Six dogs were administered 3rd ventricular infusions of artificial CSF (292 mosmol/l) alone or artificial CSF to which either NaCl, sucrose, glucose or urea was added to yield a final osmolar concentration of 500 mosmol/l. The volume of H2O drunk during 45 min of infusion was measured and blood was sampled for determination of plasma vasopressin concentration at 15 min intervals. Artificial CSF made hypertonic by addition of NaCl or sucrose stimulated water intakes of 9.0 .+-. 3.2 ml/kg (mean .+-. SE) and 7.3 .+-. 3.7 ml/kg, respectively. There was no statistical difference in the amounts drunk and the latencies. In contrast, artificial CSF containing glucose, urea or artificial CSF alone were without effect. Plasma vasopressin concentration increased significantly in response to intraventricular NaCl and sucrose but was not affected by glucose, urea or artificial CSF alone. These data apparently are compatible with an osmoreceptor mechanism mediating drinking and vasopressin release.