Perceptual Dialect Classification by Adults with High-Functioning Autism

Background: The speech signal contains information about linguistic meaning, and about indexical properties of the speaker, such as regional background (Klatt, 1989). Neurotypical (NT) adults can use indexical information in the speech signal to accurately categorize unfamiliar talkers based on dialect (Clopper & Pisoni, 2004). Categorization requires an intact perceptual system, which adults with high-functioning autism (HFA) appear to have: people with autism spectrum disorders (ASD) have normal abilities in the domains of phonological discrimination (Constantino et al., 2007), prosodic tune recognition (Jarvinen-Pasley et al., 2008) and unfamiliar talker identification (Boucher et al., 2000). By contrast, adults with HFA have problems with social aspects of language (e.g., Baron-Cohen et al., 2000) which may impair their ability to identify talkers by dialect. Indeed, one previous study (Baron-Cohen & Staunton, 1994) found that, in marked contrast to NT teens, ASD teens modeled their own dialect use on the dialect of their mother as opposed to that of their peers.

Objectives: Dialect variation sits at the intersection of linguistic and social information, and this study examined how people with HFA use those two sources of information to make dialect classification judgments about unfamiliar talkers. Can adults with HFA categorize talkers based on dialect, and if so, are their categories comparable to those produced by NT adults?

Methods: An auditory free classification procedure (Clopper & Pisoni, 2007) was used to assess the perceptual similarity structure of dialect variation in American English. Fourteen adults with HFA and 27 NT adults listened to 20 male talkers from four different dialect regions and were asked to group the talkers based on where they believed the talkers were from. All participants also completed the Autism Quotient (Baron-Cohen et al., 2004).

Results: As expected, the adults with HFA had significantly higher average Autism Quotients than the NT adults. In the dialect classification task, both the NT and HFA adults grouped the talkers into an average of 6-7 dialects. However, the NT adults were significantly more accurate than the adults with HFA in grouping talkers from the same dialect together. Further, a clustering analysis revealed that both the NT and HFA participants sorted speakers from the New England and Southern dialects into distinct groups, and both perceived a high degree of similarity between talkers from the Midland and Northern dialects. However, the HFA adults produced noisier dialect groupings, and the perceptual similarity between the more marked New England and Southern dialects and the less marked Midland and Northern dialects was higher for the HFA adults than the NT adults.

Conclusions: These findings suggest that adults with HFA can perceive relevant dialect differences in the speech signal, and can use this variation to categorize talkers by dialect. However, the differences between the HFA and NT groups suggest that the HFA adults' perceptual dialect categories are less robust than the NT adults' categories. Ongoing work is examining the extent to which HFA performance differences in perception are related to differences in social categorization of dialects.