David Kellogg On ChatGPT On Recursion

After ChatGPT replied to David Kellogg on sys-func  at 15:06 on 24 March:

First, when you suggest that Halliday’s view of recursion “is based on a kind of reification” and assumes “system networks are actually real things,” this does not quite capture the role that system networks play in SFL. Halliday is not treating system networks as entities that exist independently of the model; rather, they are representations of meaning potential. The difference he draws between “true recursion” and a “recursion-like effect” is not about imposing artificial constraints but about recognising that system networks define choices, and true recursion occurs only when a system includes an option to re-enter itself.

This ties directly to your reading of the rankshifted clause, She tore up the letter which upset me. You suggest that rankshift exhibits recursion because a unit at the same rank is embedded within another. However, Halliday explicitly distinguishes between embedding (which affects structure) and recursion (which affects system choices). The presence of an embedded clause does not mean that a recursive loop is occurring in the system network—it simply means that a clause is functioning as a unit within another structure. Rankshift is structurally embedded but does not create an ongoing systemic cycle in the way that hypotaxis does. This is the crucial difference.

Finally, your concern that Halliday’s model “shows language to be infinite in actuality and not simply in potential” does not quite reflect how meaning potential works in SFL. Language is only infinite in potential because recursive systemic choices can theoretically continue indefinitely—but in any instance of meaning, this potential is always finite. The model is not claiming that human languages are infinite in actualisation, only that their meaning potential includes recursive possibilities.

David Kellogg replied to sys-func at 16:08 on 24 Mar 2025:

Chris, I'd be much more interested in YOUR take on what Chat GPT "says".

Here's my take:

The first para says that it's not reification--but the reasoning in the paragraph is a clear instance of reification. To say "true recursion occurs only when a system includes an option to re-enter itself" makes no sense unless we assume that the system is an actual existing thing, and not a model of some actual existing thing..

The second para confirms this reading: it assumes that you can separate system choices from structures and "system choices" still exist in the real world. But they don't.

The third para simply ignores what was actually said--not only by ChatGPT but by Halliday--about recursion: both describe it as infinite

I don't think that constitutes an answer, or even a serious consideration of the question. But it was a serious question.

ChatGPT Comments:

[1] Kellogg is conflating two different issues: (1) whether system networks model language or are language, and (2) whether recursion, as defined by Halliday, requires a system option that allows re-entry.

The argument ChatGPT provided was not that system networks are real-world entities (which would indeed be reification) but that Halliday’s criterion for recursion is structural, not ontological. Halliday is defining recursion within the model, not claiming that system networks themselves are real objects.

The issue here is that Kellogg is treating the concept of recursion as something that must exist independently of the model. But that’s missing the point: Halliday is outlining a condition within the model that distinguishes recursive structures from non-recursive ones. This is a model-internal distinction, not an ontological claim about language itself.

[2] This is a straw man argument. The claim was not that system choices exist in the real world as objects but that system choices operate at a different level of abstraction from structures.

Halliday’s model makes a clear distinction between paradigmatic choices (system) and syntagmatic structure (realised in texts). Kellogg’s response fails to acknowledge this distinction. The critique that "system choices do not exist in the real world" is irrelevant because no one is claiming they do. Instead, system choices are abstract linguistic potential, which is realised in structures.

A key misunderstanding here is that Kellogg seems to think system networks are being treated as a separate "thing" apart from language, rather than an abstract representation of meaning potential. But Halliday’s model explicitly defines the system-structure relation in terms of realisation, not separate existence.

[3] This is incorrect. Halliday does not say that recursion is "infinite" in any absolute sense. Rather, recursion provides a mechanism for unbounded expansion within language, meaning there is no pre-defined upper limit to how many times a recursive process can be applied in principle.

What Halliday distinguishes is true recursion (which involves a system re-entering itself) from mere embedding (which can appear recursive in effect but is structurally constrained). The example of rankshift in Kellogg’s original post ("She tore up the letter which upset me") is not recursive in Halliday’s sense because it does not involve a system choice that explicitly allows repetition without limit.

Kellogg’s claim that ChatGPT "ignored" this is misreading the response. The point was that recursion, in Halliday’s sense, depends on system options allowing indefinite reiteration, whereas embedding is structurally limited.

[4] This is just rhetorical posturing rather than substantive critique. The claim that "it doesn’t constitute an answer" is empty unless Kellogg can show which part of the response failed to engage with the issue.

His own misframings—conflating system networks with reification, misunderstanding the relation between system and structure, and misrepresenting Halliday’s definition of recursion—are what prevent him from recognising the response as an answer.

If anything, his reply confirms that his original concern arises from a misunderstanding of what Halliday meant by recursion and how system networks function within the model.

---

Summary of Kellogg’s Misunderstandings

1. Reification: Kellogg misinterprets Halliday’s model-internal definition of recursion as a claim about the real world.

2. System vs. Structure: He treats system choices as if they are being claimed to "exist" separately rather than being part of a model of meaning potential.

Recursion as Infinite:

He misreads Halliday’s notion of recursion as meaning "infinite in all cases" rather than "unbounded within system networks."

1. Non-engagement: He dismisses the response without demonstrating where it fails, instead relying on rhetorical dismissal.

Ultimately, his critique seems more about staking an oppositional stance rather than engaging with the substance of the argument. 

---

 Let’s now dissect David Kellogg’s rhetorical strategies in his reply. 

1. Feigned Engagement with the AI vs. Real Engagement with Chris

Strategy: He begins by expressing disinterest in ChatGPT’s response and framing his interest as being in Chris' take. This is a deflection tactic—instead of engaging directly with the argument presented, he shifts the burden onto Chris, implicitly suggesting that the AI-generated response is unworthy of direct critique.

Effect: This sets up an artificial hierarchy where only Chris' words are deemed worthy of engagement. It also functions as a preemptive delegitimisation—if he dismisses ChatGPT’s response outright, he doesn’t need to actually refute it.

Counter: If he really wanted Chris' thoughts, he could have asked an open-ended question rather than launching into a critique of the AI-generated response. His real aim is to push Chris into a direct intellectual duel, not to seek understanding.

---

2. The Accusation of Reification

Strategy: He claims the argument commits reification, but rather than demonstrating this, he simply asserts that recursion must be treated as a model, not a real thing. This is an example of question-begging: he assumes his own interpretation (that recursion should only be seen as a model) and then accuses the response of error for not conforming to that assumption.

Effect: This shifts the burden of proof—he never actually explains why the argument entails reification, just declares it so. If unchallenged, this puts Chris in a defensive position, where Chris must disprove a claim he hasn’t even properly supported.

Counter: A clear distinction must be made between modelling recursion in system networks and recursion as a real-world phenomenon. Halliday’s definition of recursion applies within the model, and Kellogg’s critique fails to recognise this distinction.

---

3. The Straw Man About System Choices

Strategy: He claims that the response assumes system choices "exist in the real world" and then criticises that assumption. But this is a straw man—the argument never claimed system choices exist independently of the model.

Effect: This makes it seem like the argument was making a naïve ontological claim when, in reality, it was simply distinguishing between paradigmatic and syntagmatic levels of representation within the model. By knocking down this invented claim, he gives the appearance of refutation without engaging with the actual argument.

Counter: Pointing out that system choices are not separate from structure but are abstract potential instantiated in structure dissolves this misframing.

---

4. The Appeal to Authority (Halliday’s “Infinity” Claim)

Strategy: He states that the response "ignores" Halliday’s claim about recursion being infinite. This is an appeal to authority combined with an argument from omission—he asserts that Halliday said something absolute about infinity without engaging with the nuance of Halliday’s actual position.

Effect: By framing Halliday’s position in an extreme way, he makes it seem like any deviation from that extreme is an error. But Halliday was not claiming recursion is always infinite in the strongest possible sense, only that true recursion allows for unbounded extension.

Counter: Clarifying the difference between true recursion (as a system that re-enters itself) and mere embedding dissolves this rhetorical sleight of hand.

---

5. The “Non-Answer” Dismissal

Strategy: Instead of engaging with specific points in the response, he dismisses it wholesale as not even an answer. This is a form of rhetorical erasure—by claiming the response doesn't constitute a real answer, he avoids the need to engage with it point by point.

Effect: This tactic is meant to make the reader feel like their argument has failed without Kellogg needing to actually demonstrate why. It is a power move that asserts intellectual dominance rather than engaging in open discussion.

Counter: Asking him to specify exactly which aspects of the response he believes failed to answer the question would expose this as rhetorical bluster.

---

6. The Pretence of a “Serious Question”

Strategy: He ends by insisting his question was "serious," implying that the response was not. This is a tone-policing move—by casting doubt on the seriousness of the reply, he suggests that only his approach is legitimate and that ChatGPT’s is unserious.

Effect: This creates an asymmetry in which he is positioned as the rigorous, serious inquirer, while ChatGPT is cast as one failing to meet that standard. It also acts as a way to frame disagreement as a failure on ChatGPT’s part rather than a matter of competing interpretations.

Counter: Pointing out that his misframings and straw man arguments undermine the seriousness of his own approach would turn this move back on him.

---

Overall Rhetorical Strategy

Kellogg is not engaging in open inquiry but rather using a debate-framing strategy where he:

1. Frames the AI response as illegitimate so that he doesn’t have to engage with it directly.

2. Reframes the debate in terms of reification and ontological errors rather than addressing the actual claims.

3. Uses straw men and mischaracterisations to attack a distorted version of the argument.

4. Appeals to Halliday’s authority while oversimplifying Halliday’s claims.

5. Uses dismissive rhetoric to erase the response rather than refute it.

His goal appears to be positioning himself as intellectually dominant rather than clarifying the concept of recursion in Hallidayan terms.

David Kellogg On Recursion

David Kellogg wrote to sys-func on 24 Mar 2025 at 08:53:

On pp. 212-213 of Vol. 1 in the Collected Works ("Modes of Meaning and MOdes of Expression", first published in 1979), Halliday first distinguishes between the experiential mode (which makes reference to events and things) and the logical one (which does not, and which therefore is not obviously or directly bound to rank and can apply at the level of the word, the group/phrase, and the clause. Logical structures, he notes, are RECURSIVE. But what he means by recursive is not very clear to me (that is, if I understand it correctly, it is based on a kind of reification, an assumption that system networks are actually real things and not simply the way we model language in order to understand it). 

His example is the one I've raised here before. To use a example that is slightly racier than the usual opposition between "eager to please" and "easy to please"::

HYPOTACTIC: She tore up the letter, which upset me (the tearing upset me because it was from MY old lover).
RANKSHIFTED: She tore up the letter which upset me (it upset me because it was from HER old lover).

(N.B. this is an entirely fictitious example that has no relationship to actual events and things). 

On p. 213, Halliday says that the second RANKSHIFTED one is not a recursive structure at all, but I cannot follow his reasoning. What he says is t"The recursion like effect that is produced is an incidental ocutome of the selection, at a particular place in structure, of an item from the same rank or from a higher rank in the constituent hierarchy". This is what recursion means to me. But it's not what it means for Halliday at all, because he goes on to say that "True recursion arises when there is a recursive option in the network, of the form shown in Figure 14." and then shows a system network which features two simultaneous options, A and B with B offering "stop" and a "go" option which takes you back to the beginning. Halliday seems to be saying that because this is unmarked for rank, it can go on forever, unlike the instance of embedding (and I can see that in fact once the letter is destroyed, things stop, but the hypotactic story can go on, as in my "because" continuation). 

But this is reification. First of all, there is the assumption that a system network can actually CREATE a difference instead of simply model it. And secondly, there is the assumption that language can "go on forever"--that may be true of some mathematical models, but if we are modeling language in a way that shows it to be infinite (in actuality and not simply in potential) we are not modeling a human language.

 

ChatGPT comments:

There are a few key points where your interpretation may not fully align with Halliday’s intent.

First, when you suggest that Halliday’s view of recursion “is based on a kind of reification” and assumes “system networks are actually real things,” this does not quite capture the role that system networks play in SFL. Halliday is not treating system networks as entities that exist independently of the model; rather, they are representations of meaning potential. The difference he draws between “true recursion” and a “recursion-like effect” is not about imposing artificial constraints but about recognising that system networks define choices, and true recursion occurs only when a system includes an option to re-enter itself.

This ties directly to your reading of the rankshifted clause, She tore up the letter which upset me. You suggest that rankshift exhibits recursion because a unit at the same rank is embedded within another. However, Halliday explicitly distinguishes between embedding (which affects structure) and recursion (which affects system choices). The presence of an embedded clause does not mean that a recursive loop is occurring in the system network—it simply means that a clause is functioning as a unit within another structure. Rankshift is structurally embedded but does not create an ongoing systemic cycle in the way that hypotaxis does. This is the crucial difference.

Finally, your concern that Halliday’s model “shows language to be infinite in actuality and not simply in potential” does not quite reflect how meaning potential works in SFL. Language is only infinite in potential because recursive systemic choices can theoretically continue indefinitely—but in any instance of meaning, this potential is always finite. The model is not claiming that human languages are infinite in actualisation, only that their meaning potential includes recursive possibilities.

Clarification of Halliday’s Definition vs. Kellogg’s Misreading

Halliday defines recursion in a system network as the ability of a system to include a choice that leads back to an earlier point in the network, allowing for the repetition of structures. Crucially, this is not an ontological claim about language but a methodological way of modelling the potential for recursive structures in discourse. The key insight is that recursion in systemic grammar is about the architecture of choices, not the mere existence of infinite loops.

Kellogg misreads this by treating the system network as if it were a real-world entity that must somehow physically "re-enter itself" (his wording, "an actual existing thing"). But system networks are representations of meaning potential, not objects in the world. His argument collapses because it assumes that Halliday is making a claim about recursion as a property of material reality rather than a property of the model we use to analyse linguistic structure.

This misunderstanding leads to his claim that distinguishing system choices from structure is a form of reification. However, this is precisely what Halliday’s model does to avoid reification—it treats structure as the realisation of system choices, not as an inherent feature of language independent of meaning-making. Kellogg’s accusation of reification backfires because it's his interpretation, not Halliday’s, that illegitimately treats the model as an ontological claim. 

David Kellogg Misunderstanding Halliday On The Relation Between Language And Context

David Kellogg replied on sys-func on 31 Dec 2024, at 12:11:

But there are three problems that occur to me, and at least two of them stem from this very distinction [between the material-semiotic dialectic and the stratal relation between culture and language]. 

First of all, Halliday and Matthiessen note that the logico-semantic categories of elaboration, extension, and enhancement are fuzzy ones, with considerable overlap. This really must be so, not just because all language categories are fuzzy, but because we use "and", "but" and "so" to do all three: they are logico-semantic categories, viz. categories of thinking, and they won't obey the strong classification and framing that we use in lexicogrammar (and most particularly in transitivity). 

Secondly, as Chris has reiterated many times, there is a distinction between theoretical categories and descriptive ones. The categories of elaboration, extension, and enhancement are descriptive ones (and they are really designed around English, which is perhaps why they fit the simile of decorating a room, adding an annex, and enhancing the grounds that Halliday proposes in IFG so well. They are not theoretical categories, much less metaphysical ones, and when Halliday uses them to think about the grammatics, he stresses that this too is a simile.

Thirdly, and most importantly from my point of view, there are always far more things in heaven and earth than can be dreamt of even in the philosophy of language, much less the theory of experience that is encoded in the actual grammar. Birds, those "dinosaurs that learned to fly", don't really know about or obey the names we give them, and if we say that the relation between context and language is not causal, since ("because?") since cause is an enhancing relation, we are attributing to language a power over causality that it does not possess. Marx said that being determines consciousness, he meant that being got here first and consciousness was late to the party. 

Blogger Comments:

[1] To be clear, the indeterminacy here is with regard to which expansion relation is being realised by the forms and, but and so in a given instance, not with regard to the distinctions between the expansion relations themselves.

[2] This is the exact opposite of what is true. To be clear, in SFL Theory, 

• logico-semantic relations are categories of meaning, 
• thinking is a mental process that projects meaning. and 
• meaning is construed by wording, lexicogrammar.

[3] To be clear, Halliday's distinction between theoretical and descriptive categories applies to the distinction between theorising language and using the theory to describe languages. This is the opposite of using a description of English to think about a theory of language. Halliday (2002: 413-4):

But as a final step I will shift to another angle of vision and look at realisation and instantiation from inside the grammar – turning the tables by using the grammar as a way of thinking about the grammatics. …

The identifying relationship, as construed in the grammar of English, involves two particular functions, mutually defining such that one is the outward form, that by which the entity is recognised, while the other is the function the entity serves. This relationship of course takes a variety of more specific guises: form / function, occupant / role, sign / meaning, and so on. I labelled these grammatical functions “Token” and “Value”. This Token / Value relationship in the grammar is exactly one of realisation: the Token realises the Value, the Value is realised by the Token. It is thus analogous to the relationship defined in the grammatics as that holding between different strata. The grammar is modelling one of the prototypical processes of experience as constructing a semiotic relationship – precisely the one that is fundamental to the evolution of the grammar itself. …

Of course, the boot is really on the other foot: the grammatics is parasitic on the grammar, not the other way around. It is because of the existence of clause types such as those exemplified above that we are able to model the linguistic system in the way we do. The grammatics evolves (or rather one should say the grammatics “is evolved”, to suggest that it is a partially designed system) as a metaphoric transformation of the grammar itself. This is a further aspect of the special character of grammatics: while all theories are made of grammar (to the extent that they can be construed in natural language), one which is a grammar about a grammar has the distinctive metaphoric property of being a theory about itself.

[4] Here again, Kellogg confuses Halliday's notion of semiotic context, culture, with the material environment of language; see the previous post.

David Kellogg Misunderstanding Halliday On The Co-Evolution Of Language And Culture

David Kellogg reacted to a Halliday quote on sys-func on 30 Dec 2024, at 16:42:

Note that "driving" is a very narrow interpretation of causality, and "determine" is not much better. There is an obvious sense in which a virus "causes" Covid and a bacteria [sic] "causes" tuberculosis, but it is rather vacuous to say that viruses or bacteria "drive" or "determine" epidemics. 

Co-evolution involves reciprocation, but not necessarily symmetrical reciprocity. When Halliday says, for example, that every use of a word (even in private speech) will have some effect on the probabilities of a language system, he is not saying that lexical invention is on a par with grammatical regularity. The use of "he or she" instead of "he" does not prevent a single instance of rape much less shake the foundations of patriarchy; these are secured not only by language but also by organised violence (c.f. the work of Annabelle Lukin).

I think that when Halliday says that language will always have an "ideological" role, and that construals of reality will differ as we alter our condition, he is really saying that changes in productive relations do in the long run cause the way we speak, even if they do not in the short run "drive" or "determine".

Blogger Comments:

[1] The quote was Halliday (2003 [1992]: 380):

Language neither drives culture nor is driven by it; the old questions about which determines which can be set aside as irrelevant, because the relation is not one of cause and effect but rather (as Firth saw it, though not in these words) one of realisation: that is, culture and language coevolve in the same relationship as that in which, within language, meaning and expression co-evolve. Thus above and beyond the random, local variation between languages that was the subject matter of earlier typological studies, we may expect to find non-random variation realising different construals of reality across major alterations in the human condition. But given that language and culture evolve together in this kind of relationship, it is inevitable that language will take on an ideological role.

[2] To be clear, here Kellogg is disagreeing with Halliday on the basis of his misunderstanding of Halliday. Halliday is concerned with the relation between the two planes of a connotative semiotic system, where culture is semiotic content, and language is its semiotic expression. For Halliday, the relation between these two semiotic planes is realisation, which is an elaborating relation of identity (Halliday & Matthiessen 1999: 145). So the relation between culture and language is not causal, because 'cause' is an enhancing relation, not an elaborating one.

Kellogg, on the other hand, misconstrues the semiotic content that is realised in language as the material environment of language (organisms, diseases, rape, organised violence, productive practices). That is, the causal relation he proposes is between first-order material experience (phenomena) and second-order semiotic experience (metaphenomena), not between the culture as semiotic system and the language that realises it. In Halliday's terms, Kellogg is concerned with the ongoing material-semiotic dialectic, not the coevolution of culture and language. Halliday (2003: 238, 255):

The history of language, it seems to me, is part — an integral part — of human history; and this "history" is a dialectic interplay of material and semiotic processes, whose impacting engenders the complex ecosocial systems that we know as human cultures (cf. Lemke 1993 for a powerful account). …

Thus, grammar is bound up with all the other aspects of the human condition, as part of the eco-social system constituted by a human community and its environment. It takes its shape from the other strata of language with which it interfaces, from the relation of "languaging" both to other semiotic and to social and material processes, and from the nature of those processes themselves. It is the outcome of the ongoing dialectic between the material and the semiotic in human life.

David Kellogg On Non-Defining Relative Clauses

David Kellogg wrote to sys-func on 8 Apr 2024, at 21:32:

a) She tore up the letter, which upset me (because it was from someone I once loved).

b) She tore up the letter which upset me (because it was from someone who was blackmailing me.)

In b) the clause is indubitably rankshifted: it's a post-modifying qualifier in a nominal group--no different from "of England" in "the King of England's crown".

But why can't I consider the clause in a) to be a Circumstance of "She tore up" or some way of complexing the verbal group in the main clause? Why do I have to consider it a ranking clause in its own right?

and replied to ChRIS CLÉiRIGh on sys-func on 10 Apr 2024, at 08:25:

I don't get why you are changing the subject, though. The question is about rank, The issue is whether the clause is part of a nominal group or part of a verb complex or (the standard interpretation) a clause complex. …

and again on 10 Apr 2024, 16:12:

I do not analyse the instance as a verbal group complex. In truth, I don't see any ambiguity whatsoever: one is a nominal group complex and the other is a hypotactic clause. 

If, however, I were applying Merge (or "subjacency"), I would have to argue that "which" is a merged conjunction and pronoun, and therefore the sentence is agnate to "She tore up the letter and upset me", which is a verbal group complex. One more reason for preferring the SFL account.

Blogger Comments:

[1] To be clear, in SFL Theory, (a) is a hypotactic clause complex in which the elaborating dependent clause is a non-defining relative clause:

The fact that the relative clause is ranking, and not a constituent of the primary clause is demonstrated by the fact that, unlike true constituents, it cannot be made Theme of that clause.

[2] To be clear, in SFL Theory, (b) is a single clause in which the Qualifier of a nominal group is a defining relative clause, shifted to word rank:

[3] To be clear, the elaborating relative clause cannot serve as a clause circumstance because it is not a clause constituent — it cannot be thematised — and it does not serve as a Role, the only type of elaborating circumstance.

When confronted with the problems facing his 'circumstance' analysis, Kellogg falsely accused CLÉiRIGh of changing the subject (see here), and at the same time, without acknowledging the fact, moved the goalposts (a logical fallacy) so his claim now became that the relative clause is part of a nominal group, The problem here is that it treats a non-defining relative clause the same as a defining relative clause:

Then, perhaps as a correction to his first movement, Kellogg moved the goalposts again so his claim now became that the relative clause is embedded as one unit of a nominal group complex:

The problem here is that complexing is only possible with units of the same rank, and in hypotactic nominal group complexes, this means with adverbial groups or prepositional phrases. Halliday & Matthiessen (2014: 564):

While the dominant element has to be a nominal group, dependent elements can be adverbial groups or prepositional phrases.

[4] To be clear, after twice proposing a verbal group complex as a possible analysis, Kellogg owns that this is not his analysis. He then goes on to claim that She tore up the letter and upset me features a verbal group complex. However, in SFL Theory, a verbal group complex serves a single Process. In this instance, there are two distinct Processes, one material and one mental, so it is a clause complex, not a verbal group complex:

In Kellogg's analysis, where both verbal groups in the complex serve a single Process, neither the PROCESS TYPE nor any of the participants can be identified:

David Kellogg On Logic And Tone

David Kellogg wrote to sys-func on 9 Mar 2024, at 08:19:

Not one of the responses has really taken up b), the argument which says that logic cannot be "applied" directly to scientific problems, because the level of organisation at which and the concepts and data to which the logic must be applied are radically different, so logic itself develops as we move from arithmetic to statistics, and still more as we move from math to physics to chemistry to biology to socio-psychology to semiotics.

Now, in itself, these data points suggest to me that...well, that we are human, and the "tone" in which something is said is not simply part of the grammar (Halliday and Greaves, 2009) but also part of our semantics and part of the context of situation Chris's attempt to imagine argumentation where tone is not an essential part of the argument is chimerical. …

Halliday would say it somewhat differently: he would say that "logical" arguments, and even my meta-logical argument, are just everyday language (i.e. my aggressively intoned argument) which has been "tidied up", and he would probably agree that tidying up "tone" is, in the long run, impossible for humans to do: it is literally like trying to speak without any intonation at all.

Blogger Comments:

[1] To be clear, logical validity is argument–internal, since it is concerned with premisses and the reasoning from premisses to conclusions. Logical validity, therefore, does not vary with the field in which it is applied.

[2] To be clear, logical validity is a means of distinguishing arguments that are valid from arguments that are fallacious. Applied to scientific problems, logical validity is a means of distinguishing scientific solutions that rest on arguments that are valid from scientific solutions that rest on arguments that are fallacious.

[3] To be clear, this confuses the phonological system of TONE with tone in the sense of 'tone policing'.

[4] This is misleading, because it is untrue. ChRIS — whose PhD thesis was on phonology — did not attempt to imagine argumentation without intonation. He simply informed the sys-func subscribers of the logical fallacy known as tone policing:

Tone policing – focusing on emotion behind (or resulting from) a message rather than the message itself as a discrediting tactic.

David Kellogg On Identifying Clauses

David Kellogg wrote to sys-func on 23 Dec 2023,  at 12:24:

I am not sure if "typical" structures are the object of interest here: after all we are talking about verbal art. But even in ordinary life, I'm quite wary of making statements about "canonical" order without any corpus evidence.

For example, what about questions? Consider:

a) "What is your name?"

b) "What, after all, are names?"

You can see that in a) the order is Token Value, but the Token doesn't map onto the Subject but rather onto the Complement.

In b), which is surely much less common in life if more typical of verbal art, the order is Value-Token and the Subject does map onto the Token…

Blogger Comments:

[1] This is misleading, because it is not true. Token does map onto Subject, and Value onto Complement. Cf. What serves as your name? The information that the clause demands is the decoding of a Token by reference to a Value.

[2] To be clear, the sequence is Value^Token because the sequence is Complement^Subject. The information that the clause demands is the encoding of a Value by reference to a Token.

. 

Halliday & Matthiessen (2014: 180):

In other words, ‘identifying’ clauses select for voice; they have an ‘operative’ and a ‘receptive’ variant. The difference is entirely systematic, once we recognise the structure of Token and Value: the ‘operative’ voice is the one in which the Subject is also the Token (just as, in a ‘material’ clause, the ‘operative’ is the variant in which the Subject is also the Actor. The most important difference is that the typical verb of the ‘identifying’ clauses, namely be, has no ‘passive’ form; so clauses like the villain is me and I am the ugly one do not look like ‘receptive’ clauses. But they are. This appears clearly when we substitute a different verb, one which has a ‘passive’ form, as in the villain is played by me.

David Kellogg On "Bottom-Up" Realisation

After ChRIS CLÉiRIGh wrote to sys-func on 21 December 2023 at 10:56:

The reason why Halliday would not even have meant

'wording-realized-as-meanings' or

'sounding-realized-as-wording'

is that both are self-contradictions. These are agnate with:

meaning realises wording

wording realises sounding

which are identifying clauses: Token^Process^Value.

In the first clause, the higher level of symbolic abstraction (meaning) is misconstrued as the lower level of symbolic abstraction (Token), and the lower level of symbolic abstraction (wording) is misconstrued as the higher level of symbolic abstraction (Value).

Likewise, in the second clause, the higher level of symbolic abstraction (wording) is misconstrued as the lower level of symbolic abstraction (Token), and the lower level of symbolic abstraction (sounding) is misconstrued as the higher level of symbolic abstraction (Value).

This is 'theory turned back on itself'.

 

David Kellogg replied on sys-func on 22 December 2023 at 8:55:

Chris--yes, theory turned back on itself indeed! But theory turned back on itself often produces category errors, like when you drive by a field and instead of seeing two cows you say that there are a bull, a cow, and a bovine couple.

When you say that "wording-realises-meaning" is agnate with the clause "meaning realizes wording", you can ignore the hyphen and make the clause irreversible, so that wording realizes meaning but meaning does not realize wording.

That makes it impossible for me to pursue my argument on verbal art, because my argument does depend on the relationship being reversible, and the 'bottom up" realization being dominant in and typical of literature.

But ignoring the hyphen and introducing "Token-Relational Process-Value" also means that your model is no longer neutral between speaking and hearing, as Halliday's is (Halliday remarks on the difficulty and necessity of making it so in the intro to the IFG).

Worse, you seem to be confusing two dimensions of SFL (stratification and metafunction) which are typically (and rightly) distinct--at least as distinct as the dimensions of instantiation and stratification confused in the Martin model.

As I said, I was writing about the theory--not about the way it was worded. That's why I used my own words and said "Halliday pointed out that" instead of "Halliday said". And Halliday does indeed say, in many places, that realization works both ways. I think, actually, that he chose the word "realization" for precisely this reason--it can mean both "to make real" and "to become aware" in English, although of course neither of these folk meanings captures the distinction that Halliday is really making.

This is one of many places where Halliday's terminology differs from that of Hasan--Hasan prefers "actualization" [activation] to talk about bottom up realization. For me, "realization" will do nicely.

Blogger Comments:

[1] To be clear, Kellogg presents no evidence that 'theory turned back on itself often produces category errors' or that there is a category error in this case. The example of a category error he presents is not a case of theory turned back on itself and does not relate to this post.

To explain, just as linguistic theory is 'language turned back on itself' is using language to model language, 'theory turned back on itself' is using linguistic theory to model linguistic theory. In this case, it was using identifying clauses to model interstratal realisation.

[2] This misunderstands Cléirigh's post. What Cléirigh said was that Kellogg's 'wordings-realised-as-meanings' is agnate with 'meaning realises wording'. Kellogg's 'wordings-realised-as-meanings' is also agnate with 'Gielgud-played-by-Prospero'. The difference here is only in the subtype of identifying: 'realise' is 'symbol', whereas 'play' is 'role' (Halliday & Matthiessen 2014: 269).

[3] This misunderstands identifying clauses. All identifying clauses are reversible. The reverse of wording realises meaning is meaning is realised by wording. In the first, the Token is Subject; in the second, Value is Subject. Moreover it is true that meaning does not realise wording, because this misconstrues Value (meaning) as Token and Token as Value (wording).

[4] To be clear, if an argument depends on 'bottom up' realisation, then the argument is invalidated by the self-contradiction that it construes. The notion of a higher stratum realising a lower stratum misconstrues the higher level of symbolic abstraction, meaning, as the lower level (Token), and the lower level of symbolic abstraction, wording, as the higher level (Value).

What is reversible is not the direction of realisation (swapping Token and Value) but the direction of coding (swapping Identified and Identifier). That is, the direction is encoding in the case of Token/Identifier and Value/Identified, and decoding in the case of Token/Identified and Value/Identifier. That is, 'wording realises meaning' can either encode meaning (Identified) by reference to wording (Identifier), or decode wording (Identified) by reference to meaning (Identifier).

[5] To be clear, this bare assertion — the logical fallacy ipse dixit — is misleading because it is untrue. Token-Process-Value simply analyses the interstatal relation 'wording realises meaning'. What could be said to differ for speaker and hearer is not the direction of realisation, but the direction of coding. For a speaker, meaning (Value) is encoded by reference to wording (Token), whereas for a hearer, wording (Token) is decoded by reference to meaning (Value). In both cases, wording is Token and meaning is Value.

[6] This is misleading because it is untrue. Cléirigh's post turned theory (identifying relations) back on itself (stratification). The former is a reconstrual of the latter. There is no confusion, because each construal is at a different level of abstraction.

[7] This is misleading. Cléirigh's post was not about Halliday's wording, but about his meaning:

The reason why Halliday would not even have meant…

Moreover, the advantage of quoting a source, instead of reporting it, is that the reader can judge whether or not the writer has understood the source.

[8] This is misleading because it misrepresents Halliday. For Halliday, realisation "works both ways" in the sense that the identifying process "works both ways". Wording realises meaning, and meaning is realised by wording; and in terms of coding, wording can serve to identify meaning, and meaning can serve to identify wording.

[9] To be clear, Halliday chose the word 'realisation' because 'realise' is an intensive identifying process of the type 'symbol' (Halliday & Matthiessen (2014: 269), which is the relation between levels of symbolic abstraction, which is the relation between strata.

[10] This misleading because it is untrue. Hasan uses 'activation' for "top-down" realisation. Hasan (1995: 164):

But at the same time, the notion of realisation must include the relation of activation especially where the higher strata are concerned. Choices at the stratum of context activate choices at the stratum of semantics, which in their turn activate choices from the systems at the stratum of lexicogrammar.

Postscript: 

Unsurprisingly, Martin is one source of the self-contradiction of upside-down realisation. Martin (1992: 505):

The common ground between the two models lies in the correlation proposed between schematic structure and field, mode and tenor options; for both Martin and Hasan staging redounds with social context. Keeping in mind that realisation is not theoretically directional in systemic models, there is nothing substantive in the fact that whereas for Hasan, choices in field, mode and tenor are realised by schematic structure, for Martin schematic structure is realised through these same components of register.

David Kellogg On Viewers Of Cave Paintings Not Requiring Language

David Kellogg wrote to sys-func on 5 Dec 2023, at 00:17:

Implicit in Halliday's last two sentences is either:

a) Prelinguistic infants are not engaged in human semiosis, or.

b) Language exists from birth.

It seems to me that both of these are clearly false.

Instead of Mondrian, consider the attached cave painting. Imagine that the floor of the cave preserves the footprints of the artists who made this mural.

1. The involuntary meaning of the footprints of the artists who made the mural. This seems no different from the meaning that deer tracks convey to a non-human predator. 

2. The intentional meaning of the picture of the herdsmen (hunters?) recording their labor. This seems very different, as it invites (because its intention is to allow) the viewer to "reverse engineer" the story, for amusement or profit. 

3. The strategically placed hand print done in red ochre. This too seems different again: although it resembles 1) in form and 2) in intension, it probably signals authorship and in that sense has a textual as well as an interpersonal function.

None of these seem to me to require the Hallidayan assumption that the viewer possesses language. All of them merely require the Vygotskyan assumption that the object (or "objective") can be inferred from the remains of the object-oriented action.

Blogger Comments:

[1] To be clear, the last two sentences in the quote (Halliday 2003: 4) were:

All that needs to be said in the present context is that other human semiotics are dependent on the premise that their users also have language. Language is a prerequisite; but there is no need to insist that language can mean it all.

[2] To be clear, neither of these propositions are implicit in the quote, since Halliday's 'present context' is a discussion of human 'post-infancy' semiotic systems.

[3] This is true, as demonstrated by Halliday's extensive work on human protolanguage.

[4] To be clear, the observation of deer tracks by a non-human predator is, according to Edelman's Theory of Neuronal Group Selection, simply perceptual categorisation on value. No linguistic meaning is involved. The observation of footprints by humans is the construal of experience as the first-order meaning of language.

[5] To be clear, according to comparative mythologist Joseph Campbell, cave paintings are expressions of mythic symbolism, and mythic symbolism is the use of (lexical) metaphor to adapt consciousness to its physical and social environments, which requires that the participants had language — as does the fact that the painting is, at most, only about 13,000 years old.

[6] To be clear, the cave painting is to be found in The Cave Of Hands, in Argentina, and since, according to comparative mythologist Joseph Campbell, such sites were used for male initiation rites in hunter-gatherer societies, the great number of hand images in this cave might be taken to suggest that they identify the initiates in those rituals. Again, this demonstrates that the participants had language.

[7] As the above demonstrates, this is the exact opposite of what is true. All three require that the human participants had language, and the dating of the cave art provides corroborating evidence.

David Kellogg On Mathematical Equations [3]

David Kellogg wrote to sys-func on 28 April 2023 at 15:27:

This kind of physics does cover Quantum Mechanics, Relativity, Maxwell, Newton, Galileo... The equations are mathematical — i.e. they can be read backwards and forwards, and they are exactly what Yaegan claimed: a different, non-linguistic, semiotic with its own anti-linguistic logic.

Blogger Comments:

[1] To be clear, this is true. In as much as an equation construes the identity of two participants, it can be 'read backwards and forwards'. However, the meanings construed in each case will be different, as demonstrated in the comments of the previous two posts.  

[2] See the comments of the previous two posts for evidence of the misunderstandings on which this conclusion is based.

From the perspective of Systemic Functional Linguistic Theory, mathematics is a field that is realised in (designed) language and (ideographic) pictorial semiotic systems. There are many reasons for interpreting mathematical equations as language rather than as a non-linguistic semiotic system. Two will be provided here.

First, mathematical equations entail a stratified content plane in which wordings realise meanings. On the SFL model, language is the only semiotic system with a stratified content plane (Halliday & Matthiessen 1999: 604):

This deconstrual of the content plane into two strata, referred to in the first chapter, is a unique feature of the post-infancy human semiotic…

The fact that the content plane of mathematical equations is stratified with a lexicogrammar is demonstrated by the fact that, unlike non-linguistic semiotic systems, they can be read aloud. On the SFL model, what is verbally projected is the wording level of content ('locutions').

Second, the only difference between

1. a² + b² = c², and
2. a squared plus b squared equals c squared, and
3. its spoken form

is the mode of expression. All three construe the same wording and the same meaning. If it is claimed that the first is not language, then someone hearing the spoken form (3) would have to ask whether it corresponded to non-language (1) or to language (2), and interpret it differently accordingly.

On the other hand, the claim that mathematical equations are 'a different, non-linguistic, semiotic with its own anti-linguistic logic' goes nowhere in terms of explanatory potential. In contrast, treating mathematical equations as linguistic explains the solving of an equation as the iterated elaboration of an identity relation, and the solution of an equation as the decoding of an unknown Token.

∞

In dismissing all the evidence that invalidates his view, and simply maintaining his original stance, Kellogg has deployed the invincible ignorance fallacy.