A Summary of The Problems That Invalidate The Authors' Model Of Ideational Structure

Martin & Doran (2023: 44, 45):

By way of summarising our discussion we extend Table 1 above as Table 3 below, including English examples and filling out our paradigm for ideational structure.

The two central columns in Table 3 oppose non-recursive systems realised by non-iterative structures to recursive systems realised by iterative structures. This opposition is cross-classified by experiential as opposed to logical structure; logical structure is further classified as paratactic vs hypotactic, and within hypotactic structures are subclassified as progressive or regressive. The only structure type in the table not introduced above is non-iterative parataxis for which we suggest English correlative structures such as both…and, either…or, neither…nor (including perhaps the somewhat archaic the harder they come the harder they fall).

 

Blogger Comments:

[1] To be clear, Table 1 was from Halliday (1965):

[2] To be clear, the authors' update of Halliday (1965) is limited to the three types of non-iterative logical structures and the iterative experiential structure.

The mistaken notion of a non-iterative logical structure arose from not understanding that a two-unit complex (duplex) is specified by selecting 'stop' on the first pass through a recursive system. Having first misapplied a subjacency duplex structure to submodification in a Pitjantjatjara nominal group, the authors then applied it to other structures in other languages without providing argument as to why it was appropriate to do so.

The mistaken notion of an iterative experiential structure arose from not understanding that 'multiple Epithets' at the rank of group are realised by iterative logical structures at the rank of word. The authors compounded this error by proposing that two types of structure obtain within the one experiential structure of a nominal group.

By way of comparison, the current SFL model of structure type by metafunction is given by Halliday & Matthiessen (2014: 85):

Confusing Group Rank With Word Rank

Martin & Doran (2023: 39, 40):

To this point in our discussion we have focused on structure, without looking carefully at the systems from which our structures derive. One systemic implication arising has to do with the need to distinguish two different types of recursive system — namely recursive systems giving rise to serial logical structures (whether paratactic or hypotactic) as opposed to recursive systems giving rise to iterating elements of experiential structure. 

For English EPITHESIS for example we need to distinguish between systems underpinning indefinitely extendable regressive grading complexes (a not much more glorious history) and systems underpinning multiple Epithets (their long glorious well-documented history). 

One possibility would be to retain standard SFL recursive system notation for paratactic and hypotactic series (e.g. the grading system in Figure 4 to generate not much more glorious history) and use an ‘ⁿ’ superscript on the relevant feature for experiential iteration (e.g. the [describedⁿ] option in Figure 4 to generate long glorious well-documented history). 

Formulated along these lines Figure 4 thus includes one non-recursive DEIXIS system (from which a single Deictic function can be derived), one recursive EPITHESIS system (from which multiple Epithets can be derived) and one recursive GRADING system (from which a hypotactic series can be derived).

Blogger Comments:

[1] To be clear, as previously demonstrated, the recursive systems that specify these iterative structures are at word rank, not group rank, and the structures are logical, word complexes, not experiential.

[2] To be clear, Figure 4 confuses systems at group rank that specify multivariate structures, with the systems at word rank that specify the univariate structures that realise submodification at group rank.

Some Of The Problems With Interpreting A Nominal Group "Focus" As A Subjacency Duplex

Martin & Doran (2023: 34):

Generalisation of subjacency structure for adpositions of all kinds opens up a host of possibilities. In (25) for example the Focus structure in (1) is reinterpreted along these lines, which obviates the need for an experiential Focus Marker function realised by of. This analysis more readily acknowledges that the of is not in fact a constituent of the embedded nominal group, but rather “modifies” it. It also opens the way for acknowledging that the of does not have the possibility for internal constituency itself (a regular feature of subjacency markers), unlike many of the other words within the group — that is, it acknowledges that the of is not in fact part of the experiential constituency hierarchy that underpins SFL’s notion of rank.

Blogger Comments:

[1] Again, adpositions are not structure markers, but markers of functions (clause roles); see Matthiessen (1995: 370). Because they are not structure markers, they are irrelevant to the concerns of the paper: interpreting structure markers in terms of subjacency duplexes.

[2] To be clear, 'Focus' is Martin's rebranding of Halliday's 'Facet'. It is a particularly poor rebranding, since it uses a textual distinction for an experiential construal, and because the term already has a place in the theory as the Focus of information.

[3] To be clear, here the authors have misinterpreted a genuine structure marker (of) as an adpositional function marker and reinterpreted the misinterpretation as a subjacency duplex. Importantly, the structure marker of is not a marker of one nominal group function, such as Facet, as demonstrated by all the nominal groups with this of where the Head and Thing conflate, as in the king of Bohemia.

[4] This is misleading, because it is not true. To be clear, in SFL Theory, modification is a hypotactic relation of subcategorisation. Halliday (1985: 170):

for the purposes of the nominal group we need to take account of just one such relationship, that of subcategorisation: ‘a is a subset of x’. This has usually been referred to in the grammar of the nominal group as modification, so we will retain this more familiar term here.

Clearly, the preposition of does not subcategorise the nominal group the tops. Moreover, this analysis misinterprets a nominal group and the preposition of the following prepositional phrase as a two-unit complex (duplex).

[5] This is a very serious misunderstanding of the ranked constituency of SFL Theory. On the one hand, the word of is monomorphemic: it consists of one morpheme; that is its 'internal constituency'. On the other hand, the ranked constituency hierarchy is a model of form, not experiential meaning. What is true is that the type of structure favoured by the experiential metafunction, segmental, is based on constituency (Halliday & Matthiessen 2014: 85).

[6] To be clear, all the constituents of this nominal group are accounted for by the logical analysis:

However, the authors' problem is, again, methodological: instead of starting with function (Value) and asking how it is realised in form (Token), they start with form (Token) and ask what function (Value) it realises. In this case, instead of starting with the Facet and asking how it is realised, they start with a preposition and ask what function it realises.

On the basis of Halliday's editions of IFG (1985, 1994), the multivariate analysis of this nominal group is:

Misapplying Subjacency Duplexes To Adpositions Misunderstood As Structure Markers [3]

Martin & Doran (2023: 33-4):

Examples (7) and (8) above presented the problem of embedded clauses whose role in clause structure is signalled by the same structure markers that position nominal groups. The relevant parts of these are reworked as (23) and (24) below with a subjacency duplex realising the relevant clause function. This obviates the need to suggest, rather spuriously, that these embedded clauses are in fact a special kind of nominal group.

Blogger Comments:

[1] As previously explained, there is no problem here. The adposition marks a clause rank function, and in these cases the function is realised by an embedded. The reason why the authors mistakenly think this to be a problem is methodological: instead of starting with function (Value) and asking how it is realised in form (Token), they start with form (Token) and ask what function (Value) it realises.

Again, these adpositions are not structure markers, but markers of functions (clause roles); see Matthiessen (1995: 370). Because they are not structure markers, they are irrelevant to the concerns of the paper: interpreting structure markers in terms of subjacency duplexes.

[2] To be clear, applying existing SFL theory, the participants in (23) and (24) are each realised by an adpositional phrase consisting of an embedded and an adposition, on the model of a prepositional phrase:

Again, Martin & Doran have merely rebranded an adpositional phrase as a subjacency duplex.

[3] Again, this seriously misunderstands rankshift. There is no suggestion that a clause that is embedded in a nominal group is a "special kind of nominal group". Such a clause is shifted to the rank of word, where, like words, it serves as an element of nominal group structure (Head/Thing).

Again, the authors' misunderstanding is methodological: instead of starting with function (Value) and asking how it is realised in form (Token), they start with form (Token) and ask what function (Value) it realises.

Misapplying Subjacency Duplexes To Adpositions Misunderstood As Structure Markers [1]

Martin & Doran (2023: 32-3):

Recognition of subjacency duplexes opens up the possibility of a logical alternative to the experiential analysis of adpositions discussed in Section 2. Examples (2) and (3) are reworked as (19) and (20) below. Here participants in clause structure are analysed as being realised by subjacency duplexes, with progressive (α β#) or regressive (#β α) structures as appropriate; their α is realised by a nominal group and their β by the relevant clitic.

Blogger Comments:

[1] To be clear, subjacency duplexes were said to be motivated by relations within nominal group structure, but here they are applied to a relation between a nominal group and an adposition. 

Importantly, these adpositions are not structure markers, but markers of functions (clause roles); see Matthiessen (1995: 370). Because they are not structure markers, they are irrelevant to the concerns of the paper: interpreting structure markers in terms of subjacency duplexes.

[2] To be clear, applying existing SFL theory, the participants in (19) and (20) are each realised by an adpositional phrase consisting of a nominal group and an adposition, on the model of a prepositional phrase:

Cf Matthiessen (1995: 370):

In summary, Martin & Doran have merely rebranded an adpositional phrase as a subjacency duplex.

Nevertheless, the question is — even ignoring all the misunderstandings involved — what explanatory advantage does the subjacency duplex analysis have over existing SFL theory?

Misconstruing Word Rank Iterative Logical Structures As Group Rank (i) Iterative Experiential Structures And (ii) Non-Iterative Logical Structures

Martin & Doran (2023: 31):

The discussion in this section indicates that we need to loosen up the association of experiential and logical meaning with specific types of system and structure – as outlined in Table 2 (setting aside paratactic complexing to simplify the discussion here). 

For experiential meaning the table draws on the contrast in English between a Deictic Numerative Thing structure for which non-recursive systems insert each function once and Epithet Thing Qualifier structures for which recursive systems allow for multiple Epithets and Qualifiers independently modifying the Thing (notated with an “ⁿ” superscript below). 

For logical meaning the table contrasts duplexes deriving from non-recursive systems with series deriving from recursive ones (e.g. the contrast between Pitjantjatjara and English, for example, as far as grading is concerned). The notation allows for both progressive and regressive duplexes and series (i.e. “left-headed” α β… vs “right-headed” β α dependency structures). And duplexes are distinguished from series by marking their culminative dependent β element with “#”.

Blogger Comments:

[1] This is misleading, because it is untrue. As previously demonstrated, the authors' notion of a non-recursive system and non-iterative structure for a logical structure is based on their misunderstanding of the nature of recursive systems and the iterative structures that realise them. And the authors' notion of a recursive system and iterative structure for experiential structure is based on mistaking recursive systems and iterative structures of the logical metafunction at word rank for recursive systems and iterative structures of the experiential metafunction at group rank.

That is, what the authors presented as a non-iterative logical structure at group rank was an iterative logical structure at word rank: a two-unit hypotactic word complex. And what the authors presented as an iterative experiential structure at group rank was actually an iterative logical structure at word rank: a hypotactic word complex realising an Epithet.

[2] As previously explained, here the authors misconstrue the word rank iterative logical structures that can realise elements of group structure as iterative experiential structures at group rank. But here they also create theoretical inconsistency by proposing different structural types obtaining between different elements of the same unit. 

[3] As previously explained, here the authors misconstrue the word rank iterative logical structure that realises a submodified Epithet as a group rank non-iterative structure. But here they also create theoretical inconsistency by proposing different logical structural types obtaining between different experiential elements of the same unit.

[4] To be clear, Matthiessen (1995: 600-2) proposes a textual system of CULMINATION, at clause rank, as a written mode analogue of spoken mode INFORMATION. The authors, however, just use 'culminative' to mean 'final', without any acknowledgement of Matthiessen.

Misrepresenting Recursive Systems As Specifying Experiential Structures

Martin & Doran (2023: 30):

The possibility of multiple Epithets (English) and Qualifiers (Korean and English) indicates that too strict an association of non-recursive systems with experiential structure is not tenable. Recursive systems clearly underlie both the English Epithets and Korean Epithets and Qualifiers exemplified above.⁷

⁷ At first blush Halliday’s (1985) analysis of English Classifiers would appear to exemplify a comparable pattern. But English Classifiers do not each independently modify the Thing (the structure is serial not orbital). In fact structures of this kind would be better modelled as hypotactic word complexes recursively taxonomising the entity construed by the Thing function (see Martin et al. 2021 for discussion).

Blogger Comments:

[1] To be clear, this is misleading because it is untrue. The 'multiple Epithets' example demonstrated that the experiential function is realised by a logical structure at word rank: a paratactic or hypotactic word complex. The 'multiple Qualifiers' example featured an embedded clause complex serving as a single Qualifier.

[2] To be clear, firstly, this misconstrues experiential structure (the Classifier–Thing relation) as logical ('modify', 'serial'). Secondly, it claims that both the Classifier and the Thing are both Heads, since Martin's serial structure is 'multi-nuclear'. Thirdly, Martin's serial structure misconstrues univariate structure as parataxis ('multi-nuclear'). Fourthly, it proposes a different structure type (serial) for just two elements of a nominal group: Classifier(s) and Thing. 

[3] This is misleading. Firstly, the notion that the Classifier 'taxonomises' the Thing is Halliday (1985), not Martin (2021). Halliday (1985: 160, 164):

This is an experiential structure which, taken as a whole, has the function of specifying (i) a class of things, namely trains, and (ii) some category of membership within this class. We shall refer to the element expressing the class by the functional label Thing. … 

The Classifier indicates a particular subclass of the thing in question, e.g. electric trains, passenger trains, wooden trains, toy trains.

Secondly, the notion that a Classifier can be realised by a hypotactic word complex is Halliday (1985), not Martin (2021), as expressed in Matthiessen (1995: 665):

CLASSIFICATION is the resource for specifying successively more specific classes of a thing; it is used to represent taxonomies. Table 7-14 presents a few examples from geology.

The recursive system is, of course, logical and at word rank.

Misconstruing Nominal Group Qualifiers

Martin & Doran (2023: 29-30):

The same kind of pattern arises for both Korean and English as far as Qualifiers are concerned — multiple Qualifiers are possible, each modifying the Thing function independently of one another (and thus can be moved around with respect to one another without affecting the construal of experiential meaning). Korean examples are provided in (15) and (16) below, with the same two Qualifiers, but in reverse sequence (Martin and Shin 2021). The English translations for these examples display the same pattern, but with the Qualifiers following rather than preceding the Thing.

Blogger Comments:

[1] To be clear, based on the English translation, there is only one Qualifier in (15) and (16), each one realised by an embedded paratactic extending clause complex:

The interdependency relation is paratactic because either clause can stand on its own without the other, and the expansion relation is extension: addition because the meaning is 'X and Y'; see Halliday & Matthiessen (2014: 471-2).

The reason why Martin & Doran misconstrue this as two Qualifiers is methodological: they are concerned with assigning function to form (embedded clauses), instead of assigning forms to function (Qualifier). That is, they are decoding form instead of encoding function.

[2] To be clear, the theoretical reason why the embedded clauses 'can be moved around with respect to one another without affecting the construal of experiential meaning' is that parataxis is a symmetrical relation. Halliday (1985: 198):

In principle, the paratactic relation is logically symmetrical … 'salt and pepper' implies 'pepper and salt', so the relation is symmetrical;

Misconstruing The Multivariate Structure Of Nominal Groups

Martin & Doran (2023: 29):

This restriction contrasts with English, where alongside the adjective complexing in (13), there can be an indefinite number of Epithets as in (14).⁵

The fact that functions such as Epithets can be repeated calls into question a strict interpretation of multivariate structures as comprising elements of structure that only occur once.⁶

 ⁵ Ghesquiere (2014:53) notes that Dixon (1982:25) refers to such structures as involving “independent modification”. Breban (2010: 37–38) distinguishes “classifying adjectives”, which enter into recursive modifications of their head from “descriptive adjectives” which independently modify theirs. Tucker (1998) and Vandelanotte (2002) make a similar distinction between “coordinated adjectives” and “non-coordinated” (or “modifier-sequence”) adjectives. 

 ⁶ From the perspective of orbital structure we can have an unlimited number of Epithets, each modifying a nuclear Thing function.

Blogger Comments:

[1] To be clear, the difference between (13) and (14) is that (13) presents a paratactic word complex as serving one Epithet, whereas (14) presents a hypotactic word complex serving three Epithets, one for each adjective. Both types of representation are used in SFL. Halliday & Matthiessen (2014: 388, 397):

[2] To be clear, as previously explained, it is Systemic Functional Grammar itself that 'calls into question' the Scale-&-Category Grammar 'interpretation of multivariate structures as comprising elements of structure that only occur once'. For example, Halliday & Matthiessen (2014: 364):

Categorisation within the class is typically expressed by one or more of the functional elements Deictic, Numerative, Epithet and Classifier. They serve to realise terms within different systems of the system network of the nominal group.

[3] To be clear, as previously explained, Martin's orbital structure misconstrues the multivariate structure of the experiential metafunction as a hypotactic univariate structure of the logical metafunction. The authors betray this misconstrual here by their use of the term 'modifying', which denotes a hypotactic univariate structure, not a multivariate one. Halliday & Matthiessen (2014: 389):

Misrepresenting 'Canonical' Multivariate Structure

Martin & Doran (2023: 28):

At this point let’s look a little more critically at Halliday’s (2002 [1979]) proposal that multivariate structures realise non-recursive experiential systems and that univariate structures realise recursive logical systems. We’ll address the multivariate structure/non-recursive systems association first, by focusing on structures involving Epithets and Things in nominal group structure. Recall that in a canonical multivariate structure of this kind each variable is distinct and occurs once. As far as the Epithet function is concerned this works fine for Korean (Martin and Shin 2021). Each nominal group is limited to just one Epithet.

The only way to introduce additional description into the picture is via a paratactic word complex realising the Epithet. For example in (12) we have three adjectives forming a word complex realising just one Epithet. Note that the final adjective in the Epithet necessarily takes the suffix -n, linking it to the Thing — as shown in both (11) and (12). In order to add more adjectives Korean deploys the paratactic linker -go, since expanding the number of adjectives involves building a paratactic word complex realising the Epithet. We cannot add additional Epithets to the structure. Thus *빠른 아름다운 혁신적인 자동차 (*ppareu-n areumdau-n hyeoksinjeogi-n jadongcha), where the adjectives have the linking suffix -n rather than parataxis marking -go, is ungrammatical.

Blogger Comments:

[1] To be clear, this obsolete claim is from an exploratory paper by Halliday in 1965, at the time of Scale-&-Category Grammar, when there were no metafunctions, and so, no metafunction structures. In the theory that replaced it, Systemic Functional Grammar, the variable of a multivariate structure is the relationship between the functional elements, not the functional elements themselves. For example, the multivariate structure of the verbal group involves the different relationships that obtain between Finite, Polarity, Auxiliary and Event, but this does not preclude the possibility of there being more than one Auxiliary; Halliday & Matthiessen (2014: 397):

The experiential structure of the finite verbal group is Finite (standing for ‘Finite operator’) plus Event, with optional elements Auxiliary (one or more) and Polarity. Finite verbal groups range from short, one-word items such as ate, where the Finite is fused with the Event and there is no Auxiliary, to long strings like couldn’t have been going to be being eaten (Figure 6-13). 

[2] To be clear, in SFL Theory, it is a linker that marks a paratactic relation. If the suffix -n does not mark a paratactic relation, then it either marks a hypotactic relation, in which case it is a binder, or it marks the function Epithet — or both the relation and the function.

Misconstruing Multivariate And Paratactic Structures As Having "Heads"

 Martin & Doran (2023: 25):

For particulate structure this means bringing nuclearity into the picture and re-interpreting multivariate structures as orbital (with satellites related to a single nucleus) and univariate structures as serial (i.e. multi-nuclear). This additional variable (i.e. nuclearity) allows for the recognition of “heads” in both experiential and logical structures. The canonical “head” in an orbital structure is exemplified by the Thing in nominal groups, the Event in verbal groups or the configuration of Process and Medium in clause structure; the canonical “head” in serial structure is the α variable in hypotactic complexes – e.g. the primary tense choice in an English or Spanish verbal group (Martin et al. 2023) or the projecting clause in a reporting clause complex across languages.

Blogger Comments:

[1] To be clear, as previously explained, Martin's model of structure misconstrues multivariate experiential structure as hypotactic univariate logical structure, and so it is this misconstrual that 'allows for the recognition of "heads" in experiential structures'. However, because Martin's model of structure misconstrues univariate logical structure as paratactic univariate structures, it does not 'allow for the recognition of "heads" in logical structures', since a paratactic structure links elements of equal status, and as such, has no Head element.

[2] This is potentially misleading. To be clear, unacknowledged by Martin & Doran, the notion of Process and Medium as nucleus in the ergative model of the English clause appears in Halliday (1985: 147):

The Process and the Medium together form the nucleus of an English clause; and this nucleus then determines the range of options that are available to the rest of the clause. Thus the nucleus ‘tear + cloth’ represents a small semantic field that may be realised as a clause either alone or in combination with other participant or circumstantial functions.

Importantly, for Halliday, the meaning of 'nucleus' is distinct from the meaning of 'head', which is why he uses the different terms. The nuclear model construes a cline from the most central to the most peripheral. Halliday & Matthiessen (2014: 348):

the nucleus of ‘Process + Medium’ has an inner ring of additional participants as well as an outer ring of circumstances surrounding it: see Figure 5-40.

[3] To be clear, this is an instance of self-contradiction. If Martin models logical structure as serial (i.e. multinuclear), and it is nuclearity that 'allows for the recognition of heads', then, in a logical structure every single nucleus is a head, and there are no satellites.

[4] This is misleading, because it invites the reader to falsely credit Martin et al. (2023) with the insight that the realisation of primary tense serves as the element of the logical structure of a verbal group. Halliday (1985: 177):

The primary tense is that functioning as Head, shown as α.

Misconstruing Hypotactic vs Paratactic Logical Structure As Experiential vs Logical Structure

Martin & Doran (2023: 25, 26):

Inspired by Halliday’s paper³ and by his own work on Tagalog grammar (e.g. 1995a) and English text structure, Martin (1988, 1995b, 1996a, 2000, 2004a, 2008, 2018) proposes the correlations between kinds of meaning and types of structure outlined in Figure 3. For particulate structure this means bringing nuclearity into the picture and re-interpreting multivariate structures as orbital (with satellites related to a single nucleus) and univariate structures as serial (i.e. multi-nuclear).

 

Blogger Comments:

[1] To be clear, Halliday updated his 1979 model in Halliday (1994: 36), and it remains so twenty years later in Halliday & Matthiessen (2014: 85):

[2] To be clear, Martin's orbital structure for experiential meaning is inconsistent with the notion of multivariate structure. Halliday & Matthiessen (2014: 451):

A multivariate structure is a configuration of different functional relationships, like Theme + Rheme, Mood + Residue + Moodtag, or Actor + Process + Recipient + Goal.

Martin's model, however, interprets the multivariate structure of the experiential metafunction as just one type of relationship: that between a nucleus and satellite. To the extent that a satellite is dependent on a nucleus, this is actually a model of a hypotactic univariate iterative structure of the logical metafunction.

On the other hand, Martin's serial structure for logical meaning is partially consistent with the notion of univariate structure. Halliday & Matthiessen (2014: 451):

A univariate structure is an iteration of the same functional relationship … . Such iterative structures are unique to the logical mode of meaning; they are, as noted, formed out of logico-semantic relations.

However, in giving each nucleus equal status, this is only a model of a paratactic univariate iterative structure of the logical metafunction.

In short, Martin's model misconstrues the difference between multivariate experiential and univariate logical structures as the difference between hypotactic and paratactic logical structures.