Testing the hyperspecificity hypothesis: a neural theory of autism

Children'S Hosp Of Philadelphia, Philadelphia PA

Investigators

Linked publications, trials & patents

Abstract

DESCRIPTION (provided by applicant): The hypothesis that a core dysfunction in autism spectrum disorders (ASD) lies in the global relationships between perceptual representations of objects and object features has been central to several very prominent theories of ASD, including the weak central coherence (Frith, 1989; Happ¿ & Frith 2006) and reduced generalization (Plaisted, 2001) accounts. Despite the popularity and longevity of these theories in the literature, little success has been achieved in validating or instantiating them with respec to underlying neural mechanisms. Bridging this divide between cognitive science and brain mechanism represents a significant hurdle in ASD research, and innovation in experimental approach is essential. The hyperspecificity hypothesis (McClelland, 2000) lays the framework for just such a bridge, suggesting that sparser, less inter-connected representation of objects or events drives altered cognitive functioning in ASD by reducing the neural relationships necessary for a flexible understanding of the world. Neural encoding schemes exist along a spectrum of breadth. At sparest coding extreme are independent grandmother cell representations for objects and events. At the other end of this continuum are maximally overlapping distributed codes. The hyperspecificity hypothesis of ASD processing predicts sparser, lesser overlapping neural codes for visual perception among individuals with ASD compared to broader representations in typically developing individuals. Such narrower tuning (Gustafsson, 1997) has direct theoretical corollaries to the classical phenotypic features of ASD, including difficulties recognizing facial identities and expressions. The current proposal represents the first ever neural test of the hyperspecificity hypothesis of ASD. Testing involves application of new methods that our labs developed for implementation with fMRI and ERPs. We predict that fMRI based probe will reveal sparser neural encoding in ASD, and that indices of encoding sparseness will correlate with deficits on face recognition tests. This pursuit is one of the first to link a viable network hypothesis of ASD to an experimentally testable neural instantiation. Such a finding would propel future research into the neural etiology of the disorders, offering a neural foothold for evaluating the interactions of genetics, development and learning on the process of perception. Further, our proposal has the potential to inform treatment by elucidating the core neural mechanisms that can be the focus of new interventions.