Researchers attempting to understand how citizens process political information have advanced motivated reasoning to explain the joint role of affect and cognition. The prominence of affect suggests that all social information processing is affectively charged and prone to biases. This article makes use of a unique data set collected using a dynamic information board experiment to test important effects of motivated reasoning. In particular, affective biases should cause citizens to take longer processing information incongruent with their existing affect and such biases should also direct search for new information about candidates. Somewhat perversely, motivated reasoners may actually increase their support of a positively evaluated candidate upon learning new negatively evaluated information. Findings are reported that support all of these expectations. Additional analysis shows that these affective biases may easily lead to lower quality decision making, leading to a direct challenge to the notion of voters as rational Bayesian updaters.