Parsing voicing cues: A context effect predicted by feature cue parsing theory

Abstract

Factors including lawful assimilation conspire to produce overlapping temporal distributions of phonetic feature cues for neighboring segments. Gow’s feature cue parsing model [Gow, Perception and Physics, 65, 575–590 (2003)] suggests that listeners rely on perceptual grouping mechanisms to integrate multiple cues and align them with correct segmental positions. While this account was developed to address progressive and regressive context effects found in the perception of lawfully assimilated speech, its strong form predicts that similar context effects may be found in the perception of speech where there is similar overlap in feature cue distribution, even if such overlap is not the result of a lawful phonological process. This hypothesis was examined in a series of speeded 2AFC experiments in which voicing cues were manipulated in adjacent fricatives and stops. A regressive context effect was found in the perception of the fricative in a peace-peas voicing continuum, which was strongly affected by the voicing of a following stop (e.g., pay versus bay). Similarly, listeners showed a progressive context effect in the perception of stops in a pay-bay VOT continuum when these items were preceded by voiced versus voiceless fricatives. These results are discussed in terms of competing interpretations of assimilation context effects. [Work supported by NIH R01DC3108.]