Neural Underpinnings of Autistic Reasoning and Novel Problem Solving

Isabelle Soulieres, Ph.D.1, Michelle Dawson2, Fabienne Samson3, Elise B. Barbeau3, Cherif Sahyoun4, Thomas A. Zeffiro, MD, PhD5, and Laurent Mottron, MD, PhD6. (1) Psychiatry, Massachusetts General Hospital/ Harvard Medical School, CNY Bldg 149, 2nd floor, 149 13th street, Charlestown, MA 02129, (2) Clinique specialisee de l'autisme, Hopital Riviere-des-Prairies, Canada, (3) Riviere des Prairies Hospital, University of Montreal, Montréal, QC H1E 1A4, Canada, (4) Harvard/ MIT, (5) Neural Systems Group, Massachussetts General Hospital, Boston, MA, (6) Psychiatrie, Université de Montréal, 7070 bvd Perras, Montréal, QC H1E1A4, Canada

Background: Raven’s Progressive Matrices (RPM) is a pre-eminent measure of fluid intelligence, assaying abilities to infer rules, manage goal hierarchies, and form high-level abstractions. Autistics perform significantly better on RPM than on Wechsler intelligence scales, while no such discrepancy is found in nonautistics (Dawson et al., 2007). Neuroimaging studies of matrix reasoning and novel problem solving in nonautistics have demonstrated task-related activity in a bilateral frontoparietal network.

Objectives: To investigate the neural basis of autistics’ RPM performance using fMRI.

Methods: Fifteen autistics and 18 nonautistics participated, matched on age (mean 22 years), Wechsler IQ (mean 103), sex and manual preference. The 60 RPM problems were presented in random order in one continuous self-paced run using echoplanar imaging on a 3T MRI system. Sixty visual pattern matching problems, similar in format to RPM, but not involving complex reasoning, were presented in a separate session. Performance and task related activity were compared between groups for both problem types.

Results: While accuracy for the RPM and pattern matching problems was equal in both groups, nonautistics were 40% slower in solving RPM problems. For the RPM problems, both groups exhibited similar patterns of task-related activity in occipital cortex, posterior parietal cortex, lateral premotor cortex, insula and cerebellum; with higher activity in nonautistics in the left middle frontal gyrus and medial posterior parietal cortex, and higher activity in autistics in a left extrastriate area (cuneus, BA19). This extrastriate region was active in both groups during pattern matching problems, but only in the autistic group during the RPM problems.

Conclusions: Autistics’ faster performance in solving randomly-ordered RPM problems may indicate a net processing efficiency resulting from increased use of posterior and decreased use of frontal processing mechanisms. Areas primarily involved in perceptual processing in nonautistics may be utilized in reasoning and complex problem solving in autistics.