AbstractLinkage analysis under the two‐locus model and the admixture model was compared on pedigree data for a common disease simulated under a model of genetic heterogeneity. The ascertainment of families was designed so that the samples had a large proportion of families segregating for both disease loci. The two‐locus linkage analysis model did not demonstrate increased power of detecting linkage or more accurate estimates of the recombination fraction, θ than did the admixture model linkage analysis. When a sample was purposely chosen so thatallof the families were segregating for both loci, then the two‐locus lod score analysis was better. However, the increased power depended on assuming the correct gene frequency for the linked locus. It can be concluded that under the conditions of genetic heterogeneity examined here, testing for linkage under the admixture model is the preferred method of analysis. However, this is not a general conclusion that can apply to all two‐locus disease models. Published 1992 by Wiley