It is demonstrated that in linkage analyses with a misclassification error p of the phenotypes at one locus, a wrong specification of p leads to a bias in the estimate of the recombination fraction. In particular, if p is not taken into account at all in the analysis, the recombination fraction is overestimated. This is shown for offspring of phase‐known double backcross matings. In the light of these results, the possible effects of dichotomization of quantitative phenotypes are discussed and illustrated with an example of the linkage between hypercholesterolemia and C3.