Multidimensional scaling dataData used for cluster,non metric multidimensional scaling and mantel test correlation analysis testing for an association between visitor community and floral phenotype.Multidimensional scaling.xlsxUnivariate regression dataData used for univariate regression analysis, testing for a relationship between style length and functional visitor length.Univariate Regression.xlsxNested ANOVA: floral traitsData used in nested ANOVA analysis testing for differences in floral traits between long and short style phenotypesFloral traits_nested ANOVA.xlsxReciprocal translocation dataData from reciprocal translocation experiments testing for local adaptation in long and short style phenotypes of Nerine humilis.Reciprocal translocation.xlsxVisitor preference dataData from choice experiments conducted during reciprocal translocations. This data was used to test whether local adaptation is a consequence of the sensorial fit between flowers and visitors.Visitor Preferences.xlsxSingle visitations: contactContact data from single visitations used to test whether local adaptation is a consequence of the mechanical fit between flower and visitor morphology.Single visitations_contact.xlsxSingle visitation: pollen depositionPollen deposition data for single visitations used to test whether local adaptation is a consequence of the mechanical fit between flower and visitor morphology.Single visitation_Pollen deposition.xlsxSingle visitation: seedsetSeedset data from single visitations used to test whether local adaptation is a consequence of the mechanical fit between flower and visitor morphology.Single visitation_Seedset.xlsxSupplementary material (controls)Data used in controls (Supplementary material)

Geographic variation in floral morphology is often assumed to reflect geographic variation in pollinator communities and associated divergence in selective pressures. We studied populations of Nerine humilis (Amaryllidaceae) to assess whether geographic variation in floral form is the result of local adaptation to different pollinator communities. We first tested for associations between floral traits and visitor communities, and found that populations with similar floral morphologies were visited by similar insect communities. Mean style length in each population was also closely associated with the mean body length of the local visitor community. A reciprocal translocation experiment demonstrated that native phenotypes set more seed than translocated phenotypes. Single visitation experiments showed that native flowers received more pollen, and set more seed per visit, than introduced phenotypes in both populations. This suggests that the effectiveness of pollinator visits is determined by the degree of mechanical fit between flowers and visitors. We provide strong evidence that the observed among-population variation in floral traits is an adaptive response to geographic variation in the pollinator community.