To define conditions under which saliva lithium carbonate concentrations can be used as a guide in lithium therapy, serum (Cp) and saliva (Cs) concentrations were determined simultaneously under different circumstances in 12 patients receiving lithium, and the effect of seven variables for example, (sex, saliva flow rate) on the Cp:Cs ratio was examined. The average ratio was 0.57 if saliva was collected in the morning before breakfast and 0.45 otherwise. The Cp:Cs ratio was found to vary much more between individuals than within an individual. We propose a method that minimizes the effect of the interindividual variation on the error in the prediction of Cp from Cs by using one or two measured Cp:Cs ratios to adjust the individual ratio. Using this technique the Cs may be useful as a predictor of the Cp in monitoring lithium therapy.