Perceptual Inference in Autism Spectrum Disorders: Insights from Time-Order Effect in Tactile Discrimination

Background:

Perceptual decisions are known to be influenced by global context as much as by sensory evidence¹. Computationally, this can be cast in Bayesian terms where behavior results from the optimal combination of acquired priors and incoming sensations, such that the more ambiguous the sensory input, the larger the influence of prior knowledge. In Autism Spectrum Disorders (ASD), it has been suggested that an imbalance in the way prior information and sensory evidence are trade could explain many of the symptoms.^2,3

In a previous tactile study⁴, we showed that people with ASD were able to build up priors in a stable learning context. However and against some current theories, we found that those priors would have a stronger effect in ASD than in neurotypical (NT) controls. We thus further hypothesized that people with ASD would have difficulties adapting their priors to a new context.

Objectives:  

We used a very similar tactile task as before, so as to reveal the process of prior updating in ASD. We made use of the so-called time-order effect (TOE)⁵, whereby the percept of the first stimulus appears to be strongly biased by contextual priors.

Methods:  

We tested 20 NT adults and 17 adults with ASD (diagnosed according to the DSM V using their ADOS or/and ADI-R score) matched on age, gender and IQ. Non-painful electrical stimulations were delivered on their index finger. Participants were asked to compare the frequencies of two consecutive stimuli (F1, F2). In a first experimental phase, F1 was delivered at 30Hz. In the second phase, F1 varied from trial to trial, with a mean frequency of 26Hz. The first phase was used to induce a strong prior expectation of a fixed first stimulus at 30Hz, while the second phase, designed to elicit a TOE, was meant to trigger a process of prior updating. We recorded performances and used computational modelling to characterize perceptual inference in each group⁶. 

Results:  

NT and ASD participants showed similar performances (75% and 73% correct discriminations on average, respectively). Both groups exhibited a TOE of similar amplitude. However, model comparison showed that contrary to NT, ASD subjects were slow in adapting their perceptual priors when changing the context. We also observed a trend in ASD participants such that the higher the Autistic Quotient, the slower the updating (p = 0.052). Importantly, this effect was not due to a difference in the precision of the encoding of sensory information.

Conclusions:  

ASD people differ from NT in the way they update their perceptual expectations. They appear to be less flexible, which may point to a deficit in the way the relative weight of prior belief and sensory evidence is adjusted⁷. Further assessments should combine advanced computational models, informed task designs and neurophysiological measures in order to refine our understanding of ASD.

References

1.  Summerfield  Nat. Rev. Neurosci. (2014).

2. Pellicano Trends Cogn. Sci. (2012).

3. Van de Cruys Psychol. Rev. (2014).

4. Sapey-Triomphe IMFAR (2015).

5. Karim Front. Neurosci. (2013).

6. Sanchez PhD (2014).

7. Lawson Front. Hum. Neurosci. (2014).