Batesian mimicry is a well‐studied adaptation for predation avoidance, in which a mimetic species resembles an unpalatable model species. Batesian mimicry can be under positive selection because of the protection gained against predators, due to resemblance to unpalatable model species. However, in some mimetic species, nonmimetic individuals are present in populations, despite the benefits of mimicry. The mechanism for evolution of such mimetic polymorphism remains an open question. Here, we address the hypothesis that the abundance of mimics is limited by that of the models, leading to mimetic polymorphism. In addition, other forces such as the effects of common ancestry and/or isolation by distance may explain this phenomenon. To investigate this question, we focused on the butterfly, Papilio polytes, that exhibits mimetic polymorphism on multiple islands of the Ryukyus, Japan, and performed field surveys and genetic analysis. We found that the mimic ratio of P. polytes was strongly correlated with the model abundance observed on each of the five islands, suggesting negative frequency‐dependent selection is driving the evolution of polymorphism in P. polytes populations. Molecular phylogenetic analysis indicated that the southern island populations are the major source of genetic diversity, and the middle and northern island populations arose by relatively recent migration. This view was also supported by mismatch distribution and Tajima's D analyses, suggesting a recent population expansion on the middle and northern islands, and stable population persistence on the southern islands. The frequency of the mimetic forms within P. polytes populations is thus explained by variations in the model abundance rather than by population structure. Thus, we propose that predation pressure, rather than neutral forces, have shaped the Batesian mimicry polymorphism in P. polytes observed in the Ryukyus.