Repairs and Uninterpretable Features

Abstract

Chapter 5 develops the last-resort mechanism ℜ, through a cross-linguistic study of the repairs of person constraints by the emergence of otherwise unavailable ergatives, accusatives, PPs, and enriched DPs. They are unified as the minimal enrichments of a non-convergent syntactic structure by an Agree/Case dependency, that is, by an uninterpretable phi-feature (probe). ℜ thus extends the hypothesis that uninterpretable features create syntactic dependencies to meet Full Interpretation, from features that are lexically fixed, to those that are dynamically inserted for this reason. The point of departure is Chomsky's (1995 et seq.) proposal that syntactic content can be licensed by an 'effect on output', confined by a strong modular architecture that restricts its scope as an interface mechanism. ℜ can detect illegibility at the interfaces of syntax with PF and LF, but not problems that arise within these modules. In response, it can enrich the numeration interface between syntax and the lexicon with an uninterpretable feature, but cannot modify syntactic computation, nor search the lexicon for interpretable content. ℜ is extended to other Full Interpretation failures, notably the ergative and accusative 'dependent Case' of all transitives, construed as a response to Case licensing.