Traditional accounts of the allophony of /bdg/ in Spanish describe it as a classic case of complementary distribution: /bdg/ appear as occluded voiced stops in phrase initial position, after nasals and after homorganic laterals, and as approximants elsewhere. This study aims to answer two questions: 1) how native Spanish listeners categorize occluded voiced stops in intervocalic positions, since they do not occur natively, and 2) how listeners rate them in terms of goodness of fit to their chosen native category. Two theories of cross-language speech perception make differing predictions of how occluded intervocalic voiced stops should be assimilated. Flege's Speech Learning Model, which posits that assimilation is based on position-specific allophones, predicts that intervocalic [bdg] would be assimilated to native Spanish /ptk/, since they are more similar acoustically. On the other hand, Best's Perceptual Assimilation Model, which posits higher order invariants as the units of perception, predicts that intervocalic [bdg] would be assimilated to /bdg/, since they are more similar articulatorily. The results of a cross-language perceptual assimilation task and a goodness rating task, using Brazilian Portuguese stimuli, reveal that native Spanish listeners assimilated occluded voiced stops to their native /bdg/ categories, despite their phonotactic deviance, albeit as worse exemplars of their respective /bdg/ categories. These findings support the predictions of Best's PAM rather than Flege's SLM, thereby suggesting that the mapping of non-native contrasts to native phonological categories, not just to position-specific allophones, plays a determining role in cross-language speech perception.
Selected Proceedings of the 15th Hispanic Linguistics Symposium
edited by Chad Howe, Sarah E. Blackwell, and Margaret Lubbers Quesada Table of contents
ISBN 978-1-57473-457-7 library binding
vi + 342 pages
publication date: 2013
published by Cascadilla Proceedings Project, Somerville, MA, USA