Abstract In order to understand how the brain prepares for and executes a switch in task demand, we measured reaction time (RT), accuracy, and event-related brain potentials associated with performance in single and mixed-task blocks using a cued design. Our results show that trials which repeat in a mixed-task block (repeat trials) were more demanding than trials which repeated in a single-task block, as reflected by the presence of a RT mixing cost and by the presence of a smaller target-locked positivity (P3b) on repeat trials. Within a mixed-task block, repeat and switch trials also differed, where repeat trials showed evidence of greater preparation (larger cue-locked negativity), more efficient target processing (larger target-locked P3b), and shorter RTs. In addition, the cue-locked negativity difference remained despite equating repeat and switch trials on RT, suggesting that this negativity difference is specific to the switching process. Our results are discussed in light of existing models of task switching.