Inversion of stress-conditioned phonology in Stratal OT
Matthew Wolf
June 2012

Stratal OT posits that, on the way to the surface, words (and more generally utterances) pass through a succession of Optimality-Theoretic phonological grammars, with potentially different rankings. As a result, the same phonological string, even without undergoing any affixation, may receive different metrical parses at successive levels. In this paper, I argue that this capability of Stratal OT gives rise to typologically undesirable results involving metrically-conditioned segmental phonology. If a stress-conditioned property, e.g. aspiration of voiceless stops in the onsets of stressed syllables, is assigned with respect to the metrical parse at one level, and the result of this is protected by faithfulness at subsequent levels even as the stress is changed, the result is that the surface distribution of aspiration may not match the surface distribution of stress. The extreme possible case of this, which this paper explores, is the situation where successive levels assign exactly complementary patterns of rhythmic stress, such that every syllable stressed at one level is unstressed at the next level, and vice-versa. This would yield a language where, on the surface, voiceless stops are systematically aspirated in the onsets of unstressed syllables, inverting the attested cross-linguistic affinity between aspiration and stress. The general problem is that Stratal OT subverts any attempt to predict which kinds of segmental properties can be conditioned in stressed vs. unstressed position, because even given only markedness constraints which favor a property in [α stressed] position, Stratal OT is capable of giving rise to languages where, on the surface, the property is found in all and only the [-α stressed] positions. In connection with this point, I review several reported seemingly-paradoxical cases of stress-conditioned segmental phonology, which might be taken as evidence that Stratal OT's power of 'inversion' is needed; I argue, however, that all of these will plausibly submit to more typologically-conservative analyses. I also show why the inversion prediction does not arise in Harmonic Serialism or OT-CC: because these frameworks make use of only one grammar with one ranking, the possibility of assigning different parses at different stages of the derivation is tightly limited.
Format: [ pdf ]
Reference: lingbuzz/001547
(please use that when you cite this article)
keywords: stratal ot, stress, opacity, metrical conditioning, harmonic serialism, nganasan, huariapano, moksha, phonology
Downloaded:558 times


[ edit this article | back to article list ]