Where Does Subjective Expected Utility Fail Descriptively ?

Subjective expected utility (SEU) rests on and implies four tenets of rational preferences; transitivity, monotonicity of consequences, independence of a common consequence, and accounting equivalences. Empirical evidence against transitivity and monotonicity is reevaluated and the opposite conclusion drawn using more recent data. The more complex accounting equivalences are descriptively implausible. The three simplestidempotence, complementarity, and event commutativity-seem to be the only ones that may be descriptive. These, coupled with the postulate of an interval scale representation, result in a rank-dependent, weighted linear generalization of SEU. Further generalizations to nonbinary mixtures and to rankand sign-dependent representations are also described. Problems in testing these theories are discussed.