The Estes and Straughan (1954) two choice, non-contingent probability learning situation involves the presentation of a signal (e.g., light, buzzer) to which S responds by predicting one of two mutually exclusive and exhaustive events. The predicion of the more frequent event on trial n, E1,n , is designated A1,n, and the prediction of the less frequent event on trial n, E2,n, is designated A2,n. This learning situation is described as "non-contingent" becasue the occurrence of either event on trial n is independent of the response, i.e., P(Ei,n) = P(Ei,n|Aj,n), i,j = 1,2. The events E1 and E2 occur with probabilities π and 1-π, respectively. One purpose of the present research is to consider several models which predict behavior in the situation just described. These models will be examined for their ability to account for the choice behavior of Ss who have received extended training in one of two reinforcement schedules (levels of π) and one of two payoff conditions. In particular, these models will be evaluated for their ability to describe the mean and variability of response probabilities, as well as sevearl response probabilities conditional on previous sequences of events and responses.