Although bumble bees may show frequency dependence on simple artificial flowers, they might not exert frequency-dependent selection in real populations. First, patterns of behavior might be affected by the organization of flowers into inflorescences: encountering many flowers clustered together could affect a bee's perception of frequency. Second, even if visitation rates are well correlated with female function through seed set (Waser & Price 1981), they may be weakly coupled to other components of fitness such as outcrossing through male function (Stanton et al. 1986, 1989). In particular, if pollinators visit more flowers per inflorescence on one morph, more pollen will be transferred between flowers within inflorescences of that species (Klinkhamer & de Jong 1993; Harder et al., this volume). This will cause geitonogamous selfing in self-compatible species, and may block receipt of compatible pollen in self-incompatible species; it will also reduce the amount of pollen available for outcross seed paternity (de Jong et al. 1993; Harder & Barrett 1995). Thus, differences in the numbers of flowers visited per inflorescence between morphs can affect final reproductive success.
Artificial inflorescence experiments attempt to bridge the gap between bee-board experiments and pollinator behavior in the field; some have been used to test optimality models (Cartar & Abrahams 1996). We used artificial inflorescences to test both for frequency dependence and differences in the numbers of flowers visited per inflorescence between morphs (A. Smithson & L. Gigord, unpublished data). We made artificial inflorescences from green plastic rods 40 cm long and 1 cm in diameter, the top of which held 10 "flowers" arranged spirally around the rod 1.5 cm apart. Flowers were colored card stock "corollas," with central holes that gave access to wells inside the rod, into which we pipetted sucrose solution. We conducted two experiments in a cage, with inflorescences of purple and yellow arranged randomly on a grid at frequencies of 50 yellow:25 purple and 50 purple:25 yellow. Worker bumble bees (Bombus terrestris) from a captive colony foraged singly.
For the array with 50 yellow:25 purple, 7 out of 11 bees showed a preference for yellow. Overall results indicated a common-morph preference (G = 42.92, p < 0.001). For 50 purple:25 yellow, 3 out of 11 bees showed a preference for purple, while 4 showed a preference for yellow; heterogeneity between individuals caused deviations from expected ratios. This suggests common-morph preference combined with frequency-independent preference for yellow.
After removing data from revisited inflorescences, we found that, for most of the bees, morphs differed significantly in the number of flowers visited per inflorescence for both experiments, although overall the difference was significant only for the 50 yellow:25 purple array (F = 19.52, p < 0.001). The relationship between visitation rate and the mean number of flowers visited per inflorescence was strong, non-linear, and positive: a regression analysis explained 79% of the variance. The relationship did not differ across experiments or color, and was not influenced by revisiting empty inflorescences, as there was no difference between morphs in the proportion of revisits (A. Smithson & L. Gigord, unpublished data).
These results are consistent with a relationship between inflorescence learning and flower learning. If bumble bees interpret each flower as an individual learning event, then preference for a morph should also result in an increased number of flowers visited per inflorescence. In other unpublished experiments, we found a positive correlation between the number of flowers visited on unrewarding artificial inflorescences and with visitation rate. These results show that although pollinators may visit plants in a frequency-dependent way, fitness relationships are unlikely to be predicted by visitation rate alone. Other components of fitness may also depend on morph frequency.
Was this article helpful?