Fig. 4.

Rule selectivity of GFP neurons. (A) Responses of two exemplary GFP neurons to the presentation of the two rules, shown separately for correct and for error trials. The one on the left exhibited selectivity for the GF rule, while the one on the right preferred the IM rule. (B) Breakdown of rule preferences of rule sensitive GFP neurons for both monkeys. (C, Right) The time course of the population responses of the two pools of neurons preferring the GF rule (n = 60) and the IM rule (n = 44), respectively, tested in both the GF and the IM task. (C, Left) The contributions of individual neurons ordered according to the latency of their peak discharge rates. (D) Exemplary GFP neuron demonstrating that preferences for rules and spatial cues are yoked. This neuron preferred the gaze rule and exhibited a much stronger response to the gaze cue pointing to the target at 10° eccentricity with respect to the observer (G4) in the subsequent 500 ms of the spatial cueing period. Solid lines represent the average response in all trials and the red dashed line represents only gaze-following responses to G4. The vertical line at 400 ms denotes the onset of spatial cue. This neuron preferred nonface objects over faces in the passive viewing task. (E) Deterioration of the rule selectivity of the GFP as captured by the RSI results in erroneous decisions. (F) SVM decoding accuracy obtained from GFP rule-selective neurons for correct trails as compared to error trails. The shaded area represents SDs obtained by bootstrapping (n = 1,000). Note the clear drop in performance for error trials. Error bars and shaded areas represent SE except for F, which shows SDs. ***P < 0.001; ns. P > 0.05.