Bumble bee workers under controlled conditions unequivocally forage in frequency-dependent ways in response to variation in corolla color, showing a common-morph preference when flowers contain rewards but a rare-morph preference when they do not. Field experiments, however, showed that pollinators preferred common corolla morphs in only one of four rewarding species. No significant correlations between morph frequency and relative reproductive success have been found in rewarding plant species, although there are suggestions of a negative relationship for a study on an unrewarding species. How may the contradictions in these results be explained.?
Does pollinator behavior in the field differ from that observed in the laboratory?
One potential cause of a difference between laboratory and field behavior is that the frequency dependence observed in bumble bees is not found in other pollinator types. However, FDS has been demonstrated in many animals, including birds (Allen 1988) and a range of predatory, parasitic, and herbivorous arthropods (Sherratt & Harvey 1993). Although further experiments are required, given that frequency-dependent behavior is widespread and common, many polylectic pollinators may respond to variation in floral traits such as color in a frequency-dependent way.
Laboratory experiments typically use highly distinct color morphs. It could be argued that natural floral polymorphisms are often less distinctive to the eyes of their pollinators, and thus induce weaker frequency-dependence. Alternatively, perhaps overall preference for a morph is so strong that frequency dependence is difficult to detect. Pollinators might not respond to floral traits other than color in a frequency dependent way. In experiments using three color morphs of rewarding artificial flowers, and testing for FDS between two that are relatively similar to the eyes of bumble bees, Smithson & Macnair (1997&) still detected significant positive FDS, but this was weaker than shown on yellow and blue flowers that contrasted more strongly (Table 12.1). Overall bias is important in determining the frequency at which preference switches in favoring one morph over another (Fig. 12.1, Table 12.1), so if this bias is strong, FDS may be detectable only at more extreme morph frequencies than those used in field experiments. Further experiments are required to test pollinators' responses to other types of floral traits.
It could also be argued that the sample sizes in the above studies may be too small. If pollinators in the wild forage as they do under laboratory conditions, we expect strong preferences, which should have been detectable at the sample sizes used. However, it is certainly noteworthy that experiments with significant effects are usually the ones with the higher sample sizes!
It is possible that pollinators might react differently to natural polymorphisms because real morphs differ not only in color but also in other traits, due to pleiotropy. Pleiotropic effects of variation in corolla color have been demonstrated in several cases (e.g., Waser & Price 1981), but not in others (e.g., Jones 1996&). Clearly, for floral traits like corolla color, pleiotropic effects will depend on the biochemical pathways involved. Raguso (this volume) suggests that because floral scent compounds may be produced by the same biochemical pathways as some floral pigments, pleiotropy may be common in characters important to pollinators. Such pleiotropic effects on pollinator behavior need evaluation.
Do the predicted fitness relationships differ from those observed in the field.'
A classic example of FDS is Batesian mimicry in butterflies, in which the fitness of the unpalatable model is inversely related to its abundance relative to that of its palatable mimic (Turner 1977). Both predator and prey are mobile, and the predator either consumes the prey totally or it does not. Contrast this with the type of selection that pollinators exert on immobile plants. The constant spatial remixing of available types does not occur as it does in butterflies. Pollinators consume only part of a plant; a visit usually depletes, but rarely fully consumes, the nectar or pollen. Nectar reward may be replenished by the plant, but slowly. Reproductive success is not an absolute form of selection, and pollinators may affect many aspects of overall fitness. I argue that these fundamental differences between predator-prey and plant-pollinator systems may cause a substantial difference in the frequency dependence expected in plant-pollinator systems.
First, if pollinator visitation levels are high, pollinators will deplete the rewards. Laboratory experiments show that varying the relative amounts of nectar present in two morphs of flowers has a weaker effect on frequency-dependent behavior by bumble bees than changing availability through presence or absence of nectar (Table 12.1). This suggests that differences in reward production between two morphs will affect FDS weakly as long as flowers are not emptied by pollinators, i.e., visitation rates are low. As soon as some flowers are empty, preference patterns will change. Depletion of one morph would lead to strong frequency-independent selection, and depletion of both morphs may to lead to negative FDS. Thus, selection patterns will be expected to vary both temporally and spatially according to factors that determine pollinator abundance and overall visitation rates to a species.
Second, as noted above, if morph frequencies are constant, a preference for common corolla colors develops initially over 100 flower visits and changes little subsequently. However, if morph frequencies of rewarding species fluctuate from patch to patch, patch-hopping pollinators may develop common-morph preference only weakly because of a lag in pollinator response to the morph frequency in the current patch. Alternatively, pollinator preference might be strongly influenced by the morph frequency in the first patch visited. Color morphs are patchily distributed within plant populations (Epperson & Clegg 1986), potentially making FDS much weaker than expected.
Third, the number of flowers visited per inflorescence increases with visitation rate, both for rewarding and unrewarding inflorescences. This relationship leads to the expectation that, while selection by visitation rate alone may be positively frequency-dependent for rewarding species, selection through selfing rates and outcross male and female reproductive success may be negatively frequency-dependent. Final reproductive success may not be predicted by visitation rates alone.
For rewarding plants, positive FDS is thus not necessarily expected, either for pollinator behavior or for selection in plant populations. Furthermore, results for different fitness indices may conflict. Pollinator response to the frequency of different plant species, as opposed to different morphs, also suggests that FDS may go in different directions, depending on the fitness index considered (Stout et al. 1998). Pollinator abundance is likely to be crucial, however, in determining the likelihood of positive FDS - it is more likely if pollinators are rare, because visitation rate will be a more important variable than the number of flowers visited per inflorescence in determining fitness.
Things are different for plants that present neither nectar or pollen to visitors. These receive few visits. Pollinators soon locate alternative rewards (Nilsson 1980), thereby limiting the reproductive success of the unrewarding plants (Gill 1989). Visits also do not change the reward status of unrewarding plants. I expect pollinator preferences for unrewarding plants to be independent of variation in morph frequency from patch to patch. Because the flowers are empty, pollinators should move long distances between visits, so they will not stay long enough in a patch to detect shifts in morph frequency (Pyke 1978). Also, when visitors sample only a few flowers per plant, there is little scope for the number of flowers visited per inflorescence to differ between morphs. In populations of unrewarding plant species without mimicry, therefore, negative FDS should act straightforwardly on floral variants for such traits as color.
What can we predict about the evolutionary dynamics
Was this article helpful?