In order to test the "anti-interference" hypothesis for adaptive significance of reciprocal herkogamy, patterns of illegitimate pollination and effects of self-pollen on the legitimate pollination and fertility were investigated in a naturally pollinated experimental population of distylous Persicaria japonica. Pollen deposition was compared among the emasculation treatments, i.e., emasculation of a single flower from individual inflorescences, emasculation of all the flowers of individual inflorescences, and no emasculation control. In both morphs, considerable illegitimate pollination was found to occur in all the treatments, and there was no significant difference in incompatible pollen load among the treatments. Therefore, it is suggested that herkogamy of P. japonica can sufficiently reduce both intraflower and intra-inflorescence self-pollinations, but not interinflorescence geitonogamous and-or interclonal illegitimate pollination. Measurements of pollen load after the repetitive pollen addition showed that space on the stigma surface may not limit the legitimate pollination under natural pollination conditions. Seed sets after legitimate pollination following prior self-pollination did not differ from the controls without self-pollination. Therefore, there is little possibility that incompatible pollen load interferes with either pollination or fertilization by compatible pollen, suggesting that "anti-interference" is unimportant for adaptive significance of reciprocal herkogamy at least in P. japonica.