How listeners represent dialectal variation.

Successful speech recognition requires that listeners map acoustically variant pronunciations onto the same meaningful interpretation. As listeners are exposed to a talker’s speech, they learn the relevant mapping, and their perception shifts to accommodate the talker’s pronunciations. Recent work on the perceptual learning of idiolectal variation has made it possible to investigate precisely what it is that changes in perceptual processing as listeners adapt to speakers. These studies suggest that phonemic representations themselves are rapidly modified to reflect the idiolectal variants produced by a talker. However, dialectal variation appears to be handled differently. Several studies will be presented that address the representation of dialectal forms, including a recent experiment that directly compared learning of a dialectal variant to learning of an idiolectal variant with the same acoustic properties. Listeners adjust to talkers’ dialects, but unlike with idiolectal variation, such adjustments do not reflect an underlying modification of the phonemic representations of the canonical counterparts. The implication of these results for dialect representation, as well as for the perceptual learning process, will be discussed. [This work was supported by NIH NRSA Award No. 7 F32 HD052342‐03.]