Non-transitive Choice: Event-splitting effectsor framing effects?

A significant body of experimental evidence documents systematic violations of the transitivityaxiom of expected utility theory in a manner consistent with regret theory. In a recent paper Starmerand Sugden (in press) test a variety of alternative explanations of this evidence, including eventsplittingeffects, and conclude that anticipated regret provides the best explanation, but is sensitive tominor changes in problem representation. Humphrey (1998) shows that the apparent sensitivity ofregret-cycling to problem representation in Starmer and Sugden’s data is best explained by the eventsplittingargument they rule-out. This paper reports an experiment which discriminates betweenregret-driven and event-splitting-driven cycles in a manner which controls for the sources ofambiguity in Starmer and Sugden’s tests..