The task-switching paradigm provides a powerful tool to measure the development of core cognitive control processes. In this study, we use the alternating runs task-switching paradigm to assess preparatory control processes involved in flexibly preparing for a predictable change in task and stimulus-driven control processes involved in controlling stimulus-level interference. We present three experiments that examine behavioral and event-related potential (ERP) measures of task-switching performance in middle childhood and young adulthood under low and high stimulus interference conditions. Experiment 1 confirms that our new child-friendly tasks produce similar behavioral and electrophysiological findings in young adults as those previously reported. Experiment 2 examines task switching with univalent stimuli across a range of preparation intervals in middle childhood. Experiment 3 compares task switching with bivalent stimuli across the same preparation intervals in children and young adults. Children produced a larger RT switch cost than adults with univalent stimuli and a short preparation interval. Both children and adults showed significant reduction in switch cost with increasing preparation interval, but in children this was caused by greater increase in RT for repeat than switch trials. Response-locked ERPs showed intact preparation for univalent, but less efficient preparation for bivalent stimulus conditions. Stimulus-locked ERPs confirmed that children showed greater stimulus-level interference for repeat trials, especially with bivalent stimuli. We conclude that children show greater stimulus-level interference especially for repeat trials under high interference conditions, suggesting weaker mental representation of the current task set.