Correlated discharges in the primate prefrontal cortex before and after working memory training

Abstract

The correlation of discharges between single neurons can provide information about the computations and network properties of neuronal populations during the performance of cognitive tasks. In recent years, dynamic modulation of neuronal correlations by attention has been revealed during the execution of behavioral tasks. Much less is known about the influence of learning and performing a task itself. We therefore sought to quantify the correlated firing of simultaneously recorded pairs of neurons in the prefrontal cortex of naïve monkeys that were only required to fixate, and to examine how this correlation was altered after they had learned to perform a working memory task. We found that the trial-to-trial correlation of discharge rates between pairs of neurons (noise correlation) differed across neurons depending on their responsiveness and selectivity for stimuli, even before training in a working memory task. After monkeys had learned to perform the task, correlated firing decreased overall, although the effects varied according to the functional properties of the neurons. The greatest decreases were observed on comparison of populations of neurons that exhibited elevated firing rates during the trial events and those that had more similar spatial and temporal tuning. Greater decreases in noise correlation were also observed for pairs comprising one fast spiking neuron (putative interneuron) and one regular spiking neuron (putative pyramidal neuron) than pairs comprising regular spiking neurons only. Our results demonstrate that learning and performance of a cognitive task alters the correlation structure of neuronal firing in the prefrontal cortex.

Figure 1

A) Schematic diagram of the monkey brain with the area of recordings highlighted. Abbreviations: AS, Arcuate Sulcus; PS, Principal Sulcus. B) Successive frames illustrate the sequence of stimulus presentations. Stimuli were white squares presented on a 3×3 grid, with a spacing of 10° from each other. In the pre-training stage the animals were rewarded for maintaining fixation after the end of the second delay period. After training, the animals were presented with two choice targets (not shown) and were required to saccade to a green target if the two stimuli were matching and to a blue target otherwise.

Figure 2

A) Average noise correlation is plotted for neuron pairs with significant responses to visual stimuli. Noise correlation was computed separately for the fixation period, stimulus periods, and delay periods following the stimulus presentation, in recordings conducted prior to (N=950 pairs) and after training (N=807 pairs). Error bars represent standard error of the mean. B) Time course of noise correlation. Gray bars represent the time of stimulus presentations; vertical line, the time of reward in the pre-training condition and the time of choice targets appearance in the post-training condition. Shaded area around each curve represents the standard error of the mean. Noise correlation was computed in 100 ms bin windows; differences in average noise correlation values compared to panel A are due to the difference in integration window. C) Distribution of noise correlation values computed during the stimulus period, for all pairs of neurons before and after training. D) Average noise correlation values are plotted for groups of neurons binned based on the geometric mean of firing rate during the first stimulus presentation period, before and after training.

Figure 3

Average (and standard error of the mean) noise correlation is computed for pairs of neurons with different properties, before and after training. A) All pairs of recorded neurons, whether they responded to the task or not (N=2678 pre; 1730 post-training). B) Pairs of neurons both members of which did not respond to the stimuli at any task period (N=476 pre; 260 post-training). C) Pairs of neurons driven by the stimuli but non-selective for the visual stimuli, in any task period (N=221 pre; 239 post-training). D) Pairs of neurons driven by the stimuli and selective for the nine stimulus locations (ANOVA, p<0.05) are shown (N=209 pre; 226 post-training).

Figure 4

Average noise correlation (and standard error of the mean) is shown for the passive and active condition, after training. A) Average noise correlation for neurons in each task epoch (N=165). B) Average noise correlation during the second stimulus presentation epoch, separately for match and nonmatch stimuli.

Figure 5

A) Average value of noise correlation (and standard error of the mean) computed separately for pairs of RS units, both members of which were putative pyramidal neurons (N=821 pre; N=653 post-training). B) Average value of noise correlation for FS-RS pairs, consisting of one putative interneuron and one putative pyramidal neuron (N=119 pre; N=147 post-training). C) Average value of noise correlation for RS-RS neuron pairs grouped based on signal correlation. D) Average value of noise correlation for RS-FS pairs grouped based on signal correlation.

Figure 6

A) Noise correlation in the fixation period is plotted as a function of signal correlation (which was computed in the stimulus presentation period) for pairs of neurons with significant stimulus responses. Regression lines are shown for pairs of neurons recorded before (blue lines) and after training (red lines). B) Average noise correlation in the fixation period is plotted as a function of temporal correlation (which was computed based on averaged firing rates in successive 0.5 s periods). Regression lines are plotted as in A. C) Average values from panel A; neuron pairs were binned according to signal correlation values. Error bars represent standard error of the mean. D) Average values from panel B. E, F) Noise correlation as a function of both signal and noise correlation, before and after training, respectively. Color value in each bin represents the average noise correlation of neurons with signal and temporal correlation corresponding to this bin. Dark blue color indicates bins with no pairs.

Figure 7

A–E). Average cross-covariance of noise correlation is plotted as a function of trial number, in the fixation, first stimulus, first delay period, second stimulus, and second delay period, respectively. F-J) Average auto-covariance of noise correlation is plotted in the same fashion, in the same intervals.