The accuracy and reliability of a rapid blood alcohol estimation by means of a breath alcohol analyser has been evaluated under casualty department conditions in a series of 646 road-crash victims managed at the Dandenong and District Hospital. A higher correlation (r = 0.91) was found between the breath alcohol analyser readings in 633 casualties and those obtained by blood analysis in the police laboratory. In 13 unconscious casualties in whom a nasal breath test was performed, the correlation was lower (r = 0.76). In six casualties, the breath alcohol analyser readings showed lower alcohol concentrations than the legal limit of 0.05 g/100 mL (10.9 mmol/L), but blood analysis detected an illegal concentration. Further evaluation of the accuracy of the breath alcohol analyser in other casualty departments is necessary before it can be recommended as a screening device in States which have legislated for compulsory blood alcohol tests in adult road casualties.