A generalization of expected utility theory. The alternative outcomes are ranked from lowest to highest, and the probabilities of the outcomes are transformed in a way that ensures unlikely extreme outcomes are over-weighted. If (x1,…, xn) are the ranked outcomes and (p1,…, pn) the probabilities with which they occur, then the rank dependent expected utility function is where wi (p1, …, pn), i = 1,…, n, are the weighting functions. The motivation for this construction is to obtain a pay-off function that can explain the choices observed in the Allais paradox and other similar anomalies.