Final Paper Number 413.028
Background: Studies have found that children with Autism Spectrum Disorder (ASD) use the filler “um” at a significantly lower rate than children with Typical Development (TD), with no difference in “uh” usage. Examiners’ filler usage in similar conversational language samples has not been studied. We investigated whether differences in “um” and “uh” usage between ASD and TD children also characterize the speech of their conversational partners: the examiners.
Objectives: (1) Compare examiner usage rates of “um” and “uh” when conversing with ASD vs. TD participants; (2) Investigate whether within-group differences in examiner filler usage vary by participant age, intellectual ability, expressive language ability, or autism symptom severity.
Methods: Participants were recruited for an fMRI study. They were 7 to 17 years old, spoke native English, had full-scale IQ ≥ 70, and consisted of 83 children with ASD (68 male) and 28 with TD (12 male). Language samples consisted of transcribed Autism Diagnostic Observation Schedule (ADOS-2) Module 3 sessions. Transcription was completed by trained transcribers who were blind to the participants’ diagnostic status and intellectual abilities in accordance with the Systematic Analysis of Language Transcripts (SALT) guidelines. Four ADOS-2 tasks were chosen for analysis: Emotions; Social Difficulties and Annoyance; Friends, Relationships, and Marriage; Loneliness.
We computed three measures of filler usage: um-rate = total um ⁄ total words; uh-rate = total uh ⁄ total words; um-ratio = total um ⁄ (total um + total uh). Examiner filler usage rates between diagnostic groups were compared using Wilcoxon-Mann-Whitney tests, with effect sizes calculated using Glass rank biserial correlation coefficients. Associations between examiner um-rate and participant-level measures were measured with Kendall rank correlation coefficients, with p-values adjusted using the Benjamini-Hochberg procedure to reduce false discovery rate.
Results: There was a significant difference in examiner um-rate between ASD and TD (U = 763.0, p = .007; ASD < TD), with a medium effect size (r_pb = = -.343; Table 1). There was no significant difference in examiner uh-rate between ASD and TD (U = 1,038.0, p = .399) or in examiner um-ratio (U = 1,017.5, p = .369). Examiner um-rate was significantly positively correlated with both participants’ age (τ_b = .280, p_adj. = .002) and mean length of utterance in morphemes (MLUM) (τ_b = .220, p_adj. = .018; Table 2). There was no significant correlation between examiner um-rate and any of the following participant-level measures: full-scale IQ; number of distinct word roots; Children’s Communication Checklist (CCC-2) General Communication Composite, structural language score, pragmatic language score; ADOS-2 Social Affect total, Comparison Score.
Conclusions: Examiners use the filler “um” significantly less with children with ASD than children with TD, which mirrors previous results on “um” usage in ASD and TD. We also found that examiner “um” usage is positively associated with participant age and MLUM but not with more strenuous participant-level measures of expressive language ability and autism symptom severity. Because analyses did not control individual differences between examiners, these results should be interpreted with caution. Further analyses that account for examiner-level measures are needed.