High-Dimensional Gesture Coding for Use in Prospective Gene-Brain-Behavior Studies of ASD

Background: Early gestures predict later social communication skills, including language, prosocial behavior, and peer play (Mundy et al., 1990; Sigman & Ruskin, 1999), and infants and toddlers with autism spectrum disorders (ASD) display gesture impairments by 12 months of age (Wetherby et al., 2007). Siblings of children with ASD, who are at heightened risk for developing an autism spectrum disorder, also demonstrate social communication deficits, often in a mild form (Yirmiya et al., 2006). However, most children with ASD are not diagnosed until ~4 years old (CDC, 2016). Research on social communication deficits in children under three therefore has important implications for improving early identification. In addition, gene-brain-behavior research has been hindered by behavioral metrics that are insufficiently granular. In this exploratory study, we compare fine-grained communicative gesture codes in 2-year-old children at high familial risk for ASD who received a diagnosis (HR-ASD), to those at high risk that did not receive a diagnosis (HR-neg), and those at low risk who did not develop ASD (LR-).

Objectives: Determine whether high-dimensional gesture coding of the Communication and Symbolic Behavioral Scales (CSBS; Wetherby & Prizant, 2003) differentiates diagnostic/risk groups in overall gestures, and in subcategories of behavior regulation, joint attention, and social gestures.

Methods: Trained clinicians administered the CSBS to 44 2-year-olds (LR-: 22, 6 female; HR-neg: 11, 4 female; HR-ASD: 11, 1 female) participating in the longitudinal Infant Brain Imaging Study (IBIS; Estes et al., 2015). Videos were segmented and coded for CSBS communicative gestures by 4 independent annotators. Coders were not informed of participant diagnostic status, and all coding discrepancies were resolved by consensus. Single gestures were summed into groups: Behavior Regulation (give, reach, open hand), Joint Attention (point, show), and Social Interaction (wave, high five, nod head, shake head).

Results: Overall gestures per 10 minutes differentiated the HR-ASD group from the LR- and HR-neg groups (ps<.001, Fig.1), but did not differentiate the LR- and HR-neg groups from each other. The same pattern was evident in the Behavior Regulation and Joint Attention subcategories (ps<.01). In the Social category, gestures per 10 minutes differentiated the LR- group from the HR-ASD group (p<.05), but did not distinguish the HR-neg group from the other two groups.

Conclusions: This project lays the foundation for dimensional gesture coding that can be used in genetic and imaging research, as well as to track developmental change and treatment response. Of note, our study does not address the issue of how gestures are defined and conceptualized. Although we were able to code over 1,700 individual gestures, they may fail to capture the full extent of communicative gestures in children with ASD. In addition, we only coded fully formed gestures – developing and partially formed gestures were coded in the “other” category; our findings may therefore not capture children’s emerging skills. Our next steps include distinguishing between gestures that do and do not carry communicative intent, increasing our sample size, and coding gestures in the same infants at 12 months for use in longitudinal analyses.