Probabilistic Reversal Learning (PRL)

Flexibility under probabilistic feedback, recovery after contingency shifts, and perseverative errors.

Across blocks, the optimal side switches; you need to adapt to keep earning rewards. Note that this is a simplified version: the switch happens on a fixed schedule (every 30 trials), whereas in the classic paradigm the reversal is performance-dependent — it occurs once you have clearly learned the current contingency.

The rewarded option is not rewarded on every trial, so single outcomes are ambiguous evidence — you must integrate across trials instead of switching after every loss.

Track the relative reward rates of the two options and shift only when accumulated evidence favors the other one; both extreme stickiness and constant flip-flopping cost points.

It cleanly separates learning rate from decision noise, making it a standard probe of reinforcement learning studied in relation to dopamine and compulsive behavior.