Introduction

The remarkable diversity of the angiosperms is often attributed to their specialized reproductive associations with diverse pollinating vectors (Crepet 1984; Eriksson and Bremer 1992; Grimaldi 1999). The appeal of this argument lies in the premise that specializing on particular pollinating groups can directly result in reproductive isolation from related plant species, even in sympatry, thus generating diversity (Grant 1949; Stebbins 1970). This pollinator shift model is the dominant paradigm explaining floral diversification and is often invoked as an important plant speciation mechanism (reviewed in Kay and Sargent 2009). However, pollinator shifts, which are usually quantitative, not qualitative, often result only in weak reproductive isolation (Armbruster and Muchhala 2009). In addition, sympatric species that use different pollinators are rarely sister species, and most often also exhibit substantial postmating isolation. These observations question the validity of the link between pollinator shifts and reproductive isolation, which underlies the pollinator shift paradigm, and suggest that floral divergence associated with pollinator shifts is unlikely to be a product of selection for reproductive isolation in sympatry, except perhaps upon secondary contact. Instead, floral diversity might result largely from spatially variable influences on efficient gamete transfer (Johnson 2006), a perspective that opens the possibility of numerous mechanisms influencing the divergence of floral traits.

A hallmark of this perspective should be substantial variation in floral traits between closely related allopatrically distributed species, but also between populations of species. Numerous studies document such floral diversity (Herrera et al. 2006; Ellis and Johnson 2009; Schlumpberger et al. 2009), suggesting that geographical variation in floral traits is ubiquitous. Approximately 20 % of British plant species (Warren and Mckenzie 2001), 40 % of the Polemoniaceae (Schemske and Bierzychudek 2007), 38 % of Cape Erica species (Rebelo and Siegfried 1985), and 40 % of Protea species (Carlson and Holsinger 2010) exhibit flower color polymorphisms. Although these numbers confound intra- and interpopulation variation, they do attest to the extent of variation in just a single floral trait.