In this study we investigated microsatellite variation in Dutch roe deer (Capreolus capreolus) populations. We used 65 tissue samples from culled animals from three populations (Amsterdamse Waterleidingduinen, National Park Zuid-Kennemerland and Flevopolder). The first two are dune populations and are about 3.5 kilometers apart. In both populations, roe deer have been present since the early 1950's. In the Flevopolder roe deer were first observed in 1959. From theoretical predictions, a decrease in heterozygosity of 20ue to genetic drift could be expected in the small Amsterdamse Waterleidingduinen population, compared to the much larger Flevopolder population. However, the expected heterozygosity (He) at seven loci was 0.56 on average with no significant differences in He between populations. The probability that a decrease of 20␘r more did occur but that it went undetected is smaller than 2.2ÐAll populations were significantly differentiated from each other and showed positive FST and Rho values, suggesting limited gene flow between populations. The fact that the decrease of genetic variation was smaller than expected is probably due to gene flow between the Amsterdamse Waterleidingduinen and National Park Zuid-Kennemerland, effective population size at the time of introduction being larger than assumed, and/or because the animals which were used for stocking came from different populations. Considering the amount of variation still present in the different populations, negative effects of reduced genetic variation are currently not expected, but in both dune populations the decrease in the amount of variation will continue.