Postersession 3
Poster #: 120
Topic: Speech and language (incl. deficits)
Friday, Sep 11, 2015
15:30-17:00
1st floor
Adult listeners’ processing of indexical versus linguistic differences as reflected by the mismatch negativity
1University of Amsterdam, Netherlands
2University of Western Sydney, Surry Hills, Australia
3MARCS Institute, University of Western Sydney, Surry Hills, Australia
paola.escudero@uws.edu.au
A puzzling phenomenon in speech research is the ability of humans to comprehend speech sounds regardless of variation across speakers and dialects. Previous studies assume that during speech comprehension, differences between speakers and dialects are normalized using the same underlying mechanism (Pierrehumbert, 2002). However, dialect normalization appears to rely on lexical knowledge and exposure (White & Aslin, 2011). In contrast, speaker normalization seems to occur automatically and without lexical knowledge since not only human infants (Kuhl, 1983), but also birds can discriminate human speech sounds across different voices (Ohms, et al., 2010). The present study uses event-related potentials to test whether such different mechanisms exist in a pre-attentive discrimination paradigm.
We tested Australian English monolinguals and bilinguals and native Dutch listeners. Using a multiple-deviant oddball paradigm we assessed their perceptual sensitivity to four types of variation in vowels: in speaker identity, sex, dialect, and vowel category. We predicted that vowel category and dialect deviants will yield larger MMN amplitude than speaker identity deviants, because the vowel category and dialect deviants involved the largest spectral difference from the standard stimulus and represented a combined acoustic and linguistic change. Interestingly, all listeners showed similar results regardless of their linguistic background: changes in dialect elicited large MMN amplitude. Rather surprisingly, however, speaker identity deviants elicited larger MMN than vowel category deviants, which is not in line with previous overt vowel classification findings. The present MMN results are explained by adults’ automatic sensitivity to differences in voice quality, rather than to spectral properties of vowels.