Matthew Schaff

Department of Neuroscience

University of Pennsylvania

Pairwise correlations in macaque auditory cortex predict multi-neuronal firing patterns

Our understanding of the function of the auditory cortex derives almost exclusively from experiments that study the coding properties of single neurons. Although these studies have shed important information, they are limited because they have not considered how neuron-to-neuron (i.e., pairwise) correlations may affect population coding. We, therefore, sought to determine whether, in the primate auditory cortex, pairwise correlations contribute information to population coding above and beyond the information provided by the independent firing of auditory-cortex neurons. We recorded neuronal activity in the primary auditory cortex in a male rhesus monkey. Neuronal activity was recorded with multi-contact linear u-probe electrodes. While recording neuronal activity, the monkey passively listened to ripple stimuli. Based on either (1) the independent spiking activity of these neurons or (2) the independent spiking activity plus their pairwise correlations, we generated “Ising” models. Ising models are the maximum-entropy distributions that are consistent with experimentally observed firing rates and pairwise correlations without any mechanistic assumptions. We report that Ising models more accurately predict the experimental probability of population dynamics when pairwise correlations are taken into account along with independent spikes. This accuracy difference increases when predicting higher-order population dynamics. This study lays groundwork for future studies that determine the relative importance of correlations in processing different types of acoustic stimuli.

A pizza lunch will be served.