Viebahn, M. & Bürki, A.
University of Geneva, Switzerland
This study investigates the processing and representation of words with multiple pronunciation variants in French. For example, the French word for horse can be produced with the vowel schwa (as in cheval) or without it (as in chval). Previous research has shown that variants with schwa (i.e., canonical variants) are processed faster and more accurately compared to variants without schwa (i.e., reduced variants). In the current study, we investigate whether this processing advantage for canonical variants is at least in part due to differences in the frequency with which each variant occurs. Participants performed an auditory lexical decision task on canonical and reduced variants before performing a rating task in which the subjective relative frequency of each variant was estimated for each word. Our results show that when taking the average subjective relative frequency of each pronunciation variant into account, there is no difference in the processing of reduced and canonical variants. When taking un-averaged (i.e., participant-specific) frequency scores into account, we find an effect of subjective frequency for reduced but not for canonical variants. The implications of these results for rule-based and representational accounts of spoken-word processing will be discussed.