Testing the tuning-width hypothesis in a unified theory for autism

Clinicians rely on a triad of behavioral characteristics for the diagnosis of autism spectrum disorders: impaired social interaction, impaired language and communication, and restricted interests and repetitive movements. However, a much broader range of behaviors has been documented in people with autism. For example, individuals with autism can discriminate slightly different grating patterns better than typically developing individuals can, but they struggle to distinguish different facial emotions.

Although many theories of autism have been proposed, each focuses on only a few characteristic behaviors. The lack of a unified account greatly impedes efforts to link behaviors to the underlying physiology, anatomy, and genetics of autism. Additionally, fragmented accounts have sometimes led to opposite therapeutic recommendations.

To address this problem, Ning Qian and his colleagues recently published a theory that coherently explains 18 major behavioral observations in autism spectrum disorders. In addition to the standard triad for diagnosis, the theory takes into account style of information processing, tendency for overwhelm (sensory overload), atypical learning, poor predictive ability and hypersensitivity. The key hypothesis is that brain circuits in individuals with autism respond to a reduced range of inputs (narrower tuning widths), biasing them toward a learning style that is good at rote memorization but poor at extracting complex, fuzzy regularities from daily experience.

Qian aims to test the tuning-width hypothesis via psychophysical methods. If confirmed, the hypothesis would increase our understanding of many behaviors in autism and aid the development of more effective therapies. In particular, the theory suggests that training people on the autism spectrum to learn common social behaviors is similar to training typical people to memorize random facts such as phone numbers; both present a mismatch between learning style and task.