Recently, some evolutionary biologists have argued that selection on the male component of fitness shapes the evolution of reproductive characters in angiosperms. Floral features, such as inflorescence size, that lead to increased insect visitation without a concomitant increase in seed production are viewed as adaptations to enhance the probability offathering seeds on other plants. In tests of this "pollen donation hypothesis," male reproductive success has usually been measured indirectly by flower production, pollinator visitation, or pollen removal. We tested the pollen donation hypothesis directly by quantifying the number of seeds sired by individual genotypes in a natural population of poke milkweed, Asclepias exaltata, in southwestern Virginia. Multiple paternity was low within fruits, a fact which allowed us to use genotypes ofprogeny arrays to identify a unique pollen parent for 85% of the fruits produced in the population. Seeds sired (male success)and seeds produced (female success)were significantly correlated with flower number per plant (for male success, r = 0.32, P < 0.05; for female success, r = 0.66, P < 0.001). While the number of pollinaria removed, the usual estimator of male success in milkweeds, was highly correlated with numbers of seeds sired (r = 0.47; P < 0.001), it was even more highly correlated with numbers of seeds produced (r = 0.71, P < 0.001). Analysis of functional gender indicated that plants with many flowers did not behave primarily as males. In fact, individuals with the highest total reproductive success contributed equally as males and females. Furthermore, estimates of gender based on numbers of flowers produced or pollinaria removed overestimated the number of functional males in the population. In pollen-limited species, such as many milkweeds, proportional increases in both male and female reproductive success indicate the potential for selection to shape the evolution of large floral displays through both male and female functions.