Among size-dimorphic animals, a few clades such as hummingbirds show ‘reversed’ sexual size dimorphism: females tend to be the larger sex. What selects for this pattern? Sexual selection for flight performance could drive the evolution of smaller, more agile males, either for male-male combat or female choice for aerial courtship displays. Alternately, natural selection can select for female fecundity (e.g. egg size influences female body size), or sex differences in foraging niche could favor body size differences. The sexual selection hypotheses predict that dimorphism extends to other aspects of flight morphology (e.g. flight muscle size) whereas the natural selection hypotheses predict that male and female flight morphologies are isometric, and the niche differentiation hypothesis predicts that bill dimorphism is correlated with size dimorphism. We tested these predictions through phylogenetic comparative analyses of flight morphology, wingbeat frequency, and courtship behaviors, focused on 30 species within the "bee" hummingbird clade (tribe Mellisugini). There is no correlation between bill morphology and dimorphism. Relative to females, males tend to be smaller, have proportionately shorter wings and higher hovering wingbeat frequencies, but also longer keels and larger flight muscles. Male wingbeat frequencies are greatly elevated during aerial displays, and the species with the greatest wingbeat frequencies have the greatest dimorphism. Of the four hypotheses for dimorphism, the data best support the hypothesis that female choice for courtship displays has selected for aerial agility and small size in male hummingbirds.