Abstract
This paper argues against a (standard) markedness-based approach to rhyme phonotactics in Taiwanese since analyses in this vein overgenerate by predicting unattested VC gaps. Instead, I explain VC gaps in the following terms. The key point is that the salient places cues of release to a coda stop are absent in Taiwanese (and many other (South) East Asian languages), leaving VC transitions as the only cues and consequently impeding an accurate and reliable identification of place distinctions in coda position. As such place contrasts may be neutralized (henceforth the gaps) because there are systematic asymmetries after particular vowels in terms of their ability to signal place in the upcoming stop. Otherwise, enhancement is invoked to maintain place contrasts, resulting in vowel quality change or emergence of an excrescent schwa (confirmed by an acoustic study reported in this work). I shall show that both “repair strategies” are attested in Taiwanese and Cantonese (and perhaps Hakka) and can be captured by Steriade’s (1999) Licensing-by-Cue hypothesis, in tandem with Flemming’s (2002) Dispersion Theory of contrast.