Though extensively used in a variety of disciplines, the evolutionary pattern of microsatellite sequences is still unclear. We addressed several questions relating to microsatellite evolution by analysing historically accumulated mutation events in a large set of artiodactyl (CA) n repeats, through sequence analysis of orthologous bovine and ovine loci. The substitution rate in microsatellite £anking sequences was not di¡erent from that in intron sequences, suggesting that if intron sequences in general are selectively neutral, sequences close to microsatellites are similarly so. This observation thus does not support the idea that successful heterologous ampli¢cation of microsatellites across distantly related taxa would be due to £anking sequences generally being under some form of selection. Interestingly, the substitution rate at the ¢rst nucleotide positions £anking repeats was signi¢cantly higher than in sequences further away. Moreover, the substitution rate in repeat units in the very end of microsatellites was signi¢cantly higher than that in the middle of repeat regions. Together these observations suggest a relative instability close to the boundary between repetitive and unique sequences. We present three models that potentially could explain such a feature, all involving ine¤ciency of mismatch repair systems.