In the context of a study of wild chimpanzees, Pan troglodytes verus, we found that genotypes based on single PCR amplifications of microsatellite loci from single shed hair have a high error rate. We quantified error rates using the comparable results of 791 single shed hair PCR amplifications of 11 microsatellite loci of 18 known individuals. The most frequent error was the amplification of only one of the two alleles present at a heterozygous locus. This phenomenon, called allelic dropout, produced false homozygotes in 31% of single-hair amplifications. There was no difference in the probability of preferential amplification between longer and shorter alleles. The probability of scoring false homozygotes can be reduced to below 0.05 by three separate amplifications from single hairs of the same individual or by pooling hair samples from the same individual. In this study an additional 5.6% of the amplifications gave wrong genotypes because of contamination, labelling and loading errors, and possibly amplification artefacts. In contrast, amplifications from plucked hair taken from four dead individuals gave consistent results (error rate < 0.01%, n = 120). Allelic dropout becomes a problem when the DNA concentration falls below 0.05 ng/10 microL in the template as it can with shed hair, and extracts from faeces and masticated plant matter.