Comparing French and English listeners’ on-line perception of assimilated speech

Clayards, M. 1, 2, 3 , Gaskell, G. 4 & Niebuhr, O. 5

1 Department of Linguistics. McGill University. Montreal, Canada
2 School of Communication Sciences and Disorders. McGill University. Montreal, Canada
3 Centre for Research on Language Mind and Brain. McGill University, Montreal, Canada
4 Department of Psychology. York University. York, UK
5 Institute of Phonetics and Digital Speech Processing. Christian-Albrechts-University. Kiel, Germany

Spoken language contains many types of contextual changes in pronunciation such as assimilation across word boundaries. Languages differ in the degree and type of assimilatory changes that are allowed. For example, both French and English exhibit some assimilation of sibilants in sibilant sequences (e.g., /s/ becomes like /sh/ in “dress shop”), but they differ in the strength and directionality of these shifts. Listeners are able to compensate for these changes when hearing speech and models of spoken-word recognition differ on whether compensation for assimilatory changes are language-specific or language general. To explore this issue we compared French and English listeners on very similar tasks and stimuli. We taught English and French participants the same set of novel words that began or ended with /s/ or /sh/ (e.g. “samal”, “tamash”). After training, participants were presented with the novel words embedded in sentences that could engender assimilation. Sentences were uttered by both a French and an English speaker and both groups of listeners heard both. Target words contained a sibilant from a 7 step continuum of sibilant sounds varying between /s/ and /sh/ representing a range of possible assimilation strengths. Listeners’ perceptions of the potential assimilations were examined using a visual-world eyetracking paradigm in which the listener clicked on a picture matching the novel word. Both French and English participants were able to compensate for assimilation in these sentences but they only showed compensation for the most extreme assimilations in situations where extreme assimilations are likely to occur in their language. These results suggest that compensation for assimilation develops in response to experience with the statistics of assimilation in the listener’s native language. Patterns of compensation consistent with a more auditory based mechanism were also evident in situations where strong assimilatory patterns are unlikely to occur in the listener’s experience.