Enthusiasm for optimal foraging theory in the 1970s and 1980s stimulated much work on foraging by bees and, to a lesser extent, hummingbirds. These animals were assumed to be energetically stressed because they were nectarivorous, small, and dependent on costly forms of flight. Researchers sought to explain foraging in terms of movement patterns that saved energy. For example, Pyke (1981) derived movement rules both within and between inflorescences. He compared observed directions and distances of movements following departure from a flower to optimal predictions, under the assumption that the animals should maximize their net rate of energy intake. Pyke (and many others) also assumed that animals would have imperfect knowledge about their environments, particularly with regard to predictions as to what and where to find food in the future. In addition to using statistical rules (Pyke 1984), such animals should always sample in order to track an ever-changing world. Despite the power and appeal of this viewpoint, we now see growing evidence that simple rules and patterns alone cannot explain foraging in hummingbirds. Here, we review how learning and memory influence hummingbird foraging and how memory might affect the ways in which hummingbirds pollinate plants.
Much of a hummingbird's diet is derived from the nectar of flowers that, in turn, rely on hummingbirds for pollination. These flowers frequently provide only a few mg of sugar daily (Kodric-Brown & Brown 1978). Hummingbirds therefore must visit many flowers on each feeding bout, transferring pollen among flowers in the process. It is claimed that replenishment of this nectar, when it occurs, requires several hours (Armstrong et al. 1987). If this is the case, a hummingbird that is foraging efficiently would do well to use a strategy for avoiding those flowers it has recently emptied.
Given that there is no evidence that hummingbirds detect empty flowers before visiting (see below), either by olfactory cues or by the sight of nectar, there are several types of strategy by which a bird might make decisions about which flowers (or inflorescences, plants, etc.) to visit: (1) move to new flowers following simple decision rules based on immediate past experience, the present flowers being visited, and what can be seen from the present site; (2) visit flowers based on their visual characteristics; or (3) visit flowers based on memories of the locations of individual flowers and their physical attributes (such as color or nectar). This last strategy includes combining rules and memory in a systematic fashion (e.g., traplining; Feinsinger 1978). We shall address each of these in turn.
The application of optimal foraging theory (OFT) to investigating hummingbird foraging followed on the heels of the apparent success of applying such modeling to data on bee foraging (see Heinrich 1983). Bumble bees tend to arrive at the bottom of an inflorescence and move upwards before flying to another. Starting at the bottom might be an optimal strategy, because many vertical inflorescences display a gradient of nectar, with highest standing crops at the bottom (Pyke 1978a). Upward movement might not always represent an optimal response to nectar gradients, however. First, nectar gradients are not universal (see Corbet et ol. 1981). Second, Waddington & Heinrich (1979) found it difficult to teach bees to reverse the direction of searching from a bottom-to-top to a top-to-bottom pattern. Heinrich (1983) suggests that flower morphology may also play a part in this searching strategy. Because flowers on inflorescences often hang downwards, it would be more efficient for the bee to move up the inflorescence while foraging. In addition, moving in one direction only, either up or down an inflorescence, circumvents revisiting flowers that the bee has just emptied. So, foraging by bees on inflorescences can be described using movement rules - is this true for hummingbirds also.? In the flower mostly commonly used in foraging observations and experimental tests of hummingbirds foraging, scarlet gilia (Ipomopsis oggregoto), Pyke (1978b) found no evidence that hummingbirds (broad-tailed (Selosphorus plotycercus) and rufous (S. rufUs)) work systematically up or down an inflorescence (see also Hainsworth et ol. 1983). At this stage, then, there is no evidence that hummingbirds forage within an inflorescence using movement rules like those ofbumble bees.
Hummingbirds' movements between inflorescences have also been examined. Whereas some animals maintain a general directionality during a foraging bout (e.g., goldfish [Carassius auratus], Kleerekoper et al. 1970; European thrushes [Turdus spp.], Smith 1974), neither Pyke (1978b) nor Wolf & Hainsworth (1990) found evidence that this was how hummingbirds (broad-tailed and rufous) moved between inflorescences. Rather, a feeding hummingbird appeared to choose the next inflorescence based upon a visual impression of its proximity and/or its size. In addition, birds tended to fly further if the inflorescence on which they had just fed was oflow quality. Although Wolf & Hainsworth (1990) found that the hummingbirds were foraging neither randomly nor by following simple movement rules, they concluded "that the birds have relatively little information about the rewards in an inflorescence before a visit." However, these results seem at odds with evidence (Miller & Miller 1971; Gass & Sutherland 1985) that hummingbirds can remember good feeding locations. Wolf & Hainsworth (1990) point out that foraging strategy may vary with the spatial scale at which foraging is being examined. For example, one might describe multiple visits among flowers in an inflorescence, among inflorescences within plants, or among clumps of plants.
It is still a popularly held belief that hummingbirds have an innate preference for red flowers. Whereas considerable evidence suggests that hummingbirds exhibit color preferences, no data are available to determine whether this preference is innate (as has been demonstrated in butterflies by Weiss 1997). For example, Lyerly et al. (1950) found a significant avoidance of yellow feeders by a single Mexican violet-eared hummingbird (Calibri t. thalassinus). However, as early as 1941, Bené provided evidence that learning played a major role in flower color preference and that hummingbirds (black-chinned [Archilochus alexandri]) could be trained to visit specific food sources irrespective of their color (Bené 1941, 1945). Later investigations have supported this role of learning of color preferences in other species (e.g., Anna [Archilochus anna], Collias & Collias 1968; ruby-throated [Archilochus colubris], Miller & Miller 1971; rufous, Miller et al.
The possibility that a preference for red may arise from humming birds having a high visual sensitivity to red and low sensitivity to blue (Graenicher 1910) was convincingly refuted by Goldsmith & Goldsmith (1979), who demonstrated that black-chinned hummingbirds (Archilochus alexandri) learned to visit feeders lit by green light (546 nm) as quickly as they learned to visit feeders lit by red light (620 nm). They also showed that two different and opposing color associations could be learned simultaneously and that following experience with red feeders, the birds tended to visit red and blue (490 nm) feeders in preference to green and yellow (560 nm) feeders. Goldsmith & Goldsmith (1979) suggested that red is the least likely color to attract potential hymenopteran pollinators (although see Chittka & Waser 1997; Chittka, this volume). Thus, reduced interspecific competition between hummingbirds and the Hymenoptera may increase the relative value of red flowers and hummingbirds eventually learn this association. In addition, for hummingbirds at least, red flowers may offer a striking visual contrast against a green foliage background and may be the most conspicuous of flowers (relative to their size). This possibility has not yet been tested. Because many other floral attributes are correlated with the type of pollinator(s) plants attract, plants that offer their nectar and pollen in red flowers will also have nectar concentrations, nectar amounts, and size and shapes of flowers that differ from those of plants whose flowers are not red (Faegri & van der Pijl 1971; Thomson etal. 2000; but see Waser etal. 1996).
In hummingbirds, color preferences are learned associations between food sources and nectar amounts or concentrations (e.g., Stiles 1976; George 1980; Melendez-Ackerman et al. 1997). In the field (at least in North America), this typically leads to hummingbirds feeding preferentially on red flowers. Indeed, nearly all of the Californian flora that appear to be specially adapted for pollination by hummingbirds have red flowers. These flowers are often long, tubular, and odorless, provide relatively large amounts of nectar, and have anthers and stigma positioned such that a hummingbird is that plant's most effective pollinator (e.g., Grant 1966).
Melendez-Ackerman & Campbell (1998) attempted to dissociate morphological and color cues offered by Ipomopsis aggregata (red), I. tenui-tuba (whitish, longer, and more slender than I. aggregata ), and a hybrid of the two (intermediate in color [pink] and flower form) to foraging broad-tailed Selasphorus platycercus and S. rufous. In one experiment, they painted I. aggregata flowers to match the colors of the three plant types; birds visited red flowers more than pink or white flowers. In another experi ment, flowers of all three plant types were painted a standard red; birds showed no preference. Leaving aside possible criticisms of such painting techniques (e.g., Bennett et al. 1994), these manipulations appeared to show that the changing of flower color alone affected visitation rates to the three flower types, with the birds visiting red flowers in preference to pink or white flowers. In the first set of experiments, birds could not be using the relationship between flower color and nectar reward, as all the flowers were from the same plant. Therefore, they must have chosen flowers based on a previously learned association between color and reward. Despite some evidence that hummingbirds will extract more nectar from artificial flowers with wide corollas (Grant & Temeles 1992), birds in Melendez-Ackerman & Campbell's (1998) experiment did not choose flowers based on morphology alone (the second experiment). It may be that such features are far less conspicuous than color; with flowers 0.5 m apart, as presented in these experiments, the bird might as well probe the flower once it has chosen to fly close enough to assess its morphology (Ipomopsis aggregata, with a wider corolla than the two alternatives, produces an average of 1.8 ^l nectar per day, while I. tenuituba and the hybrid produce about 0.25 ^l nectar per day). At least under the experimental conditions, flower color seems to explain the birds' choices. However, the experiments investigating color preferences and the role of learning in the development of preferences have shown that red flowers need to reinforce their possibly more conspicuous signal with a greater reward or easier access (e.g., via a wider corolla).
In summary, hummingbirds appear to associate color and reward, and are able to change their flower visit accordingly. However, learned preferences will persist if novel flowers tend to match the previously learned associations. Therefore, in a region in which hummingbirds have learned to associate red flowers with greater reward, novel flowers should be red in order to benefit both from high conspicuousness and from birds generalizing the learned association of red flowers and reward. This may be why the Californian flora pollinated by hummingbirds is dominated by red flowers. However, unless red flowers are also more conspicuous than other flower colors in other environments, such a relationship between flower color and hummingbird pollination need not exist. To determine whether generalization of learning has influenced floral features in this way, we need to know more about which flowers are pollinated exclusively by hummingbirds, the visual background in which they are found, and the local floral diversity.
Certain male hummingbirds defend territories of hundreds or perhaps thousands of flowers (Kodric-Brown & Brown 1978; Paton & Carpenter 1984; Armstrong et al. 1987). These birds may protect nectar resources by emptying the flowers on the edges of the territory early in the day, then moving toward the center of the territory as the day progresses (exploitation defense; Paton & Carpenter 1984). In order to do this the hummingbird must at least remember which flowers or patches of flowers he has emptied most recently. There are several different spatial scales at which the hummingbird might keep track of flowers he has emptied. A bird might remember: (1) patterns of movement around his territory; (2) broad areas of flowers (e.g., the southwest corner of his territory); or (3) specific flowers. The last of these has rarely been considered, possibly because such a memory capacity seems extraordinary.
The first possibility, that a bird might remember a pattern of movement around his territory, seems well within the abilities of a hummingbird, because some birds do this on a much larger scale than a territory. The lekking hummingbird species (e.g., the hermits, Trochilidae, Phaethornithinae) are thought to trapline, i.e., to visit isolated, undefended nectar sources in a regular fashion (Feinsinger & Colwell 1978; Gill 1988; Garrison & Gass 1999). To do this, the bird must remember the sequence of nectar sources (by remembering a pattern of movement and not each of the nectar sources separately) and which was the last source he visited. Despite general acceptance that some hummingbirds do forage in this manner, there are no detailed maps of routes: spatial traplining is inferred from noting regular reappearances of marked birds, not by following them (cf. Thomson, this volume). This kind of systematic foraging would be the simplest of the three memory tasks above; it seems possible, given the evidence supporting the slightly more demanding proposal, that hummingbirds can remember patches of flowers in order to avoid them for several hours (e.g., Gass & Sutherland 1985; Wolf & Hainsworth 1991). Whether or not the birds use sequences of vectors, landmark memories, or both combined is not yet known (see Chittka et al. 1995 for an experimental test on bees).
Given that hummingbirds appear to exhibit some sort of spatial memory, much initial research focused upon the nature of learning, rather than the capacity or function of memory in natural foraging. Laboratory tests have been usefully employed to explore the nature of spatial learning in North American species of hummingbird (black-chinned, Rivoli [now magnificent, Eugenes fulgens], and blue-throated [Lampornis clemenciae], Cole et al. 1982; rufous, Brown & Gass 1993; Brown 1994). Cole et al. (1982) found that the hummingbirds they tested (males and females) learned a "shift" task more quickly than a "stay" task. In the former, birds had to choose the new feeder of a pair; in the stay task, they had to return to the familiar feeder. Whether this difference in learning rate is based on innate biases or on experience gained from the field, the birds' responses make sense if they are accustomed to foraging on flowers that are depleted in a single visit.
Cole et al. (1982) suggested that stay learning might be easier to demonstrate in an experiment and easier for birds to learn, if the rewarding locations were patches of inflorescences rather than single flowers, because each visit does not produce substantial depletion. There has been, to our knowledge, no test of this suggestion. However, we have carried out two field experiments that indirectly bear on this issue. In the first, birds fed from a single artificial flower. In the first block of trials, the flower contained too much sucrose to deplete in a single visit; in the second block of trials, the flower contained 70 ^l of sucrose, an amount birds take in a single visit. On their return, birds were presented with two flowers, 40 cm apart - the original flower and an alternative flower that differed from the original either in color or pattern. In some trials, the original flower stayed in its original location; in other trials, the alternative flower occupied the original position. When the alternative differed in color/pattern from the original flower, birds chose flowers apparently at random whether or not the original flower had been depleted in the first visit. Birds avoided the location of the original flower if it had been depleted, but did not show this avoidance behavior if the flower had not been depleted (see Fig. 7.1). In the second experiment, we attempted to increase the likelihood that birds would return to the original flower by allowing birds to visit a flower three times before offering an alternative. Birds were somewhat more likely to avoid the flower in the same location after three visits but were more likely to return to a flower bearing the same color/pattern as the flower they had visited three times. Therefore, it is difficult to teach male rufous to use a "stay" rule during foraging even if the "flowers" do not deplete. Of course, these birds can learn to stay, as is shown by their enthusiasm for using hummingbird feeders as well as the anecdotes describing birds returning from migration and hovering at the place the feeder had been hung the previous year.
Was this article helpful?