SUMMARY A model is given for line transect sampling for the purpose of estimating animal population density which makes no assumption about the value of the detection probability along the transect or about the form of the detection probability function other than continuity. In addition, animals are permitted to be available, or not, according to a process that is independent of the sampling process. The model requires sampling at two effort levels, either by varying the speed of the observers and/or their number. The unconditional maximum likelihood estimate does not exist, whereas the maximum likelihood estimate, given the total numbers detected at each effort level, exists but does very poorly. A generalization of a moment-like estimator given by Cooke (1985, Report of the International Whaling Commission 35, 319-323) is proposed instead. Asymptotic and small-sample behavior is examined. The estimator is also applied to a field experiment.