A wealth of research has amassed that points to substantial sensory and language processing deficits in autism. This proposal seeks to focus on these two dimensions of functioning—and their interaction—under a mechanistic theory of brain function known as predictive coding. This theory proposes that perception involves a process of inferring the causes of our sensory input by combining that input with our learned/evolved model of the world. Under this framework, some of the symptomatology of autism is accounted for as an atypical comparison of top-down predictions with bottom-up sensory input. In particular, it has been suggested that people with autism underweight their top-down predictions leading them to impute too much importance to their sensory input. This can produce hypersensitivity to and/or repetitive engagement with that sensory input, both of which can have profound effects on emotional and social functioning.
To explore this idea, Edmund Lalor and his colleagues aim to record electroencephalographic (EEG) responses from individuals as they listen to speech under different experimental conditions. Importantly, these experiments will involve manipulating 1) the quality (precision) of speech stimuli (e.g., by adding background noise), and 2) the quality (precision) of prior information (e.g., by using narrative, predictable speech versus random strings of words). Lalor’s team will then deploy state-of-the-art computational data analysis methods to assess the relationship between top-down predictability and bottom-up sensory signal quality.
These experiments will be carried out with neurotypical adults and with adolescents with and without a diagnosis of autism. Comparative analyses between groups will provide an important test of the hypotheses generated under predictive coding theory. In addition, the approach should provide valuable, interpretable measures of sensation, perception and language function in people with autism that can be linked to genetic and other environmental factors in future work.