In affected sibpair studies of late-onset diseases, such as prostate cancer, DNA samples for subjects' parents are often unavailable. In such situtations, genotyping errors cannot be detected by the usual methods, since all possible genotypes conform to Mendel's rules. We have studied two approaches to dealing with genotyping errors in such data, making use of multipoint marker information. First, we consider a likelihood ratio statistic to identify genotypes likely to be in error. Second, we calculate IBD probabilities given multipoint marker data using an incomplete penetrance function which allows for the presence of errors. We present the results of a simulation study comparing the power to detect a disease susceptibility gene when (a) genotyping errors are ignored, (b) genotypes likely to be in error are detected and removed, and (c) using an analytic method which allows for the presence of genotyping errors. We studied the effect of marker density, different error models, gene effect, sample size and error rate on the outcome. Genotyping errors are shown to erode the power to detect linkage. The effect may be ameliorated by use of methods to detect or allow for the presence of errors, though marker density must be quite high.