Healthy control college students. 23 subjects completed the 3-armed bandit task with oscillating probabilities. For example, the ‘blue’ stim would slowly move from 20% reinforcing to 90% then back to 20 over many trials. The other ‘red’ and ‘green’ stims would move similarly, but in different phase. See Fig 1 of the paper. This makes the task great for investigating reward processing & reward prediction error in the service of novel task set generation.