According to the Grant–Stebbins model of pollinator-driven divergence, plants that disperse beyond the range of their specialized pollinator may adapt to a new pollination system. Although this model provides a compelling explanation for pollination ecotype formation, few studies have directly tested its validity in nature. Here we investigate the distribution and pollination biology of several subspecies of the shrub Erica plukenetii from the Cape Floristic Region in South Africa. We analyze these data in a phylogenetic context and combine these results with information on pollinator ranges to test whether the evolution of pollination ecotypes is consistent with the Grant–Stebbins model.

Pollinator observations showed that the most common form of E. plukenetii with intermediate corolla length is pollinated by short-billed Orange-breasted sunbirds. Populations at the northern fringe of the distribution are characterized by long corollas and are mainly pollinated by long-billed Malachite sunbirds. A population with short corollas in the center of the range was mainly pollinated by insects, particularly short-tongued noctuid moths. Bird exclusion in this population did not have an effect on fruit set, while insect exclusion reduced fruit set. An analysis of floral scent across the range, using coupled gas chromatography-mass spectrometry, showed that the scent bouquets of flowers from moth-pollinated populations are characterized by a larger number of scent compounds and higher emission rates than those in bird-pollinated populations. This was also reflected in the clear separation of moth- and bird-pollinated populations in a two-dimensional phenotype space based on a non-metric multidimensional scaling analysis of scent data. Phylogenetic analyses of chloroplast and nuclear DNA sequences strongly supported the monophyly of E. plukenetii, but not of all the subspecies. Reconstruction of ancestral character states suggests two shifts from traits associated with short-billed Orange-breasted sunbird pollination: one towards traits associated with moth pollination, and one towards traits associated with pollination by long-billed Malachite sunbirds. The latter shift coincided with the colonization of Namaqualand in which Orange-breasted sunbirds are absent.

is characterized by three pollination ecotypes, but only the evolutionary transition from short- to long-billed sunbird pollination can be clearly explained by the Grant–Stebbins model. Corolla length is a key character for both ecotype transitions, while floral scent emission was important for the transition from bird to moth pollination.