Toddlers with ASD are better at visual search without trying harder: a pupillometric study

Abstract

Introduction: Recently, we found that 2.5-year-olds with Autism Spectrum Disorder are much better than age-matched, typically developing controls at ‘feature conjunction’ visual search (e.g. finding the target up to 2-3x as often in a fixed duration trial, Kaldy et al., 2011). But how does the ASD group achieve superior performance? One hypothesis is that they exert greater cognitive effort than typically developing children (searching more dedicatedly and/or faster), while another is that they search more efficiently. To examine this, we compared changes in pupil dilation (a standard measure of cognitive effort, Beatty, 1982) during search. Methods: A unique aspect of our search paradigm is that it does not require verbal instructions, making it ideal for populations with weak language skills. 17 typically developing toddlers and 17 with ASD participated (diagnosis was confirmed by ADOS, mean age: 29.6 +/- 4.8 months). Test stimuli consisted of four single-feature trials (color and shape; set sizes 5 or 9) and nine conjunction trials (set sizes 5, 9 or 13) in mixed blocks. The display was presented for 4 seconds, then the target item rotated; acting as feedback and reward. A Tobii T120 recorded eye movements and pupil diameter throughout. Results: There were no significant differences in pupil changes between the ASD and the typically developing group during search, or any of the other trial events (onset, offset, reward animation). Conclusion: Pupillometry indicates that toddlers with ASD do not exert greater cognitive effort (i.e. search harder) in conjunctive visual search than typically developing children (consistent with measures such as number of fixations and overall search time per trial, which are also not significantly different). Instead, the ASD group’s superior performance is likely due to perceptual and attentional factors that raise target saliency and hence the efficiency of guided search. Meeting abstract presented at VSS 2012