Dual–process models of attitudes highlight the fact that evaluative processes are complex and multifaceted. Nevertheless, many of these models typically neglect important interactions among processes that can contribute to an evaluation. In this article, we propose a multilevel model informed by neuroscience in which current evaluations are constructed from relatively stable attitude representations through the iterative reprocessing of information. Whereas initial iterations provide relatively quick and dirty evaluations, additional iterations accompanied by reflective processes yield more nuanced evaluations and allow for phenomena such as ambivalence. Importantly, this model predicts that the processes underlying relatively automatic evaluations continue to be engaged across multiple iterations, and that they influence and are influenced by more reflective processes. We describe the Iterative Reprocessing Model at the computational, algorithmic, and implementational levels of analysis (Marr, 1982) to more fully characterize its premises and predictions. Recent advances in neuroscientific methods have provided researchers with an unprecedented opportunity to examine the neural correlates of