Pollination syndromes predict the principal animal pollinators of plant species from floral traits. However, the existence of discrete pollination syndromes and the relevance of these syndromes for the effects of animal pollinators on plant reproduction are controversial and a quantitative, trait-based approach may refine the concept of pollination syndromes. Thus far, however, very few studies have simultaneously analysed the importance of signalling, accessibility and reward traits of flowers for the effects of animal pollinators on plant reproduction of multiple species. The genus Protea is an ideal study system because the members of this genus exhibit remarkable variation in functional traits and Protea species are visited by several guilds of animal pollinators, such as birds and insects. We experimentally investigated seven Protea species from the South African fynbos to assess variability in reproduction (seed set, seed mass, germination) in response to pollinator exclusion (full-exclusion, bird-exclusion) and hand-pollination (self-pollen, cross-pollen). To generalise the findings of the experiment, we conducted a meta-analysis of the effects of bird-exclusion on seed set of 14 Protea species, associating effect sizes of bird-pollination with signalling (inflorescence length and diameter), accessibility (inflorescence and style length) and reward traits (length of pollen presenter, nectar volume and sugar concentration). All Protea species were self-compatible, but only two species were to a small degree capable of autonomous selfing. Protea species varied in their dependence on animal pollinators, including species depending exclusively on birds, insects, or both for seed production. Protea species exhibited continuous variation in the effects of bird-pollinators on reproduction, challenging the accuracy of categorical pollination syndromes for predicting pollinator effects on plant reproduction. The meta-analysis showed that the quality of floral rewards, especially nectar sugar concentration, predicts the dependence of Protea reproduction on bird pollinators, whereas the signalling and accessibility traits tested in this study were not important. Beyond revealing new insights into the pollination biology of Protea species, our findings underscore the usefulness of experimental approaches on multiple plant species to test the effects of pollinator guilds on plant reproduction and contribute to a trait-based understanding of the functional determinants of plant-pollinator interactions.
