r-Epenthesis and the bigrade alternation

For Japanese verbal suffixes sensitive to the C/V status of the stem-final segment, C-stem alternants are underlying, and regular V-stem alternants result from intervocalic epenthesis of r at stem boundary (de Chene 2016). This “Analysis A” entails that any V-stem suffix not consisting of r plus its C-stem counterpart is irregular and subject to replacement. While the r-Epenthesis rule of Analysis A is naturally understood as a generalization of the r-zero alternation of three suffixes that have shown it since the eighth century, however, the innovative r-initial suffixes of other categories do not appear until the eighteenth. This lag is illuminated by the dialects of Kyūshū, where adoption of Analysis A is blocked by the “bigrade” stem alternation, which in most dialects was leveled in the seventeenth century. Building on a discussion of leveling that treats that phenomenon as a subtype of regularization, it is proposed in explanation of this “bigrade blocking” effect that the order in which alternations become subject to regularization is constrained by the phonological distance between alternants. Investigation of the possibility that the bigrade alternation and Analysis A are related by a triggering effect as well as by a blocking effect then leads to an account of the adoption of Analysis A that, similarly, relies crucially on the concept of phonological distance. Throughout, the focus is on the role of language-internal factors in determining the timing of analogical change.