Management units based on biological subpopulation structure are a critical base for wildlife management. Although subpopulation structure can be detected using landscape genetics, it is more important for successful management to understand the nature of subpopulation structure. The success rate of a management plan depends on the degree of movement among subpopulations. It is reported that the Hokkaido population of sika deer consists of 2 microsatellite DNA-based subpopulations (Ou et al. 2013), but our understanding of their stability, independency, and boundary has not been sufficed. To recognize those features we analyzed 521 samples collected from the entire mainland of Hokkaido, using GENELAND and SPAGeDi. Two microsatellite DNA-based subpopulations were identified, and the probability of each individual sample belonging to either of the two drastically changed around the boundary. This disruptive structure was not explained by the effect of geographic distance because pairwise genetic distances were significantly higher between individuals belonging to different subpopulations than between individuals belonging to the same subpopulation, even though their geographic distances were similar. Although the landscape of Hokkaido is heterogeneous, there is no clear landscape barrier preventing deer from moving between the subpopulations. These results suggest another effect than landscape for the subpopulation formation.