On the difficulties underlying Bayesian reasoning: A comment on Gigerenzer and Hoffrage

G. Gigerenzer and U. Hoffrage (1995) suggested that formulating problems in frequentistic terms would facilitate the practice of Bayesian inference. The authors of the present article propose that there are two different factors at work in this context: the use of frequency or probability statements and the use of joint or conditional sampling information. An experiment is reported demonstrating that, even with the use of frequency statements, Bayesian reasoning is more difficult with conditional than with joint information. Broader implications of the proposed distinction are briefly discussed.