Plasma-to-peritoneal transport rate of albumin (TERperit.space) was determined in eighteen patients with decompensated cirrhosis by sampling ascitic fluid after i.v. injection of 125I-labelled serum albumin. Median TERperit.space was 0.30% of the intravascular albumin mass (IVM) per hour (range 0.10-0.59). The transport rate of albumin from ascitic fluid back to plasma was measured in eight patients by plasma sampling after intraperitoneal injection of 131I-labelled serum albumin. After correction for tracer re-extravasation this back transport (median 0.31, range 0.07-0.44% IVM/h-1) was not significantly different from the simultaneously determined TERperit.space, indicating steady state with respect to albumin flux. Analysis of errors of the method was performed from experiments in pigs, and it is concluded that especially a high content of non-protein bound iodine and even minor bleeding from the abdominal puncture may lead to overestimation of TERperit.space, whereas systematic understimation seems less likely. This may besides differences in patient selection and unsteady state, account for the discrepancy between the present relatively low value and earlier reports on much higher values of TERperit.space in cirrhosis.