A connectionist model may shed light on neural mechanisms for visually guided reaching.

Abstract Formal principles of vision-based planning and control of arm movements are used to gain insight into the neurobio-logical mechanisms that underlie this important class of sen-sorimotor behavior. The primary conceptual tool used in this work has been a neurally inspired connectionist system called MURPHY that learns to reach for visual targets among obstacles, crudely based on the style of architecture and representations in sensory and motor areas of cerebral cortex. This system has provided a concrete implementation that demonstrates how areas of cerebral cortex could in principle interact to direct both sensory-locked and internally planned reaching movements. We use MURPHY'S simple, artificial "cortex" as a point of departure in the development of two high-level cortical models for visual limb control, involving the supplementary motor area (SMA), areas 5 and 7 of the posterior parietal lobe, and several visually responsive areas including V2, PO, and areas TPO and STP in the upper bank of the superior temporal sulcus.