Figure 3.

Comparison of stimulus-based RL in bandit tasks with single versus multiple states. A, Fraction of times that the monkeys in the control group chose the visual stimulus that was initially more rewarding in What blocks during the current study and in a previous study (Costa et al., 2016). Although both studies assessed stimulus-based RL, in the previous study, rewards were exclusively associated with choosing a stimulus (one reward state), whereas, in the current study, rewards were associated with either a stimulus or an action (two reward states). Choice behavior in each task is broken out by reward schedule. Solid lines indicate data from the current task and dotted lines indicate data from the previous task. Shaded regions indicate ± 1 SEM computed across sessions. B, Same procedure in A but for the VS lesion group.