Voicing Assimilation and Optimal Domains in English

Abstract

This paper examines various aspects of voicing assimilation in English. Adopting McCarthys (2004) Span Theory, this paper assumes that segments of a word are exhaustively parsed into domains for voice feature and each domain has a head segment. The non-head segments in the domain are pronounced with the head segments voice feature value. Under this assumption, the main proposal of this paper is that regressive voicing assimilation in English occurs domain-internally when the domain for voice feature contains hetero-voicing obstruents, and progressive voicing assimilation is a domain-crossing process enforced by a restriction on the syllable wellformedness. In this view, the regressive voicing assimilation is due to the head-finality. The non-occurrence of voicing assimilation observed in compounds and some affixed words is argued to be a result of the restriction that the domain for voice feature cannot cross the word boundary. The syllable wellformedness constraint called HARMS GENL, like in Lombardi (1995, 1999), plays a crucial role in explaining the situation where progressive assimilation occurs across the word boundary.