Categorical Speech Perception Across the Autism Spectrum and Its Relation to Cognitive and Language Ability
Individuals with autism spectrum conditions (ASC) show superior discrimination of certain types of perceptual information (e.g., Bonnel et al., 2003) and also a tendency to process perceptual stimuli without reference to the contextual information of the percept (Jolliffe & Baron-Cohen, 1999). Consistent with these observations, adults with ASC are known to perceive exemplars along a visual spectrum less categorically (i.e., more continuously) than controls (Soulières et al., 2007). However, this pattern has not been examined in speech perception, a domain that typically exhibits categorical perception (e.g., Abramson & Lisker, 1970).
To test whether adults with ASC demonstrate reduced categorical perception of basic speech sounds and whether this effect is independent of IQ and/or language functioning that are known to affect speech perception.
Twenty adults with High-Functioning Autism or Asperger’s syndrome (HFA/AS; mean age=27.3, SD=7.8; males=13) were compared with 20 nonverbal IQ-matched typically-developed adults (mean age=27.7, SD=9.7; males=7). Participants were given measures of autistic traits (the Autism-Spectrum Quotient; Baron-Cohen et al., 2001a), nonverbal IQ (Raven’s Advanced Progressive Matrices; Raven, Raven, & Court, 1998), verbal IQ (Mill Hill Vocabulary Scales; Raven, Court, & Raven, 1988), 5 language measures, the Autism Diagnostic Observation Schedule (Lord et al., 1999), and an ABX speech discrimination task. The A and B stimuli in the ABX task were two samples taken from a /gi/-/ki/ continuum that differed in voice-onset time by 20 ms. Categorical perception of these stimuli should be manifested in higher accuracy scores around the midway point of the continuum where the boundary between the /gi/ and /ki/ categories lie.
There was a significant main effect of Step (F(4, 152)=8.38, p<.001, η2p=.18); accuracy was significantly better at Step 30-50 ms than at Steps 10-30 ms (p<.05), Step 40-60 ms (p<.01), and 50-70 ms (p<.001). There was no significant main effect of Group (F(1, 38)=.13, p=.72, η2p=.00) or Step x Group interaction (F(4, 152)=2.13, p=.08, η2p=.05). In order to test for the influence of IQ the sample was split at the median on non-verbal IQ scores. Participants with HFA/AS with below-median nonverbal IQ were significantly less accurate across the continuum (F(12, 136)=2.50, p<.01, η2p=.18) and they did not show the peak of performance at the boundary between phonemes (p<.05) compared to participants with HFA/AS with above-median nonverbal IQ. That is, there was a diminished influence of category on the discrimination of exemplars (F(3, 39)=3.78, p<.05). This was in contrast to the median-split nonverbal IQ groups without ASC.
A diminished influence of category on the perception of exemplars occurs for speech signals across the autism spectrum as seen in the visual domain (Soulières et al., 2007). There is a significant ASC x IQ interaction for performance on this task. Further research is needed to assess whether reduced categorical perception can occur for other voicing continua in individuals with ASC. Implications in terms of speech processing and training on speech tasks in individuals with ASC need to be explored.