We have studied the reproducibility of the thirst and arginine vasopressin (AVP) responses to osmotic and hypoglycemic stimulation in healthy volunteers undergoing repeat hypertonic (855 mmol/l) saline infusion and insulin tolerance tests (ITTs). Hypertonic saline infusion caused similar mean rises in plasma osmolality, AVP, and thirst on each occasion. Linear-regression analysis defined close relationships between the slopes (r = +0.72, P less than 0.05) and the abscissal intercepts (r = +0.89, P less than 0.001) of the regression lines relating plasma osmolality (Posmol) and plasma AVP (PAVP), and the group intraindividual component of the variance for the slopes and intercepts was 7 and 0.6%, respectively. There were close correlations between the slopes (r = +0.79, P less than 0.02) and the intercepts (r = +0.84, P less than 0.01) of the regression lines relating Posmol and thirst, and group intraindividual component of the variance was 14 and 0.7%, respectively. Hypertonic saline infusion was infused on four occasions in four subjects, and the results showed that the linear regression lines relating PAVP and Posmol and thirst and Posmol were reproducible within an individual. There were similar falls in blood glucose and elevations in PAVP in both ITTs. No relationship was defined between the fall in blood glucose and either the rise in PAVP or the area under the AVP curve (AUC). The intraindividual component of the variance for the rise in AVP and the AUC was 77 and 22.5%, respectively. The AVP and thirst responses to osmotic stimulation are highly reproducible, but there is considerable intraindividual variation in the AVP response to hypoglycemia.