Two groups of rats were trained in 2-lever operant tasks that required them to follow either a Win-shift/Lose-stay rule, or else a Win-stay/Lose-shift rule. The rats in each testing condition attained a similar level of performance. Once the tasks had been learnt, half of the rats in each testing condition were given fornix lesions. These lesions produced a clear performance impairment at all inter-response retention intervals tested in those rats trained on the Win-shift/Lose-stay task, but impaired performance in rats trained on the Win-stay/Lose-shift condition only when the inter-response interval reached 10 s. The implications of these results for the working memory theory and for the temporal discontiguity theory are discussed.