The validity and reliability of plasma volume estimation using indocyanine green were investigated in five in vitro experiments and in three in vivo series. The in vitro measurements reflected real volumes with an error of about 1%. Comparative measurements in the same patients using indocyanine green or Cr51 labelled red cells differed by 1.7% (r = 0.97). The mean (SD) plasma volume difference between two successive plasma volume measurements using indocyanine green was 38 (43) ml (r = 0.99). Plasma volume measured before and about 7 min after a hyperosmolar saline bolus (100 ml, 1 molar) was increased by 223 (102) ml and 286 (49) ml when determined by indocyanine green and plasma protein changes respectively. Nevertheless, the necessity for central venous injection and arterial sampling restricts the possible application of the method to intra-operative or emergency care use.