On Halliday's 1972-6 Model

Fawcett (2010: 75):

Finally, I showed that there was one temporary phase in the development of Halliday's theory in which he showed the Scale and Category elements of the clause as serving the function of integrating the various strands of meanings that are always shown in any IFG-style analysis — very much as the Cardiff Grammar does, in general terms, and as this book argues that all systemic functional grammars should. Yet it is a model which Halliday quickly abandoned for reasons that are far from clear, inserting the 'integrating' elements in the 'interpersonal' strand of meaning instead, as we shall see in Chapter 7. And, as we shall also see in Chapter 7, this leaves the considerable problem of how these semi-semantic 'multiple structures' are to be integrated into a single structure.

Blogger Comments:

[1] To be clear, as previously demonstrated, this hangover from Halliday's pre-Systemic theory, Scale and Category Grammar, had already been re-theorised by the time of the second edition (Halliday 1978) of the paper that Fawcett used (Halliday 1977 — written 1972-6) to exemplify the "temporary phase in the development of Halliday's theory".

[2] To be clear, this "abandonment" was a direct consequence, in Halliday (1978), of identifying units of the rank scale — clause, group, word, morpheme — as the locus onto which the structures of metafunctional systems are mapped.

[3] As will be demonstrated in the examination of Chapter 7, the "problem" here, as elsewhere, is with Fawcett's difficulty in understanding Halliday's theory.

Fawcett's Argument Against Realisation Rules In System Networks

Fawcett (2010: 75): 

Next, I showed why diagrams consisting of system networks in which the realisation rules are shown as 'footnotes' on the features are not only inadequate for a large-scale grammar but that they also give a misleading picture of language. In other words, the existence of such diagrams should not be taken as evidence that the full set of components and outputs shown in Figure 4 is unnecessary.

Blogger Comments:

[1] This is misleading, because, as previously demonstrated, Fawcett did not show why the location of realisation rules (statements) in system networks is inadequate for a large-scale grammar, not least because the network he used in his argument was a small-scale network of his own devising which, for example, confuses lexical items with grammatical features.

[2] This is misleading, because, as previously demonstrated, Fawcett did not show that the location of realisation rules in system networks gives a misleading picture of language.  Moreover, Fawcett's own alternative model (Figure 4) is invalidated by the fact that it locates categories of the same level of symbolic abstraction, grammatical features, at different levels of symbolic abstraction, meaning and form, depending on whether they are located in networks or in rules.

[3] To be clear, this provides Fawcett's motivation for arguing against the location of realisation rules (statements) in system networks: the fear that it invalidates his own model, whereas, in truth, what invalidates Fawcett's model is its own internal inconsistency — including its misconstrual of the relation between realisation rules and the structures that realise them as instantiation.

On The Topological Equivalence Of SFL Architecture And Fawcett's Flowchart

Fawcett (2010: 74-5):

I then gave some of the reasons why I think he is wrong to dismiss his earlier insight that the system networks constitute the level of semantics, and I argued that the topological relationships between the different parts of the model of language summarised in Figure 4 remain intact — even when the diagram is redrawn in order to make them appear to occupy a single stratum of language, as in Figure 5. The key point is that, since the relationships remain the same, it is still possible to make a direct comparison between the output structures of the Sydney and the Cardiff Grammars.

Blogger Comments:

[1] To be clear, a principal reason why the system networks are located on the stratum of lexicogrammar, rather than semantics, is that distinguishing semantic and grammatical systems improves the explanatory power of the theory by providing the means of systematically accounting for grammatical metaphor.  But this is not the only reason. Halliday & Matthiessen (1999: 428-9):

Within systemic-functional linguistics, Fawcett (e.g. 1980) has pioneered a "cognitive model of an interactive mind". There are many fundamental similarities with the approach we are taking here, e.g. in construing an experiential system of process configuration within the content plane. However, there are two related differences of particular interest in the context of our present discussion:

(i) in Fawcett's model, there is only one system-structure cycle within the content plane: systems are interpreted as the semantics, linked through a "realisation component" to [content] form, which includes items and syntax, the latter being modelled structurally but not systemically; 

(ii) in Fawcett's model, the semantics is separate from the "knowledge of the universe", with the latter as a "component" outside the linguistic system including "long term memory" and "short term sort of knowledge". 

With respect to (i), in our model there are two system-structure cycles, one in the semantics and one in the lexicogrammar. Terms in semantic systems are realised in semantic structures; and semantic systems and structures are in turn realised in lexicogrammatical ones. As we saw in Chapter 6 in particular, grammatical metaphor is a central reason in our account for treating axis and stratification as independent dimensions, so that we have both semantic systems and structures and lexicogrammatical systems and structures. Since we allow for a stratification of content systems into semantics and lexicogrammar, we are in a stronger position to construe knowledge in terms of meaning. That is, the semantics can become more powerful and extensive if the lexicogrammar includes systems. It follows then with respect to (ii) that for us "knowledge of the universe" is construed as meaning rather than as knowledge. This meaning is in the first instance created in language; but we have noted that meaning is created in other semiotic systems as well, both other social-semiotic systems and other semiotic systems such as perception. Our account gives language more of a central integrative role in the overall system. It is the one semiotic system which is able to construe meanings from semiotic systems in general.

[2] To be clear, as previously demonstrated, Fawcett's model summarised in Figure 4 is invalidated by its own internal inconsistencies, including its misconstrual of axial realisation of form as instantiation.

[3] To be clear, as previously demonstrated, Figure 5 is not topologically equivalent to Figure 4, since the one cannot be formed from the other without tearing. Moreover, in terms of Fawcett's argument, it is a 'straw man' of Fawcett's own invention.  See the earlier post Attacking A Straw Man.

Realisation Between Strata And Realisation Between Axes

Fawcett (2010: 74):

Then, in the second half of the chapter, we surveyed the effect of Halliday's adoption of the concept that there is a higher set of system networks than those of TRANSITIVITY, MOOD and THEME. First we noted that this has led him to express increasingly strongly the view that the relationship between these system networks and the structures that are generated from them is only one of realisation by "extension" or by "analogy".

Blogger Comments:

[1] To be clear, the 'higher set of system networks' than the grammatical systems of TRANSITIVITY, MOOD and THEME are the systems of the semantic stratum, such as the interpersonal system of SPEECH FUNCTION and the ideational systems set out in Halliday & Matthiessen (1999).

[2] This is misleading. To be clear, the reason why the relation of realisation also applies to the relation between system and structure is because it is the very same relation: the fundamental semiotic relation of symbolic abstraction between a higher level Value (system) an a lower level Token (structure).

See also what Halliday actually said, in this regard, in the earlier clarifying critique: (Accusing Halliday Of) Confusing Axial Realisation With Instantiation.

Misunderstanding Realisation And Instantiation

Fawcett (2010: 74):

The new general concept that was needed to provide for these 'interstratal' relationships was realisation… . A second new basic concept was instantiation — and it is the fact that we find both a 'potential' and an 'instance' at each of the two levels within the lexicogrammar that demonstrates the presence, within the grammar, of the two levels of 'meaning' and 'form'. 

Blogger Comments:

There are several misunderstandings here.

1. In terms of theoretical architecture, the relation between potential and instance, instantiation, does not "demonstrate the presence of" meaning and form, since the relation between the levels of meaning and form is realisation (symbolic abstraction).
2. In terms of epistemology, it is not the "presence" of meaning and form that can be demonstrated, but the explanatory power of proposing such a distinction in the model.
3. In terms of theoretical consistency, by construing realisation rules as potential, and structure as instance, Fawcett's level of form confuses instantiation with the realisation relation between the paradigmatic and syntagmatic axes.

The Relation Of The Cardiff Grammar To Halliday's Evolving Theory

Fawcett (2010: 73):

One reason why Halliday's "Language as choice in social context" is important, then, is that it is to this apparently aberrant stage of his developing model that the Cardiff Grammar is most closely related. … here my purpose is simply to note the fact of this phase in Halliday's frequently changing model.

Blogger Comments:

To be clear, Fawcett here positions his theorising in 2010 as most closely related to Halliday's theorising in 1972-6 (published in 1977 and 1978).  That is, while Halliday kept learning and improving his model on the basis of that learning, Fawcett staid put, with the result that few, if any, of the theoretical insights built into Halliday's model in the last half-century are likely to be found in Fawcett's interpretation of Halliday's model.

Seriously Misrepresenting Halliday's Early Theorising

Fawcett (2010:71-3):

The example analysed in Halliday (1977) as an illustration of these principles is fairly complex, i.e., "I would as soon live with a pair of unoiled garden shears, " said her inamorata. Halliday's analysis of this text-sentence raises a number of interesting questions that would distract us from the main point, so we shall look instead at the simpler example from the second of Halliday's two papers that exemplify the type of analysis that reflects the above description, i.e, that in Halliday (1970/76b). This is shown in Figure 6. 

Notice first that there is a line of structure for each of the 'ideational' and the 'interpersonal' strands of meaning: with the usual two for the 'textual' meaning — i.e., one showing the 'thematic' structure, and one the 'information' structure. But the key feature of the diagram is the single line of structure which comes below these four lines, and which uses the names of the elements of the clause which were established as part of Scale and Category Grammar — i.e, at the time when the assumption was that the whole grammar was at the level of form. Some of the labels for the 'functions' are different from those found in later representations such as those in IFG, e.g., "Modal" and "Propositional" are later replaced by "Mood" and "Residue", but we shall find in Chapter 7 that the first four lines of Figure 6 correspond closely to the type of analysis found in all of Halliday's work since that time.

Figure 6 is exactly as it occurs in Halliday (1970/76b), with the exception of the word "COMBINED", which I have added (borrowing it from the equivalent diagram in Halliday (1977). It is clear that, even though there are few explanatory comments on the diagram in either Halliday (1970/76b) or Halliday (1977), Halliday's intention is precisely that of showing that the structures represented in the four 'strands of meaning' above are "combined" in the single integrated structure shown below them. In Halliday's words (1970/76b:24): "any element [e.g., the Subject] may have more than one structural role, like a chord in a fugue which participates simultaneously in more than one melodic line." 

Blogger Comments:

This is very misleading indeed, since it misrepresents Halliday's theorising at the time. Specifically, here Fawcett chooses to ignore the analysis in the 1978 edition of the Halliday paper — which puts a lie to the claims he makes about this stage of Halliday's theorising — and instead, returns to an earlier stage (1970) of Halliday's model.  Halliday (1978: 130):

As can be seen above, even at this stage, Halliday had recognised the categories of Scale and Category Grammar (Subject etc.) as interpersonal functions of the clause. 

As Halliday (1978: 129) makes very plain, it is the clause itself that combines the metafunctional structures:

Fifth, we shall assume that the lexicogrammatical system is organised by rank (as opposed to by immediate constituent structure); each rank is the locus of structural configurations, the place where structures from the different components are mapped on to each other. …

It follows from the above that each type of unit — clause, verbal group, nominal group etc. — is in itself a structural composite, a combination of structures each of which derives from one or other component of the semantics. A clause, for example, has a structure formed out of elements such as agent, process, extent; this structure derives from the system of transitivity, which is part of the experiential component. Simultaneously it has a structure formed out of the elements modal and propositional: this derives from the system of mood, which is part of the interpersonal component. It also has a third structure composed of the elements theme and rheme, deriving from the theme system, which is part of the textual component.

Misrepresenting Halliday In A Footnote

Fawcett (2010: 71n):

However, it must be said that one of the five introductory 'assumptions' in Halliday (1977/78) seems to reflect the second of the two positions on meaning identified in Section 4.6. This is when he writes: "Let us assume that each stratum [...] is described as a network of options." This view is clearly incompatible with that modelled in Figure 4 of Chapter 3, because there the component that specifies the 'potential' at the level of form is the realisation rules — and so not a system network. Since the other assumptions are fully compatible with the model described in Figure 4 of Chapter 3, and since this one therefore appears to be at odds with those others, we shall take it that its inclusion in the paper is evidence that even as he wrote it Halliday was toying with the possibility of the second of his two models of meaning.

Blogger Comments:

[1] To be clear, as previously explained, Halliday has had only one position on meaning.  As previously explained, the difference between the early model of more than 40 years ago (Halliday 1977/78) and the current model is that the former proposed that semantic structures were mapped onto grammatical structures, an interstratal relation, whereas the latter proposes that it is grammatical systems that are realised by grammatical structures, an axial relation.  In both models, the content strata are semantics (meaning) and lexicogrammar (wording).

[2] To be clear, as previously explained, Fawcett's model, described in Figure 4, is inconsistent with all of Halliday's models, past and present, not least because it confuses axial realisation with instantiation, and distributes phenomena of the same level of symbolic abstraction, features, across different levels, meaning and form, according to whether they are positioned in networks or rules.

[3] To be clear, Halliday's third assumption, that each stratum is described as a network of options, is not theoretically inconsistent with his other four assumptions.  In this early paper, Halliday provides no systems for either semantics or grammar, but describes the organisation of the lexicogrammatical system in his fifth assumption; Halliday (1978: 129):

Fifth, we shall assume that the lexicogrammatical system is organised by rank (as opposed to by immediate constituent structure); each rank is the locus of structural configurations, the place where structures from the different components are mapped on to each other. The ‘rank scale, for the lexicogrammar of English is:

Fawcett On The Rôle Of Lexicogrammar In Halliday (1977/8)

Fawcett (2010: 71):

Halliday then makes the point that "these are strata in Lamb's sense", and the terms in brackets are intended to show the parallels with Lamb's multi-stratal model (Lamb 1966). Of the three levels distinguished here, it is the first two in which we are interested. Let us take the "semantic" level first. Halliday writes:

Let us assume that the semantic system has four components: experiential, logical, interpersonal and textual (1977:176). 

And two paragraphs later he suggests a third basic assumption:

Let us assume that each component of the semantic system specifies its own structures, as the Output' of the options in the network (each act of choice contributing to the formation of the structure). (1977:176) 

Since Halliday never shows any structures that can be de[s]cribed as "semantic" other than the multiple structures of functional elements found in IFG (which we shall be examining in the next chapter), it is clearly these to which he is referring. In other words, choices in the system network in the experiential component result in 'structures' such as 'Agent + Process + Affected'.

What, then, is the role of the level labelled "lexicogrammar" in this approach? It is a role that is very different from that which it is assigned in Halliday's 'two levels of meaning' approach to language, in which the "lexicogrammar" includes everything from the system networks of TRANSITIVITY, MOOD, THEME etc. to as their final realisation in form. Interestingly, Halliday writes in "Language as choice in social context" that

it is the function of the lexiogrammatical stratum to map the structures onto each other so as to form a single integrated structure [my emphasis] that represents all components [of the semantics] simultaneously. (Halliday 1877:176)

Blogger Comments:

[1] To be clear, Halliday (1978: 129) explicitly states this to be the case:

It follows from the above that each type of unit — clause, verbal group, nominal group etc. — is in itself a structural composite, a combination of structures each of which derives from one or other component of the semantics.

A clause, for example, has a structure formed out of elements such as agent, process, extent; this structure derives from the system of transitivity, which is part of the experiential component. Simultaneously it has a structure formed out of the elements modal and propositional: this derives from the system of mood, which is part of the interpersonal component. It also has a third structure composed of the elements theme and rheme, deriving from the theme system, which is part of the textual component.

[2] To be clear, the approach in this early paper, written between 1972 and 1976, is different from Halliday's later revised model.  In this paper, structures produced by semantic systems are mapped onto units of the grammatical rank scale.  In the current model, these semantic systems are reconstrued as lexicogrammatical systems, chiefly in order to systematically account for grammatical metaphor.  Semantic systems were later set out in Halliday & Matthiessen (1999).

[3] As previously explained, this is a (motivated) misunderstanding of Halliday by Fawcett.  Halliday has never proposed a "two levels of meaning" approach to language.  His approach has consistently been to propose two levels of content: meaning (semantics) and wording (lexicogrammar).

Fawcett's Claim That His Model Is 'Fully Compatible' With Halliday (1977/8)

Fawcett (2010: 70):

The general picture that it [Halliday (1977/78)] gives of the nature of language and of how the grammar works is fully compatible with the picture given in Chapter 3. That is, the system networks of TRANSITIVITY, MOOD, THEME and so on are presented as being at the level of semantics, and their realisations are integrated in a single structure at the level of form. (However, Halliday there terms it the 'lexicogrammatical level'; this is a little confusing, since Halliday later uses the term "lexicogrammar" in a sense that includes the system networks.) Thus the paper begins with the words:

Let us assume that the semantic system is one of three levels, or strata, that constitute the linguistic system: 

Semantic (semology)
Lexicogrammatical (lexology: syntax, morphology and lexis)
Phonological (phonology and phonetics). 

(Halliday 1977:176) 

Blogger Comments:

[1] This is demonstrably false, as shown in dozens of previous posts.  The picture given in Chapter 3, as represented in Figure 4, is inconsistent in its own terms (e.g. confusing axial realisation with instantiation) and inconsistent with Halliday's model — even with the embryonic model presented in 1977.

[2] For once, this is actually true.  At this early stage of theorising, when Halliday was encoding the semantics (Value) by reference to the lexicogrammar (Token), he located the systems from which 'grammatical structures derive' — TRANSITIVITY, MOOD and THEME — at the level of semantics.  That is, at this early stage, Halliday (virtually) conflated the axial relation of realisation between system and structure with the stratal relation of realisation between semantics and lexicogrammar.  This situation was rectified within the next decade, largely motivated by the need to account systematically for grammatical metaphor.  Later still, Halliday & Matthiessen (1999: 429) explain:

… in our model there are two system-structure cycles, one in the semantics and one in the lexicogrammar. Terms in semantic systems are realised in semantic structures; and semantic systems and structures are in turn realised in lexicogrammatical ones. As we saw in Chapter 6 in particular, grammatical metaphor is a central reason in our account for treating axis and stratification as independent dimensions, so that we have both semantic systems and structures and lexicogrammatical systems and structures. Since we allow for a stratification of content systems into semantics and lexicogrammar, we are in a stronger position to construe knowledge in terms of meaning. That is, the semantics can become more powerful and extensive if the lexicogrammar includes systems.

[3] This is misleading in a way that supports Fawcett's own model.  Unlike Fawcett, Halliday does not propose a level of form.  As the quote that Fawcett himself provides makes clear, Halliday proposes the three strata of semantics, lexicogrammar and phonology.

[4] To be clear, Halliday (1978: 128-9) proposes systems for both semantics and lexicogrammar:

Third, we will assume that each stratum, and each component [metafunction], is described as a network of options, sets of interrelated choices having the form 'if a, then either b or c'.

with the lexicogrammatical stratum system organised by rank:

Fifth, we shall assume that the lexicogrammatical system is organised by rank (as opposed to immediate constituent structure); each rank is the locus of structural configurations, the place where structures from the different [metafunctional] components are mapped onto each other.

Theoretical Indeterminacy, Computer Modelling And The Value Of Fawcett's Framework

Fawcett (2010: 69):

Halliday has written at different times in terms of both frameworks, and also in terms that suggest that the boundary between the two is indeterminate (Halliday 1996:29). While I recognise the 'indeterminacy' that is bound to be found in living systems such as natural human languages, I think that it is right to accept the challenge of trying to make the model sufficiently explicit to be incorporated in a computer model of language. And doing this in turn suggests the value of recognising the component 'modules' of Figure 4 in Chapter 3.

Blogger Comments:

[1] As previously demonstrated, this is misleading because it strategically misrepresents Halliday.  Halliday has never espoused the misunderstanding in Fawcett's model (Figure 4) of realisation rules as a lower level of abstraction ('form potential') than the features they apply to.

[2] This is misleading because it strategically misrepresents Halliday (1996), which is concerned with the theoretical indeterminacy of the nature and location of the boundary between semantics and grammar, not the boundary between Fawcett's framework and Halliday's.  Halliday (2002 [1996]: 411):

But there is another opening-up effect which is relevant to the present topic: this concerns the nature and location of the stratal boundary between the grammar and the semantics. This is, of course, a construct of the grammatics; many fundamental aspects of language can be explained if one models them in stratal terms, such as metaphor (and indeed rhetorical resources in general), the epigenetic nature of children’s language development, and metafunctional unity and diversity, among others. But this does not force us to locate the boundary at any particular place. One can, in fact, map it on to the boundary between system and structure, as Fawcett does (system as semantics, structure as lexicogrammar); whereas I have found it more valuable to set up two distinct strata of paradigmatic (systemic) organisation. But the point is that the boundary is indeterminate – it can be shifted; and this indeterminacy enables us to extend the stratal model outside language proper so as to model the relationship of a language to its cultural and situational environments.

[3] Strictly speaking, living systems are biological, whereas languages are semiotic, though both can be construed as evolving, complex adaptive systems.

[4] To be clear, making a model of human language explicit is distinct from adapting a model to the constraints of computer technology; and this is distinct from claiming that such an adapted computer model also applies to language as a natural phenomenon.  And these are distinct from the questions of whether a model is internally self-consistent or consistent with the phenomenon being modelled.

[5] To be clear, this is a non-sequitur.  Merely adapting a theoretical architecture to the constraints of computer technology does not, of itself, "suggest the value" of Fawcett's particular method of doing so, nor does it say anything about its value as a model of human language.

[6] To be clear, the architecture of SFL theory is dimensional, not modular, and Fawcett's model (Figure 4) is a flowchart for text generation.  Moreover, as previously demonstrated, it is internally inconsistent, confusing the realisation relation between the paradigmatic and syntagmatic axes with instantiation relation between system and instance (inter alia).

On 'Form Potential' Being Limited To Realisation Rules

Fawcett (2010: 69):

But when the 'form potential' is not given a specific identity in overall diagrams of how language works, as is the case in diagrams where language is represented as a system network with the realisation rules presented as 'footnotes' on the features, there is a temptation to see the processes described in Figure 4 and exemplified in Appendix A as all occurring within one 'level', as in Figure 5.

Blogger Comments:

[1] To be clear, Fawcett limits 'form potential' to realisation rules only, and excludes systems (potential) from his 'form potential'.  In SFL theory, however, grammatical form is theorised as a rank scale of units — clause, group/phrase, word, morpheme — each of which is the entry condition to a system of potential.  Because SFL is a functional theory, the potential of such forms is modelled in terms of the functions (Senser, Finite, Theme etc.) that their formal constituents serve.

[2] To be clear, given the above, 'form potential' is "given a specific identity in overall diagrams of how language works", whether it is defined in Fawcett's narrow sense of realisation rules or in the broader SFL sense of rank-ordered systems with realisation statements located at their place of application.  This is because the theoretical distinction between realisation statements and system networks does not depend on their formal arrangement but on the different theoretical functions they serve.

[3] To be clear, this is a statement about Fawcett's model (Figure 4), not about the architecture of SFL theory, despite being presented as such.  As previously explained — see Attacking A Straw Man — Figure 5 is Fawcett's reworking of his own Figure 4 that he falsely attributes to Halliday and falsely claims to be topologically equivalent to his Figure 4.

As previously explained, in Figure 4, Fawcett confuses realisation with instantiation; and in Figure 5, Fawcett misattributes his own confusion to Halliday.

'Form Potential' As A Separate Component

Fawcett (2010: 69):

When the 'form potential' component of a systemic functional grammar is shown as the separate component that it undoubtedly is (as in Figure 4), this helps to make it clear that the system networks are a different component from the realisation rules — and one that is at a higher level.

 Blogger Comments:

[1] To be clear, here Fawcett indulges in the logical fallacy of circular reasoning (circulus in probando):

the reasoner begins with what he or she is trying to end up with; sometimes called assuming the conclusion.

That is, Fawcett's argument is if P then P, where

P: 'form potential' is shown clearly as a separate component from system networks.

[2] To be clear, Fawcett's logical fallacy here might be interpreted as a variant of proof by repeated assertion.  Moreover, as Halliday (1994: 89) points out: 'you only say you are certain when you are not'.

[3] As previously explained, Fawcett's model (Figure 4) misconstrues the same level of symbolic abstraction, grammatical features, as different levels of abstraction, depending on whether they figure in system networks or in realisation rules.  Moreover, Fawcett's model misconstrues the relation between realisation rules and what realises them — i.e. realisation — as the instantiation relation between potential and instance.

Fawcett's Justification For Realisation Rules As A Separate Component

Fawcett (2010: 69):

To summarise: when simple realisation statements are written under the features in network diagrams, these are best regarded as an informal version of the full realisation rule. Such diagrams may have the laudable effect of focussing attention on the system networks themselves — but they bring with them the unfortunate side-effect that they make the realisation rules appear to be relatively minor 'footnotes' to the features in the networks. And they are not. In a fully explicit theoretical model of how language works, therefore, it is necessary to show the realisation rules as a separate component, as was done in Halliday's first generative grammars (as cited above), in other early systemic grammars such as the very large one described in Hudson (1971), and in all versions of the Cardiff Grammar (e.g., in Figure 2 of Appendix B). Including the conditions on realisation within the system network has the further disadvantage that it muddles two aspects of language: (1) choices between meanings and (2) their realisation at the level of form.

Blogger Comments:

[1] To be clear, this is not a summary of Fawcett's previous argument, nor is it logically entailed by what has been argued; see previous posts.  Instead, it is a new bare assertion of what Fawcett assesses as "best", unsupported by evidence or argument.

[2] To be clear, the system network is the formalism of Systemic Functional Linguistic Theory, and embodies the theory's fundamental perspective on language: meaning as choice.  Fawcett's claim here is that it is "laudable" to focus attention on the theory's formalism.

[3] To be clear, this, again, is not a summary of Fawcett's previous argument, nor is it logically entailed by what has been argued; see previous posts.  Instead, it is a new bare assertion of what Fawcett assesses as "unfortunate", unsupported by evidence or argument.

Moreover, representing realisation statements in system networks has the advantage of being consistent with theory and displaying how the conditions of their application relate to all other choices within the entire system.

[4] To be clear, this is not entailed by the previous argument, nor by the new bare assertions that are misrepresented here as a summary of the argument.  As previously explained, Fawcett's model of realisation rules as a separate component (Figure 4) misconstrues one level of symbolic abstraction, grammatical features (in systems and rules) as two levels (meaning and form), and misconstrues the realisation relation between the paradigmatic and syntagmatic axes as the instantiation relation between potential and instance.

[5] This is misleading.  On the one hand, Halliday's "first generative grammars" did not model realisation rules along the lines of Fawcett's model (Figure 4), and on the other hand, Halliday's "first generative grammars" didn't model realisation rules as a component, because the architecture of SFL theory is dimensional, not componential (modular).

[6] This is another example of the informal logical fallacy known as

argumentum ad populum (a.k.a. appeal to widespread belief, bandwagon argument, appeal to the majority, appeal to the people) wherein a proposition is claimed to be true or good solely because a majority or many people believe it to be so.

[7]  To be clear, here Fawcett assesses Halliday's model in terms of the misunderstandings in his own.  Specifically, on the basis of his own misunderstanding of features (in systems and rules) as different levels of symbolic abstraction, he assesses Halliday's model as muddled because it doesn't mistakenly assign them to different levels of abstraction.  Moreover, in SFL theory, realisation statements specify the function of form, not form, as demonstrated by the realisation statement 'Finite^Subject' for polar interrogative MOOD.

To be clear, in SFL theory, grammatical form is modelled as a rank scale (clause, group/phrase, word, morpheme).  Each form (rank unit) is structured in terms of function, with function being the meaning encoded by the wording.  (wording = Identifier/Token, meaning = Identified/Value)

In the absence of grammatical metaphor, meaning and wording are congruent (in agreement).  It is grammatical metaphor that motivates the distinction between meaning (semantics) and wording (lexicogrammar).

Fawcett's Approach To Realisation Rules As "The Only One That Is Workable In A Large-Scale SF Grammar"

Fawcett (2010: 68):

This second approach [i.e. Fawcett's realisation rules as form potential] is in fact the only one that is workable in a large-scale SF grammar. The reason is simple: it is that the number of realisation rules that require conditions grows as the grammar is extended to cover the less frequent linguistic phenomena. Thus it often happens that an action in building a part of the structure is dependent on the co-selection of one or more other features. 

As the coverage of the grammar grows fuller, then, it has to encompass more and more exceptions to the general rule, and the place of the general concept of 'conditions on realisation rules' becomes correspondingly more important. It is interesting to study the nature of the realisation rules presented in Fawcett, Tucker & Lin (1993) from this viewpoint.

Blogger Comments:

[1] To be clear, there may be many viable approaches to realisation rules, but Fawcett's is not one of them.  This is because, as previously demonstrated, his approach (Figure 4) confuses realisation with instantiation, and misconstrues one level of symbolic abstraction — grammatical features (in systems and rules) — as two distinct levels (meaning and form).

[2] This misunderstands the SFL notion of system.  A well-formed system, in principle, covers all linguistic phenomena of the domain stipulated by its entry condition, across all frequencies — or more precisely, since a system models potential, across all probabilities, since feature frequencies in texts are instances of feature probabilities in the system.

[3] As previously demonstrated, Fawcett's argument in this regard is based on one of his own system networks (Figure 2, Appendix 1), which is inconsistent both with SFL theory, misconstruing 'deixis' as a system of 'thing', and with the principles of a system network, misconstruing lexical items such as 'student' as grammatical features.  That is, Fawcett has merely demonstrated his own inability to devise system networks and to locate realisation statements in them at their point of application.

[4] Yes.  It is.

Summary Of The Argument For Realisation Rules As A Separate Component

Fawcett (2010: 68):

To summarise so far: the insistence that realisation rules must not contain conditional features so that they can be simple enough to be written in on the system network makes the additional 'wiring' in the network quite complex, and the greatly preferable alternative is to place all of the realisation rules together in a separate component — i.e., the component that specifies the 'form potential' — as shown in Figure 4 and demonstrated in Figure 2 of Appendix A.

Blogger Comments:

[1] To be clear, whether realisation rules are included in system networks, or listed separately, are just two ways of representing the same linguistic complexity.  As previously demonstrated, Fawcett's argument does not determine which representation is preferable, since it is based on a network of his own (Figure 2 of Appendix A) which is not consistent with the principles of a system network.

[2]  This a non-sequitur.  To be clear, representing realisation rules separately from system networks does not entail theorising them as separate component in the model (Figure 4).

More importantly, locating realisation rules as a component of form potential is inconsistent with the dimensions of Fawcett's own model.  Firstly, it misconstrues features and the rules that apply to them as different levels of symbolic abstraction, meaning and form.  And secondly, it misconstrues the realisation relation between the paradigmatic axis (realisation rules) and syntagmatic axis (structure) as instantiation (potential to instance).

On The Value Of Conditional Features In Realisation Rules

Fawcett (2010: 68):

This example of the alternative approaches to a relatively simple part of the grammar demonstrates clearly the value of the use of conditional features in realisation rules. But it also underlines the value of respecting the distinction between the use of the system network notation for representing systemic relationships of choice, and the mis-use of them (as it seems to me to be) to represent conditions on realisation. It is clearly preferable in the case we are considering here, as it is in any model, to have different notations for the two different concepts. This is why, in Appendix B, system networks are used to model choices in meaning (as in Figure 1) and tables are used to model the realisation rules (as in Figure 2). Indeed, this follows the pattern established in Halliday's early grammars (e.g., 1969/81 and 1970/76b).

Blogger Comments:

[1] As previously demonstrated, the example Fawcett uses violates the principles of genuine system networks, and as a consequence, arguments made using it are thus examples of the 'straw man' logical fallacy, and do not apply to system networks.

[2] To be clear, the point at issue is how the conditions on the activation of realisation statements are to be represented.  In SFL theory, the conditions are represented by the wiring of networks and the location of features in the network.  In Fawcett's model (Figure 4), realisation rules are incongruently located at a different (lower) level of symbolic abstraction (form) than the features (meaning) to which they apply.

[3] To be clear, this is the opposite of what is true.  Fawcett's example does not underline the value of this distinction, since it actually demonstrates that it is even possible to incorporate realisation rules into his own misunderstanding of a system network; see previous post.

[4] These are bare assertions, unsupported by evidence or argument.

[5] As previously noted, in this matter, Fawcett misrepresents a presentational limitation in early publications as a theoretical distinction.

Rewiring Fawcett's Network To Include Realisation Rules

Fawcett (2010: 67-8):

There is an alternative solution, and we shall explore it here briefly in order to demonstrate that it is not a desirable answer to the question of how best to model conditions on realisations. It is to model the conditions by the use of the conventions of a system network. Continuing with the example from Appendix B, we would need to extend the existing relatively simple network in Figure 1 in the following ways. We would need to add (1) a right-opening 'and' bracket after each of [mass], [singular] and [plural], and (2) a right-opening 'or' bracket after [near]. Then (3) a line would need to be drawn from each of the three 'and' brackets associated with [mass] and [singular] to a new left-opening 'or' bracket, with (4) a further line running from the latter to a new left opening 'and' bracket. This would also be entered by a line from the right-opening 'or' bracket by [near] (5). Then (6) a dummy feature (standing for the meaning 'near-and-singular-or-mass') would need to be inserted to the right of the left-opening 'and' bracket. This would be a case of what is termed a 'gate', i.e., a feature that is in the system network but which is not part of a system.* Next, we would need to draw a line from the right-opening 'and' bracket by [plural] to a second new left-facing 'and' bracket (7), and (8) this would also be entered by a line from the second branch of the right-opening 'or' bracket' by the feature [near]. Then (9) a second 'dummy' feature would be placed to the right of this left-opening 'and' bracket, standing for the meaning 'near-and-plural'. As a result of the addition of all this new 'wiring' it would be possible to insert two realisation rules which would not have conditions attached to them, i.e., one that stated that the feature 'near-and-singular-or-mass' would be realised by the item this, and one that said that ' near-and-plural' is realised by these.

* Clearly, this concept is an anomaly in a systemic grammar; see Fawcett, Tucker & Lin (1993:126) for a discussion of the concept of 'gate', which is widely used in the computer implementation of Halliday's version of SFG in the Penman Project to minimise the use of conditions on realisation rules (e.g., Mann & Matthiessen 1983/85). However, its theoretical status requires further clarification, discussion and justification before it is given the status in the theory that is accorded to the concept of a system.

Blogger Comments:

[1] Here again Fawcett uses his own network, which, as previously demonstrated, violates the principles of the system network, in order to argue against the inclusion of realisation statements in genuine system networks.  However, what Fawcett actually demonstrates is that it is even possible to include realisation rules in such a network — at least, for those that specify grammatical items rather than structural realisations.

[2] To be clear, in rewiring Fawcett's network, there is no need for "a right-opening 'and' bracket" after the features [singular] or [plural], in this example, because only one wire extends from each of these features.

                     

On The Timing Of The Application Of Realisation Rules

Fawcett (2010: 66-7):

In practice the simplest workable solution is that no realisation rules should be applied until the traversal of the network has been completed, and to make the resulting selection expression of features available to each realisation rule, as it is applied. This is what is done in the Cardiff Grammar and, as we shall see in the next chapter, in the computer implementation of Halliday's grammar too.

Blogger Comments:

To be clear, here Fawcett confuses system networks, the SFL formalism that models human language as potential, with the flowcharts he uses for text generation by computer.  In doing so, he presents his solution to his own technical problem as a solution to a non-existent theoretical problem (as demonstrated in previous posts).

Fawcett's Argument Against Realisation Rules In System Networks

Fawcett (2010: 66-7):

Let us look at a simple example, taken from the little grammar in Appendix B. Consider the realisation rule for the feature [near] in Figure 2 of Appendix B. The rule states that, if either [singular] or [mass] is also chosen, the realisation is that the deictic determiner (dd) will be expounded by the item this, but that if the feature [plural] is co-selected it will be expounded by these. 

In this particular example, the conditional features happen to occur in a sub-network that is 'higher' on the page than the one in which the feature [near] occurs. This might lead you to think that this makes it possible for the realisation rule for the feature [near] to fire as soon as it is chosen, on the grounds that the grammar already 'knows' whether the conditional features have or have not been chosen. However, the features that function as 'conditions' could equally well occur in a part of the network to be traversed later, so that we cannot proceed on this assumption.* 

* Neither approach would be acceptable in the Sydney Grammar, however, because there is a strong insistence on the concept that, in principle, all systems are entered simultaneously. If this is the case, the grammar would not know whether a possible conditional feature had or had not been co-selected at the time when the feature [near] was chosen. (The computer implementation of the Cardiff Grammar currently operates on the assumption that the 'higher' system networks are traversed before the lower ones, but they could be reformulated if it ever became possible to apply the computational concept of 'parallel processing' to system networks.)

Blogger Comments:

[1] Figures 1 and 2 of Appendix A:

To be clear, this network is not consistent with the principles of SFL theory.  For example, the network

• confuses Thing with Deictic + Thing in a nominal group,
• misunderstands delicacy, in that it presents examples of mass nouns as more delicate features of the feature [mass], and examples of count nouns as more delicate features of the feature [count], and
• presents grammatical classes (mass, count) of nouns as semantic features.

That is, Fawcett's argument that realisation rules cannot be located in system networks only applies to misunderstandings of system networks, such as this devised by Fawcett himself.  As such, Fawcett's argument  here is merely another deployment of the Straw Man fallacy.

A system network, featuring realisation statements, that is consistent with the principles of SFL, from Halliday & Matthiessen (2014: 366), is presented below for comparison:



[2] As previously explained, Fawcett misunderstands system networks as flowcharts, and it is this misunderstanding that leads him to be concerned with the temporal order of feature selection.  To be clear, system networks are networks of relations.  It is the instantiation process that unfolds in time, during logogenesis, not traversals of system networks.