ISCA Workshop on Plasticity in Speech Perception (PSP2005)

Senate House, London, UK
June 15-17, 2005

Short- and Medium-Term Plasticity for Speaker Adaptation Seem To Be Independent

Holger Mitterer

MPI for Psycholinguistics, Nijmegen, The Netherlands

At least two forms of plasticity in speech perception serve speaker adaptation. Norris, McQueen, and Cutler (2003) showed that exposure to ambiguous phonemes in words leads to a recalibration of phonetic categories: If a listener encounters an ambiguous fricative, and the lexical context indicates the fricative's identity, as in legiXlature, listeners are more likely to perceive the sound 'X' as an instance of the 's'- category in other, later, words too. This adaptation is speaker-specific and does not necessarily generalize to fricatives produced by other speakers. While this form of plasticity occurs on a medium time frame over a couple of minutes, another instance of speaker adaptation can occur even faster: Ladefoged and Broadbent (1957) showed that listeners adapt their phoneme boundaries in a compensatory way on the basis of the formant range encountered in a carrier phrase: Listeners are more likely to perceive a low vowel (i.e., higher F1) if the carrier sentence contained comparably a low F1. Most vowel normalization studies using the Ladefoged and Broadbent paradigm have used meaningful carrier sentences. The question arises whether the short-term plasticity observed in vowel-normalization may profit from lexical information, which provides the phonetic category of the vowels encountered in the carrier phrase. Accordingly, the lexical status of the constituents of the carrier phrase was manipulated in two vowel-normalization experiments. Simultaneously, F2 range in the carrier phrase was manipulated and the effects on the perception of a F2-vowel continuum was assessed. In Experiment 1, the carrier phrase contained high, low, front, and back vowels. Listeners adapted their vowel categories in a compensatory way for the F2 range in the carrier, independent of the lexical status of its constituents. In Experiment 2, the carrier phrase contained only high-front vowels, in order to minimize the information about the formant range of the speaker. Nevertheless, lexical status of the constituents of the carrier phrase again failed to influence vowel normalization. The perception of the test continuum still indicated compensation for the F2 range in the carrier phrase, but only for the category boundary between the high-front and highmid vowel, which is close to the high-front vowels presented in the carrier. The boundary between a high-mid vowel and a high-back vowel was, in contrast, unaffected. This indicates, first, that vowel normalization is vowel-specific and, second, that short- and medium-term plasticity for speaker adaptation seem to be independent.

Full Paper

Bibliographic reference.  Mitterer, Holger (2005): "Short- and medium-term plasticity for speaker adaptation seem to be independent", In PSP2005, 83-86.