David Rose On The Need For (Discourse) Semantics As A Separate Stratum

David Rose replied to Mick O'Donnell on sys-func on 12/7/22 at 21:36:

Re your earlier comment about ‘the need for Semantics as a separate stratum’... I think that takes us back to two first principles...

One is that features in systems are no more substantial than the probabilities of their realising structures being re-instantiated... more so for more general and less so for more delicate features.

The other is that the content planes of lg and other modalities have evolved to semioticise social relations, for which semiosis is a necessary condition. So the social and semiotic are two sides of the same coin.

The semiotic systems of most species are reproduced genetically. In some species these are augmented by learnt systems, i.e. culturally reproduced.

Lg evolved in humans in tandem with the complexity of our social relations. It required on one hand sufficient flexibility to manage variability in our ancestors’ sociality, and on the other a mechanism for replication fidelity to ensure its reproduction across multiple generations (on the model of genetic reproduction).

The tension between these selection pressures produced a bifurcation in lg’s content plane, still recapitulated in the transition from protolg to mother tongue in infants.

On one hand, LG systems provided the replication fidelity for reproduction over deep time (analogous with the reproductive role of DNA). Their features cement untold millions of re-instantiations, more so at higher ranks. One result is their more general features remain common across lgs (as humans share 85% of our DNA with mice, 60% with insects, and 50% with plants). It is probably also why grammatical systems encode such stable models of experience and exchange.

While stability is essential for cultural reproduction, it is insufficient for negotiating unfolding contingencies in human social relations. That is the task for which DS systems evolved (analogous with the role of RNA in managing intracellular functions).

Both LG and DS systems are necessary for human semiosis on phylogenetic and logogenetic timescales, but their functions are complementary. The fact that we are still trying to model the semantic stratum on what we known of LG systems is, I think, an accident of history. From 2500 years of grammatics, cryptogrammatics has struggled to emerge in the last 80, and DS systems are another order of covert reactances.

Blogger Comments:

[1] To be clear, Halliday & Matthiessen (1999: 237) identify grammatical metaphor as the primary motivation for a semantic stratum:

If the congruent pattern had been the only form of construal, we would probably not have needed to think of semantics and grammar as two separate strata: they would be merely two facets of the content plane, interpreted on the one hand as function and on the other as form.

[2] To be clear, features in systems are defined by their relations to other features, and each has a probability of being instantiated, varying for register, and actualised as frequencies in texts.

[3] It is undoubtedly true that semiosis is a necessary condition for semiotising social relations.

[4] To be clear, not all semiotic systems are social semiotic systems, as demonstrated by perceptual semiotic systems, which are somatic, not social. 

[5] This is a bare assertion, unsupported by evidence. The difficulty here lies in being able to distinguish, in Halliday's terms, systems that involve the exchange of value (social systems) from systems that involve the exchange of symbolic value (semiotic systems).

[6] To be clear, it is language that enabled the enactment of social relations to complexify beyond those of the other chimpanzees.

[7] To be clear, if it were these factors that produced the stratification of the content plane, then they should have done so for all species' protolanguages, not just human protolanguage.

[8] This is a false analogy, since it correlates generality with proportion (%).

[9] This is misleading, because it is untrue. On the one hand, it misunderstands evolution: systems do not evolve in order to perform tasks (purpose); they evolve because they successfully perform tasks (reason). On the other hand, it is the grammar (MOOD) that construes the semantics (SPEECH FUNCTION) that makes symbolic negotiation possible.

[10] This a false analogy, since RNA is at the same level of abstraction as intra-cellular functions, whereas semantics is a higher level of symbolic abstraction than lexicogrammar.

[11] To be clear, the sense in which semantics and grammar are complementary is as complementary perspectives on the same phenomenon, the content plane of language, varying in terms of symbolic abstraction. With regard to the ideational metafunction, Halliday & Matthiessen (1999: 237) write:

Of course, what we are recognising here as two distinct constructions, the semantic and the grammatical, never had or could have had any existence the one prior to the other; they are our analytic representation of the overall semioticising of experience — how experience is construed into meaning.

[12] This is misleading. On the one hand, it is the grammar that construes the semantics. On the other, the false implication is that Martin's discourse semantic systems are not 'modelled in terms of what we know of lexicogrammatical systems'. To be clear, Martin's discourse semantic systems derive from non-structural textual grammar (cohesion) and from SPEECH FUNCTION, which is construed by the interpersonal grammar of MOOD and KEY. Evidence here.

David Rose Misrepresenting Hasan's System As A Synthesis Of Martin's Systems

David Rose replied to Mick O'Donnell on sys-func on 12/7/22 at 9:14:

What differs between the MESSAGE features in Figure 24.4 below and the mood features that realise them is their systemic valeur. The valeur of MOOD features is shaped by similarities and contrasts in the mood structures that realise them, exemplified in Table 24.4.

The valeur of message features is shaped by three factors as far as I can see. System G features by their position in exchange structures: verify/enquire etc. System F features by appraisal: precise/tentative; by field: specify/explain; again by exchange structure: prompted/unprompted; and again by field: global/particular...

In sum, the MESSAGE system attempts to synthesise the complementarity of two discourse semantic systems, NEGOTIATION and APPRAISAL. …

Blogger Comments:

[1] To be clear, the critical distinction between the semantic features of the SPEECH FUNCTION system and the lexicogrammatical features of the MOOD system is that each system of features is located at a different level of symbolic abstraction.

[2] This is misleading. To be clear, the valeur of each MOOD system feature is its relations to other features in the system; see, for example, Halliday & Matthiessen (1999: 96, 509). This misrepresentation of 'valeur' serves as a pretext for the misrepresentations that follow.

[3] To be clear, the valeur of each of Hasan's SPEECH FUNCTION system features is its relations to the other features in the system, and as such, they are not "shaped by" other systems such as APPRAISAL or FIELD, the latter being Martin's misunderstanding of ideational semantics as register misunderstood as context.

[4] This is very misleading indeed, because it misrepresents Hasan's (earlier) system as an attempt to synthesise Martin's (later) systems, one of which, NEGOTIATION, is Martin's rebranding of SPEECH FUNCTION.

Mick O'Donnell Confusing Stratification With Instantiation And Delicacy

Mick O'Donnell replied to David Rose on sys-func on 11/7/22 at 22:07:

David Rose wrote: The semantic networks appear to re-systemicise the LG systems.

But that is exactly the role of semantics, a more abstract representation of the utterance. I also note the "systemicisation" in the semantics may reflect a totally different way choices are realised in the grammar. For instance, the choices in the speech functional network reflect lexicogram choices in clause mood and mood-tag, but also in intonation, and in one case the clause complex. So, the SEMANTIC NETWORK generalises over various forms across the grammar, which is what one would want.

Blogger Comments:

[1] This confuses the hierarchy of stratification with the cline of instantiation. To be clear, the semantic stratum is more abstract than lexicogrammatical stratum, whereas the utterance is the text, an instance of the system.

[2] This confuses the hierarchy of stratification (symbolic abstraction) with the scale of delicacy (generalisation).

Mick O'Donnell On Semantics–Grammar Mappings

 Mick O'Donnell replied to David Rose on sys-func on 11/7/22 at 22:01:

Firstly, one example from a non-Martin model does not validate your claim that ALL Hallidayan models map 1-to-1 from Semantics to Grammar.

Secondly, Yes, I agree that there is an apparent one to one mapping from Speech Function to Grammatical Realisation in Geoff's model. But in this case, I take this as a good thing. If we take seriously the claim that differences in form relate to differences in meaning, there should be semantic differences between the different grammatical metaphors.

The problem really stems from whether you are looking at the mapping of partial semantic specifications (which should have alternative realisation possibilities) or whether you are looking at mapping full semantic specifications (which should constrain the possibilities down to a single form).

In things I have looked at, it is often the case that two sentences are ideationally the same, but textually (informationally) different (e.g., John gave Mary a book vs Mary was given a book by John; or John might be leaving vs. John is possibly leaving). Just looking at the ideational meaning in isolation, there are multiple ways the meaning can be realised. But when one takes ideational, interpersonal and textual meaning all together at once, the [sic] will ideally be a single utterance that realises that meaning.

(exceptions here:

i) I am ignoring here the possibility that some models allow Context of Situation to "skip" semantics and affect lexico-grammatical choices without mediation of semantic choices. Add to this Context of Culture difference (which may lead to different ways in which a meaning is mapped onto form)

ii) some differences in form may not have any correlation in meaning, e.g., is there a meaning difference between "the man that I saw" and "the man who I saw"? Or between [sic]

In the case of Geoff's network, the difference in ways to ask for information are not so much cross-metafunction, but intra-metafunction: the different sub-choices reflect different interpersonal strategies for getting information. And as this network offers a complete speech-functional account of the possibilities, the end-points can be attached to definite lxg realisations.

We only get a many-to-one mapping between semantics and form when we work on partial semantic descriptions (e.g., demand:information). As I say above, a complete specification of meaning should result in a complete specification of form (barring as I said above models where context skips semantics to directly affect form choices)

Blogger Comments:

[1] To be clear, Halliday's theory distinguishes between 'compact' and 'dispersed' grammatical realisations of semantic systems, and adds that compactly realised systems 'may become dispersed in their realisation through grammatical metaphor'  (Halliday & Matthiessen 2014: 666-7).

[2] To be clear, the speech function network under discussion is from Hasan (1983), and its gloss includes one-to-many mappings between semantic features and grammatical realisations; see the previous post.

[3] Here again O'Donnell follows Fawcett in misconstruing the semantics-grammar relation as a relation between meaning and form. To be clear, in SFL Theory, grammatical form is modelled as a rank scale, and forms (e.g. nominal group) are interpreted in terms of their function in realising meaning (e.g. Phenomenon). In the absence of metaphor, semantics and grammatical function agree (are congruent).

[4] To be clear, the notion of different forms realising different meanings is independent of grammatical metaphor, and grammatical metaphor is not a meaning-form relation, but an incongruence between the function of a form (grammar) and the meaning that it realises (semantics).

[5] To be clear, here O'Donnell introduces two 'red herrings': agnate structures and the integration of metafunctional meanings into a single syntagm, as evidence on the question of whether or not there are one-to-one mappings between semantic and lexicogrammatical features.

[6] To be clear, the notion of semantics being "skipped" is nonsensical. Linguistic strata are the same phenomenon viewed at different levels of symbolic abstraction. In terms of the ideational metafunction, Halliday & Matthiessen (1999: 237) write:

Of course, what we are recognising here as two distinct constructions, the semantic and the grammatical, never had or could have had any existence the one prior to the other; they are our analytic representation of the overall semioticising of experience — how experience is construed into meaning.

[7] To be clear, SPEECH FUNCTION is an interpersonal system, and since the network from Hasan (1983) is part of her SPEECH FUNCTION system, it is necessarily "intra-metfunction".

[8] To be clear, although O'Donnell begins by continuing his argument against Rose's false claim that Halliday's theory posits a one-to-one mapping between semantic and grammatical features (first paragraph), he soon forgets what side of the argument he is on, and switches to arguing for Rose's claim, and ends by concluding in its favour (final paragraph). The switch begins when O'Donnell assesses what he wrongly accepts as a one-to-one mapping as 'a good thing' (second paragraph).

David Rose Misrepresenting Speech Function

David Rose wrote to sys-func on 11/7/22 at 22:00:

As the systems below show, the grammatico-semantic model assumes a bijective or one-to-one relation between LG features and semantic features. What differs is the intrastratal relations between the features. The semantic networks appear to re-systemicise the LG systems.

Crucially, the names given to the semantic features imply moves in unfolding exchanges. They are therefore potentially complementary with the exchange systems first proposed by Margaret Berry and further developed in ET as the discourse semantic system of NEGOTIATION. 

Blogger Comments:

[1] This misleading, because it is not true. As Table 24.4 details, the semantic feature 'alternative' has multiple systemic lexicogrammatical realisations.

 

[2] To be clear, a semantic network represents a higher level of symbolic abstraction than a lexicogrammatical network.

[3] The implication here is misleading. To be clear, Halliday's system of SPEECH FUNCTION, part of which Hasan (1983) develops in Figure 24.4, has always been concerned with 'moves in unfolding exchanges'. The title of the chapter on mood in IFG, has always been 'clause as exchange'.

[4] This is very misleading indeed. The original work is by Halliday and Hasan. The reason why the work of Berry and Martin is "complementary" is that both are based on the original work. Rose's wording falsely suggests that they are parallel independent developments. For a close examination of Martin's system of NEGOTIATION, see here (English Text) and here (Working With Discourse).

David Rose On Incongruent Grammatical Realisations And 'Wording'

David Rose replied to Mick O'Donnell on sys-func on 11/7/22 at 20:10: 

I’m happy to be mistaken but my understanding is that incongruent grammatical realisations of speech functions are treated as delicate features in message semantic systems. That is, each feature in the semantic systems has a particular grammatical realisation. … 

One issue with your synopsis below, is a suggested correlation of form and meaning with grammar and semantics. This is also suggested by the oft-used terms ‘wording’ vs ‘meaning’, as though grammar was not meaningful in itself. What makes grammar meaningful is axis... features realised as structures. The features are the meanings of the structures. Or rather the features and structures are two faces of the meanings, that MAKH showed us are made by the grammar.

Blogger Comments:

[1] This is clearly true, as 140+ posts on this blog demonstrate.

[2] This is a serious misunderstanding. Incongruent realisations are relations between strata, whereas delicacy is the ordering principle of the system of a stratum. Delicacy is the ordering of systemic features that may be realised congruently or incongruently on the stratum below.

[3] This is misleading, because it is untrue. To be clear, the "oft-used" distinction 'wording vs meaning' (Halliday 1970) does not correspond to the distinction between form and meaning. The term 'wording' is used in SFL Theory as the folk equivalent of 'lexicogrammar', and it refers to the interpretation of form in terms of its function of realising meaning, as when a formal constituent of a clause, a verbal group, is interpreted as serving the function of Process of a clause.

[4] This is misleading, because it is not true that axis makes grammar meaningful. This is demonstrated by the fact that axis is a dimension of every stratum, including phonology. As Halliday & Matthiessen (2014: 20, 32) explain, on each stratum, the dimension of axis has two orders: the paradigmatic (the dimension of system) and the syntagmatic (the dimension of structure).

[5] This is misleading. In SFL theory, grammatical systems (e.g. PROCESS TYPE) and the structures that realise them (e.g. Senser ^ Process) are both interpreted in terms of the meanings they realise.

[6] Rose is here drawing on Martin's misunderstanding of stratification as 'all strata make meaning', which is a statement of semogenesis, not stratification. This confusion leads Martin (1992) to misconstrue even phonology as a level of meaning. Evidence here.

Mick O'Donnell On Semantics And Grammar

Mick O'Donnell replied to David Rose on Sys-func on 11/7/22 at 17:57:

I would not think any Hallidayan would ever posit a 1-to-1 mapping between Semantics and Grammar. That would invalidate the need for Semantics as a separate stratum. …

So, the mapping between Semantics and Grammar always needs to be seen as a many-to-many mapping.

- language provides multiple forms for realising a particular meaning.

- a particular form can express alternative meanings

(but yes, "Construing Experience" can be read (misread?) to suggest a one to one mapping between semantic process type selection and Transitivity selection in the grammar. Something I and other have argued against in the past. And not something that I think was intended by the authors, Christian and MAKH).

Blogger Comments:

[1] To be clear, it is metaphor that motivates semantics as a stratum of content. Halliday & Matthiessen (1999: 237):

If the congruent pattern had been the only form of construal, we would probably not have needed to think of semantics and grammar as two separate strata: they would be merely two facets of the content plane, interpreted on the one hand as function and on the other as form.

[2] Here O'Donnell misconstrues the distinction between semantics and grammar as a distinction between meaning and form. To be clear, the distinction in SFL Theory is between meaning and wording, where wording is form (the rank scale) interpreted in terms of its function of realising meaning; e.g. verbal group (form) as Process (function).

[3] This is misleading. On the one hand, PROCESS TYPE is a grammatical system within the larger system of TRANSITIVITY (Halliday & Matthiessen 2014: 219, 354-5), so, in this sense, the notion of a 'one to one mapping' of 'process type selection' to 'transitivity selection' is nonsensical.

On the other hand, in the absence of grammatical metaphor, the process type served by a verbal group is congruent with its semantic value. Moreover, in the case of grammatical metaphor, the incongruence is not between one process type and another, but between a process and some other element, as when a Process is realised grammatically by a Range participant, as he had a bath.

[4] To be clear, O'Donnell misunderstands the system of PROCESS TYPE, largely because he fails to recognise verbal group complexes. See, for example:

Mick O'Donnell On Process Types [1]

Mick O'Donnell On Process Types [2]

Mick O'Donnell On The Number Of Process Types

Mick O'Donnell On Process Types

Mick O'Donnell On The Difference Between 'Should Not Fail To See' And 'Should See'

David Rose Misrepresenting Halliday On Stratification

David Rose wrote to sys-func on 11/7/22 at 11:53:

To clarify, the grammatico-semantic model assumes a bijective or one-to-one relation between LG features and semantic features, while the discourse semantic model allows for multivalued or one-to-many relations between DS features and LG features. The latter is consistent with MAKH’s model of congruent/incongruent relations between semantic features and LG features. This was described in Ch10 of IFG1/2 but was replaced in IFG3/4 by ‘‘transgrammatical semantic domains’, which relocates multivalued realisation from an interstratal to an intrastratal relation within semantics.

Blogger Comments: 

[1] To be clear, by the 'grammatico-semantic' model, Rose means the 'standard model' of SFL Theory: the stratification of the content plane into lexicogrammar and semantics.

[2] This is very misleading, because it is very untrue. On the 'standard model', one semantic system may be realised across many lexicogrammatical systems. Halliday & Matthiessen (2014: 667) term this 'dispersed realisation':

[3] To be clear, the discourse semantic systems of IDENTIFICATION, IDEATION and CONJUNCTION are Martin's rebrandings of Halliday & Hasan's (1976) textual lexicogrammatical systems of REFERENCE, LEXICAL COHESION and COHESIVE CONJUNCTION. Unsurprisingly, the grammatical realisations of these discourse semantic systems are not specified in either Martin (1992) or in Martin & Rose (2007); see here and here.

[4] On the one hand, one-to-many relations between discourse semantics and lexicogrammar is a different matter from congruent/incongruent relations, since the latter, but not the former, is the contrast between non-metaphorical and metaphorical realisations.

On the other hand, this is misleading, because it is untrue. Martin's model is not consistent with Halliday's model of congruent/incongruent relations, because it does not identify what constitutes congruent versus incongruent grammatical realisations of its discourse semantic systems. (NEGOTIATION is Martin's rebranding of Halliday's semantic system SPEECH FUNCTION.)

[5] This is very misleading, because it is very untrue. 'Transgrammatical semantic domains' did not replace 'congruent/incongruent relations' in IFG3/4. Halliday & Matthiessen (2014: 665):

Table 10-1 shows how semantic units are mapped onto grammatical ones. The principle is that of rank-based constituency – semantic unit a ↘ grammatical unit m; the key grammatical unit is the clause, as shown diagrammatically in Figure 10-1. But while this is the foundation on which the relationship between semantics and lexicogrammar is based, there are two other principles affecting this relationship, making it more complicated but also extending the meaning potential of language: 

(i) transgrammatical semantic domains – domains of meaning extending across different grammatical units, and 

(ii) metaphor – incongruent realisational relations between semantics and lexicogrammar.

[6] This is very misleading, because it is very untrue. As the term 'transgrammatical semantic domains' suggests, it describes interstratal relations: 'semantic domains that range over more than a single grammatical unit' (ibid.)

David Rose Misrepresenting Halliday On Grammatical Metaphor

David Rose wrote to sys-func on 11/7/22 at 10:26:

This diagram is a good illustration of the grammatico-semantic model, with a semantic ‘figure’ realised by a grammatical clause and a semantic ‘participant’ by a grammatical nominal group. It must then explain grammatical metaphor by stratifying the semantic stratum into two ‘domains’, with the diagram showing a more abstract ‘Value’ realised by a less abstract ‘Token’.

In the discourse semantic model, stratification is between the strata of DS and LG. So a semantic figure may be realised ‘congruently’ by a clause, or metaphorically by a nominal group. This paper proposes the term ‘symbolising’ for the latter metaphorical type of relation between DS and LG.

Martin, J. R. (2020). Metaphors we feel by: stratal tension. Journal of World Languages, 6(1-2), 8-26.

Blogger Comments:

[1] To be clear, the diagram is Figure 6-16 from Halliday & Matthiessen (1999: 289):

[2] To be clear, the 'grammatico-semantic' model is Halliday's original stratification of the content plane.

[3] To be clear, here Rose misunderstands Figure 6-16. The diagram simply represents the semantics of the metaphorical wording engine failure as a congruent meaning (figure) realised by a metaphorical meaning (participant).

[4] This is misleading, because what Rose describes is actually Halliday's ('grammatico-semantic') model of grammatical metaphor. Halliday & Matthiessen (1999: 280):

The notion of a semantic figure originates in Halliday's model of ideational semantics (Halliday & Matthiessen 1999), and did not feature in Martin's model of discourse semantics (Martin 1992). The terms 'figure' and 'element' — from Halliday & Matthiessen (1999) — are assumed at times in Martin & Rose (2007), without either explanation or attribution, though neither term is identified as a unit in their model of ideational discourse semantics.

[5] To be clear, 'symbolising' is a poor choice of term for a subtype of realisation, if only because, in SFL Theory (Halliday & Matthiessen 2014: 269), realisation is a subtype of 'symbol'.

[6] To be clear, in order for 'stratal tension' to be a valid model of grammatical metaphor, the model must provide a systematic contrast between stratal tension and its absence. This is not provided by Martin's (1992) model of ideational discourse semantics, since it is a rebranding of lexical cohesion (textual lexicogrammar) and does not feature discourse semantic analogues of clause complex, clause and group — its only unit being the 'message part' (Martin 1992: 385).

David Kellogg On Rank, Axis And Metafunction

David Kellogg wrote to sys-func on 5/7/22 at 6:54:

Chris--your inventory of the architecture of SFL doesn't include RANK (or AXIS). Three possible reasons occur to me....

a) The Cardiff grammar is more strictly Firthian and works with "pieces" of language rather than with the clear hierarchy of units. We can actually see this in the diagramme because the arrow that connects "realization rules/statements" to "system networks" isn't realization or activation but rather "re-entry". So there actually isn't a rank scale. (This is also implicit in the rather vague expression "richly described tree structure", which my old syntax teacher Andy Radford would have heartily approved of....)

b) "Rank" is an organizing principle within Structure (just as System and Structure are principles within Axis). The inventory only includes the basic concepts from the IFG intro and not the points of origin, principles or meta-principles which order them. (It's an inventory and not an actual building!)

c) "Rank" isn't all pervasive in the way that, say, metafunction is. For example, Halliday and Matthiessen are agnostic about whether rank exists across the level of context or even semantics, but they have no such doubts about system, structure, instantiation, or metafunction. (On the other hand, H&M are agnostic about whether metafunction exists at the level of phonology and phonetics and I am much less so....)

Blogger Comments:

[1] To be clear, the architectural dimensions are those provided by Halliday & Matthiessen (2014: 20):

[2] To be clear, as dimensions, rank and axis are local, rather than global. Halliday & Matthiessen (2014: 32):

[3] For a close examination of Fawcett's theory of functional syntax, the Cardiff Grammar, see here.

[4] To be clear, paradigmatic and syntagmatic are orders of the local dimension axis; see [2].

[5] See [1].

[6] This is misleading, because it is untrue. Halliday & Matthiessen (2014: 87):

At the rank of tone group:

• the system of TONALITY — choice in the distribution of tone groups — realises the distribution of information units (textual metafunction);
• the system of TONICITY — choice in the placement of tonic prominence — realises the culmination of New information (textual metafunction);
• the system of TONE — choice in the major pitch movement — realises the system of KEY (interpersonal metafunction);
• the systems of TONE SEQUENCE and TONE CONCORD realise the systems of TAXIS and LOGICO-SEMANTIC TYPE (logical metafunction).

Mick O'Donnell Confusing SFL's Architecture Of Language With Models

After ChRIS CLÉiRIGh wrote to sys-func on 28/6/22 at 11:54:

When someone uses the architecture of SFL, they mean what the theoretical architecture means. Inconsistencies arise when the architecture is not understood.

Mick O'Donnell responded on sys-func on 28/6/22 at 19:37:

Your language implies that you believe that there is one architecture of SFL. I have always enjoyed the plural nature of SFL, with multiple alternative architectures to choose from.

To me, the biggest threat to SFL as a continuing school is the attitude that only the word of God (MAKH) is true, and everything else is heresy. …

Pluralism is good. Even if it involves different choices in fundamental architecture. Choice is good. What is not good is continual sniping at those who choose to differ from Halliday in details (but not in fundamentals).

Blogger Comments:

[1] To be clear, SFL does propose only one architecture for language. Halliday & Matthiessen (2014: 20) outline its dimensions as follows:

Importantly, all models proposed within SFL Theory use this architecture, including those of Martin and Fawcett. Importantly, because these dimensions are precisely specified, they can be used to identify models that misunderstand the architecture. For example, it can be seen that Fawcett's model (2010: 36) — which O'Donnell supports — misunderstands stratification and confuses axis with instantiation: 

[2] To be clear, here O'Donnell confuses the architecture with models using that architecture. The 'multiple alternatives' are the models, not the architecture.

[3] For me, "the biggest threat to SFL" is the inability of its community of users to understand the theory and, consequently, to detect misunderstandings of the theory.

[4] To be clear, here O'Donnell betrays his own religious orientation by projecting it onto others. Importantly, SFL Theory is a scientific theory, in the sense that its theoretical architecture of language is systematically organised and precisely specified. This means that models within the theory are subject to scientific criteria, not just chosen like the sacred text of a preferred religious sect. 

[5] To be clear, here O'Donnell reduces all the reasoned argumentation on this blog, using hard-won knowledge of SFL Theory, to "continual sniping'. The validity of the argumentation is of no interest to O'Donnell. The pedagogical value of blog is of no interest to O'Donnell.

[6] To be clear, the type of pluralism that O'Donnell advocates is not good. It exemplifies just the type of anti-intellectualism deplored by Asimov:

David Rose On Jim Martin's Self-Correction

David Rose wrote to sys-func on 28/6/22 at 9:01:

This is an opportunity to repeat a correction that JRM has repeated several times since 1992, that this para from ET erroneously equates his genre and register with MAKH’s context of culture and context of situation.

In JRM’s model, systems at the strata of genre, register and language are all instantiated as text, hence ‘all strata instantiate’. A text is at once a pattern of selections in genre, register and language (and other modalities). In this model, there is no situation without text.

Blogger Comments:

[1] This is misleading. Rose gives no references to any of the several corrections, and Genre Relations (Martin & Rose 2007: 10-1, 16) — 15 years after Martin (1992) — explicitly equates Martin's stratum of register to Halliday's instance of context (situation) and Martin's stratum of genre to Halliday's system of context (culture), with a realisation relation between system and instance:

See also Jim Martin On Context, Instantiation & Stratification

[2] To be clear, this is a serious inconsistency in Martin's model. Having distinguished language from context, his context is nevertheless instantiated as text (an instance of language, not context).

[3] The word 'hence' here is misleading, because it falsely implies a conclusion reasoned from premisses. Rose merely repeats the meaning of his previous clause as a generalisation. More importantly, 'all strata instantiate' is merely Martin's attempt to understand instantiation; cf. 'all strata make meaning' (semogenesis) as his attempt to understand stratification. Importantly, instantiation is a relation between system and instance, and a distinct dimension from stratification, so each can be described without reference to the other.

[4] This is misleading. The situation-text relation is a feature of SFL Theory: the realisation of context in language at the instance pole of the cline of instantiation. Martin's model, on the other hand, does not include the theoretical category of situation, except in its misconstrual as register; see [1] above.

David Rose Misrepresenting Ruqaiya Hasan's Research

David Rose replied to Geoff Williams on sys-func on 24/6/22 at 11:06: 

Could we describe as two of the most influential pedagogic applications of her research: One oriented to social analysis, using message semantics to demonstrate statistically the class basis of pedagogic codes, including your research in early literacy. And the other oriented to discourse analysis, with cohesion and schematic structure laying a foundation for discourse semantics and genre, that ultimately led to genre pedagogy?

Blogger Comments:

[1] Here Rose suggests that the value of Hasan's work is merely that it led to Martin's work. That is, for Rose, the value of Hasan's original research on lexicogrammatical cohesion and Generic Structure Potential is simply that Martin was able to rework it to derive his models of discourse semantics and genre pedagogy.

[2] This is misleading. To be clear, 'schematic structure' is Martin — e.g. Martin (1992: 505) — not Hasan. For Hasan's critique of Martin's 'schematic structure' — and genre — see her The Conception Of Context In Text (1995).

David Rose On Martin's Argument For Genre And Register

David Rose wrote to sys-func on 21/6/22 at 17:25:

Also in the spirit of lessening confusion (pace Chris), JRM’s 1992 argument for positioning genre ‘above’, configuring selections in field, mode and tenor, is long and very heteroglossic, but this little para presents one angle...

Overall it would appear that "rhetorical purpose" is the wild card in contextual description, being variously categorised under field (Halliday 1965), tenor (Gregory 1967), mode (Halliday 1978a, 1985/1989) and as a separate contextual variable in its own right (Firth 1950 — effects, Ure & Ellis 1977 — role, Fawcett 1980 — pragmatic purpose). The main reason for this is that purpose is difficult to associate with any one metafunctional component of the lexicogrammar or discourse semantics. The effect of a text is the result of all components of its meaning. This makes associating the notion of rhetorical purpose with Bakhtin's more global notion of speech genres an attractive one (cf. Gregory 1982).

H&H … apply the term register from the perspective of language, looking ‘up’ at field, mode and tenor. JRM applies it from the perspective of field, mode and tenor, looking ‘down’ at language (and other modalities). Hence register and genre theory.

Blogger Comments:

[1] Some other examples of Rose 'lessening confusion' can be viewed here.

[2] To be clear, heteroglossia does not guarantee a valid argument. Martin's argument for locating genre above field tenor and mode (misconstrued as register) is examined at

Why The Argument For Register And Genre As Context Strata Is Invalid 

Theoretical Inconsistencies In Modelling Genre And Register As Context Strata

Problems With The Non-Argument For Register As Context

Problems With The Non-Argument For Genre As Context

Problems With 'Genre As A Pattern Of Register Patterns'

Seven Problems With The First Justification For A Genre Stratum

Two Problems With The Second Justification For A Genre Stratum

Eight Problems With The Third Justification For A Genre Stratum

Two Problems With The Fourth Justification For A Genre Stratum

Two Problems With The Fifth Justification For A Genre Stratum

[3] This very paragraph (Martin 1992: 501) is closely examined at Confusing Context With Text Type.

[4] This is misleading, because it misrepresents the difference between Halliday & Hasan's view of register and Martin's view of register as a difference in (trinocular) perspective. It is not true that Halliday & Hasan "apply the term register from the perspective of language, looking ‘up’ at field, mode and tenor". Rather, Halliday & Hasan associate the contextual features of a situation type with the linguistic features of a register, which thus can viewed either 'down' from context or 'up' from language. Halliday & Hasan (1976: 22):

The linguistic features which are typically associated with a configuration of situational features — with particular values of the field, mode and tenor — constitute a register.

Martin, on the other hand, simply misapplies the term 'register' to the situational features instead of the linguistic features.

[5] The word 'hence' is misleading, because it gives the false impression that the validity of 'register and genre theory' has been supported by the text that precedes it.

David Rose Falsely Implying That Ruqaiya Hasan Endorsed Martin's Model

David Rose replied to Mick O'Donnell on sys-func on 21/6/22 at 13:54: 

Affiliationally (!), it’s flattering to be called the Martin-Rose model, but you’d have to add a lot more names with hyphens. JRM just tried to give me a leg up as a co-author, as a good teacher does (Halliday & Matthiessen 1999, 2004). It would be more accurate to call it the Hasan-Martin model, as so much of it is built on her work. ET mentions her 197 times, with 15 of her publications cited.

Blogger Comments:

[1] To be clear, the proportionalities here are

Rose : Martin ::

Matthiessen : Halliday

[2] This is very misleading indeed. What is true is that Martin rebranded Hasan's work on cohesion — principally Halliday & Hasan (1976) — as his own model of discourse semantics. What is not true is that Martin's derived model is consistent with Hasan's original work, and what is emphatically not true, is that Hasan endorsed Martin's misunderstandings of her work.

With regard to Martin's model that confuses language varieties with context, Hasan wrote a protracted piece that identified the problems with Martin's approach. See:

Hasan, R. 1995. The Conception of Context in Text. In P. H. Fries & M. Gregory (eds.), Discourse in Society: Systemic Functional Perspectives, Meaning and Choice in Language: Studies for Michael Halliday, 183-283. Norwood, New Jersey: Ablex.

David Rose On Inexplicit Realisations And Lexis

David Rose replied to Mick O'Donnell on sys-func on 21/6/22 at 12:26:

… So how do we model field elements that are not explicitly realised in a text? That is so complex that SFL has avoided trying to describe lexical items, beyond their very general relations, as ‘sets’ vs systems, as delicate grammar, and as lexical relations in discourse.

It’s long worried me, as I would like to describe how we learn to read (comprehend written texts), and that can’t be done without tackling lexis.

In this paper on reading metaphor I’ve suggested we technicalise the term connote for these kinds of relations.

Rose, D. (2021). Reading metaphor: Symbolising, connoting and abducing meanings. Linguistics and Education, 64, 100932.

Blogger Comments:

[1] To be clear, what Rose, following Martin, regards as 'field' is ideational semantics in SFL Theory. Halliday & Matthiessen (1999: 407):

… Martin, where"field" corresponds to what has been discussed here in terms of ideational semantic networks in the ideation base …

[2] To be clear, this is a matter of inexplicit realisations of semantic features being instantiated in a text.

[3] To be clear, this is not simply a matter of lexis. A lexical item is the synthetic realisation of the most delicate grammatical features, just as the phoneme /b/ is the synthetic realisation of the phonological features [voiced, bilabial, stop]. Modelling this phenomenon systemically entails identifying the system, with the choice of in/explicit realisation, that has been instantiated in the text (data).

[4] To be clear, lexis as most delicate grammar means that lexical items are specified by the most delicate grammatical features, and lexical sets are the paradigmatic organisation of the lexical items thus specified. 'Lexical relations in discourse', on the other hand, is the use of paradigmatic and syntagmatic relations between lexical items to create textual cohesion. In Rose's source, Martin (1992), the notion of lexis as most delicate grammar is confused with lexical cohesion; evidence here.

[5] To be clear, Rose does not understand the SFL notion of 'lexical item'. See, for example:

• Misconstruing (Abridged) Grammatical Structures As Lexical Items
• Mistaking (Partial) Grammatical Units For Lexical Items

[6] To be clear, Rose does not understand the SFL notion of 'metaphor'. See, for example:

• Foreshadowing Misunderstandings Of Grammatical Metaphor
• Reducing All Grammatical Metaphor To Elemental Metaphor
• Misunderstanding Ideational Metaphor
• Misrepresenting Ideational Metaphor
• Misrepresenting Elemental Ideational Metaphor As Logical Metaphor
• Misunderstanding Ideational Metaphor

David Rose Strategically Misrepresenting Karl Popper

David Rose replied to Mick O'Donnell on sys-func on 20/6/22 at 19:15

...well, it’s been 30 years so far since ET. But is resolution or refutation really necessary? A model is valid if it works, according to Popper and MAKH. Maybe we should just be looking at differences in appliability, and labelling models by those criteria, or by internal features that afford them.

and to Fran Christie at 21:49:

Keeping in mind that the lexicogrammar is at the base of every model. Then perhaps the grammatico-semantic model is good for large scale statistical analyses of corpora as RH pioneered, while the discourse semantic model is good for metafunctional analyses of text structuring.

Blogger Comments:

[1] This is misleading, because it is untrue. As a logician, Popper ascribed validity to arguments, not to scientific theories. Nor did Popper regard scientific theories as verifiable. Theories remain provisionally corroborated while experiments fail to falsify them. 

[2] Here Rose is actually deploying the logical fallacy that Popper drew specific attention to:

If A then P;

P;

Therefore A.

Popper's point is that, if theory A predicts phenomenon P, and phenomenon P is observed through experiment, this does not prove that theory A is true. That is, just as the appliability of the Ptolemaic model of the solar system did not prove that model true, the appliability of Martin's model does not prove his model "valid".

Importantly, Martin's stratification model is not falsifiable by experiment, and so, is not scientific in Popperian terms. Moreover, it is invalidated by its own internal contradictions, as demonstrated elsewhere on this blog and here.

[3] To be clear, the 'grammatico-semantic' model is Halliday's model of the semantics that the lexicogrammar construes. The discourse semantic model, on the other hand, is Martin's (1992) rebranding of Halliday's lexicogrammatical systems of cohesion and semantic system of speech function. Evidence here.