Gain and Abstraction

Abstract

Gain and normalization are fundamental computational mechanisms that can support various functions in the brain. Numerous neuronal mechanisms support gain control, including divisive inhibition, short-term plasticity of synapses, and subcortical neuromodulators. Gain control allows inputs from the retina and the positions of the eyes in their sockets, the head, and the hands to affect the magnitude of responses to visual inputs in multiple brain regions, particularly the parietal cortex. Gain control mechanisms can shift coordinate representations; for example, from visual space to head space to hand space, or recognize an object as the same when it is viewed from different directions. The mechanisms of translation and object invariance are the neuronal basis of abstraction, a process of ignoring features that are not essential for recognizing entities. Gain control allows judgment of distances independent of locomotion speed. Attention may be viewed as internalized gain control.