Studies have shown that consumer preference can be changed by the choice context: joint evaluation vs. separate evaluation. Based on such studies, this study examines how consumer knowledge moderates the context effect on preference. In Study 1, participants are asked to evaluate DVD home theater systems with two attributes; one is easy to evaluate, the other difficult to evaluate. Joint-separate preference reversal happens because the hard-to-evaluate attribute in separate evaluation becomes easy to evaluate in joint evaluation. Results show that preference reversal happens in the lowknowledge group, but not in the high-knowledge group. In Study 2, only hard-to-evaluate information is given to participants. Participants are asked to evaluate attractive or unattractive computer monitors in different evaluation modes. Because the high-knowledge group has reference information for the product, they are hypothesized to evaluate products based on their own reference information in separate evaluation, but not in joint evaluation. Consequently, even the high-knowledge group shows preference reversal between two evaluation modes. Implications and future research directions are discussed.