A water-escape version of the radial-arm maze was used to assess rat spatial working memory performance. Intact females and ovariectomized females receiving a physiologically low dose, physiologically moderate dose or no estradiol replacement were studied. Subjects were given seven trials a day for 12 days. Females receiving moderate dose estradiol made fewer errors than the other three groups during the latter portion of testing. As trials progressed within a session, the elements of information to be remembered increased. Assessment of individual trials revealed that when the demand on an animal's working memory system was limited to one to four elements of information, the three groups with estrogen (including intact females) maintained successful performance, whereas the ovariectomized females made more errors. However, when the demand on an animal's working memory system was increased to six elements of information, only the moderate dose estradiol females maintained successful performance. These data suggest that, although moderate levels of estradiol replacement are the most beneficial for working memory function, even low-dose estradiol replacement can act to protect working memory systems from the decline seen with the removal of ovarian hormones.