In our daily life, we continuously monitor others' behaviors and interpret them in terms of goals, intentions, and reasons. Despite their central importance for predicting and interpreting each other's actions, the functional mechanisms and neural circuits involved in action understanding remain highly controversial. Two alternative accounts have been advanced. Simulation theory assumes that we understand actions by simulating the observed behavior through a direct matching process that activates the mirror-neuron circuit. The alternative interpretive account assumes that action understanding is based on specialized inferential processes activating brain areas with no mirror properties. Although both approaches recognize the central role of contextual information in specifying action intentions, their respective accounts of this process differ in significant respects. Here, we investigated the role of context in action understanding by using functional brain imaging while participants observed an unusual action in implausible versus plausible contexts. We show that brain areas that are part of a network involved in inferential interpretive processes of rationalization and mentalization but that lack mirror properties are more active when the action occurs in an implausible context. However, no differential activation was found in the mirror network. Our findings support the assumption that action understanding in novel situations is primarily mediated by an inferential interpretive system rather than the mirror system.