Using tritiated arginine-8-vasopressin [3H]AVP, vasopressin-specific binding sites were detected on human platelet membranes. One class of high-affinity binding sites was characterized with an equilibrium dissociation constant of 1.01 +/- 0.06 nM and a maximal binding capacity of 100 +/- 10 fmol/mg of protein (n = 12). Highly significant correlations were found between the relative agonistic (r = 0.87, P = 0.002) or antagonistic (r = 0.99, P = 0.007) vasopressor activities of a series of 13 AVP structural analogues and their relative abilities to inhibit [3H]AVP binding to platelet receptors whereas no such relationship existed when antidiuretic activities were considered (r = 0.28, P = 0.47). AVP did not stimulate cyclic AMP production of human platelets; on the contrary, high AVP concentrations (10(-6) M) inhibited cyclic AMP production measured in basal and prostaglandin E1-stimulated conditions. AVP caused intact platelet aggregation with a half-maximal aggregation (EC50) of 28 +/- 2 nM. This effect was more potently reversed by the specific vascular antagonist d(CH2)5Tyr(Me)AVP (pA2 = 8.10 +/- 0.23) than by the specific renal antagonist d(CH2)5IleuAlaAVP (pA2 = 6.67 +/- 0.12). The pA2 values of these two antagonists in platelets are in close agreement with the pKi values obtained in competition experiments (respectively 8.59 and 6.93) and with pA2 values reported in the literature for their in vivo antivasopressor activity (respectively 8.62 and 6.03). The observation that human platelets bear AVP receptors belonging to the vascular class suggests that platelet receptors can be used to further explore the role of vasopressin in cardiovascular homeostasis.