Probing Symptoms of Autism in a Thin Slice

Background: Interest in big phenotypic data, motivates development of behavioral probes for use in community settings outside the research lab. While the Autism Diagnostic Observation Schedule (ADOS; Lord et al, 2000) is a touchstone for measuring autism’s core symptoms, its required intensive training and lengthy administration duration make it hard to deploy broadly in community settings.

Objectives: Inspired by appraisal from brief, “thin slices” of interaction (Grossman, 2015), and by rapid scoring of subtle pragmatic behaviors using semi-structured probes (Simmons, Paul and Volkmar, 2014), we constructed a 5-minute series of semi-structured probes and gestalt ratings, to score social communication behaviors known to differ in autism. Here we describe their performance.

Methods: Verbally fluent, high-functioning participants comprised a group of autistic children (n=14) and an age-matched group of typically developing controls (TDC; n=20). Ages ranged from 6 to 13 years (M=8.7 (SD=1.7)), for each group, respectively. Diagnostic status was confirmed by expert clinicians using comprehensive evaluation including ADOS. Verbal and full-scale IQ differed slightly between groups (FSIQ: ASD, 103.4 (12.3); TDC, 108.8 (16.3); d=0.37).

Probes were delivered by trained research assistants. They targeted (1) gaze-following joint attention; (2) informativity of speech and gestures, when narrating a story shown in a previously viewed cartoon video; (3) help-seeking communication, after being tasked with writing using a secretly broken pen; and (4) integration of gesture with speech, when describing a previously-manipulated toddler’s shape-sorting toy.

Probes comprised series of increasingly supportive prompts, providing opportunities to demonstrate varying levels of abilities, and facilitating rapid and reliable scoring of participant responses, from video, afterward. Scores matched probe hierarchies, and also included three gestalt ratings for awkwardness, quality of rapport, and quality of responses. Scores took integer values from 0, to 2, 3, or 4. Higher scores indicated lower performance (e.g., more prompting required or greater awkwardness).

Results: The ASD group scored a higher sum of probe scores (possible range = [0, 36], ASD: M=11.7 (SD = 10.0), TDC: 6.7 (3.7), d=0.66, p<.001), indicating worse performance. Speech-gesture integration differed between groups (r=.50, see Fig 1). In the ASD group, speech-gesture integration varied negatively with verbal IQ (r=-.61). There were marginal (p<.1) group differences in overall rapport (r=.35), amount of information in narration (r=.30), and prompt level required before requesting help (r=.42).

Response to subtler joint attention bids was associated with lower ADOS Social Affect (SA; r=.46, p<.05) and total scores (r=.39). Overall quality of responses increased marginally with age (r=-.31). Amount of information provided in narration was related to full-scale (FS; r=-.44), nonverbal (r=-.45), and marginally to verbal IQ (r=-.35).

Within the ASD group, response to subtler joint attention prompts likewise correlated with ADOS SA (r=.59) and marginally with ADOS total scores (r=.50). Awkwardness was marginally related to FSIQ (r=.59).

Conclusions: Group differences and relationships to autism symptom severity suggest the promise of using combined gestalt impressions and structured probes to measure a variety of core behaviors from a brief interaction.