Misrepresenting Halliday (1994) On Its Representation Of Clause Structure

Fawcett (2010: 113-4):

Halliday's introductory writings often imply that each line of structure corresponds to one of his three (or four) 'metafunctions', i.e., broad types of  'meaning'. There are four such metafunctions when he splits the 'ideational' metafunction into the 'experiential' and the 'logical', e.g., as described in Halliday (1977/78:128), but in his introductory writings he frequently omits the 'logical' and only illustrates three (as in Figure 7). The interesting point here, then, is that the four 'strands of meaning' shown in Figure 7 — and in all equivalent representations in IFG — do not correspond to Halliday's four 'metafunctions'. Thus the THEME and INFORMATION lines of analysis both belong in his 'textual' metafunction; the MOOD analysis (which itself contains two lines) is the 'interpersonal' metafunction; and the TRANSITIVITY is 'experiential'. The fourth 'metafunction' (i.e., the 'logical') is omitted in Figure 7, because Halliday only introduces it when there is a 'logico-semantic relationship' between two or more clauses.

Blogger Comments:

[1] This is misleading.  Halliday explicitly states — over and over, across many publications — that each line of structure corresponds to the meanings of a separate metafunction.  For example, Halliday (1994: 35):

[2] To be clear, the logical metafunction is concerned with relations between units in complexes all along the rank scale, whether clauses in clause complexes, groups or phrases in group or phrase complexes, words in word complexes, or morphemes in morpheme complexes.

[3] This is misleading. To be clear, Fawcett's Figure 7 illustrates only three metafunctional strands of meaning: textual, interpersonal and experiential, and misrepresents information as a structure of the clause (instead of the information unit). Reminder:

[4] This is misleading. To be clear, there are no representations in IFG that are equivalent to Fawcett's Figure 7; that is, there are no representations showing all three lines of clause structure in one diagram, though there are several that show both MOOD and TRANSITIVITY. The nearest equivalents appear in Appendix 1, an analysis of the 'Silver' text, but these are not confined to clause structure, and additionally demonstrate both cohesion and information analyses, neither of which constitutes clause structure, as well as the analysis of grammatical metaphor.  For example, Halliday (1994: 371):

Clearly, Fawcett confuses this series of composite 'text analysis' diagrams in IFG with diagrams merely illustrating clause structure.

[5] This is misleading.  Mood structure represents only one line of meaning, interpersonal meaning, but the structure itself involves functional constituency.  For example, the Mood block of finite clauses consists of the Subject and Finite elements (± modal Adjuncts), and the Residue can consist of the elements Predicator, Complement(s) and circumstantial Adjunct(s).

Misrepresenting The Conflation Of Functional Elements

Fawcett (2010: 113):

Let us now examine Figure 7 more closely. It show four clearly separated lines of box diagrams (one containing two lines of structure within it), such that each of the four major lines corresponds to one of four major types of 'meaning'. The two basic assumptions upon which such a diagram rests are (1) that there should be a separate structural representation for each major strand of meaning (including two completely distinct structures within 'textual' meaning), and (2) any one 'element' of the clause is likely to consist of the 'conflation' of two or more 'functions' from two or more different lines of meaning. For example, the description in Halliday (1994:30f.) of the duke in the duke gave my aunt this teapot states that, in his approach, the three 'functions' of 'Subject', 'Actor' and 'Theme' are typically — but not necessarily — mapped onto one another to form a single 'element' of the clause (just as the same three 'functions' are in Figure 7).

Blogger Comments:

[1] Reminder:

A functional analysis (with syntagm) that is consistent with SFL Theory is presented below for contrast:

we	would	visit	Mrs Skinner	every Sunday
Theme	Rheme
Subject	Finite	Predicator	Complement	Adjunct
Mood		Residue
Actor	Process		Scope	Extent
nominal group	verbal group		nominal group	prepositional phrase

[2] This is misleading. The clause has only one textual structure (Theme^Rheme).  Information is not a system of the clause, but of the information unit — whether or not the information unit happens to be coterminous with the clause.

[3] To be clear, the "element" that is the conflation of different functions is a formal constituent of the clause, represented as a unit of the syntagm.  For example, it is the initial nominal group onto which the three functional elements of Theme, Subject and Actor are mapped.

The 'Structure Conflation' And 'Element Conflation' Models

Fawcett (2010: 112-3):

In the latter paper, Halliday writes that 

the clause has a number of different but simultaneous constituent structures [my emphasis] according to which set of options [each of which corresponds to one 'metafunction'] is being considered. (1969/81:143) 

This concept of "simultaneous structures" is the concept that underlies the representation in Figure 7, and all of the similar representations in IFG and the many derived works. Notice, however, that the words "simultaneous structures" in the description of the clause cited above make a very much stronger claim about the size of the 'simultaneous' units than do the 'realisation statements' on the immediately preceding page of the same paper. Table 1 on page 142 of Halliday (1969/81) simply show that pairs and trios of 'functions' are sometimes brought together by a conflation statement to form a single element of the clause, the meanings of such elements being derived from two or three different 'metafunctions'. (It gives no rule that conflates whole structures.) Halliday describes this second and less ambitious type of 'conflation' lower on the same page as the quotation above when he says that 

each element [my emphasis] of [the] structure is a complex of functions, a set of structural 'roles' specified as realisations of the options selected [in the system network]. (Halliday 1969/81:143) 

These two statements correspond to two different models of 'conflation' that will be introduced later in this chapter, and we shall refer to them as the 'structure conflation' and the 'element conflation' models. The second concept could of course be included within the first, in principle — but we shall find that in practice this does not happen. This difference between the two types of 'conflation' will be crucial, as the argument to be presented in this chapter unfolds. 

Blogger Comments:

[1] To be clear, Halliday (1969), Options and Functions in the English Clause, is a very early exploratory paper, published 31 years before the first edition of Fawcett's volume.

[2] To be clear, the 'simultaneous structures' of the clause are the three function structures, varying according to metafunction, that are integrated into a single syntagm — a syntagm being a string of clause constituents (different classes of group ± prepositional phrase).  Halliday & Matthiessen (2014: 74):

The clause, as we said, is the mainspring of grammatical energy; it is the unit where meanings of different kinds, experiential, interpersonal and textual, are integrated into a single syntagm.

Halliday & Matthiessen (2014: 212) provides the following explanatory example:

[3] To be clear, the single "element" to which a complex of functions — a set of structural rôles — is assigned is a clause constituent, a group or phrase of the syntagm; see Figure 5-1 above. It is likely that it is Halliday's use of the term 'element' for 'constituent' in the 1969 paper that is Fawcett's motivation for citing such an early paper instead of a more recent, up-to-date one.

[4] To be clear, in SFL Theory, the rank scale unit 'clause' serves as the entry condition for three simultaneous metafunctional systems, each of which specifies a different structural realisation that is mapped onto a syntagm of units of the rank below (group/phrase).  On this model, there is no need for a "rule that conflates whole structures".

[5] To be clear, the notion of 'structure conflation' misrepresents Halliday's model; see [2] and [4] above.

[6] It will be seen in later posts that Fawcett's argument is invalidated by the confusion caused by its omission of the notion of a syntagm as integrating the different metafunctional meanings.

Misrepresenting Halliday (1994) On Structure And Misanalysing A Transitivity Structure

Fawcett (2010: 111-2):

In the text-descriptive strand of the work in the Sydney Grammar, the functional structure of a text that consists of a simple clause is modelled in the way shown in Figure 7. It is a set of box diagrams, each of which represents several simultaneous 'strings' of elements (using the term loosely). These 'box diagram' representations of structure represent a simple constituency relationship of the 'flat tree' type, i.e., of the type with only one layer of branching. We should be clear from the start, then, that the use of 'box diagrams' for representing structure is not a 'structurally neutral' form of representation, such as Halliday's writings seem to imply would be desirable, but a type of constituency — as Halliday himself in fact recognises (1994:16).

 

Blogger Comments:

[1] To be clear, the analysis in Figure 7 is misleading in two ways.  Firstly, INFORMATION is not a system of the clause, but of the information unit, which is only coterminous with the clause in the unmarked case.  Secondly, and more importantly, it misrepresents the transitivity structure of the clause, which, on the SFL model, can be represented as:

we	would visit	Mrs Skinner	every Sunday
Actor	Process	Scope	Extent

That is, the clause is not transitive, but a scoped intransitive, with Mrs Skinner as the Scope/Range of the Process, not the Goal/Medium and We as the Actor/Medium, not the Actor/Agent. 

In other words, Mrs Skinner is not the medium through which the visiting process unfolds, but the domain over which the visiting process takes place; and We is not the external agent of the visiting process, but medium through which the visiting process unfolds.

Trivially, "Circumstance" — like 'participant' — is not an element of function structure, but a class of functions (and hence not capitalised).

[2] This is misleading. To be clear, 'flat' constituency refers to rank, not function structure.  Halliday (1995[1993]: 273):

'Rank' is constituency based on function, and hence 'flat,' …

[3] This seriously misrepresents Halliday's view.  Halliday (1994: 35-6) first outlines the different types of structure favoured by the different metafunctions:

and then explains:

It is the segmental type of structure, with clearly separated constituent parts organised into a whole, that has traditionally been taken as the norm in descriptions of grammar; the very concept of 'structure', in language, has been defined in constituency terms. This partly because the kind of meaning that is expressed in this way: experiential meaning has been much more fully described than meaning of the other kinds. But there is also another reason, which is that constituency is the simplest kind of structure, from which other, more complex types can be derived; it is the natural one to take as prototypical… .

For both these reasons, in the remaining chapters of Part One (Chapters 3-5) I shall use the constituent-type descriptions of structure, merely pointing out now and again where they do not tell the whole story. As far as possible, each strand of meaning in the clause will be described independently in its own terms. 

[4] This is misleading. What Halliday actually says is that constituency is but one type of structure. Halliday (1994: 16):

Constituency is a form of structural organisation; in other words, it is part of the mechanism whereby meanings are put into effect. It is an extremely simple but powerful device, whereby parts are built into wholes, and these again as parts into larger wholes, but with different organic configurations at each step. … But precisely because it is such a powerful device, it is important to make clear that it is not sufficient by itself. We shall see in Chapters 3-5, when we consider the different kinds of meaning that are constructed in language, that these make use of constituency for their expression in different ways and to varying degrees. As one explores language more deeply, constituency gradually slips into the background, and explanations come more and more to involve other, more abstract kinds of relationship. For this reason it is essential not to to conceive of a language as an inventory of structures, each set up as a whole consisting of some ordered arrangements of parts.

Fawcett's Summary Of His Argument So Far

Fawcett (2010: 106):

We noted in Chapter 5 that what is missing in "Systemic theory" is an account of those concepts that are required to describe the instances of language at the level of form — i.e., the concepts of "Categories". But we have seen in the present chapter that, despite initial appearances, these concepts are still in use in IFG — if only in the background. 

There are two main reasons for the very considerable differences between the concepts presented in "Categories" and "Systemic theory": (1) the changes to the theory in the 1960s (as outlined in Chapter 4), which have removed the concept of the system from the level of form, and (2) the unexplained lack in "Systemic theory" of a section on the outputs at this level. The lack of a specification of a theory of the 'syntax of instances' in IFG means that Halliday has not made a statement about this aspect of the theory since 1961. Yet, as we saw in the last section, this is an essential part of a full theory of syntax.

Blogger Comments:

[1] This is very misleading. As previously explained, "Systemic theory" (Halliday 1993) is an encyclopædia entry that sets out the architecture, history and development of Systemic Theory in 3 pages.  It stresses that system, not structure, is the fundamental organising concept of the theory.  (By 'instances at the level of form', Fawcett means grammatical structure.)

[2] This is very misleading.  As previously explained, IFG (Halliday 1994) provides the grammatical structures that realise grammatical systems, organised according to rank scale, because rank units constitute the entry conditions to the systems that specify grammatical structures.

[3] This is misleading. To be clear, the single most important reason for the differences in "Categories" (Halliday 1961) and "Systemic theory" (Halliday 1993) is that they outline different grammatical theories: the former sets out Scale-&-Category Grammar, while latter sets out Systemic Functional Grammar.

[4] This is very misleading, because it misrepresents Fawcett's model as Halliday's.  The 'level of form' is Fawcett's model, not Halliday's, and the absence of the concept of system at the level of form is Fawcett's model, not Halliday's.  Halliday's model distinguishes linguistic strata as semantics, lexicogrammar and phonology/graphology, with systems posited for each stratum.  Grammatical form is modelled as a rank scale of units, with each unit providing the entry condition to systems of functional features at that rank.

[5] This is misleading.  'Syntax' is Fawcett's model, not Halliday's. Halliday explicitly rejects the "syntax" approach to grammatical theorising, as he made clear in the first two editions of IFG. Halliday (1994: xiv):

The Need For An Explicit Theory Of Systemic Functional Syntax

Fawcett (2010: 104-5):

The practical demonstration of this fact (as I take it to be) is that this aspect of a systemic functional theory of syntax is needed in two important areas of application. The first is that of specifying what a computer needs to know, in order to analyse a string of incoming words into the syntactic structure that relates them. … The second area of application is the analysis of texts by humans. It is not surprising that broadly the same set of concepts is required in both cases, and this is why a book that is written to help people to analyse texts invariably makes at least some use of a theory of instances. Indeed, this is precisely why we find Halliday making such frequent but informal use of the concepts of 'class of unit' and 'element of structure' in IFG …

The fact that computer models of systemic functional grammars cannot simply be turned into natural language understanding machines by reversing them underlines this book's main argument, i.e., the argument that there is a need for an explic[i]t theory of systemic functional syntax. It provides evidence from this important sub-field of computational linguistics research that we can place alongside the less explicit evidence from the needs of the text analyst. In other words, it is a clear demonstration of the need for the sort of theory of 'instances of syntax' that is to be presented in Part 2 of this book.

Blogger Comments:

[1] To be clear, Fawcett's argument is that an explicit theory of systemic functional syntax is needed because it is needed for computational applications.

[2] To be clear, as IFG demonstrates, grammatical structures can be analysed by humans without resort to a theory of syntax.

[3] To be clear, in Fawcett's model (Figure 4), selection expressions (meaning) and structures (form) are misunderstood as instances, as previously explained.

[4] This is misleading. To be clear, in SFL Theory, formal units are modelled as a rank scale of clause, group/phrase, word and morpheme, each of which serves as the entry condition to systems that specify structures at that rank. This is why IFG, which outlines the structures at the ranks of clause and group, is organised in terms of the rank scale. The notion of 'class of unit' is most relevant at the rank of group, because different classes of group — nominal, verbal, etc. — have different elements of structure.

The Two Processes Of Generation And Understanding

Fawcett (2010: 104):

The reason why the two processes of generation and understanding cannot be treated as mirror images of each other is that each of the two processes of generation and understanding involves a different kind of 'problem-solving'. This arises because the two processes operate in different directions. More specifically, the evidence that is taken into account when deciding to make one analysis rather than another when parsing the syntax of a text is different from the evidence that is drawn on to determine the choices in generation. In parsing, the available linguistic data are those that are observable in the surrounding text at the level of form; in generation in the decisions are taken at the level of meaning (or in a higher component), so that the data that affect choices are at the level of semantics (or above it).

Blogger Comments:

[1] To be clear, here Fawcett is not concerned with modelling language, but with the problems that a computational linguist needs to solve when trying to use a theory of language for the purposes of text generation and parsing.  Moreover, the theory of language being used is his own 'flowchart' model ('syntax', 'level of form', 'level of semantics'), which, as previously demonstrated, is inconsistent with the architecture of SFL Theory.

[2] To be clear, when applied to language, rather than computation, this is at odds with SFL theory, where structural interpretations are made on the basis of systemic contrasts. Halliday & Matthiessen (2014: 49):

Giving priority to the view ‘from above’ means that the organising principle adopted is that of system: the grammar is seen as a network of interrelated meaningful choices. In other words, the dominant axis is the paradigmatic one: the fundamental components of the grammar are sets of mutually defining contrastive features. Explaining something consists not in stating how it is structured but in showing how it is related to other things: its pattern of systemic relationships, or agnateness (agnation).

[3] To be clear, this confuses linguistic data with linguistic theory. Linguistic data are what is to be modelled, whereas 'level of form' and 'level of meaning' are dimensions of Fawcett's theoretical model (Figure 4).

In terms of SFL Theory, lexicogrammar is a purely abstract level of representation. Halliday & Matthiessen (1999: 613-4):

For these to be possible you need a semiotic of a different kind, one that allows for a purely abstract level of representation "in between" the two faces of the sign, To put this another way (as we did at the beginning of the book), the sign has to be deconstructed so that, instead of content interfacing directly with expression, the relationship is mediated by a systematic organisation of form (a lexicogrammar).

The Concepts Required For The Structural Description Of A String Of Words

Fawcett (2010: 104):

For the purposes of analysing a string of words in a text, it is necessary to be able to specify the concepts that are required for the structural description of that string of words — and to be able to do so, moreover, without drawing at every stage on one's knowledge of the procedures by which the string of words was generated. In other words, it must be possible to carry out the process of analysing the syntax of a text (traditionally known as parsing) independently of the process of generation.

Blogger Comments:

To be clear, in the case of SFL Theory, the concepts that are required for the structural description of a string of words in a text are specified in all four editions of IFG:

1. Halliday (1985)
2. Halliday (1994)
3. Halliday & Matthiessen (2004)
4. Halliday & Matthiessen (2014).

On The Unjustifiable Approach Of Halliday (1993)

Fawcett (2010: 104):

One might expect, in principle, that it would be sufficient to characterise the 'form potential', and that the theory required to describe the instances would follow automatically from this — and if this was so the approach taken in "Systemic theory" would be justified. But in practice this does not provide the framework of concepts that is required in the applications to which a model of language is put. In other words, it is not enough to model language in its generative mode (or 'at rest'); the theory must supply the concepts that are required for modelling the use of language in both the generation and the analysis of text-sentences — and it must do this for each level of language that is recognised in the theory.

Blogger Comments:

[1] This is misleading. This presents Fawcett's model, Figure 4, as if it represents the architecture of SFL Theory.  By 'form potential' Fawcett means realisation rules (divorced from system networks), and by 'instances' Fawcett means syntagmatic structures (rather than instances).

[2] This is misleading. "Systemic theory" (Halliday 1993) is merely a brief article in an encyclopædia that sets out in 3 pages the history, basic concepts, and development of Systemic Theory, written for readers of encyclopædias, rather than an academic publication for scholars seeking a deep understanding of the theory.  Fawcett's complaint is that a 3-page summary does not set out the theory in all its complexity.

[3] This is misleading. The systems and realisation statements that specify grammatical structures are set out in Lexicogrammatical Cartography (Matthiessen 1995), with some also incorporated in the third (and fourth) editions of IFG (Halliday & Matthiessen 2004 (and 2014)).  The grammatical structures that realise these systems are set out in all editions of IFG.

[4] To be clear, Fawcett's compliance with this stipulation amounts to splitting one level of abstraction (systems and their realisation rules) into two levels (meaning and form), in the case of the left-hand column of Figure 4, and misunderstanding selection expressions and syntagmatic structure as instances, in the case of the right-hand column of Figure 4.

Misrepresenting Halliday On Theory Exposition

Fawcett (2010: 103-4):

The fact that "Systemic theory" omits any specification of the part of the theory that would be used in a description of the outputs from the grammar leaves the reader in a highly unsatisfactory position. Since the paper is entitled "Systemic theory", this omission seems to imply that the theory does not need to specify these concepts. On the other hand, it is just possible that Halliday has omitted them on the grounds that if one specifies the 'form potential' in the realisation statements, there is no need to specify the outputs, on the grounds that this is what the realisation statements do.

Whatever Halliday's position on this issue, I wish to make clear that my view is that a theory does indeed have the responsibility to specify these concepts explicitly. In other words, it is the task of a theory of SF syntax to specify both the apparatus that generates the text-sentences that are the outputs from the grammar (the realisation component) and the concepts that are required to model those outputs. Moreover, both of these must be treated as int[e]gral parts of the theory, as the use of the theory for modelling the generation and understanding of language in computers shows clearly. 

Blogger Comments:

[1] This is misleading. To be clear, "Systemic Theory" is an article in an encyclopædia that outlines the origins, architecture and development of the theory in 3 pages; it is oriented to theory rather than text description.  On the other hand, the "outputs of the grammar" had been previously set out in 350+ pages of the first edition of IFG (Halliday 1985), a text oriented to text description rather than theory.

[2] This is misleading, because it misrepresents Halliday's model (realisation statements in system networks) as Fawcett's model (realisation rules at the level of form).

[3] This is misleading, because these concepts are set out in great detail in the four editions of IFG (1985, 1994, 2004, 2014).

[4] This is misleading. To be clear, the grammatical systems that are realised as grammatical structures are set out in Lexicogrammatical Cartography (Matthiessen 1995) and the grammatical structures that realise grammatical systems are set out in the four editions of IFG.  Moreover, the most important grammatical systems are also included in the third and fourth editions of IFG.

[5] This is misleading, because it misrepresents Halliday's model (Systemic Functional Theory) in terms of Fawcett's approach (a theory of syntax). As Halliday (1994: xiv) makes clear: