- Email: [email protected]
Exploring the ‘Most Effective Pollinator Principle’ with Complex Flowers: Bumblebees and Ipomopsis aggregata
Annals of Botany 88: 591±596, 2001 doi:10.1006/anbo.2001.1500, available online at http://www.idealibrary.com on
Exploring the `Most Eective Pollinator Principle' with Complex Flowers: Bumblebees and Ipomopsis aggregata M A R G A R E T M . M AY F I EL D {{, N I C KO L A S M . WA S E R *{} and MA RY V. P R I C E {} {Rocky Mountain Biological Laboratory, Crested Butte, CO 81224, USA, {Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA and }Department of Biology, University of California, Riverside, CA 92521, USA Received: 4 April 2001 Returned for revision: 11 May 2001 Accepted: 19 June 2001 Published electronically: 17 August 2001 The `most eective pollinator principle' implies that ¯oral characteristics often re¯ect adaptation to the pollinator that transfers the most pollen, through a combination of high rate of visitation to ¯owers and eective deposition of pollen during each visit. We looked for the expected positive correlation between quantity and quality of visits in Ipomopsis aggregata, whose red, tubular ¯owers are considered to be adapted to hummingbirds. Hummingbirds were indeed the most common ¯oral visitors in 5 years of observation. However, long-tongued bumblebees deposited on average three-times as much outcross pollen per visit to virgin ¯owers, and elicited four-times as much seed production, as did hummingbirds. Hence visitors that are relatively infrequent, and unexpected given the `pollination syndrome' of the plant, can be surprisingly good pollinators. One interpretation of this observation is that natural selection favours a specialized ¯oral morphology that excludes all but a single type of visitor, but that there are constraints on achieving this outcome. An alternative is that selection favours some degree of ¯oral generalization, but that ¯owers can retain features that adapt them to a particular type of pollinator in spite of this # 2001 Annals of Botany Company generalization. Key words: Bumblebees, ¯oral evolution, hummingbirds, Ipomopsis aggregata, pollination syndromes, pollinator eectiveness, scarlet gilia, seed set, specialization, temporal variation.
I N T RO D U C T I O N Angiosperm ¯owers are strikingly diverse. They vary in colour, scent, size, morphology, and in the type and amount of reward they oer as a bribe to pollinating animals (Proctor et al., 1996). Their morphology can also be complex. For example, some ¯owers sequester rewards for pollinators within narrow corolla tubes or spurs or behind barricades that are dicult to breach, present sex parts in highly speci®c locations, or exhibit bilateral symmetry or asymmetry. Such complex features are generally interpreted as adaptations for attracting and exploiting certain types of pollinators (e.g. large bees vs. moths vs. birds), and for excluding other types (Grant and Grant, 1965; Crepet, 1984). This view is codi®ed in the concept of `pollination syndromes', a dominant organizing theme in pollination biology (Faegri and van der Pijl, 1971; for recent commentary see Armbruster et al., 2000; Ollerton and Watts, 2000). Further, the syndrome concept implies that plant species that specialize on a given type of pollinator will converge on a suite of phenotypic features that adapts them to the morphology, sensory and nutritional physiology, and behaviour of this pollinator. Against this backdrop, ecologists in the ®eld observe that many complex ¯owers receive visits from several types of animals. Stebbins (1970, p. 318) proposed a resolution to this paradox of apparent phenotypic specialization in the * For correspondence at: Department of Biology, University of California, Riverside CA, 92521, USA. Fax 1 909 787 4286, e-mail [email protected]
0305-7364/01/100591+06 $35.00/00
face of ecological generalization by suggesting that ` . . . the characteristics of the ¯ower will be molded by those pollinators that visit it most frequently and eectively'. In this view, the contribution to ®tness by a given type of ¯ower visitor is broken down into two multiplicative components, the number of ¯ower visits and the amount of reproductively-eective (e.g. compatible) pollen transferred to a ¯ower per visit. These `quantity' and `quality' components of pollination (sensu Waser and Price, 1983; Herrera, 1987, 1989) together provide an estimate of overall pollinator eectiveness [sometimes called pollinator importance (Olsen, 1997) or pollinator eciency (Schemske and Horvitz, 1984)] for a given species of animal visiting ¯owers of a given species of plant. An extension of Stebbins' `most eective pollinator principle' is that the quantity and quality components of eectiveness will be positively related across dierent types of visitors to ¯owers of a given plant species. The logic is as follows: all else being equal, any heritable change in ¯oral expression which increases the quality of each visit by a frequent visitor, or conversely increases the frequency of visits by a pollinator of high per-visit quality, should spread by natural selection even if there is a trade o in the form of reduced eectiveness of other visitors. Indeed, it is possible to read Stebbins' (1970) original statement of the most eective pollinator principle as predicting precisely this process (Olsen, 1997; Aigner, 2001), and Waser et al. (1996) independently codi®ed it in a simple model of ¯oral evolution in a constant environment. # 2001 Annals of Botany Company
592
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
Numerous studies have now con®rmed Stebbins' (1970) implication that ¯ower visitors can vary substantially in their overall eectiveness (e.g. Primack and Silander, 1975; Schemske and Horvitz, 1984; Jennersten and Morse, 1991), although they do not always do so (e.g. Motten et al., 1981; Tepedino, 1981; Waser and Price, 1990). It seemed to us of value to take the additional step of testing the proposal that the quantity and quality components of eectiveness will be positively related, using a plant species whose complex ¯owers clearly conform to one of the more striking pollination syndromes. To this end we measured pollination eectiveness of bumblebees visiting Ipomopsis aggregata (Pursh) V. Grant, a quintessential North American `hummingbird ¯ower', and compared this to the eectiveness of hummingbirds, which are predominant visitors in most years. Our results indicate that bumblebees are surprisingly good pollinators of I. aggregata, and we suggest some new ways of looking at the most eective pollinator principle and ¯oral adaptation.
M AT E R I A L S A N D M E T H O D S The study system In western North America, numerous unrelated plant species have repeatedly evolved a suite of ¯oral traits characteristic of the hummingbird pollination syndrome: relatively abundant but dilute nectar concealed within deep tubes, bright (usually reddish) colouration, lack of distinct landing platform, and absence of strong scent (Grant and Grant, 1968). A typical example is the scarlet ( skyrocket) gilia, Ipomopsis aggregata (Polemoniaceae), a monocarpic herbaceous perennial found in montane regions from southern Arizona, New Mexico and Texas to British Columbia (Grant and Wilken, 1986). At the Rocky Mountain Biological Laboratory (RMBL) in western Colorado, USA, plants grow for 2±7 years or more as vegetative rosettes, then die after producing between ten and several hundred red, tubular, protandrous, self-incompatible ¯owers. The predominant visitors to I. aggregata ¯owers at RMBL are broad-tailed and rufous hummingbirds (respectively Selasphorus platycercus Gmelin and S. rufus Gmelin). However, insects in several orders also visit, including solitary bees and syrphid and muscoid ¯ies that collect or eat pollen, and hawkmoths and butter¯ies that feed on nectar (Waser, 1982). Queens of Bombus appositus (Cresson) also visit I. aggregata ¯owers in some summers. With the longest tongues ( proboscides) of any bumblebee at RMBL, these animals obtain a good energetic reward when hummingbird visits are rare and nectar accumulates in the ¯owers, even though they cannot reach all the nectar ( for complete analysis see Pleasants and Waser, 1985). During early July 1997, B. appositus queens foraged in substantial numbers at I. aggregata. We observed them ¯ying systematically from ¯ower to ¯ower, landing on each ¯ower by grasping the re¯exed tips of the petals, and with proboscides extended pressing their heads into the opening of the corolla, where sex parts also are located (Fig. 1).
Per-visit eectiveness At the peak of bumblebee visitation in 1997, we exposed `virgin' ¯owers of I. aggregata to single visits by B. appositus, S. platycercus and S. rufus. Within a 4 m 4 m study plot we randomly chose ®ve large plants, each with 50±100 ¯owers open at a time. We removed all the ¯owers that were already open, carefully excised the immature anthers from elongated buds, and then placed a wire cage covered with ®ne nylon netting over each plant. These procedures ensured that stigmas would carry no pollen until visited by a pollinator. When emasculated ¯owers became femalereceptive 2±3 d later we removed the cages and waited for visitors. Observation over 6 d (10±12, 17±19 July), for 1±7 h per day between 0800 and 1800 h (total observation 30 h) was rewarded by 14 ¯ower visits by B. appositus queens, 13 visits by S. platycercus and 15 visits by S. rufus, all concentrated in the afternoon hours. During each of these visits, the animal foraged ®rst in a natural fashion on other plants outside of the experimental plot, then ¯ew to one or more experimental plants. Bumblebees visited fewer ¯owers per experimental plant (range 1±5; mean 2.3) than did hummingbirds (range 1±14; mean 5.6). After each pollinator had left the experimental plants we removed the corollas of the ¯owers visited so that they would not be revisited later, and continued observations. After each observation period we replaced the cages over plants. We returned 24 h after each ¯ower had been visited to excise its stigma. Stigmas were mounted on microscope slides in glycerine-gelatine medium stained with basic fuchsin
F I G . 1. A queen of B. appositus visiting and pollinating a red `hummingbird' ¯ower of I. aggregata.
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower' (Kearns and Inouye, 1993), and pollen loads counted at 100 magni®cation. Pollen tubes in I. aggregata grow to the ovary and fertilize ovules within 24 h (Waser and Price, 1991), so removal of stigmas did not aect fruit set, and we were able to return 3 to 4 weeks after a ¯ower had been visited to collect the resulting fruit and count mature seeds. Annual variation in rates of visitation The amount of pollen transferred to a virgin ¯ower after a single visit, and the resulting seed set, both estimate the quality component of pollinator eectiveness. In addition, we wished to estimate the quantity component, i.e. rate of visitation, to I. aggregata ¯owers by various pollinators. To this end we utilized three permanent study sites at the RMBL, including the site of the experiment described above. Sites were 170±550 m apart. For 5 weeks during peak ¯owering of I. aggregata in 1997, we recorded visits to 15±20 individually numbered plants at each site 3 d per week for 1.5 h each day between 0800 and 1800 h, for totals of 4.5 h per site per week. Open ¯owers on numbered plants were counted before each census so that results could be expressed as visits per ¯ower h ÿ1 by each species of visitor observed during a census. Results could then be compared to those from censuses carried out in the same fashion and at the same study sites in 1996, 1998, 1999 and 2000. R E S U LT S
Pollen load
Bumblebees were suprisingly good pollinators of I. aggregata. In a single visit to a virgin ¯ower they delivered on average three-times as much pollen to stigmas as did hummingbirds (respective means of 112.7 vs. 38.1 pollen grains; Fig. 2). This dierence was highly statistically signi®cant, whereas pollen delivery did not dier signi®cantly between the two species of hummingbird (Table 1). Following a single visit by B. appositus, 57 % of ¯owers produced expanded fruits, whereas fruit set was only 31 and 13 % following pollination by S. platycercus and S. rufus,
593
respectively. Again the dierence between bumblebees and hummingbirds was signi®cant, whereas that between the two species of hummingbird was not, based on categorical analysis (Table 1). Furthermore, expanded fruits resulting from bumblebee visits contained 1.7-times as many seeds per fruit as those resulting from hummingbird visits (respective means of 7.63 vs. 4.58 seeds). Although this dierence was not signi®cant, it is of much larger magnitude than the dierence between the two hummingbirds (Table 1). Combining the eects of dierent pollinators on the probability of fruit expansion and seed set per fruit, single visits by bumblebees yielded on average fourtimes the seed production of single visits by hummingbirds (respective grand means of 4.36 vs. 1.08 seeds). Bumblebees also were reasonably frequent visitors to I. aggretata ¯owers in summer 1997. Averaged across the entire ¯owering season, they visited ¯owers about a third to a quarter as frequently as hummingbirds (Table 2). Other insects, in particular syrphid ¯ies, also contributed a substantial minority of all visits. Because syrphids were foraging for pollen rather than nectar, and so avoided our experimentally emasculated ¯owers, we were unable to assess their contribution to pollination. DISCUSSION Several workers have reported recently on pollination interactions in which there is at best a weak relationship between quality and quantity components of pollinator eectiveness (e.g. Pettersson 1991; Kwak 1993; Olsen, 1997; GoÂmez and Zamora, 1999). Most such studies seem to T A B L E 1. Analyses of pollen delivery, fruit set and seed set resulting from single visits to I. aggregata ¯owers Source of variation in pollen load Pollinator type Bees vs. birds Broad-tails vs. rufous Error
SS
d.f
F
P
184.86 182.01 1.28 638.34
2 1 1 39
5.65 11.12 0.08
0.007 0.002 0.781
Chi-square
d.f.
P
1 2 1 1
0.031 0.060 0.025 0.273
150
Source of variation in fruit set
100
Intercept Pollinator type Bees vs. birds Broad-tails vs. rufous
4.66 5.62 5.04 1.20
Source of variation in seeds per fruit
SS
d.f.
F
P
Pollinator type Bees vs. birds Broad-tails vs. rufous Error
0.84 0.78 0.00 7.42
2 1 1 11
0.62 1.16 0.00
0.556 0.305 0.998
A 50
0
B. appositus
B
B
S. platycercus
S. rufus
Pollinator F I G . 2. Pollen delivery to virgin stigmas of ¯owers of I. aggregata (means + 1 s.e.) by queen bumblebees (B. appositus) vs. two species of hummingbird (S. platycercus and S. rufus). Dierent letters indicate signi®cantly dierent means.
The analyses of pollen load per stigma and of seed set per fruit are single-factor ANOVAs; that for seed set is based only on non-zero values. Raw values for pollen counts and seed set were square-root transformed to achieve normality and homoscedasticity of model residuals. The analysis of fruit set is a categorical analysis of the number of expanded vs. unexpanded fruits. In each case the eect of pollinator species is subdivided with orthogonal contrasts into eects of bumblebees vs. hummingbirds (`Bees vs. birds') and of broad-tailed vs. rufous hummingbirds (`Broad-tails vs. rufous').
594
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
T A B L E 2. Rates of visitation to I. aggregata ¯owers at three permanent study sites at the RMBL by hummingbirds, bumblebees (B. appositus queens) and other pollinators, across 5 years of study Mean visits per ¯ower h ÿ1 Year
Hummingbirds
Bumblebees
Other pollinators
Total
1996 1997 1998 1999 2000
0.0363 0.0145 0.0156 0.0266 0.0603
0.0001 0.0041 0 0.0017 0
0.0001 0.0077 0.0094 0.0072 0.0066
0.0365 0.0263 0.0250 0.0355 0.0669
The protocol for observation in all years was similar to that described in the text for 1997. The `other pollinators' comprised solitary bees and swallowtail butter¯ies in summer 1996, solitary bees, syrphid ¯ies and muscoid ¯ies in 1997, 1998 and 1999, and muscoid ¯ies in 2000.
involve relatively generalized ¯owers that do not clearly exhibit features of a pollination syndrome and that are visited only by insects. It is more perplexing to ®nd that insects were the most eective per visit in transferring pollen to the ¯owers of I. aggregata, considering that these ¯owers conform closely to a hummingbird pollination syndrome and hummingbirds are the most frequent visitors overall. The wide margin by which bumblebees outperformed hummingbirds in pollen transfer and in eliciting fruit and seed set is especially remarkable. Readers may object that our estimates of pollinator eectiveness are misleading. Any such estimate based on rate of visitation and per-visit pollen transfer is incomplete when pollinators dier in additional ways, as they may do (Waser, 1983). For example, I. aggregata ¯owers visited by bumblebees might receive less outcross pollen, and more self pollen from other ¯owers on the same plant (geitonogamy), compared to ¯owers visited by hummingbirds. Such a dierence would reduce seed set in nature (Waser and Price, 1991; de Jong et al., 1992), but would not show up in our experiments with emasculated ¯owers. However, our small sample (see Materials and Methods: per-visit eectiveness) suggests that on average B. appositus queens visited less than half as many ¯owers per plant as did S. platycercus and S. rufus; a large sample (Waser, 1982) con®rms this, and the dierence is statistically signi®cant. Thus bumblebees should transfer proportionately less geitonogamous pollen than hummingbirds. A similar objection is that bumblebees and hummingbirds might visit at dierent times of day, and pollen might be available for transfer only at certain times. However, our observations here, and earlier ones (N. Waser, unpubl. res.), indicate no obvious interspeci®c dierences in times of activity between about 0700 and 1900 h, with activity of both bumblebees and hummingbirds highest in the afternoon. Polleniferous male-phase ¯owers and receptive female-phase ¯owers are available in this period, because most new ¯owers of I. aggregata open (with anthers dehiscing pollen) in the afternoon, and most stigmas become receptive then as well (M. Price and N. Waser, unpubl. res.). Indeed, the pollen delivered to our experimental ¯owers is a
natural indicator of the pollen available to be exported from unmanipulated ¯owers that were visited by pollinators before they entered our experimental plot. Probably the most serious concern is that our measures of pollinator eectiveness miss subtle dierences in the export of pollen, i.e. in success of plants through the male sex function. This is a complex topic of which we venture a summary. Assume that individual pollinators visit I. aggregata for a period sucient for their pick-up and deposition of pollen at ¯owers to reach equilibrium. Due to `pollen carryover' (e.g. Price and Waser, 1982), pollen deposited on a female-phase ¯ower will come from some number of previously visited male-phase ¯owers, with the number likely to dier between bumblebees and hummingbirds due to dierent carryover properties. Nonetheless, if at equilibrium a bumblebeee deposits on average X pollen grains on a stigma and a hummingbird on average Y, then X and Y are the expected values of total pollen export for each male-phase ¯ower previously visited by the pollinator in question. In our system, X 110 pollen grains and Y 40, suggesting that bumblebee visits contribute more to male reproductive success, just as they do to female reproductive success. However, this analysis ignores the possibility that bumblebees pick up more pollen than hummingbirds in the ®rst visit to a male-phase ¯ower, and lose most of it (e.g. because they groom it from their bodies), yet still deposit more on female-phase ¯owers. If this is true, the critical issue is whether ¯owers can expect many visits during their male phase, in which case hummingbirds might export more pollen overall, because bumblebees would quickly remove all pollen but lose most of it. Here we can only oer the information that I. aggregata ¯owers remain in male phase for 1.5±2 d and receive only about one pollinator visit per day (Campbell et al., 1991; Waser and Price, 1991)Ð conditions which favour the interpretation that bumblebees are superior in overall pollen export. To go beyond this will require dicult experiments with captive bumblebees to determine the dynamics of their pollen transfer (e.g. Price and Waser, 1982; Thomson et al., 1986). We have not attempted such experiments. There is a possibility, therefore, that dierences in total pollen export from each ¯ower will render hummingbird visits of highest quality overall. However, this possibility seems unlikely, and we now proceed to discuss some implications of our original conclusion that visit quantity and quality are not positively related in our I. aggregata populations. What are the possible explanations for this in such complex ¯owers? It seems to us that one general possibility involves constraint in the design of ¯owers. If natural selection consistently favours a ¯oral morphology that limits the types of animals that can gain access to nectar or other rewards, it seems certain that some degree of exclusivity will be achieved. However, there may be limits to this process, in large part because the animals do not cooperate. Pollinators themselves are often selected to be opportunistic foragers, and will exploit ¯owers on which they can realize an energetic pro®t, even if they do not mechanically `®t' very well. Furthermore, the sensory abilities of opportunistic foragers are likely to be broadly tuned, so ¯owers also may be unable to evolve `private
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower' signals' detectable only by speci®c types of animals (Waser et al., 1996; Waser and Price, 1998). In our example, bumblebees possess visual systems that allow them to see the red ¯owers of I. aggregata (Chittka and Waser, 1997). They visit these ¯owers opportunistically when hummingbird visits are infrequent and the corollas therefore ®ll with nectar, allowing the bumblebees to make a good energetic pro®t (Pleasants and Waser, 1985). An alternative possible explanation for the lack of correlation between quantity and quality components of pollinator eectiveness is that natural selection may often favour generalized ¯ower morphologies that admit a diversity of visitors (Waser et al., 1996). Generalization may be adaptive, especially when the availability of any given pollinator ¯uctuates through time and space, as often occurs (e.g. Fishbein and Venable, 1995; Kwak and Velterop, 1997; C. M. Herrera, 1988; J. Herrera, 1988; Fenster and Dudash, 2001). In our example, the rate of hummingbird visitation to I. aggregata ¯owers at the RMBL varied four-fold over a 5 year study (Table 2), perhaps because of episodes of unusual mortality of hummingbirds on subtropical wintering grounds or during migration to and from temperate-zone breeding grounds (Calder and Calder, 1992; Russell et al., 1994). Annual variation in hummingbird visitation indicates that these animals are an uncertain resource from the plant's perspective. This, and the fact that individual plants enjoy only a single season of reproduction, suggests the value of `backup' pollinators, however surprising and apparently maladapted to the ¯owers. Indeed, within our 5 year study, the lowest rate of hummingbird visitation to I. aggregata occurred in 1997 (Table 2), but this was partially compensated for by B. appositus and other insect visitors. Bumblebees probably contributed as much as, or slightly more than, hummingbirds to seed production overall during 1997, judging from the product of visits per ¯ower h ÿ1 and seeds per visit from initial visits to virgin ¯owers (0.0041 4.36 0.018 seeds h ÿ1 for bumblebees vs. 0.0145 1.08 0.016 seeds h ÿ1 for hummingbirds). If this second explanation is true, and some generalization of I. aggregata ¯owers in use of pollinators is adaptive, how can it also be that the ¯owers exhibit such apparent specialization for one type of visitor, namely hummingbirds? Perhaps appearances are misleading and some subtle features of I. aggregata ¯owers actually adapt them for bumblebees. Examples of ¯owers as adaptive compromises have been proposed by Macior (1986), Hurlbert et al. (1996), Sahley (1996), and Lange and Scott (2000). Here one might imagine either that a ¯ower possesses some features adapting it to attract or exploit one type of pollinator and other features adapting it to another type, or that individual features are compromises (e.g. a corolla of intermediate length and width). Alternatively, perhaps it is simply unrealistic to assume that phenotypic specialization for one pollinator necessarily incurs a strong trade o in the ability to attract or make use of other pollinators. We assumed such a trade o in a model of ¯oral evolution which suggested that quality and quantity components of pollinator eectivness would evolve to be positively correlated (Waser et al., 1996), and the assumption is implicit in discussions of ¯oral specialization that
595
permeate pollination biology. However, if trade os are in fact weak, striking ¯oral features might evolve that adapt ¯owers to a subset of pollinators, without losing the ability to use other pollinators. Robinson and Wilson (1998) develop this idea for the evolution of phenotypic specialization in general, and Aigner (2001) explores it in detail for the special case of ¯oral evolution. In this view ¯owers will not be compromises because no complex features are needed to exploit one or more types of pollinator. Applied to I. aggregata, the hypothesis is that a long, narrow corolla is needed to ensure that hummingbirds, with narrow bills and rapid visits to ¯owers, will usually contact anthers and stigma; whereas no special adaptations are needed to make use of bumblebees, whose large hairy bodies and relatively long time spent in ¯owers ensure excellent contact with sex parts. These considerations raise exciting possibilities for future study of the ¯oral phenotype and of ¯oral adaptation. For example, there is some evidence to support the intuitive idea that the `®t' between ¯ower and pollinator aects pollinator eectiveness (Nilsson, 1988; Fenster, 1991; Campbell et al., 1996; Johnston and Steiner, 1997; Cresswell, 1998), but there are also examples (including the one discussed here) that raise questions about this idea (Cresswell and Galen, 1991; Wilson, 1995). Perhaps it is time to re®ne our thinking about the ®t between ¯owers and pollinators and to ®nd more quantitative methods for measuring it, for example with morphometric approaches developed for animals (e.g. Rohlf and Marcus, 1993). In addition, a static measure of ®t may not suce to tell us how animals of dierent morphologies pollinate the same ¯ower, since behaviour also plays a critical role. We advocate experimental manipulation of ¯oral features to explore how they interact with the behaviour of visitors to determine pollen transfer (Wilson, 1995; Campbell et al., 1996; Hurlbert et al., 1996; Temeles, 1996), and the inclusion both of visitors to which the ¯ower appears adapted, and those to which it does not. We predict many surprising results. AC K N OW L E D G E M E N T S Thanks to Paul Aigner, Hylton Mayer, Cathy Koehler, Kate Sharaf and Natasha Thorne for ®eld assistance, to Paul Aigner for discussions on ¯oral specialization and trade os, to Sarah Corbet, Kristina Jones and David Causton for editorial insights, and to the National Science Foundation (U.S.A.) and Reed College for ®nancial support. L I T E R AT U R E C I T E D Aigner PA. 2001. Optimality modelling and ®tness tradeos: when should plants become pollinator specialists?. Oikos (in press). Armbruster WS, Fenster CB, Dudash MR. 2000. Pollination `principles' revisited: specialization, pollination syndromes, and the evolution of ¯owers. Det Norske Videnskaps-Akademi I Matematisk Naturvidenskapelige Klasse, Skrifter, Ny Serie 39: 179±200. Calder WA, Calder LL. 1992. Broad-tailed hummingbird (Selasphorus platycercus). In: Poole A, Stettenheim P, Gill F, eds. The birds of North America. Philadelphia: The Academy of Natural Sciences, 1±16. Campbell DR, Waser NM, Price MV. 1996. Mechanisms of hummingbird-mediated selection for ¯ower width in Ipomopsis aggregata. Ecology 77: 1463±1472.
596
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
Campbell DR, Waser NM, Price MV, Lynch EA, Mitchell RJ. 1991. Components of phenotypic selection: pollen export and ¯ower corolla width in Ipomopsis aggregata. Evolution 45: 1458±1467. Chittka L, Waser NM. 1997. Why red ¯owers are not invisible to bees. Israel Journal of Plant Sciences 45: 169±183. Crepet WL. 1984. Advanced (constant) insect pollination mechanisms: patterns of evolution and implications vis-aÁ-vis angiosperm diversity. Annals of the Missouri Botanical Garden 71: 607±630. Cresswell JE. 1998. Stabilizing selection and the structural variability of ¯owers within species. Annals of Botany 81: 463±473. Cresswell JE, Galen C. 1991. Frequency-dependent selection and adaptive surfaces for ¯oral character combinationsÐthe pollination of Polemonium viscosum. American Naturalist 138: 1342±1353. de Jong T, Waser NM, Price MV, Ring RM. 1992. Plant size, geitonogamy, and seed set in Ipomopsis aggregata. Oecologia 89: 310±315. Faegri K, Pijl L van der. 1971. The principles of pollination ecology, 2nd edn. Oxford: Pergamon. Fenster CB. 1991. Selection on ¯oral morphology by hummingbirds. Biotropica 23: 98±101. Fenster CB, Dudash MR. 2001. Spatiotemporal variation in the role of hummingbirds as pollinators of Silene virginica. Ecology 82: 844±851. Fishbein M, Venable DL. 1995. Diversity and temporal change in the eective pollinators of Asclepias tuberosa. Ecology 77: 1061±1073. GoÂmez J, Zamora R. 1999. Generalization vs. specialization in the pollination system of Hormathophylla spinosa (Cruciferaceae). Ecology 80: 796±805. Grant KA, Grant V. 1968. Hummingbirds and their ¯owers. New York: Columbia University Press. Grant V, Grant KA. 1965. Flower pollination in the phlox family. New York: Columbia University Press. Grant V, Wilken DH. 1986. Taxonomy of the Ipomopsis aggregata group (Polemoniaceae). Botanical Gazette 147: 359±371. Herrera CM. 1987. Components of pollinator `quality': comparative analysis of a diverse insect assemblage. Oikos 50: 79±90. Herrera CM. 1988. Variation in mutualisms: the spatiotemporal mosaic of a pollinator assemblage. Biological Journal of the Linnean Society 35: 95±125. Herrera CM. 1989. Pollinator abundance, morphology, and ¯ower visitation rate: analysis of the `quantity' component in a plantpollinator system. Oecologia 80: 241±248. Herrera J. 1988. Pollination relationships in southern Spanish Mediterranean shrublands. Journal of Ecology 76: 274±287. Hurlbert AH, Hosoi SA, Temeles EJ, Ewald PW. 1996. Mobility of Impatiens capensis ¯owers: eect on pollen deposition and hummingbird foraging. Oecologia 105: 243±246. Jennersten O, Morse DH. 1991. The quality of pollination by diurnal and nocturnal insects visiting common milkweed, Asclepias syriaca. American Midland Naturalist 125: 18±28. Johnson SD, Steiner KE. 1997. Long-tongued ¯y pollination and evolution of ¯oral spur length in the Disa draconis complex (Orchidaceae). Evolution 51: 45±53. Kearns CA, Inouye DW. 1993. Techniques for pollination biologists. Niwot: University Press of Colorado. Kwak MM. 1993. The relative importance of syrphids and bumblebees as pollinators of three plant species. Proceedings of the Section of Experimental and Applied Entomology of the Netherlands Entomological Society, Amsterdam 4: 137±144. Kwak MM, Velterop O. 1997. Flower visitation by generalists and specialists: analysis of pollinator quality. Proceedings of the Section of Experimental and Applied Entomology of the Netherlands Entomological Society, Amsterdam 8: 85±89. Lange RS, Scott PE. 1999. Hummingbird and bee pollination of Penstemon pseudospectabilis. Journal of the Torrey Botanical Society 126: 99±106. Macior LW. 1986. Floral resource sharing by bumblebees and hummingbirds in Pedicularis (Scrophulariaceae) pollination. Bulletin of the Torrey Botanical Club 113: 101±109. Motten AF, Campbell DR, Alexander DE, Miller HL. 1981. Pollination eectiveness of specialist and generalist visitors to a North
Carolina population of Claytonia virginica. Ecology 62: 1278±1287. Nilsson LA. 1988. The evolution of ¯owers with deep corolla tubes. Nature 334: 147±149. Ollerton J, Watts S. 2000. Phenotype space and ¯oral typology± towards an objective assessment of pollination syndromes. Det Norske Videnskaps-Akademi I Matematisk Naturvidenskapelige Klasse, Skrifter, Ny Serie 39: 149±159. Olsen KM. 1997. Pollination eectiveness and pollinator importance in a population of Heterotheca subaxillaris (Asteraceae). Oecologia 109: 114±121. Pettersson MW. 1991. Pollination by a guild of ¯uctuating moth populations: option for unspecialization in the bladder campion, Silene vulgaris. Journal of Ecology 79: 591±604. Pleasants JM, Waser NM. 1985. Bumblebee foraging at a `hummingbird' ¯ower: reward economics and ¯oral choice. American Midland Naturalist 114: 283±291. Price MV, Waser NM. 1982. Experimental studies of pollen carryover: hummingbirds and Ipomopsis aggregata. Oecologia 54: 353±358. Primack RB, Silander JAJ. 1975. Measuring the relative importance of dierent pollinators to plants. Nature 255: 143±144. Proctor M, Yeo P, Lack A. 1996. The natural history of pollination. Portland, OR: Timber Press. Robinson BW, Wilson DS. 1998. Optimal foraging, specialization, and a solution to Liem's paradox. American Naturalist 151: 223±235. Rohlf FJ, Marcus LF. 1993. A revolution in morphometrics. Trends in Ecology and Evolution 8: 129±132. Russell RW, Carpenter FL, Hixon MA, Paton DC. 1994. The impact of variation in stopover habitat quality on migrant rufous hummingbirds. Conservation Biology 8: 483±490. Sahley CT. 1996. Bat and hummingbird pollination of an autotetraploid columnar cactus, Weberbauerocereus weberbaueri (Cactaceae). American Journal of Botany 83: 1329±1336. Schemske DW, Horvitz CC. 1984. Variation among ¯oral visitors in pollination ability: a precondition for mutualism specialization. Science 225: 519±521. Stebbins GL. 1970. Adaptive radiation of reproductive characteristics in angiosperms I: pollination mechanisms. Annual Review of Ecology and Systematics 1: 307±326. Temeles EJ. 1996. A new dimension to hummingbird-¯ower relationships. Oecologia 105: 517±523. Tepedino VJ. 1981. The pollination eciency of the squash bee (Peponapis pruinosa) and the honey bee (Apis mellifera) on summer squash (Cucurbita pepo). Journal of the Kansas Entomological Society 54: 359±377. Thomson JD, Price MV, Waser NM, Stratton DA. 1986. Comparative studies of pollen and ¯uorescent dye transport by bumble bees visiting Erythronium grandi¯orum. Oecologia 69: 516±566. Waser NM. 1982. A comparison of distances ¯own by dierent visitors to ¯owers of the same species. Oecologia 55: 934±944. Waser NM. 1983. The adaptive nature of ¯oral traits: ideas and evidence. In: Real LA, ed. Pollination biology. New York: Academic Press, 241±285. Waser NM, Price MV. 1983. Optimal and actual outcrossing in plants, and the nature of plant-pollinator interaction. In: Jones CE, Little RJ, eds. Handbook of experimental pollination biology. New York: Van Nostrand Reinhold, 341±359. Waser NM, Price MV. 1990. Pollination eciency and eectiveness of bumble bees and hummingbirds visiting Delphinium nelsonii. Collectanea Botanica (Barcelona) 19: 9±20. Waser NM, Price MV. 1991. Reproductive costs of self pollination in Ipomopsis aggregata: are ovules usurped?. American Journal of Botany 78: 1036±1043. Waser NM, Price MV. 1998. What plant ecologists can learn from zoology. Perspectives in Plant Ecology, Evolution and Systematics 1: 137±150. Waser NM, Chittka L, Price MV, Williams N, Ollerton J. 1996. Generalization in pollination systems, and why it matters. Ecology 77: 279±296. Wilson P. 1995. Selection for pollination success and the mechanical ®t of Impatiens ¯owers around bumblebee bodies. Biological Journal of the Linnean Society 55: 355±383.
Exploring the `Most Eective Pollinator Principle' with Complex Flowers: Bumblebees and Ipomopsis aggregata M A R G A R E T M . M AY F I EL D {{, N I C KO L A S M . WA S E R *{} and MA RY V. P R I C E {} {Rocky Mountain Biological Laboratory, Crested Butte, CO 81224, USA, {Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA and }Department of Biology, University of California, Riverside, CA 92521, USA Received: 4 April 2001 Returned for revision: 11 May 2001 Accepted: 19 June 2001 Published electronically: 17 August 2001 The `most eective pollinator principle' implies that ¯oral characteristics often re¯ect adaptation to the pollinator that transfers the most pollen, through a combination of high rate of visitation to ¯owers and eective deposition of pollen during each visit. We looked for the expected positive correlation between quantity and quality of visits in Ipomopsis aggregata, whose red, tubular ¯owers are considered to be adapted to hummingbirds. Hummingbirds were indeed the most common ¯oral visitors in 5 years of observation. However, long-tongued bumblebees deposited on average three-times as much outcross pollen per visit to virgin ¯owers, and elicited four-times as much seed production, as did hummingbirds. Hence visitors that are relatively infrequent, and unexpected given the `pollination syndrome' of the plant, can be surprisingly good pollinators. One interpretation of this observation is that natural selection favours a specialized ¯oral morphology that excludes all but a single type of visitor, but that there are constraints on achieving this outcome. An alternative is that selection favours some degree of ¯oral generalization, but that ¯owers can retain features that adapt them to a particular type of pollinator in spite of this # 2001 Annals of Botany Company generalization. Key words: Bumblebees, ¯oral evolution, hummingbirds, Ipomopsis aggregata, pollination syndromes, pollinator eectiveness, scarlet gilia, seed set, specialization, temporal variation.
I N T RO D U C T I O N Angiosperm ¯owers are strikingly diverse. They vary in colour, scent, size, morphology, and in the type and amount of reward they oer as a bribe to pollinating animals (Proctor et al., 1996). Their morphology can also be complex. For example, some ¯owers sequester rewards for pollinators within narrow corolla tubes or spurs or behind barricades that are dicult to breach, present sex parts in highly speci®c locations, or exhibit bilateral symmetry or asymmetry. Such complex features are generally interpreted as adaptations for attracting and exploiting certain types of pollinators (e.g. large bees vs. moths vs. birds), and for excluding other types (Grant and Grant, 1965; Crepet, 1984). This view is codi®ed in the concept of `pollination syndromes', a dominant organizing theme in pollination biology (Faegri and van der Pijl, 1971; for recent commentary see Armbruster et al., 2000; Ollerton and Watts, 2000). Further, the syndrome concept implies that plant species that specialize on a given type of pollinator will converge on a suite of phenotypic features that adapts them to the morphology, sensory and nutritional physiology, and behaviour of this pollinator. Against this backdrop, ecologists in the ®eld observe that many complex ¯owers receive visits from several types of animals. Stebbins (1970, p. 318) proposed a resolution to this paradox of apparent phenotypic specialization in the * For correspondence at: Department of Biology, University of California, Riverside CA, 92521, USA. Fax 1 909 787 4286, e-mail [email protected]
0305-7364/01/100591+06 $35.00/00
face of ecological generalization by suggesting that ` . . . the characteristics of the ¯ower will be molded by those pollinators that visit it most frequently and eectively'. In this view, the contribution to ®tness by a given type of ¯ower visitor is broken down into two multiplicative components, the number of ¯ower visits and the amount of reproductively-eective (e.g. compatible) pollen transferred to a ¯ower per visit. These `quantity' and `quality' components of pollination (sensu Waser and Price, 1983; Herrera, 1987, 1989) together provide an estimate of overall pollinator eectiveness [sometimes called pollinator importance (Olsen, 1997) or pollinator eciency (Schemske and Horvitz, 1984)] for a given species of animal visiting ¯owers of a given species of plant. An extension of Stebbins' `most eective pollinator principle' is that the quantity and quality components of eectiveness will be positively related across dierent types of visitors to ¯owers of a given plant species. The logic is as follows: all else being equal, any heritable change in ¯oral expression which increases the quality of each visit by a frequent visitor, or conversely increases the frequency of visits by a pollinator of high per-visit quality, should spread by natural selection even if there is a trade o in the form of reduced eectiveness of other visitors. Indeed, it is possible to read Stebbins' (1970) original statement of the most eective pollinator principle as predicting precisely this process (Olsen, 1997; Aigner, 2001), and Waser et al. (1996) independently codi®ed it in a simple model of ¯oral evolution in a constant environment. # 2001 Annals of Botany Company
592
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
Numerous studies have now con®rmed Stebbins' (1970) implication that ¯ower visitors can vary substantially in their overall eectiveness (e.g. Primack and Silander, 1975; Schemske and Horvitz, 1984; Jennersten and Morse, 1991), although they do not always do so (e.g. Motten et al., 1981; Tepedino, 1981; Waser and Price, 1990). It seemed to us of value to take the additional step of testing the proposal that the quantity and quality components of eectiveness will be positively related, using a plant species whose complex ¯owers clearly conform to one of the more striking pollination syndromes. To this end we measured pollination eectiveness of bumblebees visiting Ipomopsis aggregata (Pursh) V. Grant, a quintessential North American `hummingbird ¯ower', and compared this to the eectiveness of hummingbirds, which are predominant visitors in most years. Our results indicate that bumblebees are surprisingly good pollinators of I. aggregata, and we suggest some new ways of looking at the most eective pollinator principle and ¯oral adaptation.
M AT E R I A L S A N D M E T H O D S The study system In western North America, numerous unrelated plant species have repeatedly evolved a suite of ¯oral traits characteristic of the hummingbird pollination syndrome: relatively abundant but dilute nectar concealed within deep tubes, bright (usually reddish) colouration, lack of distinct landing platform, and absence of strong scent (Grant and Grant, 1968). A typical example is the scarlet ( skyrocket) gilia, Ipomopsis aggregata (Polemoniaceae), a monocarpic herbaceous perennial found in montane regions from southern Arizona, New Mexico and Texas to British Columbia (Grant and Wilken, 1986). At the Rocky Mountain Biological Laboratory (RMBL) in western Colorado, USA, plants grow for 2±7 years or more as vegetative rosettes, then die after producing between ten and several hundred red, tubular, protandrous, self-incompatible ¯owers. The predominant visitors to I. aggregata ¯owers at RMBL are broad-tailed and rufous hummingbirds (respectively Selasphorus platycercus Gmelin and S. rufus Gmelin). However, insects in several orders also visit, including solitary bees and syrphid and muscoid ¯ies that collect or eat pollen, and hawkmoths and butter¯ies that feed on nectar (Waser, 1982). Queens of Bombus appositus (Cresson) also visit I. aggregata ¯owers in some summers. With the longest tongues ( proboscides) of any bumblebee at RMBL, these animals obtain a good energetic reward when hummingbird visits are rare and nectar accumulates in the ¯owers, even though they cannot reach all the nectar ( for complete analysis see Pleasants and Waser, 1985). During early July 1997, B. appositus queens foraged in substantial numbers at I. aggregata. We observed them ¯ying systematically from ¯ower to ¯ower, landing on each ¯ower by grasping the re¯exed tips of the petals, and with proboscides extended pressing their heads into the opening of the corolla, where sex parts also are located (Fig. 1).
Per-visit eectiveness At the peak of bumblebee visitation in 1997, we exposed `virgin' ¯owers of I. aggregata to single visits by B. appositus, S. platycercus and S. rufus. Within a 4 m 4 m study plot we randomly chose ®ve large plants, each with 50±100 ¯owers open at a time. We removed all the ¯owers that were already open, carefully excised the immature anthers from elongated buds, and then placed a wire cage covered with ®ne nylon netting over each plant. These procedures ensured that stigmas would carry no pollen until visited by a pollinator. When emasculated ¯owers became femalereceptive 2±3 d later we removed the cages and waited for visitors. Observation over 6 d (10±12, 17±19 July), for 1±7 h per day between 0800 and 1800 h (total observation 30 h) was rewarded by 14 ¯ower visits by B. appositus queens, 13 visits by S. platycercus and 15 visits by S. rufus, all concentrated in the afternoon hours. During each of these visits, the animal foraged ®rst in a natural fashion on other plants outside of the experimental plot, then ¯ew to one or more experimental plants. Bumblebees visited fewer ¯owers per experimental plant (range 1±5; mean 2.3) than did hummingbirds (range 1±14; mean 5.6). After each pollinator had left the experimental plants we removed the corollas of the ¯owers visited so that they would not be revisited later, and continued observations. After each observation period we replaced the cages over plants. We returned 24 h after each ¯ower had been visited to excise its stigma. Stigmas were mounted on microscope slides in glycerine-gelatine medium stained with basic fuchsin
F I G . 1. A queen of B. appositus visiting and pollinating a red `hummingbird' ¯ower of I. aggregata.
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower' (Kearns and Inouye, 1993), and pollen loads counted at 100 magni®cation. Pollen tubes in I. aggregata grow to the ovary and fertilize ovules within 24 h (Waser and Price, 1991), so removal of stigmas did not aect fruit set, and we were able to return 3 to 4 weeks after a ¯ower had been visited to collect the resulting fruit and count mature seeds. Annual variation in rates of visitation The amount of pollen transferred to a virgin ¯ower after a single visit, and the resulting seed set, both estimate the quality component of pollinator eectiveness. In addition, we wished to estimate the quantity component, i.e. rate of visitation, to I. aggregata ¯owers by various pollinators. To this end we utilized three permanent study sites at the RMBL, including the site of the experiment described above. Sites were 170±550 m apart. For 5 weeks during peak ¯owering of I. aggregata in 1997, we recorded visits to 15±20 individually numbered plants at each site 3 d per week for 1.5 h each day between 0800 and 1800 h, for totals of 4.5 h per site per week. Open ¯owers on numbered plants were counted before each census so that results could be expressed as visits per ¯ower h ÿ1 by each species of visitor observed during a census. Results could then be compared to those from censuses carried out in the same fashion and at the same study sites in 1996, 1998, 1999 and 2000. R E S U LT S
Pollen load
Bumblebees were suprisingly good pollinators of I. aggregata. In a single visit to a virgin ¯ower they delivered on average three-times as much pollen to stigmas as did hummingbirds (respective means of 112.7 vs. 38.1 pollen grains; Fig. 2). This dierence was highly statistically signi®cant, whereas pollen delivery did not dier signi®cantly between the two species of hummingbird (Table 1). Following a single visit by B. appositus, 57 % of ¯owers produced expanded fruits, whereas fruit set was only 31 and 13 % following pollination by S. platycercus and S. rufus,
593
respectively. Again the dierence between bumblebees and hummingbirds was signi®cant, whereas that between the two species of hummingbird was not, based on categorical analysis (Table 1). Furthermore, expanded fruits resulting from bumblebee visits contained 1.7-times as many seeds per fruit as those resulting from hummingbird visits (respective means of 7.63 vs. 4.58 seeds). Although this dierence was not signi®cant, it is of much larger magnitude than the dierence between the two hummingbirds (Table 1). Combining the eects of dierent pollinators on the probability of fruit expansion and seed set per fruit, single visits by bumblebees yielded on average fourtimes the seed production of single visits by hummingbirds (respective grand means of 4.36 vs. 1.08 seeds). Bumblebees also were reasonably frequent visitors to I. aggretata ¯owers in summer 1997. Averaged across the entire ¯owering season, they visited ¯owers about a third to a quarter as frequently as hummingbirds (Table 2). Other insects, in particular syrphid ¯ies, also contributed a substantial minority of all visits. Because syrphids were foraging for pollen rather than nectar, and so avoided our experimentally emasculated ¯owers, we were unable to assess their contribution to pollination. DISCUSSION Several workers have reported recently on pollination interactions in which there is at best a weak relationship between quality and quantity components of pollinator eectiveness (e.g. Pettersson 1991; Kwak 1993; Olsen, 1997; GoÂmez and Zamora, 1999). Most such studies seem to T A B L E 1. Analyses of pollen delivery, fruit set and seed set resulting from single visits to I. aggregata ¯owers Source of variation in pollen load Pollinator type Bees vs. birds Broad-tails vs. rufous Error
SS
d.f
F
P
184.86 182.01 1.28 638.34
2 1 1 39
5.65 11.12 0.08
0.007 0.002 0.781
Chi-square
d.f.
P
1 2 1 1
0.031 0.060 0.025 0.273
150
Source of variation in fruit set
100
Intercept Pollinator type Bees vs. birds Broad-tails vs. rufous
4.66 5.62 5.04 1.20
Source of variation in seeds per fruit
SS
d.f.
F
P
Pollinator type Bees vs. birds Broad-tails vs. rufous Error
0.84 0.78 0.00 7.42
2 1 1 11
0.62 1.16 0.00
0.556 0.305 0.998
A 50
0
B. appositus
B
B
S. platycercus
S. rufus
Pollinator F I G . 2. Pollen delivery to virgin stigmas of ¯owers of I. aggregata (means + 1 s.e.) by queen bumblebees (B. appositus) vs. two species of hummingbird (S. platycercus and S. rufus). Dierent letters indicate signi®cantly dierent means.
The analyses of pollen load per stigma and of seed set per fruit are single-factor ANOVAs; that for seed set is based only on non-zero values. Raw values for pollen counts and seed set were square-root transformed to achieve normality and homoscedasticity of model residuals. The analysis of fruit set is a categorical analysis of the number of expanded vs. unexpanded fruits. In each case the eect of pollinator species is subdivided with orthogonal contrasts into eects of bumblebees vs. hummingbirds (`Bees vs. birds') and of broad-tailed vs. rufous hummingbirds (`Broad-tails vs. rufous').
594
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
T A B L E 2. Rates of visitation to I. aggregata ¯owers at three permanent study sites at the RMBL by hummingbirds, bumblebees (B. appositus queens) and other pollinators, across 5 years of study Mean visits per ¯ower h ÿ1 Year
Hummingbirds
Bumblebees
Other pollinators
Total
1996 1997 1998 1999 2000
0.0363 0.0145 0.0156 0.0266 0.0603
0.0001 0.0041 0 0.0017 0
0.0001 0.0077 0.0094 0.0072 0.0066
0.0365 0.0263 0.0250 0.0355 0.0669
The protocol for observation in all years was similar to that described in the text for 1997. The `other pollinators' comprised solitary bees and swallowtail butter¯ies in summer 1996, solitary bees, syrphid ¯ies and muscoid ¯ies in 1997, 1998 and 1999, and muscoid ¯ies in 2000.
involve relatively generalized ¯owers that do not clearly exhibit features of a pollination syndrome and that are visited only by insects. It is more perplexing to ®nd that insects were the most eective per visit in transferring pollen to the ¯owers of I. aggregata, considering that these ¯owers conform closely to a hummingbird pollination syndrome and hummingbirds are the most frequent visitors overall. The wide margin by which bumblebees outperformed hummingbirds in pollen transfer and in eliciting fruit and seed set is especially remarkable. Readers may object that our estimates of pollinator eectiveness are misleading. Any such estimate based on rate of visitation and per-visit pollen transfer is incomplete when pollinators dier in additional ways, as they may do (Waser, 1983). For example, I. aggregata ¯owers visited by bumblebees might receive less outcross pollen, and more self pollen from other ¯owers on the same plant (geitonogamy), compared to ¯owers visited by hummingbirds. Such a dierence would reduce seed set in nature (Waser and Price, 1991; de Jong et al., 1992), but would not show up in our experiments with emasculated ¯owers. However, our small sample (see Materials and Methods: per-visit eectiveness) suggests that on average B. appositus queens visited less than half as many ¯owers per plant as did S. platycercus and S. rufus; a large sample (Waser, 1982) con®rms this, and the dierence is statistically signi®cant. Thus bumblebees should transfer proportionately less geitonogamous pollen than hummingbirds. A similar objection is that bumblebees and hummingbirds might visit at dierent times of day, and pollen might be available for transfer only at certain times. However, our observations here, and earlier ones (N. Waser, unpubl. res.), indicate no obvious interspeci®c dierences in times of activity between about 0700 and 1900 h, with activity of both bumblebees and hummingbirds highest in the afternoon. Polleniferous male-phase ¯owers and receptive female-phase ¯owers are available in this period, because most new ¯owers of I. aggregata open (with anthers dehiscing pollen) in the afternoon, and most stigmas become receptive then as well (M. Price and N. Waser, unpubl. res.). Indeed, the pollen delivered to our experimental ¯owers is a
natural indicator of the pollen available to be exported from unmanipulated ¯owers that were visited by pollinators before they entered our experimental plot. Probably the most serious concern is that our measures of pollinator eectiveness miss subtle dierences in the export of pollen, i.e. in success of plants through the male sex function. This is a complex topic of which we venture a summary. Assume that individual pollinators visit I. aggregata for a period sucient for their pick-up and deposition of pollen at ¯owers to reach equilibrium. Due to `pollen carryover' (e.g. Price and Waser, 1982), pollen deposited on a female-phase ¯ower will come from some number of previously visited male-phase ¯owers, with the number likely to dier between bumblebees and hummingbirds due to dierent carryover properties. Nonetheless, if at equilibrium a bumblebeee deposits on average X pollen grains on a stigma and a hummingbird on average Y, then X and Y are the expected values of total pollen export for each male-phase ¯ower previously visited by the pollinator in question. In our system, X 110 pollen grains and Y 40, suggesting that bumblebee visits contribute more to male reproductive success, just as they do to female reproductive success. However, this analysis ignores the possibility that bumblebees pick up more pollen than hummingbirds in the ®rst visit to a male-phase ¯ower, and lose most of it (e.g. because they groom it from their bodies), yet still deposit more on female-phase ¯owers. If this is true, the critical issue is whether ¯owers can expect many visits during their male phase, in which case hummingbirds might export more pollen overall, because bumblebees would quickly remove all pollen but lose most of it. Here we can only oer the information that I. aggregata ¯owers remain in male phase for 1.5±2 d and receive only about one pollinator visit per day (Campbell et al., 1991; Waser and Price, 1991)Ð conditions which favour the interpretation that bumblebees are superior in overall pollen export. To go beyond this will require dicult experiments with captive bumblebees to determine the dynamics of their pollen transfer (e.g. Price and Waser, 1982; Thomson et al., 1986). We have not attempted such experiments. There is a possibility, therefore, that dierences in total pollen export from each ¯ower will render hummingbird visits of highest quality overall. However, this possibility seems unlikely, and we now proceed to discuss some implications of our original conclusion that visit quantity and quality are not positively related in our I. aggregata populations. What are the possible explanations for this in such complex ¯owers? It seems to us that one general possibility involves constraint in the design of ¯owers. If natural selection consistently favours a ¯oral morphology that limits the types of animals that can gain access to nectar or other rewards, it seems certain that some degree of exclusivity will be achieved. However, there may be limits to this process, in large part because the animals do not cooperate. Pollinators themselves are often selected to be opportunistic foragers, and will exploit ¯owers on which they can realize an energetic pro®t, even if they do not mechanically `®t' very well. Furthermore, the sensory abilities of opportunistic foragers are likely to be broadly tuned, so ¯owers also may be unable to evolve `private
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower' signals' detectable only by speci®c types of animals (Waser et al., 1996; Waser and Price, 1998). In our example, bumblebees possess visual systems that allow them to see the red ¯owers of I. aggregata (Chittka and Waser, 1997). They visit these ¯owers opportunistically when hummingbird visits are infrequent and the corollas therefore ®ll with nectar, allowing the bumblebees to make a good energetic pro®t (Pleasants and Waser, 1985). An alternative possible explanation for the lack of correlation between quantity and quality components of pollinator eectiveness is that natural selection may often favour generalized ¯ower morphologies that admit a diversity of visitors (Waser et al., 1996). Generalization may be adaptive, especially when the availability of any given pollinator ¯uctuates through time and space, as often occurs (e.g. Fishbein and Venable, 1995; Kwak and Velterop, 1997; C. M. Herrera, 1988; J. Herrera, 1988; Fenster and Dudash, 2001). In our example, the rate of hummingbird visitation to I. aggregata ¯owers at the RMBL varied four-fold over a 5 year study (Table 2), perhaps because of episodes of unusual mortality of hummingbirds on subtropical wintering grounds or during migration to and from temperate-zone breeding grounds (Calder and Calder, 1992; Russell et al., 1994). Annual variation in hummingbird visitation indicates that these animals are an uncertain resource from the plant's perspective. This, and the fact that individual plants enjoy only a single season of reproduction, suggests the value of `backup' pollinators, however surprising and apparently maladapted to the ¯owers. Indeed, within our 5 year study, the lowest rate of hummingbird visitation to I. aggregata occurred in 1997 (Table 2), but this was partially compensated for by B. appositus and other insect visitors. Bumblebees probably contributed as much as, or slightly more than, hummingbirds to seed production overall during 1997, judging from the product of visits per ¯ower h ÿ1 and seeds per visit from initial visits to virgin ¯owers (0.0041 4.36 0.018 seeds h ÿ1 for bumblebees vs. 0.0145 1.08 0.016 seeds h ÿ1 for hummingbirds). If this second explanation is true, and some generalization of I. aggregata ¯owers in use of pollinators is adaptive, how can it also be that the ¯owers exhibit such apparent specialization for one type of visitor, namely hummingbirds? Perhaps appearances are misleading and some subtle features of I. aggregata ¯owers actually adapt them for bumblebees. Examples of ¯owers as adaptive compromises have been proposed by Macior (1986), Hurlbert et al. (1996), Sahley (1996), and Lange and Scott (2000). Here one might imagine either that a ¯ower possesses some features adapting it to attract or exploit one type of pollinator and other features adapting it to another type, or that individual features are compromises (e.g. a corolla of intermediate length and width). Alternatively, perhaps it is simply unrealistic to assume that phenotypic specialization for one pollinator necessarily incurs a strong trade o in the ability to attract or make use of other pollinators. We assumed such a trade o in a model of ¯oral evolution which suggested that quality and quantity components of pollinator eectivness would evolve to be positively correlated (Waser et al., 1996), and the assumption is implicit in discussions of ¯oral specialization that
595
permeate pollination biology. However, if trade os are in fact weak, striking ¯oral features might evolve that adapt ¯owers to a subset of pollinators, without losing the ability to use other pollinators. Robinson and Wilson (1998) develop this idea for the evolution of phenotypic specialization in general, and Aigner (2001) explores it in detail for the special case of ¯oral evolution. In this view ¯owers will not be compromises because no complex features are needed to exploit one or more types of pollinator. Applied to I. aggregata, the hypothesis is that a long, narrow corolla is needed to ensure that hummingbirds, with narrow bills and rapid visits to ¯owers, will usually contact anthers and stigma; whereas no special adaptations are needed to make use of bumblebees, whose large hairy bodies and relatively long time spent in ¯owers ensure excellent contact with sex parts. These considerations raise exciting possibilities for future study of the ¯oral phenotype and of ¯oral adaptation. For example, there is some evidence to support the intuitive idea that the `®t' between ¯ower and pollinator aects pollinator eectiveness (Nilsson, 1988; Fenster, 1991; Campbell et al., 1996; Johnston and Steiner, 1997; Cresswell, 1998), but there are also examples (including the one discussed here) that raise questions about this idea (Cresswell and Galen, 1991; Wilson, 1995). Perhaps it is time to re®ne our thinking about the ®t between ¯owers and pollinators and to ®nd more quantitative methods for measuring it, for example with morphometric approaches developed for animals (e.g. Rohlf and Marcus, 1993). In addition, a static measure of ®t may not suce to tell us how animals of dierent morphologies pollinate the same ¯ower, since behaviour also plays a critical role. We advocate experimental manipulation of ¯oral features to explore how they interact with the behaviour of visitors to determine pollen transfer (Wilson, 1995; Campbell et al., 1996; Hurlbert et al., 1996; Temeles, 1996), and the inclusion both of visitors to which the ¯ower appears adapted, and those to which it does not. We predict many surprising results. AC K N OW L E D G E M E N T S Thanks to Paul Aigner, Hylton Mayer, Cathy Koehler, Kate Sharaf and Natasha Thorne for ®eld assistance, to Paul Aigner for discussions on ¯oral specialization and trade os, to Sarah Corbet, Kristina Jones and David Causton for editorial insights, and to the National Science Foundation (U.S.A.) and Reed College for ®nancial support. L I T E R AT U R E C I T E D Aigner PA. 2001. Optimality modelling and ®tness tradeos: when should plants become pollinator specialists?. Oikos (in press). Armbruster WS, Fenster CB, Dudash MR. 2000. Pollination `principles' revisited: specialization, pollination syndromes, and the evolution of ¯owers. Det Norske Videnskaps-Akademi I Matematisk Naturvidenskapelige Klasse, Skrifter, Ny Serie 39: 179±200. Calder WA, Calder LL. 1992. Broad-tailed hummingbird (Selasphorus platycercus). In: Poole A, Stettenheim P, Gill F, eds. The birds of North America. Philadelphia: The Academy of Natural Sciences, 1±16. Campbell DR, Waser NM, Price MV. 1996. Mechanisms of hummingbird-mediated selection for ¯ower width in Ipomopsis aggregata. Ecology 77: 1463±1472.
596
May®eld et al.ÐBumblebees Pollinating a `Hummingbird Flower'
Campbell DR, Waser NM, Price MV, Lynch EA, Mitchell RJ. 1991. Components of phenotypic selection: pollen export and ¯ower corolla width in Ipomopsis aggregata. Evolution 45: 1458±1467. Chittka L, Waser NM. 1997. Why red ¯owers are not invisible to bees. Israel Journal of Plant Sciences 45: 169±183. Crepet WL. 1984. Advanced (constant) insect pollination mechanisms: patterns of evolution and implications vis-aÁ-vis angiosperm diversity. Annals of the Missouri Botanical Garden 71: 607±630. Cresswell JE. 1998. Stabilizing selection and the structural variability of ¯owers within species. Annals of Botany 81: 463±473. Cresswell JE, Galen C. 1991. Frequency-dependent selection and adaptive surfaces for ¯oral character combinationsÐthe pollination of Polemonium viscosum. American Naturalist 138: 1342±1353. de Jong T, Waser NM, Price MV, Ring RM. 1992. Plant size, geitonogamy, and seed set in Ipomopsis aggregata. Oecologia 89: 310±315. Faegri K, Pijl L van der. 1971. The principles of pollination ecology, 2nd edn. Oxford: Pergamon. Fenster CB. 1991. Selection on ¯oral morphology by hummingbirds. Biotropica 23: 98±101. Fenster CB, Dudash MR. 2001. Spatiotemporal variation in the role of hummingbirds as pollinators of Silene virginica. Ecology 82: 844±851. Fishbein M, Venable DL. 1995. Diversity and temporal change in the eective pollinators of Asclepias tuberosa. Ecology 77: 1061±1073. GoÂmez J, Zamora R. 1999. Generalization vs. specialization in the pollination system of Hormathophylla spinosa (Cruciferaceae). Ecology 80: 796±805. Grant KA, Grant V. 1968. Hummingbirds and their ¯owers. New York: Columbia University Press. Grant V, Grant KA. 1965. Flower pollination in the phlox family. New York: Columbia University Press. Grant V, Wilken DH. 1986. Taxonomy of the Ipomopsis aggregata group (Polemoniaceae). Botanical Gazette 147: 359±371. Herrera CM. 1987. Components of pollinator `quality': comparative analysis of a diverse insect assemblage. Oikos 50: 79±90. Herrera CM. 1988. Variation in mutualisms: the spatiotemporal mosaic of a pollinator assemblage. Biological Journal of the Linnean Society 35: 95±125. Herrera CM. 1989. Pollinator abundance, morphology, and ¯ower visitation rate: analysis of the `quantity' component in a plantpollinator system. Oecologia 80: 241±248. Herrera J. 1988. Pollination relationships in southern Spanish Mediterranean shrublands. Journal of Ecology 76: 274±287. Hurlbert AH, Hosoi SA, Temeles EJ, Ewald PW. 1996. Mobility of Impatiens capensis ¯owers: eect on pollen deposition and hummingbird foraging. Oecologia 105: 243±246. Jennersten O, Morse DH. 1991. The quality of pollination by diurnal and nocturnal insects visiting common milkweed, Asclepias syriaca. American Midland Naturalist 125: 18±28. Johnson SD, Steiner KE. 1997. Long-tongued ¯y pollination and evolution of ¯oral spur length in the Disa draconis complex (Orchidaceae). Evolution 51: 45±53. Kearns CA, Inouye DW. 1993. Techniques for pollination biologists. Niwot: University Press of Colorado. Kwak MM. 1993. The relative importance of syrphids and bumblebees as pollinators of three plant species. Proceedings of the Section of Experimental and Applied Entomology of the Netherlands Entomological Society, Amsterdam 4: 137±144. Kwak MM, Velterop O. 1997. Flower visitation by generalists and specialists: analysis of pollinator quality. Proceedings of the Section of Experimental and Applied Entomology of the Netherlands Entomological Society, Amsterdam 8: 85±89. Lange RS, Scott PE. 1999. Hummingbird and bee pollination of Penstemon pseudospectabilis. Journal of the Torrey Botanical Society 126: 99±106. Macior LW. 1986. Floral resource sharing by bumblebees and hummingbirds in Pedicularis (Scrophulariaceae) pollination. Bulletin of the Torrey Botanical Club 113: 101±109. Motten AF, Campbell DR, Alexander DE, Miller HL. 1981. Pollination eectiveness of specialist and generalist visitors to a North
Carolina population of Claytonia virginica. Ecology 62: 1278±1287. Nilsson LA. 1988. The evolution of ¯owers with deep corolla tubes. Nature 334: 147±149. Ollerton J, Watts S. 2000. Phenotype space and ¯oral typology± towards an objective assessment of pollination syndromes. Det Norske Videnskaps-Akademi I Matematisk Naturvidenskapelige Klasse, Skrifter, Ny Serie 39: 149±159. Olsen KM. 1997. Pollination eectiveness and pollinator importance in a population of Heterotheca subaxillaris (Asteraceae). Oecologia 109: 114±121. Pettersson MW. 1991. Pollination by a guild of ¯uctuating moth populations: option for unspecialization in the bladder campion, Silene vulgaris. Journal of Ecology 79: 591±604. Pleasants JM, Waser NM. 1985. Bumblebee foraging at a `hummingbird' ¯ower: reward economics and ¯oral choice. American Midland Naturalist 114: 283±291. Price MV, Waser NM. 1982. Experimental studies of pollen carryover: hummingbirds and Ipomopsis aggregata. Oecologia 54: 353±358. Primack RB, Silander JAJ. 1975. Measuring the relative importance of dierent pollinators to plants. Nature 255: 143±144. Proctor M, Yeo P, Lack A. 1996. The natural history of pollination. Portland, OR: Timber Press. Robinson BW, Wilson DS. 1998. Optimal foraging, specialization, and a solution to Liem's paradox. American Naturalist 151: 223±235. Rohlf FJ, Marcus LF. 1993. A revolution in morphometrics. Trends in Ecology and Evolution 8: 129±132. Russell RW, Carpenter FL, Hixon MA, Paton DC. 1994. The impact of variation in stopover habitat quality on migrant rufous hummingbirds. Conservation Biology 8: 483±490. Sahley CT. 1996. Bat and hummingbird pollination of an autotetraploid columnar cactus, Weberbauerocereus weberbaueri (Cactaceae). American Journal of Botany 83: 1329±1336. Schemske DW, Horvitz CC. 1984. Variation among ¯oral visitors in pollination ability: a precondition for mutualism specialization. Science 225: 519±521. Stebbins GL. 1970. Adaptive radiation of reproductive characteristics in angiosperms I: pollination mechanisms. Annual Review of Ecology and Systematics 1: 307±326. Temeles EJ. 1996. A new dimension to hummingbird-¯ower relationships. Oecologia 105: 517±523. Tepedino VJ. 1981. The pollination eciency of the squash bee (Peponapis pruinosa) and the honey bee (Apis mellifera) on summer squash (Cucurbita pepo). Journal of the Kansas Entomological Society 54: 359±377. Thomson JD, Price MV, Waser NM, Stratton DA. 1986. Comparative studies of pollen and ¯uorescent dye transport by bumble bees visiting Erythronium grandi¯orum. Oecologia 69: 516±566. Waser NM. 1982. A comparison of distances ¯own by dierent visitors to ¯owers of the same species. Oecologia 55: 934±944. Waser NM. 1983. The adaptive nature of ¯oral traits: ideas and evidence. In: Real LA, ed. Pollination biology. New York: Academic Press, 241±285. Waser NM, Price MV. 1983. Optimal and actual outcrossing in plants, and the nature of plant-pollinator interaction. In: Jones CE, Little RJ, eds. Handbook of experimental pollination biology. New York: Van Nostrand Reinhold, 341±359. Waser NM, Price MV. 1990. Pollination eciency and eectiveness of bumble bees and hummingbirds visiting Delphinium nelsonii. Collectanea Botanica (Barcelona) 19: 9±20. Waser NM, Price MV. 1991. Reproductive costs of self pollination in Ipomopsis aggregata: are ovules usurped?. American Journal of Botany 78: 1036±1043. Waser NM, Price MV. 1998. What plant ecologists can learn from zoology. Perspectives in Plant Ecology, Evolution and Systematics 1: 137±150. Waser NM, Chittka L, Price MV, Williams N, Ollerton J. 1996. Generalization in pollination systems, and why it matters. Ecology 77: 279±296. Wilson P. 1995. Selection for pollination success and the mechanical ®t of Impatiens ¯owers around bumblebee bodies. Biological Journal of the Linnean Society 55: 355±383.