Clinical observations have suggested that lithium may exert adverse effects on memory. The difficulty in achieving empirical consensus regarding this issue has reflected several methodological problems: diversity of research designs, heterogeneous samples, lack of control groups and the possible confounding of memory test scores by variables such as depression, other acute psychopatho-logies, organicity, treatment duration, and age. The diversity of memory tests in terms of the complexity and modality of the stimuli as well as the types of memory assessed (immediate, short- and long-term, logical, visuo-practic) has further complicated the comparison of results across studies. Furthermore, the administration of test batteries has been limited by patients’ fatigue and the severity of their illness, and by the time required to complete testing. Hence, the use of test norms may be restricted. Suggestions are made for the selection of appropriate memory tests, patients sampling and data analysis. The authors discuss the difficulties inherent in blind studies and in matched-group designs examining the effects of lithium on memory. Conclusions point to the advantages of prospective within-subject designs with repeated testing in which patients serve as their own controls.