In cats under chloralose anaesthesia the cerebral ventricles were perfused from a cannula in the lateral ventricle. The effluent was collected either from the cisterna or from the aqueduct. When acetylcholine was added to the perfusion fluid and its destruction was prevented by the addition of an anticholinesterase, its recovery in the internal effluent was incomplete and irregular. This could be explained mainly by the devious route the perfusion fluid had to take in order to reach the cisterna and the unavoidable mixture with the cerebrospinal fluid of the subarachnoidal space. When collection was from the aqueduct 90% or more of the acetylcholine added to the perfusion fluid was recovered. In the presence of an anticholinesterase in the perfusion fluid, acetylcholine appeared in the effluent. Neostigmine was more effective than eserine or dyflos. The acetylcholine originated mainly from structures lining the lateral and third ventricle because the amounts in the effluent from the aqueduct were only a little less than those in the cisternal effluent. When injected intravenously, eserine was found to pass more readily into the perfusion fluid than neostigmine, probably because eserine passes the blood-brain barrier more readily than neostigmine. The method provides a quantitative approach for the study of the blood-brain barrier for pharmacologically active substances.