Restrictiveness has long been a major concern within the framework of transformational generative grammar. Substantive proposals for limiting the descriptive power of individual transformations began to appear in the very early 1960’s, and such constraints as the A-over-A condition and the principle of recoverability of deletion date from that time. The goal then, as now, was to provide the basis for an account of the selection by the language learner of the correct grammar on the basis of limited data. To the extent that complicated and seemingly idiosyncratic properties could be abstracted out of transformations, their effects deduced from the interactions of general principles, the field moved closer to that goal. Beginning with the publication in 1973 of Chomsky’s “Conditions on Transformations”, the goal, perhaps for the first time, began to seem attainable. The articles collected here, largely inspired by the program of “Conditions on Transformations”, represent a series of investigations and proposals on restrictiveness, dating from the mid 1970’s to the mid 1980’s.
KeywordsPhrase Structure Negative Evidence Descriptive Power Case Assignment Null Operator
Unable to display preview. Download preview PDF.
- 1.When Kupin and I were developing our simplified model of phrase structure, we were somewhat perplexed by the fact that the notion of phrase marker we were arriving at did not (unlike the Chomsky (1955) model) depend on phrase structure derivations. Recall that a phrase marker in that conception was the set of all strings occurring in any of the equivalent phrase structure derivations of a particular terminal string. For us, a (reduced) phrase marker was any set of strings satisfying various well-formedness conditions. We simply remained perplexed about this difference, somehow overlooking the obvious possibility illuminated by the later work of Stowell (1981) and Pesetsky (1982) that phrase structure derivations (hence, phrase structure rules) are artifactual.Google Scholar
- 2.As noted in Chapter 8, while the Case filter treats lexical NP and WH trace in a similar fashion, it does not treat them identically. In particular, the usual adjacency requirement on Case assignment in English is relaxed. See Epstein (1987) for an account of this asymmetry.Google Scholar