A dominant account of perseverative errors in early development contends that such errors reflect a failure to inhibit a prepotent response. This study investigated whether perseveration might also arise from a failure to inhibit a prepotent representation. Children watched as a toy was hidden at an A location, waited during a delay, and then watched the experimenter find the toy. After six observation-only A trials, the toy was hidden at a B location, and children were allowed to search for the toy. Two- and 4-year-olds' responses on the B trials were significantly biased toward A even though they had never overtly responded to this location. Thus, perseverative biases in early development can arise as a result of prepotent representations, demonstrating that the prepotent-response account is incomplete. We discuss three alternative interpretations of these results, including the possibility that representational and response-based biases reflect the operation of a single, integrated behavioral system.