Revealed preference theory provides analytic tools to measure the inconsistency embedded in a decision-maker’s choices, thus their deviation from a rational benchmark. We apply this theory together with neuroimaging techniques to trace the neural mechanisms that generate inconsistent behavior. We develop a novel index that measures the severity of inconsistency for each choice, and directly trace its neural correlates using fMRI. In addition, we attemp to develop a computational model of choice suggesting that variability in the computation of value can explain our findings and is therefore one possible source of inconsistent choice behavior.