A large majority of human microsatellite markers are longer than their homologues in chimpanzees, suggesting that more expansion mutations have occurred in the lineage leading to humans. However, such a length difference has also been explained as arising from the selection of unusually long microsatellites as genetic markers. In order to resolve this controversy and to establish the true source of the observed length differences, we have now conducted the necessary reciprocal study. We have compared the lengths of size-selected markers cloned from chimpanzees between this species and humans. We find that of 19 markers which were informative and polymorphic in both species, 13 are longer in humans. This result is incompatible with ascertainment bias being the sole explanation for the inter-specific length differences. We estimate that dinucleotide repeat microsatellites are an average of 3.2 repeat units longer in humans than in chimpanzees, implying a mutational bias in favour of microsatellite expansions and a higher average genome-wide microsatellite mutation rate in the human lineage.