Microsatellites are highly mutable elements in eukaryotic genomes. Although they are widely used by researchers as genetic markers and tools for evolutionary studies, the full mutation mechanisms and factors affecting the mutation rate of microsatellites are not well understood. Microsatellite loci shared by human, chimpanzee, gorilla, orangutan and rhesus macaque genomes were sorted into four different perfection status groups named Perfect, Imperfect, Perfect-Compound, and Imperfect-Compound. We found that compound microsatellite loci generally had no significant effect on mutation rates, while imperfect microsatellite loci significantly lowered mutation rates compared to perfect ones ( P = 1 × 10⁻⁴). The significant difference resulted from a small amount of interruption (1–2 bp) within microsatellites, especially when the loci were shorter than 14–15 repeats. Furthermore, real perfect loci were used to compare with split up ‘perfect loci’ actually obtained from the imperfect group. We found that the mutation rates were significantly different from each other for small numbers of repeats, especially in 7–9 repeats ( P < 0.05). This suggests that an imperfect locus should not be considered as two or more separated perfect loci. This also raises the question that the algorithms currently used to find microsatellite loci based on mismatch penalties may result not only in heterogeneous microsatellite data sets with heterogeneous numbers of imperfect loci but also in data sets that are biased in mutation rate because of these imperfect loci.