Periodic Reporting for period 1 - PreCoASD (Assessing the predictive coding accounts of autism spectrum disorders)
Reporting period: 2019-07-01 to 2021-06-30
Autism Spectrum Disorder (ASD) affects more than seven million individuals in the European Union. Yet, this neurodevelopmental condition remains poorly understood. ASD is defined by persistent deficits in social interactions and communication, and by restricted interests and repetitive patterns of behaviors. Recent theories offer potential accounts of ASD. These theories are cast in the predictive coding framework, which assumes that the brain constantly generates predictions about its environment. These predictions, or priors, constitute a kind of internal representation of our environment. Using priors, we know what to expect and we can more easily infer the meaning of stimuli or anticipate. Our perception does not only depend on the sensory information, but also on prior knowledge. The relative contributions of priors and sensory inputs depend on their precisions (i.e. weight). A suboptimal balance of prior and sensory precisions could be at the core of ASD. Predictive coding theories suggested that ASD might be characterized by a low prior precision, a high sensory precision and/or an inflexible ratio of precisions.
The main goal of this research project was to test these theories. Specifically, we aimed at a better understanding of how autistic individuals learn and adjust their priors and at characterizing their underlying neural mechanisms. We also aimed at determining whether sensory prevision is atypical in ASD, at the self-reported, behavioral and neural levels. Finally, we aimed at determining whether atypical predictive mechanisms could underlie the autistic symptoms and co-occurring problems. Altogether, our main objective was to refine the predictive coding theories of ASD and to contribute to a better understanding of the mechanisms underlying the symptomatology of ASD.
The main conclusions of the project indicate that predictive mechanisms are sometimes atypical in ASD, but autistic individuals are able to learn a prior. Sensory precision tends to be higher in ASD for certain visual features. Difficulties in adjusting predictions and atypical sensory sensitivity were associated with the number of autistic traits and with some other symptoms of ASD.
The main goal of this research project was to test these theories. Specifically, we aimed at a better understanding of how autistic individuals learn and adjust their priors and at characterizing their underlying neural mechanisms. We also aimed at determining whether sensory prevision is atypical in ASD, at the self-reported, behavioral and neural levels. Finally, we aimed at determining whether atypical predictive mechanisms could underlie the autistic symptoms and co-occurring problems. Altogether, our main objective was to refine the predictive coding theories of ASD and to contribute to a better understanding of the mechanisms underlying the symptomatology of ASD.
The main conclusions of the project indicate that predictive mechanisms are sometimes atypical in ASD, but autistic individuals are able to learn a prior. Sensory precision tends to be higher in ASD for certain visual features. Difficulties in adjusting predictions and atypical sensory sensitivity were associated with the number of autistic traits and with some other symptoms of ASD.
In the first work package focused on priors, we conducted two behavioral studies and one neuroimaging experiment. First, we developed and tested two visual tasks to better understand how autistic individuals learn and adjust priors (both implicitly and explicitly). These studies showed that autistic individuals could learn priors, but had difficulties to adjust them according to the context. We also found that autistic participants were biased by their priors, contrary to the hypothesis suggesting hypo-priors in ASD. These behavioral experiments enabled us to develop a neuroimaging experiment where we used functional Magnetic Resonance Imaging (fMRI) to characterize the brain regions that were activated when signaling priors or prediction errors (i.e. mismatch between priors and sensory inputs). We were able to precisely identify these regions in an analysis relying on computational models. We observed a few group differences in some regions that were involved in signaling priors and prediction errors, which might explain why autistic individuals could have enhanced sensations of surprise. We also used another non-invasive neuroimaging method, called Magnetic Resonance Spectroscopy (MRS), to measure the concentrations of two neurotransmitters in a brain region that encodes priors. We observed that individuals who had a higher ratio in the neurotransmitters of interest had more difficulties to make predictions. As the autistic group had an excess of one of these neurotransmitters in this region, it might contribute to their difficulties to learn predictions in certain contexts. In total, these three experiments included 163 participants with or without ASD.
In a second work package focused on sensory precision, we conducted experiments assessing visual sensitivity and responsiveness at the self-reported, behavioral and neural levels. Participants were 49 adults with and without ASD. As expected, at the self-reported level, questionnaires indicated more atypical sensory sensitivity, more hypersensitivity and responsiveness in ASD. At the behavioral level, autistic participants were more sensitive and responsive to certain simple visual stimuli than controls. At the neural level, we acquired data using electroencephalography (EEG), combined with a method to implicitly determine the threshold from which the brain starts responding to a visual stimulus. The analysis of the EEG is not finished yet, but should contribute to better understand the neural mechanisms related to the hypersensitivity encountered in ASD.
Finally, in a last work package focused on the integration of the results mentioned above with the symptoms of ASD, we investigated whether atypical predictive processes could be related to the high intolerance of uncertainty experience by autistic individuals, and how it was related to the symptoms of ASD. Among other results, individuals who were not much influenced by priors reported a more atypical sensory sensitivity. Unexpectedly, the ability to learn predictions was not related to the self-reported intolerance of uncertainty. Using online questionnaires involving 426 adults with and without ASD, we also explored how intolerance of uncertainty was related to autistic traits, social difficulties, repetitive behaviors, sensory sensitivity, anxiety and sleep problems. All these results will be integrated together to suggest a refined predictive coding theory of ASD.
These results have been exploited and disseminated as scientific articles in peer-reviewed journals (3 articles are already published and 3 are in preparation), oral presentations as invited speaker (7) and poster presentations in international conferences (3).
In a second work package focused on sensory precision, we conducted experiments assessing visual sensitivity and responsiveness at the self-reported, behavioral and neural levels. Participants were 49 adults with and without ASD. As expected, at the self-reported level, questionnaires indicated more atypical sensory sensitivity, more hypersensitivity and responsiveness in ASD. At the behavioral level, autistic participants were more sensitive and responsive to certain simple visual stimuli than controls. At the neural level, we acquired data using electroencephalography (EEG), combined with a method to implicitly determine the threshold from which the brain starts responding to a visual stimulus. The analysis of the EEG is not finished yet, but should contribute to better understand the neural mechanisms related to the hypersensitivity encountered in ASD.
Finally, in a last work package focused on the integration of the results mentioned above with the symptoms of ASD, we investigated whether atypical predictive processes could be related to the high intolerance of uncertainty experience by autistic individuals, and how it was related to the symptoms of ASD. Among other results, individuals who were not much influenced by priors reported a more atypical sensory sensitivity. Unexpectedly, the ability to learn predictions was not related to the self-reported intolerance of uncertainty. Using online questionnaires involving 426 adults with and without ASD, we also explored how intolerance of uncertainty was related to autistic traits, social difficulties, repetitive behaviors, sensory sensitivity, anxiety and sleep problems. All these results will be integrated together to suggest a refined predictive coding theory of ASD.
These results have been exploited and disseminated as scientific articles in peer-reviewed journals (3 articles are already published and 3 are in preparation), oral presentations as invited speaker (7) and poster presentations in international conferences (3).
As ASD concerns 1 to 2% of the population, there are large societal implications of research projects focused on a better understanding of this condition. Overall, this project contributes to elucidating the mechanisms that underlie the heterogeneous symptoms of ASD. It also opens the path to refining the promising predictive coding theories of ASD. Our studies presented ASD as a difference of perception, with advantages or disadvantages depending on the context. Hopefully, it could also contribute to raising awareness among the global population to promote more tolerance about neurodiversity. It may also have an impact on the practice of clinicians and other professionals working with autistic individuals. For instance, it can contribute to identifying the contexts that will favor a better and more flexible learning or less sensory overwhelming for individuals with ASD. Future research is encouraged to continue shedding light on the causes of ASD and enlightening the symptoms. It would also be useful if future studies would help to strengthen the bridge between the basic research we conducted and the clinical practice by conducting translational research (e.g. develop and test psycho-education programs, develop and test smart games that would rely on the principles revealed by our research, establish tighter links to neurostimulation and/or neuropharmacological approaches).