Supplementary Figure 5: Dopamine predicts Wrong Go selections prior to the error being signaled.

Average unsmoothed dopamine release from experiment 1 (mean ± S.E.M.) recorded during successful (filled line) and wrong (dotted line) Go trials (trials where the animal correctly initiated an action but selected the wrong lever) aligned to the time of (a, d) cue onset, (b) head exit from nose-poke or (c) the first lever press (when the error is signaled). Lower panels show average discriminability between the Correct and Wrong Go trials for each timepoint (shaded area = population of 1000 permuted sessions; line: *, p < 0.05 permutation tests, corrected for multiple comparisons). (d-e) Comparison of dopamine release on successful and wrong Go trials as a function of speed of action initiation. (d) Successful (blue lines) or Wrong (red lines) Go trials sorted by “fast” (< 1 s, filled line) or “slow” (> 1 s, dotted line) nose-poke exit latency. (e) Regression weights for trial accuracy and action initiation latency for Go trials (a.u., arbitrary units). On average, rats were both significantly slower to exit the nose poke (p = 0.03, W₇ = 1, Wilcoxon Signed Rank Test) and to move from the nose poke to make a lever press on Wrong trials (p = 0.02, W₇ = 0, Wilcoxon Signed Rank Test). Nonetheless, these differences could not fully explain the different patterns of dopamine release on the Correct and Wrong trials.