The authors evaluated the extent to which the Subdural Pharmacotherapy Device (SPD), chronically implanted over the frontal cortex to perform periodic, localized muscimol-delivery/CSF removal cycles, affects overall behavior, motor performance, electroencephalography (EEG) activity, and blood and CSF neurochemistry in macaque monkeys. Two monkeys were used to adjust methodology and 4 monkeys were subjected to comprehensive testing. Prior to surgery, the animals' behavior in a large test chamber was monitored, and the motor skills required to remove food pellets from food ports located on the walls of the chamber were determined. The monkeys underwent implantation of the subdural and extracranial SPD units. The subdural unit, a silicone strip integrating EEG electrodes and fluid-exchange ports, was positioned over the right frontal cortex. The control unit included a battery-powered, microprocessor-regulated dual minipump and radiofrequency module secured to the cranium. After implantation, the SPD automatically performed periodic saline or muscimol (1.0 mM) deliveries at 12-hour intervals, alternating with local CSF removals at 6-hour intervals. The antiepileptic efficacy of this muscimol concentration was verified by demonstrating its ability to prevent focal acetylcholine-induced seizures. During SPD treatment, the monkeys' behavior and motor performance were again monitored, and the power spectrum of their radiofrequency-transmitted EEG recordings was analyzed. Serum and CSF muscimol levels were measured with high-performance liquid chromatography electrochemical detection, and CSF protein levels were measured with turbidimetry. The SPD was well tolerated in all monkeys for up to 11 months. The behavioral study revealed that during both saline and muscimol SPD treatment, the monkeys could achieve the maximum motor performance of 40 food-pellet removals per session, as before surgery. The EEG study showed that local EEG power spectra were not affected by muscimol treatment with SPD. The neurochemical study demonstrated that the administration of 1.0 mM muscimol into the neocortical subarachnoid space led to no detectable levels of this compound in the blood and cisternal CSF, as measured 1-125 minutes after delivery. Total protein levels were within the normal range in the cisternal CSF, but protein levels in the cortical-site CSF were significantly higher than normal: 361 ± 81.6 mg/dl. Abrupt discontinuation of 3-month, periodic, subdural muscimol treatments induced withdrawal seizures, which could be completely prevented by gradually tapering off the subdural muscimol concentration from 1.0 mM to 0.12-0.03 mM over a period of 2 weeks. The monkeys' general health and weight were maintained. Infection occurred only in one monkey 9 months after surgery. Long-term, periodic, transmeningeal muscimol delivery with the SPD is essentially a safe procedure. If further improved and successfully adapted for use in humans, the SPD can be used for the treatment of intractable focal neocortical epilepsy affecting approximately 150,000 patients in the US.