The (relative) importance of pollinator‐mediated selection for evolution of flowers

Perhaps the most intriguing example, surely the most aesthetic, of plant evolution is the interaction of fl owers and their pollinators, which has sparked a burst of studies spanning diverse disciplines, from animal behavior to plant genetics. Starting in the late 1700s ( Sprengel, 1793 ) and emphasized by Darwin (1862) , pollination was accepted as the driver of fl oral variation, making pollinatormediated selection the major paradigm for fl ower evolution. Indeed, numerous studies support the hypothesis that adaptation to pollinators commonly contributes to fl oral diversity ( Harder and Johnson, 2009 ). Pollinators are considered to select not only for a single traits, but also for sets of correlated traits (pollination syndromes; Fenster et al., 2004 ). Pollinators are also thought to be agents of divergent selection that can lead to speciation and maintain species boundaries, based on divergent trait values (e.g., Schemske and Bradshaw, 1999 ; Hopkins and Rausher, 2012 ). Plant fi tness has been quantifi ed in many ways, including pollinator visitation rate, fruit and seed production (female fi tness), and seeds sired (male fi tness). No matter how fi tness is measured, whether female, male, or cumulative fi tness, it is used for quantifying natural selection, which is the standardized covariance of a trait with fi tness (selection gradient, β ) ( Lande and Arnold, 1983 ; Kingsolver et al., 2001 ). Harder and Johnson (2009) summarized many studies and found ample evidence for natural selection on fl oral traits. However, the extent to which this selection is mediated by pollinators is still in question. As an example, in their meta-analysis Harder and Johnson (2009) showed that the mean selection gradient was signifi cantly diff erent from zero only when fi tness was quantifi ed as seed production, rather than as pollination (see Harder and Johnson, 2009 : fi g. 4a). While this low average value does not mean that all individual gradients diff er signifi cantly from zero, it does raise questions regarding the quantitative value of the selection gradient, which is, on average, very small. We propose that low selection values refl ect two possibilities. One possibility is that pollinator-mediated selection is relatively high but is not detected, due to masking by other agents of selection. Th e other possibility is that pollinatormediated selection does not exist or is too small to be detected. In this essay, we highlight the two alternatives, and while we do not dismiss pollinators as the major putative drivers of fl oral evolution, we think that expanding this view to consider other agents of evolution may improve our understanding of the actual role of pollinators in shaping fl owers.