Darwin's vexing contrivance: a new hypothesis for why some flowers have two kinds of anther

Heteranthery, the presence of two or more anther types in the same flower, is taxonomically widespread among bee-pollinated angiosperms, yet has puzzled botanists since Darwin. We test two competing hypotheses for its evolution: the long-standing ‘division of labour' hypothesis, which posits that some anthers are specialized as food rewards for bees whereas others are specialized for surreptitious pollination, and our new hypothesis that heteranthery is a way to gradually release pollen that maximizes pollen delivery. We examine the evolution of heteranthery and associated traits across the genus Clarkia (Onagraceae) and study plant–pollinator interactions in two heterantherous Clarkia species. Across species, heteranthery is associated with bee pollination, delayed dehiscence and colour crypsis of one anther whorl, and movement of that anther whorl upon dehiscence. Our mechanistic studies in heterantherous species show that bees notice, forage on and export pollen from each anther whorl when it is dehiscing, and that heteranthery promotes pollen export. We find no support for division of labour, but multifarious evidence that heteranthery is a mechanism for gradual pollen presentation that probably evolved through indirect male–male competition for siring success.