Figure 11.

The five-parameter model applied to data from the first two blocks of Experiment 2. (A) Applied to the first 80 trials, during which subjects only experienced one trial type, there were no differences between conditions in mean estimated value of LR+, LR−, or β; however, R0rew < 0 (similar to a punisher) for subjects in the Reward-First condition, while R0pun > 0 (similar to a reward) for subjects in the Punish-First condition. (B) Looking at individual data, for subjects in the Reward-First condition, who only experienced reward-based trials over the first 80 trials, there was no significant correlation between estimated values of R0rew and performance; (C) but for subjects in the Punish-First condition, higher (more positive) values of R0pun were associated with better performance.