Active touch, exploratory movements, and sensory prediction

Abstract

The relation between somatosensory input and motor output is asymmetric. Somatosensation is associated with every movement an animal makes, but movement is not required for somatosensation. This symposium paper proposes a classification scheme for movement, in which movements are placed along a continuum that describes the role that somatosensory information plays during the movement. Fine sensorimotor control - manipulation and exploration - are found to fall to one extreme of the spectrum, and exploratory movements in particular are shown to possess characteristics that clearly distinguish them from other varieties of movement. Specifically, the exploratory process must permit animals to extract an object's features independently of the sequence of movements executed to explore the object. Based in part on our work on the rat vibrissal system, we suggest that exploration of objects may consist of two complementary levels of sensorimotor prediction operating in parallel. At the cognitive level, the animal might move so as to perform hypothesis testing about the identity or nature of the object. The particular hypothesis tests chosen by the animal might be implemented through sequences of control-level predictions that could be generated at the level of the brainstem and cerebellum.