Figure 2
From: Uncertainty of treatment efficacy moderates placebo effects on reinforcement learning
Behavioral measurements over the course of the probabilistic RL task and their correlation with individual RL parameter posterior means. (a, b) The development of reinforcement contingency is shown within each stimulus pair for optimal choices and for collected reward, respectively. For each trial, response values (0 or 1; top) and individual cumulative means (mean of all response values from trial 1 to the depicted trial number; bottom) were averaged across all participants. With increasing trial number, participants learned to make more optimal choices, particularly for the 80:20-pair. Shaded area indicates the standard error at each trial. The pair is designated by opacity of the lines. (c, d) The same procedure as in (a) and (b) was applied to reaction times (c), indicating faster choices with increasing trial number, and the stay probability (d), i.e., probability of three equal consecutive choices given two equal preceding choices. (e) Pearson’s product-moment correlations between RL parameters and individual means for count of optimal choices (purple), collected reward (magenta), reaction time (red), and stay probability (orange). RL parameters are represented on the x-axis, Pearson’s product-moment correlation coefficients on the y-axis. To obtain the temporal correlations, a sliding window of 20 trials was moved in steps of 1, where the first trial window (at trial 20) consisted of trials 1 to 20. Shaded areas represent the uncorrected 95% confidence interval at for each trial window.
