Foreign accented speech: Encoding and generalization

Speech produced by a non-native talker of a language is influenced by the talkers native language, and the result is Foreign accented speech (FAS). Previous research regarding talker variability and normalization has found that indexical talker characteristics are encoded into memory along with phonetic information, and implicitly helps subsequent speech perception. Are the overarching properties of FAS encoded in the same manner? The current study explores what talker information is encoded into memory, and how learned talker qualities improve subsequent speech recognition. Listeners were exposed to one speaker for four experimental sessions via a battery of tests measuring speech intelligibility. On the fifth day, listeners were given a post-test in signal correlated noise which measured speech intelligibility for either the talker they had been trained on, a similarly accented talker, or a talker with a different, unrelated accent. Three control groups received only the post-tests. In sentential, exposure to a talker with an accent seemed to improve performance for talkers who have a similar accent. This was not the case for single word contexts. Implications onto models of speech perception are discussed.