A neural model to study sensory abnormalities and multisensory effects in autism.

Computational modeling plays an increasingly prominent role in complementing critical research in the genetics, neuroscience, and psychology of autism. This paper presents a model that supports the notion that weak central coherence, a processing bias for features and local information, may be responsible for perception abnormalities by failing to "control" sensory issues in autism. The model has a biologically plausible architecture based on a self-organizing map. It incorporates temporal information in input stimuli, with emphasis on real auditory signals, and provides a mechanism to model multisensory effects. Through comprehensive simulations the paper studies the effect of a control mechanism (akin to central coherence) in compensating the effects of temporal information in the presentation of stimuli, sensory abnormalities, and crosstalk between domains. The mechanism is successful in balancing out timing effects, basic hypersensitivities and, to a lesser degree, multisensory effects. An analysis of the effect of the control mechanism's onset time on performance suggests that most of the potential benefits are still attainable even when started rather late in the learning process. This high level of adaptability shown by the neural network highlights the importance of appropriate teaching and intervention throughout the lifetime of persons with autism and other neurological disorders.