Simulation of visual cortex development under lid-suture conditions: enhancement of response specificity by a reverse-Hebb rule in the absence of spatially patterned input.

Abstract

In this report, I show that a reverse-Hebb synaptic modification rule leads to the enhancement of response specificity of simulated visual cortex neurons in the absence of spatial patterning of the afferent activity. Although it is clear that receptive fields in the visual cortex can be modified by experience, many studies have shown a substantial increase of response specificity in cats deprived of pattern vision by lid suture, leading some to conclude that receptive field properties are essentially hard-wired. The hard-wired vs. experience-dependent controversy can be resolved by assuming that while Hebb-type plasticity is responsible for developmental synaptic changes, the organization of presynaptic activity which exists under conditions of visual deprivation is sufficient to drive the neurons towards greater specificity (Linsker 1986a-c; Miller 1989, 1992; Miller et al. 1989). As a reverse-Hebb rule enhances response specificity by balancing the push-pull system of ON- and OFF-center afferents, the sufficient condition is that the activity of ON- and OFF-center retinal ganglion cells be negatively correlated, a condition which will be met by diffuse illumination as seen through sutured eyelids. Unlike the models of Linsker and Miller and colleagues, which are based on a standard-Hebb rule, the model presented here does not require the presence of a "Mexican hat" spatial patterning of the afferent correlations, which has not been observed experimentally.