We show that information overload contributes to confirmation bias. In an experiment, we vary the difficulty of information processing as subjects receive a sequence of signals of an unknown state. In the treatment condition, the preceding signal disappears as the next signal appears. In the control condition, the preceding signal remains visible. We find stronger confirmation bias among subjects in the treatment condition. Our results provide empirical support for models that emphasize the role of limited information processing in confirmation bias (Wilson (2014), Leung (2018), Jehiel and Steiner (2018)).