Preference reversal has been frequent in tests with hypothetical, or small real, payoffs concerning lotteries as well as claims redeemable at different future dates. PR is tested here for the latter case with nontrivial payment levels and subjects likely to deal with decisions of this type. It is found (i) that replacing hypothetical by real-payment tests reduces PR rates from 62 % to I 5 % for subjects 'predicted' to reverse preferences, and (2) that the real overall PR rate (before correcting for response errors) is I9%, much lower than in earlier studies.