Fig. 6: Behavioral performance, reaction times of the monkeys, and multi-linear regression behavioral model reflecting differences in the neural representational geometries.
A Left: average behavioral performance across sessions for Monkey 1 (n = 65 sessions) and Monkey 2 (n = 77 sessions). Both monkeys performed the task with high accuracy. The error bars indicate the confidence interval at 95% of confidence level. Right: average reaction time across sessions for Monkey 1 (n = 65 sessions) and Monkey 2 (n = 77 sessions). A significant difference emerged in the average reaction time between the two monkeys. The error bars are the standard error of the mean. n.s. not significant: chi-square test, p-value = 0.93. *: Mann–Whitney U test, two-sided, p-value = 10−15. B Mean behavioral performance across sessions for Monkey 1 (n = 65 sessions) computed separately for each rule and shape. The x-axis indicates the rule, and the y-axis is the mean performance averaged across sessions. The visual cue of each condition is indicated at the bottom of the plot. On the left(right), the visual cue order reflects shape(rule). n.s. not significant (left, chi-square test, p-value = 0.78; right, chi-square test, p-value = 0.11). C Mean reaction time across sessions for Monkey 1 (n = 65 sessions). As in B, the x-axis indicates the rule, and the y-axis is the reaction time averaged across sessions. The error bar is the standard error of the mean. n.s. not significant: Mann–Whitney U test, two-sided, p-value = 0.05; **: Mann–Whitney U test, two-sided, p-value = 0.02. D The same as in B but for Monkey 2 (n = 77 sessions): n.s. not significant (left, chi-square test, p-value = 0.06; right, chi-square test, p-value = 0.22). E The same as in C but for Monkey 2 (n = 77 sessions): n.s. not significant: Mann–Whitney U test, two-sided, p-value = 0.28; **: Mann–Whitney U test, two-sided, p-value = 10−10. F Weights, averaged across 100 models, of the three independent factors predicting the reaction time of single trial in a multi-linear regression model. Each weight is normalized to the total sum of the weights (see Supplementary Fig. 2).The error bars are the 2 standard deviations of weights across 100 models. The variance explained (r-squared) by the models is 12% and 18% for Monkey 1 and Monkey 2, respectively. *: Mann–Whitney U test, two-sided, previous response (p = 10−32); rule (p = 10−34); shape (p = 10−34). Source data are provided as a Source Data file.
