Pollinator shifts drive increasingly long nectar spurs in columbine flowers.

Directional evolutionary trends have long garnered interest because they suggest that evolution can be predictable. However, the identification of the trends themselves and the underlying processes that may produce them have often been controversial. In 1862, in explaining the exceptionally long nectar spur of Angraecum sesquipedale, Darwin proposed that a coevolutionary 'race' had driven the directional increase in length of a plant's spur and its pollinator's tongue. Thus he predicted the existence of an exceptionally long-tongued moth. Though the discovery of Xanthopan morgani ssp. praedicta in 1903 with a tongue length of 22 cm validated Darwin's prediction, his 'race' model for the evolution of long-spurred flowers remains contentious. Spurs may also evolve to exceptional lengths by way of pollinator shifts as plants adapt to a series of unrelated pollinators, each with a greater tongue length. Here, using a species-level phylogeny of the columbine genus, Aquilegia, we show a significant evolutionary trend for increasing spur length during directional shifts to pollinators with longer tongues. In addition, we find evidence for 'punctuated' change in spur length during speciation events, suggesting that Aquilegia nectar spurs rapidly evolve to fit adaptive peaks predefined by pollinator morphology. These findings show that evolution may proceed in predictable pathways without reversals and that change may be concentrated during speciation.