Two methods for predicting steady-state serum lithium level were compared prospectively in in-patients suffering from affective disorder. A single-point prospective administration model that required a single 24-hour serum lithium level, following a test dose produced statistically similar predictions of the observed steady-state lithium levels as did a model that required 12- and 36-hour levels. However, the latter two-point method produced significantly more accurate predictions from clinical interpretation. Although the two-point approach is preferable, the single-point method is clinically acceptable if its limitations of accuracy are taken into consideration.