Lack of Neural Specialization for Speech in Children with Autism Spectrum Disorder

Background: fMRI studies have demonstrated abnormal cortical processing of social auditory stimuli in adults with Autism Spectrum Disorder (ASD).  Specifically, a voice-selective region in the superior temporal sulcus/superior temporal gyrus (STS/STG) is hypoactive to voiced sounds in adults with ASD, suggesting deficit in processing human voiced versus non-voiced sounds.  However, research from our laboratory suggests that this STS/STG region is most sensitive to human voiced sounds that are communicative (e.g. speech), rather than being equally sensitive to all voiced sounds.  This raises the possibility that reduced STS/STG activity in response to voiced versus non-voiced sounds observed in adults with ASD may be driven by altered response to particular categories of voiced sounds, rather than all voiced sounds.

Objectives: (1) Investigate whether children with ASD show hypoactivation in STS/STG in response to voiced versus non-voiced sounds, and (2) Investigate whether children with ASD and typically-developing (TD) children differ in response to different categories of human voiced sounds.

Methods: Data was collected from 38 participants (18 ASD) matched on age, gender, and verbal IQ.  Participants heard both voiced and non-voiced sounds while in the fMRI scanner.  There were 3 categories of voiced sounds: non-native human speech, human communicative vocalizations (e.g. laughing), and human non-communicative vocalizations (e.g. coughing).  Non-vocal stimuli consisted of human non-communicative non-vocal sounds (e.g. walking).  Each condition was presented for 20 seconds, 5 times in a block design.

Results: We performed a 2 x 2 whole-brain ANCOVA analysis to identify regions exhibiting significant Condition (voiced versus non-voiced) x Group (TD versus ASD) interaction (q<0.001, k>36).  No significant voxels were observed.  A more liberal threshold (p<0.05, k>12), revealed a group difference to voiced versus non-voiced sound in solely anterior insula. To determine whether the groups responded differently to different human voiced sounds we did an ROI analysis, with the ROI functionally defined as bilateral STS/STG that responded to voiced versus non-voiced human sounds in both TD and ASD children (q<0.001 and k>36).  A repeated measures ANOVA with sounds as a within-subject factor and diagnosis as a between-subject factor was conducted on beta values from the bilateral STS/STG ROI.  There was a significant group x sound interaction in both hemispheres (right: F(2,35) = 3.089, p<0.05; left: F(2,35) = 3.821, p<0.026).  Follow-up analyses showed a main effect of sound for TD children (TD: F(1,20) = 7.873, p<0.001; ASD:F(1,16) = 0.972, p<0.389) but not for ASD children.  Paired-samples t-tests revealed that TD children’s STS/STG responded more robustly to speech than to both non-speech voiced categories bilaterally (Right:  t(20) = 5.538, p<0.000; Left: t(20) = 4.878, p<0.0001).

Conclusions: Our results didn’t support previous findings that individuals with ASD show hypoactivation in the STS/STG to voiced versus non-voiced human sounds.  However, we found between-group differences in response to different categories of human voiced sounds within this voice-sensitive region.  Specifically, the TD children differentiated between three voiced conditions, responding most strongly to speech, unlike the ASD group.