhaps more pervasive example: Phonological Words at the Phonological Level are not necessarily in a one‐to‐one relationship with constituent boundaries at the Morphosyntactic Level, as can be illustrated with the following example from Dutch, as pronounced in certain varieties in the Netherlands:

(19) Ik wou  dat  hij kwam.

I want.PST COMP he come.PST

‘I wish he would come.’

An alignment of the morphosyntactic analysis with the phonological analysis would be as follows (the symbol ‘—’ is used to indicate the beginning and the end of a fragment that is not further analysed in detail, see 1.5 below):

(20) (Cli: [  (Npi: ‐ik‐ (Npi)) (Vpi: —wou— (Vpi))

(U i: [IP i: [ (PP i: (PW i: —kυau— (PW i)) (PP i))

(Clj: [ (Gwi: —dat— (Gwi)) (Npj: —hij— (Npj))

(PP j: [ (PW j: —dati— (PW j))

(Vpj: —kwam— (Vpj)) ] (Clj)) ]   (Cli))

(PW k: —kυam‐ (PW k)) ] (PP j)) ] (IP i)) ] (U i))

This example shows that the first Phonological Phrase (PP i) corresponds to the first two syntactic constituents of the main clause, while the second corresponds to the embedded clause. Within the second Phonological Phrase there are two Phonological Words (PW j) and (PW k), one corresponding to the first two syntactic constituents of the embedded clause, the second corresponding to the single remaining syntactic constituent.

(p.19) 1.3.3 Primitives Introduction

The various operations creating the levels just discussed make use of sets of primitives which serve as the building blocks for their respective levels of application. The rules that constitute the operations within the grammar (formulating and encoding) combine these primitives in order to produce the various levels of representation. Primitives used in Formulation

The operation of formulation has to produce two different levels of representation: the Interpersonal Level and the Representational Level. For each of these levels, similar (although distinct) primitives are relevant. These will be presented in brief here: full detail will be given in Chapters 2 and 3.

First of all, the set of primitives contains Frames which define the possible combinations of elements at the Interpersonal Level and at the Representational Level for a certain language. Despite their language‐specific nature, the inventory of frames is expected to be partly predictable in terms of crosslinguistically valid typological hierarchies. Relevant distinctions captured by frames at the Interpersonal Level include, for example, the expressive or communicative nature of Discourse Acts, the encoded configurations of information structure, and the rhetorical functions of Discourse Acts. At the Representational Level frames capture such matters as quantitative and qualitative valency, the combinations of semantic categories allowed, and possible modification structures.

Secondly, this set of primitives contains Lexemes; these are given in phonemic form, although for ease of exposition we shall generally simply use orthographic form. Within the set of Lexemes a distinction is made between those that function at the Interpersonal Level (e.g. interjections, proper names, illocutionary adverbs, performative expressions, etc.) and those that function at the Representational Level. Lexemes are independent units that have to be associated with the aforementioned frames (see García Velasco and Hengeveld 2002 for discussion of this issue). In the implementation of the grammar the frames are selected first, and only after that are lexemes inserted. This reflects the choice the Speaker often has in describing one and the same entity through a variety of lexemes with different connotations and/or denotations. It also provides a natural framework for understanding the phenomenon of coercion, through which lexemes that are strongly associated with a particular frame can be forced for expressive purposes into a frame that is usually coupled with lexemes of another meaning class.

(p.20) Thirdly, this set of primitives contains interpersonal and representational operators, which represent grammatical expressions in terms of their pragmatic or semantic content respectively. The classification of these operators will be addressed extensively in Chapters 2 and 3. Here we just give a number of examples. At the Interpersonal Level, mitigation is an operator at the illocutionary layer, reportative is an operator at the layer of the Communicated Content, approximation (‘sort‐of’) is an operator at the layer of the Ascriptive Subact, and identifiability operates at the layer of the Referential Subact. At the Representational Level, examples are subjective modality at the layer of the Propositional Content, tense at the layer of the State‐of‐Affairs, number at the layer of the Individual, and phasal aspect at the Property layer. Primitives used in Morphosyntactic Encoding

The Morphosyntactic Level is organized on the basis of morphosyntactic templates for Linguistic Expressions, Clauses, Phrases, and Words which are stored as part of the set of primitives relevant for the operation of Morphosyntactic Encoding. The inventory of templates has to be specified for each language individually, although again the expectation is that crosslinguistically valid generalizations will make this inventory largely predictable on the basis of a limited number of parameters.

The second set of primitives relevant at the Morphosyntactic Level consists of grammatical morphemes, which are unmodifiable elements such as Auxiliaries, Particles, and Affixes. These grammatical morphemes have to be introduced at the Morphosyntactic Level, since they occupy slots in the morphosyntactic configuration, which is determined at this level. To give an example at the clause layer: in Dutch the main verb normally occurs in second position in a Clause, but when an auxiliary verb is present, this Auxiliary occupies the second position and the main verb occurs in final position, as illustrated in (21) and (22):

(21) Karel won   de wedstrijd.

Karel win.PST.SG DEF game

‘Karel won the game.’

(22) Karel heeft   de  wedstrijd gewonnen.

Karel have.PRS.3.SG DEF game  win.PTCP

‘Karel has won the game.’

Examples like these clearly show that it is impossible to determine the order of constituents without taking grammatical morphemes into account. Similar examples can be given at the Phrase and the Word layers.

(p.21) Often, various semantic distinctions map onto a single grammatical morpheme. For instance, the accusative case in a certain language may be triggered by the semantic function Undergoer, but also by various types of Modifier, or it may be lexically triggered by certain verbs or adpositions. The other way around, a single semantic category may map onto various morphosyntactic categories, as when the form of the accusative expressing the Undergoer argument is dependent on the noun class of the head of that Undergoer argument.

Grammatical morphemes are introduced in their phonemic form when they are regular and predictable. They are introduced by means of a Morphosyntactic Operator, the third set of primitives at the Morphosyntactic Level, in those cases in which their final form is not fully predictable and has to be selected from a suppletive paradigm. Morphosyntactic Operators can thus be considered to be placeholders for actual forms or sets of forms. In assigning names to Morphosyntactic Operators we will generally use labels similar to the ones used in glosses, so as to enhance readability. It is important to realize, however, that these names could just as well be represented by numerical codes, like 581, since they trigger forms, and at this level no longer represent meanings. Primitives used in Phonological Encoding

The Phonological Level is organized on the basis of phonological templates for Utterances, Intonational Phrases, Phonological Phrases, Phonological Words, Feet, and Syllables, which are stored as part of the set of primitives relevant for the operation of Phonological Encoding. The inventory of templates has to be specified for each language individually, and certain languages may lack entire layers altogether (for example, Vietnamese has been claimed to lack the layer Phonological Word, cf. 5.6), although again the expectation is that crosslinguistically valid generalizations will make this inventory largely predictable on the basis of a limited number of parameters.

The second set of primitives consists of the suppletive forms that correspond to the Morphosyntactic Operators introduced in the previous section and to unpredictable forms from the paradigms of lexemes. Suppletive forms are introduced at the Phonological Level, since in many languages the form of a morpheme may be affected by the morphosyntactic configuration in which it occurs. Bakker (2005: 3) cites the following example of this phenomenon from Yagua (Payne 1990: 30):


  1. a. Sa‐juuy  Anita.

    3.SG.SBJ‐fall Anita

    ‘Anita fell.’

  2. (p.22)
  3. b. Anita Ø‐juuy.

    Anita 3.SG.SBJ‐fall

    ‘Anita fell.’

In Yagua, the subject‐agreement prefix on the verb is sa‐ when the Subject term occurs in postverbal position (23a), but it is Ø‐ when the Subject occurs in preverbal position (23b). This means that in this language the form of the third singular subject marker 3.SG.SBJ can only be determined after the constituent order of the Clause is established.

A third set of primitives potentially relevant at the Phonological Level consists of Phonological Operators. These anticipate aspects of the articulatory, signed, or orthographic output that are not a direct reflection of an Interpersonal, Representational, or Morphosyntactic Operator. A good example of a phenomenon for which such Phonological Operators are necessary is intonation. The Phonological Level will distinguish such operators as r(ising) and f(alling) or h(igh) and l(ow), typically applying these to the syllable layer. Depending on the language type, some or all syllables will be marked by such an operator; in tone languages, each syllable will be in principle be marked with an operator (5.7). The ‘digital’ information given by these Phonological Operators provides instructions to the Output Component which then will perform phonologically insignificant but phonetically necessary operations to ensure a smooth ‘analogue’ intonation contour. Generalizations

There are certain correspondences across the three sets of primitives. Within each set there is a subset of units with a structuring function: the frames used in Formulation and the Templates in Morphosyntactic and Phonological Encoding all serve the purpose of providing an overall organizing structure for their respective levels. Within each set of primitives there is furthermore a subset of units in phonemic form: the Lexemes used in Formulation, the Grammatical Morphemes used in Morphosyntactic Encoding, and the Suppletive Forms used in Phonological Encoding all contribute to the cumulative segmental specification of the underlying representations. Finally, within each set of primitives there is a subset of operators: Interpersonal and Representational Operators are relevant to the operation of Formulation, Morphosyntactic and Phonological Operators to the operation of Encoding.

1.3.4 Levels and primitives

for a simple illustration of how a single constituent gets different representations at each level, using different sets of primitives, consider the example in (p.23) (24). The constituent these bananas is represented in four different ways within FDG, as provisionally indicated in (25).

(24) (I like) these bananas.


  1. a. IL (+id RI)

  2. b. RL (prox m xi: [(fi:/bə‘na:nə/N(fi)) (xi)ϕ])

  3. c. ML (Npi: [(Gwi: this‐pl(Gwi)) (Nwi:/bə‘na:nə/‐pl (Nwi))] (Npi))

  4. d. PL (PP i: [(PW i:/ði:z/(PW i)) (PW j:/bə‘na:nəz/(PW j)) ] (PP i))

At the Interpersonal Level (IL, 25a), the constituent is characterized as having a referential function (R). The referent is furthermore assumed by the Speaker to be identifiable (+id) by the Addressee. At the Representational Level (RL, 25b) the constituent is characterized as designating more than one (m) Individual (x) with a Property (f) and in terms of the location of its referent (prox). The Property (f) is specified by the Nominal (N) Lexeme /bə‘nc:nə/. At the Morphosyntactic Level (25c) the constituent is characterized as being a Noun Phrase (Np), which consists of a Grammatical Word (Gw) and a Nominal Word (Nw). At this level a Morphosyntactic Operator is introduced, for convenience here shown as ‘this’, which acts as a placeholder in the appropriate syntactic position. The Representational Operator m is converted into the Morphosyntactic Operator Pl(ural), which occurs twice, since it has to be expressed on each of the two Words making up the Noun Phrase. At the Phonological Level (25d) the appropriate plural forms of the Words are introduced, in the case of the noun by adding the appropriate form of the plural suffix, in the case of the determiner by selecting the appropriate suppletive form corresponding to the combination of Morphosyntactic Operators. The Phonological Level in this case consists of one Phonological Phrase (PP) containing two Phonological Words (PW).

1.3.5 Implementation Introduction

The various levels of organization are related to each other through rules of Formulation and Encoding, in a dynamic implementation of the grammar (cf. Bakker 2001,2005). Two principles are crucial in this implementation, and these are discussed in the following sections. Depth first

The depth/first principle was proposed in Bakker (1999) in the context of FG and is adopted in FDG, but with a somewhat different interpretation. In defining its role within the grammar, recall that a basic assumption in FDG is that a grammatical model will be more efficient the more it resembles (p.24) language production in the individual. There is a consensus in the psycholinguistic literature that language production is incremental, in the sense that prelinguistic conceptualizations arise gradually through time (in microseconds, it should be said) and that material is sent ahead for encoding before the entire communicative intention has been fully developed (Levelt 1989: 24–7; see also Mackenzie 2000, 2004b and Harder 2007). In accordance with this, information from a certain level is sent down to a lower level as soon as the necessary input information for that lower level is complete. The grammar would slow down considerably if first the Interpersonal Level had to be fully specified, and second the Representational Level had to be filled in completely, so that only then could the morphosyntactic configuration be determined, which after that would be mapped onto a phonological configuration. This is not how language production in the individual works, and it would therefore, given the basic assumption mentioned above, not lead to a very efficient model of grammar either. (Fortescue 2004: 169 warns of the dangers of ‘hybrid models’, oriented partly to pattern and partly to process: our model is a pattern model that is inspired by process without seeking to model the latter.)

As an example, consider the effect of specifying an illocutionary value at the Interpersonal Level (cf. Risselada 1993: 78–86). As soon as an Imperative (IMP) frame has been selected for the Discourse Act, there are potentially important consequences at all subsequent levels of representation: (i) at the Representational Level, the State‐of‐Affairs frame will have to designate a controlled State‐of‐Affairs, and the first argument will have to include the Addressee; (ii) at the Morphosyntactic Level, in some languages a specific constituent order is used, or there may be special imperative auxiliaries or morphological markers; (iii) at the Phonological Level, there may be specific prosodic patterns that are used with Imperatives (cf. 5.4). All this means that the selection of an Imperative frame at the Interpersonal Level may trigger a whole range of specifications at subsequent levels, both in terms of formulation and of encoding, irrespective of the specification of further elements at the Interpersonal and lower levels.

Note, however, that evidence is also available that there is also a role in language production for processes which involve looking ahead to a unit‐final element: cf. Hannay and Martínez‐Caro (2008) for the notion of working up to a clause‐final position in syntax, Fortescue (2007: 340–1) for morphological processes that involve ‘backtracking’ from a word‐final position, and again Levelt (1989: 401–5) for look‐ahead in phonology. In our modelling of morphosyntax, too, based as it is on observations about the patterning of linguistic units, we will see that counting forwards from an initial position and backwards from a final position are both called for. In all these (p.25) ways, and others, we may observe a general analogy between production processes and the sequence of steps involved in a pass through the model of FDG. Maximal depth

The principle of maximal depth states that only those levels of representation that are relevant for the build‐up of (a certain aspect of) an utterance are used in the production of that (aspect of the) utterance. This principle, too, is meant to speed up the implementation of the grammar. It avoids the vacuous specification of levels of representation that are irrelevant to the production of the utterance at hand.

Following up on the example in the previous section, this means that in a certain language there may be a direct connection, circumventing the Representational Level, between the Interpersonal Level and the Morphosyntactic Level in those cases in which the Imperative frame has to be mapped onto a specific clausal template. Similarly, there may be a direct connection, circumventing the Representational Level and the Morphosyntactic Level, between the Interpersonal and Phonological Levels when the Imperative frame is mapped onto a specific prosodic pattern. In this way, superfluous steps in passing on information within the top‐down procedure are avoided. Looking at this from a bottom‐up perspective, it means that the expression of underlying structures is potentially based on information from all higher levels, not just from the next one up.

Having seen something of the architecture of FDG and of its implementation in the analysis of various phenomena, let us now place it in its broader context.

1.4 FDG in its broader context

1.4.1 Introduction

Functional Discourse Grammar is so called because it adheres to the principles of linguistic functionalism and takes the Discourse Act as its basic unit of analysis. As we have seen, it is a grammatical model that constitutes one component of an overall theory of verbal interaction and aims to be equally valid for all types of language. As a result the notions of functionalism, language typology, language modelling, and Discourse Act all play a central role in FDG. Accordingly, Section 1.4.2 will deal with functionalism, contrasting it with formalism and indicating how FDG is to be located with respect to these two major schools of linguistic thought and to two closely related theories. Section 1.4.3 will turn to linguistic typology, and consider both the influence of typology upon the theory of FDG and the role it could play in (p.26) typological work. Section 1.4.4, finally, will present FDG as a form‐oriented function‐to‐form model, showing how it relates to psycholinguistic work on speech production and giving some indications on how a practising linguist can work with FDG. In all three sections, the Discourse Act will play a central role.

1.4.2 Functionalism

FDG occupies a position halfway between radically functional and radically formal approaches to grammatical analysis. Functionalism refers here 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 (Dik 1986). Radical functionalism is an extreme form of this standpoint, denying the cognitive reality of linguistic structure and seeing linguistic form as an ephemeral manifestation of the language user's attempt to achieve his/her communicative purposes. Radical functionalists tend to support a usage‐based linguistics, one which typically involves the detailed examination of corpus data and the extraction of inductive generalizations which typically pertain only to the language under examination. Patterns discerned in these data are seen as emergent rather than as reflecting any kind of structure. A major statement of this position is Hopper (1987: 142), who takes a view of structure as ‘always provisional, always negotiable, and in fact as epiphenomenal’.

Formalism, by contrast, is strongly committed to the existence of mental structure, the foundations of which are typically regarded as innate. The deeper properties of linguistic phenomena cannot from this perspective be understood directly from data. Rather, the utterances in an actual text or transcript of speech reflect (quite imperfectly, it is generally believed) an underlying system that is governed by rules that predict the form taken by idealized linguistic units. Radical formalism is in our terms an extreme manifestation of this standpoint, one that limits linguistic study to the investigation of this covert system, totally independent of the uses to which it is put. For a critique of both radical positions and a plea for the recognition of both flexibility (i.e. variability) and rigidity (i.e. the requirement for rules) in the make‐up of language, see Givón (2002: ch. 2).

The position taken by FDG lies between these extremes. FDG, like formalist models, seeks to describe the knowledge that underlies a language user's potential to communicate in his/her language in an explicit and highly formalized way. The language user is seen as having knowledge both of units (e.g. lexemes, auxiliaries, syntactic constituents, phonemes) and of the ways in which these units may be combined (into Discourse Acts, Propositions, (p.27) Clauses, and Intonational Phrases). This knowledge displays a large degree of stability, such that it can be compared across languages, revealing universal trends in linguistic structure, as studied in language typology. However, FDG takes the position that this knowledge of units and their combination is instrumental in interpersonal communication and has arisen as a result of historical processes: forms that have served Speakers well through the ages have sedimented into the repertory now available to language users and are well‐adapted to their purposes. The forms that are at language users' disposal are variable across languages, but do not vary without limits. Rather, the limits on variation are set by the range of communicative purposes displayed by all language users and by the cognitive constraints they are subject to. FDG thus offers not only an inventory of forms but also seeks to clarify how these are combined in verbal interaction.

The two sides of the dualist position taken by FDG, i.e. its orientation to both form and function, may perhaps be compared to different ways of analysing the bicycle (here disregarding the fact that, while a bicycle is an artefact purposefully invented to satisfy certain needs, language has evolved naturally). One aspect of an FDG‐style analysis of a bicycle would be to give a complete and descriptively adequate account of this phenomenon, i.e. one that accurately covers all necessary properties for an object to count as a bicycle: a frame with certain geometrical and engineering properties, a handlebar, pedals, a chain, etc., and of course two wheels (with their various characteristics). The account would make a distinction between allowable variation (for instance in the overall size of the bicycle or in the relative size of the wheels) and impermissible variation (without pedals and chain, the object is a child's scooter rather than a bicycle; with fewer or more than two wheels it is not a well‐formed bicycle, but a monocycle or tricycle, for example). A description of other non‐criterial properties of a bicycle, such as a bell or lights, would be added for completeness. These, then, would be elements of the formalist side of the FDG account.

What is missing from this description is any indication of how the bicycle is used for human purposes (transportation, diversion, competition, etc.). In principle, such a function‐free description is possible, but is less enlightening: it offers no answer to the question why bicycles have been designed to have two wheels, just as the formalist account offers no answer to the question why languages have evolved to have the properties that they do have. What is more, it fails to show how the variation in the weight and structure of bicycles depends upon the uses to which they are put: a bicycle designed to carry shopping will not be suitable for racing and vice versa. Similarly, the formalist account does not clarify how linguistic structures co‐vary with the purposes to which they are put in communication.

(p.28) The both‐function‐and‐form approach (such as FDG adopts within linguistics) offers an understanding of why the unmarked form of the cycle is the bicycle. The monocycle offers the advantages of small size and light weight and of consequently being extremely manoeuvrable in traffic; on the other hand, it is relatively dysfunctional in being highly unstable (for the untrained user), in being largely limited to even surfaces and in not offering the possibility of transporting goods. The tricycle is highly stable, can be used in a range of environments and for the transportation of goods; however, it is relatively heavy, obstructive, and difficult to manoeuvre. In this light, the bicycle emerges as a perfect compromise, being of moderate weight, fairly stable (in use), appropriate for flat and inclined surfaces, and offering reasonable facilities for transportation.

This example typifies the FDG approach to linguistic forms. FDG recognizes that the forms taken by utterances are variable but that the variation is limited by the (communicative rather than transportational!) needs of users. Let us give a couple of linguistic examples that will be developed in later chapters. There is a strong tendency for the principal units of verbal interaction (Discourse Acts) to contain one element with the pragmatic function Focus (for details, see Only under rather special circumstances will it contain more than one Focus (as in such multiple wh‐questions as Who gave what to whom?). In a sense it might indeed seem more efficient to cram many Foci into one Discourse Act, and some languages, such as English, do not forbid this from happening, although others do. Why there should be a preference for one‐Focus Discourse Acts is something to be explained in terms of human communicative practices. Similarly, the units describing States‐of‐Affairs will across languages tend to contain one or two essential participants (arguments) (for details, cf. Chapter 3); certain languages also permit predications with three or more arguments, and certain languages permit predications without any arguments. Again, the question arises why this should be (as with the two wheels of the bicycle). Whereas a formalist description confines itself to a mere observation of this regularity, the approach taken by FDG calls for an explanation in terms of human cognition and communication.

FDG shares with the formalist approach, but not with certain more radical functionalist approaches, that it is concerned with the criterial properties of the language under description. Just as the laws of a country may require that a bicycle be provided with a bell and lights, so social conventions may require linguistic utterances to display certain properties. To the extent that these are not criterial to the functioning of language, these properties fall outside the scope of an FDG. Thus FDG will not concern itself directly with the impact of genre distinctions on linguistic form: the distinction between the style of an official letter and that of an informal e‐mail, for example, falls (p.29) outside the scope of FDG, since this concerns norms of communication rather than properties of the language system. However, where norms do impinge upon the system, e.g. through the introduction of systematic oppositions that reflect interpersonal relations (honorific morphology, pronouns of intimacy and distance), these must be accounted for.

The above will have made clear that FDG is what Butler (2003) refers to as a structural‐functional grammar, a term which nicely captures its intermediary status. While accepting that grammar is shaped by use, FDG holds ‘that in synchronic terms the grammar of a language is indeed a system, which must be described and correlated with function in discourse’ (Butler 2003: 30). This standpoint brings FDG into a close relation with two other structural‐functional approaches, Systemic‐Functional Grammar (SFG; Halliday and Matthiessen 2004) and Role and Reference Grammar (RRG; Van Valin and LaPolla 1997, Van Valin 2005). Although there is no room here for a detailed comparison (but see Butler 2003 for an exhaustive comparison of FG, SFG, and RRG, and Gonzálvez García and Butler 2006 for a mapping in multidimensional space of these three approaches and eight others), we may venture the hypothesis that FDG occupies a position intermediate between SFG, which stands closer to radical functionalism, for example in embracing the study of genre, and RRG, which stands closer to radical formalism, for example in seeing itself as first and foremost a theory of syntax (Van Valin 2001: 172). A brief consideration of the differences among the three approaches may help to clarify the aims and ambitions of FDG.

A characteristic feature of work in SFG is its orientation to the use of language in social contexts; as explained above, FDG limits itself to systematic grammatical reflections of social meanings. SFG furthermore takes ‘the text rather than the sentence’ (Halliday 1994: 4505) to be the object of linguistic description. This does not apply to FDG, which is not a ‘discourse grammar’ in the sense of a grammar of discourse (if such an entity is attainable at all) deriving from text‐linguistic analysis. Rather, FDG wishes to understand those systematic properties of the Discourse Act (the minimal unit of communication) that require reference to its being situated within an interactive Move by the language user. FDG also differs from SFG in concentrating on the individual‐psychological rather than the social dimension of the language user, although the two aspects are of course closely connected in that social interaction is mediated through individual psychologies. A final difference is one of emphasis: whereas FDG has a strong typological orientation, seeking to provide a general theory of linguistic resources, SFG is more centrally concerned with the description of individual languages, only recently having devoted some attention to implications of particular language descriptions for crosslinguistic generalization (cf. Caffarel and Matthiessen 2004). What (p.30) emerges is a picture of SFG as an approach that shares FDG's general aims but is less oriented to cognition and more to the analysis of texts in their social context: indeed many followers of Halliday prefer the contraction SFL, Systemic‐Functional Linguistics, dropping the reference to Grammar.

Nevertheless, a comparison of FDG with its predecessor, FG, does suggest a certain rapprochement with SFG, especially with respect to the work of Fawcett (2000, 2007). In giving equal emphasis to the Interpersonal Level and the Representational Level, FDG shares Halliday and Matthiessen's (2004: 29) concern with the omnipresent dual functionality of language as ‘making sense of our experience, and acting out social relationships’. As for the textual metafunction, which Halliday and Matthiessen (2004: 30) see as an ‘enabling or facilitating function’ reflecting our ability to ‘build up sequences of discourse, organizing the discourse flow and creating cohesion and continuity’, FDG proposes that grammatically relevant textual relations will be accounted for within each of the levels: at the Representational Level, States‐of‐Affairs may be grouped into Episodes, and at the Interpersonal Level, Discourse Acts are grouped into Moves. Indeed both the terms ‘interpersonal’ and ‘move’ are inspired by SFG‐oriented work.

Role and Reference Grammar (RRG), on the other hand, positions itself closer to the formalist end of the spectrum in taking the syntactic unit of the clause as its object of attention, whereas to FDG, being oriented to the Discourse Act, the clause is merely one possible syntactic form. In RRG, the clause receives a single representation, both syntactic and semantic, with information structure being overlaid upon it where relevant. Although FDG recognizes the need for all three types of structure, these are not collapsed into one representation but pertain to three levels of analysis. Apart from a range of technical differences, the central distinction between RRG and FDG is that the former's point of departure is the predicate as a syntactico‐semantic unit, whereas the latter sees itself as providing an analysis of the Discourse Act as an interactional unit, with predicates being introduced into the emerging structure where called for.

Whatever the differences of emphasis and execution, FG (as FDG's predecessor) incorporates various aspects of RRG, while proponents of RRG underline the close relationship between their theory and FG. In particular, Van Valin and LaPolla stress that both share a strong typological orientation (1997: 14), that both assume that levels of analysis are structured by means of layering (1997: 46), and that various FG analyses, such as Rijkhoff's layered view of the Np, have been important to RRG proposals (1997: 640). The close relationship between RRG and FG also emerges from Butler's (2003) assessment, and his more recent work on the comparison of models (Gonzálvez García and Butler 2006) shows this to apply equally to FDG. The RRG view of semantic functions (p.31) (cf. Van Valin 2004), for example, has been very influential on the proposals made here for the treatment of semantic functions (cf. Chapter 3).

Finally, brief mention should be made of a model that displays various similarities to our own, Autolexical Syntax as developed by Sadock (1991). In that theory, the lexicon plays a central role, forging connections between autonomous representations of semantics, syntax, and morphology. Although his theory is unlike ours in separating morphology and syntax and in lacking interpersonal and phonological levels of analysis, it shares our rejection of a derivational model, our commitment to multiple orthogonal representations of linguistic phenomena, and our interest in mismatches between the levels. Where Sadock's model goes further than ours, as presented in the present book, is in presenting and indeed concentrating on the interface conditions between the various levels he proposes; we recognize the importance of such interfaces, but for reasons of space will not enter into a discussion of them in this book.

For all the fruitful overlaps with other approaches, the central point is that FDG sees itself primarily as a grammar of the Discourse Act. Its goal is to describe and, as far as possible, explain the formal properties (syntactic, morphological, and phonological) of Discourse Acts from a functionalist perspective. These formal properties reflect in various ways the dual purposes of the language communicator: to interact successfully and to impart propositional information. The former is modelled at the Interpersonal Level of the grammar, the latter at the Representational. Together these ‘formulating’ levels form the input to the ‘encoding’ levels (the Morphosyntactic Level and the Phonological Level) which yield corresponding structures. The nature of these levels will be dealt with in detail in Chapters 2 to 5.

1.4.3 Typology

Linguistic typology, the study of the principles underlying variation across the languages of the world, is an essential source of inspiration for FDG. Linguistic typology is oriented to laying bare limitations on variation, otherwise known as linguistic ‘universals’, by formulating statements that purport to be true of all languages. Since data is not currently (and is unlikely ever to be) available about all languages, typologists typically, although not exclusively, work with a principled sample of languages (cf. Rijkhoff et al. 1993). Even so, the problems arise that the amount of information available in descriptive grammars varies enormously from language to language and that the methodology inevitably compares languages that are used in very different social circumstances, some of them moreover having a written form while others do not. However, even with these difficulties, it has been possible to elaborate a large body of universals, each of which stands in need of explanation. FDG is a theory that (p.32) is capable of providing a framework for the enunciation and comparison of universals and of offering lines of explanation.

A universal, despite its name, very rarely takes the form of a statement such as ‘All languages are …/ have …’. Rather, they make implicational statements that apply within the grammars of all languages, of the form A ⊂ B, to be understood as saying that property A stands higher in the hierarchy than B, in other words, if a language has property B it will also have property A. This permits of languages with neither A nor B, both A and B, or A only, but excludes languages with B but without A. The universal A ⊂ B could thus be formulated as a negative statement that ‘No language with B will lack A’. Such implication statements may be rendered more complex: where A ⊂ B and B ⊂ C, then A ⊂ B ⊂ C. Where there are multiple implications of this sort, we speak of an implicational hierarchy. The more complex an implicational hierarchy, the more language types are excluded: where n is the number of properties in the hierarchy, the number excluded (m) can be calculated as follows: m = 2n − (n + 1). For this reason, typologists attempt to strengthen their claims by formulating maximally complex hierarchies. Very often, implicational hierarchies do not apply to all languages in the sample; then it is necessary to formulate statistical implicational hierarchies, which indicate the percentage of the sample for which the hierarchy is true.

The hierarchies apply in principle to all domains of linguistic organization (see van Lier 2005 on their applicability in various areas). In FDG, as mentioned above, a strict division is made between four levels of analysis within the grammar; within each, the options available are subject to being organized into implicational hierarchies. At the Phonological Level, the inventory of phonemes can, at least to some extent, be treated in this manner: the nearly absolute universal /n/ ⊂ / m/ ⊂ /ŋ/ indicates that languages pattern systematically in their inventory of (at least these) nasal phonemes. At the level of morphosyntax, we find universals that apply to the relative distance between affixes and stems: Hengeveld (1989) shows that morphological distinctions pertaining to a range of verbal categories are ordered in terms of relative distance from the stem. The hierarchy in (26):

(26) qualitative aspect/agentive modality ⊂ tense/realis‐irrealis/quantitative aspect/negation ⊂ evidentiality ⊂ illocution ⊂ mitigation‐reinforcement

is here to be understood as meaning that morphology with the leftmost meaning has greater proximity to the stem than morphology with the next meaning along, etc.; languages with deviant orderings of morphology are excluded by the hypothesis expressed in this hierarchy. In an FDG framework, two observations should be made about this hierarchy. Firstly, it applies not only to (p.33) morphology but also to the syntax of auxiliary verbs, as for example in English They should (evidentiality) have (tense) begun to (qualitative aspect) work, and indeed FDG treats morphology and syntax at one level, the Morphosyntactic Level. Secondly, the hierarchy also reflects the organization of the grammar into levels: the first three meanings are accounted for at separate layers within the Representational Level, and the last two (the most peripheral) at separate layers of the Interpersonal Level. As we shall see (2.7, 3.3), FDG distinguishes two types of ‘evidentiality’, reportativity (a category of the Interpersonal Level) and evidentiality proper (a category of the Representational Level).

There are also typological hierarchies that apply more purely to semantics. Here the hierarchies indicate, not as in phonology the presence or absence of categories and the implications derivable from those, but rather the degree to which the forms of the language in question enforce semantic distinctions. The classical example in the area of semantics is the hierarchy of colour distinctions developed by Berlin and Kay (1969) and refined in the intervening decades (for quite a radical revision, see Kay et al. 1997). Their original observations, as they relate to semantic distinctions available in languages, may be represented as the semantic hierarchy (27) (these semantic distinctions are said by Berlin and Kay to correlate with physiological properties of the human perception system; this has however been challenged by Saunders and van Brakel 1997):

(27) black and white ⊂ red {⊂ green ⊂ yellow} or {⊂ yellow ⊂ green} ⊂ blue ⊂ brown ⊂ purple and pink and orange and grey

However, this hierarchy is not purely semantic but also morphological in that multimorphemic words for colours (e.g. ‘sky colour’ for blue) are excluded.

We find a similar combination of semantic and morphosyntactic considerations in Hengeveld (1992)'s discovery of hierarchical relations in the way the languages in his sample express the semantics of non‐verbal predication. In the area of non‐verbal predication, the following implicational hierarchy is proposed:

(28) Locative ⊂ Property ⊂ Status ⊂ Possessive

A prediction made by this hierarchy is that, if there is conflation in expression, languages will show the same form for constructions adjacent on this hierarchy. In Turkish, for example, all four types of predication are expressed in the same manner (i.e. total conflation), whereas Babungo uses one strategy for Locative and one other for the remaining three meanings. Spanish introduces a further complication, using one form (estar) for Locative and contingent Properties, but another for inherent Properties and all remaining constructions in the hierarchy. These observations, concerning semantic distinctions, (p.34) clearly pertain to the Representational Level of FDG. Hengeveld (1992) points out that distinctions in this area are also sensitive to the communicative status of the construction as [± presentative], with regard to Locative and Possessive. In FDG, the last‐mentioned generalization relates to the Interpersonal Level, where it is understood in terms of the informational status of Subacts (see 2.7.2), with [+ presentative] applying where one Subact is concurrently Topic and Focus. Thus Turkish uses a special strategy for [+ presentative] constructions that involves the introduction of a copula, cf. (29), whereas the formal distinctions in Babungo are indifferent to the presence or absence of presentativeness, cf. (30):

(29) Bahçe‐de  köpek var.

garden‐LOC dog  COP

‘There's a dog in the garden.’ (Hengeveld 1992: 118)

(30) Zŭ wī   lŭu shɔ̀.

wife POSS.3.SG COP there

‘He has a wife.’

“His wife is there.”

Implicational hierarchies also apply more generally to the Interpersonal Level. As will be reported in greater detail in, Hengeveld et al. (2007) have found that the crosslinguistic comparison of the formal marking of illocutionary distinctions in various languages of Brazil reveals a set of interlocking hierarchies. For example, the presence of a content Interrogative (like English wh‐questions) in a language predicts the presence of a polar Interrogative (like English yes/no‐questions), but not vice versa.

The question now arises as to the status of these hierarchies. Although they can be distinguished at each of the levels of analysis in FDG, they do not in themselves form part of the description of individual languages. Rather, they are derived from such descriptions, so that their theoretical status is that of generalizations that permit empirically falsifiable predictions. In other words, they are hypotheses about possible and impossible language systems, since each hierarchy, as mentioned above, excludes certain combinations of values. In addition, they allow linguistic constructions to be categorized in terms of markedness: those that are most restricted in their occurrence, i.e. those that are rightmost in the representation of the hierarchy, are said to be marked. Markedness can manifest itself in various ways: lesser frequency in use, longer forms (i.e. a greater number of phonemes), more syncretism, less suppletion, etc.

The use of hierarchies has manifested its value in intralinguistic studies, too. It is to be expected that linguistic forms, in extending their meanings (p.35) diachronically, will gradually move from unmarked to more marked items on the hierarchy. For instance, the extension of the meaning of English will from Desiderative to Future to Evidential can be seen as following the first three stages of the above‐mentioned hierarchy (cf. Goossens 1987). In language acquisition and language attrition, as well, the expectation is that the order in which semantic distinctions are gained or lost respectively as formally marked categories will follow relevant hierarchies (cf. Boland 2006 and Keijzer 2007 respectively for evidence in this regard). Quantitative studies of individual languages, too, such as that of Pérez Quintero (2002) on subordination in English, may be expected to reflect the hierarchies in that more marked, hierarchically lower, categories will be statistically less prevalent than unmarked, hierarchically higher categories. These are thus examples of various ways in which FDG can impact the study of individual languages while remaining under the general inspiration of language typology.

The hierarchies that emerge from crosslinguistic and intralinguistic investigations are more than mere descriptive generalizations, however. The hypothesis must be that the hierarchies, although deriving from distinctions made within linguistic systems, reflect aspects of the cognition that drives linguistic communication. The assumption is that the crosslinguistically most widespread distinctions, i.e. those leftmost on the hierarchies, are those with the greatest degree of communicative salience and/or cognitive or physical simplicity. In phonology, for example, from the presence of implosive consonants in a phoneme inventory we can predict the presence of explosive consonants, but not vice versa:

(31) explosive ⊂ implosive

This is generally felt to reflect the greater articulatory complexity of implosives (superimposing an ingressive airstream on the basically egressive airstream of speech). In morphosyntax, we find that languages that have Subject assignment differ systematically with respect to the semantic functions of the units that can undergo Subject assignment, roughly according to the following hierarchy (the semantic functions Locative, Undergoer, and Actor will be explained in 3.6.2 and the details of Subject assignment in 4.4.3):

(32) Actor ⊂ Undergoer ⊂ Locative

In other words, a language permitting Subject assignment to a unit with a Locative function will also permit Subject to be assigned to Undergoer, but the reverse implication does not hold; examples of such languages are, respectively, English, which does permit Locative Subjects of the recipient type, and French, which does not. The validity of the hierarchy may be understood in terms of the anthropocentricity of language: if Subject assignment is a matter (p.36) of perspective‐taking, speakers as active human beings will most naturally take the vantage point of an Actor; seeing a State‐of‐Affairs from the viewpoint of an Undergoer, and then of a Locative requires ever greater cognitive effort.

In pragmatics, a prominent distinction, to which we shall return in, concerns the identifiability of referents. In FDG we distinguish between identifiability for the Speaker and identifiability for the Addressee; the former is equivalent to specificity and the latter to definiteness. In the typological and Optimality Theory literature (cf. Comrie 1989; Aissen 2003) it has been observed that the pragmatic notion of definiteness and the semantic notion of animacy interact in determining case‐marking in ways that can best be captured using hierarchies. Thus Aissen (2003: 437) proposes the following hierarchy for definiteness:

(33) Personal pronoun ⊂ Proper name ⊂ Definite NP ⊂ Indefinite specific NP ⊂ Non‐specific NP

Although FDG does not recognize Np at the Interpersonal Level (but at the Morphosyntactic Level), the Aissen hierarchy comprises various notions that are central to distinctions made at the FDG Interpersonal Level. In FDG terms, we might break the hierarchy down into two as follows, on the assumptions (to be supported in Chapters 2 and 3) that personal pronouns are to be shown as units with an identifiable/specific operator and an abstract head; proper names as units with an identifiable/specific operator and a lexical head; and the remaining categories as all requiring insertion of a lexical head at the Representational Level:


  1. a. [+id, +s] ⊂ [−id, +s] ⊂ [−id, −s]

  2. b. Interpersonal abstract ⊂ Interpersonal lexical ⊂ Representational lexical

FDG derives much of its inspiration from typological work. At the same time, it can provide a coherent model for the kind of language description that feeds into typological investigations. The application of a framework such as FDG, with its multilayered mode of description, to a wide range of languages will permit more reliable comparisons of language systems. Current typological and language‐comparative work tends to eschew particular grammatical models, and indeed Dryer (2006) has recognized that most of this activity is based upon what he calls, following Dixon (1997), ‘basic linguistic theory’. By this is simply meant ‘traditional grammar, modified in various ways by other theoretical traditions over the years’ (Dryer 2006: 212). For Dryer, the functional factors identified by grammarians serve only retroactively to explain instances of language change; he denies that the user of a linguistic system has any access to such factors. In his view, knowing a grammar involves no more (p.37) than knowing a set of brute facts: the language user is unable to draw the kinds of generalization achieved by typologists. While this is clearly correct, it leaves a number of phenomena unaccounted for. If, as we have suggested, language systems do not vary without limit and the differences between them as well as the changes they undergo can be described and circumscribed in terms of implicational hierarchies, those hierarchies must be tapping into matters of general cognitive relevance. Similarly, if—as has often been found—there is a correlation between hierarchical position and frequency in use, this again suggests that the hierarchies are reflecting cognitive preferences. If FDG sees the hierarchies distilled from applications of its principles to various languages as having explanatory relevance, that is because they together define a space within which linguistic activity is constrained to operate.

1.4.4 Language modelling

The predecessor of FDG, Functional Grammar (FG), proclaimed itself to be a quasi‐productive model of the natural language user (Dik 1997a: 1; for detailed presentations of FG see Siewierska 1991 and García Velasco 2003). This was to be interpreted as meaning that the various steps in the grammar should be understood as having a loose parallelism with the temporal sequence of actions conducted by a language user in producing language. Thus the formulation of a communicative intention was seen as being carried out in anticipation of the Addressee's interpretation of the linguistic unit. Encoding was then a matter of linguistic choices judged by the Speaker to be likely to have the desired communicative effect upon the Addressee. For FG, the primary linguistic choice was that of the lexical items. These brought with them various frames, which were fitted together into an underlying predication. This procedure was made fully explicit in the computer model of FG (Dik 1992), which similarly generated linguistic expressions by building upwards from a lexical frame. To the basically semantic underlying predication were added operators and functions which further specified the meaning until every formal property of the linguistic unit could be accounted for.

FDG is like FG in emphasizing the parallels with language production (cf. 1.2.2). However, FDG differs sharply from FG in its architecture, taking not the minimal unit (the lexical predicate) but the Discourse Act as the essential constituent of the entire communicative event initiated by the Speaker as its point of departure. FG was justifiably criticized for treating communicative notions like Topic and Focus (pragmatic functions) as appendages to a semantically complete representation: terms inherited a semantic function from the predicate frame into which they were inserted, e.g. Agent; they could then be adorned with a syntactic function, e.g. Subject; and, finally, to this AgentSubject could be appended the pragmatic function Topic. This suggested (p.38) a primacy of semantics and syntax over pragmatics that ran counter to the principles of functionalism. FDG reverses this by giving pride of place to the Discourse Act.

In FDG's view, each desire to communicate linguistically involves the appearance of a corresponding intention, which is modelled as taking place in the Conceptual Component. This is the impulse that drives the ‘motor’ of the grammar. The intention involves a decision to expend linguistic energy, to perform one or more acts in pursuance of the Speaker's desire to influence the thinking and action of the Addressee. These acts typically do not occur in isolation, but form part of a longer‐term strategy, and as such are known as Discourse Acts. The grammatical form taken by Discourse Acts (which is the ultimate object of FDG) is often influenced by the presence of preceding and following Discourse Acts. For this reason, Discourse Acts are modelled as combining into Moves where there is grammatical justification for doing so. Moreover, the form of each Move may be influenced by preceding and following Moves. Detailed justification for this approach and the hierarchical structure that follows from it will be given in Chapter 2. Every unit analysed in FDG will thus involve the Discourse Act.

This entails a significant difference between FDG and most other models of grammar. Most grammars see themselves as offering accounts of the clause or the sentence, i.e. syntactic units; although there are countless in‐depth studies of smaller syntactic units (the noun phrase or the adpositional phrase, for example), and certain so‐called text grammars have sought to extend the range of grammatical study to larger units, this is typically done against the background of the clause as the essential unit of analysis. In practice, this was also true of FG. However, FG did purport from the earliest days to take the linguistic expression (Dik 1978: 15) as its object of analysis, foreshadowing FDG's orientation to units both larger and smaller than the clause. Note in particular that the sentence, as a ‘discourse unit whose composition and complexity is subject to cultural variation and rhetorical fashion’ (Miller and Weinert 1998: 42), plays very little part in FDG. Although easy to recognize in standardized written languages, it has no straightforward counterpart in oral languages or even in the oral use of languages that do have written forms. As such, it will not be treated within FDG.

What kind of a language model is FDG? It is a fundamental characteristic of functionalist grammars that they seek to relate language form to language function. Those approaches that have attempted to detect the functions underlying the formal distinctions made in language, such as FG, may for this reason be classified as ‘form‐to‐function’: they seek to account for formal properties of syntactic units in terms of their functions in communication. FDG, however, takes a rather more complex position, what we might call a (p.39) form‐oriented ‘function‐to‐form’ approach. It is form‐oriented in providing, for each language analysed, an account of only those interpersonal and representational phenomena which are reflected in morphosyntactic or phonological form. It is ‘function‐to‐form’ in positing a range of functions flowing from the Speaker's communicative intentions, for example a language‐specific set of Illocutions. More specifically, as shown above, communicative intentions are translated in the process of formulation into one or two rather complex functional representations, and these in turn provide the input to the section of the grammar that deals with the formal aspects of utterances, known as encoding.

Formulation involves the strategic arrangement of the communicative intention, which itself is a dynamic, strategic entity, into a temporal sequence of (in principle, discrete) Discourse Acts that may themselves form part of a temporal sequence of larger Moves. As will become clear in Chapter 2, formulation may be restricted to this process, but this usually applies only in the case of relatively simple or ritualized Discourse Acts; in such cases we shall say that only the Interpersonal Level is involved. However, formulation will typically also bring into play the Representational Level, which displays the semantics of the content communicated through the Discourse Acts; Chapter 3 enters into the full detail of this aspect of the formulation process. Formulation is thus distributed over two levels, and deals with the conversion of conceptual material into the functional categories made available by the language system being used by the speaker.

In a simple function‐to‐form approach, there would be no need for the formulation levels. Then one could progress directly from cognitive intentions to encoding, say in the manner of a phrasebook that tells its users how to ask for a beer or complain about cockroaches in their hotel bedroom without giving them any knowledge of the language. It is because of the obvious inadequacy of a direct linking of cognition to expression, which does no justice to the speaker's knowledge of the formulating potential of his/her language, that FDG adopts a form‐oriented function‐to‐form approach. How does this approach work in practice? What it entails is that, for each language examined, the grammarian will consider all its formal properties (variation in constituent order, the repertory of morphological elements, the distribution of particles, the impact of intonation contours, etc.). Then a determination is made which of these grammatical characteristics regularly reflect distinct communicative intentions, for example the desire to indicate the source of one's observations about reality. If the language possesses a set of forms which reveal this aspect of communicative activity, that is a prima facie indication that formulation will make available a semantic category of evidential distinctions; further grammatical analysis will uncover the precise location of the units to (p.40) which these distinctions apply. More generally, those formal distinctions that pertain to constitutive elements of the communicative intention will be seen as encoding the results of formulation of the Interpersonal and Representational Levels.

In addition, languages will display characteristics that cannot be brought into correspondence with distinct communicative intentions (cf. Moutaouakil 2004). These will be regarded as ‘a‐functional’ and will be accounted for as autonomous characteristics of the Morphosyntactic Level and/or the Phonological Level. For example, the languages of the world divide fairly evenly into those with a syntactic head ‐ modifier arrangement and those with a modifier ‐ head arrangement (in FG terms, Postfield and Prefield languages; cf. Dik 1997a: 397). These arrangements are fairly stable, but have been shown to reverse gradually over millennia (with various mixed structures arising in the interim). The very fact that no preference has arisen for one arrangement over the other strongly suggests that it is a‐functional. Neither appears, to use a genetic metaphor, to be better adapted than the other. The Postfield/Prefield distinction will therefore not be derived from deeper communicative motivations, but will be regarded as an autonomous setting at the Morphosyntactic Level, relating to a preference for using unit‐initial or unit‐final positions as basic. Several alternative constituent orders within a language, however, will be seen as flowing from communicatively motivated distinctions, in keeping with the long tradition of such observations and explanations within functionalist grammar.

A crucial aspect of the FDG methodology is that the process of formulation as reflected at the Interpersonal and Representational Levels will not, for any one language, make distinctions that are not reflected in the language in question: thus in a language in which evidentiality is not reflected morphosyntactically (or phonologically), it will simply not be indicated as an operator at formulation. Although the theory of FDG makes evidentiality operators available as an ‘etic’ option (since there are demonstrably languages in which relevant distinctions are made), for each language it has to be determined whether they are pertinent, whether they are ‘emic’: and similarly for every other category introduced in this book. In keeping with the discussion of implicational hierarchies above, none of the distinctions made in FDG inherently carries a claim to universality. In the discussion of parts‐of‐speech, for example, in 3.7, it will be seen that these, too, are subject to a hierarchy such that the theory allows for languages which make no distinctions in this respect at all.

A further crucial aspect of the FDG methodology, inherited from its predecessor FG, is that it constrains potential analyses of linguistic phenomena to those that do not involve the postulation of transformations and filters (p.41) (Dik 1997a: 18–24). These two restrictions ensure that no underlying structures arise that are later discarded. This is warranted from a psycholinguistic point of view, since in this way underlying structures are ‘recoverable from their outward manifestations’ (Dik 1997a: 23). Furthermore, by applying these constraints there is a limit on possible hypotheses concerning the analysis of a linguistic phenomenon, which strongly enhances the testability of these hypotheses.

1.4.5 On using FDG

It may be useful to conclude this section with some reflection on the uses to which FDG can be put by practising linguists of various kinds. Our principal hope is that FDG will offer a structured framework within which it will be possible to enunciate and test linguistic hypotheses. Because FDG provides an overall perspective on linguistic phenomena, comprising four levels of analysis, and being integrated into a four‐component model of the natural language user, it will be possible to articulate those phenomena more clearly and with greater sensitivity to their place in the overall scheme of linguistic things. As Jackendoff (2002: 18), in presenting an encompassing framework for language study, rightly emphasizes, ‘Any adequate theory must begin with the fact that even the simplest sentences contain …rich …structure’ and ‘If one wishes to join the conversation about the nature of language, one must recognize and acknowledge this complexity’.

As we stated above, FDG seeks to bring order to this complexity by providing formalizations of its claims. The purpose of these formalizations is to provide a rigorous framework in which linguistic claims can be enunciated, and then tested, substantiated, or disproved, and then submitted to further refinement or sophistication. At the same time, it provides a structure for the observation of linguistic phenomena, and in this way is involved in the entire cycle of research, from observation to prediction, to the testing of prediction through further observation, which leads to new predictions, and so on. A clear example of this cycle pertains to the notion of layering within FDG: the principle of layered structure was first developed in FG (cf. Hengeveld 1989, inspired by Foley and Van Valin 1984) for what is now roughly speaking the Representational Level, and only later with the emergence of FDG was the hypothesis formulated that a similar degree of layering might be found at the Interpersonal Level (Hengeveld 2004a and references cited there); this in turn engendered the expectation that notions established for the Representational Level such as operator and modifier could be equally relevant for the Interpersonal Level (Hengeveld 2004b); and later, this led to the proposal that the FDG notion of layering could also link up with the already generally accepted hierarchical organization of morphosyntax and phonology (as in this (p.42) book). In short, the parallelisms between the levels themselves have arisen from the cycle of observation and prediction, and provide the basis for new cycles.

Although FDG provides precise representations for its claims, the formalisms it uses should not be confused with the formal languages employed by truth‐conditional semanticists and in radical formalism. Ultimately, while every effort is made to keep them mutually consistent, clear and usable, the representations are but a means to the end of insightful analysis of linguistic phenomena.

As is implicit in what has just been said, one form of work within FDG will engender proposals for the development and improvement of the theoretical apparatus. Of particular interest for the further advancement of FDG will be future studies on the interface issue, the question of how best to connect the four concurrent representations that characterize the current model. It is to be expected that mismatches across the various levels will be of particular importance in this enterprise.

As for research primarily oriented to using rather than reforming the model, we may differentiate various emphases. FDG seeks to provide a framework for typological and language‐contrastive and language‐contact work that is neutral with respect to any specific language type. At the same time, it can be employed for the description of individual languages (but always with a view to the implications for other languages), as well as the growth and decline of languages in contexts of acquisition and attrition. FDG lends itself to the investigation of the crosslinguistic distribution of interpersonal, representational, morphosyntactic, and phonological categories, but also to the detailed examination of individual phenomena within a single language. For various FDG treatments of phenomena in single languages, see the articles in van Staden and Keizer (fc.), Hengeveld and Wanders (fc.) and Hattnher and Hengeveld (2007).

As we emphasized above, FDG, despite its name, is not a functionally oriented Discourse Grammar (in the sense of an account of discourse relations). Rather, it is an account of the inner structure of Discourse Acts that is sensitive to the impact of their use in discourse upon their form. From this viewpoint, there is little to be gained from an application of FDG as a tool for the inductive examination of texts or segments from the transcription of speech. As Butler (2004) points out, the proper relation between functionalist theories and corpora is for the former to provide hypotheses which can be tested against data; for a fine FDG example, see Anstey (2006). The description of data in corpora need be no more exhaustive than is necessary for the analytical task at hand. As we will show in the next section, FDG provides for the possibility of formally simplifying non‐essential aspects of its representations with a view to focusing on the essential questions.

(p.43) 1.5 Notational conventions

As the formalisms for the various levels are further elaborated in the following chapters, they will become increasingly detailed and complex. Such detail is necessary for the model to achieve precision and predictive power. In order to enhance readability, we apply a number of special conventions within the formalisms.

The first concerns the use of different typefaces for variables at the different levels of analysis, tacitly applied in the preceding text: capitals at the Interpersonal Level (e.g. ‘M’ for Move), lower case at the Representational Level (e.g. ‘ep’ for episode), title case at the Morphosyntactic Level (e.g. ‘Np’ for Noun Phrase), and small capitals at the Phonological Level (e.g. ‘PP’ for Phonological Phrase). At all levels operators are given in lower case and functions in title case.

Secondly, in many cases not all details are necessary for the analysis of the phenomenon at hand. For these situations we use a special symbol ‘‐’ to indicate the beginning and the end of a fragment that is not further analysed in detail. Thus, if we are just interested in the nature of the relation between Discourse Acts within a Move, the analysis of (35) may be as in (36):

(35) Watch out, because there is a bull in the field.

(36) (MI: [(AI: ‐watch out‐ (AI)) (AJ: ‐there is a bull in the field‐ (AJ))Expl] (MJ))

in which it is indicated that AJ is grammatically encoded as an explanation of AI, but no further claims are made as to the internal structure of the Discourse Acts.

Similarly, at the Representational Level, if one is interested in the semantic functions of arguments irrespective of their internal complexity, (37) may be represented as in (38):

(37) My neighbour bought a book about bullfighting.

(38) (fi: [(fj: buy (fj)) (xi: ‐my neighbour‐ (xi))A (xj: —a book about bullfighting‐ (xj))U] (fi))

in which it is indicated that xi is the Actor and xj the Undergoer in the buying State‐of‐Affairs, but no details are given about the internal structure of these two descriptions of Individuals.

In a similar way morphosyntactic representations may be simplified if, for example, one is interested in Phrases rather than in Words; and phonological representations may be simplified if, for example, one wants to concentrate on Phonological Phrases rather than on Phonological Words. An example of (p.44) the former is given in (20a) above, in which (Npi: (Nwi: ikPro (Nwi)) (Npi)) is reduced to (Npi: ‐ik‐ (Npi)), and so on. An analogous reduction of (20b) would yield the following phonological representation:

(39) (ui: [(IP i: [(PP i: ‐kυau‐ (PP i)) (PP j: ‐datikυam‐ (PP j))] (IP i))] (U i))

Another important convention applied within FDG formalizations concerns the use of subscripts for variables. In presenting general frames and templates we use numerical subscripts, indicating that the variable is uninstantiated. In representations of actual examples we use alphabetical subscripts, indicating that the variable is instantiated. Thus, the general frames underlying (36) and (38) are (40) and (41) respectively:

(40) (M1: [(A1) (A2)Expl] (M1))

(41) (f1: [(f2) (x1)A (x2)U] (f1))

Square brackets are used to keep elements together that are in a non‐hierarchical relationship with respect to one another, but together are hierarchically subordinate to a higher layer, as in (40), where the two Discourse Acts are non‐hierarchically related, but together within the scope of the (M1)‐variable, or as in (41), where three semantic categories are in a non-hierarchical relation, but all three within the scope of the (f1)‐variable. Finally,+ curly brackets are used in cases in which it is desirable to explicitly indicate the optionality of elements.

In running text, words are capitalized when they are used as technical terms as applied within the FDG framework. Thus, we use capitals for analytical units such as Move, Propositional Content, Verb Phrase, and Phonological Word, but also for operators and functions such as Past, Undergoer, Subject, etc., even though the latter are not capitalized in representations.

One aspect of our glossing conventions requires brief comment. When giving examples from languages other than English, we supply a morphemic gloss according to the principles of the Leipzig Glossing Rules, <http://www.eva.mpg.de/lingua/files/morpheme.html>; the glosses listed as an appendix to those rules have been used, supplemented by many more, as required by the data examined (see the list of Abbreviations and Symbols preceding this chapter). The morphemic gloss is followed by a idiomatic translation of the example into English between single quotation marks. In addition, but only where this aids the understanding of the example, we also provide a more literal translation between double quotation marks. For a case in point, see (1) in this chapter.

(p.45) 1.6 Structure of the book

This book offers a complete overview of Functional Discourse Grammar, understood as the Grammatical Component of the full theory of verbal interaction set out in Figure 2 above. The next four chapters deal at length with the four levels of analysis, beginning with the two formulation levels and moving on to the two encoding levels. Accordingly, Chapter 2 deals with the Interpersonal Level and Chapter 3 with the Representational Level. The overall structure of these two chapters is, as far as this is possible, the same: the aim is to bring out the default relations that hold between the inner workings of each level. Chapter 4 presents the Morphosyntactic Level, working down from the highest to the lowest layers of analysis, and the same technique applies in Chapter 5, devoted to the Phonological Level.

The data examined in the following chapters have been drawn from a variety of sources. The principal source has been a wide range of grammatical descriptions of the languages of the world; our intellectual debt to the authors of those descriptions is enormous. Alongside consultation of our own intuitions about languages we know well, we have also had recourse to the Internet as a source of data. For aesthetic reasons, and also in the knowledge that URLs can change or disappear overnight, we have in such cases simply used the indication ‘Internet’.

Since this book is primarily oriented to the presentation of a theory rather than the analysis of data (although we hope to persuade the reader that the theory promises interesting analyses), the linguistic examples are cited above all to illustrate the potential of FDG. Our analyses of particular phenomena should therefore be taken as indicative rather than as representing any claim to a definitive FDG statement (if such were even possible) about the phenomena in question.