Constancy of (acoustic) relative timing measures in phrase-level utterances.

Abstract

The hypothesis that acoustic measures of relative speech timing remain constant across large changes in speaking rate was tested for fluent utterances produced by normal and neurogenically disordered speakers. For all speakers, relative timing tended to be statistically invariable across conversational and fast speaking rates. This finding was considered especially interesting in the case of speakers with dysarthria and apraxia of speech, because absolute measures of speech timing associated with these disorders are typically aberrant. A post hoc analysis of relative variability across speakers showed the ratio measures to be consistently less variable than measures of absolute interval duration. The greater similarity across speakers of ratio measures, as compared to interval measures, was taken to mean that characterizing the temporal structure of articulatory sequences in terms of relative timing is an appropriate way to capture an important aspect of the biological constraints associated with speech production.