Visual Motion Prediction and Verbal False Memory Performance in Autistic Children

Background: Recent theoretical accounts propose that atypical predictive processing has a causal role in many aspects of the autism phenotype, from sensory processing to social symptoms (e.g., Sinha et al., 2014). Moreover, difficulties making predictions may be related to reduced contextual processing, as the ability to make good predictions requires using contextual information about the situation at hand, and generalising from similar situations experienced previously. Despite the potential of these new theoretical accounts, there is relatively little empirical evidence for disordered prediction in autism.

Objectives: In this study we tested three hypotheses. First, we hypothesised that autistic children would make less precise (i.e., more variable) predictions than typically developing children in visual extrapolation tasks, and that group differences would become more pronounced as children were required to make predictions over longer timescales. Second, we hypothesised that autistic children would be less susceptible to false memories, in line with reduced contextual processing. Third, we hypothesised that reduced precision in the prediction task would be related to increased sensitivity in the false memory task.

Methods: The study was pre-registered prior to data collection (https://osf.io/pv4w6/register/565fb3678c5e4a66b5582f67). Thirty autistic children aged 6-14 years and 30 typically developing children matched in age and non-verbal IQ completed visual extrapolation and verbal false memory tasks to assess predictive abilities and contextual processing, respectively. In the visual extrapolation tasks, children were asked to predict when an occluded car would reach the end of a road and when an occluded set of lights would fill up a grid. In the false memory task, children were asked to remember word lists and were then given a recognition test including semantically related, non-presented items. Where non-significant group differences were obtained, we used Bayesian analyses to quantify evidence for the null hypothesis.

Results: In the extrapolation tasks, autistic children made predictions that were just as precise as those made by typically developing children, across a range of occlusion durations. In the false memory task, the autistic and typically developing children did not differ significantly in their discrimination between items presented in a list and semantically related, non-presented items, although the data were insensitive, suggesting the need for larger samples. Insofar as there were relationships between performance in the two tasks, they were opposite to our hypothesis, with children who were more sensitive in the false memory task showing less variable performance in the extrapolation tasks. This relationship may reflect domain-general factors (e.g., motivation) or the involvement of a memory component.

Conclusions: Our findings provide an important test of rapidly developing theories and suggest that autism is not characterised by pervasive impairments in prediction. Prediction difficulties may well become apparent when making predictions between events that are more weakly associated, such as those involved in social interactions, or when it is important to decide what information is relevant to use for prediction. Further studies will be required to assess the relationship between predictive processing and context use in autism, and to establish the conditions under which predictive processing may be impaired.