Fawcett's Argument On Hypotaxis vs Embedding [1]

 Fawcett (2010: 327):

(1a, i-iii) He left the room before / after / while they voted.

(3a, i-v) He left the room, then / (,) and then / (,) and they voted.

Firstly, I suggest that all linguists would agree that (la, i) — which was earlier labelled simply (la) — has a close systemic relationship with (la, ii-iii). Notice, though, that one result of recognising this rather obvious fact is to show us that the parallel between (la) and (3a) is less close than one might otherwise assume. 

Blogger Comments:

To be clear, this is the beginning of Fawcett's argument which seeks to show that dependent clauses are better analysed as embedded clauses than as ranking clauses in hypotactic clause complexes. To this end, he aims to show that dependent clauses are more closely related to prepositional phrases serving as clause Adjuncts than they are to ranking clauses in paratactic clause complexes. His method will be to compare expressions of time across the different grammatical domains. 

However, Fawcett misleads by beginning with a false correspondence. While (1a) features expressions of 'earlier' vs 'later' vs 'same' time, (3a) features only expressions of 'later' time.

That is, Fawcett begins his argument that dependent clauses do not parallel their paratactic counterparts by choosing examples that are not parallel, instead of those that are, such as:

But see later posts for further clarifications.

Inconsistencies In The Data Used For Fawcett's Argument

Fawcett (2010: 327):

Here is a set of examples, all of which are potential text-sentences (though these too have been adapted from the original example in order to create 'minimal pairs'). I suggest that they provide a useful basic set of data for exploring this area of the grammar.

(1a, i-iii) He left the room before / after / while they voted.

(2a, i-iii) He left the room before / after / during the vote.

(1b, i-iii) Before / after / while they voted, he left the room.

(2b, i-iii) Before / after / during the vote he left the room.

(3a, i-v) He left the room, then / (,) and then / (,) and they voted.

(3b, i-n) He left the room, but earlier (on) / first/ before that / before his departure / etc. they (had) voted. 

(3c, i-n) He left the room and in that period / during that time / during his absence / etc. they voted. 

(3a, v-n) He left the room, and later (on) / afterwards / after that / after his departure / etc. they voted.

Blogger Comments:

[1] To be clear, Fawcett is arguing the question of hypotaxis vs embedding on the basis of how temporal distinctions are realised in the grammar. However, his argument is unnecessarily complicated by inconsistencies in the organisation of the data in terms of temporal features:

[2] To be clear, (3biv) is inconsistent with the other members of this set, since it construes a concessive relation rather than a temporal relation.

This is because before his departure functions as a circumstantial Adjunct, whereas the others — earlier (on), first, before that — function as conjunctive Adjuncts that realise the temporal relation.

So, in the case of the clause with the circumstantial Adjunct, the expansion relation is expressed only through but, which in this instance, signifies a concessive causal relation; see Halliday & Matthiessen (2014: 478).

[3] To be clear, all the members of (3c) include circumstantial Adjuncts — in that period, during that time, during his absence — and the expansion relation is expressed only through and, which in these instances, signifies positive additive extension; see Halliday & Matthiessen (2014: 471).

[4] To be clear, (3aviii) is inconsistent with the other members of this set, since it construes positive additive extension rather than temporal enhancement.

Again, this is because after his departure functions as a circumstantial Adjunct, whereas the others — later (on), afterwards, after that — function as conjunctive Adjuncts that realise the temporal relation.

So, in the case of the clause with the circumstantial Adjunct, the expansion relation is expressed only through and, which in this instance, signifies positive additive extension; see Halliday & Matthiessen (2014: 471).

Evidence For Hypotaxis vs Embedding

Fawcett (2010: 326-7):

I suggested earlier that a SF linguist should approach a problem of this sort by asking: "What evidence is there that we should give priority in the system networks to one of these relationships over the other?" Specifically, one should first collect together a body of examples that are closely related systemically to each of (la), (2a) and (3a). Then one should examine them to see how far they provide evidence 

(1) that the relationship in (la) is one of 'hypotaxis' rather than embedding, and 

(2) that the grammar does indeed make similar options available when relating clauses 'paratactically' to the options that are available when the relationship is 'dependent' (whether interpreted as 'hypotaxis' or embedding). 

Next, one should make a judgement as to which systemic relations should be given priority in the system networks (i.e., those that model meaning potential within the lexicogrammar), and finally one should make proposals as to how any relationship that has not been modelled in the grammar should be handled in the overall model.

Blogger Comments:

To be clear, (1a) is He left the room before they voted. In SFL Theory, the second clause is analysed as a ranking clause that is dependent on the first clause. In Fawcett's Cardiff Grammar, it is analysed as embedded as Adjunct in a single clause. That is, unrecognised by Fawcett, in his model, the clause is analysed as rankshifted so as to function in a ranking clause.

Importantly, it is not the data that determines whether the clause is dependent or embedded, but the theory, since the clause can be analysed either way, depending on the theory. That being the case, it is a question — if both are consistent with theory — of which theoretical construal has the more explanatory potential.

Fudging The Data

 Fawcett (2010: 326):

(I have slightly altered the wording of Huddleston's examples to create 'minimal pairs' that make the relevant contrasts fully explicit.)

(1a) He left the room before they voted. 

(2a) He left the room before the vote. 

(3a) He left the room, then they voted.

Huddleston's grammatical analysis of (la) — which is broadly similar to mine — is to treat it as a single clause in which the embedded clause before they voted functions as an Adjunct that identifies the 'time position' of the event of 'leaving' by relating it to an event that is already known to the addressee (the 'voting' event), in the same way that before the vote does in (2a). Indeed, Halliday and M&M would agree with Huddleston and me that, when the event of 'voting' is nominalised as in (2a), it serves this function and is therefore an Adjunct. So why, we might ask, do they not also treat before they voted in (la) as an Adjunct? Essentially, their approach is to interpret (la) as a relating of two events (rather than as a 'main' event that is located in time by relating it to another event) — and to claim that this 'relating' can be achieved either 'paratactically', as in (3a), or 'hypotactically' as in the second interpretation of (la).

Blogger Comments:

In SFL Theory, the three instances are analysed as follows:

[1] This is misleading, on two counts. On the one hand, Fawcett's slight alteration of the data does not create 'minimal pairs', and on the other hand, it does not make the relevant contrasts fully explicit. Instead, the contrast it makes explicit is one of time: 'earlier' ((1a) and (1b)) vs 'later' (1c), which is the irrelevant to the distinction between hypotaxis and embedding.

[2] To be clear, this is bare assertion unsupported by reasoned argument.

[3] This is not misleading, because it is essentially true.

Agreeing With Huddleston's Conclusion

Fawcett (2010: 324):

There is much in Huddleston's review, therefore, with which I disagree. However, it will be clear that I share his conclusion that most of the uses to which 'hypotaxis' has been put by Halliday are better handled by a simple embedding relationship (but one by co-ordination; see Section 11.9 of Chapter 11). And I also share his view (though for a set of reasons that only partly overlap with his) that "the constraints this [i.e., the requirement of 'accountability at all ranks'] imposes on the grammar have numerous unsatisfactory consequences" (Huddleston 1988:141). Indeed, I would also agree with his statement (though again for reasons that only partly overlap with his) that

the unsatisfactory nature of the constituent structures given in [IFG] derives in very large measure from their foundation in rank theory. (Huddleston 1988:155)

Blogger Comments:

To be clear, agreeing with someone's conclusion is not reasoned argument. Fawcett has still not provided Huddleston's actual argument, and it is clear from previous posts that Fawcett does not understand hypotaxis, especially with regard to how unit complexes relate to the rank scale.

The rhetorical purpose of providing Huddleston's conclusion before providing his argument is to prime the reader to prejudge the argument through prior alignment with the conclusion.

Misrepresenting Halliday As Theoretically Inconsistent And Intellectually Dishonest

Fawcett (2010: 323n):

¹⁶ As with many other concepts, the idea that all groups can be handled as word complexes is one for which at least some support can be gleaned from Halliday's writings (e.g., "a group is in some respects equivalent to a word complex (IFG pp. 179-80), and "a group is the expansion of a word" (p. 180). Indeed, Halliday sometimes gives the concept of a group as a word complex more weight than the concept of a group as the expression of a semantic unit with its own set of functional elements. One clear case is his treatment of quality groups when they fill the modifier in a nominal group, as in the case of very small in some very small ones (IFG pp. 192 and 194). Thus very small is said to be a 'complex' of two words that are 'hypotactically related' rather than a group of words. Yet the same words would be a group for Halliday if they filled a Complement, as in The egg was very small. We can guess that the reason why he adopts this somewhat inconsistent position is that it at least has the virtue (from his viewpoint but not mine) that it avoids having to recognise such examples as yet another cased of the unwanted phenomenon of 'rank shift' — which, if acknowledged, would be further evidence against the concept of the 'rank scale'.

Blogger Comments:

[1] This is misleading in a trivial way. The idea that all groups can be handled as word complexes does not find "at least some support" in Halliday's writings; it is Halliday's model.

[2] To be clear, with the nominal group, which is Fawcett's chief concern, Halliday gives equal weight to the logical ("word complex") and experiential ("its own set of functional elements") structures. With the verbal group, which is lacking from Fawcett's Cardiff Grammar, the logical structure is of greater value, because it 'embodies the single most important semantic feature of the English verb, its recursive tense system' (Halliday & Matthiessen 2014: 398). It is the groups that are of no concern to Fawcett — adverbial, conjunction, preposition — that are structured only logically.

[3] To be clear, the (ideational) semantic unit that is realised by the group is the element: participant, process, circumstance, relator; see Halliday & Matthiessen (1999: 177-226).

[4] To be clear, quality groups do not feature in Halliday's model.

[5] This is misleading. In such instances, very small  is a subcomplex within the logical structure of the nominal group, corresponding to the Epithet in the experiential structure. Halliday (1994: 194):

To be clear, very small is a "group of words", but in this instance, it does not constitute an entire nominal group (of words).

[6] To be clear, in this instance, very small  is once again a subcomplex within the logical structure of the nominal group, once again corresponding to the Epithet in the experiential structure. The difference in this case is that the Epithet conflates with the Head of a nominal group with no Modifier:

In a functional grammar, where the view is 'from above', it is the function of words that is criterial, not the words as forms ("the same words").

[7] This is misleading, because Halliday does not adopt an inconsistent position, as demonstrated above. Once again, as throughout this book, Fawcett has falsely assumed that he correctly understands Halliday's model.

[8] This is misleading on three counts. Firstly, both examples involve internal bracketing, not rankshift, and so the question of recognising rankshift does not arise. Secondly, rankshift is not an "unwanted phenomenon"; for example, it provides a systematic explanation of how the complexity of written language (lexical density) is brought about. Thirdly, rankshift is not "evidence against the concept of the rank scale" because it is consistent with the notion of a rank scale, with the rank scale providing the yardstick by which to identify the shift.

But it must be said, the really nasty thing here is that Fawcett has falsely accused Halliday of intellectual dishonesty: of trying to conceal what Fawcett wrongly claims to be a problem for the model. In this book of 12 chapters, only 3 were devoted to expounding Fawcett's model, but all 12 were devoted to misrepresenting Halliday's model in a way that favoured Fawcett's argument for his own model.

Misunderstanding Huddleston On Halliday's Group

Fawcett (2010: 322-3):

But Huddleston goes even further, apparently dismissing the concept of the 'group' (or 'phrase') altogether. So it is yet more surprising that he states:

Clause, word and morpheme are commonplace categories in syntactic theory; it is the group, among Halliday's basic ranks, that has no significantly close analogue in non-rank grammars, and it is difficult to see that this innovation has anything to recommend it" (Huddleston 1988:155).

Here Huddleston seems to be denying the value of the concept of the group (or 'phrase'). Presumably he would replace them with various types of 'word complex', each built up of simple 'modifier-head' relationships.¹⁵ Yet he himself uses the concept of 'phrase' throughout Huddleston & Pullum (forthcoming). It is therefore hard to understand what prompted him to make the statement cited above, unless he has moved from that position to a more conventional one in the period since 1988. But if the above remarks stand, he would not only do away with the 'rank scale', but also with any units other than the clause, the word and the morpheme. This would indeed be radical.

¹⁵ This passage reads as if Huddleston is suggesting that the concept of the 'group' (or 'phrase') is an "innovation" by Halliday, whereas it is in fact a commonplace of descriptive grammars — including Huddleston's own current grammar of English. Perhaps the "innovation" is intended to be the placing the group on the 'rank scale'? In any case, the syntax in Huddleston & Pullum (forthcoming) makes full use of familiar concepts such as 'noun phrase', 'adjectival phrase' and 'prepositional phrase' (Huddleston, personal communication 2000).

 

Blogger Comments:

[1] To be clear, here Huddleston uses the logical fallacy of relevance known as argumentum ad populum: a proposition is claimed to be false solely because a majority or many people believe it to be so.

And from another angle, the absence of a component from Newton's Universal Gravitation in Einstein's General Relativity is clearly no argument against Einstein's theory.

[2] To be clear, here Huddleston uses the logical fallacy of relevance known as the argument from incredulity: 'I cannot imagine how it could be true; therefore it must be false'.

[3] This is misleading. To be clear, it is only the group that Huddleston identifies as Halliday's innovation — not the phrase.

[4] To be clear, the innovation is simply Halliday's concept of the 'group'. As demonstrated above and previously, Fawcett does not understand the distinction between group and phrase.

Huddleston's Third Problem For The Rank Scale

Fawcett (2010: 321-2):

The third problem for the 'rank scale' that Huddleston's review brings out is the fact that the completive of a 'prepositional group/phrase' is practically always filled by one or more nominal groups. The problem is that 'phrase' and 'group' are at the same 'rank' on the 'rank scale'. As Huddleston says (1988:154), "for the prepositional phrase always to involve rankshift is contrary to the spirit of rankshift, since rankshift represents a departure from the normal or unmarked state of affairs".

Blogger Comments:

To be clear, if this were really a problem, it would also be a problem for nominal and adverbial groups since the Postmodifier of both groups is almost always realised by a rankshifted unit — "contrary to the spirit of rankshift". Clearly, whether or not rankshift "represents a departure from the normal or unmarked state of affairs" depends on the grammatical environment in which it occurs.

Misrepresenting Halliday (1965) On Hypotaxis

Fawcett (2010: 321):

Finally, we should note that the position taken by Halliday himself on this issue is exploratory rather than dogmatic. He writes that "hypotaxis is in some ways intermediate between parataxis and [...] rankshift; in fact all hypotaxis could be regarded as rankshift" (1965/81:40). And over twenty years later (IFG p. 216) he introduces this aspect of his model with hedging expressions such as "the tendency is [...]", "we shall assume [...]", "we shall interpret [...]".

Blogger Comments:

[1] To be clear, Halliday takes a scientific approach by proposing hypotheses and then providing arguments to demonstrate why and how such hypotheses are valuable.

[2] This is deliberately misleading through selective omission — the logical fallacy known as contextonomy.  What Halliday (1981 [1965]) actually wrote in this pre-Systemic paper was the following: 

Hypotaxis is in some ways intermediate between parataxis and multivariate structure involving rankshift; in fact all hypotaxis could be regarded as rankshift, although this would result in very different bracketing of constituents and would obscure the important respects in which hypotaxis differs from rankshift and resembles parataxis. We may summarise the types of structure dealt with as follows:

Fawcett's Argument Against Expansion And Projection

Fawcett (2010: 321n):

¹² Many of those who accept Halliday's approach to 'paratactic' and 'hypotactic' relations in structure would claim that the categorisations of relations between clauses (and other units) into 'projection' and 'expansion' (and then of the latter into 'elaboration', 'extension' and 'enhancement') are independently self-justifying. But there are in fact different ways of analysing all of these phenomena that many others, including myself, consider to be more insightful. See Section 11.9 of Chapter 11 and Fawcett (1996) and (in press) for examples of these.

Blogger Comments:

[1] This is misleading, because no-one who understands Halliday's model — including its epistemological assumptions — would claim that the logico-semantic relations of expansion and projection are "independently self-justifying". To be clear, the theoretical value of these distinctions lies in their explanatory potential, which is substantial. As Halliday & Matthiessen (1999: 127) point out:

expansion and projection are trans-phenomenal categories in the sense that they are manifested over the system as a whole — not merely in different logical environments across ranks but also experientially.

For example, expansion and projection are manifested in relational processes:

• elaboration as intensive,
• extension as possessive, and
• enhancement and projection as circumstantial.

Similarly, expansion and projection are manifested in circumstances:

• elaboration as Role,
• extension as Accompaniment,
• enhancement as Extent, Location, Manner, Cause, Contingency, and
• projection as Matter, Angle.

Moreover, expansion is manifested in the textual systems of cohesive conjunction and lexical cohesion. In the case of the latter:

• elaboration is manifested as repetition, synonymy, hyponymy,
• extension is manifested as meronymy, and
• enhancement is manifested as collocation (in many cases).

See also Table 10-3 in Halliday & Matthiessen (2014: 670-2).

[2] To be clear, this is a bare assertion unsupported by argument.

[3] To be clear, Section 11.9, How embedding and co-ordination can replace 'hypotaxis' and 'parataxis', clearly has considerably less explanatory potential than the logico-semantic relations of expansion and projection (see [1]). Moreover, clarifications and critiques of the section can again be viewed here:

• How The Cardiff Grammar Handles Hypotactic And Paratactic Projection
• How The Cardiff Grammar Handles Hypotactic Extension And Enhancement
• How The Cardiff Grammar Handles Hypotactic "Elaboration"
• How The Cardiff Grammar Handles Paratactic Expansion
• Summary Of How The Cardiff Grammar Handles Hypotaxis And Parataxis

[4] To be clear, Fawcett (in press) is still unpublished 21 tears after the first edition of this volume.

Seriously Misrepresenting Huddleston On Univariate Structure

Fawcett (2010: 320-1):

The above argument goes rather further in exploring the implications for the concept of the 'rank scale' of Halliday's proposals than does Huddleston's review, perhaps in part because he concentrates on different matters (relations between words and groups, and issues of left and right branching). While I shall not attempt to summarise these here, I shall cite his interesting concluding words. These deserve attention because he was, as we noted in Section 5.1 of Chapter 5, one of the small team (with Halliday, Hudson and Henrici) who worked on these problems in the 1960s. Indeed, it was Huddleston who wrote one of the key S&C papers on this topic (Huddleston 1965/81). Since he was so closely involved, we should give due weight to his statement that

historically, the layered univariate structure was introduced in the context of an attempt to solve certain problems stemming from the total accountability requirement of the rank model. [...] "The problems we have been discussing [layering in 'paratactic' and 'hypotactic' structures] are created by the [rank scale] model". (Huddleston 1988:151)

The clear implication of this revealing statement is that the purpose of introducing the concepts of 'paratactic' and 'hypotactic' relations in structure to the theory was less their inherent insightfulness than to shore up the ailing concept of the 'rank scale'. This is therefore secondary evidence for the position taken here, i.e., that the concept of the 'rank scale' is ultimately not a useful one. Moreover we now have a replacement for it, i.e., the concept of a set of units, each of which is capable of filling several elements of one or more higher units in a tree representation, supplemented for text analysis by filling probabilities.

Blogger Comments:

[1] As previously demonstrated, Fawcett's exploration of the implications on the rank scale proceeds from misunderstandings of unit complexes with regard to the rank scale.

[2] Again, to cite a conclusion without the argument on which it is based is a fallacious use of the argument from authority (argumentum ad verecundiam).

[3] This is very misleading indeed. Huddleston's point is that if a rank scale is assumed, then total accountability at all ranks entails layered univariate structures in the model.  Fawcett misrepresents this as meaning univariate structures were devised primarily as an attempt "to shore up the ailing concept of the rank scale".

[4] This is misleading, because it is untrue. Fawcett has merely misrepresented Huddleston — see [3] — in a way that agrees with his desires.

[5] To be clear, Fawcett ranks his units on a scale from higher to lower, but denies that his model includes a rank scale.

The Case Of The Missing Unit Complexes

Fawcett (2010: 320, 320n):

Is there any reason why Halliday should not include in his representations the 'unit-complexes' that the description implies? The reason why he does not address this question is, I would guess, that the additions to the 'rank scale' that such structures imply would threaten its viability because of the unacceptable quantity of 'singulary branching' that the maintenance of the principle of 'accountability at all ranks' would introduce. They are therefore unwelcome in Halliday's theory, both as 'units on the rank scale' and as the representation of such units on paper.¹⁰

¹⁰ A possible alternative answer might in principle be that, while this would suit 'hypotactic' structures quite well, it would not suit 'paratactic' structures, because co-ordination does not involve a 'modifier-head' relationship (as is recognised in all theories). However, given Halliday's view of the centrality of the 'rank scale' in the theory, it seems more likely that the reason suggested in the main text is the right one.

Blogger Comments:

[1] This is misleading, because it is untrue. In all his representations Halliday does include the unit complexes that the description makes explicit ("implies"). For example, in the following representation (Halliday 1994: 216), the unit complex that is "included" is the clause complex:

[2] To be clear, one possible reason why Halliday does not address this question is that, because he writes at a level to be comprehensible to an intelligent 12-year-old, he would be justified in assuming that, for most readers, the question would not arise.

"The Theoretical Core Of The Problem That 'Unit Complexes' Raise"

Fawcett (2010: 320):

We come now to the theoretical core of the problem that 'unit complexes' raise for the concept of the 'rank scale'. Halliday's descriptions of these additional layers of structure are always in terms of what he terms their "elements", e.g., the 'hypotactic' ones are said to be in a series of 'modifier-head' relationships to each other (IFG p. 217-8), represented by the symbols "α β γ" etc. But these elements are not treated as elements of a unit (as in the case of the modifier and head of a nominal group), but as elements that are somehow able to function in relation to each other without the use of the concept of 'unit'. Yet in SFL the concepts of 'unit' and 'element' are mutually defining. In other words, an element is by definition an element of the structure of something, and in SF theory that "something" is a unit. Yet in IFG we are presented with 'elements' that are shown in the diagrams as relating directly to each other, without any indication of the unit of which the element is a component — or indeed of the unit that fills the element. In other words, the concept of a 'unit' is in fact covertly present, even through it is not referred to at any point.

Blogger Comments:

[1] This is potentially misleading. On the one hand, a hypotactic structure is not a series of "modifier-head" relationships. A hypotactic structure comprises one Head and one Modifier, either of which may nest a hypotactic or paratactic subcomplex. In a clause structured as α β γ, α constitutes the Head and β γ the Modifier, as demonstrated in Halliday (1994: 217). And, on the other hand, modification is only a first step in understanding hypotaxis. Halliday (1994: 218):

As a first step, therefore, we can interpret the relationship between these clauses as one of modification … . The concept of modification needs to be enriched by allowing for systematic alternatives along two separate dimensions: (i) type of INTERDEPENDENCY or TAXIS; (ii) the LOGICO-SEMANTIC RELATION.

[2] This misrepresentation is misleading. To be clear, on the one hand, the elements of a unit complex are the units that are complexed. For example, the elements of a clause complex are clauses. Taxis is a relation between forms (units on the rank scale). This is distinct from elements of a (multivariate) function structure of a unit, such as Predicator as a functional element of clause structure. And on the other hand, these units (elements) relate to each other in the unit complex. Viewed in terms of constituency, a clause complex is the "unit" of which the clauses are "components".

Misunderstanding The Place Of Unit Complexes On The Rank Scale

Fawcett (2010: 319):

However, it makes little difference which way one decides on this matter, because my essential point stands in either case. This is that the effect on the concept of the 'rank scale' of introducing 'parataxis' and 'hypotaxis' is that it greatly increases the number of units on the 'rank scale'. Indeed, since in Halliday's model there can be unit complexes of both the 'paratactic' and the 'hypotactic' types above each basic unit, the number of units on the 'rank scale' is increased to at least twelve. Moreover, since either the 'paratactic' or the 'hypotactic' structure may come above the other (or indeed between two instances of the other) there may be even more layers still. The model with which Halliday's account of 'univariate' structures leaves us is therefore one which has, in principle, potentially very many layers of unit complexes (either 'paratactic' or 'hypotactic' or both) above each of the four basic 'units' of the 'rank scale'.

Blogger Comments:

As previously explained, this is misleading because it is untrue. Parataxis and hypotaxis are relations that obtain between units in unit complexes at one of the four ranks of the rank scale. Here Fawcett explicitly demonstrates the extent to which he does not understand either the rank scale or taxis.

Misunderstanding Huddleston On Minimal Bracketing

 Fawcett (2010: 319):

I agree with Huddleston (1988:148-9), therefore, when he says that "hypotactic univariate structures do not lend themselves satisfactorily to the minimal bracketing principle", i.e., to a representation of such structures as "α β γ" . 

Huddleston devotes four pages to a discussion of 'maximal' vs. 'minimal' bracketing and an examination of Halliday's illustrative analyses (pp. 148-51), and his strongly worded conclusion (which seems to be justified) is that "the α β γ structure is unmotivated and inconsistently applied" (1988:151).

Blogger Comments:

[1] To be clear, the 'minimal bracketing principle' is ranked constituent analysis, which construes formal constituency as a rank scale; see Halliday (1994: 20-8). Hypotactic univariate structures, on the other hand, are complexes of units of the rank scale. There is no inconsistency between the two notions: the former defines the units which are complexed in the latter.

[2] To be clear, here Fawcett glosses 'to the minimal bracketing principle' as 'to a representation of such structures as α β γ'. This demonstrates that Fawcett does not understand Huddleston's argument; see [1].

[3] To be clear, Fawcett does not provide Huddleston's argument. Endorsing Huddleston's conclusion without providing his argument amounts to a fallacious use of the argument from authority. 

Misrepresenting Halliday On Hypotactic Structures

Fawcett (2010: 319):

I should add, however, that Halliday would probably not agree that every additional element in a 'hypotactic' unit complex (after the first two) adds a new layer of structure — even though diagrams such as those referred to above clearly imply that they do. In Halliday (1965/81:34) he says that "a hypotactic structure is better thought of as a chain of dependencies". Indeed, 'box diagrams' such as those in the lower halves of Figures 7-2 and 7-3 on p. 217 of IFG show the elements "α β γ" as a set of adjacent symbols — so implying that they are all elements of the same unit. This in turn raises the question of whether it is possible to have a recursive 'modifier-head' relationship between three elements α, β and γ, because the β element has to function as both a head (to γ) and a modifier (to α), which is arguably illogical. Halliday certainly intends this interpretation (1965/81:36), but I find his reasons for its desirability unpersuasive, and I would analyse all of his examples in terms of embedding.

Blogger Comments:

[1] This is misleading, because it is not true that the diagrams — in Halliday (1994: 216-9) — imply that a complex with more than two units necessarily entails what Fawcett calls 'an additional layer of structure'. Clause complexes may be simply linear — e.g. α β γ — but they may involve the nesting of sub-complexes with the complex. Halliday & Matthiessen (2014: 442):

Many clause complexes are linear sequences … But we also often find internal bracketing, or nesting. This is where what is being linked by a logico-semantic relation is not a single clause but rather a ‘subcomplex’ – a clause nexus in its own right.

We can show nesting in either of two ways. (i) The nesting can be represented explicitly as internal bracketing – e.g. 1 ^ 2(α ^ β); (ii) or it can be represented as a simple string – e.g. 1 ^ 2α ^ 2β.

[2] To be clear, Figures 7-2 and 7-3 both represent instances of nesting:

[3] To be clear, there is no "implication" here. The α β γ symbols in Figures 7-2 and 7-3 symbolise the hypotactic structure of a subcomplex within each clause complex. Each symbol categorises a clause (unit) in the clause (unit) complex. This demonstrates again that Fawcett does not understand hypotactic structures or their diagrammatic representations.

[4] To be clear, the structure α β γ is a "recursive modifier-head relationship". In a linear α β γ clause complex, the dominant α is Head and the dependent units are Modifier. Halliday (1994: 216):

[5] This is misleading, because it is untrue that Halliday intends Fawcett's misinterpretation. In a linear structure, there is no Head-Modifier relation within the Modifier. This only occurs when the Modifier itself is a nested subcomplex, as illustrated in Figures 7-2 and 7-3. It is clear from this that Fawcett does understand the notion of nesting in complexes.

[6] Clearly, Fawcett is unpersuaded by Halliday's argument because he does not understand his model of hypotactic structures, as demonstrated above. This incomprehension explains why Fawcett would analyse all of Halliday's examples in terms of embedding.

Misunderstanding Halliday's Representations Of Hypotactic Structures

Fawcett (2010: 318-9):

But matters are in fact even more complex, because recursive structures can occur within 'unit complexes' of both the 'paratactic' and the 'hypotactic' types (as is illustrated on pp. 216-9 of IFG). Thus there can be more than one layer of structure between each 'basic rank', e.g., as in the three layers of hypotactically related clauses in the two examples in Figure 2 in Section 2.6.1 of Chapter 2, and similarly in Halliday's example of I don't mind if you leave as soon as you've finished, shown in Figure 7-2 on p. 217 of IFG. (This example would have essentially the same representation as (a) in Figure 2.)

Blogger Comments:

[1] To be clear, the only analyses that match Fawcett's description are Figures 7-2 and 7-3 in Halliday (1994: 217):

[2] This is misleading. The layers of structure in a clause complex are at the rank of clause — not "between" ranks.

[3] To be clear, the two examples in Figure 2 present one layer of structure — α β γ— not three.

[4] To be clear, Halliday's Figure 7-2 presents two layers of structure, not three. See [1] above.

[5] This is misleading, because it is untrue. Figure 7-2 presents two layers of structure, whereas Figure 2 presents only one layer. 

∞

It is clear, from the above, that Fawcett is unable to understand Halliday's representations of hypotactic structure.

Seriously Misrepresenting Halliday On Unit Complexes

Fawcett (2010: 318, 318n):

However, so far as the two 'rank scale' concepts of (1) 'accountability at all ranks' and (2) 'rank shift' are concerned, Halliday treats the 'unit complexes' as if they were not part of the 'rank scale'. He does not consider, therefore, that every clause should be analysed as serving a function in a clause complex, that every group should be seen as filling an element in a group complex, and so on. (And yet, as we shall shortly see, this is precisely what he does say, at some points.) Although he does not state in IFG why it is undesirable to treat 'unit complexes' as 'units' on the 'rank scale', we can infer that the reason is the additional layers of 'singulary branching' that would occur — because one of his reasons for introducing 'hypotaxis' in 1965 was to avoid the "somewhat artificial increase in 'depth' in number of layers [introduced by embedding]." ⁸

⁸ A defender of Halliday's position might be tempted to offer a modified model of the standard column of units on the 'rank scale', in which each type of unit complex was placed beside its equivalent basic unit rather than above it. But this would not resolve the problem, because it would leave the relationship between a 'unit' and its equivalent 'unit complex' undefined. There is in fact no alternative, in Halliday's framework, to accepting unit complexes as additional units on the 'rank scale'.

Blogger Comments:

[1] This is misleading, because it is the opposite of what is true. In SFL Theory, a clause complex is a complex of clauses, and 'clause' is a unit on the rank scale; a group complex is a complex of groups, and 'group' is a unit on the rank scale; etc.

[2] This is misleading. On the one hand, in SFL Theory, clauses in a clause complex do not function as elements of a clause complex because a clause complex is not a higher rank than the clause. Likewise, groups do not function as ("fill") elements of a group complex because a group complex is not a higher rank than the group.

On the other hand, in SFL Theory, a secondary clause in a clause complex functions either as an expansion of the primary clause in terms of elaboration, extension or enhancement or as a projection of it. Likewise, a secondary group in a group complex functions either as an expansion of the primary group in terms of elaboration, extension or enhancement or as a projection of it.

[3] This is misleading, because here Fawcett falsely implies an inconsistency in IFG where there is none.

[4] To be clear, Halliday frequently advised his students to write at a level that could be understood by an intelligent 12-year-old. We can infer from this that he assumed that an intelligent 12-year-old would understand that a unit complex is a complex of rank scale units.

[5] This is not only irrelevant, given the above, but also misleading. In this pre-Systemic paper, there can be no singulary branching in hypotactic structures. Halliday (2002 [1965]: 235):

In constituent terms, all hypotactic structures can in fact be represented as binary; that is, as having no more than two constituents at a single layer.

[6] This is very misleading indeed, because the quote from Halliday (1965) is not concerned either with Halliday's reason for introducing hypotaxis or with embedding. The "somewhat artificial increase in 'depth' in number of layers" is a disadvantage that results from treating hypotactic structures 'as having no more than two constituents in a single layer' (see [5]). And it is for this reason that Halliday (2002 [1965]: 235) instead proposes:

In the present analysis, hypotactic structures are not being treated as exclusively binary, but are considered as capable of extension on one layer as well as by layering one within another.

[7] As previously explained, unit complexes are complexes of units located on the rank scale. A clause complex is a univariate structure at the rank of clause.

[8] To be clear, in SFL Theory, there is no "relationship" between a unit and its equivalent unit complex, except in the tautological sense that one is a complex of the other. On the other hand, units in a unit complex are related in terms of interdependency and logico-semantic types.

[9] This is misleading, because it is untrue. The alternative "to accepting unit complexes as additional units on the 'rank scale' " is simply understanding that unit complexes are complexes of rank scale units.

Misrepresenting Halliday On Unit Complexes

 Fawcett (2010: 318):

Firstly, then, we need to note that in Halliday's current model any such "unit complex" is located above the equivalent "basic unit" on the 'rank scale'. (These terms were introduced by Huddleston himself, in his important contribution in Huddleston 1965/81:46.) In other words, Halliday's model states 

(1) that above the clause there is the 'clause complex' (the term "sentence" being permitted as an equivalent when referring to written rather than spoken text); 

(2) that between the clause and the group/phrase there is the 'group/phrase complex'; 

(3) that between the group/phrase and the word there is the 'word complex'; and 

(4) that between the word and the morpheme there is the 'morpheme complex'. 

In principle, then, the introduction of 'unit complexes' to the theory doubles the number of 'units' on the 'rank scale'. Moreover, there is no doubt that a unit complex is a type of 'unit' in Halliday's theory, because he refers to it as a 'unit' — writing, for example, that "the clause complex is the only grammatical unit that we shall recognise above the clause" (IFG p. 216).

Blogger Comments:

[1] This is misleading, because it is not true. A unit complex is the structural relation of units at the rank of the unit:

• a clause complex is the structural relation of units at clause rank;
• a group/phrase complex is the structural relation of units at group/phrase rank;
• a word complex is the structural relation of units at word rank; and
• a morpheme complex is the structural relation of units at morpheme rank.

[2] To be clear, in SFL Theory, sentence is a unit on the stratum of graphology, not lexicogrammar. Halliday & Matthiessen (2014: 436):

… the clause complex is realised graphologically as a ‘sentence’, in the way that this has evolved, over the centuries, as a unit in the written language. The sentence is the highest unit of punctuation on the graphological rank scale and has evolved in the writing system to represent the clause complex as the most extensive domain of grammatical structure. We will use the term sentence to refer only to this highest-ranking graphological unit of punctuation.

[3] This is misleading, because it is not true. As explained above, unit complexes are located at the rank of the units in the complex.

[4] To be clear, a unit complex is a unit in the sense of unit with univariate structure.

[5] To be clear, in SFL Theory, a clause complex is not 'above' the clause in terms of the rank scale, as demonstrated, for example, by the location of the clause complexing system at clause rank; Halliday & Matthiessen (2014: 438):

Misrepresenting Halliday On Layers Of Structure

Fawcett (2010: 317, 317n):

Secondly, Huddleston discusses in considerable detail certain problems that result from the concept of 'layered univariate structures' within 'parataxis' and 'hypotaxis'. We first met these concepts in Section 2.6.1 of Chapter 2. Figure 2 in that section (which is taken directly from Halliday 1965/81) illustrates in a simplified form the concept that 'hypotactic' structures add new layers to the overall structure. Moreover, the top halves of the diagrams in Figures 7-2 and 7-3 on p.217 of IFG similarly suggest that each 'hypotactic' relationship adds a further layer of structure. And the 'paratactic' structures of Halliday's model, with their elements of "1 2 3", necessarily also add more layers to the overall structure.

Blogger Comments:

[1] To be clear, Fawcett does not identify any of the"problems" resulting from layered univariate structures that Huddleston raises. He does, however, provide his own arguments, but as will be seen, these do not withstand close scrutiny.

[2] This is misleading. To be clear, Figure 2 only displays one level of structure: α β γ. Fawcett here simply misinterprets clauses displayed on three different lines as three levels of structure.

[3] This is misleading. To be clear, Fawcett here simply misinterprets clauses displayed on different lines in the top halves of Figures 7-2 and 7-3 as multiple levels of structure. On the other hand, the layers of structure are explicitly represented — not suggested — in the bottom halves of Figures 7-2 and 7-3.

[4] This is misleading. Clearly, the paratactic structure 1 2 3 construes only one level of structure.

Huddleston's Suggestion That Halliday Should Treat Hypotactic Clauses As Embedded

Fawcett (2010: 317):

Firstly, Huddleston suggests that in He assumed that she was guilty, the clause that she was guilty functions as a Complement, just as too much is a Complement in He assumed too much. Similarly, he suggests that in He left before the vote was taken, the clause before the vote was taken functions as an Adjunct, just as before the debate does in He left before the debate. Huddleston's point is that Halliday should treat the dependent clause in such examples as functioning as an element of the matrix clause (i.e., as embedded) and not as a clause that is 'hypotactically' related to the rest of the main clause in a 'modifier-head' relationship. The relevance of this for the 'rank scale' is that, if his position is accepted, the amount of 'rank shift' in the grammar is thereby increased enormously, and the predictions made by the 'rank scale' concept are consequently weakened. We shall shortly consider more closely both Huddleston's reasons for taking the position that he does on this matter, and Matthiessen and Martin's reply. As you may have noticed, Huddleston's position is essentially the same as my own, as described in Section 11.9 of Chapter 11.

Blogger Comments:

 [1] To be clear, merely reporting Huddleston's suggestions, without the reasoning on which they are based, is not argument.

But note that the above clause that she was guilty can interpreted as a Complement in SFL Theory, if it is construed as a pre-projected fact: He assumed (the fact) that she was guilty. Importantly, the distinction between embedding and hypotaxis provides the means of distinguishing between pre-projected facts and projected ideas, whereas the approach advocated by Huddleston and Fawcett does not.

[2] This is misleading. The rank scale derives from taking a functional approach to formal constituency: the minimal bracketing of ranked constituent analysis (Halliday 1994: 20-8). Its "predictions" are not weakened by any increase in the amount of rankshift. On the contrary, the rank scale provides the principled means of accounting for rankshift in functional terms.

[3] This is misleading, because it is not possible to look more closely at reasons which have not yet been provided. That is, Fawcett falsely implies that the bare assertions he provides here, from Huddleston, are reasoned arguments.

[4] To be clear, Fawcett's Section 11.9 is titled 'How embedding and co-ordination can replace hypotaxis and parataxis'. Clarifications and critiques of the section can be viewed here:

• How The Cardiff Grammar Handles Hypotactic And Paratactic Projection
• How The Cardiff Grammar Handles Hypotactic Extension And Enhancement
• How The Cardiff Grammar Handles Hypotactic "Elaboration"
• How The Cardiff Grammar Handles Paratactic Expansion
• Summary Of How The Cardiff Grammar Handles Hypotaxis And Parataxis

Fawcett's Reasons For The Lack Of Discussion Of Solutions To Problems In SFL

Fawcett (2010: 312-3, 313n):

We need to ask why there has been so little discussion of alternative systemic functional solutions to problems in SFL. The reason is partly the example that is set by Halliday himself. He, like many others, has a strong dislike for the type of supposedly 'hard-nosed' combative argumentation that was so popular in the heyday of Chomskyan linguistics. This may be at least part of the reason why Halliday has only rarely responded to criticisms, and why he hardly ever comments adversely on alternative proposals from within SFL — and so why there is so little 'debate' in SFL. 

On the rare occasions when he does reply to a criticism, his typical response is to concede courteously that the point needs consideration (as he did with respect to Matthews' idea of treating Linkers such as and as "markers"), while at the same time continuing to assert the value of the original concept. The problem is that, with the passage of time and the repeated re-presentation of the original concept (both his own works and, often, in the various introductions to his ideas by others) the criticism gets forgotten and the original concept, despite its weaknesses, survives.

Blogger Comments:

[1] To be clear, the reason there has been so little discussion of the problem of the rank scale is that the only people who think it a problem are those who cannot understand it. On the other hand, alternative solutions to non-problems have been proposed, not only by Fawcett, but also by others, most notably Martin.

[2] This speculation confuses debate, in general, with one particular type of debate — the type exemplified by Matthews (1966) and Fawcett (2010: 238n, 256n):

… its existence is therefore an embarrassment for the 'rank scale' concept …

We might note that the data that we are about to consider are yet another serious source of embarrassment for the concept of the 'rank scale'.

[3] This is misleading, because it is untrue. When Halliday later formulated Systemic Functional Grammar, he did indeed treat these as structure markers. Halliday & Matthiessen (2014: 611):

The logico-semantic relation is marked by a conjunction – either by a non-structural one that is used only in this way, i.e. only cohesively, such as for example, furthermore, consequently; or by a structural one whose prototypical function is to mark the continuing clause in a paratactic clause nexus. … the latter are simply analysed as structure markers and are obligatorily thematic as structural Theme.

[4] To be clear, as this blog has demonstrated, this is not true in the case of either the rank scale or structure markers.

∞

It might be mentioned that Fawcett and his fellows are not so keen to have Fawcett's "solutions" discussed. See, for example, here.

Misrepresenting IFG On Conjunction Groups

Fawcett (2010: 311-2):

We have seen that, in his response to Matthews' comments, Halliday (1966) allows that Matthews may have a valid point with respect to Linkers such as and. In IFG, however, he makes an alternative proposal (p. 211). He introduces a new class of group, the 'conjunction group', which is to fill a Linker or Binder. He is right that this is needed (at least for Linkers), but one wonders whether the proposal has the additional attraction of enabling him to handle Linkers within the 'rank scale' rather than as 'markers', and so to defend the original 'rank scale' concept. However, while Binders occasionally require an internal structure (as Appendix B shows), Linkers do not. (I assume here that Halliday would treat and so, etc. as a single item, as I would). Halliday's new class of group does not help here, nor does it help with Adjuncts that express logical relations such as therefore and however. Moreover, the structure that he suggests for the 'conjunction group' is simply "β α". The problem here is that this makes it a 'hypotactic' relationship, so that an example such as immediately after is treated as a 'word complex' rather than a group. In other words, such a structure does not constitutes a group in Halliday's theory, but a 'unit complex' that occurs between a simple word and a simple group.

Blogger Comments:

[1] This is misleading, because it is untrue, since this is not an alternative proposal. Linkers and binders are classes of conjunctions, which serve as the Head of conjunction groups, which can function as structure markers in continuing clauses of a clause complex.

[2] This misunderstanding is misleading, because it is untrue. For Halliday, linkers and binders are word classes, classes of conjunction, whereas conjunction groups consist of conjunctions and serve as structural Themes or conjunctive Adjuncts at clause rank. In short, group rank units — conjunction groups — don't realise ("fill") word rank units — linkers and binders — except where rankshift is possible.

[3] This is misleading because it misrepresents theorising with the rank scale as defending the rank scale.

[4] To be clear, in SFL Theory, binders and linkers are classes of words, which therefore consist of morphemes. Any internal structure of a word is a configuration of functional elements served by morphemes.

[5] To be clear, this non-sequitur is a bare assertion unsupported by argument. Moreover, it is demonstrably untrue, if only because conjunction groups are proposed as the formal units that serve the clause function of conjunctive Adjunct.

[6] This is misleading, because it is untrue. Halliday (1994: 211):

Conjunctions also form word groups by modification, for example even if, just as, not until, if only. These can be represented in the same way, as β ^ α structures (or α ^ β in the case of if only). Note however that many conjunctive expressions have evolved from more complex structures, e.g. as soon as, in case, by the time, nevertheless, in so far as. These can be treated as single elements without further analysis. They are themselves, of course, subject to modification, e.g. just in case, almost as soon as.

To be clear, in SFL Theory, the 'Head Modifier 'structure is the logical structure of all groups: nominal, verbal, adverbial, conjunction and preposition. Halliday & Matthiessen (2014: 451):

Parataxis and hypotaxis are general relationships that are the same throughout the grammar: they define complexes at any rank (clause complex, group or phrase complex, word complex; in addition hypotaxis defines the logical organisation of groups.

The logical structure models the group along the lines of a word complex (Halliday & Matthiessen 2014: 362), which is why they are called groups (of words), not phrases. By the same token, it is because prepositional phrases cannot be modelled as logical structures that they are called phrases, not groups.

Misrepresenting Halliday (1966) On Reasons For The Value Of The Rank Scale

 Fawcett (2010: 311):

However, the main thrust of his defence of the 'rank scale' is to make a set of claims for the general insightfulness of working with the concept of the 'rank scale' — as a means of relating units and their associated system networks to each other, as a hypothesis that raises interesting questions about language, and as a tool that may be useful in fields of applied linguistics such as translation and others that involve text analysis. 

In other words, the defence is an empirical one: if the concept proves to be useful in advancing our understanding of how language works, it is a hypothesis that is worth making. And it has to be said that, whatever one's view of the 'rank scale' today, there is little doubt that, for those doing descriptive work in the theory and those using it in various fields of application in the 1960s and 1970s, the 'rank scale' did indeed provide a framework for describing languages that was an advance over unrestricted constituency relations. And, despite the drawbacks of its excessively strong claim, it has for several decades provided one dimension of the 'matrix' model of Systemic Functional Grammar (the other dimension being Halliday's four 'metafunctions', i.e., the 'experiential', the 'logical', the 'interpersonal' and the 'textual'). 

From this viewpoint, it may seem to some that it is 'politically' important to defend the 'rank scale' concept.

 

Blogger Comments:

[1] This is potentially misleading. To be clear, what Halliday (2002 [1966]: 120) argued was as follows:

We must then ask whether it is worth making in the first place, and basically there are two grounds for thinking that it is: its descriptive advantages, and the questions that follow from it. Among these, I suggest, are the following.

It defines a point of origin for structures and systems, so that the assignment of any item to a given rank, as also the assignment of the structures and systems themselves, becomes an important step in generalisation. To show that a system operates at a given rank is the first step in stating its relationship to other systems; likewise to assign an item to a given rank is the first step in stating the systemic and structural relations into which it may enter and those which it may embody within itself. On the structure axis, rank is a form of generalisation about bracketing, and makes it easier to avoid the imposition of unnecessary structure. It also serves frequently to distinguish between similar structures, for example between defining and non-defining relative clauses in English. It may contribute towards a significant measure of depth (Huddleston 1965). It provides a point of reference for the description at other levels, such as phonology. These and other considerations suggest to me that the rank hypothesis, if valid, leads to a gain in descriptive power.

Among the further questions that would follow from it are these. If some such form of hierarchical organisation is universal, is the number of units also a universal, or is it a typological variable (in either case it is of interest)? Are certain paradigmatic or syntagmatic relations universally associated with specific ranks? Is there any statistical association among the relative frequencies of items of different rank? Is there any type of aphasia characterised by progressive, rank-by-rank loss of grammatical structures? Is there any reason why different languages have institutionalised different grammatical units in their orthographies or the same unit in different ways?

[2] This is misleading, because it is not true that "the defence is an empirical one" since empirical means 'based on what is experienced or seen rather than on theory'.

[3] To be clear, 'excessively strong' is an interpersonal Epithet, merely an enactment of Fawcett's attitude, instead of reasoned argument, supported by evidence.

[4] This is misleading, because it falsely reduces the architecture of SFL Theory to just two dimensions: rank and metafunction. The four dimensions unknown to Fawcett — stratification, instantiation, axis and delicacy — are included with these two by Halliday & Matthiessen (2014: 32):

[5] This is very misleading indeed, not least because the word 'political' can mean 'done or acting in the interests of status or power within an organisation rather than as a matter of principle'. 

To be clear, Halliday (1966) provides reasons why grammars with a rank scale have an explanatory advantage over rank-free grammars.