Selective adaptation of English consonants using real speech

Abstract

Most English consonant-vowel (CV) syllables have other CV syllables embedded within them. For example, splicing sufficient energy from the onset of [k ha] yields [p ha], splicing into [ma] or [va] yields [ba], and splicing into [ǰa] yields [da]. We spliced successively longer segments from naturally spoken CV syllables to produce sequences of CV syllables which varied in discrete acoustic steps from [k ha] to [p ha], from [ǰa] to [da], from [ma] to [ba], from [fa] to [ba], and from [va] to [ba]. Random presentation of syllables in each series resulted in identification functions with typically sharp phoneme boundaries. For example, in the seven-syllable [ǰa]-[da] series (where [ǰa] was the original or first syllable), there were98% [ǰa] responses to Syllable 3, but only 12% [ǰa] responses (88% [da] responses) to Syllable 5. Following the identification test, subjects listened to 180 repetitions of either the first or last syllable in the test series, and were again required to identify randomly presented syllables from the test series. A shift in the phoneme boundary toward the repeated (adapting) syllable was observed for 11 of the 12 repeated syllables. Repeated presentation of [ǰa], for example, resulted in fewer [ǰa] responses to syllables in the [ǰa]-[da] series, compared to performance on the previous identification test. Likewise, repeated listening to [da] resulted in a decrease in [da] responses. Adaptation was selective in that syllables near the phoneme boundary were most affected by the adapting syllable. A shift in the phoneme boundary was also observed for two different continua when the adapting stimulus contained an acoustic feature identical to syllables in the test series. Thus, selective adaptation was found along a [ma]-[ba] continuum following repeated presentation of [na] and following repeated presentation of nasal resonance removed from its syllable context. A second major result, observed in five different experiments, was an asymmetrical adaptation effect. A greater shift in the phoneme boundary was observed following repeated presentation of the first syllable in each series (e.g., [ǰa]) than for the final embedded syllable (e.g., [da]). The results were discussed in terms of two different models of-selective adaptation. © 1975 Psychonomic Society, Inc.