Two-stage experiments allow results to be analysed at the end of the first stage and the second stage to be omitted if the preliminary conclusions are clear-cut. They thus offer the potential to reduce the number of experimental animals. However, using standard P values to assess the significance of results at the end of either stage will lead to an increase in risk of false positive conclusions. This paper provides a possible protocol for two-stage experiments and a method for adjusting P values. It is shown that, for experiments with reasonable power (>80%), the expected reduction in animal numbers will be at least 20%.