BACKGROUND Standardised expression of results of bronchoalveolar lavage (BAL) is problematical in the absence of a validated "denominator" of epithelial lining fluid dilution. The suitability of albumin in BAL fluid has been investigated in groups of clinically stable asthmatic and control subjects. METHODS Absolute levels of albumin in BAL fluid were measured in a preliminary study of 21 asthmatic and 10 control subjects. In a more complex study designed to investigate the origin of albumin sampled at BAL in nine asthmatic and seven control subjects, radiolabelled albumin was injected intravenously five minutes before BAL. RESULTS In the preliminary study levels of albumin in BAL fluid were very similar, with a geometric mean value of 44 (95% CI 35-54) micrograms/ml BAL supernatant for the asthmatic subjects and 41 (95% CI 33-52) micrograms/ml for the controls. The majority of control and asthmatic subjects in the radiolabel study exhibited minimal flux of albumin from the circulation into the BAL aspirate. This finding was not uniform, however, and in a third of the asthmatic subjects an albumin flux equivalent to > 20% of the measurable albumin was found in two or more aliquots of a 3 x 60 ml lavage. CONCLUSIONS The results of this investigation into the source of albumin sampled at BAL suggest that, in general, albumin would be a reasonable reference solute for normalising the degree of dilution of BAL fluid in the groups studied. The origin of albumin was not always restricted to the bronchopulmonary segment under investigation, however, with significant leakage from the blood compartment in some individuals despite the consistency of absolute levels observed in the preliminary study.