I recently found an interesting study that bears on the doctor example. Christensen and Bushyhead (1981) find that when asked to make clinical judgments, doctors usually take base rates into account quite accurately, even though when they are asked to explicitly do statistical problems involving base rates they usually get them wrong.