When closely related species breed in sympatry, and where hybrids have lower fitness, reinforcement theory predicts that selection should favour mechanisms that reduce the probability of interspecific matings. If this situation arises among species that exhibit resource defence polygyny where males and females of different species reside in the same territories, there may be some conflict between mate choice based on territory-holding ability (sexual selection) and mate choice for correct species. We investigated this in a population of fur seals where three species are sympatric and where some females breed in the territories of heterospecific males, and where interspecific matings and hybrid pups are observed. The territorial status of males and the birthing sites of females were determined during daily observations, as were the movements of males and females, the location of matings and mating partners. DNA extracted from skin samples was used to determine paternities using DNA fingerprinting and the mtDNA genotype of individuals. Individuals were also classed on the basis of species-typical phenotype. We found that extra-territory inseminations (ETIs) were significantly more prevalent (67%) when territorial males and resident females were of different phenotype than when of similar phenotype (27%), but mtDNA genotype had no effect on the rate of ETIs. ETIs were probably by males with the same phenotype, as pups born to these females in the following season had the same phenotype as their mothers, suggesting they were not hybrids. These results suggest that within the resource defence polygynous mating system of these sympatric fur seals, female mate choice is more influenced by male phenotype than genotype. Contrary to our predictions, our study indicates that potential conflict between mate choice based on sexual selection and species recognition is unlikely, because females have some capacity to discriminate between males both within and between species on phenotypic traits additional to those under sexual selection. Although at least 25% of the pups born in this study were hybrid, this study can only support reinforcement theory if hybrids have reduced fitness. The fitness of hybrids among the species studied is currently unknown.