Preference reversal is a systematic change in the preference order between options when different response methods are used (e.g., choice vs. judgment). The present study focuses on procedures used to elicit preferences according to an evaluability hypothesis. Two experiments compared joint vs. separate evaluations and explicit vs. non-explicit joint evaluations. Subjects had to express preferences between high-variance gambles (HVGs) and low-variance gambles (LVGs) either by choosing one gamble to play in a lottery or by assigning gambles minimum selling prices. We show that HVGs are preferred in both choice and pricing conditions when gambles are evaluated separately, and LVGs are preferred in both choice and selling conditions when gambles are evaluated in pairs: i.e., when the evaluation mode is held constant, classic preference reversal disappears. These results support the evaluability hypothesis, and suggest that preferences depend on whether subjects are allowed to compare the options they are asked to choose from or judge, independently of the nature of the scale (i.e., attractiveness vs. minimum selling price) they are required to adopt. Copyright (C) 2004 John Wiley Sons, Ltd.