A novel Bayesian approach to replication studies, allowing for locus heterogeneity, is introduced. Compared with currently used approaches to replication, it offers a natural way to accumulate evidence across independently collected data sets and yields more interpretable results. Using for replicates (one as initial study and the other three as replication studies) from Problem 2 of the Genetic Analysis Workshop 11 data, we show the performance of this method. All four disease susceptibility loci (D1G009, D1G024, D3G045, D5G035) are identified and accurately mapped, with no false positive signals.