Human therapeutic valproic acid (VPA) levels could be maintained in the mouse for a period of 1 week by constant rate application via subcutaneously implanted osmotic minipumps. Also, the concentrations of VPA metabolites observed in mouse plasma were similar to those seen in human plasma. The drug application could be prolonged by replacing exhausted pumps with freshly-filled devices. Removal of the implanted pumps and measurement of the decay of the drug levels revealed that the half-life of the main plasma metabolite 2-en(2-propyl-2-pentenoic acid) exceeded that of VPA. This result was confirmed by constant-rate application of this metabolite; the plasma clearance of 2-en (as calculated from the steady-state levels observed) was found to be lower than that of VPA. Brain levels of VPA and 2-en during steady-state were 3-10 per cent of corresponding plasma levels. The blood-brain kinetics of 2-en following administration of VPA were similar to those observed following application of 2-en itself. VPA was cleared faster from the brain than from the plasma, while 2-en was more persistent in the brain than in the plasma. Our results indicate that controlled, constant-rate application of drugs such as VPA, via implantable osmotic minipumps, may be a valuable procedure for a number of pharmacological and toxicological studies, particularly where persistent drug levels must be maintained for extended time periods.