next up previous contents
Next: Some later developments Up: Linguistics Lecture 2 Previous: Early evidence for transformational   Contents

Syntax and semantics

Chomsky's theories have from the beginning been criticised for their neglect of semantics and pragmatics. It is thought that semantic processing may occur in terms of propositional representations; pragmatics in terms of situational models.

Chomsky's original model (mid-1960s) (see fig. [*]) assumed an Interpretive semantics as follows -

Syntactic component: Deep Structure (Transformations) Surface Structure

DS is ``interpreted by'' Semantic component

SS is ``operated on by'' Phonological component

Syntax and semantics are autonomous, independent systems. Syntactical processing is independent of, and occurs prior to, semantic (and pragmatic) processing.

Fodor's position: there are modules in the linguistic system (as stereoscopy, colour, in the visual system); in particular, syntactic processing is modular (i.e. domain specific, obligatory operation - not consciously controlled, information encapsulation, a consequence of the fact that only end-product is accessible to the central processing system).

In the late-1960s to mid-1970s an alternative model (fig. [*]) proposed a Generative semantics (after much bitter controversy it was finally abandoned - Gardner, MNS.)

Figure: Generative semantics - Liles, 1975, p. 169, fig. 8.2

Generative semantics model: Conceptual structure, i.e., semantic tree-representation (Transformations) Surface structure (Phonological rules) Phonetic structure

DS was ``provided by'' the semantic component. Thus the simple DS was abandoned in favour structures deeper and closer to the semantic representations. (George Lakoff and J. R. Ross Is Deep Structure Necessary?, 1976) Lakoff went on to develop Cognitive Linguistics which denies the autonomy of syntax. See Article 37 by Tomasellow in Bechtel et. al., 1999.

At the same time the interpretive semanticists including Chomsky transferred more of the syntax into the semantic component so that deep structure moved closer to the surface structure and included more lexical elements.

More radically semantic models (eg. Johnson-Laird, Mental Models, 1983) have proposed that a semantic (propositonal) representation is constructed directly without an intervening syntactic representation. Every syntactic rule has a corresponding semantic rule.

Also so with Gerald Gazdar ... syntax is linked to pragmatics.

Within the generative grammar tradition has been the Lexical-functional theory of Joan Bresnan. It has no transformational component and syntactic information is placed in the lexicon. Syntax is linked to lexical semantics.

All such models which assume that semantic and pragmatic processing influence syntactic analysis, violate the modularity condition of Fodor and others.

Semantic processing is more of an interdisciplinary area - may be done next semester. See Richard Montague's formal approach to semantics.

next up previous contents
Next: Some later developments Up: Linguistics Lecture 2 Previous: Early evidence for transformational   Contents