Syntactic Theory


The T-Model

Language can be conceived of as an association between form and meaning. When we speak, there is clearly a sound side (In sign languages there is no sound side, but there is clearly a form side - the visual sign. The sound side of language is studied in Phonology) and a meaning side (which is studied in Semantics).

In GB, this elementary fact about language is captured through saying that utterances and sentences have a phonological form (PF) and a logical form (LF). However, language seems to be more than that. It does not seem to be unreasonable to pose that all languages have something like a morphology (even though this is controversial), and it is certainly true that all languages have a syntax.

In the early 1980 a model was developed that suggests that sentences are represented differently (that is, formally they look differently) at different stages of their generation. The assumption is that there is a state that sentences start out with, and that they unfold piece by piece to be audibly and comprehensively uttered in the end. Different levels of representation were proposed, namely Deep Structure (DS) and Surface Structure (SS), in addition to LF and PF. All of these levels are related by a general rule, called Move which says: "Move anything anywhere". This rule sounds very general and broad and does not seem to be appropriate for something that is structured in a certain way like natural languages seem to be. Nevertheless, it makes sense, since it interacts with a number of restrictions and requirements to make sure that only well-formed sentences will be uttered in the end.

Because of its T-like shape the model is called "T-Model" or, if you turn it around for obvious reasons, "Y-Model":

Note that this model is supposed to capture the generation of one sentence at a time. Different modules kick in at different stages of the derivation of a sentence. Starting with DS, -roles are distributed and X-Bar-Theory makes sure that phrases project in a uniform fashion. Then, at SS, Case-theory makes sure that all structural cases are assigned to phonologically realized NPs so as not to violate the Case Filter. Note that movement needs to take place between DS and SS, since the discharging of thematic roles is a very local matter whereas the assignment of Case can be a non-local matter (thinking, for example, of NP-movement or, more drastic, Wh-movement). At SS traces and other phonetically empty elements are still represented. As these elements are not spelled out, PF is "blind" to them. On the other hand, we have elemtens such as explitives which are certainly represented at PF, but LF cannot "see" them, since they don't refer, i.e. they don't have a referential meaning. Thus there are elements represented at SS which are not present either at LF or at PF. Both Quantifiers such as every and some and Wh-words such as who and what are assumed to undergo movement from SS to LF, so that they have a particular scope at LF which they don't have in their linear order (i.e. in the order at which they are spelled out). Within GB, there used to be a controvery as to where (i.e. at which level of representation) binding theory applies.

You can compare the T-Model to an assembly line in a factory, say, a car factory: in the assembly line each worker (or robot) is supposed to fulfill a very specific task - no more, no less. One worker puts wheels on the carriage, another one adds the motor, again another one puts the chairs in, and so on. Every time some component has been added to the body, the assembly line carries it to the next worker. In a similiar fashion, a sentence is consecutively assembled in the T-Model (the linguistic model, of course - which is not to be confused with the famous car by an American car manufacturer ... ): At each point a specific device - a theory - fulfills a certain task. One theory takes care of the discharging of thematic roles, another one takes care of the proper scheme of phrases, another one makes sure that each overt NP is assigned Case, and so forth. Note that the order of theory-application is crucial: -theory must apply before Case-theory in order for the machine to properly work (you don't want the wheels to be added to a car prior to the axle). Furthermore, within GB it has been assumed that multiple theories apply at a single level of representation.

The following example of a passive sentence shows you how it works; it is a derivation that starts with the lexicon and ends at LF. The PF-branch is left out, since the discussion as to what the morphology does is a seperate one:

 1) lexical item, part of the Lexicon

see V, [---NP]

 2.1) X-Bar (DS)

[VP see-n [NP every cat]]

 2.2) -roles (DS)

(the internal argument of see receives the PATIENT-role; no external argument due to passive morpheme -n; in accordance with Burzio's Generalization, the NP does not receive Accusative-Case.)

[VP see-n [NP every cat]]

 3.1) EPP (SS)

[IP [VP be [VP see-n [NP every cat]]]]

 3.2) Case-Theory (SS)

(the NP moves from COMP,V to SPEC,I to receive Nominative-Case by finite I)

[IP [NP every cat]i [VP be [VP see-n ti]]]

 4) Quantifier Raising (LF)

[CP [QP everyj] [IP [NP tj cat]i [VP be [VP see-n ti]]]]

Note that the Projection Principle is valid throughout the derivation.

Another brief but important remark on the T-Model: both PF and LF are not strictly the Semantics and the Phonology of a sentence, but syntactic representations of a sentence. LF and PF are the interfaces to what a sentence means and how it is phonetically realized. Thus LF and PF give you the formal outputs of a derivation of a sentence.

A final note concerning the role of morphology within GB: Marantz & Halle (1993) have proposed a morphological theory that mediates between SS and PF. It is called Distributed Morphology and mainly deals with inflectional morphology. I leave it at this remark. In general, it was assumed in GB that inflectional morphology such as affix lowering applies on the way from SS to PF.