EXPERIMENT CODE for the six experiments in the manuscript.
Experiment 1 - Tested all six rule types at three incentive levels (low, med, high). 16 learning trials. The only difference between the two versions is the addition of a "testscore" term saved to the database in v0.1 so that it does not have to be computed posthoc as nCorrect/nTestQuestions. So they are essentially the same experiment, one just saves an extra term for convenience (but there is no information that isn't also included in v0.0). Better performance on test phase increases PROBABILITY of fixed $10 bonus.
Experiment 2 - Double the learning trials so that there are 32 learning trials now. Test two rule types (II, IV) and two incentive levels (low, high). Design same as Exp 1.
Experiment 3 - Incentive manipulation affects MAGNITUDE not PROBABILITY of reward. Subjects are assigned to a low or high incentive condition - so they can earn either BLUE bonus tickets or GOLD bonus tickets. More answers correct above chance on the test phase equals more bonus tickets earned, which means a larger magnitude bonus.
Experiment 4 - Within subjects experiment. Each subjects performs 4 blocks of the task. 2x2 within subjects design crossing two incentive levels (low, high) and two rule types (II,IV). Design same as Exp 3.
Experiment 5 - Similar to Exp 4. Rule varied BETWEEN subjects and incentive varied WITHIN subjects. Preregistered. Rules I, II, IV. Incentives: low, high.
Experiment 6 - Category recognition rather than category learning task. This version of the task required no learning or rule discovery; instead, participants implement a given rule to count how many members of a category are displayed.