Language Identification from Nonsegmental Cues
Kay Atkinson · The Journal of the Acoustical Society of America · 1968
Low-pass filtering was used to extract intonation and durational cues from a tape recording of short utterances drawn from English and Spanish poetry, prose, natural speech, nursery rhymes, and dramatic dialogues. This tape, which thereby excluded segmental information, was used in listening tests with both simple identification and ABX designs. Forty-three subjects were asked to discriminate between English and Spanish from these cues alone. Speaker differences, listener differences, lengths of utterances, and the order of presentation of utterances were considered in the experimental design. Results from the identification test indicate that subjects can significantly discriminate between languages without reference to segmental cues. These results demonstrate that different nonsegmental acoustic properties characterize different languages. Various factors, which contribute to the ease of identification, were also revealed, e.g., subjects made the smallest percentage of errors in identifying poetry and the greatest percentage of errors in response to prose and nursery rhymes. Possible reasons for this are suggested in the paper. Alternative hypotheses concerning the acoustic cues that signal the difference between the languages are also suggested.