Cognitive impairment is underrecognized among patients with bipolar disorder and may represent not only effects of the illness but also adverse effects of its treatments. Among these, lithium is the best-studied mood stabilizer. As its cognitive effects are mixed and not well-known, we assessed reported effects of lithium on cognitive performance. MEDLINE, PsycINFO, and EMBASE databases (1950 to December 2008) were queried with the keywords lithium, cognit*, neurocognit*, neuropsych*, psycholog*, attention, concentration, processing speed, memory, executive, and learning. Database searches were supplemented with bibliographic cross-referencing by hand. The literature search was conducted independently by 2 authors (A.P.W. and T.S.W.) during August and September 2008, and questions about appropriate inclusion or exclusion were resolved between them by consensus. Of 586 reports initially identified as being of potential interest, 12, involving 539 subjects, met our inclusion criteria: (1) cognitive performance compared between subjects taking lithium and comparable subjects not taking lithium; comparability was assured by: (2) patients with the same affective disorder diagnoses in euthymic or remitted status or healthy volunteers; (3) groups of similar age and sex; (4) similar intelligence, education, or occupation; (5) similar distribution of other concurrent psychotropic drugs; and (6) cognitive abilities (outcomes) assessed with performance-based measures. Standardized mean-difference effect size (ES), corrected for small-sample bias (Hedges' g), was computed for cognitive tasks in each study. ES estimates were transformed so that positive values indicate poorer performance by lithium-treated subjects. Infrequently, when means and standard deviations were not provided, ES was estimated from reported values of t, F, or z tests. For analysis, similar neurocognitive tests were grouped a priori based on the cognitive domains they aimed to assess. We identified 12 studies involving 276 lithium-treated and 263 similar or the same subjects, lithium-free. Lithium was taken for a mean duration of 3.9 years by affective disorder patients and 2.5 weeks by healthy volunteers, yielding a mean daily trough serum concentration of 0.80 mEq/L. Overall, lithium treatment was associated with small but significant impairment in immediate verbal learning and memory (ES = 0.24; 95% CI, 0.05-0.43) and creativity (ES = 0.33; 95% CI, 0.02-0.64), whereas delayed verbal memory, visual memory, attention, executive function, processing speed, and psychomotor performance were not significantly affected. Selectively, among the 326 affective-disorder patients, in addition to these overall impairments, long-term lithium treatment also was associated with even greater impairment in psychomotor performance (ES = 0.62; 95% CI, 0.27-0.97), with no evidence of cognitive improvements. Lithium treatment appears to have only few and minor negative effects on cognition.