1. A growth model for brown trout, developed almost 20 years ago, has been used to investigate growth potential in at least 40 populations over a wide geographical range. The chief disadvantages of the model are: it is based on growth data for only 55 hatchery trout kept in tanks without strict control of temperature and oxygen, it is not continuous and is restricted to the range 3.8-19.5-degrees-C, it requires six parameters and only one of these can be interpreted biologically. 2. For the new model, growth data were obtained for an additional 130 trout bred from wild parents and kept in tanks at five constant temperatures (range +/- 0.1 or 0.2-degrees-C) and 100% oxygen saturation. The new model is continuous over the range 3.8-21.7-degrees-C and has five parameters, all of which can be interpreted in biological terms. It was fitted to growth data for individual fish and was an excellent fit (P < 0.001, R2 > 0.99) to the data for the 55 trout of the original experiment, the 130 trout of the new experiment and both experiments combined. The procedure for applying the model to field data is critically examined and a suitable test for maximum growth potential is described. The model ceases to be robust when mean temperatures are estimated over periods of more than 3 months. 3. Although parameter estimates for the new model are similar for the original and new experiments, they are significantly different. An iterative exercise, varying common and different parameters, showed this to be the result of slight differences between two parameters; the optimum temperature for growth and the growth rate of a 1-g fish at this temperature. Possible reasons for this are discussed and it is concluded that these differences have a negligible effect on values predicted from the model.