How does a person assess the outcome of a choice? Kőszegi and Rabin (2006) proposed a reference-dependent utility theory in which the overall utility of an outcome is composed of two components: an intrinsic consumption utility attached to the choice itself and a gain-loss utility which is reference dependent. Using functional magnetic resonance imaging, we investigated the way people weigh consumption utility and gain-loss utility to assess an outcome. Trials in Session 1 were to monitor neural responses to multiple types of money-food reward bundles. Trials in Session 2 comprised an initial expectancy phase, when rewards in each bundle were not fully disclosed and needed to be predicted, and a subsequent outcome phase, when actual amounts of rewards were revealed. We found that mPFC and OFC track expected consumption utility for money-food rewards bundles. Generally, reference-dependent experienced utility was also observed in the same regions. mPFC also encodes gain or loss signals that were computed during experience of actual rewards.