The POSITIONING System Network

Doran, Martin & Zappavigna (2025: 81-2):

Figure 3.3 brings together the options for rendering built in the previous chapter with those of tendering in this chapter to show the full system of POSITIONING.

ChatGPT Comments:

 Here are some theoretical tensions and confusions in this model that stand out:

1. Heterogeneous criteria of system delicacy
The network presents “POSITIONING” as a unified system, yet the delicacy expansions combine very different dimensions of meaning. The initial branching (render vs tender) sets up the primary contrast, but subsequent sub-systems shift criteria inconsistently:

• address vs note concerns interpersonal orientation versus commentary,

• placement (external vs internal) invokes discourse-structural positioning,

• proposition vs proposal introduces metafunctional clause type,

• open vs complete invokes structural polarity,

• purview introduces epistemic stance.
  What is presented as systemic delicacy is in fact a heterogeneous aggregation of distinctions that originate in different metafunctions and strata.

2. The tendering/rendering ambiguity
The diagram signals both a disjunctive and a conjunctive relationship between render and tender. This produces a conceptual inconsistency: are rendering and tendering mutually exclusive choices, or are they simultaneous dimensions of meaning? The attempt to code both at once collapses two theoretically distinct relations:

• paradigmatic opposition (choosing one or the other), and

• simultaneous activation (both may apply to a single instance).
  In practice, rendering and tendering often co-occur (e.g. “I think you’re wrong” both evaluates an existing stance and advances a counter-claim). The network’s simultaneous disjunctive/conjunctive representation fails to clarify how these dimensions intersect, instead producing analytical indeterminacy.

3. The purview problem
The addition of speaker purview versus listener purview exemplifies the model’s attempt to taxonomise epistemic stance as if it were a structural variable of the same order as mood or polarity. The resulting options (share, assert, pose, air) artificially discretise what is in practice a gradient of interpersonal alignment. More significantly, this construct duplicates resources already modelled in SFL (mood, modality, projection) without clarifying how these interact, leading to redundancy rather than theoretical integration.

4. Systemic incoherence
Although each sub-system is presented as if of the same systemic order (“rendering,” “placement,” “completion,” “purview”), they draw from divergent theoretical bases:

• speech function (proposition vs proposal),

• stance (support vs reject),

• logical completion (open vs complete),

• epistemicity (purview).
  The overall appearance of comprehensiveness masks the fact that the network is a collage of heterogeneous categories. The combination of inconsistent logics undermines systemic coherence.

5. Terminological inflation
By gathering these distinctions under the heading “POSITIONING,” the model creates the impression of a novel interpersonal system. Yet all the included options correspond to resources already theorised within canonical SFL: speech function, mood, modality, appraisal, and thematic organisation. The rebranding substitutes terminological novelty for theoretical clarity, risking obfuscation rather than refinement of the interpersonal system.

Tendering Confusions: Duplicating Semantics at the Context Plane

Doran, Martin & Zappavigna (2025: 80-1):

With this discussion of purview, we have completed the set of options for the system of POSITIONING. Leaving aside repositioning, Figure 3.2 brings together the network of TENDERING established in this chapter. This network says that when tendering, one may tender a proposition or a proposal and that this proposition or proposal may be open or complete, may have speaker purview or not, and may have listener purview or not.

ChatGPT Comments:

In Doran, Martin & Zappavigna (2025: 80–1), the authors present Figure 3.2 as a system network for tendering, summarising the options discussed throughout the chapter. The network is described as showing that a proposition or proposal may be open or complete, may have speaker purview or not, and may have listener purview or not. At first glance, this seems like a tidy synthesis of the chapter’s material. However, closer scrutiny reveals a number of theoretical and conceptual issues.

Duplicating semantic categories at the context plane

The first issue is that the network seems to duplicate categories that properly belong to the content plane—specifically, the semantics of SPEECH FUNCTION—at the level of context. Speaker and listener purview are treated as options in tenor, that is, the context plane, yet many of the distinctions the authors draw (e.g., complete vs open propositions, hedged vs asserted modalities) are realised in the semantics and grammar of the content plane. By elevating these distinctions to context, the network conflates the strata: the same semantic distinctions are effectively represented twice, once as grammatical/semantic realisations, and again as contextual “purview” options.

Ambiguity of ‘tendering’ as a system label

The term tendering itself is also problematic. Its scope and status within the network are unclear: is it meant to denote the act of proposing a proposition/proposal generally, or a particular mode of interpersonal negotiation in discourse? The authors do not explicitly distinguish between these readings, leaving the reader to infer whether tendering is an operational category, a functional label, or a meta-concept. This undercuts the explanatory power of the network and makes it difficult to see how it could be applied consistently.

Overextension of the network logic

Finally, by combining all options (open/complete, speaker/listener purview) in a single system network, the authors risk suggesting that these are discrete, combinable choices in every instance. In practice, however, many of these options are mutually constraining: open propositions are typically associated with lower speaker purview, for example. Treating them as fully independent variables, as the network implies, flattens the nuanced interplay of internal and external purview discussed earlier in the chapter.

Conclusion

While the network provides a visually compact summary of the chapter, it does so at the cost of theoretical clarity. It conflates semantic realisations with contextual options, obscures the status of the tendering category, and overstates the independence of different purview options. In short, it presents a tidy system that may look neat on paper but does little to illuminate the complex, stratified nature of interpersonal meaning in context.

Tendering: Semantic Features Networked As A System Of Context

Doran, Martin & Zappavigna (2025: 65):

At this point, we can present an initial network of different resources for tendering, as shown in Figure 3.1. This network says that when tendering, one may table a proposition or a proposal, and each of these may be either open or complete.

ChatGPT Comments:

This system network exemplifies a serious stratal confusion: the authors present a contextual system (tenor: tender) whose features—proposition vs proposal, open vs complete—are clearly semantic. This is analogous to proposing a grammatical system (e.g. mood) in which the features are phonological (e.g. tones). It inverts the logic of realisation by treating options that should be realised by context-sensitive semantics as if they were chosen within the context itself. Moreover, “tendering” is not a situational parameter but a semantic phase within an unfolding exchange—it belongs to the semantics, not the context. The attempt to redescribe interpersonal semantic distinctions as contextual systems not only muddies the metafunctional picture, but introduces a kind of phantom network: one that appears systemic but floats free of any coherent stratal grounding. The result is a mislocated reclassification of semantic options that ultimately undermines the integrity of both strata.

Context in Crisis: Theoretical Faultlines in Genre Modelling

Doran, Martin & Zappavigna (2025: 49, 53):

In terms of our understanding of the relation between genre and register (i.e., how our model of tenor fits in), the distinction between tendering and rendering offers a way of seeing key shifts in interpersonal meanings at each stage. … Indeed this distinction in tendering and rendering gives us a sense of the differences between the various persuasive genres. Discussions, as we have seen, will tend to tender opposed positions, and then render one with support and one with rejection. Expositions on the other hand, which give only one side of the argument, will tend to tender one position and render it with support via a series of tendered arguments. By contrast, challenges tender a position at the beginning of the text and immediately reject it, supporting this rejection with a series of tendered arguments.

By exploring rendering, we are also able to explore prosodies of interpersonal meaning that permeate texts, such as those that occur in story genres. … The distinction between tendering and rendering thus gives us a way of viewing genres interpersonally, in terms of how they negotiate social relations.

Reviewer Comments:

[1] This raises the problematic nature of the authors' model of genre and register, first articulated in Martin (1992). In that work, Martin (p495) proposed modelling context of culture as genre and context of situation as register, with genre as a higher stratum of context, realised by register as a lower level. The main problems with this framework are as follows:

1. It misconstrues the instantiation relation of potential to instance, between culture and situation, as a realisation relation between levels of symbolic abstraction.
2. It misconstrues a subpotential of language, register, as a system of contextual potential, and despite this 'register' being mapped onto an instance of context (situation).

Genre itself, however, is a confusion of different theoretical locations:

1. As a type of text, it corresponds to 'text type' — that is, to register viewed from the instance pole of the cline of instantiation.
2. As a taxonomy of culturally-valued uses of language (narratives, anecdotes etc.), it corresponds to the contextual system of (rhetorical) mode.
3. As structure (e.g. Thesis > Argument > Reinforcement), it is the semantic structure of text types that realise mode selections of a situation type.

ChatGPT Comments:

[1] This passage foregrounds a long-standing confusion in Martin’s theoretical architecture — namely, the stratification of genre above register as levels of context. In this model (Martin 1992: 495), genre is treated as context of culture, realised by register as context of situation. However, this misconstrues the instantiation relation between culture and situation (as potential to instance) as if it were a relation of realisation between strata — conflating ontological categories.

At the same time, register is also misconstrued — not as a subpotential of language (a theory of variation within language potential), but as if it were itself a contextual system. Despite this, register is paradoxically mapped onto the instance of context — the situation — leading to a collapsed mapping in which systemic variation is treated as structural environment. The confusion is threefold: (i) register is mislocated ontologically, (ii) genre is simultaneously a cultural taxonomy, a rhetorical mode, and a text type, and (iii) realisation is incorrectly imposed across what are in fact instantiation relations.

[2] Within this confused framework, the interpersonal functions of tendering and rendering — which are semantic-level choices realised in mood, modality, evaluation and so on — are then projected onto genre structures, as if these rhetorical stages are themselves shaped by interpersonal semantics. The authors don’t quite claim that genres tender and render, but rather that they are realised as different patterns of tendering and rendering at the register level. Yet this move flattens the relationship between semantics and text structure, mistaking textual prosody and argument structure (which are patterns of realisation over time) for the organisation of contextual genre types.

[3] This projection of interpersonal semantics into genre structure confuses system and structure: it treats meanings that arise from semantic selections in interaction as if they directly define genre classes, bypassing the register variables (field, tenor, mode) that would more appropriately model such meanings. The result is a genre typology grounded not in situation types but in imagined interpersonal trajectories, driven by the authors’ invented taxonomy of positioning (tendering, rendering, support, rejection), which is nowhere anchored in the stratified model they attempt to build on.

[4] Overall, the “distinction between tendering and rendering” does not offer a new way of “viewing genres interpersonally” so much as it exposes the structural weaknesses of a genre theory built on theoretical conflations. It reflects a persistent attempt to retrofit novel terminology (here, from their system of positioning) into a model of discourse that is already buckling under conceptual strain.

[5] Social relations (e.g. mother-child) are not negotiated; they are semiotically enacted. Meanings are negotiated.

Problems With The Second 'Rendering' System Network

Doran, Martin & Zappavigna (2025: 42-3):

Adding these choices into our system network gives us Figure 2.2. In this figure the joint square bracket and brace means and/or. That is, one may choose to render or tender or both. Thus this network says that in tenor, one may tender a position or render a position or both at the same time. If rendering, this may address or note the position (with the various subtypes described for Figure 2.1 above), and any of these renderings may be done externally or internally.

Reviewer Comments:

[1] To be clear, this network misrepresents a system of potential language options as a system of tenor, the interpersonal context that is realised by such language choices.

[2] To be clear, this network has no entry condition or realisation statements.

Problems With The First 'Rendering' System Network

Doran, Martin & Zappavigna (2025: 37):

We can pull together the discussion so far as the set of preliminary options for rendering shown in the system network in Figure 2.1. 

 

This network outlines that when speaking, we have the option of tendering positions or rendering them. If rendering, we can address what has previously been said by enacting a stance of some kind, or we can simply note it and not give away our feelings. If we address these meanings, we can either support or reject them, and we can either do this in a way that directly confers this support or rejection (what Knight (2010a) calls ‘communing’ and ‘condemning’ affiliation), or defers it through laughter. These resources allow us to negotiate meanings used to enact our social relations in a nuanced way.

Reviewer Comments:

To be clear, the authors present this network as a model of tenor, that is, of who is taking part in cultural terms. However, as the authors themselves acknowledge, this network outlines options when speaking, which means that it is a model of language, not tenor, since tenor is context, not language. As this entire book makes plain, the authors do not understand the distinction between context and language, nor the realisation relation between them.

Moreover, the feature 'render' is the authors' rebranding of 'respond' in the semantic system of SPEECH FUNCTION, with 'support' a rebranding of 'expected' and 'reject' a rebranding of 'discretionary'. Halliday & Matthiessen (2014: 136):

Of the other features in the authors' network, Figure 2.1, 'note' and 'defer' refer to responses that do not realise speech functions:

• 'note' refers to a backchannelling response, realised by protolanguage (Mm);
• 'defer' refers to laughing in response.

The Problem With The System Of Positioning

Doran, Martin & Zappavigna (2025: 23):

How we put forward positions and react to them is described in Chapters 2 and 3 within a system called POSITIONING. Chapter 2 focuses on resources for rendering meanings and Chapter 3 focuses on resources for tendering. 

When we tender meanings, we will describe different ways in which we can position others to respond. For example, when Jodie’s mother asks at the beginning of the conversation ‘What are you on about Jodie?’, she asks this as a genuine question. With this question she positions Jodie as the one who has the knowledge for this exchange. 

By contrast, later on in the conversation, Jodie notes that her not being allowed to go to the pub contrasts with her friend Billie, who is also six, getting to go to the pub. Her mother notes that this is because it was her Daddy’s birthday. But Jodie is not convinced and insists by saying ‘and did I get to go’. Although this is a question and grammatically an interrogative, Jodie is not genuinely asking for information – both Jodie and her mother know full well that she was not allowed to go to the pub. Rather, Jodie is using this to emphasise her point that she thinks there is a double standard at play. 

 Reviewer Comments:

[1] As previously demonstrated, tendering and rendering are moves in an exchange that realise the system of SPEECH FUNCTION, and so are interpersonal potential at the level of semantics, not at the level of context (tenor). The system of POSITIONING is thus interpersonal semantics misunderstood as tenor.

[2] To be clear, in interpreting the meanings of the exchange in terms of the knowledge of the interlocutors, the authors have adopted a cognitive perspective on language. However, this is inconsistent with SFL, the theory the authors are concerned with expanding. To be clear, in taking a 'language-based approach to cognition', SFL models 'knowledge' as meaning. Halliday & Matthiessen (1999: ix-x):

It seems to us that our dialogue is relevant to current debates in cognitive science. In one sense, we are offering it as an alternative to mainstream currents in this area, since we are saying that cognition "is" (that is, can most profitably be modelled as) not thinking but meaning: the "mental" map is in fact a semiotic map, and "cognition" is just a way of talking about language. In modelling knowledge as meaning, we are treating it as a linguistic construct: hence, as something that is construed in the lexicogrammar. Instead of explaining language by reference to cognitive processes, we explain cognition by reference to linguistic processes.

Misunderstanding Language (Circumstance) As Field (Property)

Doran, Martin & Zappavigna (2025: 20-1):

Turning to properties, these can be qualitative (e.g., he served fast) or spatio-temporal (e.g., he served to his opponent’s backhand). There is also the option of arraying properties (e.g., he served faster in the first set) and if arrayed, then gauged (e.g., he served on average at 137 km/h). These choices are outlined in Figure 1.10.

 

Reviewer Comments:

[1] To be clear, this misconstrues language as context. The "qualitative property" fast is a Manner circumstance, and the "spatiotemporal property" to his opponent’s backhand is a Location circumstance.

[2] Again, this misconstrues language as context. The "property" faster is "arrayed" by the Location circumstance in the first set.

[3] Again, this misconstrues language as context. The "gauge" on average at 137 km/h is the Comment Adjunct on average plus the circumstantial Adjunct of Extent at 137 km/h.

Misunderstanding Language As Dynamic And Static Perspectives On Field

Doran, Martin & Zappavigna (2025: 19-20):

For both the dynamic and static perspective there are further options in play (Figure 1.9).

An activity can be presented as a single event (an ‘unmomented activity’, e.g., Alcaraz won the match) or it can be divided into moments of that activity (a ‘momented’ activity, e.g., Alcaraz won the first set, Rudd won the second, Alcaraz won the third and also won the fourth). If momented, the relationship can be one of implication where the unfolding activities are related by contingency where one necessarily follows another (e.g., Djokovic hit the line judge with the ball between points and so was automatically defaulted); alternatively they can be related by expectancy, where the unfolding is not born of necessity, but is probabilistic or expectant (as in the 2022 US Open final example above – Alcaraz won the first set, Rudd won the second, Alcaraz won the third and also won the fourth). 

For all these possibilities there is also the option of construing an activity as cyclical (e.g., players play the Australian Open, then the French Open, then Wimbledon, then the US Open, before again playing the Australian Open...); or the activity can be linear, and if linear then culminative (e.g., the serve hit the back wall) or unending (e.g., Tennis is played around the world). 

Turning to a static perspective, an item can be noted individually (e.g., ace) or taxonomised, via classification (e.g., his second serve was an ace) or via composition (e.g., my first racquet was made of wood and natural gut).

Reviewer Comments:

[1] To be clear, this misconstrues language as context. An "unmomented activity" is a clause realising a figure, and a "momented activity" is a paratactically extending clause complex realising a sequence. In Martin (1992), such activity sequences are located in field, but in Martin & Rose (2007), they are located in experiential discourse semantics.

[2] This again misconstrues language as context. This is a paratactically enhancing clause complex realising a sequence. The relation between the clauses is cause, not contingency.

[3] This again misconstrues language as context. This is a paratactically enhancing clause complex realising a sequence. The paratactic and hypotactic relations between the clauses are temporal.

[4] This again misconstrues language as context. The "culminative" and "unending" linear activities are both clauses realising figures.

[5] This again misconstrues language as context. The individual item is a lexical item, the classification taxonomy is a set of hyponymically related lexical items, and the composition taxonomy is a set of meronymically related lexical items.

Why Mass, Presence, And Association Cannot Be Principles Of Instantiation

Doran, Martin & Zappavigna (2025: 18):

Instantiation is typically considered in terms of the coupling of choices as conditioned by varying probabilities in different texts, text types and diatypes. 

But recognising the concepts of mass, presence, and association (i.e., status/contact) as metafunctionally diversified syndromes of choice offers a perspective on instantiation that can help explain why certain choices co-occur in certain situations. That is, mass, presence and association can be considered principles of instantiation – principles underpinning the co-selection and arrangement of features across strata and metafunctions. This can help us move away from relatively ad hoc explanations as to why particular choices are taken up in particular texts or situations and move us into considering different texts and the various domains they enter into in terms of a multidimensional set of principles. 

For example, scientific writing tends to involve significant interlocking networks of activity, taxonomy and property, but relatively little evaluative language in comparison to other disciplines (e.g. Halliday & Martin, 1993). We could explain this in terms of science’s aim for very strong ideational mass (technicality) but relatively weak interpersonal mass (iconisation). We could also describe the fact that it regularly aims to link theory to data as illustrating a wide range of ideational presence (iconicity).

Reviewer Comments:

[1] Here the authors misrepresent their own misunderstanding of instantiation as the "typical" view. As a process, instantiation is the selection of features and the activation of their realisation statements in logogenesis. At the system pole of the cline of instantiation, every feature of system has a probability of instantiation relative to other features in the system, and at the instance pole, texts vary by the relative frequency of feature selection.

The midway point of variation on the cline can be viewed from the system pole as register ("diatype"), or from the instance pole as text type, with registers varying in terms of the probability of feature selection, and text types varying in terms of the frequency of feature selection.

Importantly, at ontogenetic and phylogenetic timescales, the probabilities of feature selection in the system are altered by the frequencies of feature selection in the instances.

[2] To be clear, even if mass, presence and association provided a theoretically valid model of context, they could not "be considered as principles underpinning the co-selection and arrangement of features across strata and metafunctions" simply because instantiation is not an inter-stratal relation. Instantiation is the relation between system and instance at each stratum. There is no instantiation relation between context and language. Context and language are related by realisation, elaborating identity, so there is no causal (enhancing) relation between them.

[3] This is misleading, because this is simply the imposition of terms that do not provide explanation. That is,

• it does not explain why science writing has less evaluation, it just labels it as strongly technical;
• it does not explain why science writing has less evaluation, it just labels it as weakly iconised.

[4] As demonstrated here, the authors' notion of iconicity misunderstands a congruent relation between ideational lexicogrammar and ideational semantics as an iconic relation between the ideational content of language (text) and the ideational dimension of context ("what it is talking about").

Seriously Misunderstanding Stratification

Doran, Martin & Zappavigna (2025: 17, 10):

Realisation is a hierarchy of abstraction, with higher strata realised by patterns of meaning at lower ones. … The realisation hierarchy we assume here was outlined in Figure 1.3 above (with genre as a more abstract pattern of register patterns, register as a more abstract pattern of discourse semantic patterns and so on).

 Blogger Comments:

This seriously misunderstands stratification. Less importantly, realisation is not a hierarchy, but the relation between levels of symbolic abstraction, as between axes or between strata. The hierarchy in this case is the hierarchy of strata: stratification.

Most importantly, higher strata are not "realised by patterns of meaning at lower ones". On the one hand, the function of the stratification hierarchy is to relate meaning, semantics, to other levels of symbolic abstraction: lexicogrammar and phonology below, and the context of language above. The authors' misunderstanding reflects Martin's mantra 'all strata make meaning' which confuses stratification (realising meaning) with semogenesis (making meaning).

On the other hand, the "patterns" of each stratum are the systems of each stratum. The term "patterns" suggests a confusion with instantiation: the patterns of instantiation that distinguish the variants on the cline of instantiation from each other.

Misrepresenting The Metafunctions

 Doran, Martin & Zappavigna (2025: 10, 11):

The other main dimension of analysis we need to introduce here is metafunction. As developed by Halliday (e.g. 1973, 1978) this dimension refers to the different kinds of meaning that language can make — ideational resources for construing our world, interpersonal resources for enacting social relations and textual resources for composing information flow (Figure 1.4). 

More technically, metafunction refers to the tendency for linguistic systems to bundle together, with some sets of systems highly interdependent on each other, and others highly independent of each other. For grammar, this is reflected in the fact that sets of options within the ideational system of TRANSITIVITY are relatively interdependent on each other, but they are relatively independent of those within the interpersonal system of MOOD or the textual system of THEME (Martin & Matruglio, 2013/2020; Matthiessen, 1995).

Reviewer Comments:

This is misleading. To be clear, it is the principles underlying the metafunctions that "bundle" linguistic systems, not the reverse. The metafunctions give rise to the systems, not the reverse. The systems are devised according to the metafunctions, not the reverse.

ChatGPT Comments:

You're absolutely right to challenge the reversal of theoretical dependency in this excerpt. The authors misrepresent the metafunctions as emergent properties of the systems themselves, rather than as organising principles that motivate the construction of those systems. Here’s an extended critique you could add to your review, in line with your original comment but deepening the theoretical clarification:

---

Misrepresenting Metafunction as Emergent from Systemic Groupings

The authors’ characterisation of metafunction as a “tendency for linguistic systems to bundle together” misrepresents a foundational concept in Systemic Functional Linguistics. This account mistakenly presents metafunction as an emergent property of patterns within the system network — as if metafunction were something inferred from systemic interdependence, rather than the organising principle according to which those systems were constructed in the first place.

In the SFL model, metafunctions are axiomatic: they represent the fundamental types of meaning that language is evolved to enact — ideational, interpersonal, and textual. These are not discovered by observing bundles of grammatical systems; they are the theoretical starting point from which those systems are modelled and interpreted. The systems of transitivity, mood, and theme, for example, are organised according to their metafunctional roles. It is not the case that we discover the metafunctions by observing which systems “bundle”; rather, we posit metafunctional orientations and then describe systems in terms of the kinds of meaning they realise.

To say, then, that metafunction refers to "the tendency for linguistic systems to bundle together" reverses this theoretical dependency. It obscures the fact that SFL system networks are metafunctionally motivated: system organisation is guided by the distinct functions that language must serve in social life — to represent experience, enact social relations, and manage textual flow. This confusion risks eroding the explanatory power of metafunctional organisation and could easily mislead readers into thinking metafunctions are merely taxonomic or descriptive generalisations.

A more accurate formulation would be:

In SFL theory, metafunction refers to the three fundamental orientations of meaning that language is evolved to make — ideational, interpersonal, and textual — which in turn provide the organising principles for constructing grammatical systems. Each metafunction is associated with its own network of systems (e.g. transitivity, mood, theme), and the principle of metafunctional complementarity ensures that clauses simultaneously realise meanings from all three metafunctions in a stratified and integrated structure.

Until this theoretical directionality is clarified — that metafunctions give rise to systems, not the reverse — the account presented in this section risks misconstruing one of the most essential dimensions of the SFL framework.