Why 'Syntagmatic Probabilities' Are No Replacement For The Rank Scale

Fawcett (2010: 338):

These systemic probabilities and the model's ability to vary them play a major role in the computer generation of text in the Cardiff Grammar. But from the viewpoint of the text analyst — whether a human or a computer — what is needed is the 'realisation' of these systemic probabilities as structural probabilities. In other words, probabilities that are ultimately semantic and paradigmatic have to be expressed in terms of probabilities that are formal and syntagmatic. And, within the wide range of syntagmatic probabilities at the level of form, is the particular set which states the relative likelihood that a given unit will fill a given element of another unit (or an element of the same class of unit higher in the structure). It is this aspect of the syntagmatic probabilities that replaces the concept of the 'rank scale'.

Blogger Comments:

[1] To be clear, the notion of systemic probabilities realised as 'structural/syntagmatic probabilities' is inconsistent with Fawcett's model (Figure 4, p36), because probability is the quantification of potential, whereas Fawcett (incongruously) models syntagmatic structure as instance.

[2] To be clear, syntagmatic probabilities can not replace the concept of 'the rank scale' because the rank scale is a model formal constituency, whereas syntagmatic probabilities, as form-function relations, are not.

Realisation Operations

Fawcett (2010: 281):

The general concept of 'realisation' is made specific through five major types of realisation operation. As we saw in Section 9.2 of Chapter 9, it is they, together with the potential structures, that specify the 'form potential' of a language.

Notice, however, that when they are applied (i.e., to a selection expression of features generated on a traversal of a system network, as described in Appendix A), they generate syntactic structures. The first four operations directly generate four of the relationships in syntax to be described below. And the last two provide the framework for generating structures with the recursion of co-ordination, embedding or re-iteration. Thus the realisation operations in the grammar are directly related to the relationships in the syntax of an output from the grammar — while not, as I emphasised in Chapter 9, being the same as them. In other words, we need both a theory of 'syntax potential' and a theory of syntactic instances'.

Blogger Comments:

[1] To be clear, it is the operations that are specific, not realisation. Realisation is the same in each case: the relation between two levels of symbolic abstraction.

[2] To be clear, potential structures do not feature in the representation of Fawcett's model (Figure 4):

[3] To be clear, in Fawcett's text-generation algorithm, realisation operations are misconstrued as:

(i) the form that realises the meanings of system networks and

(ii) the potential that is instantiated as syntactic structure.

[4] To be clear, Fawcett's argument for the distinction between 'syntax potential' and 'syntactic instances' is simply that the two are related, but not the same. That is, he does not provide any argument as to why, or how, syntactic structures can be understood as instances of realisation operations.

The Concept Of 'System' Has No Rôle In Systemic Functional Syntax

Fawcett (2010: 279-80, 280n):

The fourth "fundamental category" in "Categories" was, of course, the concept of system. It was Halliday's re-interpretation of this term in 1966 as 'choice between meanings' that made it the fundamental concept of a new model of language, and so of a new theory of 'meaning' (as we saw in Chapters 3 and 4). It therefore has no role in the present model of syntax.²

² It would be possible to envisage a model with a set of system networks that represented choices at the level of 'pure' form such that these were 'predetermined' by choices made at a higher level of 'semantics'. Hudson's work (e.g., Hudson 1971) is presented as a systemic model of syntax of just this type (with no ambition to model choices between meanings), but this is not the direction in which Halliday has led Systemic Functional Linguistics. I would claim that the fact that the Cardiff Grammar can indeed operate with system networks that are explicitly intended to model choices in meaning and that can be directly realised in syntax at the level of form vindicates Halliday's original hunch in the 1960s that the system networks of TRANSITIVITY, MOOD, THEME etc. should be regarded as modelling choices between meanings.

Blogger Comments:

[1] To be clear, the concept of system has no rôle in Fawcett's Theory of Systemic Functional Syntax.

[2] To be clear, Fawcett has provided no evidence of this "fact". He has provided no system networks of meaning, and no realisation rules that specify how choices in systems are realised as his structures. Because his structures are not those of SFL Theory, he cannot claim that they realise the SFL systems of TRANSITIVITY, MOOD and THEME.

[3] To be clear, even if this "fact" were demonstrated, it would not vindicate something that is not true (see [4] below). More importantly, such a "fact" would not distinguish a model in which the system-structure relation is modelled axially — paradigm to syntagm — as in SFL Theory, from a model in which it is modelled stratally — meaning to form —  as in the Cardiff Grammar, because, in both models the relation between system and structure is the same: realisation (symbolic abstraction). 

However, this is undermined by the fact that the Cardiff Grammar (Figure 4) misunderstands the axial distinction between paradigm and syntagm as the distinction between potential and instance, as previously explained.

[4] This is misleading, because it misrepresents Halliday's "original hunch" on the theoretical location of these systems. From the first formulation of these systems, they have been located on the stratum of lexicogrammar. However, because SFL Theory models lexicogrammar in terms of the meanings they realise, these systems are interpretations of lexicogrammatical form (the rank scale of constituents) as meaning. As Halliday (1985: xix, xx) explains:

A Common Framework For Comparing SFL Theory And The Cardiff Grammar

Fawcett (2010: 274-5):

The starting point for the summaries that follow must be the framework for a modern SF grammar that we established in Chapter 3. As Figure 4 in Section 3.2 of that chapter showed, its two principle characteristics are (1) that it consists essentially of the two levels of meaning and form, and (2) that there is at each level (a) a component that specified the potential, and (b) an Output' from the grammar, i.e., the instance at that level.

As we have seen, an additional advantage of this formulation of the model is that it is at a sufficiently high level of generalisation to provide a common framework in which we may compare the Sydney the Cardiff Grammars. Moreover, its ability to provide this common framework is not affected by Halliday and Matthiessen's increasing commitment, culminating in Halliday & Matthiessen (1999), to the idea of having a 'two-level' model of 'semantics' (as we saw in Section 4.6 of Chapter 4). In that model, you will recall, there is both the level of 'meaning potential' that Halliday recognised in the early 1970s as the semantics (e.g., Halliday 1971/73a:41-2), and a level of 'semantics' that is higher than this, roughly equivalent to Martin's (1992) 'discourse semantics'. The proposal that this common ground exists follows directly from statements of Halliday's from the late sixties to the present, such as:

In a functional grammar, [...] a language is interpreted as a system of meanings, accompanied by forms through which the meanings can be expressed" (Halliday 1994:xix).

Indeed, it can be argued that all of the concepts that are required in a modern SF grammar follow from accepting the need to recognise the appropriate 'division of labour' between the two levels of semantics and form in such a model.

Blogger Comments:

Reminder:

[1] To be clear, as previously demonstrated, it is the theoretical architecture of the Cardiff Grammar, as represented in Figure 4, that both invalidates the theory and illustrates the extent to which Fawcett does not understand the dimensions of SFL Theory. For example, Figure 4

• misrepresents the axial relation of realisation (paradigmatic system-syntagmatic structure) as the instantiation relation between potential and instance;
• misrepresents a selection expression as an instance of a system network; an instance of a potential system network is an actual system network; a selection expression can be both potential and instance, as demonstrated by the selection expression for the phoneme /b/: [voiced, bilabial, stop] which characterises both the phoneme as potential and the phoneme in a text;
• misrepresents structure as an instance of realisation rules, despite the structure being specified as the realisation of the realisation rules;
• misrepresents a system network and realisation rules as different levels of symbolic abstraction (meaning and form), despite the fact that both the network and the realisation rules include the features of the same level of abstraction (meaning); and
• conflates content (syntax) and expression (phonology/graphology) as the same level of symbolic abstraction (form).

[2] This is misleading, because it is untrue. Even ignoring all the theoretical inconsistencies embodied in Figure 4, it cannot provide a common framework for comparing SFL Theory with the Cardiff Grammar, if only because the two models assume different principles of stratification: meaning/wording/sounding (SFL Theory) vs meaning/form (Cardiff Grammar).

[3] This is misleading, because it is untrue. As pointed out in the examination of Section 4.6, SFL Theory has a 'one-level' model of semantics (Halliday & Matthiessen 1999) and a 'one-level' model of lexicogrammar (Halliday 1985, 1994, Halliday & Matthiessen 2004, 2014). The motivation for this misrepresentation is Fawcett's desire for his model to replace the SFL model of grammatical structure.

[4] To be clear, while it is true that Martin presents his discourse semantics as a stratum above lexicogrammar, the truth, nevertheless, lies elsewhere. As demonstrated here, Martin's (1992) discourse semantics is largely Halliday & Hasan's (1976) lexicogrammatical cohesion, misunderstood and rebranded as his own invention.

[5] This is misleading because it is untrue. The Halliday quote — which does not appear on the cited page — does not suggest that Fawcett's Figure 4 offers a common ground for comparing his model with SFL Theory. It merely acknowledges that in SFL Theory, lexicogrammatical forms are interpreted in terms of the meanings they realise. Cf Halliday (1994: xvii):

[6] To be clear, in SFL Theory, the 'division of labour' is not between meaning and form, but between meaning, wording and sounding, with wording being the interpretation of lexicogrammatical form (e.g. nominal group) in terms of its function in realising meaning (e.g. Senser).

Why Nothing Is Gained By Modelling Dependency Relations In Syntactic Representations

 Fawcett (2010: 249):

The question is therefore: "What is gained by modelling these dependency relations in the syntactic representation too?" The answer is that nothing is gained. In a SF grammar it is simply not the task of a syntactic structure to show that the presence of one element 'depends' on the presence of another. To attempt to do so leads one to ask inappropriate questions, since this is not where dependency is located. The type of 'dependency' that is important in syntax is the relationship of componence, i.e., that between the elements and the unit of which they are the elements. Without this theoretical concept no SF grammar would be complete — and nor could it be implemented in a computer.

Blogger Comments:

[1] This is misleading, because it is untrue. What is gained is the structural realisation of the systemic options in the system of taxis (interdependency).

[2] To be clear, any 'SF Grammar' that does not show the structural realisations of interdependency relations is deficient on the syntagmatic axis, and so compromises its explanatory potential.

[3] To be clear, on the one hand, as previously demonstrated, "inappropriate questions" only arise from theoretical misunderstandings. On the other hand, in SFL Theory, interdependency relations are accounted for both systemically and structurally, since structure realises systemic choices.

[4] This is misleading, because componence is not a type of dependency. Moreover, as previously observed, Fawcett's 'componence' confuses formal constituency with function-form relations.

[5] This is misleading, to the extent that it implies that, unlike Fawcett's Cardiff Grammar, SFL Theory does not already include function-form relations in its architecture.

Why Dependency Relations Are Undesirable In Fawcett's Model Of Syntax

Fawcett (2010: 249):

However, it can be shown that it is not in fact necessary — nor even desirable — to model sister dependency relations in syntax. As I pointed out in Section 10.3.3 of Chapter 10, the supposed 'dependency' of a "modifier' on the "head' on a unit (e.g., in a nominal group) is a 'second order' concept, and ultimately an uninsightful one. In a SF grammar the relationship between the two elements is more appropriately seen as an indirect one, because each element realises its own aspect of the meaning of a referent, and the relationship is therefore at the level of meaning rather than form. In other words, in a SF grammar, any relationship of apparent 'syntactic sister dependency' is already expressed, in a natural manner, in the dependency relations by which one system is dependent on another in the system network of the language's meaning potential. For the systemic grammarian this is the true location of dependence (or 'dependency').

Blogger Comments:

[1] This is misleading. On the one hand, Fawcett's argument is invalid because it relies on his own misunderstandings of theory; see further below. On the other hand, interdependency is valuable for its explanatory potential across all grammatical ranks. This is demonstrated at group rank, for example, where the logical structure of the verbal group realises the system of TENSE in English. Halliday (1994: 198-9):

[2] See the examination of Section 10.3.3 — Confusing Structure With (Transcendent) Ideational Denotation — for a dissection of Fawcett's first-stated misunderstandings on this matter.

[3] Here Fawcett repeats his confusion of logical structure (dependency relations) with ideational denotation ("each element realises its own aspect of the meaning of a referent").

[4] To be clear, on the one hand, 'meaning and form' are levels in Fawcett's model (Figure 4), not in SFL Theory. On the other hand, in SFL Theory, dependency relations obtain at both levels of content, semantics and lexicogrammar — with mismatches constituting instances of grammatical metaphor — and are modelled in systems that specify their structural realisations.

[5] This is misleading. In SFL Theory, dependency relations are modelled as systems, not by dependency relations between systems. For example, 

Strictly speaking, system networks do not construe dependency relations between systems. However, it could be said that entry to a system "depends on" the selection of its entry condition.

Misrepresenting "The Main Use Of" The Rank Scale

Fawcett (2010: 242):

In practical terms, then, the main use of the 'rank scale' concept has been as a model that makes predictions that guide the text analyst as to how the units of a text-sentence relate to each other — though these have sometimes caused problems for the analyst. However, statements of 'filling probabilities', as in Appendix B, meet the same need in a more effective manner.

All of the probabilities discussed so far are instantial probabilities, i.e., probabilities that certain patterns will occur in instances, i.e. in text-sentences. They are moreover probabilities at the level of form. In contrast with these are the probabilities on features in system networks, which we might refer to as potential probabilities, these being at the level of meaning. See Section 2 of Appendix C for a discussion of the relationship between the two.

 

Blogger Comments:

[1] This is misleading, because it is untrue. In SFL Theory, the rank scale is the means by which formal constituency is modelled. Most importantly, in a systemic functional grammar, each rank is the entry condition for the systems that specify function structures at that rank.

[2] This is misleading, because it is untrue. As previously demonstrated, all the "problems" raised by Fawcett derive from his confusing the rank scale of forms with function-form relations.

[3] This is misleading, because it is untrue. As previously demonstrated, Fawcett's 'filling probabilities' are concerned with the relation between function and form, not with formal constituency. As such, they do not "meet the same need in a more effective manner".

[4] This is misleading, because it is untrue. As previously demonstrated, Fawcett's 'filling probabilities' are concerned with the relation between function (element) and form (unit) — not with form alone.

[5] To be clear, as Section 2 of Appendix C (p315) explains, in Fawcett's model, potential probabilities are paradigmatic probabilities, whereas instantial probabilities are syntagmatic probabilities. That is, this reflects Fawcett's confusion of axis (paradigmatic/syntagmatic) with instantiation (potential/instance), as previously demonstrated in the examination of Figure 4 (p36):

Fawcett's Argument For 'Secondary Structure'

Fawcett (2010: 219-20):

In developing a SF grammar, both for very large computer implementations and for text analysis, priority must be given to the "most delicate" possible of structural descriptions, because a full account of the meaning potential of a unit (such as a clause) requires statements about each element in its own right. It is the individual elements of a unit that carry the different meanings that are the focus of interest for a functional grammarian. 

For example, the fact that the main mood meanings of a clause are realised by the configuration of the Subject and Operator is best shown by stating the semantic feature that generates this configuration in an analysis of the meaning potential of the clause, as in Figure 10 in Chapter 7 (rather than introducing an additional layer of structure as part of its syntax, as in the case of "Mood + Residue" in IFG).

Blogger Comments:

[1] To be clear, Systemic Functional Theory is a model of language as a phenomenon. Text analysis is an application of the theory, and computer implementations require the modification of the theory to accommodate the limitations of computers.

[2] To be clear, on the one hand, this priority flatly contradicts Fawcett's previous argument against what he regards as the "most delicate" structures; see the previous post. On the other hand, this priority is inconsistent with SFL Theory, because it argues for the view 'from below' (structure and form) rather than the view 'from above' (system and function).

[3] To be clear, the focus of interest for a Systemic Functional grammarian is the system that specifies structural realisations.

[4] To be clear, as previously noted, in Systemic Functional Linguistic Theory, structural realisations are specified systemically as realisation statements. Halliday & Matthiessen (2014: 162):

[5] To be clear, as previously noted, Fawcett's Figure 10 confuses systemic features of the clause with elements of structure:

Moreover, the only semantic feature that corresponds to the configuration of Subject and Operator is the unexplained 'information giver'. Significantly, Fawcett does not provide the semantic networks from which these features are derived.

Confusing Structure With (Transcendent) Ideational Denotation

Fawcett (2010: 216-7):

What is the relationship between one element and its 'sister' elements in a unit? This has been a major focus of interest for some grammarians, leading to arguments about whether such relationships are those of 'daughter dependency' or 'sister dependency' (e.g., Hudson 1976). (For a brief comparison of the two, see Section 11.2 of Chapter 11.) Here I offer a new answer to the question asked above. It is one that follows directly from the adoption of the framework outlined in Chapter 2 and exemplified in Appendix A.

Let us take as an example the relationship between a modifier and the head in the English nominal group. In the framework of a systemic functional grammar the relationship is not, I suggest, the direct one that form-centred grammarians consider it to be. In formal and traditional grammars, it is simply assumed that what the modifier modifies is the head. Here, however, the general function of the modifiers in a nominal group is regarded as being to describe the referent. (See Fawcett (in press) for the sub-types of 'description' that the various sub-types of modifier express, e.g., 'colour modifiers', 'affective modifiers', 'general epithet modifiers' etc.) Similarly, the function of the head of the nominal group is (assuming that it is a noun) to state the 'cultural classification' of the referent. The referent is thus the object to which the nominal group refers, and it is the function of the noun at the head of the nominal group to express what Lyons (1977:206-7) terms the "denotation" (of some class of 'thing'). In other words, the head realises one type of meaning that relates to the referent, while the modifier realises another. So both the modifier and the head relate, via the meanings they express, to the referent — but they are related only indirectly to each other. Thus a modifier does not in fact 'modify' (or 'describe') the head; it modifies (or describes) the referent which the head denotes.

This general principle applies to all 'sister' relationships between elements, and it applies to all units. From this viewpoint, the question of whether an element is dependent on a 'sister' element such as the 'head' or on a 'mother' unit is beside the point; the 'dependency' is not in fact 'syntactic' at all, and what we observe in syntax is the realisation of dependence in the system networks.

Blogger Comments:

[1] To be clear, this confuses relations between formal units with relations between functional elements.

[2] To be clear, this confuses structure with ideational denotation ("describing a referent").  In SFL Theory, a structure is the relation between the elements of a unit. Halliday & Matthiessen (2014: 451):

Note that, although it is the functions that are labelled, the structure actually consists of the relationships among them.

In SFL Theory, which takes an 'immanent' view of meaning, the ideational denotation of a wording (lexicogrammar) is its ideational meaning (semantics). However, in other theories, which take a 'transcendent' view of meaning, the ideational denotation is to a domain outside language.

[3] To be clear, as Fawcett explicitly makes clear by the wording "meaning that relates to the referent", his model takes a 'transcendent' view of meaning, a view that is seriously inconsistent with the epistemological assumptions from which SFL Theory is constructed.

[4] To be clear, this is inconsistent with SFL Theory, in which Modifier–Head is a univariate structural relation. Halliday & Matthiessen (2014: 390):

We refer to this kind of structure as a univariate structure, one which is generated as an iteration of the same functional relationship (cf. Halliday, 1965, 1979): α is modified by β, which is modified by γ, which is ... .

[5] To be clear, if 'dependency' is a syntagmatic relation between formal units, then it is 'syntactic'.

[6] To be clear, this is inadvertently consistent with the architecture of SFL Theory — or, at least with the general principle that syntagmatic structures realise paradigmatic selections.

Misrepresenting Halliday On Nominal Group Structure In A Footnote

 Fawcett (2010: 213-4, 213n-4n):

The main characteristic of an element is that it is defined functionally, rather than positionally. This should surely be a founding principle of a functional approach to syntax — and yet the tradition of using positional labels still lingers on in many functional grammars.¹⁷

¹⁷ For example, terms such as 'pre-deictic' and 'pre-numerative' (as found in Halliday 1994:195-6) are simply positional labels. Rather similarly, the terms 'premodifier' and 'postmodifier' (Halliday 1994:194-5), signal a positional meaning more strongly than they signal a functional meaning. This is because, at Halliday's primary level of delicacy in the analysis of a nominal group, the term 'modifier' means little more than 'anything other than the head'. So these terms give virtually no information about the element's function. 

However, I must admit to retaining one traditional 'positional' label in the Cardiff Grammar, i.e., 'preposition'. As said in Section 10.2.6, I would have preferred an explicitly functional label, but terms such as 'relator' are not specific enough, and we retain the traditional term "preposition" both because there is a lack of a clear alternative and because it is so strongly established. We define it as functioning to express a 'minor relationship with a thing'. Note that in the description of English one item that actually occurs 'postpositionally' is included as a 'preposition' — i.e ago, as in five years ago.

Blogger Comments:

[1] This is misleading (and hypocritical) because, contrary to the implication, Fawcett's own approach to functional syntax includes elements that are defined positionally. As can be seen from the previous post, Appendix B (p304), lists:

• 'Starter' and 'Ender' as elements of the clause,
• 'finisher' as an element of the quality group,
• 'quantity finisher' as an element of the quantity group,
• 'starter' and 'ender' as elements found in all groups, and
• 'Opening Quotation mark' and 'Closing Quotation mark' as elements of a text.

[2] This is misleading, because it is untrue. The terms 'pre-Deictic' and 'pre-Numerative' are not "simply positional labels", since they identify the functions 'Deictic' and 'Numerative' respectively. However, the function 'pre-Deictic' — unlike 'post-Deictic' — is not an element in Halliday's model, and the 'pre-Numerative' was later termed 'extended Numerative' (Halliday & Matthiessen 2004: 333).

[3] This is misleading, because it is untrue. The term 'pre-Modifier' and 'post-Modifier' do not signal positional meaning more strongly than functional meaning, since they identify the function 'Modifier' and, through prefixes, additionally provide the location of the modification relative to the Head.

[4] This misrepresents Halliday. To be clear, in SFL Theory, delicacy is a dimension of paradigmatic system, not syntagmatic structure.

[5] To be clear, here Fawcett misleads through his ignorance of the notion of univariate structure: 'an iteration of the same functional relationship' (Halliday & Matthiessen 2014: 390, 451).

[6] As can be seen from the four preceding points, this is the direct opposite of what is actually true.

[7] To be clear, 'preposition', as a class of word, is a unit of form, so renaming it with a function label would be theoretically inconsistent. In SFL Theory, a preposition functions as a minor Process/Predicator of a prepositional phrase.

[8] To be clear, in SFL Theory, the experiential function of a preposition, minor Process, relates to the function of a nominal group, Range, in a prepositional phrase, whereas Thing is an experiential function of a word in a nominal group.

[9] To be clear, the word ago is an adverb, not a preposition, and five years ago is a nominal group, not a prepositional phrase:

"Using The Cardiff Grammar As The Baseline For Constructing A Modern Theory Of SF Syntax"

Fawcett (2010: 185-6, 186n):

In Chapters 10 and 11 we turn to the concepts that are required for the specification of 'instances of syntax'. As we have seen, these concepts are drawn on in a computer model of parsing such as that described in Weerasinghe & Fawcett (1993), Weerasinghe (1994) and Souter (1996). However, these concepts are also referred to in the realisation rules, and are in that sense presupposed by them.

As will by now clear, we shall be using the Cardiff Grammar rather than the Sydney Grammar as the baseline for constructing a modern theory of SF syntax.⁵ 

⁵ Apart from the reasons that follow from our findings in Chapter 7, there are two more reasons for this. Firstly, the Cardiff Grammar has taken the revolutionary proposals for changes to the theory made by Halliday in the 1960s (as summarised in Chapter 4) significantly further than the Sydney Grammar has. It has full implementations of (1) explicitly semantic system networks, (2) the concept of lexis as "most delicate grammar", (3) the integration into the system networks of the meanings realised in intonation and (4) the integration of meanings realised in punctuation. Secondly, the Cardiff Grammar provides a much fuller specification than the Sydney Grammar does of the syntactic concepts that are required, both for language in general and for the description of English in particular — especially in its recognition of classes of group and cluster that are not provided for in the Sydney Grammar.

Blogger Comments:

[1] As previously explained, Fawcett's 'instances of syntax' are actually structures, not instances. Fawcett's model (Figure 4) confuses the realisation relation between the paradigmatic and syntagmatic axes with the instantiation relation between potential and instance.

[2] To be clear, Fawcett's Cardiff Grammar is his model of syntax. Halliday's "Sydney Grammar" — SFL Theory — models syntax (and morphology) as a rank scale, but Halliday (1985: xiv) explains why the term 'syntax' is inappropriate for a functional grammar:

[3] To be clear, Halliday's work in the 1960s was concerned with Scale and Category Grammar, not with Systemic Functional Grammar. That is, Fawcett's claim is actually that his Cardiff Grammar takes the proposals of Halliday's superseded theory further than Halliday's current theory does.

[4] To be clear, Fawcett does not present any of his 'explicitly system semantic networks' in this entire publication. In SFL terms, such networks are actually the grammatical networks, and can be found in Halliday & Matthiessen (2004, 2014). For genuinely semantic networks, see Halliday & Matthiessen (1999).

[5] To be clear, Fawcett has nowhere demonstrated, in this publication, how he models 'lexis as most delicate grammar', and it does not figure in his theoretical architecture (Figure 4). Moreover, since Fawcett locates grammatical systems at his level of meaning, his model is committed to lexis as most delicate semantics, not grammar.

[6] To be clear, Fawcett does not present any of the system networks of the meanings realised in intonation in this entire publication. For an SFL approach to intonation, see Halliday & Greaves (2008).

[7] To be clear, Fawcett does not present any of the 'integration of meanings' realised in punctuation in this entire publication.

[8] This misleading. The 'Sydney Grammar' (SFL Theory) does not provide any specifications 'of the syntactic concepts that are required, both for language in general and for the description of English in particular', largely because SFL Theory is not a theory of syntax; see [2] above.

Fawcett's Summary Of His Comparison Of His Operations With SFL's Realisation Statements

Fawcett (2010: 185):

Let me summarise. Leaving aside the "Split" and "Expand" operations of the Sydney Grammar, which are either unworkable or unnecessary, the Sydney Grammar has an equivalent for every realisation operation in the Cardiff Grammar except the first (though these are not always in a one-to-one relationship, as we have seen). These realisation operations are important concepts in the theory, as their treatment in both Halliday (1993) and Fawcett, Tucker & Lin (1993) clearly demonstrates. 

However, these 'operation' concepts are a part of the grammar itself, so that they are relevant only indirectly to the outputs from the grammar — i.e., to a description of the structure of the text-sentences that are the instances of the potential specified in the grammar. Essentially, their function is to generate the relationships between the categories that we shall establish in Chapter 10. It is in Chapter 11 that we shall meet the relationships again. And it is perhaps significant that the first concept to be discussed there — that of 'rank' — has no equivalent among the realisation operations and will be rejected, while all of those to be considered in Sections 11.2 to 11.8 do have such a relationship.

Blogger Comments:

[1] This is misleading, because it is untrue, as previously demonstrated.

[2] This is misleading, because it falsely presents Halliday's original model ("the Sydney Grammar") as if it were the derivative model (Fawcett's Cardiff Grammar).

[3] To be clear, this is both a non-sequitur and untrue. It is a non-sequitur because their being part of the grammar does not logically entail that realisation 'operations' are only indirectly relevant to the structures they specify; for example, the PROCESS TYPE system is also part of the grammar, and yet it is "directly relevant" to the experiential structure of the clause. And it is untrue because 'operations' (realisation statements) "directly" specify how system selections (paradigmatic axis) are realised structurally (syntagmatic axis).

[4] To be clear, here again Fawcett misunderstands the realisation relation between system and structure as the instantiation relation between system and instance — and he does so despite the fact that his term 'realisation operations' explicitly identifies the relation as realisation, not instantiation.

[5] To be clear, the claim that Fawcett's realisation operations (listed below) generate relationships between categories will be tested in the examination of Chapters 10 and 11.

1. Insert a unit (to fill an element). 

2. Locate an element at a place in a unit.

3. Conflate an element or Participant Role with an existing element. 

4. Expound an element by an item.

4a. Fetch a name to expound an element. 

5. Prefer certain features on re-entry to the system network , including preselection. 

6. For an element, re-enter the system network.

[6] To be clear, in SFL Theory, rank is not an "equivalent" of realisation statements. The rank scale provides the entry conditions to the systems of each rank, and realisation statements accompany features in those networks. Moreover, the rank scale is the means by which SFL Theory models form — i.e. syntax and morphology — and so is the theoretical dimension that makes Fawcett's model of syntax redundant. So it is hardly surprising that Fawcett rejects the 'concept' of rank. 

Misrepresenting Halliday (1993) On The Realisation Statement 'Expand'

Fawcett (2010: 184):

The last realisation operation that requires a comment is Halliday's Operation (e) of "Split". This has no equivalent in the Cardiff Grammar. A "Split" operation is only needed in a grammar which represents both 'primary' and 'secondary' structures (as introduced in "Categories" and still used regularly for some aspects of structure in IFG). Halliday introduces it to enable the grammar first to generate what he would term a 'primary structure' (such as "Mood + Residue" in Figure 7 in Chapter 7), and then to 'split' the 'Mood' element into the two elements of its supposedly 'secondary' structure, i.e., into "Subject + Finite". Halliday does not explain why he thinks it desirable to generate the "Mood" element first and then to split it into two, but we can presume that the intention is to give expression to the idea that the "Mood + Residue" structure is 'primary' and the "Subject + Finite" structure is 'secondary'. Nor does Halliday explain what would actually happen in practice when a "Split" operation is carried out.

The reason why the Cardiff Grammar has no such realisation operation is, of course, that we recognise only one degree of 'delicacy' in the structures of the clause, so that there is no need for an operation whose function is to add another line of analysis to the representation. Instead, we treat the Subject and the Operator (Halliday's "Finite") as direct elements of the clause.

 

Blogger Comments:

[1] Trivially, Halliday (1995 [1993]: 272) does not propose an Operation of "Split", but a realisation statement of 'expand':

(e) 'Expand' an element into a further configuration (e.g., expand mood into subject+ finite);

[2] Non-trivially, the grammar does not first "generate" a Mood element and then expand it into Subject and Finite elements. This is to mistake a system network of relations for an algorithmic procedure. The identity relation between system and structure is intensive (elaboration), not circumstantial (enhancement: temporal).

[3] This is misleading. It is only Fawcett who regards these structures as 'primary' and 'secondary'.

[4] This is misleading. On the one hand, Halliday does explain why a Mood element is necessary. For example, the presence of the Mood element realises the feature 'indicative' (Halliday 1994: 74). On the other hand, the realisation statement 'expand' is not used in the system network of MOOD; Halliday & Matthiessen (2014: 162):

[5] To be clear, Halliday's explains that this realisation statement expands an element into a further configuration. 

[6] To be clear, in SFL Theory, delicacy (hyponymy) is a sub-type of elaboration (Halliday & Matthiessen 1999: 145) and a dimension of systems, not structures. In contrast, the relation of Subject and Finite to the Mood element is composition (meronymy), a subtype of extension (Halliday & Matthiessen 1999: 146).

[7] This is misleading, since Subject and Finite are "direct" elements of the clause in SFL Theory also. Or more precisely, the notion of 'direct' is irrelevant here, deriving, as it does, from Fawcett's misunderstanding of a system network as a sequenced algorithm; see [2] above.

Misunderstanding Axial Realisation As Instantiation

Fawcett (2010: 174):

Thus most of this chapter will be wholly oriented to the theoretical-generative strand of work in SFL. It might at first appear that Chapters 10 and 11 are oriented to the text-descriptive strand. However, they are only 'text-descriptive' in the sense that the emphasis is on the text-sentences that are the output from the operation of the grammar, and it is these that constitute the texts that are analysed in 'text-descriptive' work. These two chapters are also in fact primarily theoretical, in that the concepts covered there are the theoretical concepts that are required in the description of the structure of English that is needed by a computer for parsing a text — and, indeed, by a human text analyst. The concepts that characterise instances are therefore as much a part of the overall theory as the concepts required in the grammar itself to generate those outputs.

Blogger Comments:

[1] To be clear, Fawcett's theory of language is actually concerned with what computers and linguists require for parsing a text.

[2] As previously explained, Fawcett equates 'instance' with 'output', and consequently misunderstands structures (syntagmatic axis) as instances of realisation rules (paradigmatic axis).

Fawcett's Theories Of Syntax Potential & Instances Of Syntax

Fawcett (2010: 174):

In the rest of the present chapter I shall specify the theory of syntax potential that is required in a modern, large-scale systemic functional grammar — i.e., the concepts that are required in the grammar itself. Then Chapters 10 and 11 will provide an account of the theory of instances of syntax — i.e., the part of the theory that underlies the description of text-sentences. Finally, Chapter 12 will summarise the current theory in relation to its antecedents, including "Categories", offering an evaluation of its importance today.

Blogger Comments:

[1] As previously explained, for Halliday (1985 & 1994: xiv), the term 'syntax' suggests an approach to theorising that is inconsistent with the approach to SFL Theory, 'such that a language is interpreted as a system of forms, to which meanings are then attached'. In SFL Theory, 'syntax' is modelled as a rank scale. Fawcett's Cardiff Grammar does not include a rank scale.

[2] As previously explained, Fawcett's 'instances of syntax' are not instances of potential, but syntagmatic structures that realise realisation rules. That is, Fawcett's model (Figure 4) confuses the realisation relation between the paradigmatic and syntagmatic axes with the instantiation relation between potential and instance.

[3] To be clear, the term 'text sentences', like the architecture of Fawcett's model (Figure 4), confuses instance (text) with structure (sentence). In SFL Theory, 'sentence' is theorised as a rank unit of graphology.

[4] Fawcett's evaluation of the importance of his own model will be evaluated in the examination of his final chapter.

Misunderstanding Instantiation

Fawcett (2010: 172):

As we saw in Chapter 3 — and most clearly in Figure 4 in Section 3.2 of that chapter — any adequate theory of language must provide, for each level of language, for both the potential and the instances of that potential. In some theories of language, of course, it is assumed that it is precisely the task of the theory of the potential (the grammar) to specify the instances that can be generated from the potential (the sentences of the language). But I have argued — both in Section 6.5 of Chapter 6 and more fully in Fawcett (1994a) — that, if we want a theory that can be used for modelling both the understanding and the generation of language texts, we need to recognise that we require a different set of concepts when faced with the task of analysing an incoming string of words (i.e., parsing) from those that are needed when we are trying to model the grammar in use for generation. In the process of generation there is no equivalent of the problematical task of parsing, as is demonstrated in Fawcett (1994a).

Blogger Comments:

[1] Reminder:

To be clear, Fawcett's model misconstrues instantiation as a relation between system and selection expression, at the level of meaning, and a relation between realisation rules and structure, at the level of form. As explained in many previous posts, a selection expression can be viewed as potential or instance, as shown by the phonological selection expression [voiced, bilabial, stop], which is synthetically realised by the phoneme /b/, whether as potential or as instance. As also explained in many previous posts, the relation between realisation rules and structure is the relation between the paradigmatic and syntagmatic axes: realisation, not instantiation.

[2] To be clear, both the grammar and 'the sentences of language' can be viewed as potential or instance. For example, 'the sentences of language' can be viewed as potential (grammatical systems) or as instances (in texts). It appears that Fawcett misunderstands 'instance' to simply mean 'output'.

[3] To be clear, in SFL Theory, the system is both the means by which speakers make meaning and the means by which addressees understand it. The different requirements of text generation and text parsing derive from trying to accommodate the fact that computers are not humans. A theory of language models the language of humans; a model of text generation and parsing by computers is not a model of human language.

The Concept Of Rank In Fawcett's Models

Fawcett (2010: 165-6):

The concept of rank from "Categories" is retained in "Some proposals" — at least in relation to the clause and the group, which form the core of the model of syntax in all SFL descriptions (with far less work on the proposed 'morpheme-word' relationship). However, this small syntactic 'rank scale' is interpreted in a very different way from "Categories", because it is seen as the realisation of an equivalent semantic relationship between a 'situation' and the 'things' and 'qualities' that are its 'elements' at that higher level. And, as we shall see in Part 2, later work in this version of the theory was to reduce the role of the concept of a 'rank scale' to the point where it no longer has any status in the theory at all. … In the theory of syntax to be presented in Part 2 the concept of the 'rank scale' is replaced by the concept of probabilities in the relations between elements and units.

Blogger Comments:

[1] As previously noted, Fawcett's "Some proposals" (1974) was oriented to Halliday's first theory, Scale and Category Grammar, after it had been superseded by Halliday's second theory, Systemic Functional Grammar.

[2] This is potentially misleading, because SFL is not a theory of syntax. Halliday (1985: xiv):

[3] To be clear, this aspect of Fawcett's model is included in his Figure 12 (p210):

In SFL Theory, the semantic unit congruently realised by a clause depends on metafunction:

• figure (ideational)
• proposition/proposal (interpersonal)
• message (textual)

and ideationally, the semantic unit congruently realised by a group is an element (Halliday & Matthiessen 1999).

[4] To be clear, SFL Theory it is the rank scale that models the linguistic phenomena deemed 'syntax' in other theories.

[5] To be clear, in SFL Theory, rank is the ordering principle of structure (syntagmatic order), whereas probability is the quantification of system (paradigmatic order). This will be elaborated further in the examination of Fawcett's model in Part 2.

The Implications Of The Argument For A Theory Of Syntax For SFL [2]

Fawcett (2010: 157-8):

However, since the current situation is that there is not yet a published work in any version of SFL that demonstrates how to use the features in system networks for the systemic analysis of texts, IFG-style analyses may still have a role to play for a while. The system networks for the Sydney Grammar have of course now been published (in Matthiessen 1995), but these networks are not in fact designed for use in text analysis, and they are often hard for the reader to interpret. Moreover there are no published guidelines as to how to use system networks for the task of making analyses of real-life texts, and no published examples of such analyses, other than those in Fawcett (1999) and in Figure 10 of Section 7.9. The publication of Fawcett (in press) and Fawcett (forthcoming a), as described at the end of Section 7.8, make available the Cardiff Grammar versions of (a) the functional structure of text-sentences and (b) system networks that are equivalent in their coverage (but more explicitly semantic) to the Sydney Grammar networks found in Matthiessen (1995). Moreover Fawcett (forthcoming a) contains guidelines that show the analyst how to go about making such analyses, and it contains many such examples.

Blogger Comments:

[1] This is still misleading, because it is still untrue. As previously noted, Halliday & Matthiessen (2004) provides, for each metafunction, clause analyses which display the features selected.

[2] To be clear, the structures exemplified in IFG represent an integral dimension of SFL Theory.

[3] To be clear, in SFL Theory, system networks are the formal means by which language is modelled, and which specify the structures that Fawcett equates with text analysis.

[4] To be clear, here Fawcett presents his own difficulty in interpreting system networks as a problem with the networks themselves.

[5] This is misleading, because it is untrue. As previously noted, Figure 10 does not include the networks from which its features are derived, and misrepresents paradigmatic features of the entire clause as individual elements of syntagmatic structures ("strands of meaning").

[6] As previously noted, these promised works — Fawcett (in press) and Fawcett (forthcoming a) — remain unpublished, 20 years after the first edition of this publication.

The Implications Of The Argument For A Theory Of Syntax For SFL [1]

Fawcett (2010: 157):

The main purpose of this chapter has been to clarify the status of the representations of clauses given in IFG and the derived works. We have also seen, in Section 7.9, that there is an alternative and more truly systemic means of representing the multifunctional nature of language, i.e., a representa[t]ion of the systemic features. In systemic functional grammar such a representation is inherently more revealing than any structural representation. As Halliday himself says, "the system takes priority; the most abstract representation [...] is in paradigmatic terms. Syntagmatic organisation is interpreted as the 'realisation' of paradigmatic features" (Halliday 1993:4505).

The consequence of having both a representation at the level of meaning and a representation of the single integrated structure — something that is required, as we have seen, in the Sydney model as well as the Cardiff model — is that the many lines of structure in an IFG analysis are, from both the theoretical-generative and the text-descriptive viewpoints, redundant. They are redundant because the analysis in terms of the features chosen in the system networks already displays clearly the different strands of meaning, as in the example of such an analysis in Figure 10.

Blogger Comments:

[1] We have also seen, in the examination of Section 7.9, that Fawcett's "alternative and more truly systemic means of representing the multifunctional nature of language" — Figure 10 — confuses systemic features with structural elements, thereby attributing features of the whole clause to elements of clause structure. These theoretical inconsistencies, alone, invalidate Fawcett's model.

[2] This is true, and Fawcett has repeatedly demonstrated that he does not understand what Halliday means by it. Despite the system taking priority in SFL Theory, Fawcett has focused instead on structure, not just in this chapter, not just in this entire book, but in his reworking of Halliday's theory — even to the extent of misrepresenting systemic features as structural elements (Figure 10).

[3] This is misleading, because it is untrue, as has been demonstrated previously, over and over. In the "Sydney model" (Halliday's version of his own theory), there is no need for an "integrated structure", since it is the realisation of the metafunctional clause structures in a syntagm of clause constituents, groups and phrases, that provides the skeleton onto which the functions are mapped. As previously explained, Fawcett's misunderstanding arises from his confusing formal constituency (the rank scale) with function structure. As will be seen, Fawcett argues against the theoretical value of a rank scale.

[4] This is misleading, because it is untrue. Fawcett's alternative analysis (Figure 10) does not make metafunctional clause structures redundant, because his semantic analysis, which displays paradigmatic features, does not present an analysis of syntagmatic structure:

An Addendum To The Argument For A Theory Of Syntax For SFL

Fawcett (2010: 156-7):

As an addendum to the line of argument set out here, we should note that there are important theoretical differences between the Cardiff Grammar and the Sydney Grammar (within the general framework summarised in Figure 6 in Section 4.9 of Chapter 4). The most important difference is in way in which the various 'strands of meaning' are represented. While the Cardiff Grammar represents the meanings systemically — i.e., in terms of the features that have been chosen in the system networks that constitute the 'meaning potential' of the language — Halliday represents the meaning by 'functions'. His use of this concept for representing meanings can be traced back to the stage in the development of the theory in which it was believed that the realisation of the features in the system networks necessarily took place in two stages: in the first the grammar would use the 'selection expression' of features that were chosen on a traversal of the system network to create an unordered string of 'functions' (and of conflated functions), and in the second these 'functions' would be ordered in sequence.²⁶

²⁶ See Berry (1977:29-31) for an introductory account of this version of the model. The fullest implementation by far of the concept of realisation in two stages is that described in Hudson (1971), which had 76 such 'functions' (these being explicitly at the level of form, however). In a later work Hudson reduced them to just three (Hudson 1976), and this constitutes a little subsidiary evidence for the position taken here, i.e., that it is unnecessary for a systemic grammar first to generate 'functions' and then to order them.

Blogger Comments:

[1] This is true, as the careful examination on this blog demonstrates. For example, the "Sydney Grammar" is the theory that Halliday painstakingly created, whereas the "Cardiff Grammar" is merely Fawcett's alteration of the theory that Halliday painstakingly created. More importantly, as demonstrated over and over again, Fawcett repeatedly misrepresents Halliday's model, while his own model, Figure 4, is invalidated by its own internal inconsistencies.

[2] This is very misleading indeed. The discussion in Section 4.9 is not concerned with Halliday's "Sydney Grammar", but with an earlier development (c1970) that was soon superseded, as previously explained. Figure 6, in which Fawcett has added the label 'COMBINED' to Halliday's diagram, otherwise displays the earlier model:

[3] To be clear, as this blog demonstrates, this is not the most important difference between the Cardiff Grammar and Halliday's version of his own theory, but it nicely illustrates the type of misunderstandings that invalidate Fawcett's version of Halliday's theory. As previously explained, Fawcett represents 'strands of meaning' — that is, structures — as systemic features.  As well as confusing the syntagmatic with the paradigmatic axis, this results in features of the clause being  represented as features of clause constituents.

Halliday's 'functions', on the other hand, are elements of function structure, with each structure of the clause being the relation between elements.

[4] To be clear, this "concept for representing meanings" is the syntagmatic dimension of the theory. As previously explained, Fawcett's model (Figure 4) misrepresents the relation between system and structure as a (stratified) relation between potential and instance:

[5] To be clear, other people's interpretations of earlier versions of Halliday's theory do not represent the "Sydney Grammar". That is, the inclusion of this in Fawcett's argument constitutes the logical fallacy known as a red herring.

[6] This is misleading, because it misrepresents Halliday's theory. To be clear, the grammar does not first generate "functions" — i.e. elements of function structure — and then order them. In SFL Theory, the grammar is modelled as a system of choices, with different choices specifying the different structures that realise them.