Speaker: Canaan Breiss (UCLA)
Title: Between Grammar and Lexicon: New Experimental Evidence for Lexical Conservatism
Time: Monday, November 30th, 5pm – 6:30pm
Abstract: In this talk I will discuss my dissertation research on Lexical Conservatism (Steriade 1997, et seq.), a theory of the relationship between lexicon and grammar which holds that markedness-improving phonological alternations are enabled by the presence of phonologically-optimizing morphologically-related forms in the lexicon. For example, English cómpensate undergoes rightward stress shift when affixed with -able, to yield compénsable, while phonologically-similar ínundate does not (ínundable, *inúndable): Lexical Conservatism holds that this is due to the presence of a phonologically-optimizing morphologically-related form compéns-(atory), while there is no similar form with a stem allomorph inúnd-. Noting this correlation, however, does not provide the detailed information necessary for a fully fleshed-out phonological model, nor a thorough understanding of how the grammar interacts with the lexicon when forming novel words. Drawing on two experiments on English and a third on Spanish, I demonstrate that Lexical Conservatism is robust in the laboratory setting, but holds as a probabilistic tendency rather than as a rule. That is, I find that we observe both the markedness-avoiding behavior pointed out by the original discussion of Lexical Conservatism in Steriade (1997), but that the likelihood of this behavior is responsive to processing factors like the accessibility of the Remote Base, as manipulated by priming. This implies a dynamic trading relationship between the phonological grammar and the lexicon that is not well-captured by extant theories of lexicon-phonology interaction. I discuss which of these findings we might want to incorporate into a phonological theory, and propose a model couched in a Maximum Entropy framework (Goldwater & Johnson 2003) to account for the phonological facts, while allowing principled integration of lexical characteristics that are, I argue, better thought of as non-phonological.