Extended Data Fig. 9: Variability in neural activity over subjects and relation to avoidance behavior.

(a) Projections of avoid-trial neural activity onto the five coding dimensions (Fig. 4a) for each subject. (b) Same as a for error trials. (c) To test whether variability in neural activity over subjects relates to behavioral variability, we correlated activity in the avoid dimensions to the overall task performance and task 2 shuttle angle over subjects (see following panels). Panel c shows the relation of the magnitude of activity in the avoid 1 dimension (time-averaged over the 3 s window preceding avoid start and averaged over all avoid trials) and overall performance (% avoid trials in active avoidance sessions). We found no clear correlation for this or any of the following analyses (Pearson correlation coefficient = 0.07). (d) Same as c for the avoid 2 dimension (Pearson correlation coefficient = −0.07). (e) Relation between magnitude of activity in the avoid 1 dimension during task 2 (time-averaged over the 3 s window preceding avoid start and averaged over all task 2 avoid trials) and the task 2 shuttle angle (see Extended Data Fig. 1d; Pearson correlation coefficient = −0.04). (f) Same as e for the avoid 2 dimension (Pearson correlation coefficient = −0.02).