Cyclic Output
The final issue I will consider, if only briefly, is the ‘‘timing’’ of the relation between syntax on the one hand and semantics and phonology on the other. In the Government-Binding ‘‘T-model,’’ there is one specific syntactic level, S-Structure, which connects to the LF and PF interface modules. In the LSLT model, on the other hand, the interface is the T marker, which includes all of the syntactic structures created in the course of the derivation. Given that the minimalist approach to structure building is much more similar to that in the LSLT model than to that in any of the intervening models, the possibility presents itself that interpretation in the minimalist model also could be more like that in the LSLT model, distributed over many structures. Interestingly, already in the late 1960s and early 1970s, arguments were occasionally made for such a model of interpretation, for phonological interpretation as well as semantic interpretation. For example, Bresnan (1971) argued that the phonological rule responsible for assigning English sentences their intonation contour applies cyclically, following each cycle of transformations, rather than at the end of the entire syntactic derivation. Similar proposals for semantic phenomena involving anaphora were put forward by Jackendoff (1969). Jackendoff showed that given standard assumptions about derived constituent structure, the rule assigning a reflexive its antecedent must apply before certain transformations. He showed that the right result is obtained if the reflexive rule operates at the end of each syntactic cycle. In Lasnik 1972, I suggested, on largely conceptual grounds, that the scope of negation is also determined cyclically (i.e., at the end of each syntactic cycle). And in Lasnik 1976, I argued that strong crossover can be accounted for if the scope rule determining bound variable anaphora is cyclic in the same sense. In recent work, Chomsky (2000, 2001) suggests a more general instantiation of this distributed approach to phonological and semantic interpretation, based on proposals by Epstein (1999) and Uriagereka (1999). At the end of each cycle (or ‘‘phase’’ in Chomsky 2000, 2001), the syntactic structure thus far created is encapsulated and sent o¤ to the interface components for all phonological and semantic interpretation. Thus, while there are still what might be called PF and LF components, there are no levels of PF and LF. There is just one cyclic derivation, which, in effect, encodes what the much earlier T-markers did.
