Phonetic convergence in shadowed speech: a comparison of perceptual and acoustic measures

Phonetic convergence is highly variable across studies, measures, and analyses. The current paper describes a study that examined multiple acoustic measures in concert with a perceptual measure of phonetic convergence. The study employed a shadowing task in which multiple talkers shadowed words from a set of models. Across different scales of analysis, the acoustic measures were highly variable, yielding inconsistent results. Perceptual assessment of phonetic convergence provided a measure that was more stable, reliable, and valid than any single acoustic attribute. Mixed-effects regression modeling assessed the relative contributions of each acoustic attribute to perceived phonetic convergence on a word-by-word basis. This study demonstrates the utility of an approach that combines acoustic and perceptual measures of phonetic convergence.