The correlation between serum and saliva lithium levels in patients undergoing lithium prophylaxis therapy is discussed. Although a high correlation can be obtained, there is an unacceptably high individual variation of paired results. Two strategies were used in an attempt to improve the reliability of saliva as a predictor of serum levels. A naturally occurring marker in saliva and serum was used. This showed high correlation but still did not prevent large sporadic differences. Serially paired results in individual patients also showed occasional excessive variation. Although saliva lithium has been used for monitoring when serum assessment is impracticable, it is considered that it is, as yet, an unreliable technique.