Animals with sham-operations or medial prefrontal cortex lesions were trained in a task which required memory for short or long lists of items (spatial locations). On any one trial a rat is presented with 2, 4, 6, 8 or 10 items (spatial locations) on a 12-arm radial maze followed 15 min later by a win-shift test comprising a choice between a place previously visited and a novel place. Sham-operated animals showed an increase in errors as a function of set size (2 to 8 items) followed by a decrease in errors with a set size of 10 items suggesting the use of both retrospective and prospective memory codes. In contrast, animals with medial prefrontal cortex lesions made most of their errors for the longest list length reflecting an inability to shift from a retrospective to prospective memory code. The results are interpreted as support for a medial prefrontal cortex role in mediating a prospective code perhaps via knowledge systems based on temporal information.