Experimental investigations of non-expected utility have primarily concentrated on decision under risk Uprobability triangles"). The literature suggests, however, that ambiguity is one of the main causes for deviations from expected utility (EUt. This article investigates the descriptive performance of rank-dependent utility (RDU) in the context of choice under ambiguity. We use the axiomatic difference between RDU and EU to critically test RDU against EU. Surprisingly, the RDU model does not provide any descriptive improvement over EU. Our data suggest other "framing" factors that do provide descriptive improvements over EU.In the past decades, students of individual choice have demonstrated that preference behavior deviates from Expected Utility theory (EU). In response, various generalizations and alternatives to EU have been put forward. Axioms that are thought to be descriptively invalid have been weakened to accommodate the observed violations. For most theorists, the independence axiom is the major culprit, exemplified by the Allais and Ellsberg paradoxes. In the early 1980s, several new models have been proposed that weaken the independence axiom, thus generalizing EU. Two major kinds of transitive generalizations have emerged: betweenness models and rank-dependent models.Recently, empirical research is cumulating that tests whether the new proposals are able to explain the experimental facts found thus far (Camerer and Ho, 1994). Most of this research can be summarized using so-called probability triangles. These triangles employ three outcomes that are kept constant. Gambles are constructed by varying the probabilities assigned to these outcomes. Most of the theories considered can be tested by observing preference behavior on selected pairs of gambles from these triangles. The overall picture emerging from this research is that none of the generalized expected utility theories can explain all the systematic violations that have been discovered. The betweenness condition, for example, is found to be systematically violated (Camerer and Ho, 1994). The rank-dependent models (RDU), especially the new version of prospect theory, Cumulative Prospect theory, are at the moment probably the leading contender for best descriptive theory. However, some remarks must be made.