Male hooded rats were preoperatively trained to perform an operant delayed alternation (DA) task. The rats were tested, for 10 days each, on DA at 5-, 10-, 20-, and 30-s delays. The rats were then randomly assigned to serve as control subjects or receive a medial septal lesion. Following surgical recovery the rats were retested on DA in the exact manner as they were tested preoperatively. Prior to surgery the groups did not differ, but all rats improved over days of testing, and performed better at short delays than at long delays. In contrast, group differences were obvious following surgery, with the septal-lesioned subjects showing a clear impairment. The control subjects ranged between 85% correct (end of 5-s delay) to 65% correct (end of 30-s delay), while the septal-lesioned rats ranged between 72% correct (end of 5-s delay) to 50% correct (end of 30-s delay). The results are discussed in terms of a working memory impairment produced by the septal lesions.