Behavioral Ecology Advance Access originally published online on July 31, 2006
Behavioral Ecology 2006 17(6):905-910; doi:10.1093/beheco/arl024
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Models of optimal foraging and resource partitioning: deep corollas for long tongues
a Estación Experimental de Zonas Áridas (CSIC), General Segura 1, 04001 Almeria, Spain b Mediterranean Institute for Advanced Studies (IMEDEA, CSIC-UIB), Miquel Marquès 21, 07190 Esporles, Mallorca, Spain
Address correspondence to M.A. Rodríguez-Gironés. E-mail: rgirones{at}eeza.csic.es.
Received 20 February 2006; revised 2 June 2006; accepted 12 June 2006.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
We model the optimal foraging strategies for 2 nectarivore species, differing in the length of their proboscis, that exploit the nectar provided by 2 types of flowers, differing in the depths of their corollas. When like flowers appear in clumps, nectarivores must decide whether to forage at a patch of deep or shallow flowers. If nectarivores forage optimally, at least one flower type will be used by a single nectarivore species. Long-tongued foragers will normally visit deep flowers and short-tongued foragers shallow flowers, although extreme asymmetries in metabolic costs may lead to the opposite arrangement. When deep and shallow flowers are randomly interspersed, nectarivores must decide, on encounter with a flower, whether to collect its nectar or continue searching. At low nectarivore densities, the optimal strategy involves exploiting every encountered flower; however, as nectarivore densities increase and resources become scarce, long-tongued individuals should start concentrating on deep flowers and short-tongued individuals on shallow flowers. Therefore, regardless of the spatial distribution of flowers, corolla depth can determine which nectarivore species exploit the nectar from each flower type in a given community. It follows that corolla elongation can evolve as a means to keep nectar thieves at bay if short-tongued visitors are less efficient pollinators than long-tongued visitors.
Key words: competition, habitat selection, nectar concealment.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
When coexisting individuals compete for an ensemble of available resources, variability in their ability to exploit the different resources normally translates, through the operation of optimal foraging, into resource partitioning (Rosenzweig 1981
, 1991). Resource partitioning has been described in natural communities of nectarivores (Heinrich 1976a; Pyke 1982; Harder 1985; Graham and Jones 1996; Irwin 2000; Stang et al. 2006), and the conditions leading to it have been modeled when nectarivores differ in the relative efficiency with which they can exploit co-occurring flower types (Possingham 1992; Rodríguez-Gironés 2006).
Possingham (1992) and Rodríguez-Gironés (2006) consider communities where all visitors can extract the same amount of nectar from flowers but differ in the time they require to do so. In many communities, nectarivores differ in the length of their tongues or proboscis and plants in the depth of their flowers' corolla tubes or spurs, so that flower visitors differ in the amount of nectar they can extract from flowers. In this paper, we consider the conditions under which optimal foraging strategies lead to an association between a nectarivore's proboscis length and the corolla depth of the flowers it visits. In practice, long-tongued visitors can extract more nectar than short-tongued visitors from deep corolla tubes and need less time to do so (Inouye 1980). Nevertheless, to study the extent to which nectar availability per se can lead to resource partitioning, we assume that there is no difference in the amount of time that the different flower visitors need to exploit flowers. (Taking this difference into account would strengthen resource partition.)
Previous models of resource partitioning in nectarivores (Possingham 1992; Rodríguez-Gironés 2006) assume that individuals choose the type of flowers they search. This makes sense when the distribution of flowers is patchy, so that individuals essentially decide where to forage, but it is unclear to what extent the results generalize to the situation where nectarivores come across a sequence of flowers and must decide, for each flower they encounter, whether to visit it or to continue searching. For this reason, we develop 2 foraging models, a patch model and a prey model, to study the effect of the spatial distribution of resources on the expected patterns of resource partitioning.
|
GENERAL MODEL |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
Both models consider a community with 2 nectarivore species and 2 flower types. Flowers differ in the depth of the structure containing nectar (corolla tube), and nectarivores differ in the depth from which they can extract the nectar column (tongue length). Nectarivorous species will be referred to as "bees," although the results apply equally to any other taxa. Flowers are thus divided in shallow and deep flowers, and bees are divided in short- and long-tongued species. The number of i flowers (i = 1 for shallow or 2 for deep flowers) is Fi, and the number of j bees (j = X for short- or Y for long-tongued bees) is Bj. Table 1 lists all variables used.
|
The corolla tube of i flowers is ci mm deep, and the tongue of j bees is tJ mm long. We assume that c1
tX tY c2. On arrival at a flower, j bees consume any nectar within a distance tj of the corolla tube opening.
The rate at which flowers secret nectar depends on factors such as the age of the flower and the time of day (McDade and Weeks 2004). It can decrease as time elapses (Cruden et al. 1983; Castellanos et al. 2002) or remain constant and cease abruptly (Cruden et al. 1983). In our models, we assume that the nectar column in i flowers raises at a constant rate ri until the column is completely full. This is the most conservative assumption, in the sense that it is the least likely to induce resource partitioning.
Depending on the spatial distribution of deep and shallow flowers, bees will encounter uninterrupted bouts of same-type flowers, random sequences of flowers of each type, or something in between. For simplicity, we consider only the 2 extremes of this continuum, and we ignore the possibility that flowers have a tendency to alternate.
In the models considered below, we will assume that bees visit flowers at random and that the duration of the intervals between consecutive visits by any 2 j bees to a randomly chosen i flower follows an exponential distribution with parameter 
ij, which depends on the foraging behavior of the 2 bee species. Given these assumptions, it is possible to determine (Appendix) the average nectar volume that a j bee will consume on arrival at an i flower, Iij:
 | (1) |
 | (2) |
 | (3) |
j. We denote by kjf the metabolic rate of j bees while flying and by kijp their metabolic rate while foraging (probing) at i flowers. To simplify the algebra, it will be convenient to define the net gain between consuming the nectar of an i flower and ignoring it,
 | (4) |
). We further define the profitability of an i flower to a j visitor as eij/hij. |
FLOWERS WITH CLUMPED DISTRIBUTION (PATCH MODEL) |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
First, consider the case in which the distribution of each plant species is clumped and bees exploiting a patch encounter flowers of a single type. Bees must decide whether to forage at patches of shallow or deep flowers. Let pij be the fraction of j bees foraging at i flowers. Within patches, flowers are visited at random, and the duration of the intervals between consecutive visits by j bees on an individual i flower follow an exponential distribution with parameter
ij,
 | (5) |
ij is the rate at which j bees encounter flowers while searching in a patch of i-type flowers.
Except for the difference in expected intake rate (Equations 1–3 of this article vs. Equations 2 and 7 in Possingham [1992]), this model is equivalent to that proposed by Possingham (1992). Following the logic of the ideal free distribution (Fretwell and Lucas 1970), individual bees forage at the flower type on which they maximize their expected payoff (Dreisig 1995; Robertson and Macnair 1995; Ohashi and Yahara 2002). If j bees exploit both shallow and deep flowers, their expected payoff must be the same at both flower types:
 | (6) |
 | (7) |
{0, 1} (Possingham 1992).
Solution types
In Possingham's model, both flower types are always visited by some nectarivores because flowers accumulate an infinite amount of nectar if they remain unvisited (Possingham 1992). When, as in the present model, flowers can only hold a finite volume of nectar, it is possible that a flower type is never visited. Let êij be the maximum (modified) gain that a j bee can obtain when foraging at i flowers. This is obtained by substituting, in Equation 4, Iij for the amount of nectar that a j bee would be able to collect from an i flower full with nectar. Type i flowers will remain unvisited if, when all bees concentrate on the other flower type, i', a hypothetical individual exploiting i flowers obtained a lower payoff than its cospecific. That is, if
 | (8) |
ij), large handling time (hij), low capacity for holding nectar (ici), or high extraction cost (kijp). Reduced competition (fewer nectarivores) will tend to increase the right-hand side of Equation 8 (through its effect on ei'j), leaving its left-hand side unaltered (because êij is independent of the number of nectarivores), and will also favor the existence of unused flowers.
When both flower types are used, there is resource partitioning: at least one flower type is exploited only by one bee species (Possingham 1992).
Correlation between tongue length and corolla depth
Short- and long-tongued bees harvest the same nectar volumes at shallow flowers, whereas long-tongued bees obtain more nectar than short-tongued bees from deep flowers. In general, therefore, long-tongued bees will be relatively more efficient than short-tongued bees at exploiting deep flowers, and when both flower types are exploited, short-tongued bees should exploit shallow flowers and long-tongued bees should exploit deep flowers. However, the correlation between tongue length and corolla depth need not be perfect, depending on the relative abundance of the 2 flower types and the 2 bee species; short-tongued bees may also exploit deep flowers, or long-tongued bees may also exploit shallow flowers. In addition, the profitability of a flower type depends on a number of parameters that might alter the relative efficiency of the 2 bee species.
From Equations 4 and 7 (in particular, because Eij is an increasing function of Iij and ij and a decreasing function of hij and kijp), it is easy to see that for the 2 isolines to intersect, at least one of the following 3 conditions must be satisfied: (1) 1Y – 2Y > 1X – 2X, (2) h1Y – h2Y < h1X – h2X, or (3) k1Yp – k2Yp<k1Xp – k2Xp. These conditions imply that, relative to short-tongued bees, long-tongued bees are more efficient at detecting and exploiting shallow than deep flowers. Equations 4 and 7 also imply that the vertical intercept of the Y isoline increases when 1Y – 2Y increases, when h1Y – h2Y decreases, and when k1Yp – k2Yp decreases. From these two results, it follows that, if 1Y – 2Y 1X – 2X, h1Y – h2Y 
h1X – h2X, and k1Yp – k2Yp k1Xp – k2Xp, the Y isoline is completely under the X isoline. (This is a sufficient, not a necessary condition.) When the Y isoline is below the X isoline, long-tongued bees are preferentially associated with deep flowers: all long-tongued bees forage at deep flowers and/or all short-tongued bees forage at shallow flowers.
Figure 1 shows how asymmetries in the handling time and/or the metabolic costs associated with nectar extraction might swap the sign of the correlation between tongue length and corolla depth. As the metabolic cost of long-tongued bees foraging at deep flowers increases (all other parameters remain fixed), the system shifts from an equilibrium where short-tongued bees forage at shallow flowers and long-tongued bees at deep flowers to a region where there is resource partitioning, but where 2 equilibria are possible, to a region where all short-tongued bees forage at deep flowers and long-tongued bees use both flower types, eventually reaching a point with total (and reversed) habitat segregation, with long-tongued bees foraging at shallow flowers and short-tongued bees foraging at deep flowers.
|
2Y. Each dashed line corresponds to a different value of |
FLOWERS WITH RANDOM DISTRIBUTION (PREY MODEL) |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
In this situation, nectarivores search their environment for individual plants rather than clumps of a favored species. Encounters with shallow and deep flowers follow 2 independent Poisson processes, and
ij is the rate at which a j visitor encounters i flowers. (Notice that ij has slightly different meaning in the 2 models.) On encounter with an i flower, bees must decide whether to exploit it or to continue searching for another flower. If the probability that a j bee exploits an i flower on encounter is qij (the condition q1j + q2j = 1 need not, and generally will not, apply), then the problem involves finding the values of qij that maximize the expected net energy intake rate (or similar currency) of X and Y visitors.
The expected net energy intake rate of a j individual accepting i flowers with probability qij is (Stephens and Krebs 1986)
 | (9) |
 | (10) |
{0, 1} for all i and j, the so-called "zero–one rule" (Stephens and Krebs 1986). Essentially, a prey should be accepted if its profitability exceeds the payoff that the forager can expect to obtain by excluding this prey type from its diet and rejected otherwise. What happens when the profitability of a prey type is exactly the same as the payoff that the forager can expect to obtain by excluding it from the diet? In this case, the forager obtains the same payoff regardless of the proportion of prey of this type it consumes on encounter. The optimal strategy is therefore undetermined and the zero–one rule breaks down. When the profitability of prey is independent of the foraging strategy of predators, only a coincidence can make the profitability of a prey type exactly equal the expected payoff obtained by excluding this prey type from the diet. The probability that this happens when parameter values are chosen at random is zero, and this possibility is therefore regarded as a mathematical curiosity (a so-called "degenerate" scenario), with no biological relevance.
This is not the case when the foraging strategy of predators determines the profitability of prey: what in the standard prey model was a mere "pathology" may now become an equilibrium condition. The model may have an internal solution, if there is at least one qij with 0 < qij < 1, or an external solution if the zero–one rule is satisfied. To find the optimal foraging strategy, we first check whether the model has an external solution. An external solution is a set 
= {q1X, q1Y, q2X, q2Y}, with qij {0, 1} for all i and j, such that: (1) if qij = 0, then the profitability of i flowers to j bees is lower than the expected payoff of j bees, both of them calculated at ; and (2) if qij = 1, then the expected payoff of j bees exceeds the payoff that a hypothetical j bee would obtain if it did not visit j flowers (which is calculated setting qij = 0 in Equation 9, leaving the eij unaltered). There are 16 possible external solutions, but most of them can be automatically discarded because the most profitable flower type must always be included in the diet (Stephens and Krebs 1986).
If the model has no external solution, then the zero–one rule is broken. This can happen because the profitability of a flower type decreases with the frequency with which that flower type is visited. Visiting a flower type may be unprofitable when it is heavily exploited, but if the same flower type is systematically avoided, nectar may accumulate in its corolla until its exploitation becomes worthwhile.
Solution types
Given that q1j + q2j = 1 need not hold, we must work with 4 independent variables. This makes it impractical to use the isoline method described above. Instead, one can find the solution as the equilibrium of a system of differential equations. This is tantamount to assuming that the probabilities of visiting flowers on encounter change through time (t) in the direction that increases intake rate (Mesterton-Gibbons 1992). An ecological equilibrium is a set where no individual bee can increase its intake rate by modifying the values of qij. Given that a system is in state = {q1X, q1Y, q2X, q2Y}, we can define 
ij as
 | (11) |
 | (12) |
> 0 and subject to 0 qij 1. This set of differential equations can be solved numerically. There is no simple relationship between the parameters of the model and the equilibrium values of the qij, and many combinations can be obtained by introducing asymmetries in metabolic costs (as in Figure 1). Typically, when competition is scarce, all bees exploit every flower they encounter. As competition increases, either the long-tongued bees start avoiding shallow flowers or the short-tongued bees start avoiding deep flowers (possibly both). There is a set of conditions for which bees exploit the nonpreferred flowers with finite probability (0 < q1Y < 1 or 0 < q2X < 1), but when competition for resources is sufficiently high, the nonpreferred flowers are totally avoided (q1Y = 0 or q2X = 0). Figure 2 shows 2 examples of the relationship between the ecological equilibrium and the density of bees. Increased metabolic cost of flying shifts curves C and D to the right.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
Differences in the ability of 2 nectarivore species to exploit the nectar of 2 co-occurring flower types will lead to resource partitioning, in the sense that at least one nectarivore will refrain from exploiting one flower type. This result was originally derived by Possingham (1992)
and generalized by Rodríguez-Gironés (2006) when nectarivores differ in the duration of flower visits. The models that we have developed here extend these findings in 2 directions: flower visitors differ in the amount of nectar they can collect from flowers, not in the time they require to exploit them, and foragers may have to decide whether to exploit flowers on encounter.
When foragers choose a patch type to search for food and encounter a single type of flower within that patch, the predictions of the model are essentially the same as those of Possingham (1992): at least one nectarivore will specialize on a single flower type, whereas the other nectarivore may act as a specialist or a generalist, depending on the conditions of the model. Under most conditions, long-tongued bees exploit deep flowers and short-tongued bees shallow flowers (regardless of which bee species, if any, exploits both flower types), but this correlation need not always hold, and it can be reversed by introducing asymmetries in the exploitation costs (Figure 1).
When flowers are encountered at random, there need not be resource partitioning. In particular, if competition for nectar is scarce, long- and short-tongued bees forage indiscriminately at shallow and deep flowers. As competition increases, short-tongued bees start avoiding deep flowers and long-tongued bees shallow flowers, avoidance being complete when competition is sufficiently high. As in the previous case, it is in principle possible to swap the correlation between tongue length and the corolla depth of the preferred flowers by introducing asymmetries in metabolic costs, but for most realistic parameter values, long-tongued bees will prefer deep flowers and short-tongued bees shallow flowers.
Because our results are based on conservative assumptions concerning the pattern of nectar secretion and the relationship between tongue length and handling time at deep flowers, resource partitioning should be more prevalent than the model predicts. In some colonial species, such as honeybees, Apis mellifera, and bumblebees, Bombus spp., individual foragers tend to specialize on a single flower type (Heinrich 1976b). Although the reasons for this specialization lie beyond the scope of this paper (but see Darwin 1876; Laverty 1980; Lewis 1986), specialization implies that individuals are searching for particular flower types. When this is the case, the patch model should apply regardless of the spatial distribution of flowers. Once again, this factor should increase the prevalence of resource partitioning.
Short-tongued bees will stop visiting deep flowers as soon as long-tongued bees keep the nectar column of most flowers beyond their reach. On the other hand, long-tongued bees will always encounter some nectar in shallow flowers, no matter how many short-tongued bees are there, and therefore, competition must be intense before long-tongued bees specialize on deep flowers. In general, although a long tongue can be seen as an evolutionary specialization, at the ecological level, long-tongued bees will normally behave as generalists, whereas the short-tongued bees behave as specialist foragers (Harder 1985; Graham and Jones 1996; Borrell 2005; Stang et al. 2006). In flowers, however, evolutionary and ecological specialization go hand in hand because deep flowers are visited by fewer pollinator species.
It is normally assumed that deep corolla tubes evolved because plants that "compelled the moths to insert their probosces up to the very base, would be best fertilised" (Darwin 1862, p. 202). Although this hypothesis has received considerable empirical support (Nilsson 1988; Johnson and Steiner 1997; Alexandersson and Johnson 2002), there are reasons to doubt the universality of this mechanism: flowers with longer corolla tubes are not always better at exporting or receiving pollen than flowers with shorter corolla tubes (Herrera 1993; Lindberg and Olesen 2001; Lasso and Naranjo 2003). Our results suggest an alternative explanation that corolla elongation may be favored because it deters floral parasites from visiting flowers (Rodríguez-Gironés and Santamaría 2005).
|
APPENDIX |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
Calculating the expected amount of nectar encountered by a j individual at i flowers, Iij
Arrival times of X and Y visitors are independent and exponentially distributed, so if we select an i flower at random, the probability that the time since the departure of the last X visitor is between
X and X + dX, and the time since the departure of the last Y visitor is between Y and Y + dY is
 | (A1) |
X, X) be the amount of nectar within reach of a j visitor at an i flower, given that the time since the departure of the last X visitor was X and the time since the departure of the last Y visitor was Y. The amount of nectar that a j individual can expect to encounter at i flowers, Iij, is:
 | (A2) |
X, X) (measured as the depth of the nectar column consumed) are calculated as follows.
Shallow (i = 1) flowers
Both short- and long-tongued bees can reach the bottom of the corolla tube; hence, every visitor depletes the nectar of the flower, and the amount of nectar a visitor encounters is the amount that has been produced since the last visit, or the amount that the flower can hold if the time that the flower has remained unvisited exceeds the time that the flower requires to refill its corolla tube with nectar. Hence,
 | (A3) |
Deep (i = 2) flowers, short-tongued (X) visitors
For this calculation, we must take into account the fact that short-tongued bees can only consume nectar if the time since the last long-tongued bee visited the flower exceeds (tY – tX)/rB (this is the time required for the nectar column to rise from tY, where it is left by long-tongued visitors, to tX, where short-tongued bees first reach the nectar).
 | (A4) |
Deep (i = 2) flowers, long-tongued (Y) visitors
The calculation is done as before, leading to
 | (A5) |
|
ACKNOWLEDGEMENTS |
|---|
This research was funded by the Spanish Ministry of Science and Technology (project INVASRED, REN2003-06962). We thank Göran Arnqvist and 2 anonymous reviewers for comments on a previous version of the manuscript.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONGENERAL MODELFLOWERS WITH CLUMPED...FLOWERS WITH RANDOM DISTRIBUTION...DISCUSSIONAPPENDIXREFERENCES |
|---|
Alexandersson R and Johnson SD. (2002) Pollinator-mediated selection on flower-tube length in a hawkmoth-pollinated Gladiolus (Iridaceae). Proc R Soc Lond B Biol Sci 269:631–6.[Medline]
Borrell BJ. (2005) Long tongues and loose niches: evolution of euglossine bees and their nectar flowers. Biotropica 37:664–9.[CrossRef]
Castellanos MC, Wilson P, Thomson JD. (2002) Dynamic nectar replenishment in flowers of Penstemon (Scrophulariaceae). Am J Bot 89:111–8.
Cruden RW, Hermann-Parker SM, Peterson S. (1983) Patterns of nectar production and plant-pollinator coevolution. In Elias TS and Bentley BA (Eds.). Biology of nectaries(Columbia University Press, New York) pp. 81–125.
Darwin C. (1862) On the various contrivances by which British and foreign orchids are fertilised by insects and on the good effects of intercrossing. (John Murray, London).
Darwin C. (1876) The effects of cross- and self-fertilisation in the animal kingdom. (John Murray, London).
Dreisig H. (1995) Ideal free distribution of nectar foraging bumblebees. Oikos 72:161–72.[CrossRef]
Fretwell SD and Lucas HL. (1970) On territorial behavior and other factors influencing habitat distribution in birds. I. Theoretical development. Acta Biotheor 19:16–36.
Graham L and Jones KN. (1996) Resource partitioning and per-flower foraging efficiency in two bumble bee species. Am Midl Nat 136:401–6.[CrossRef]
Harder LD. (1985) Morphology as a predictor of flower choice by bumble bees. Ecology 66:198–210.[CrossRef]
Heinrich B. (1976a) Resource partitioning among some eusocial insects: bumblebees. Ecology 57:874–89.[CrossRef]
Heinrich B. (1976b) The foraging specializations of individual bumblebees. Ecol Monogr 46:105–28.[CrossRef]
Herrera CM. (1993) Selection on floral morphology and environmental determinants of fecundity in a hawkmoth-pollinated violet. Ecol Monogr 63:251–75.[CrossRef]
Inouye DW. (1980) The effect of proboscis and corolla tube lengths on patterns and rates of flower visitation by bumblebees. Oecologia 45:197–201.[CrossRef]
Irwin RE. (2000) Hummingbird avoidance of nectar-robbed plants: spatial location or visual cues. Oikos 91:499–506.[CrossRef]
Johnson SD and Steiner KE. (1997) Long-tongued fly pollination and evolution of floral spur length in the Disa draconis complex (Orchidaceae). Evolution 51:45–53.[CrossRef][ISI]
Lasso E and Naranjo ME. (2003) Effects of pollinators and nectar robbers on nectar production and pollen deposition in Hamelia patens (Rubiaceae). Biotropica 35:57–66.[CrossRef][Medline]
Laverty TM. (1980) The flower-visiting behaviour of bumble bees: floral complexity and learning. Can J Zool 58:1324–35.
Lewis AC. (1986) Memory constraints and flower choice in Pieris rapae. Science 232:863–5.
Lindberg AB and Olesen JM. (2001) The fragility of extreme specialization: Passiflora mixta and its pollinating hummingbird Ensifera ensifera. J Trop Ecol 17:323–9.[CrossRef]
McDade LA and Weeks JA. (2004) Nectar in hummingbird-pollinated neotropical plants I: patterns of production and variability in 12 species. Biotropica 36:196–215.[ISI]
Mesterton-Gibbons M. (1992) Games nectarivores play: a perspective on Possingham. Ecology 73:1913–5.[CrossRef]
Nilsson LA. (1988) The evolution of flowers with deep corolla tubes. Nature 334:147–9.[CrossRef]
Ohasi K and Yahara T. (2002) Visit larger displays but probe proportionally fewer flowers: counterintuitive behaviour of nectar-collecting bumble bees achieves an ideal free distribution. Funct Ecol 16:492–503.[CrossRef]
Possingham HP. (1992) Habitat selection by two species of nectarivore: habitat quality isolines. Ecology 73:1903–12.[CrossRef]
Pyke GH. (1982) Local geographic distributions of bumblebees near Crested Butte, Colorado: competition and community structure. Ecology 63:555–73.[CrossRef]
Robertson AW and MacNair MR. (1995) The effects of floral display size on pollinator service to individual flowers of Myosotis and Mimulus. Oikos 72:106–14.[CrossRef]
Rodríguez-Gironés MA. (2006) Resource partitioning among flower visitors: extensions of Possingham's model. Evol Ecol Res 8:765–83.
Rodríguez-Gironés MA and Santamaría L. (2005) Resource partitioning among flower visitors and the evolution of nectar concealment in multi-species communities. Proc R Soc Lond B Biol Sci 272:187–92.[Medline]
Rosenzweig ML. (1981) A theory of habitat selection. Ecology 62:327–35.[CrossRef]
Rosenzweig ML. (1991) Habitat selection and population interactions—the search for mechanism. Am Nat 137:S5–28.[CrossRef]
Stang M, Klinkhamer PGL, van der Meijden E. (2006) Size constraints and flower abundance determines the number of interactions in a plant-flower visitor web. Oikos 112:111–21.[CrossRef][ISI]
Stephens DW and Krebs JR. (1986) Foraging theory. (Princeton University Press, Princeton NJ).
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||