Assimilation and morpheme boundaries in Mgira

Phonological assimilation phenomena are often restricted by morphological boundaries. This paper presents data from Mgira in order to argue that both root boundaries and affix boundaries serve to block assimilation processes such as nasal place assimilation and postnasal voicing. The result is that assimilation may obtain, but only within morphemes, not across them. This is analysed as confirmation of the power of the CrispEdge constraints proposed by Itô and Mester (1999) to take precedence over general markedness constraints, blocking feature spreading across morphological boundaries. The analysis is claimed to yield the same effect as a Morpheme Structure Constraint, limiting the force of a constraint to within the boundaries of morphemes.