Auditory cortex: representation through sparsification?

Abstract

The recent discovery of combination-sensitive neurons in the primary auditory cortex of awake marmosets may reconcile previous, apparently contradictory, findings that cortical neurons produce strong, sustained responses, but also represent stimuli sparsely.

Figure 1

Sparse coding (A) Neural codes are often assumed to be dense. In a dense code, a large proportion of neurons produce action potentials in response to every stimulus, and so activity (indicated by gray shading) is approximately evenly spread throughout the neural population. (B) In contrast, in a sparse code, a small, changing subset of neurons produces strong responses to each stimulus (black shading), whereas the rest of the population is quiescent or weakly active (white and pale shading). The resulting exponential response distribution of firing rate and the statistical independence between neurons reduce the redundancy of the neural code, while maximizing the information content of each action potential.