Fig. 4: Chosen Side and Trial Outcome decoding accuracy is predicted by behavioral strategy only in OFC.
A–G Behavioral measurements that were entered as regressors in omnibus Generalized Linear Models (GLM). The group sizes were the same across all measurements: nM2 = 18, nOFC = 24. The center of the boxplot is the median. Latency measures (A–C) were calculated as medians of all trials for each session. WinStay and LoseShift (D, E) were calculated as the proportion of total trials that were WinStay and LoseShift at the session level. Perserveration Index and Flexibility Index (F, G) were calculated using formulas described in the Methods. H Beta coefficients for regressors where significant area (M2 vs. OFC) interactions were found in predicting decoder accuracy of Chosen Side. Modeling was performed only on the accuracy from the bin right before choice (highlighted in yellow, Fig. 2). I Post-Hoc GLM performed on sessions from rats with lenses in M2. J Post-Hoc GLM performed on sessions from rats with lenses in OFC. Strategy indices were significant predictors of Chosen Side decoding accuracy in OFC, but not in M2. K Beta coefficients for regressors where significant area (M2 vs. OFC) interactions were found in predicting Trial Outcome. Similarly to (H), analyses were only performed on the bin before the reward cue (highlighted in yellow, Fig. 2). L Post-Hoc GLM performed on sessions from rats with lenses in M2. M Post-Hoc GLM performed on sessions from rats with lenses in OFC. Strategy indices were significant predictors of Trial Outcome decoding accuracy in OFC, but not in M2. Source data are provided as a Source Data file.
