We address the problem of microsatellite genotyping errors associated with polymerase chain reaction (PCR) amplification from degraded and dilute template DNA and provide suggestions for improving the accuracy of genotype data in studies using older museum specimens as a source of DNA. In the course of a population genetics study of African indigobirds (Vidua spp.), we used replicate PCR to evaluate genotyping reliability for nine microsatellite loci in relation to PCR fragment length and DNA template concentration (DNA extracted from the calamus of one vs. two wing feathers). Complete amplification failure and the dropout of one allele from heterozygous genotypes were the predominant problems encountered. For samples with heterozygous genotypes, allele dropout occurred in 19.2 and 12.1% of PCR using extracts derived from one and two feathers, respectively. The amplification of artifact bands was less frequent (affecting 4.9 and 1% of positive PCR reactions with one- and two-feather extracts, respectively). Those results indicate that multiple replicates per sample and locus are required to obtain accurate genotype data from museum feather samples. Although higher DNA concentration improved success, PCR fragment size had a much stronger influence on the success and repeatability of microsatellite amplification, which suggests that the accuracy and efficiency of genotyping can be improved most easily by designing primers that amplify smaller DNA fragments.