The affected-pedigree-member (APM) method of linkage analysis is a nonparametric statistic for testing for nonindependent segregation of a marker to affected members of a pedigree. We present here results of a simulation study evaluating the power of the APM method to detect linkage. We have systematically explored, by computer simulation, the effect of a variety of factors on the power to detect linkage using the single-locus APM statistic. These factors include mode of inheritance, marker polymorphism, the distance between marker and disease, phenocopy rate, heterogeneity, and misspecified marker allele frequencies. We also evaluated the relative power obtained under fixed-structure sampling and sequential sampling. For a dominant disease, sequential sampling led to increased power as compared to fixed-structure sampling, while for a recessive disease, there was no clear advantage in sampling beyond nuclear families.