Gradient phonotactics in Optimality Theory
Arto Anttila, Department of Linguistics, Stanford University
Gradient phonotactics means that lexical items can be phonologically
more or less well-formed depending on the segment combinations they
contain. The evidence is of two kinds: (i) some segment combinations
are statistically overrepresented, others statistically
underrepresented in the lexicon; (ii) novel words ("wug words") show
gradient acceptability based on their phonotactic structure. Based on
dictionary evidence from the Austronesian language Muna, Coetzee and
Pater (2006) put forward an optimality-theoretic approach to gradient
phonotactics. I will examine C&P's analysis in detail, show where it
succeeds and fails, and generalize their approach by introducing the
notion of T-ORDER which captures the implicational
universals hidden in factorial typologies and turns out useful in the
study of gradient phonotactics, and beyond.
Back to the
Phonology Workshop home page