The affected sib-pair method can be applied to investigate linkage between a marker locus and a disease. Several statistics have been proposed to test if the observed pattern of marker alleles shared identically by descent (ibd) is compatible with the null hypothesis of no linkage. Here, we consider different optimality criteria for sib-pair linkage tests. While for recessive inherited diseases the mean test is found to be uniformly (in Θ) most powerful, it can also be shown that, irrespective of the mode of inheritance, the mean test is the locally optimal test.