BACKGROUND Hypoperfusion around the seizure foci has been reported in interictal CBF studies of partial epilepsy. Hypoperfusion may reflect some interictal pathophysiology characteristic of epilepsy, but the mechanism remains to be elucidated. In order to elucidate the controlling system of rCBF in patients with CPS (complex partial seizure), we measured by positron emission tomography rCBF and CO2 vasoreactivity performing a hyperventilatory task, and made comparisons between the affected and contralateral sides of the temporal lobes in CPS patients and the temporal lobe of normal volunteers. We also measured PaCO2, pH, PaO2 and respiratory rate at rest to elucidate the CBF maintaining system. METHODS 15 volunteers and 12 medically controlled patients with CPS were examined. The affected side of the brain of the patients was determined by dominance of EEG paroxysm. METHODS For measuring rCBF, the Shimadzu HEADTOME-IV system (8mm FWHM) and H2(15)O (740-1480 [MBq]/scan) bolus injection method were employed. First a rest scan, then a hyperventilatory task scan was performed, at 20 min. intervals between each scan. During the rest scans, the subjects were required to close their eyes, which were covered through the transmission scan and all the other scans. They were monitered with EEG to insure that they were not asleep. Room lights were dimmed, room sounds were kept to a minimum and noises were damped with earplugs. During the hyperventilatory tasks, the subjects were asked to inspire deeply at a rate of 15/min. for 4.5 min. from 3 min. before the scan up to the end of the 1.5 min. scan, and monitored with a respiratory band around the abdomen. Arterial blood gas pressures (PaCO2, pH and PaO2) were also measured with a blood gas analyzer (Radio Meter ABL 330) at the beginning and at the end of the scans. [HCO3-] was calculated on the formula of Henderson-Hasselbalch. Vascular response to PaCO2 change: VrCO2 was defined as follows: VrCO2 = 100 x (CBFh-CBFr)/CBFr/(PaCO2h-PaCO2r) the subscripts h: hyperventilatory r: resting conditions. RESULTS Blood gasses, rCBF and corrected-rCBF at rest: In the patients, PaCO2 (mmHg) = 46.11 +/- 2.32, pH = 7.374 +/- 0.020, [HCO3-] (mEq/ml) = 25.98 +/- 1.15, CBF (ml/100g/min) = 45.51 +/- 9.33, corrected-CBF (ml/100g/min) = 40.06 +/- 7.27. In the volunteers, PaCO2 = 40. 67 +/- 2.37, pH = 7.400 +/- 0.021, [HCO3-] = 24.36 +/- 1.66, CBF = 48.76 +/- 8.54, corrected-CBF = 48.02 +/- 9.04. PaCO2 and [HCO3-] in the patients were significantly higher than those in the volunteers. pH and corrected-CBF were significantly lower than those in the volunteers. CBF showed no difference. VrCO2 in the temporal lobe: In the affected side of the patients, VrCO2 = 1.976 +/- 0.54. In the contralateral side of the patients, VrCO2 = 2.145 +/- 0.667. In the temporal lobe of the volunteers, VrCO2 = 2.557 +/- 0.898. In both sides of the temporal lobe of the patients, VrCO2 was significantly lower than those in the volunteers. CONCLUSIONS The effect of anticonvulsant drugs and the suppressive mechanism around the foci are supposed to reduce CBF of CPS patients. We suppose another mechanism, which is vasoconstriction. A significant difference was noted in CO2-vasoreactivity in the temporal lobe between patients and volunteers. Also no difference was noted in rCBF in the temporal lobe between the volunteers and the patients, when the arteries were maximally constricted with the HIV-task. These two results suggest that vasoconstrictive mechanism may be among the factors for the hypoperfusion seen in the temporal lobe of the patients. In patients, PaCO2, [H+], [HCO3-] were significantly higher compared with the volunteers, which suggests the mechanism of respiratory acidosis. Owing to this hypercapnia, global CBF in the patients is maintained in compensation for the hyperfusion and vasoconstriction seen around the foci. CO2 acts as an anticonvulsant and also reduces glucose metabolites which accumulate around the foci.