UC Irvine

A noncontingent success (NCS) reinforcement schedule for binary prediction is one in which the subject has the same probability (δ) of being correct regardless of which response he makes. These schedules may be contrasted with the more commonly studied noncontingent event (NCE) schedules in which the event probabilities are not contingent on the subject’s choice of response, but the probability of his being correct is. The NCS schedules are examined here in connection with the problem of deciding experimentally between the linear and N element models for probability learning. It is shown that for mathematical reasons there is essentially no possibility of making such a decision on the basis of experiments with NCE schedules. Predictions for NCS schedules are then derived from the two models, and an experiment with two such schedules (δ = .8 and δ = 1.0) is reported. The results unequivocally support the N element model over the linear model, but under δ = 1 contingencies subjects generate patterned response sequences-"superstitious solutions"-that cannot be explained by any of the current models. © 1969.