Immediate Constituent Analysis vs Ranked Constituent Analysis

Halliday (1994: 27-8):

Chomsky's 'Generative' Grammar

Halliday (1994: xxviii):

How Formal Syntax Obscures The Different Perspectives On The Process

Halliday & Matthiessen (1999: 466-7):

The difference between the two perspectives has been obscured for two reasons. In the first place, syntacticians have continued to tend to fail to recognise the unity of the verbal group, favouring instead a Predicate-based constituent, the Verb Phrase, with the Auxiliary then detached within it or located outside it altogether; furthermore, parts of the verbal group complex have been exported to a higher node (e.g. seem, continue). In the second place, the preference for immediate constituency rather than rank-based models of constituency effectively masks the generalisations that can be made about any one rank.

The View That Syntax Distorts Semantics: Universalist Arguments

Halliday & Matthiessen (1999: 447, 447n):

Such analyses ['surface x is really deep y'] were often supported by universalist arguments such as "negation is a verb in certain languages, so it is reasonable to claim that it is really a verb in all". … they tended to make deep structure, the 'real' structure, look like predicate logic. Surface structure came out looking like a (transformationally) twisted version of logical structures. But predicate logic had been derived from one particular area of the grammar, a simplified version of the experiential aspect of the clause; it could be used as an idealised model of certain types of figure, for the purpose of explicit rule-based reasoning, but it was not intended to be a tool for analysing the entire semantic structure of a natural language.² This view has largely been abandoned and the notion of a semantically irresponsible surface structure is no longer generally held.

 

² In terms of our model of a stratified metalanguage, we can see that taking the categories of predicate logic (or any other logical systems) to be linguistic ones constitutes a strata! slippage: categories from the level of representation in the metalanguage are imported into the theoretical account of the object language.

Blogger Comments:

In Chomskyan linguistics, this is the relocation of the universals of Platonic realism — which are said to exist outside and independently of human minds — into the res cogitans of Cartesian dualism.

(World-Oriented) Formal Semantics Viewed From (Mind-Oriented) Cognitive Semantics: Lakoff

Halliday & Matthiessen (1999: 423):

Writing from a different orientation within cognitive semantics, Lakoff (e.g. 1988) is also very critical of what he calls "objectivist metaphysics" (metaphysical realism). He presents a detailed critique of this position, also noting the problem that arises if meanings are located in the world:

To view meaning as residing only in the relationship between symbols and external reality is to make the implicit claim that neither colour categories, nor any other secondary category, should exist as meaningful cognitive categories. Yet colour categories are real categories of the mind. They are meaningful, they are used in reason, and their meaning must be accounted for. But the mechanism of objectivist cognition cannot be changed to accommodate them without giving up on the symbolic category of meaning. But to do that is to abandon the heart of the objectivist program, (p. 132)

(World-Oriented) Formal Semantics Viewed From (Mind-Oriented) Cognitive Semantics: Jackendoff

Halliday & Matthiessen (1999: 423):

Jackendoff (1983: Ch. 2), working within cognitive semantics, argues that the account of 'aboutness' in formal semantics is simplistic:

... I will take issue with the naive (and nearly universally accepted) answer that the information language conveys is about the real world, (p. 24) ... If indeed the world as experienced owes so much to mental processes of organisation, it is crucial for a psychological theory to distinguish carefully between the source of environmental input and the world as experienced. For convenience, I will call the former the real world and the latter the projected world (experience world or phenomenal world would also be appropriate), (p. 28) ...  

It should now be clear why we must take issue with the naive position that the information conveyed by language is about the real world. We have conscious access only to the projected world — the world as unconsciously organised by the mind; and we can talk about things only insofar as they have achieved mental representation through these processes of organisation. Hence the information conveyed by language must be about the projected world. We must explain the naive position as a consequence of our being constituted to treat the projected world as reality.

According to this view, the real world plays only an indirect role in language: it serves as one kind of fodder for the organising processes that give rise to the projected world. If this is the case, we must question the centrality to natural language semantics of the notions of truth and reference as traditionally conceived. Truth is generally regarded as a relationship between a certain subset of sentences (the true ones) and the real world; reference is regarded as a relationship between expressions in a language and things in the real world that these expressions refer to. Having rejected the direct connection of the real world to language, we should not take these notions as starting points for a theory of meaning. Thus an approach such as that of Davidson (1970), which attempts to explicate natural language semantics in terms of Tarskian recursive theory of truth, is antithetical to our own inquiry, (pp. 29-30)

The projected world in Jackendoff's account is the result of creative acts of perception: it is constructed as a model of sensory input, but with the significant addition of information from the perceptual system itself.

(World-Oriented) Formal Semantics Viewed From (Mind-Oriented) Cognitive Science

Halliday & Matthiessen (1999: 422-3):

It would seem that formal semantics is quite far from the concerns of cognitive science; but formal semantics is often carried out within the broad program that cognitive scientists adopt. Writing from a cognitive standpoint, Johnson-Laird (1983) assesses model-theoretic semantics as follows:

The power of model-theoretic semantics resides in its explicit and rigorous approach to the composition of meanings. ... Some idealisations definitely complicate the 'ecumenical' use of model-theoretic semantics as a guide to psychological semantics. One such idealisation is the mapping of language to model without any reference to the human mind, and this omission gives rise to certain intractable difficulties with the semantics of sentences about beliefs and other such prepositional attitudes. These difficulties are readily resolved within the framework of mental models, (p. 180)

... model-theoretic semantics should specify what is computed in understanding a sentence, and psychological semantics should specify how it is computed, (p. 167)

Johnson-Laird's own psychological approach is one based on mental models.

Formal Semantics: The Model Theoretic Approach To Meaning

Halliday & Matthiessen (1999: 422):

The model theoretic approach to meaning is a transcendent one: meanings are located in the world outside of language. This is seen as a strength by the proponents of the model theoretic approach. Dowty et al (1981) criticise various other theories of meaning for ignoring the "inherent aboutness of language", its relationship to the real world:

Any theory which ignores this central property, it is argued, cannot be an adequate theory of natural language. Examples would be theories which, in effect, give the meaning of a sentence by translating it into another language, such as a system of semantic markers or some son of formal logic, where this language is not further interpreted by specifying its connection to the world. The approach of Katz and his coworkers seems to be of this sort (Katz and Fodor, 1963; Katz & Postal, 1964), as is that of Jackendoff (1972) and of the framework of Generative Semantics (Lakoff, 1972; McCawley, 1973; Postal, 1970).

Formal Semantics: Meaning As Extensional And Intensional

Halliday & Matthiessen (1999: 422):

In modelling meaning, formal semantics models it as both extensional and intensional: extensional meaning is cast in terms of sets of entities in a model of some world, and intensional meaning is the relation between a linguistic expression and its extension in any possible world. Formal semantics has thus been concerned with how semantic representations can be related to some extra-linguistic world model. Dowty et al (1981: 5) emphasise the relationship between language and the world in their introductory characterisation of truth conditional semantics:

"... truth conditional semantics, in contrast to other approaches mentioned [including Katz & Fodor, Jackendoff, and generative semantics, MAKH & CM], is based squarely on the assumption that the proper business of semantics is to specify how language connects with the world — in other words, to explicate the inherent "aboutness" of language."

This is, of course, an ideational orientation: the focus here is on representational meaning only. One central method has been to build models of the world in set theoretic terms, and to relate these to linguistic expressions — the model theoretic approach.

Formal Semantics As Syntagmatically Organised

Halliday & Matthiessen (1999: 422):

In its relation to syntax, the formal approach to semantics is compositional, being formulated in terms of syntagmatic organisation: the meaning of larger syntagmatic units such as sentences is built up as functions of smaller units such as nouns and verbs. This semantic composition is in tandem with the syntax, so what becomes important is the nature of the syntactic rules themselves, not just the result of their application. Montague used a form of categorial grammar in formalising the rules of syntax for this purpose.

The World-Oriented Tradition: The Origin Of Formal Semantics

Halliday & Matthiessen (1999: 419):

The development of logic is of interest in itself in the present context as a tradition of 'semiotic design' whereby logical systems were developed out of the resources that have evolved in language. This design of logical systems was earned out to meet certain restricted purposes having to do with reasoning; but once such systems had been developed, philosophers and linguists turned them back on language and used them as systems for representing and theorising meaning. In the present century, when formal semantics was developed for underpinning logical systems, this was then taken as a model for work on the semantics of natural languages.

A Systemic Grammar

Halliday & Matthiessen (1999: 3-4):

A systemic grammar is one of the class of functional grammars, which means (among other things) that it is semantically motivated, or "natural". In contradistinction to formal grammars, which are autonomous, and therefore semantically arbitrary, in a systemic grammar every category (and "category" is used here in the general sense of an organising theoretical concept, not in the narrower sense of 'class' as in formal grammars) is based on meaning: it has a semantic as well as a formal, lexicogrammatical reactance. [The reactance of a category is its distinctive treatment.]

Looked at from the formal angle, of course, this means that it is likely to appear complex; many of the categories are "cryptotypic", manifested only through a long chain of realisations (a "realisational chain"). Hence it takes a long time and a great deal of effort to get such a grammar off the ground in any context (such as natural language processing) requiring total explicitness. Once airborne, however, because it is semantically natural it has considerable potential as the basis on which to represent higher level organisation — provided, that is, such organisation is interpreted in linguistic terms, as meaning rather than as knowledge.

The Functional Status Of Quoted And Reported Clauses

Halliday & Matthiessen (2014: 303):

What about the function of … to keep quiet, it’s half past ten? In formal grammar what is said is treated as a ‘noun clause object of the verb say’, meaning a clause that is rankshifted by nominalisation. But functionally this clause is not rankshifted; it functions as the secondary clause in a ‘clause complex’, being either (a) directly quoted, as in (he said) ‘I’m hungry’, or (b) indirectly reported, as in (he said) he was hungry. This means that such sequences consist of two clauses. (Only the primary clause is a ‘verbal’ one, of course; the other may be a process type of any kind.)

Formal-Cognitivist Alliance

Halliday & Matthiessen (1999: 596n):

Alongside this cognitivist approach, there is a material one embodied in formal approaches to semantics, where the ‘aboutness’ of linguistic expressions is taken as central and these expressions are interpreted in terms of models of possible worlds. However, in this respect there is a formal-cognitive alliance: meaning is interpreted not as something in its own right but as something outside language, either a mental construct (concepts, ideas etc) or a material one (referents in the real world or a formal model of a possible world).

Systemic Grammar Vs Formal Grammar

Halliday & Matthiessen (1999: 3-4):

A systemic grammar is one of a class of functional grammars, which means (among other things) that it is semantically motivated, or “natural”. In contradistinction to formal grammars, which are autonomous, and therefore semantically arbitrary, in a systemic grammar every category (and “category” is used here in the general sense of an organising theoretical concept, not in the narrower sense of ‘class’ as in formal grammars) is based on meaning: it has a semantic, as well as formal, lexicogrammatical reactance. [The reactance of a category is its distinctive treatment.]