This paper tests a novel implication of the original version of prospect th
eory (Kahneman and Tversky, 1979): that choices may systematically violate
transitivity. Some have interpreted this implication as a weakness, viewing
it as an anomaly generated by the `editing phase' of prospect theory which
can be rendered redundant by an appropriate re-specification of the prefer
ence function. Although there is some existing evidence that transitivity f
ails descriptively, the particular form of non-transitivity implied by pros
pect theory is quite distinctive and hence presents an ideal opportunity to
expose that theory to test. An experiment is reported which reveals strong
evidence of the predicted intransitivity. It is argued that the existence
of this new form of non-transitive behaviour presents a fresh theoretical c
hallenge to those seeking descriptively adequate theories of choice behavio
ur, and a particular challenge to those who seek explanations within the co
nventional economic paradigm of utility maximisation.