Plasma colloid osmotic pressure (COP) has been calculated from both serum albumin concentration and plasma total protein concentration. These values have been compared to those measured directly using a membrane-transducer oncometer in a group of normal subjects, in a group of critically-ill patients with a variety of primary diagnoses and in a group of hypovolaemic patients before and after plasma volume replacement with 6% hydroxyethyl starch solution. In the normal samples, COP calculated from albumin (COPalb) underestimated the measured COP (COPm) by mean of 2.0 mmHg (p less than 0.002), with correlation coefficient r = 0.39(n/s). Similarly, the COPalb underestimated COPm by a mean of 5.7 mmHg (p less than 0.001) in the critically ill patient group; r = 0.38 (p less than 0.02). Furthermore, in the patients receiving plasma volume replacement serum albumin concentration fell by 13.1% (p less than 0.001) whilst COPm increased by 11.5% (p less than 0.002). In the normal subjects COP calculated from total protein concentration (COPtp) underestimated the COPm by 1.5 mmHg (p less than 0.02) with r = 0.65 (p less than 0.01). Conversely, in the patient samples, mean COPtp overestimated COPm by 3.5 mmHg (p less than 0.001) with r = 0.73 (p less than 0.002). We conclude that COPalb is an inadequate estimate of COPm particularly in patients where its use may have important clinical consequences. COPtp provides a reasonable estimate of COPm in normal subjects but in patients samples, where albumin: globulin ratio is low COPtp overestimates substantially in many cases. We advocate the direct measurement of COP in critically-ill patients.