1. Extravascular albumin in human tissue was determined by direct immunochemical analysis. Samples of muscle, skin and intestinal tract obtained at surgery were extracted with a solution of 0· 9% sodium chloride plus 0· 1% desoxycholate and total albumin of the extracts was determined by radial diffusion in agarose plates containing antibody to human albumin. Occluded blood in tissues was estimated from the haemoglobin content of the extracts and the albumin due to blood calculated. This was used to obtain by difference the extravascular albumin content of the tissues. The extravascular albumin content of muscle and skin from seven patients was (mean and range) 3·6 (2·7–6·4) and 8·4 (6·5–12) mg/g, respectively. The albumin content of intestinal tissue (five patients) ranged from 4·3 to 9·0 mg/g (mean 6·3), and of stomach wall (three patients), from 3·5 to 5·4 mg/g (mean 4·4). Total extravascular albumin in humans is estimated to be at least 4 g/kg.
2. The extravascular (EV) albumin of human tissue is similar to that of rats. It is concluded that calculation of the extravascular albumin from the intercepts and slopes of the decay curve of injected radioactive iodine-labelled albumin leads to underestimation of the extravascular mass by 20–40%. The validity of the conventional multicompartmental model of albumin metabolism in the human is questioned.