Please come to this colloquium featuring a phonetic study by our department’s own Richard Wright. Abstract after the fold.
November 19, 2010
This talk concerns the role of training-stimulus variation in adult learners’ acquisition of non-native speech sounds. It describes a non-native word learning experiment designed to differentiate the beneficial roles of inter-talker (indexical) variation, and intra-talker (style) variation.
Several studies have shown that adult learners of non-native speech sounds benefit from intensive training involving multiple talkers (multi-talker training) and multiple positional/allophonic variants of the trained sounds (e.g. Logan, Lively, & Pisoni, 1991; Lively, Logan, & Pisoni, 1993; Yamada, 1993). In these studies adult learners’ ability to discriminate perceptually challenging sounds at the segmental level (American English /?/-/l/ in the cited cases) improved with high-variability training. Importantly, the learners’ perceptual learning generalized to the trained sounds in new words and to new talkers. The perceptual training also generalized to long-term production improvements (Bradlow, Pisoni, Akahane-Yamada, & Tokhura, 1997; Bradlow et al, 1999). Wang et al (1999) extended these findings to the suprasegmental level by training American English listeners to identify Mandarin tones. Again, the benefits of multi-talker perceptual training showed long-term effects, although with some complexities related to different dimensions of tones (Wang et al, 2002). These results are generally interpreted as evidence that this training produces highly generalized perceptual learning and results in long-term modifications of the learners’ perceptual system. The multi-talker aspect of the training is sometimes further interpreted as evidence of the importance of indexical, talker-specific, variation in training. One obvious shortcoming in the research to date is a confound between indexical variation and style variation in speaking. That is, each talker brings his/her interpretation of the speaking task to the recording resulting in a wide range of styles (ex: Bond & Moore, 1994). Therefore, it remains unresolved whether the multi-talker effect is due to inter-talker variation, to intra-talker variation, or to some combination thereof. There is evidence that at the very least speech rate variation has some effect for memory for spoken words (Bradlow, Nygaard & Pisoni, 1999).
In this experiment style variation was fully crossed with indexical variation. Subjects were trained in one of four sets of stimuli (1 single-talker single-style, 2 single-talker dual-style, 3 dual-talker single-style, 4 dual-talker, dual-style). Subjects were trained using two sets of nonsense words: 1 composed of native sounds, and containing non-native speech sounds [y] or [x] that were presented randomly paired nonsense shapes (Fribbles
http://titan.cog.brown.edu:8080/TarrLab/stimuli). Subjects were probed at each quartile of the training to monitor learning (10 presentations of each word-shape combination). Subjects were then tested in a generalization task with new talkers. Results of this experiment indicate an important role for style variation in the ability of learners to generalize to novel exemplars of trained words.
The University of Washington is committed to providing access, equal opportunity, and reasonable accommodations in its services, programs, activities, education, and employment for individuals with disabilities.
To request disability accommodations, please contact the Office of the ADA Coordinator in advance. 545-6450 (voice); 543-6452 (TDD); firstname.lastname@example.org (e-mail).