Figure 1.

Experimental design. (a) Time bandit task, where each option was randomly mapped to the [Q, W, O, P] keys on the keyboard, with a different mapping each round. Participants completed 40 rounds (each containing 20 trials), where we manipulated time pressure (b) and payoff conditions (c) in a crossed, within-subject design. (b) In unlimited time rounds, participants could take as long as they wanted to make each selection and received positive feedback (happy face) and were shown the value of the acquired payoff for 400 ms (feedback period). In limited time rounds, participants were only given 400 ms to make each selection. If they exceeded the time limit, they earned no rewards and received negative feedback (sad face) with the value of the payoff they could have earned crossed out. We used the same feedback period duration of 400ms before the next trial automatically began and participants were shown the choice screen. Inputs during the feedback period had no effect. (c) Each payoff condition specifies a normal payoff distribution for each option, with the means and variances described numerically in Table 1. The reward distributions are designed to compare how differences in reward expectations and differences in uncertainty influence choices, where IGT refers to a payoff distribution inspired by the Iowa Gambling Task (see “Methods”). Dots and the Tukey boxplots describe 100 randomly drawn payoffs, while the half violin plots show the generative distribution, with the diamond indicating the mean.