Since the introduction of disporsable columns containing polymercoated charcoal, trials of charcoal haemoperfusion have been undertaken in various drug intoxications. We studied the efficiency of direct haemoperfusion (DHP) for the removal of paraquat (1, 1'-dimethyl 4, 4'-bipyridilium dichloride: GramoxonR) from circulating solution. In vitro, the concentration of paraquat 5, 10, and 100 ppm in 4 L dialysate were prepared. The flow rate through the column of 100 and 200 ml/min was obtained by the use of pump. In 100 ml/min flow rate, 1-7% of paraquat remained after three hours perfusion. In 200 ml/min flow rate, however, paraquat was not detected in one and a half hours after the start of perfusion (Limit of detection is less than 0.04 ppm). In three years, we experienced 18 cases of herbicidal paraquat intoxications. In 10 of these 18 patients, DHP with hydrocoating charcoal were performed after gastric lavage, forced diuresis, and administration of Fuller's earth. Six of these 10 patients died within 28 days after injestion of Gramoxon quantity varing between 5 to 80 mis. Four of these 10 survive without pulmonary complication and with mild to moderate kidney and/or liver dysfunction during their courses. These complications, however, disappeared complitely within one month. We conclude that charcoal haemoperfusion itself is effective instrument for removing paraquat from solution in vitro, and also for clinical use in some extend. Our experiences, however, indicate that patients who ingested large dose, who were not treated in initial phase, showed poor prognosis inspite of DHP therapy.