Resistance to warfarin and an increased vitamin K requirement appear to be pleiotropic effects of the same allele (Rw2). In a natural population containing resistant individuals where the use of warfarin is discouraged the change in the frequency of resistance should reflect the relative fitnesses of the three possible genotypes. A large polymorphic population of rats was extensively poisoned with warfarin and the level of resistance monitored regularly for a period of 18 months after withdrawal of the poison. During this period the proportion of resistant animals in live-capture samples decreased significantly from approximately 80 per cent to 33 per cent. This decline is consistent with a hypothesis of reduced fitness of both Rw2Rw2 and Rw1Rw2 genotypes relative to Rw1Rw1 under natural conditions. The relative fitnesses of these genotypes were calculated using an optimisation method based on least squares analysis. These estimates were: Rw2Rw2 (0.46), Rw1Rw2 (0.77) and Rw1Rw1 (1.00). Homozygous resistant individuals were found in some of the samples, confirming that the Rw2 allele does not act as a recessive lethal, although it must be extremely disadvantageous. Some heterogeneity was observed in the proportion of resistant animals in samples taken from different areas of the farm building complex. This could reflect stochastic processes influencing the Rw2 allele frequency in small peripheral populations.