Do Children with Autism Show Reduced Susceptibility to Visual Illusions?

Background:  Many theories of autistic perception propose that individuals with autism are less susceptible to perceptual biases than typically developing individuals.  One class of evidence in support of this claim comes from visual illusions, with reports of reduced susceptibility in autistic individuals for a range of different illusions (Happé, 1996; Ishida et al., 2013; but see also Ropar & Mitchell, 1999, 2001). It is generally assumed that differences in responses to visual illusions reflect differences at the level of the percept. However, it is possible that group differences instead reflect differences in higher-level decision-making strategies.

Objectives: We aimed to measure the perceptual biases of children with autism and typically developing children as purely as possible, by minimising the influence of cognitive factors such as decision rules.

Methods: We used a 2-alternative-forced-choice method with a roving pedestal (Morgan, Melmoth & Solomon, 2013) to quantify internal noise and decision biases for Ebbinghaus stimuli in 20 children with autism aged 6 to 14 years and 26 typical children matched in age and non-verbal ability.  Children were presented with a reference stimulus and two comparison stimuli (see Figure 1), and asked to identify which comparison stimulus had a central circle most similar in size to that of the reference stimulus.  One comparison stimulus was a pedestal, which had a central circle either 5% larger or 5% smaller than the reference stimulus. The other comparison stimulus had a central circle that was an increment larger than the pedestal. The pedestal size (+5%, -5%) was randomly interleaved throughout the task, so that children did not know which of the two comparison stimuli was the pedestal on a given trial.  Children completed this task in two context conditions: once with small surrounding circles on the reference and large surrounding circles on the comparison stimuli (S-L), and once with large surrounding circles on the reference and large surrounding circles on the comparison stimuli (L-S; see Figure 1).  The data were fit with a cumulative normal psychometric function using the maximum likelihood estimate technique, modelling the effect of context condition as an equivalent pedestal with no effect on internal noise. A difference score was computed between the biases in each context condition, in order to determine the extent of bias for each participant. 

Results: Children with autism differed from typical children neither in their levels of internal noise nor in their degree of perceptual bias.

Conclusions:  Our findings suggest that children with autism are just as susceptible to the Ebbinghaus illusion as typical children when decisional and response biases are minimised.  Our results are inconsistent with theories proposing reduced contextual integration in autism (e.g., Frith & Happé, 1994).  Furthermore, our findings suggest that previous reports of reduced susceptibility to illusions arise from differences in response or decisional criteria. We are currently collecting data with the Muller-Lyer illusion to investigate the generalisability of this result.