Binary choice and the supply of memory

In a binary choice experiment, the subject predicts on a series of trials whether a particular random event will or will not occur. The results of such experiments show systematic departures from expected value maximization, but to a degree that decreases as the reward for successful prediction increases. This paper proposes that the mechanism underlying these results involves the supply of memory. Subjects are influenced by the outcomes of recent trials, but tend to respond to the evidence in a larger number of trials as reward (and hence attention) are increased. Broader conceptual issues surrounding the notion of rational choice are also discussed.