One of the fathers of pollination biology, Hermann Müller, clearly saw the danger of studying pollination without knowledge of insect biology, as the quotation at the head of this chapter shows. Unfortunately, Müller's (1873) warning remains apropos. Relatively few of the systems that have served as models in pollination biology have been studied with equal emphasis on botany and zoology. Too many studies include only a superficial treatment of pollinator behavior, relying instead on longstanding truisms about what the animals do, and why. Pollination biology will benefit greatly if it can replace this casual approach with a tradition that is more rigorous and includes quantitative and experimental study of behavior where appropriate, along with an appreciation of recent advances in behavior and cognitive biology of insects and other pollinators.
With regard to the ethological isolation paradigm, it will help to step back and examine the central assumptions of specialization and flower constancy. A reasonable place to begin is the theory of foraging behavior (see Pyke 1984; Stephens & Krebs 1986; Parker & Maynard Smith 1990). In particular, the theory of optimal diets considers a forager searching an environment that contains food items of different values for fitness. The value or "utility" of each food item is taken as its expected mean return in terms of calories or some other nutrient; the variance around this expectation also may constitute part of the value (Kacelnik & Bateson 1996; Smallwood 1996). The forager is assumed to have accurate knowledge of the values of different items and the search costs associated with them. The optimal behavior is to specialize on the most valuable item if this is encountered sufficiently often, but if not, to expand the diet to include the second-ranked item. Depending on the rate of encounter with first-and second-ranked items, the diet will be expanded further to the third-ranked item, and so on. These predictions can be modified for conspicuous (rather than cryptic) food items, in which case the forager should choose a path through the environment that yields a maximum rate of caloric or nutrient return. Depending on the spatial arrangement of food items this optimal path may involve visits to a single item or a mixture (Mitchell 1989).
This summary shows that a continuum from complete specialization to complete generalization can be expected of the same animal, depending on properties of the "prey" community. This conclusion assumes that constraints of cognition (and of physiology or morphology) are not absolute. Available evidence suggests that this is a reasonable assumption for pollinating animals (Waser et al. 1996). Even specialized oligolectic solitary bees appear usually to be constrained to particular plant species only in terms of pollen collection (due apparently to digestive physiology of larvae), and not in terms of nectar collection, which they tend to undertake at many flowers. Hence it is not surprising to find that pollinators such as nectar-collecting bees often behave in reasonable agreement with predictions of first-generation diet models (Pyke 1984, and references therein). Still, it has become clear that the models are incomplete, for example because foraging behavior is often moderately (if not absolutely) constrained by features of pollinator morphology, physiology, or cognition (for more details see Weiss, this volume; Menzel, this volume; Dukas, this volume). The behavior of flower constancy, properly defined as a propensity to visit the same type of flower as last visited irrespective of the value of alternatives, in particular suggests cognitive constraints such as those involving memory retrieval (for further development see Chittka et al. 1999). Several second-generation foraging models incorporate such constraints (e.g., Hughes 1979; McNair 1981; Dukas & Ellner 1993; Kunin & Iwasa 1996). These models predict for example that constancy will break down at low flower density.
The discussion so far concerns small temporal and spatial scales, e.g., a single foraging bout or single day or single meadow. What is predicted on larger scales.? Generalization is predicted as an optimal strategy so long as the abundances of preferred plant species fluctuate sufficiently in time and space (Waser et al. 1996). For example, if the pollinator lifespan exceeds the flowering of a single plant species, generalization must ensue over an individual's lifetime. Similarly, if the pollinator's habitat affinities and geographic range incompletely overlap those of a single plant species, generalization must occur across pollinator populations (e.g., Herrera 1988, 1995).
Several conclusions follow. We should be unsurprised to find that many pollinators are opportunistic and generalized in choice of flowers; selection has molded pollinator physiologies and behaviors in floral environments that often vary substantially in time and space, where strict specialization may be disadvantageous. Conversely, an observation of short-term specialization does not necessarily indicate a morphological, physiological, or cognitive constraint. A strong constraint is indicated if the specialization is fixed throughout the lifetime of a forager and across individuals. In addition, the intriguing behavior of flower constancy is not predicted of an unconstrained forager, and should not be absolute even in a constrained forager, but rather should depend on ecological context. The overall conclusion from animal behavior is this: oligolecty and constancy are not necessary directional outcomes of evolution. Any view of these as traits of "advanced" pollinators (e.g., Crepet 1984) should be replaced by curiosity about features of the immediate floral environment, and of past environments, that affect the behavioral machinery of foraging animals to cause generalization vs. various forms of specialization (Waser 1986; Waser & Price 1998; Wcislo & Cane 1996).
Nature is the final arbiter; what do we see from direct observation of foraging pollinators. The evidence to date is that generalization and opportunism are common in pollinating insects of several orders (most studies have focused on Hymenoptera, and on Lepidoptera; Weiss, this volume) and in pollinating vertebrates (such as birds). To the references assembled by Waser et al. (1996), limited space allows me to add only a few recent ones. Memmott (1999) showed that the interactions between flowers and insect visitors in a British meadow form a highly connected web. Although some insects are specialist visitors, the plants they visit often attract more generalized insects as well, and these often connect the plants to the rest of the web (see also Jordano 1987). Careful study by Cotton (1998) showed that hermit hummingbirds are as generalized as non-hermits, contrary to the usual assumption. Momose et al. (1998) presented an impressive community-wide analysis of pollination in a lowland tropical forest, showing that many pollinators are generalized and opportunistic in flower use. Fleming & Holland (1998) found that the obligate mutualism between senita cactus and senita moth in northern México is supplemented by pollination from generalist solitary bees. Similarly, Kwak & Velterop (1997) documented pollination by generalists to an endangered plant species in Holland and France, in addition to pollination by a specialist bee; they also showed that the faunal composition of generalists changed through time and space. Olsen (1997) showed that the most effective pollinators of a native composite in Texas (those producing the most seed from a single visit) were among the least abundant of 10 different pollinators, seemingly at odds with Stebbins's (1970) "most effective pollinator" principle. Finally, Leebens-Mack et al. (1998) found that yucca moths are not so specialized to two sympatric yucca species as to prevent hybridization between them (see also Webber 1960).
This last example brings us directly back to the topic of the chapter. Even a relatively small proportion of incompletely constant generalist pollinators may begin to genetically connect related plant species that grow in sympatry (compare May & Anderson 1987), so long as the species are otherwise interfertile. The transition from an unconnected set of plant-pollinator interactions to a web of genetic connection may be nonlinear (compare Green 1994). Such genetic connection will hinder progress toward complete sympatric speciation of taxa that otherwise might so diverge. Similarly, if secondary contact occurs between sister species that are newly diverged in parapatry or allopatry, increasing gene flow from generalist pollinators may foster hybridization and possibly fusion into a single species (e.g., Arnold 1997).
Was this article helpful?