Traditionally, retrospective estimates of blood alcohol concentration (BAC) have been made by assumption of a steady rate of elimination, some considerable variation in which has been reported. During recent years, in contradiction of Widmark's hypothesis, it has been confirmed that the metabolic rate of alcohol is not constant but varies, in a predictable way, with BAC. The present study considers this dichotomy and shows that the experimental evidence hitherto used in support of Widmark's hypothesis can, more satisfactorily, be interpreted on the basis of first-order absorption kinetics together with Michaelis–Menten elimination. It is thus demonstrated that a range of apparent linear rates, from as little as 120 mg l −1 h −1 to over 200 mg l −1 h −1, can be seen in a single subject. The implications for retrospective estimation of level are examined and a simple procedure is proposed to enable a more realistic assessment to be made than is given by the historical linear assumption.