Sodium valproate is often used with phenytoin when epilepsy cannot be controlled by a single drug. Sodium valproate depresses phenytoin protein binding and so invalidates plasma phenytoin monitoring as a means of determining precise phenytoin dosage requirements. Plasma and saliva phenytoin and plasma valproate concentrations were measured in 42 patients with epilepsy receiving both drugs. Phenytoin protein binding was also measured by ultrafiltration in 19 of these patients and 19 patients taking phenytoin alone. Saliva phenytoin concentration bore the same close correlation to unbound (therapeutically active) phenytoin in patients receiving both drugs as it did in patients receiving phenytoin alone, whereas plasma total phenytoin did not. The same therapeutic range for saliva phenytoin (4-9 mumol/1; 1-2 microgram/ml) was therefore valid in both groups. The depression of phenytoin binding was directly related to the plasma concentration of valproate both in random samples taken from the 42 patients and in samples taken throughout the day in two of these patients. This was confirmed in vitro. Even when the concentration of valproate is known the degree of binding cannot be predicted. Saliva rather than plasma monitoring of phenytoin treatment is therefore valuable in the presence of valproate and with reduced phenytoin binding from any cause.