Disfluencies Distinguish the Speech of Children with Autism Spectrum Disorder
Objectives: Compare and categorize disfluency rates in the speech of children with ASD and typical development during a semi-structured interview, and assess correlations with clinical and parent ratings of social impairment.
Methods: Research-reliable PhD-level clinicians administered the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2012) to 30 children with ASD and 14 typically developing controls that were matched on age, IQ, and sex ratio (Table 1). The conversation and reporting section of the ADOS was extracted and all utterances, were transcribed verbatim. Two reliable raters blind to diagnostic status coded disfluent behaviors, which were classified as stuttering-like disfluencies (SLD), repetitions (REP) or atypical disfluencies (AD; see Table 2). Interjections, revisions, and abandoned phrases (ABD) were analyzed independently. We used Welch’s t-tests to compare group means for total disfluency level (TDL), frequency per type of disfluency, and proportion of disfluencies. Clinicians assigned each child a calibrated severity score based on behavior during the ADOS, and parents completed the Social Responsiveness Scale (SRS; Constantino, 2005).
Results: Participants in the ASD and TDC groups produced an equal number of total words with no significant differences in TDL (ASD=.09, TDC=.08, p=ns). However, the ASD group produced higher rates of SLD (t=-1.99, p=.05) as compared to the TDC group. No significant differences were found in rates or counts of AD, REP, ABD, interjections, or revisions, although the percentage of children that produced abandoned phrases in the ASD group (63%) was marginally higher than the TDC group (36%, Chi-square=2.94, p=.09). In the ASD group, TDL correlated positively with ADOS severity scores (Pearson=.39, p=.03), as did REP (Pearson=.44, p=.02). In the TDC group, higher SLD rates were associated with more social impairment as measured by the SRS (Pearson=-.56, p=.04).
Conclusions: Higher rates of certain disfluency types may contribute to the perception of impaired pragmatic language in ASD. Consistent with previous studies, the ASD group in the current study produced higher rates of SLDs than the TDC group, and disfluency rates correlated with clinical ratings of autism severity. Interestingly, 30% of participants in the ASD group produced prolongations (initial, mid-word, and/or final) versus 0% in the typical group. This latter, atypical finding warrants further exploration and may indicate enhanced diagnostic specificity when compared to other disfluency types. Transcription and coding of an extended sample is currently underway, and will be complete by May 2016.