Research in animal neuropsychology is providing an exciting new generation of behavioral tests for mice that promise to overcome many of the limitations of current high-throughput testing, and provide direct animal homologues of clinically important measures in human research. Set shifting tasks are some of the best understood and widely used human neuropsychological tasks, with clinical relevance to traumatic brain injury, schizophrenia, autism, obsessive compulsive disorder, trichotillomania, and many other disorders. Here we report the first successful modification of a human set shifting neuropsychological task, the Intra-Dimensional Extra-Dimensional (IDED) task, for use with mice. We presented mice with a series of compound discrimination and reversal tasks where one stimulus dimension consistently cued reward. Task performance improved with a new set of compound stimuli, as did reversal performance--indicating the formation of a cognitive-attentional set. We then overtrained a subset of the mice, and presented control and overtrained mice with a new compound discrimination where a novel stimulus dimension cued reward. As is the case in human control subjects, control mice persisted in responding to the now-incorrect stimulus dimension, performing poorly on this extra-dimensional shift compared with the previous intra-dimensional shift, thereby validating the task as a measure of set shifting. Furthermore, overtrained mice were impaired on this extra-dimensional shift compared with controls, further validating the task. The advantages and disadvantages of the IDED task compared to high-throughput approaches are discussed.