The Role of Prior Knowledge in Perceptual Inference Is Preserved in Autism Spectrum Disorder

Background:

Several researchers have exploited the Bayesian framework to propose an imbalance between bottom-up (sensory) and top-down (prior) information processing in Autism Spectrum Disorder (ASD). One recent Bayesian account about ASD is the ‘weak priors’ hypothesis. This hypothesis can be traced back to principles of older perceptual theories about ASD (Enhanced Perceptual Functioning and Weak Central Coherence), suggesting that perception of individuals with ASD is more determined by sensory input instead of by prior knowledge. However, studies empirically testing the ‘weak priors’ theory have generated contradictory results and alternative accounts of ASD exist that do not assume impaired acquisition of prior knowledge per se.

Objectives:

Here we aim to provide clear evidence for or against the ‘weak priors’ hypothesis by obtaining a distinct measurement of prior knowledge using Mooney images. On first viewing, Mooney images are experienced as meaningless black and white patches. After exposure to the source image however, prior knowledge causes the patches to (re)organize and be perceived as a meaningful whole. Hence, we can investigate the role of priors in Typically Developing (TD) individuals varying in ASD-like traits (study 1) and ASD individuals (study 2) by looking at improvements in recognition accuracy for Mooney images, before and after exposure to their source image.

Methods:

In the first study, 282 TD individuals (40 males, mean age 18.6 ± 1.91) performed the experiment and the Autism-Spectrum Quotient (AQ) questionnaire. In the second study, the experiment was completed by 23 adolescents with ASD (19 males, mean age 14.04 ± 1.49) and 24 matched TD individuals (20 males, mean age = 14.38 ± 1.28). The experiment in both studies consisted of six experimental blocks, each made up by three phases of 10 trials (10 different Mooney images). In the first phase, participants were asked to identify the object in the Mooney images (open response). Then the source images were presented in random order. After this exposure phase participants were presented with the same Mooney images as in the first phase and again had to try to recognize the image.

Results:

Participants of the first study had an average AQ score of 17.73 (range of 3-43), comparable with other studies, but there was no difference between the accuracy data of individuals with lower and higher AQ. Results of the second study also showed no difference between the accuracy data of TD and ASD individuals. GLMM analyses confirmed both findings.

Conclusions:

Our findings consistently did not support a difference in the impact of prior knowledge in perception, as a function of varying ASD-like traits or a clinical ASD diagnosis. These findings, thus, suggest that the fast formation and application of priors is preserved in ASD and go against the ‘weak priors’ hypothesis. Earlier studies reporting an intact use of priors in ASD are consistent with this result, while other studies do claim to have found evidence for weaker priors in ASD. Future research will have to systematically evaluate the differences in settings and priors concerned.