Most studies of microsatellite evolution utilize long, highly mutable loci, which are unrepresentative of the majority of simple repeats in the human genome. Here we use an unbiased sample of 2,467 microsatellite loci derived from alignments of 5.1 Mb of genomic sequence from human and chimpanzee to investigate the mutation process of tandemly repetitive DNA. The results indicate that the process of microsatellite evolution is highly heterogeneous, exhibiting differences between loci of different lengths and motif sizes and between species. We find a highly significant tendency for human dinucleotide repeats to be longer than their orthologues in chimpanzees, whereas the opposite trend is observed in mononucleotide repeat arrays. Furthermore, the rate of divergence between orthologues is significantly higher at longer loci, which also show significantly greater mutability per repeat number. These observations have important consequences for understanding the molecular mechanisms of microsatellite mutation and for the development of improved measures of genetic distance.T he human genome is composed of 40-50% repetitive DNA, an important class being simple tandem repeats or microsatellite DNA sequences (1). Microsatellites are iterations of short (1-6 bp) sequence motifs, repeat numbers generally being less than 30 (2). They are spread over the genome with an estimated average density of one locus per 2-30 kb, the frequency being dependent on the criteria used for defining a microsatellite locus (1, 3). Similar microsatellite densities have also been documented for other eukaryotic genomes (4). Microsatellites are instrumental as genetic linkage markers in genome mapping projects (5) and have also found widespread use for evolutionary and population genetics studies in many species (6, 7), including humans (8).Microsatellites differ from most other DNA sequences in their high degree of polymorphism, with heterozygosities commonly exceeding 70%. As they generally seem to be free of selective constraints, it is evident that the extensive degree of genetic variability requires a high underlying mutation rate. Estimates of the human genomic mutation rate are in the range of 10 Ϫ4 to 10 Ϫ2 per meiosis (9), several orders of magnitude higher than that of unique DNA sequences (10, 11). It is commonly assumed that microsatellite mutations arise from replication slippage (slipped strand mispairing), a process thought to result in the gain or loss of one or a few repeat units (12)(13)(14). A microsatellite locus may initially evolve from the random occurrence of a few repeat units embedded within unique sequence and subsequently, after several steps of repeat expansion, reach a stage of an appreciable number of repeats. In theory, such expansion can proceed indefinitely in the absence of selection unless there are mechanisms that direct mutations toward contractions (15)(16)(17) or that make loci collapse (e.g., point mutations or large deletions; ref. 18).Based on the assumption that replication slippage is the main cause...