Using quantitative autoradiography, we investigated the effect of intracarotid infusions of hyperosmolar mannitol solutions on capillary permeability and blood flow. Capillary permeability, expressed in terms of a blood-to-tissue transfer constant (K), was determined in two rat brain tumor models by measuring the entry of 14C-alpha aminoisobutyric acid into brian tumor, into brain tissue adjacent to tumor, and into cortex. Cerebral blood flow was determined by measuring the uptake of 14C-iodoantipyrine in one rat brain tumor model. Blood flow was examined in the same regions as K, as well as in the corpus callosum. Before mannitol administration, K values in both Walker 256 (W256) carcinosarcoma and C6 gliomas were much higher than those in cortex. C6 gliomas were about three times more permeable than were W256 tumors. There was a direct correlation between tumor size and increased capillary permeability. Mannitol at a concentration of 1.37 M did not increase the K values for either tumor or adjacent tissue. At 1.6 M, mannitol increased the K values for both tumors (1.7-fold in C6 glioma and 13-fold in W256) as well as for adjacent tissue. At both concentrations, mannitol markedly increased cortical K values in all groups: by 48-to 72-fold at 1.37 M and by 90- to 105-fold at 1.6 M. The net effect of the mannitol was to reverse the tumor-to-cortex permeability relationship. Cortical blood flow increased modestly after intracarotid mannitol administration on both sides of the brain. These data provide little justification for using intracarotid mannitol during chemotherapy of human brain tumors.