Pollinators may mediate selection on traits affecting pollinator attraction and effectiveness, and while nonadditive effects of traits influencing the two components of pollination success are expected when seed production is pollen limited, they have been little studied. In a factorial design, we manipulated one putative attraction trait (number of flowers) and one putative efficiency trait (spur length) previously shown to be subject to pollinator-mediated selection in the deceptive orchid Dactylorhiza lapponica. Removal of half of the flowers reduced pollen removal, proportion of flowers receiving pollen, fruit set, and fruit mass compared to unmanipulated plants, while spur-tip removal increased fruit set and fruit mass but did not affect pollen removal or proportion of flowers receiving pollen. The effect of spur-tip removal on fruit mass was stronger among plants with intact number of flowers compared to plants with experimentally reduced number of flowers. The results demonstrate that number of flowers and spur length are direct targets of selection and may affect female fitness nonadditively. More generally, they show that the adaptive value of a given trait can depend on floral context and illustrate how experimental approaches can advance our understanding of the evolution of trait combinations.