Figure 5. Predicted versus optimal reward per trial (RTrial).

(A) Scatterplots of RTrial_predict:RTrial_max ratio as a function of session index. Each session was represented by two ratios, one for each reward context. (B) 97% RTrial_max contours for all sessions, computed using the best-fitting HDDM parameters and experienced coherences and reward ratios from each session. Light grey: LR-Left blocks; Dark grey: LR-Right blocks. (C) The monkeys’ adjustments (blue: LR-Left blocks, red: LR-Right blocks) were largely within the 97% RTrial_max contours for all sessions and tended to cluster in the me over-biased, z under-biased quadrants (except Monkey F in the LR-Right blocks). The contours and monkeys’ adjustments are centered at the optimal adjustments for each session.

Figure 5—figure supplement 1. Predicted versus optimal reward rate (RR).

Same format as Figure 5. Mean RR_predict:RR_max ratio across sessions = 0.971 for monkey F, 0.980 for monkey C, and 0.980 for monkey A.