Identifying traits and agents of selection involved in local adaptation is important for understanding population divergence. In southern Sweden, the moth‐pollinated orchid Platanthera bifolia occurs as a woodland and a grassland ecotype that differ in dominating pollinators. The woodland ecotype is taller (expected to influence pollinator attraction) and produces flowers with longer spurs (expected to influence efficiency of pollen transfer) compared to the grassland ecotype. We examined whether plant height and spur length affect pollination and reproductive success in a woodland population, and whether effects are non‐additive, as expected for traits influencing two multiplicative components of pollen transfer. We reduced plant height and spur length to match trait values observed in the grassland ecotype and determined the effects on pollen removal, pollen receipt, and fruit production. In addition, to examine the effects of naturally occurring variation, we quantified pollinator‐mediated selection through pollen removal and seed production in the same population. Reductions of plant height and spur length decreased pollen removal, number of flowers receiving pollen, mean pollen receipt per pollinated flower, and fruit production per plant, but no significant interaction effect was detected. The selection analysis demonstrated pollinator‐mediated selection for taller plants via female fitness. However, there was no current selection mediated by pollinators on spur length, and pollen removal was not related to plant height or spur length. The results show that, although both traits are important for pollination success and female fitness in the woodland habitat, only plant height was sufficiently variable in the study population for current pollinator‐mediated selection to be detected. More generally, the results illustrate how a combination of experimental approaches can be used to identify both traits and agents of selection.
