The standard theory of choice—based on value maximization—associates with each option a real value such that, given an offered set, the decision maker chooses the option with the highest value. Despite its simplicity and intuitive appeal, there is a growing body of data that is inconsistent with this theory. In particular, the relative attractiveness of x compared to y often depends on the presence or absence of a third option z, and the “market share” of an option can actually be increased by enlarging the offered set. We review recent empirical findings that are inconsistent with value maximization, and present a context-dependent model that expresses the value of each option as an additive combination of two components: a contingent weighting process that captures the effect of the background context, and a binary comparison process that describes the effect of the local context. The model accounts for observed violations of the standard theory and provides a framework for analyzing context-dependent preferences.