Recent studies on phonological productivity have shown that native speakers' phonological knowledge not only includes statistical patterns reflected in the lexicon, but also patterns that cannot be gleaned from the lexicon. Based on the results of a wug test, the authors argue in this paper that Taiwanese speakers' phonological knowledge of the tone pattern of double reduplication is the combined result of lexical statistics and a priori knowledge that they bring to the task of learning, which causes the speakers to know both more and less than the lexical patterns. The phonologically opaque tone sandhi pattern is only variably productive in novel words, indicating that the speakers have underlearned this exceptionless pattern in the lexicon. The productivity of the opaque sandhis, however, also shows signs of both overlearning and proper learning from the lexicon: speakers prefer shorter tones as sandhi tones on nonfinal syllables, which could not have been deduced from lexical statistics; but the lexical frequencies of tones and tonal melodies also have an effect on productivity. Finally, transparent tonal patterns such as floating High docking in double reduplication are properly learned from the lexicon and are entirely productive in novel words. The authors also provide a stochastic OT grammar based on the dual listing/generation model of Zuraw (2000) to account for the wug test results.
Proceedings of the 27th West Coast Conference on Formal Linguistics
edited by Natasha Abner and Jason Bishop
Table of contents