Fig. 6: Neural representation of the support-stay, conflict-shift model.
From: Striatal arbitration between choice strategies guides few-shot adaptation
a Task configuration8. The numbered S marks within the T-maze indicate the position of photo-beam sensors, which represent the boundaries between different sections within a single trial. b Distribution of neurons encoding decision and task variables (n = 201 neurons from 3 rats). c Behavioral dynamics profiles from rats and fitted SSCS model (n = 3 rats). CI, CE, and Conditional CE from the left. The difference between CE-C and CE-S was examined using a linear mixed-effects model, with the number of S+ repetitions (NS) and type (CE-C or CE-S) as fixed factors and subject as a random effect. Estimated marginal means were compared with degrees of freedom adjusted by the Satterthwaite method for main effects and simple main effects tests. d Regression weights from the multi-trial history regression analysis from rats and fitted SSCS model (n = 3 rats). βS and βC from the left. Regression weights were tested against zero using a linear mixed-effects model, with the trials in the past (τ) as the fixed factor and the subject as a random effect. Estimated marginal means were compared with degrees of freedom adjusted by the Satterthwaite method for post-hoc tests. e, f Example temporal dynamics sorted and colored by the value of the corresponding decision variable, Modulations of firing rates are listed in the order of prepare, action, reward, and update section from the left; e Normalized firing rate of support bias neurons. f Normalized firing rate of conflict bias neurons. All statistical tests were two-sided and corrected for multiple comparisons using the Benjamini–Yekutieli procedure. Error bars indicate mean ± s.e.m. See Supplementary Table 2 for full statistical information. Source data are provided as a Source Data file.
