The number of genotypes investigated per population is important for the reliability of diversity studies. The objective of this study was to determine the sample size for the identification of differences among populations of an outcrossing autotetraploid species, alfalfa (Medicago sativa), using codominantly coded SSR markers. One hundred and twenty genotypes from each of two closely related populations were analysed with two markers. Twenty random subsamples for each of three sample sizes (10, 20 and 40 genotypes) were built. Compared to the populations with 120 genotypes, alleles that were no longer present in subsamples with 40 genotypes were mainly rare, whereas abundant alleles were also excluded in subsamples with 10 genotypes. F (ST) values for pairs of subsamples between the two populations were always significantly different based on 40 genotypes, whereas for 10 genotypes more than half of the pairs were not significantly different. We concluded that 40 genotypes are a reasonable sample size for diversity studies with closely related populations of tetraploid alfalfa investigated with SSR markers. Twenty genotypes may be an economical alternative for large scale studies, but 10 genotypes were a too low number for reliable results.
