An Extrusional Approach to *p-/w- Variation in Sino-Tibetan

There are a surprisingly large number of Tibeto-Burman [TB] and Sino-Tibetan [ST] roots that show interchange between a labial stop and the labial semivowel /w/. These are not regular correspondences, where a given language consistently has a stop, while another consistently has a w. Neither can the distribution of stop vs. semivowel reflexes be correlated very neatly with particular subgroups of TB. Certain subgroups, notably Qiangic and Kamarupan, are split down the middle, with stop and semivowel reflexes equally common and distributed randomly. Naxi, genetically quite close to Loloish, usually has stops, while Loloish itself favors semivowels. Some languages (e.g. Lepcha) have doublet formations, with both stop and semivowel allofams descending from the same etymon. This phenomenon has been one of the most vexatious in TB comparison, and it is clear that Benedict was never satisfied with any of the many ‘explanations’, often mutually contradictory, that are offered in the text and notes of STC. In this paper I try to sharpen the theoretical issues involved in choosing one line of explanation over another, opting eventually for an ‘extrusional’ analysis.