Résumé
In this study we quantify the relative rates and amounts of between-morph (legitimate) cross-pollination, within-morph (illegitimate) cross-pollination and self-pollination in a natural population of distylous Gaertnera vaginata (Rubiaceae) endemic to the oceanic island of La Réunion in the Indian Ocean. Since pollen size distributions of each morph overlap we develop a method, based on chi-squared comparisons of observed to expected pollen size distributions, to quantify pollen transfer proficiencies. This method, unlike previous methods, does not assume an equal chance of pollen transfer for pollen grains in the region of overlap. The probability that a single pollen grain from a short-styled plant undergoes legitimate pollen transfer to long-styled flowers (4.19×10−2) was more than twice that for illegitimate transfer to other short-styled plants (1.56×10−2). In contrast, the probability that a single pollen grain from a long-styled flower undergoes legitimate pollen transfer to short-styled flowers (1.65×10−2) was no greater than that of illegitimate pollen transfer to other long-styled plants (2.15×10−2). The transfer proficiencies of individual pollen grains are thus highly asymmetric. However, because long-styled flowers produce significantly more pollen, the ratio of legitimate to illegitimate pollen on the stigmas of short-styled flowers (1.85) markedly exceeded that for long-styled flowers (1.11). Short-styled flowers incurred significantly higher levels of self-pollination than long-styled plants, emphasizing the importance for self-pollination to be removed before comparing stigmatic pollen loads. The method we propose could be applied to other distylous species with pollen size overlap between morphs and thus allow for a more widespread study of the pollination ecology of heterostylous species.