On The Questions And Possibilities That Martin’s Approach Opens Up

Bateman (1998: 25):

It also requires, as with many things, an appropriate attitude when working with the book: English Text is clearly at its best when engaged with, rather than considered passively as a (potential) source of wisdom. Certainly, one of the most important aspects of the book is exactly the kinds of questions that it raises. When an attitude is adopted that foregrounds the questions and possibilities Martin’s approach opens up, the results are almost always beneficial.

Blogger Comments:

As demonstrated on this blog, and in more detail here, the kinds of questions and possibilities that Martin's approach in English Text opens up arise from deep misunderstandings of the global dimensions of SFL theory: stratification (evidence here), instantiation (evidence here) and metafunction (evidence here).

Moreover, any theoretical value in Martin's work lies in the original works that he has misunderstood and rebranded as his own:

• Halliday's semantic SPEECH FUNCTION (rebranded as Martin's discourse semantic NEGOTIATION);
• Halliday & Hasan's grammatical REFERENCE (rebranded as Martin's discourse semantic IDENTIFICATION);
• Halliday & Hasan's grammatical ELLIPSIS-&-SUBSTITUTION (also rebranded as Martin's discourse semantic IDENTIFICATION);
• Halliday & Hasan's grammatical cohesive CONJUNCTION (reinterpreted as Martin's discourse semantic CONJUNCTION, now rebranded as CONNEXION); and
• Halliday & Hasan's LEXICAL COHESION (rebranded as Martin's discourse semantic IDEATION).

As Bateman has demonstrated in his review of Martin's English Text, he has not bothered to check the provenance of its content, nor has he been able to identify any of the wealth of theoretical inconsistencies that arise from Martin's misinterpretations. In recommending Martin's work, Bateman has done a great disservice to the academic community.

On The Breadth And Systematic Interrelationships In 'English Text'

Bateman (1998: 25): 

One of the difficulties in approaching English Text is at the same time one of the principal reasons for making the effort—that is, its breadth. Work in the complex area of discourse organisation and function typically achieves a fragmentary view at best, focusing in on particular areas and phenomena. English Text goes beyond this, presenting both detail and systematic interrelationships that successfully evoke a feel for the semantics of discourse and its realisation in lexicogrammatical patterns as a complex interacting whole.

Blogger Comments:

[1] This is very misleading.  Martin's model of discourse semantics is largely confined to the textual metafunction, and only part of that.  This is because Martin's systems of IDENTIFICATION, CONJUNCTION and IDEATION are misunderstandings of Halliday & Hasan's (1976) lexicogrammatical cohesion, rebranded and misrepresented as semantic stratum systems of three different metafunctions.  (The one exception is NEGOTIATION, which is Martin's rebranding of Halliday's interpersonal semantic system of SPEECH FUNCTION.)

[2] This is very misleading.  Martin's systems of IDENTIFICATION, CONJUNCTION and IDEATION do not specify how their paradigmatic features are realised syntagmatically, and do not specify how they are realised at the level of lexicogrammar.  Moreover, due to Martin's misunderstanding of theoretical concepts, the model creates mismatches between semantics and lexicogrammar, even in the absence of grammatical metaphor.

[3] Given all the misunderstandings that invalidate Martin's model, it is not true to claim that it evokes even a feeling for such matters successfully.

∞

The arguments that constitute the evidence for all the above claims can be read here.

On The Theoretical Value Of Martin's Stratification Of The Content Plane

Bateman (1998: 23-4):

The availability of the two strata, lexicogrammar and discourse semantics, is presented as beneficial in most areas that Martin addresses. The value of stratification, and of allowing the description to range across two (or more) strata, is argued in all of the areas that Martin addresses in English Text. The ability to distinguish systematically between a semantic unit and a range of possible grammatical units or patterns of such units is of fundamental importance—particularly given the very richly elaborated view of grammar assumed. Opening up the realisational relation between a semantic description and a grammatical description significantly empowers grammatical metaphor as an explanatory device.

Blogger Comments:

[1] This is misleading to the extent that it gives the false impression that the stratification of the content plane into lexicogrammar and semantics is Martin's proposal rather than Halliday's (whose considerably prior notion of grammatical metaphor depends on such a stratification).

[2] To be clear, Martin argues for the value of his stratification because it allows him to create his own "module" within SFL theory by rebranding his misunderstandings of others' work as his own systems:

• negotiation is Martin's rebranding of Halliday's speech function;
• identification is Martin's rebranding of Halliday & Hasan's grammatical reference (and ellipsis-&-substitution);
• conjunction is Martin's rebranding of Halliday & Hasan's grammatical conjunction; and
• ideation is Martin's rebranding of Halliday & Hasan's lexical cohesion.

Martin's argument for stratification, however, presented in Chapter 1, does not survive close scrutiny; see, for example, Why The Argument For A 'Discourse' Semantic Stratum Is Invalid.

In brief, Martin provides three motivations for his stratification:

• semantic motifs
• grammatical metaphor
• cohesion

Of the three, only grammatical metaphor motivates a distinct semantic stratum, but this is Halliday's argument, not Martin's, and it does not motivate a specifically discourse semantics. Moreover, Martin misunderstands grammatical metaphor (evidence here), and reinterprets it as a process, a "texturing interface" (p401) between his discourse semantic systems and lexicogrammar (ellipsis-&substitution, Theme and collocation), thereby undermining his own argument for it as a motivation for stratification.

Of semantic motifs, Martin proposes (p16) setting up an attitude network to generalise what he sees as commonalities between some mental, behavioural and relational clauses.  For some of the misunderstandings involved, see The Problems With Semantic Motifs As A Motivation For Stratification.  Even so, no mention of semantic motifs is made beyond Chapter 1, and Martin's experiential system, ideation, does not address the points he raises — which is hardly surprising, given that it is a rebranding of a textual grammatical system, lexical cohesion.

Regarding cohesion as a motivation for stratification, Martin's argument involves misrepresenting the source of his ideas: Halliday & Hasan's (1976) model of cohesion.  See, for example:

• Misrepresenting Cohesion
• Misrepresenting The Domain Of Cohesion
• Misconstruing 'Cohesion Within The Sentence'
• Not Recognising The 'Continuity' Between Clause Taxis And Conjunctive Cohesion

[3] To be clear, Martin's discourse semantic units are:

• exchange and move (interpersonal system of negotiation)
• participant (textual system of identification — i.e. rebranded cohesive reference etc.)
• message (logical system of conjunction — i.e. rebranded cohesive conjunction), and
• message part (experiential system of ideation — i.e. rebranded lexical cohesion).

The problems here are with Martin's rebranding of cohesive systems. Most importantly, because Martin treats these non-structural systems as structural, the units he proposes are not units with internal structure, but units that relate to other units. This creates an inconsistency in the meaning of unit, since all other units in SFL theory are units with internal structure. So these discourse semantic units are not units in the same sense as lexicogrammatical units, such as clause, phrase and group.

Martin's naming of units further reveals his theoretical misunderstandings.  For example, the use of an experiential category, participant, for a textual system, identification, is consistent with his misunderstanding of textual reference with ideational denotation, as explained elsewhere on this blog.  On the other hand, Martin's use of textual categories, message and message part, for ideational systems, conjunction and ideation, is inconsistent with his own theorising, but (metafunctionally) consistent with the source of his ideas, the textual systems of cohesive conjunction and lexical cohesion.  Moreover, in naming the two message and message part, Martin misrepresents a relation of interdependency as one of composition (whole-part).

[4] This is very misleading indeed.  Firstly, as already mentioned, Martin does not understand grammatical metaphor (evidence here).  Secondly, Martin does not open up "the realisational relation between a semantic description and a grammatical description" since he provides no realisation statements that relate his discourse semantic systems to lexicogrammatical systems. Thirdly, because Martin does not specify congruent lexicogrammatical realisations of his discourse semantic systems, he provides no means of identifying congruent vs metaphorical realisations using his model.  That is, if anything, Martin's model undermines grammatical metaphor rather than "empowering" it — the direct opposite of Bateman's claim.

On Martin's Controversial Move To Explicitly Consider Semantics

Bateman (1998: 23):

It is also certainly not the case that Martin’s approach is uncontroversial, even within his own functional tradition. The need for a move to explicitly consider semantics has sometimes been downplayed, aided in part by the rich functional semantic flavour that permeates grammatical descriptions such as that found in Halliday’s IFG; here “the grammar is infused with meaning, and a stratal distinction between grammar and semantics systematically blurred.’’[p33]. 

Blogger Comments:

[1] Here Bateman again frames the matter in terms of agreement among linguists instead of in terms of theoretical validity.  The distinction between agreement and validity is encapsulated in the following cartoon of B. Kliban:

[2] This is potentially misleading, since Martin does not "explicitly consider semantics", because his discourse semantics, as previously explained, is merely a combination of Halliday's textual lexicogrammar (rebranded cohesion systems) and Halliday's previously theorised semantic system (rebranded speech function).

[3] To be clear, as Halliday & Matthiessen (2014: 49) explain:

Being a ‘functional grammar’ means that priority is given to the view ‘from above’; that is, grammar is seen as a resource for making meaning — it is a ‘semanticky’ kind of grammar. But the focus of attention is still on the grammar itself.

[4] This quote from Martin demonstrates his misunderstanding of stratification, which is also encompassed in his mantra 'all strata make meaning', which confuses stratification ('all strata') with semogenesis ('make meaning'), and leads him to interpret every stratum as a stratum of meaning.  See, for example, Misrepresenting Stratification.

To be clear, SFL theory distinguishes between meaning (semantics) and wording (lexicogrammar), principally because the distinction is essential for the systematic understanding of grammatical metaphor.

On Martin's Text Analysis And The Differences Between Identification And Negotiation

Bateman (1998: 12):

The chapter concludes with detailed analyses of two longer texts, both drawn from earlier research into the development of children’s narrative, a brief discussion of the relation between the circumstances of location and manner and discourse identification, and an orienting summary of the differences between IDENTIFICATION and NEGOTIATION as discussed in the previous chapter.

Blogger Comments:

[1] For some of the serious problems with Martin's analyses of texts [3:88] and [3:89], unnoticed by Bateman, see:

• The Analysis Of Text [3:88]: An Attempt To Hide A Theoretical Inconsistency
• Eight Problems With Martin's Six Notes To His Analysis Of Text [3:88]
• Eight Problems With Martin's Nine Notes To His Analysis Of Text [3:89]

[2] For 19 of the problems with Martin's 2-page discussion, unnoticed by Bateman, see:

• Confusing Circumstances Of Location With Demonstrative Reference Items
• Mistaking Manner For Extent And Confusing Circumstances With Comparative Reference Items

[3] For some of the theoretical misunderstandings and inconsistencies in the "orienting summary", unnoticed by Bateman, see:

• Metafunctional Inconsistencies
• Some Of The Theoretical Problems With Participant As The Entry Condition To The System Of Identification
• Misconstruing 'Multivariate' & Metafunctional Inconsistencies (1)
• Misconstruing 'Multivariate' & Metafunctional Inconsistencies (2)
• Martin's Insight That Interruptions Depend On There Being Something To Interrupt
• The Commonality Of Reference Structures And Negotiation Structures

On Martin's Understanding Of Stratification

Bateman (1998: 9-10):

Thus, for example, whereas some accounts of interaction previously have needed to go inside the structure of individual discourse moves, … the discourse semantics of English Text can instead leave much of the work here to grammar and retain simpler congruent move realisations as an independent clause or incongruent move realisations as sequences of sentences related conjunctively. … 

As Martin writes:

“The general point here is that if the grammar, or phonology for that matter ..., does the work, so be it. The model developed here does not dualise meaning and form so does not have to re-state the contributions made by phonology and lexicogrammar to text structure at the level of semantics. Equally important, ... is the fact that negotiation provides just one of four perspectives on text structure...’’ [p56] ... “[T]he modular approach to discourse means that no one component is responsible for accounting comprehensively for textual relations.’’[p268]

Blogger Comments:

[1] The thought that didn't occur to Bateman here is that Martin's model is based on a fundamental misunderstanding of the architecture of SFL theory, in as much as he misunderstands both strata and metafunctions as interacting modules, instead of levels of symbolic abstraction and organising principles, respectively.

[2] The thought that didn't occur to Bateman here is that Martin does not understand that strata represent different levels of symbolic abstraction.  Because of this, discourse semantics cannot "leave much of the work to grammar".  Instead, the grammar realises semantics, and if the theorising is to be consistent, the semantics that the grammar realises needs to be identified. 

[3] The thoughts that didn't occur to Bateman here are that Martin confuses 'marked' with 'incongruent' and proposes incongruent grammatical realisations that are not metaphorical.

[4] The thought that didn't occur to Bateman here is that Martin mistakes all strata — even phonology — as levels of meaning; hence the claim that his model "does not dualise meaning and form".  Martin's error derives from mistaking semogenesis — his "all strata make meaning" — for stratification, where meaning, wording and sounding constitute different levels of symbolic abstraction.  See also:

• Confusing Semogenesis With Stratification
• Confusing Semogenesis With Levels Of Symbolic Abstraction
• Confusing Semogenesis With Stratification

On Martin's Modularity

Bateman (1998: 9):

This leads to an important strand in the discussion that also runs through many of the later chapters: the concept of modularity. English Text distributes its linguistic work across lexicogrammar and the four components of discourse semantics. What some approaches would therefore have tried to include in their exchange structure analysis, Martin separates out, allowing other components to contribute. This simplifies the overall account in a number of ways, and also opens up the significant topic of interaction between modules.

Blogger Comments:

The thought that didn't occur to Bateman here is that the architecture of Systemic Functional Linguistic theory is not modular, but dimensional; see also here.  Halliday & Webster (2009: 231):

In SFL language is described, or “modelled”, in terms of several dimensions, or parameters, which taken together define the “architecture” of language. These are 

(i) the hierarchy of strata (context, semantics, lexicogrammar, phonology, phonetics; related by realisation); 

(ii) the hierarchy of rank (e.g. clause, phrase/group, word, morpheme; related by composition); 

(iii) the cline of instantiation (system to instance);  

(iv) the cline of delicacy (least delicate to most delicate, or grossest to finest);  

(v) the opposition of axis (paradigmatic and syntagmatic);  

(vi) the organisation by metafunction (ideational (experiential, logical), interpersonal, textual).

Misunderstanding the architecture of the theory in this way creates serious theoretical inconsistencies in Martin's model, as will be demonstrated on this blog in the critiques of Bateman's review of Martin's Chapter 6, and has already been demonstrated here.

On Martin's 'Incongruence'

Bateman (1998: 9):

Martin also suggests problems with Ventola’s definition of move complexes as being realised as paratactic clause complexes and instead relies on a stratal mismatch (incongruency) to allow a discourse move that can be realised congruently by a single full clause, but incongruently as a sequence of clauses or even sentences related conjunctively (cf. Chapter 4).

Blogger Comments:

There are two thoughts that did not occur to Bateman here, both of theoretical importance:

1. Martin (pp58-9) confuses congruence with unmarkedness, as explained at Confusing Unmarkedness And Congruence.
2. In proposing such incongruent relations between discourse semantics and lexicogrammar, Martin is unwittingly proposing that such realisations are metaphorical — a claim that he does not make (and could not be maintained in a manner consistent with theory).

On Martin's System Of Negotiation

Bateman (1998: 8):

Martin adopts a number of different starting points for his discussions, providing good insights into different ways of approaching discourse semantics. His starting point for Chapter 2 [pp31-92] on NEGOTIATION is the grammatical systems of MOOD: i.e., that part of clause grammar that describes basic interactional functions such as assertion, question, imperative, tags, finite vs. nonfinite, dependent vs. independent, and polarity (cf. IFG: Chapter 4; Matthiessen, 1995: Chapter 5, pp391-433). The system network for NEGOTIATION that Martin derives must provide motivations for selections from the MOOD system as are appropriate for particular dialogue moves. The basic framework set up is one that describes the semantic unit ‘text’ as an exchange of meanings. Such exchanges are organised into configurations of discourse moves; the particular configurations of moves that are found establish a notion of exchange structure.

Blogger Comments:

[1] This misrepresents Martin's theoretical inconsistencies as a virtue.  The different starting point for the system of NEGOTIATION reflects its different source (interpersonal semantics) from the other discourse semantic systems (textual lexicogrammar).

[2] Trivially, this confuses the grammatical system of MOOD with the semantic system of SPEECH FUNCTION.

[3] To be clear, "the system network for NEGOTIATION that Martin derives" is not displayed as an integrated network, but instead consists of:

• Figure 2.10 (p49): Ventola's (1987) extension of Berry (1981);
• Figure 2.11 (p49): Martin's augmentation of it with a less delicate system; and
• Figure 2.23 (p81): Martin's addition a more delicate system (parallel with Ventola's) which only applies to classroom registers.

[4] This misunderstands the relation between strata.  Semantic systems don't "motivate" grammatical selections; (selections in) semantic systems are realised by (selections in) grammatical systems, and the relation between the selections may be congruent or incongruent (metaphorical), and vary in terms of probability according to register.

[5] This is misleading on two counts.  On the one hand, it misrepresents the relation between the systems of NEGOTIATION and MOOD.  Martin's model (p50) distinguishes two ranks, exchange and move.  NEGOTIATION is the system of exchange rank, and it is not realised by any grammatical system.  Halliday's SPEECH FUNCTION is the system of move rank, and it is realised by the the grammatical system of MOOD.

On the other hand, since the relation between SPEECH FUNCTION selections and MOOD selections was theorised by Halliday, long before Martin, it falsely attributes Halliday's ideas to Martin.  Moreover, Martin (1992: 36) misunderstands the realisation relation between SPEECH FUNCTION and MOOD, as demonstrated here.

[6] This confuses Halliday's highest semantic unit, text, with Martin's highest discourse semantic unit, exchange.

On Martin's 'Synoptic Vs Dynamic' Distinction

Bateman (1998: 8):

There is also an important difference with respect to the status of text as ‘product’ or as ‘process’. Of the regions discussed only NEGOTIATION has a significantly developed ‘static’ component — hence the multivariate structure adopted for exchanges. The remainder are more dynamic and concerned with the unfolding of texts, giving a very different flavour to the discourse structures used. The dynamic aspect is also taken up at some length in the discussion of genre [pp550–557]. Reconciliation of these differing perspectives and structures remains an exciting area for future work, just as it does for lexicogrammar.

Blogger Comments:

[1] Here Bateman's wording conceals Martin's confusion of the realisation of system as (synoptic) structure with the (dynamic) instantiation of the system in logogenesis.  See for example:

• Confusing Realisation And Instantiantion
• Misconstruing Realisation And Instantiation

[2] This is misleading.  The difference lies in the different intellectual sources of Martin's theorising.  The system of NEGOTIATION is taken from the previous work of colleagues on (structural) interpersonal semantics, whereas the systems of IDENTIFICATION, CONJUNCTION and IDEATION are taken from the previous work of colleagues on (non-structural) textual lexicogrammar.

[3] This is misleading.  The discussion of 'dynamic perspectives' on genre is very brief (three paragraphs), and does little more than refer to Ventola's flowchart of a service encounter (Figure 7.22, pp554-5).

On The Type Of Structure And Direction Of Dependency For Discourse Systems

Bateman (1998: 8):

Table 5.21 [p332] also sets out for each type of semantic structure the direction of dependencies typically found among its covariately related units. We can therefore see that each region, and hence each chapter, discusses its own particular ‘discourse structural’ configurations and the system network of alternatives that gives rise to those configurations.

Blogger Comments:

[1] This is misleading.  Table 5.21 provides very little differentiation of "discourse structures":

Table 5.21. Type of structure and direction of dependency for discourse systems

	STRUCTURE	DIRECTION OF DEPENDENCY
NEGOTIATION	multivariate & covariate	prospective & retrospective
IDENTIFICATION	covariate	retrospective (occasionally cataphoric)
CONJUNCTION	covariate	retrospective (occasionally cataphoric)
IDEATION	covariate	prospective & retrospective

[2] This is misleading.  Martin's system networks provide no realisation rules that specify structural realisations.

Favourably Misrepresenting Martin's Discourse Structures

Bateman (1998: 7-8):

These diverse discourse structures are developed in detail and with copious examples in their own particular chapters, but a number of useful overview tables are also given towards the end of Chapter 5. For example, the semantic units participating in discourse structures are summarised in Table 5.18 [p325] thus:

Table 5.18. Discourse semantics: units proposed in English Text
interpersonal	textual	ideational: logical	experiential
exchange
move		message
	participant		message part

As within the grammar, different labellings are proposed within each metafunction. That is: the region of NEGOTIATION, which is concerned with the interpersonal construction of text as an unfolding ‘exchange’, links together discourse ‘moves’; the region of CONJUNCTION, which is concerned with how meanings are combined together to form larger ‘logically’ related meanings, links ‘messages’; the region of IDENTIFICATION, concerned with establishing reference to discourse entities, links ‘participants’; and the region of IDEATION, concerned with ‘the company that words keep’, links ‘message parts’.

Blogger Comments:

[1] The term 'diverse' here is misleading.  Martin (1992: 331) describes all his structures as covariate, with the addition of multivariate structure in the case of his interpersonal meaning; see previous post.  Importantly, 'covariate' (Lemke 1985) is not a type of structure, as later acknowledged by Lemke (1988: 159) himself.

[2] The notion of 'units participating in discourse structures' conceals an inconsistency in Martin's structure types, deriving from his confusing structured units with units in a relation.  To be clear, Martin's interpersonal units are structured units, like the clause in the grammar, whereas the units of Martin's other metafunctions are units in relations to other units, like Halliday's cohesive relations that Martin has misunderstood, rebranded and relocated, stratally and metafunctionally, from lexicogrammar to discourse semantics.

[3] This echoes Martin's confusing of the realisation relation between system (negotiation) and structure (exchange), on the one hand, with the instantiation of potential in logogenesis (unfolding of text), on the other.  See, for example, Confusing Realisation And Instantiantion.

[4] The use of 'links' here continues the concealment of the inconsistencies in Martin's notion of structure, as identified above in [2].  The relation between 'exchange' and 'move' is one of constituency: an exchange consists of moves.  The notion of NEGOTIATION "linking" moves falsely construes interpersonal structure as consistent with

• CONJUNCTION "linking" messages, 
• IDENTIFICATION "linking" messages, and
• IDEATION "linking" message parts.

The notion of 'links' is consistent with the notion of 'cohesive ties': the non-structural relations in the systems of cohesion that Martin has rebranded, relocated and misunderstood as discourse semantic structural relations.

On Martin's Four Discourse Semantic Systems

Bateman (1998: 6-7):

The four regions of discourse semantics addressed in depth are NEGOTIATION, IDENTIFICATION, CONJUNCTION and IDEATION. NEGOTIATION concerns those resources of discourse semantics that are responsible for the construction of dialogue and interaction. IDENTIFICATION captures the semantic resources for referring to and tracking discourse participants. CONJUNCTION develops the ‘logic’ of English text in terms of those resources by which semantic messages are combined into larger complex messages, and through which messages are related to previously expressed messages as a text or dialogue unfolds. And IDEATION attempts to motivate those systematic selections from groups of ‘related’ lexical items that bring about a text’s ‘lexical’ cohesion. Each of these therefore illustrates a distinct general kind of meaning carried in discourse and can be broadly allocated to a particular metafunction: NEGOTIATION is broadly interpersonal, IDENTIFICATION textual, CONJUNCTION logical, and IDEATION experiential. Consequently, as with grammar, it appears the case that diverse kinds of discourse structure may be usefully posited for each area.

Blogger Comments:

[1] To be clear, in theoretical terms, these four are proposed systems of a proposed discourse semantic stratum.

[2] To be clear, this is the apt subtitle of the chapter on CONJUNCTION. For some of the violations of logic in this chapter, see the 103 critiques here.  Martin abandons the most general categories of expansion — elaboration, extension and enhancement — and because Martin's CONJUNCTION is a largely a rebranding, relocating and misunderstanding of Halliday's system of grammatical cohesion, it doesn't include the other major logico-semantic system, projection, at all. This undermines any systematic account of grammatical metaphor, even if it redefined less precisely as 'interstratal tension'.

[3] The use of 'broadly' here is an unjustified hedge.  Metafunctions are distinct perspectives on meaning, and Martin specifically assigns each of his systems to each of the metafunctions.

[4] The proposed unit of Martin's textual system of IDENTIFICATION is the experiential category 'participant'.  Martin's system is a rebranding, relocating and misunderstanding of Halliday's grammatical system of cohesive reference, as demonstrated here.

[5] Martin's logical system of CONJUNCTION is a confusion of the structural relations between clauses in complexes (logical metafunction) and non-structural cohesive conjunctive relations (textual metafuction).

[6] Martin's experiential system of IDEATION is a confusion of lexical cohesion (textual metafunction), and misunderstandings of 'lexis as most delicate grammar', inter alia, as demonstrated here.

[7] This is misleading.  As Martin (1992: 331) points out after describing all four of his discourse semantic systems:

All the discourse structures introduced to this point have been covariate ones, with the exception of the multivariate interpretation of the exchange introduced in Chapter 2. But even there the multivariate approach presented only a partial picture; covariate tracking and challenging structures had to be developed to fill out the picture. Lexical relations are also covariate structures — message parts depend semantically on each other, and depending message parts are themselves depended on.

That is, for the most part, the "diverse kinds of discourse structure…posited for each area" is the one kind, covariate — which was later repudiated as a structure type by its originator, Lemke, several years before Martin's publication, as previously shown.  For a critique of the above extract, see here.