Abstract | dc.description.abstract | Flower herbivory and pollination have been described as interactive processes that influence each other in their effects on plant reproductive success. Few studies, however, have so far examined their joint effects in natural populations. In this paper we evaluate the influence of flower damage and pollination by the hummingbird Oreotrochilus leucopleurus on the fecundity of the Andean monkey flower Mimulus luteus. We performed a 2x2 factorial experiment, with artificial clipping of lower petals and selective exclusion of the hummingbird as main factors. In spite of the relatively low proportion (27.5%) of the variance in seed production accounted for by the full factorial model, artificial damage and hummingbird exclusion, as well as their interaction, were highly significant, indicating nonadditive effects of factors on plant fecundity. In the presence of hummingbirds, undamaged flowers had a seed production that was 1.7-fold higher than for damaged flowers, suggesting that the effect of flower damage on female reproductive success occurs probably as a consequence of hummingbird discrimination against damaged corollas. This result indicates that the impact of flower herbivory on plant fecundity was contingent on the presence or absence of hummingbirds, suggesting that pollinators may indirectly select for undamaged and probably resistant flower phenotypes. A second interaction effect revealed that undamaged flowers produced 78.5% more seeds in the absence of rather than in the presence of O. leucopleurus, raising the question of the ecological mechanism involved. We suggest that the strong territorial behavior exhibited by the bee Centris nigerrima may confine the foraging activities of the remaining bee species to safe sites within exclosures. Overall, our results provide evidence that hummingbird pollination and flower herbivory have interdependent effects on M. luteus fecundity, which indicates that it will be difficult to predict their ecological and evolutionary consequences unless interactions are analyzed in an integrated form. | en |