Failures of the reduction principle in an Ellsberg-type problem

Segal (1987) suggested that the Ellsberg paradox might be explained in terms of individuals mentally representing the decision problem as a two-stage lottery which they evaluated according to a non-expected utility model. This paper describes an experiment involving an explicitly two-stage analogue to an Ellsberg-type problem. This design substantially reduces the frequency of classic Ellsberg behaviour, but reveals other systematic violations of conventional theory. The paper discusses the particular patterns of choice and raises the more general problem of modelling individual decisions when the reduction principle does not hold.