Decreased Experience-Driven Optimization of Eye-Movement Patterns in People with Autism Spectrum Conditions

Background:

Pellicano and Burr (2012) proposed that predictive coding theory may help to understand the mechanism behind the cognitive phenotype of autism, i.e. the diminished ability to generate and employ predictions in the processing of sensory data. In other words, people with autism may see the world without the “glasses of experience”- and in consequence may have difficulty ignoring noise inherent in sensory input and distinguishing signal from noise. This may also manifest itself in the difficulty with the generalization of experience and inflexibility of thinking. Predictive coding account of autism also explains enhanced sensory discrimination abilities (a consequence of very realistic sensory experience), local information bias at the expense of global information or diminished susceptibility to perceptual illusions, which are caused by strong top-down expectations distorting the perception of certain images.

Objectives:

We wanted to investigate the ability of participants with ASC (autism spectrum conditions) to use top-down expectations to optimize their eye-movement patterns. In a previous study with typically developing participants, we showed “economies of experience” in eye-movement patterns, i.e. optimization of eye-movement patterns aimed at more efficient and less costly visual processing. We showed that eye-movement dispersion, velocity, and the number of fixations decreased with each stimulus presentation. This suggests that eye-movement patterns can evolve to facilitate the optimal processing of a given stimulus via experience-driven perceptual learning. The aim of this study was to test whether this experience-driven optimization would be present to a similar extent in participants with ASC.

Methods:

21 participants with ASC diagnoses confirmed by ADOS, and 23 age-, IQ- and gender-matched typically developing (TD) participants took part in the study. Participants’ eye movements were recorded using a remote eye-tracking device SMI RED250Mobile, with a sampling rate of 250 Hz. Participants looked at Mooney-type, degraded stimuli that were difficult to recognize without prior experience, but easily recognizable after exposure to their undegraded versions, in a design similar to a Loth, Gómez and Happé (2010) study. Participants first saw the degraded version of the stimulus, which was used to record the eye-movements without prior experience of the stimulus. Next, participants saw the original, disambiguated photograph, immediately followed by the degraded version again. These presentations were used to disambiguate the degraded stimulus and enable perceptual learning, to make sure that the participants would be able to link the degraded and undegraded stimuli. Finally, after a delay, participants saw the degraded version again, interspersed with other stimuli, which was used as a measure of perception with prior experience and the outcome of perceptual learning.

Results:

We compared the first presentation (pre-experience) and the last presentation (post-experience). We performed a mixed-model analysis, controlling for intelligence, age and ASC diagnosis, of several eye-tracking measures. We found that participants with autism used their experience to optimize their eye-movement patterns to a smaller degree compared to TD participants.

Conclusions:

This study provides evidence in support of the predictive coding account of autism and identifies atypical perceptual learning in people with ASC, who may underuse top-down expectations to optimize eye-movement patterns.