Fig. 7: Network connectivity governed the evolution of outcome encodings during AL.

a Comparisons of the noise correlation between different neuron pairs. Each data point denotes the averaged value of noise correlation across one group of neuron pairs recorded from one session. (***: maximum P = 2.09e-12, two-sided, n = 59). b The Pearson correlations between noise correlations of 7a neurons in the current day and the signal correlations of these neurons in the subsequent day. Different colors denote different learning courses. Each dot denotes the result from two consecutive days. c The correlation between the changes in noise correlations and the changes in signal correlations of 7a neurons across two consecutive days. d Schematic illustration of noise correlation calculation for the within-type pairs. This analysis only included neurons that belonged to the same type (CN, EN, or non-selective) on the current day but transitioned to different types on the next day. In day 2, neuron type 1 and type 2 are outcome-selective (S.), while neuron type 3 is non-selective (N.S.). Colored solid and dashed arrows indicate the within-future-type pairs and across-future-type pairs, respectively. e Noise correlation distributions for an example session. f An ROC analysis was used for quantify the difference in noise correlation between within-future-type pairs and across-future-type pairs. The area under curve (AUC) value quantifies the predictive power of noise correlation for neuron ensemble evolution. The red and gray lines indicate the ROC curves for outcome-selective and outcome-non-selective neurons, respectively. Data are from the same example session shown in (E). g Comparison of the predictive power for neuron ensemble evolution between noise correlation and signal correlation. Each dashed line connects data from the same day. Red and gray dots represent results from outcome-selective and outcome-non-selective neurons, respectively. The red horizontal line represents the average value. The green dots denote data from the example session. Noise correlation was significantly different between within-future-type and across-future-type pairs for neurons which were outcome-selective on the second day, but no such difference was observed for neurons which were outcome-non-selective on the second day. (***P = (3.1e-5, 7.7e-5), paired t-test, two-sided, n = (12, 20)). h Schematic illustration noise correlation calculation the for the across-type pairs. This analysis only included neurons that were initially of different types on the current day but merged to the same type on the following day. i, j Difference in noise correlations between within-future-type pairs and across-future-type pairs in another example session. k Comparison of the predictive power for neuron ensemble evolution between noise correlation and signal correlation. (***P = (5.3e-5, 7.1e-5), paired t-test, two-sided, n = (8, 25)). Data in (j, k) were shown in the same format as in (f, g).