Within- and between-subject variability in serum and salivary lithium concentrations in nine psychiatric inpatients on stable drug regimens undergoing therapy has been assessed using criteria for determining biologic variation. This allows separation of analytic from other measures of variance. There were marked differences in inter- and intra-subject variance for serum and salivary lithium concentrations for serum/salivary ratios. Moreover, these variances were greater for salivary lithium than for serum concentrations. The results have been used to assess analytical performance, the usefulness of the therapeutic range, and the reference change interval. Despite greater variance for salivary concentrations, predicted serum concentrations from predetermined serum/saliva ratios were in good agreement with actual concentrations in most subjects.