Auditory Statistical Learning in Children with ASD Relates to Verbal IQ: An ERP Study

Background:  

Approximately 30% of children with autism spectrum disorder (ASD) remain minimally verbal (MV) despite receiving intervention (Anderson et al., 2007).  Examining cognitive skills that underlie language learning may help reveal reasons for this impairment.  One crucial precursor to language development is auditory statistical learning (ASL), where infants are able to detect word boundaries by determining transitional probabilities between syllables in the speech stream (Kuhl, 2004). Although impaired ASL in high-functioning children with ASD has been documented using fMRI (Scott-Van Zeeland et al., 2010), no studies have examined ASL in minimally verbal children.  Electroencephalography (EEG) represents an ideal methodology for studying this population because it gathers information about cognitive processing without requiring overt responses.  

Objectives:  

Evaluate ERP evidence of ASL in MV and verbal children with ASD, compared with typically developing children.  Examine the relationship between ASL and language ability.  

Methods:  

30 children with ASD (17 MV; age 5-10) and 7 age-matched TD children participated in an ASL event-related potential (ERP) paradigm. Participants heard a 4-minute acoustic stream of 4 pseudo-words (constructed from 12 syllables), followed by a test phase in which pseudo-words (high transitional probabilities between syllables) were interspersed with non-words (low transitional probabilities). The mean ERP amplitude 400-600ms after word offset was calculated for each condition, for 4 regions of interest (ROI; left/right, frontal/temporal).  ASD group analysis was performed using repeated measures ANOVA, with correlations to examine relationship between ERP results, verbal and non-verbal IQ. ERP results were directly compared across groups (verbal ASD, MV ASD, TD) using t-tests.  

Results:  

Repeated measures ANOVA yielded a main effect of region (F=5.01, p=.007) but no condition effect or region-by-condition interaction, indicating that ASD participants as a group did not show ERP evidence of ASL.  Group comparisons showed that the MV ASD group showed significantly larger amplitude differences in: left frontal ROI compared with verbal ASD (t=-2.5, p=.02), left and right temporal ROI compared with TD (t=2.23, p=.04; t=3.11, p=.005).  There were no significant differences between verbal ASD and TD groups.  Amplitude difference was negatively correlated with VIQ (r=.46, p=.01), but was not related to NVIQ or chronological age.

Conclusions:  

There is variability across individuals with ASD in the magnitude of their ERP response to an ASL paradigm, where participants with a lower VIQ show more condition differentiation.  TD participants performed similarly to verbal ASD participants with neither group differentiating conditions. ASL may be a more immature process, and so not reliably detectable in participants with a higher developmental age.  Minimally verbal children with ASD may continue perform this cognitive process but fail to develop other necessary cognitive-linguistic processes.  Because ERP evidence of ASL is related to VIQ but not NVIQ, it is likely related specifically to language ability rather than overall developmental level.  Research including a mental age matched comparison group as well as longitudinal research is necessary to fully interpret these findings.  Further research into ERP correlates of ASL as well as other language-related cognitive processes will further elucidate pathways to language impairment in minimally verbal children with ASD.