Assessing Tactile Perceptual Inference and Learning in Autism Spectrum Disorders

Background:  

Recently, some Bayesian theories were suggested to explain how our perception of the world emerges and how it could be altered in Autism Spectrum Disorders (ASD)^1,2. In these theories, the representation of a stimulus would be influenced by both the precision of the encoding of the stimulus and the priors we have on incoming sensations. Priors correspond to our internal references that are dynamically adjusted through implicit or explicit learning. Thanks to priors, we are quickly able to identify a stimulus as being part of a category. Nonetheless, priors also bias our perception: the information received will be perceived as closer to the priorthan it is.

In ASD, the perception of the world might be less biased than in neurotypical (NT) individuals. This perception, which one could refer to as “too close to reality” would explain subsequent impairments such as failures in generalizing and making suitable predictions. Two hypotheses were suggested to explain this particular perception in ASD: 1-Hypo-prior hypothesis: priors would be too blurred and would poorly influence perception¹; 2-Sensory accuracy hypothesis: sensory information would be encoded very precisely and would have an important weight², resulting in a too acute representation of the inputs.

Objectives:  

Our current project aims to better understand prior construction and their influence on perception in ASD, and to test whether the hypo-prior hypothesis prevails in ASD.

Methods:  

We investigated tactile frequency discrimination and the influence of the global perceptual context on behavioral performance, with NT (n=20) and ASD participants (n=16). At each trial, participants received two successive non-painful electrical finger stimulations to be compared. Given the range of delivered frequencies, they were able to build-up a prioron the forthcoming frequency of stimulation, centered on the mean value of the already perceived frequencies. In NT subjects, such a design results in a perceptual bias: the first stimulation (to be memorized) is underestimated when it is larger than the average of the delivered frequencies; and overestimated when smaller. This task reflects how subjects learn about the global range of stimulations and how this strongly influences their perceptual decisions.

Results:  

As expected, NT participants showed a perceptual bias, also known as a time-order effect (Figure 1-a). However and surprisingly, ASD participants showed an even stronger effect (Figure 1-b). Our preliminary interpretation is that in ASD, the stability of the learning context would favor the construction of “hyper priors”.

Conclusions:  

This study contributes to elucidate the mechanisms underlying perceptual learning in ASD. We hypothesize that in ASD, the influence of the priors can either be too strong (“hyper prior”) in a stable learning context, or too weak (“hypo prior”)¹ in a changing learning context. Neuroimaging studies will be performed to test these hypotheses and to explore the neural correlates of prior construction.

1. Pellicano, E. & Burr, D. When the world becomes ‘too real’: a Bayesian explanation of autistic perception. Trends Cogn. Sci.16(2012).

2. Brock, J. Alternative Bayesian accounts of autistic perception: comment on Pellicano and Burr. Trends Cogn. Sci.16(2012).