NONSYLLABIC ANALYSIS OF VOICE ASSIMILATION IN POLISH

This article argues (1) that voice assimilation in Polish is governed by linear adjacency of Laryngeal nodes rather than by syllable structu re, and (2) that [voice] should be regarded as binary, not privative. It is shown that past research has depended too heavily on the direct syllabic conditioning of voice assimilation. In reality, prosodic cond itioning is indirect and is mediated by Sonorant Default. The correct application of several rules requires that adjacency be defined at the largyneal tier, which is possible only if [voice] is regarded as a bi nary rather than a privative feature.