The conjunction fallacy? A reply to Bar-Hillel

In her commentary on our article, Bar-Hillel (1991) argues that our normative interpretation of the Lindaproblem is inaccurate and that the representativeness heuristic can account for all of the relevant data. I argue that the normative interpretation of the Linda problem remains slippery and that representativeness fails to account for much of the data. Furthermore, substantial evidence supports our suggestion that subjects use different underlying models in estimating likelihoods in different con problems.