Sixteen patients each received infusions of 1 g of mannitol per kg over 5 to 10 minutes, and serial determinations of intracranial pressure (ICP), systemic arterial blood pressure (SABP), central venous pressure, cerebral perfusion pressure (CPP), hematocrit, hemoglobin, serum Na+, K+, osmolarity, and fluid balance were carried out for 4 hours. Urine output was replaced volume for volume with 5% dextrose in 0.45% NaCl solution. We tested the hypothesis that patients with high (greater than or equal to 70 torr) CPP would respond less well to mannitol by either ICP or CPP criteria than patients with low (less than 70 torr) CPP. The rationale for this hypothesis was based upon the association of low CPP with autoregulatory vasodilatation, whereas high CPP is associated with vasoconstriction. If mannitol should work by a vasoconstriction mechanism, the ICP effects should be most apparent under conditions of low CPP. Those patients with CPP greater than or equal to 70 torr responded relatively poorly to mannitol, with ICP decreasing from 25 +/- 4 to 17 +/- 5 (SE) mm Hg at 45 minutes postinfusion. Patients with CPP less than 70 responded with ICP declining from 35 +/- 5 to 13 +/- 4 mm Hg. The initial SABP was 81 +/- 5 mm Hg in the CPP less than 70 group and immediately rose to 90 +/- 7 at 15 minutes postmannitol. The SABP increase correlated with ICP (r = -0.40, P less than 0.01), but not when CPP greater than or equal to 70. SABP was significantly higher in the latter group (105 +/- 6) and increased to 107 +/- 8 postmannitol. The ICP decrease began immediately with the SABP increase. No mannitol "rebound" occurred in these patients. Measures of acute volume expansion all correlated with ICP (Na+, r = -0.67, P less than 0.001; hematocrit, r = -0.27, P less than 0.01; serum osmolarity, r = 0.32, P less than 0.05) when CPP less than 70 torr. None correlated with ICP when CPP greater than or equal to 70. These data suggest that mannitol infusion is at least partly dependent upon hemodynamic mechanisms that allow vasoconstriction to occur with reduction in cerebral blood volume and that little may be gained by using mannitol when CPP greater than or equal to 70 either by SABP, ICP, or CPP criteria because vasoconstriction is already nearly maximal. This mechanism is not exclusive of other potential mechanisms of action. Mannitol "rebound" may be a function of net dehydration, hemoconcentration, and SABP decline.