Canonical neural computation in autism spectrum disorders
David Heeger, Ph.D.
New York University
Marlene Behrmann, Ph.D.
Carnegie Mellon University
Computational neuroscience posits that there is a fundamental organizing principle, called the ‘canonical circuit’, for both the architecture and the function of higher-level brain areas. Two lines of evidence are suggestive of specific changes in canonical cortical circuits in autism spectrum disorders: hypersensitivity to sensory stimuli, and a high prevalence of seizure activity.
Instead of focusing on deficits in a specialized brain area or system (such as the mirror system or the language processing system), David Heeger and colleagues at New York University, Carnegie Mellon University (Marlene Behrmann), and Stanford University (Tony Norcia) propose that generic, brain-wide circuit malfunctions are at the root of autism spectrum disorders. According to this idea, the general malfunction unfolds developmentally such that the end result may be most evident in a particular brain area or system, but the underlying cause can be measured at the computational and circuit level.
To test and refine this hypothesis, Heeger’s team plans to focus on the visual system. This is because there is already an extensive body of research on the visual system as it relates to autism. There are also well-developed computational theories and sensitive experimental methodologies for characterizing neural computations in the visual system. The proposed research is framed specifically in terms of the ‘normalization model’ — a model that explains neural processing throughout the visual system, as well as key aspects of attention, multi-sensory integration, decision-making, audition and olfaction.
Heeger and colleagues plan to develop and validate protocols to assess neural computations in people across the autism spectrum. Techniques will include behavior/psychophysics, functional magnetic resonance imaging and visual-evoked potentials. These experiments will be complemented by theoretical work to explain the empirical findings. If the researchers find positive evidence for specific computational deficits, a future phase of their research will be devoted to detailing the underlying mechanisms.