Cross-correlations between three units in cat primary auditory cortex

Here we use a modification of the Joint-Peri-Stimulus-Time histogram (JPSTH) to investigate triple correlations between cat auditory cortex neurons. The modified procedure allowed the decomposition of the xy-pair correlation into a part that is due to the correlation of the x and y units with the trigger unit, and a remaining 'pair correlation'. We analyzed 16 sets of 15-minute duration stationary spontaneous recordings in primary auditory cortex (AI) with between 11 and 14 electrodes from 2 arrays of 8 electrodes each that provided spontaneous firing rates above 0.22 sp/s and for which reliable frequency-tuning curves could be obtained and the characteristic frequency (CF) was estimated. Thus we evaluated 11,282 conditional cross-correlation functions. The predictor for the conditional cross-correlation, calculated on the assumption that the trigger unit had no effect on the xy-pair correlation but using the same fraction of xy spikes, was equal to the conventional pair-wise correlation function between units xy. The conditional correlation of the xy-pair due to correlation of the x and/or y unit with the trigger unit decreased with the geometric mean distance of the xy pair to the trigger unit, but was independent of the pair cross-correlation coefficient. The conditional pair correlation coefficient was estimated at 78% of the measured pair correlation coefficient. Assuming a geometric decreasing effect of activities of units on other electrodes on the conditional correlation, we estimated the potential contribution of a large number of contributing units on the measured pair correlation at 35-50 of that correlation. This suggests that conventionally measured pair correlations in auditory cortex under ketamine anesthesia overestimate the 'true pair correlation', likely resulting from massive common input, by potentially up to a factor 2.