Voice contrast and culumative faithfulness in Luwanga nouns

Abstract

Luwanga has a seemingly allophonic surface distribution of voiced and voiceless obstruents. This commonplace distribution typically requires the proposition that segments are specified as either [±voice] underlyingly, with their counterparts derived via phonological rule. Drawing evidence from consonant alternations in Class 9/10 nouns and their derivatives, obstruents contrast for [voice], at least in stem-initial position. Elsewhere, voice is non-contrastive. The outcome of this alternation, although transparent, cannot be captured in a standard constraint-based optimality theoretic framework and instead requires machinery employed to address surface opacity. This paper illustrates that the result of competing pressures to remain faithful to the underlying segmental structure, as well as to a consonant’s specification for [voice], is the seemingly transparent but analytically opaque retention of marked structure. We illustrate that this type of cumulative faithfulness is best addressed via one of two evaluative mechanisms capable of capturing additive effects, namely Local Constraint Conjunction and Harmonic Grammar.