Dynamical Systems Analysis of Hand Movement Organization in Autism: Association of Movement Organization with Problem-Solving and Symptom Severity

Background: Since Asperger's first descriptions, clinicians have reported striking differences in the conversational gestures produced by individuals with autism spectrum disorders (ASD), noting that gestures are “clumsy” and “inappropriate.” Currently, gestural quantity, quality, and integration with speech contribute to 1/3 of diagnostic items on the ADOS. The limited empirical literature focuses primarily on declarative (attention-drawing) pointing gestures in preschoolers. Furthermore, work from our lab and elsewhere indicates that individuals with ASD gesture as frequently as their peers. Objectives: Dynamical systems theory indicates that behaviors at multiple timescales (e.g., from milliseconds to minutes) can reliably drive behaviors at other timescales. For example, new cognitive structures emerge as a consequence of multi-scale behaviors. By examining hand movements as participants solve a computer-based “gears” puzzle, we assess (1) the contribution of multi-scale behaviors to solving the puzzles, and (2) group differences in behavior structures. If individuals with ASD show distinctive patterns of organization in their hand movements, this may provide some insight both into cognitive differences and into clinicians' observations of gesture impairments. Methods: Participants were 15 high-functioning adolescents with ASD and 14 TD adolescents matched for age, gender, IQ, and receptive vocabulary (p's > .18). Participants completed a computer-based gears task and were videotaped explaining their reasoning after each trial. The ASD group was as accurate for smaller gear systems, but less accurate for 7-gear systems, p = .006. Hand movements during explanation were tracked and examined to assess relative structure (e.g., organization), and the prediction of structure at one timepoint based on the previous timepoints. Specifically, two-dimensional X,Y hand positions of participants as they explained and gestured about their gear puzzle solutions were coded. X,Y position data were subjected to detrended fluctuation analysis (DFA), which can measure activity in a nested multiscale structure. After dividing the integrated time series into bins, a least-squares line was fitted for each bin. The root mean square error (RMSE) of the time series within the bin is the quantity of interest; this measures the activity of the system at a particular scale (i.e., bin size). This procedure is repeated and RMSE is calculated for multiple bin sizes. The scaling relation between RMSE and bin size typically follows a power-law, indexed by the power-law exponent, H. Results: The structure of hand movements over successive trials of the task, in DFA, showed a significant group by time interaction, p = .03. Unlike controls, the ASD group showed no increase in structure over time. For the ASD group, DFA value across trials was also significantly associated with symptom severity (ADOS repetitive behaviors), r =.63, p = .02. DFA was uncorrelated with age and IQ, p's > .35, suggesting it was not simply indexing general abilities. Conclusions: Low-level organization of hand movements was related to gears task performance (e.g., the timing of new concept learning; Dixon & Bangert, 2005). These analyses suggest (1) meaningful differences between ASD and control groups in coordinating motor movements, and (2) associations between this coordination and autistic symptomatology.