In rodents, working memory is a representation of an object, stimulus, or spatial location that is typically used within a testing session, but not between sessions, to guide behaviour. In this review we consider a number of the tasks used to assess this type of memory in the rodent, and highlight some of their limitations. Although the concept of working memory as applied to rodents has its origin in the experiments of David Olton and Werner Honig in the 1970s, many earlier experiments assessed the same type of memory under the guise of delayed reaction or alternation paradigms. We revisit these early tasks, and also consider the nature of working memory used on maze tasks, operant box based tasks, and non-spatial delayed non-matching to sample paradigms.