Discrimination of Dialect from Temporal Patterns of the Speech Signal

Abstract

To determine whether dialect discrimination could be made on the basis of temporal sequencing (or rhythm) of the speech signal, speech samples were collected from 90 Ss, identified as black, white, and Mexican-American. Samples were matched on basis of similarity of context. Similar samples were recorded, using a triad test format. Two racially similar and one different speaker, e.g., black, black and white, formed a triad. 24 ads were put together and recorded. An electronic logic system (Schmitt trigger design) generated a noise analogy (temporal pattern) for the speech signal 24 noise-analogy triads and 24 speaker triads were presented to groups of black, white, and Mexican-American listeners to determine whether ethnically identified listener groups could distinguish speakers' patterns. Our ethnically distinct Ss were unable to distinguish between dialects on the basis of temporal patterns alone. There was a tendency for all listeners to detect the Mexican-American speech at the chance level.