Surplus Nectar Available for Subalpine Bumble Bee Colony Growth

PLANTÐINSECT INTERACTIONS Surplus Nectar Available for Subalpine Bumble Bee Colony Growth

1,2 SUSAN E. ELLIOTT

Environ. Entomol. 38(6): 1680Ð1689 (2009) ABSTRACT Mutualisms may cause coupled population expansion or decline if both partners re- spond to variation in the otherÕs abundance. Many studies have shown how the abundance of animal mutualists affects plant reproduction, but less is known about how the abundance of plant mutualists affects animal reproduction. Over 2 yr, I compared reproduction of the bumble bee, Bombus appositus, across meadows that varied naturally in ßower density, and I compared reproduction between fed colonies and unfed control colonies. Colony reproduction (gyne, worker, and male production) was constant across meadows that varied naturally in ßower density. Forager densities per ßower did not vary among meadows, and daily nectar depletion was consistently low across meadows, suggesting that bees had ample nectar in all meadows. However, colonies directly fed with supplemental nectar and pollen generally produced over twice as many gynes as control colonies. Feeding did not affect male or worker production. Although colonies may beneÞt from food supplementation at the nest, it is possible that they may not beneÞt from additional ßowers because they have too few workers to collect extra resources.

KEY WORDS Bombus appositus, bumble bee, Delphinium barbeyi, food limitation, ßower density

Mutualisms are often considered from a unilateral and the costs of partner exploitation (Johnson and perspective, emphasizing the more sessile or larger Steiner 1997, Herrera 2000, Morales 2000, Egger and partner, whose Þtness is often easier to measure (e.g., Hibbett 2004, Ness et al. 2004, Rudgers and Strauss the plant in a plant-pollinator mutualism: Cushman 2004). In plantÐpollinator mutualisms, unless the pol- and Beattie 1991). Knowing how a visitor affects a host linator also acts as a seed predator, there is likely little mutualist can help explain the ecology and evolution to no cost of extra pollinator visits for seed production of the host, but if the visitor responds to changes in its (but see Young and Young 1992). Similarly, there is partner species (e.g., by an increase or decrease in probably no cost to bees for living in areas where there abundance), this could alter the nature (and our in- are surplus ßowers unless bee parasites are positively terpretation) of their relationship (Bronstein 1994b, correlated with ßower density (Carvell et al. 2008). Thompson 2005). For example, little is known about The beneÞts of pollinator abundance to female plant how the Þtness or population growth of mutualistic reproduction and changes in population growth mycorrhizal fungi, nitrogen-Þxing microbes, and pho- (through effects on seed production) may saturate tosynthetic zooxanthellae vary with the frequency of when abiotic resources become limiting for seed pro- their mutualistic partners (Bever 1999, Simms and duction (Burd 1994, Ashman et al. 2004). Plants may Taylor 2002, Hay et al. 2004). In plantÐpollinator mu- compensate for pollinator shortages if they can self- tualisms, ßowering plant reproduction often varies pollinate without inbreeding costs, reproduce asexu- with the number and diversity of pollinators (Pellmyr ally, or endure unfavorable periods via longevity or and Thompson 1996, Gomez et al. 2007, Sahli and dormancy by living longer or having seed banks (Pake Conner 2007). However, the sensitivity of pollinator and Venable 1996, Morgan et al. 2005). In contrast, reproduction to changes in ßoral resources or other most insect pollinators are short-lived organisms, so environmental variables is poorly understood, espe- they must reproduce even when resources are scarce. cially in North America (National Resource Council With very few exceptions, bee pollinators cannot re- 2006). produce without harvesting pollen and nectar to feed The outcomes of mutualisms are often highly con- themselves and their offspring (Michener 2007). text dependent (Bronstein 1994a). The beneÞts of However, is pollinator per-capita reproduction food- mutualistic associations can vary with partner quality, limited? resource availability, the presence or absence of pred- Positive correlations between bee and ßower den- ators, competition for access to a mutualistic partner, sities could indicate that bees have higher reproduction, recruitment, or survival in areas with more ßow- ers (Steffan-Dewenter et al. 2002, Westphal et al. 1 Corresponding author: Pinyon Publishing, Montrose, CO 81403 (e-mail: [email protected]). 2006), but these relationships could also be inßuenced 2 Dartmouth College, Biology, Hanover, NH 03755. by other limiting factors such as parasitism and nest

0046-225X/09/1680Ð1689$04.00/0 ᭧ 2009 Entomological Society of America December 2009 ELLIOTT:BUMBLE BEE NECTAR SURPLUSES 1681

Fig. 1. Study meadows (a) in the East River Valley, Gunnison National Forest, CO, and total number of D. barbeyi inßorescences (b) in 200-m radii around each captive colony location (1Ð3; GPS coordinates ϭ 1: 2Њ56Ј8Љ E, 38Њ49Ј33Љ N; 2: 2Њ56Ј12Љ E, 38Њ49Ј21Љ N; 3: 2Њ56Ј17Љ E, 38Њ49Ј13Љ N). site availability. Recent studies from low-elevation ar- male offspring production) was limited by ßoral re- eas (Ͻ200 m) conÞrm that bee reproduction of soli- sources, by asking the following questions. tary and colonial bee species is greater in areas with 1. Does colony reproduction vary across meadows higher natural or experimental levels of ßoral re- that vary naturally in ßower density? If bee colony sources (Goulson et al. 2002, Pelletier and McNeil density is independent of ßower density and if 2003, Greenleaf 2005, Williams and Kremen 2007, colony reproduction is food-limited, colonies in Carvell et al. 2008). However, because bees at high meadows with more ßowers should produce more elevations have shorter growing seasons and conse- offspring, assuming that ßower number represents quently less time to establish nests, collect resources, pollen and nectar availability (Tepedino and Stan- and produce offspring, their reproduction may instead ton 1982). Alternatively, if meadows with more be time-limited (Pyke 1982). To determine whether ßowers have disproportionately more colonies, subalpine bumble bees have surplus ßowers for colony ßower availability per colony might be lower in reproduction, I measured bumble bee reproduction meadows with more total ßowers. If ßower avail- across meadows that varied naturally in ßower density ability per colony is greater in some meadows, and in response to supplemental feeding at the nest. ßowers in those meadows should have lower per- Time limitation of bumble bee reproduction was not ßower pollinator visitation rates, and nectar in addressed experimentally but is discussed within the those meadows should be depleted less quickly context of the experimental Þndings of this study. than in meadows with more ßowers per colony. Although bumble bees visit a variety of ßowers, they 2. Is bee colony reproduction food-limited? If bee may be functionally specialized if they derive most of reproduction is food-limited, colonies that are fed their resources from the most rewarding ßower spe- at the nest with supplemental nectar and pollen cies (Dramstad and Fry 1995, Stenstrom and Bergman should have higher reproduction than unfed con- 1998, Goulson and Darvill 2004). The focal bumble bee trol colonies. Also, fed colonies may reduce for- in this study, Bombus appositus L. (Apidae) may visit aging to avoid costs associated with foraging out- at least 20 ßower species throughout the season in the side the colony (Dukas and Morse 2003) or to study area (Pyke 1982), but it collects much of its beneÞt the colony by devoting more time to brood pollen and nectar from the abundant perennial wild- care or colony defense (Cartar 1992). ßower, Delphinium barbeyi Huth (Ranunculaceae) (Elliott 2009). Delphinium barbeyi, the second most Materials and Methods abundant ßower species in these meadows, is in bloom in the middle of season while B. appositus colonies are In 2006 and 2007, I studied reproduction in the provisioning and producing adult workers, gynes, and long-tongued bumble bee, B. appositus (Apidae), in males. During this time, D. barbeyi accounts for the subalpine meadows in the East River Valley near the majority of pollen collection and ßower visitation by Rocky Mountain Biological Laboratory in Gothic, CO B. appositus (described below). I tested the hypothesis (2,952- to 2,969-m elevation; Fig. 1). B. appositus is one that B. appositus reproduction (i.e., worker, gyne, and of the three most abundant bumble bee species in this 1682 ENVIRONMENTAL ENTOMOLOGY Vol. 38, no. 6 study area (Pyke 1982, Elliott 2009). Inseminated len collection by B. appositus is dominated by these bumble bee queens emerge from hibernation in early two species (S1). I used a one-way analysis of variance spring (late May to early June) and establish new (ANOVA) to compare the proportions of cream and nests, where they incubate their Þrst brood. About 6 orange pollen loads carried by foragers returning to wk later, the Þrst worker cohort emerges from their the colonies (described below) among the three study cocoons, and forager density increases sharply. During meadows, using colony as the unit of replication. this increase in bee abundance, the common perennial Bumble Bee Colonies. In 2006 and 2007, I moni- wildßower, Delphinium barbeyi (Ranunculaceae), ac- tored captive B. appositus colony reproduction in the counts for 91.2% of the pollen collected by B. appositus three focal meadows. To obtain colonies, I set out and 94.6% of the ßoral visits made by B. appositus Ϸ100 wooden nest boxes in the study meadows, pro- (median contributions, S1). viding each box with cotton for nest insulation. Each In the study area, D. barbeyi occurs in meadows and year, Ϸ10% of boxes succeeded in attracting emerging forest clearings in distinct patches, and D. barbeyi B. appositus queens from the study meadows to the ßower density varies over three orders of magnitude nest boxes. I also started colonies in the laboratory among 0.5-ha meadow plots (0.8Ð26.2 ßowers pro- using nest-searching queens collected from the study duced/m2; Elliott 2008). With 13.6 Ϯ 0.5 D. barbeyi meadows (2006: three colonies, one per meadow, inßorescences per plant (mean Ϯ SE, n ϭ 420 plants), 2007: one colony, used in one meadow only). As soon each bearing 25.4 Ϯ 0.8 ßowers per inßorescence (n ϭ as Þeld or laboratory colonies laid their Þrst brood, I 372 inßorescences), D. barbeyi accounts for 14.0 Ϯ 4% distributed the colonies evenly among the three study of all ßowers produced (mean Ϯ SE, n ϭ 6 meadows; meadows. In each study meadow, all colonies were Elliott 2008). On average, ßowers contain 1.8 Ϯ 0.05 ␮l placed in the same fenced and covered shelter (to of nectar per ßower in the morning (before pollinator reduce damage from porcupine, bear, and rain). visits begin, n ϭ 512 ßowers) with 36.1 Ϯ 0.7% sugar Therefore, this study tests the effects of ßower density concentration (n ϭ 34 ßowers; Elliott 2008). Flowers on colony growth and reproduction (as in Goulson et are protandrous, with anthers dehiscing over a period al. 2002, Pelletier and McNeil 2003, Thomson 2004, of ϳ3 d (unpublished data). Greenleaf 2005) and not on colony establishment. The queens in the captive colonies fed their offspring with nectar and pollen obtained in the meadows. The study 1. Does Colony Reproduction Vary Across Meadows colonies contributed to a small proportion of ambient That Vary Naturally in Flower Density? bee density (S2). The shelters were fenced enclosures I compared B. appositus colony reproduction in that reduced, but did not completely eliminate, por- three meadows that varied naturally in D. barbeyi cupine and bear disturbances. After excluding dis- density. The meadows were separated by Ϸ400 m turbed colonies, there were seven (2006) and eight (Fig. 1). Although large-bodied bees like B. appositus (2007) control colonies (Fig. 2). Once per week at are physiologically able to ßy up to 9.8 km (Goulson night, I counted the number of individuals of each and Stout 2001), in this study system, bumble bees caste present in each colony. forage primarily within a 100-m radius (Elliott 2009). Colony Reproduction. After colonies were aban- Therefore, bees were unlikely to ßy among focal doned at the end of the season, I quantiÞed three meadows, each separated by Ͼ300 m. In addition, all aspects of offspring production per colony: number of bees in this study that were marked at the colony and gyne cocoons, number of worker cocoons, and num- relocated in foraging meadows were found in the ber of males. I counted all cocoons produced (which meadows nearest their respective colonies even remain intact after adult individuals emerge), and I though all meadows were sampled equally, suggesting measured the diameter of each cocoon to the nearest that they did not ßy among study meadows. 0.1 mm. Cocoons that had housed gynes were roughly Flower Density. In 2007, I counted the number of three times larger than worker or male cocoons, as D. barbeyi ßowering inßorescences in a 200-m radius with most Bombus species that feed offspring individ- around each shelter that housed captive bee colonies ually (Goulson 2003). Gyne cocoons ranged from 9.6 (one shelter per focal meadow; Fig. 1). Flower num- to 18.8 mm in diameter and nonqueen cocoons ranged ber per inßorescence did not vary with inßorescence from 5.6 to 9.9 mm in diameter. For intermediate sizes, density (r ϭ 0.09, P ϭ 0.8, n ϭ 10 meadows, 60 inßo- I assigned the caste based on the cocoon size, relative rescences averaged per meadow). This analysis and all to the other cocoons in the colony. I measured gyne subsequent analyses were performed with JMP v. 4.04 cocoon diameter as a proxy for gyne body size, which (SAS Institute 2001). Although D. barbeyi density is can affect diapause survival (Beekman et al. 1998). I correlated with total ßower density in these meadows, quantiÞed male production as the sum of all males the most abundant ßower species, Potentilla pulcher- seen in the colonies during weekly night censuses. rima (Rosaceae) only contributed to 2.6% of all B. This is a conservative estimate of male production appositus visits and therefore probably does not con- because males leave the colonies 2Ð4 d after emerging tribute strongly to colony reproduction (Elliott 2008, from their cocoons (Kearns and Thomson 2001), and 2009). In 2007, I compared corbicular pollen loads I only visited colonies once per week. I estimated carried by B. appositus in each meadow. Potentialla worker production as the number of small (nongyne) pulcherrima pollen is bright orange and easily distin- cocoons minus the number of males observed during guished from D. barbeyi’s cream-colored pollen. Pol- night censuses. Therefore, worker production is December 2009 ELLIOTT:BUMBLE BEE NECTAR SURPLUSES 1683

Fig. 2. Number of live B. appositus individuals observed per colony during night censuses over the season in three study meadows (see Fig. 1) for control colonies (open) and fed (closed) colonies (aÐf). Arrows point to disturbances [P ϭ porcupine, U ϭ nest usurpation by U(A) (B. appositus) or U(P) (Psithyrus sp.), and B ϭ bear damage]. The ? points to a colony that was found after the start of the study so its starting size is unknown. In 2007, the fed colony with bear damage had the same colony size and is thus plotted over a control colony without damage. slightly inßated because I probably missed seeing ows. A sample size of 12 observation bouts per meadow some males if they dispersed between censuses. With provides sufÞcient power to detect among-meadow vari- ϭ ϭ ϭ the exception of successful social nest parasitism, ation in visitation rates (F1,11 2.9, P 0.002, n 12 bumble bees that enter colonies that are not their own meadows). Before each 15-min observation period, I are killed or chased out by resident bees (Kearns and recorded the number of open inßorescences I was ob- Thomson 2001). Thus, all workers, males, and new serving. I also counted the number of open ßowers per gynes observed in the colony were most likely pro- inßorescence from 20 inßorescences per meadow to cal- duced by that colony. culate per-ßower pollinator visitation rates. I recorded I used multivariate analyses of variance (MANOVAs; visits by all bee and bird species, although B. appositus one for each year) to test whether the number of made up the majority of all ßower visits (S3). For each gynes, workers, and males per colony varied among visitor, I recorded the species and the number of ßowers meadows. I did not use ßower density as a covariate they visited. I used a one-way ANOVA to test whether because by using meadow as a categorical predictor B. appositus per-ßower visitation rates (i.e., percent of variable, I could detect differences among meadows open ßowers visited by B. appositus per minute) varied that may not vary linearly with ßower density. I used among the three study meadows, using observation pe- a MANOVA because reproduction among castes may riod as the unit of replication. be correlated (Pelletier and McNeil 2003). Given a In 2007, to determine whether per-ßower nectar avail- marginally signiÞcant effect in 1 yr, I used separate ability varied among meadows, I compared pollinator one-way ANOVAs to test how the number of gynes, visitation rates per ßower (all pollinator species) and workers, and males per colony varied among mead- daily nectar depletion per ßower across the three meadows. I could not test for meadow effects on average ows. I collected ßowers to measure nectar availability gyne cocoon diameter because some meadows had and depletion on 3 d (5, 18, and 25 July) in the middle only one colony that produced gynes. I Þrst separated of the D. barbeyi blooming period. Each day, I collected analyses by year and then combined years to assess ßowers from different plants spread throughout each overall effects. meadow in the mornings (before 0800 hours when D. Foraging Behavior and Nectar Availability. In 2007, to barbeyi pollinators become active) and in the late after- determine whether foragers were distributed in propor- noons (after 1600 hours when most foraging on D. bar- tion to ßower availability, I calculated ßower visitation beyi has slowed or stopped). I collected 20 ßowers per rates in Ϸ20-m2 patches (12 15-min observation bouts meadow at each time period on each day, but I excluded between 0900 and 1500 hours, spread over 3 wk in the ßowers that were damaged so total sample size per middle of the D. barbeyi blooming period). I recorded meadow varied from 32 to 48 morning ßowers and 24Ð38 visitation rates in the three focal meadows plus nine afternoon ßowers per meadow (total n ϭ 217 ßowers). adjacent meadows. However, because I never saw I picked ßowers from the middle of inßorescences to marked (captive colony) bees in the surrounding mead- standardize ßower age. I kept the ßowers in a cooler and ows, I only report visitation rates in the three focal mead- extracted the nectar (from the two nectar spurs per 1684 ENVIRONMENTAL ENTOMOLOGY Vol. 38, no. 6

ßower) within 24 h, using 2-␮l capillary tubes. Nectar but for worker production, I could not assume equal reabsorption in unvisited ßowers in D. barbeyi has not variances, so I did not pool variances. been reported in published studies. To determine To determine whether individuals in fed colonies for- whether meadows varied in nectar availability and aged more frequently (e.g., because of increased energy whether nectar availability dropped over the course of a to fuel foraging efforts) or less frequently (e.g., to reduce day, I used a two-way ANOVA to test the effects of foraging costs) than unfed control colonies, I compared meadow and time period (morning versus afternoon) on foraging activity of fed and control colonies. In 2006, I nectar volume per ßower. I also tested whether nectar recorded foraging activity for 30 min per colony on 1 d depletion varied among meadows (meadow ϫ time pe- toward the end D. barbeyi’s blooming period. In 2007, I riod interaction), using ßower as the unit of replication. increased observation time to 12 30-min observation periods per colony, spread over 2 wk at peak D. barbeyi bloom. For each colony, I divided the total number of 2. Is Bee Colony Reproduction Food-Limited? entries into and exits out of the colony by the number of In 2006 and 2007, to test whether colonies were food- individuals and by the 30 min of observation time to limited, I fed two colonies (one from each of the Þrst two calculate an activity rate per individual per minute. I used meadows) and compared their reproduction and activity two-tailed t-tests to compare activity rates between fed levels with the control (unfed) colonies (described and control colonies, with colony as the unit of replica- above). Because of a limited sample size, I did not feed tion, analyzing each year separately. multiple colonies to maximize control colony sample size to establish baseline variation in reproduction across meadows and to avoid inßating forager number (and Results resource depletion) by feeding multiple colonies. I also 1. Does Colony Reproduction Vary Across Meadows fed one colony in the third meadow each year, but they That Vary Naturally in Flower Density? were destroyed by porcupine and bear (2006 and 2007, respectively) early in their development and were there- Despite Þve-fold variation in ßower abundance fore excluded from analyses. Each week, I fed the col- (Fig. 1), bee reproduction was relatively constant onies with 50 ml of sugar-water (50:50 sucrose and water, among meadows (2006: F ϭ 0.4; df ϭ 6,4; P ϭ 0.9; 2007: with 1 ml of honey mixed in) with two plastic pipette F ϭ 5.7; df ϭ 6,4; P ϭ 0.056; Fig. 3; Table 1). During feeders. The colonies typically drained the feeders in 3 d. both years, control colonies produced 0Ð12 gynes I added Colorado wildßower honey (Ambrosia Honey, (median ϭ 0 gynes per colony), and average gyne Parachute, CO) so that natural nonsugar constituents production did not vary among meadows (2006: F ϭ would also be present (Baker and Baker 1982). The 50-ml 0.4; df ϭ 2,4; P ϭ 0.7; 2007: F ϭ 0.5; df ϭ 2,5; P ϭ 0.6; nectar volume is comparable to the total amount of both years: F ϭ 0.5; df ϭ 2,12; P ϭ 0.6; Fig. 3a; Table nectar held within one third of all of the D. barbeyi 1). Colonies produced 0Ð27 workers (median ϭ 11 ßowers in an average meadow (given an average of 3,856 workers per colony), and average worker production inßorescences per meadow ϫ 25 ßowers per inßores- did not vary across meadows for both years combined cence ϫ 1.8 ␮l of nectar per ßower). In 2007, to deter- (F ϭ 0.5; df ϭ 2,12; P ϭ 0.6; Fig. 3b; Table 1) and in 2006 mine whether colonies were limited by both nectar and alone (F ϭ 0.1; df ϭ 2,4; P ϭ 0.9; Fig. 3b; Table 1). pollen, I also gave the two fed colonies supplemental However, in 2007, colonies in the meadow with in- pollen. Each week, I gave the colonies Ϸ500 mg of honey termediate-high ßower density (Fig. 1) produced beeÐcollected pollen moistened with artiÞcial nectar marginally more workers than colonies in the other (Kearns and Thomson 2001), which is slightly more pol- two meadows (F ϭ 6.1; df ϭ 2,5; P ϭ 0.06; Fig. 3b; Table len than all of the workers in a colony would bring in on 1). Males were seen in 5 of 16 control colonies during an average day (19 Ϯ 5 corbicular pollen loads per day; night censuses from the 2 yr, with a maximum of Þve unpublished data; 21 mg pollen per load, Heinrich 1979). males observed per colony (median ϭ 0 males per I used pollen collected from honey bees that were for- colony). In 2006, very few males were seen in any of aging in similar subalpine habitats in Colorado. The food the meadows, but in 2007, males were seen more was provided inside the colonies so individuals from frequently in the meadow with intermediate-high other colonies could not collect or consume it. ßower density (2006: F ϭ 0.1; df ϭ 2,4; P ϭ 0.9; 2007: I compared four metrics of offspring production F ϭ 7.0; df ϭ 2,5; P ϭ 0.04; both years: F ϭ 2.7; df ϭ 2,12; between fed and control colonies: number of gyne P ϭ 0.1; Fig. 3c; Table 1). cocoons, worker cocoons, and males censused, and In 2007, B. appositus ßower visitation rate (i.e., per- gyne cocoon diameter (described above). First, I used cent of D. barbeyi ßowers visited by B. appositus per MANOVAs (one for each year) to test whether the minute) did not vary among meadows (F ϭ 1.1; df ϭ number of gynes, workers, and males per colony var- 2,33; P ϭ 0.4; Fig. 4). Similarly, ßower visitation rate of ied among meadows. Given signiÞcant differences in all visitor species combined (S3) did not vary among both years, I tested the effect of feeding on these four meadows (F ϭ 0.9; df ϭ 2,33; P ϭ 0.4; Fig. 4). Although variables with separate two-tailed t-tests for each year nectar volume per ßower was highest in the meadow and for both years combined. For males and gynes, I with the fewest D. barbeyi ßowers (F ϭ 7.6; df ϭ 2,211; used pooled estimates of variance for the two treat- P ϭ 0.02; Fig. 4), nectar depletion did not vary among ments (LeveneÕs test of the null hypothesis that equal meadows (meadow ϫ time period: F ϭ 1.7, df ϭ 2,211; variance are equal were all nonsigniÞcant at ␣ ϭ 0.05), P ϭ 0.2; Fig. 4). Available nectar decreased by 28.4% December 2009 ELLIOTT:BUMBLE BEE NECTAR SURPLUSES 1685

Fig. 3. Bombus appositus colony reproduction across meadows (unfed control colonies; aÐc) and between control and fed treatments (dÐf) in terms of queens (a and d), workers (b and e), and males (c and f) produced per colony. over the course of the day (time effect: F ϭ 14.5; df ϭ 2. Is Bee Colony Reproduction Food-Limited? 1,211; P ϭ 0.0002; Fig. 4). In individual years, there were some signiÞcant dif- The proportion of cream-colored pollen loads brought back by foragers, indicating D. barbeyi pollen ferences in reproduction between control and fed ϭ ϭ ϭ ϭ use instead of orange P. pulcherrima pollen, varied colonies (2006: F 8.0; df 3,5; P 0.02; 2007: F ϭ ϭ among meadows (F ϭ 13.3; df ϭ 2,6; P ϭ 0.006). The 8.5; df 3,5; P 0.02; Fig. 3; Table 1). Although all four lowest D. barbeyi use was in the meadow with the most fed colonies produced gynes, only 8 of the 15 control D. barbeyi ßowers, which was also the meadow with colonies produced gynes (Fig. 3d). In 2006, fed col- intermediate P. pulcherrima density in 2004 (mean Ϯ onies produced Þve times more gynes than control SE percent cream pollen loads per colony: meadow colonies (t ϭ 3.0; df ϭ 7; P ϭ 0.02; Table 1). However, 1 ϭ 42.7 Ϯ 7.5%, meadow 2 ϭ 85.6 Ϯ 6.5%, meadow 3 ϭ in 2007, feeding did not signiÞcantly affect gyne pro- 96.4 Ϯ 9.2%). duction (t ϭ 0.3; df ϭ 9; P ϭ 0.8; Table 1). The

Table 1. Mean ؎ SE B. appositus offspring production (queen cocoons, worker cocoons, and male sightings) compared across three low density; see Fig. 1) and between fed andcontrol (unfed) colonies ؍ medium-high density, 3 ؍ high density, 2 ؍ subalpine meadows (1

Meadows or Queen Worker Male Factor Year N colonies feeding cocoons cocoons sightings Meadows 2006 1 2 1.5 Ϯ 1.5 18.0 Ϯ 9.0 0.0 Ϯ 0.0 2 2 0.0 Ϯ 0.0 13.0 Ϯ 8.0 0.0 Ϯ 0.0 3 3 1.3 Ϯ 1.3 15.0 Ϯ 3.6 0.7 Ϯ 0.7 2007 1 2 0.0 Ϯ 0.0 5.0 Ϯ 1.0 0.0 Ϯ 0.0 2 3 1.7 Ϯ 0.7 16.3 Ϯ 4.4 3.7 Ϯ 0.9 3 3 4.0 Ϯ 4.0 7.3 Ϯ 7.3 0.7 Ϯ 0.7 Both years 1 4 0.8 Ϯ 0.8 11.5 Ϯ 5.3 0.0 Ϯ 0.0 2 5 1.0 Ϯ 0.5 15.0 Ϯ 3.6 2.2 Ϯ 1.0 3 6 2.7 Ϯ 2.0 11.2 Ϯ 4.0 0.7 Ϯ 0.4 Fed versus control 2006 F 2 5.0 Ϯ 1.0 14.0 Ϯ 1.0 1.5 Ϯ 0.5 C 7 1.0 Ϯ 0.7 15.3 Ϯ 3.1 0.3 Ϯ 0.3 2007 F 2 3.0 Ϯ 0.0 16.5 Ϯ 1.5 1.6 Ϯ 0.7 C 8 2.1 Ϯ 1.5 10.1 Ϯ 3.4 1.5 Ϯ 0.5 Both years F 4 4.0 Ϯ 0.7 15.3 Ϯ 1.0 1.5 Ϯ 0.3 C 15 1.6 Ϯ 0.8 12.5 Ϯ 2.3 1.0 Ϯ 0.4 1686 ENVIRONMENTAL ENTOMOLOGY Vol. 38, no. 6

Discussion In contrast to the wealth of studies that test for pollen limitation of plant reproduction, very few studies have tested for ßower limitation of pollinator reproduction, creating a skewed understanding of plantÐpollinator mutualisms. A handful of studies from low elevations suggest that bumble bee pollinator reproduction is limited by ßoral resources (Goulson et al. 2002, Pelletier and McNeil 2003, Greenleaf 2005). However, in subalpine meadows, although bumble bee reproduction was marginally food-limited (i.e., fed colonies generally produced more gynes), there was no evidence that reproduction was ßower-limited (i.e., nectar was not depleted to a large degree and reproduction did not vary across a natural gradient in ßower density). One explanation for this pattern may be that unless bumble bee colonies have ample workers, they may not be able to take full advantage of abundant ßowers. Natural ßower density may not have affected col- Fig. 4. Comparison of pollinator abundance and nectar ony reproduction if colony density increased propor- availability per ßower for D. barbeyi growing in three sub- tionally with ßower density, resulting in areas with alpine meadows in Gunnison National Forest, CO (in de- more ßowers harboring more foragers (Steffan-Dew- creasing order of ßower abundance, see Fig. 1) showing (a) enter et al. 2002, Westphal et al. 2006). However, ßower visitation rate (percent of ßowers visited per minute) and (b) ambient nectar volume per ßower measured before relationships between forager and ßower densities (AM) and after (PM) pollinator foraging. Error bars repre- may not hold if pollen and nectar production per sent ϮSE around meadow averages. SigniÞcance levels for ßower varies across habitats (Cartar 2004, Goulson et among-meadow comparisons are reported in text. al. 2007). In 2007, forager abundance per D. barbeyi ßower was constant across meadows, conÞrming that nonsigniÞcant effect of feeding on gyne production forager density was proportional to ßower density. for both years combined was caused by a control Also, although recapture rates were low overall, colony in 2007 with very high reproduction (i.e., 12 marked bees from the captive colonies contributed to gynes; with this outlier excluded, t ϭ 3.9; df ϭ 16; P ϭ a greater proportion of ßower visitors in the meadow 0.001; Fig. 3d). Feeding did not affect gyne cocoon with the fewest D. barbeyi ßowers (S2). Therefore, diameter in either year (2006: control ϭ 10.9 Ϯ 0.5 mm, there may have been fewer natural bee colonies where fed ϭ 11.9 Ϯ 0.6 mm, t ϭϪ1.2; df ϭ 5; P ϭ 0.3; 2007: there were fewer ßowers. However, more extensive control ϭ 11.7 Ϯ 0.2, fed ϭ 11.5 Ϯ 0.3 mm, t ϭ 0.4; df ϭ markÐrecapture studies are needed to test the hypoth- 4; P ϭ 0.7; both years: control ϭ 11.3 Ϯ 0.3 mm, fed ϭ esis that natural bumble bee colony densities are pro- 11.7 Ϯ 0.4 mm, t ϭ 0.9; df ϭ 9; P ϭ 0.4). Worker portional to ßower densities. production also did not vary between control and fed Colony reproduction also may have been constant colonies in either year (2006: t ϭ 0.4; df ϭ 6.9; P ϭ 0.7; across meadows if all meadows had similar nectar (and 2007: t ϭ 1.8; df ϭ 7.9; P ϭ 0.1; Fig. 3e; Table 1). Males pollen) availability. Because daily D. barbeyi nectar were found in the colonies at night in all four fed depletion was minimal, it is unlikely that bees ex- colonies, yet they were only observed in one third of hausted the available nectar in 2007. For example, in the control colonies (Fig. 3f). However, there were no 2005, bees depleted D. barbeyi nectar from a starting statistically signiÞcant differences in the number of level of 0.68 Ϯ 0.03 ␮l per ßower down to 0.23 Ϯ 0.04 males censused between control and fed colonies in ␮l per ßower over the course of the day (Elliott 2008). either year (2006: t ϭ 2.0; df ϭ 7; P ϭ 0.08; 2007: t ϭ Therefore, the average 1.3 Ϯ 0.1 ␮l of nectar remaining 0.8; df ϭ 7; P ϭ 0.9; Table 1). With both years com- in D. barbeyi ßowers at the end of the day in 2007 was bined, there were no statistically signiÞcant effects of probably not too deep within the nectar spur for B. feeding on gyne, worker, or male production (gynes: appositus to reach. The three nectar sampling dates t ϭ 1.4, df ϭ 17; P ϭ 0.2; workers: t ϭ 1.1; df ϭ 17; P ϭ included the period of colony growth up to peak 0.3; males: t ϭ 0.6, df ϭ 17; P ϭ 0.6; Fig. 3dÐf; Table 1). colony size (Fig. 1, Þrst three census dates). Later in In 2006, individuals in fed colonies made more en- the season, if less nectar was available, it would have tries and exits in and out of their nest boxes than primarily affected eclosion, but not provisioning, of individuals in control colonies (mean Ϯ SE activity per gynes and males. However, a greater proportion of individual: fed ϭ 0.38 Ϯ 0.06, control ϭ 0.06 Ϯ 0.03, t ϭ gynes did not emerge from their cocoons from than 4.8; df ϭ 7; P ϭ 0.0021). However, in 2007, individual control colonies (95% conÞdence interval for the activity rates in fed colonies did not differ from control mean percent of gynes that did not eclose per colony: colonies (fed ϭ 0.60 Ϯ 0.36, control ϭ 0.57 Ϯ 0.18, t ϭ control ϭϪ13.1Ð41.6%, fed ϭϪ23.7Ð48.7%). There- 0.08, df ϭ 8; P ϭ 0.9). fore, there was probably ample nectar available for December 2009 ELLIOTT:BUMBLE BEE NECTAR SURPLUSES 1687 colony growth and after the period of colony growth, males, than unfed control colonies (Pelletier and Mc- bee colonies may have needed less nectar. Neil 2003). Reduced resource availability in conven- Food limitation for bumble bee colonies depends on tional agricultural habitats versus natural habitats or both nectar and pollen availability. In 2006, foragers in ßower rich suburban habitats also reduced colony fed colonies were six times more active than foragers reproduction of B. vosnesenskii in California (Green- in control colonies (i.e., more movements in and out leaf 2005) and B. terrestris the United Kingdom (Goul- of the hive, scaled by colony size). Fed colonies may son et al. 2002). In another United Kingdom study, have used the extra nectar to increase foraging efforts, worker production was higher when B. terrestris col- enabling them to harvest even more resources, to onies were placed near ßower-rich oil-seed rape Þelds produce more gynes. In contrast, in 2007, colonies that versus ßower-poor wheat Þelds, but rich resources were fed with extra nectar and pollen, instead of nec- were also associated with higher social parasitism tar only, were not more active than control colonies. (Carvell et al. 2008). Because Carvell et al. (2008) If foraging activity increases when colonies need more terminated the colonies before they reached gyne pollen (Plowright et al. 1993, Rasheed and Harder production, it is not known if the social parasitism 1997), this could also explain why bees in fed colonies would have outweighed the positive effects of ßoral in 2007 were not more active (entry/exit rate) than resources on overall colony reproduction. Finally, in bees in control coloniesÑi.e., they may have already California, proximity to honey bee competitors re- had plenty of pollen so they did not need to leave the duced B. occidentalis gyne and male production colony and forage more (Cartar 1992, Weinberg and (Thomson 2004). In all of these systems, colonies had Plowright 2006). Future studies could compare am- longer foraging seasons and grew to be larger than the bient pollen and nectar availability and the relative colonies in this study. pollen and nectar requirements for colony reproduc- Subalpine bees may be at one end of a food limi- tion. tation gradient, being less food-limited than other In the feeding experiment, there was conßicting bees. As the length of the growing season increases at evidence for nectar (2006) or nectar and pollen lower elevations (and as summer daylength increases (2007) limitation of bee reproduction. In 2006, gyne at higher latitudes), colonial bees have more time to (but not worker or male) production increased with build large colonies, and multivoltine solitary bees nectar additions. However, in 2007, supplemental nec- have more time to produce additional generations tar and pollen additions did not affect reproduction. (Minckley et al. 1994, Goodwin 1995, Thiele 2005, de The feeding effect size was so low that 80% power to la Hoz 2006, Packer et al. 2007). Because the growing detect a signiÞcant effect of feeding on gyne produc- season at high elevation is shorter than at low elevation would require a sample size of 73 colonies, which tion, high elevation bumble bee reproduction may be would require many years of study, given the low time-limited (Pyke 1982) instead of food-limited. For success rates of rearing captive colonies (Kearns and example, in this study, colonies produced a maximum Thomson 2001). Adding few nonambient colonies per of only 21 workers (Fig. 2), whereas the average num- meadow is ideal because ambient colony densities are ber of workers alive at any given time in the Quebec unknown and increasing captive colony number per study ranged from 30 to 49 workers per control colony meadow could inßate natural levels of resource de- (Pelletier and McNeil 2003). Small colonies may never pletion. However, future work could increase sample have enough foragers to collect extra resources even sizes by studying colonies in more meadows and/or if they are readily available in nearby ßowers. Food more ambient colonies per meadow. It is important to supplementation not only provides more total re- note that supplemental food could have had a stronger sources, but they are also resources that workers can effect on natural colonies if my captive colonies had use while staying in the colony and maintaining colony been smaller than natural colonies. Small colony sizes defense and brood care (Cartar 1992). Therefore, the (i.e., few workers per colony) may prevent colonies beneÞts of food additions for reproduction may not from taking advantage of extra ßowers and reduce the necessarily mirror the beneÞts of ßower addition for overall demand for food. Also, additional ßoral re- bee reproduction. sources might have had a larger effect on colony es- Recent pollinator shortages have heightened tablishment than on colony growth, because ßower awareness that pollinator declines may be linked to availability per bee is lower early in the season when declines in plant populations (Buchmann and Nabhan colonies are becoming established than later during 1996, Kearns et al. 1998). For example, both bees and colony growth (Elliott 2009). plants that require pollinators for outcrossing have The few studies that have tested whether bumble declined in Britain and The Netherlands (Biesmeijer bee colony productivity in the wild is sensitive to food et al. 2006). However, this study suggests that natural supplementation or natural variation in ßoral re- ßoral resources (in particular, nectar resources) are sources suggest that at lower elevations (Ͻ200 m), not the sole driver of subalpine bumble bee repro- bumble bees are food-limited (Goulson et al. 2002, duction. Although bumble bee colonies that were fed Pelletier and McNeil 2003, Thomson 2004, Greenleaf with supplemental nectar generally produced more 2005, Carvell et al. 2008). For example, in the only offspring than control colonies, colonies in meadows other feeding experiment with Þeld colonies, fed col- with more ßowers (which contained surplus nectar at onies of Bombus impatiens and B. ternarius colonies in the end of the day) did not produce more offspring Quebec, Canada, produced more gynes, workers, and than colonies in meadows with fewer ßowers. Also, 1688 ENVIRONMENTAL ENTOMOLOGY Vol. 38, no. 6 supplemental nectar feeding did not increase worker Cushman, J. H., and A. J. Beattie. 1991. Mutualisms: assess- production, and these subalpine colonies stayed very ing the beneÞts to hosts and visitors. Trends Ecol. Evol. small compared with lower elevation colonies. Thus, 6: 193Ð195. areas with small colonies may not be able to use sur- de la Hoz, J.D.T. 2006. Phenology of Bombus pennsylvanicus plus natural ßoral resources because they have a lim- sonorus Say (Hymenoptera: Apidae) in central Mexico. ited number of foragers per colony. This study high- Neotrop. Entomol. 35: 588Ð595. lights the need for more studies that examine food and Dramstad, W., and G. Fry. 1995. Foraging activity of bumblebees (Bombus) in relation to ßower resoruces on ar- ßower limitation of pollinator reproduction in multi- able land. Agric. Ecosyst. Environ. 53: 123Ð135. ple locations and habitats. Dukas, R., and D. H. Morse. 2003. Crab spiders affect ßower visitation by bees. Oikos 101: 157Ð163. Egger, K. N., and D. S. Hibbett. 2004. The evolutionary Acknowledgments implications of exploitation in mycorrhizas. Can. J. Bot. Rev. 82: 1110Ð1121. I thank G. Entsminger and J. Mahmoudi for help moni- Elliott, S. E. 2008. Reciprocal beneÞts in a plant-pollinator toring bees in the Þeld and R. Irwin, D. Inouye, and J. Thom- mutualism. PhD dissertation, Dartmouth College, Hanover, son for providing nest boxes. J. Bronstein, L. Burkle, R. NH. Calsbeek, R. Cox, G. Entsminger, R. Irwin, M. McPeek, and Elliott, S. E. 2009. Subalpine bumble bee foraging distances D. Peart provided insightful comments on the manuscript. and densities in relation to ßower availability. Environ. This work was funded by the ExplorerÕs Club Exploration Entomol. 38: 748Ð756. Fund, the Rocky Mountain Biological Laboratory Snyder Gomez, J. M., J. Bosch, F. Perfectti, J. Fernandez, and M. Fund, and the National Science Foundation (R. E. Irwin and Abdelaziz. 2007. Pollinator diversity affects plant repro- S.E.E., DDIG-0608144). The experiments in this study com- duction and recruitment: the tradeoffs of generalization. ply with all Gunnison County regulations. Oecologia (Berl.) 153: 597Ð605. Goodwin, S. G. 1995. Seasonal phenology and abundance of early-, mid- and long-season bumble bees in southern References Cited England, 1985Ð1989. J. Apicult. Res. 34: 79Ð87. Goulson, D. 2003. Bumblebees: their behaviour and ecol- Ashman, T. L., T. M. Knight, J. A. Steets, P. Amarasekare, M. ogy. Oxford University Press, New York. Burd, D. R. Campbell, M. R. Dudash, M. O. Johnston, S. J. Goulson, D., and J. C. Stout. 2001. Homing ability of the Mazer, R. J. Mitchell, et al. 2004. Pollen limitation of bumblebee Bombus terrestris (Hymenoptera: Apidae). plant reproduction: ecological and evolutionary causes Apidologie 32: 105Ð111. and consequences. Ecology 85: 2408Ð2421. Goulson, D., and B. Darvill. 2004. Niche overlap and diet Baker, H. G., and I. Baker. 1982. Studies of nectar-consti- breadth in bumblebees; are rare species more specialized tution and pollinator-plant coevolution, pp. 131Ð171. In in their choice of ßowers? Apidologie 35: 55Ð63. M. H. Nitecki (ed.), Biochemical aspects of evolutionary Goulson, D., W.O.H. Hughes, L. C. Derwent, and J. C. Stout. biology. University of Chicago Press, Chicago, IL. 2002. Colony growth of the bumblebee, Bombus terres- Beekman, M., P. van Stratum, and R. Lingeman. 1998. Dia- tris, in improved and conventional agricultural and sub- pause survival and post-diapause performance in bum- urban habitats. Oecologia (Berl.) 130: 267Ð273. blebee queens (Bombus terrestris). Entomol. Exp. Appl. Goulson, D., J. L. Cruise, K. R. Sparrow, A. J. Harris, K. J. 89: 207Ð214. Park, M. C. Tinsley, and A. S. Gilburn. 2007. Choosing Bever, J. D. 1999. Dynamics within mutualism and the main- rewarding ßowers; perceptual limitations and innate tenance of diversity: inference from a model of interguild preferences inßuence decision making in bumblebees frequency dependence. Ecol. Lett. 2: 52Ð62. and honeybees. Behav. Ecol. Sociobiol. 61: 1523Ð1529. Biesmeijer, J. C., S.P.M. Roberts, M. Reemer, R. Ohlemuller, Greenleaf, S. S. 2005. Local-scale and foraging-scale habi- M. Edwards, T. Peeters, A. P. Schaffers, S. G. Potts, R. tats affect bee community abundance, species richness, Kleukers, C. D. Thomas, et al. 2006. Parallel declines in and pollination services in Northern California. Princeton pollinators and insect-pollinated plants in Britain and The University, Princeton, NJ. Netherlands. Science 313: 351Ð354. Hay, M. E., J. D. Parker, D. E. Burkepile, C. C. Caudill, A. E. Bronstein, J. L. 1994a. Conditional outcomes in mutualistic Wilson, Z. P. Hallinan, and A. D. Chequer. 2004. Mu- interactions. Trends Ecol. Evol. 9: 214Ð217. Bronstein, J. L. 1994b. Our current understanding of mutualisms and aquatic community structure: the enemy of tualism. Qtly. Rev. Biol. 69: 31Ð51. my enemy is my friend. Annu. Rev. Ecol. Evol. Syst. 35: Buchmann, S., and G. P. Nabhan. 1996. The forgotten pol- 175Ð197. linators. Island, Washington, DC. Heinrich, B. 1979. Bumblebee economics. Harvard Univer- Burd, M. 1994. BatemanÕs principle and plant reproduction: sity Press, Cambridge, MA. the role of pollen limitation in fruit and seed-set. Botan- Herrera, C. M. 2000. Flower-to-seedling consequences of ical Rev. 60: 83Ð111. different pollination regimes in an insect-pollinated Cartar, R. V. 1992. Adjustment of foraging effort and task shrub. Ecology 81: 15Ð29. switching in energy-manipulated wild bumblebee colo- Johnson, S. D., and K. E. Steiner. 1997. Long-tongued ßy nies. Anim. Behav. 44: 75Ð87. pollination and evolution of ßoral spur length in the Disa Cartar, R. V. 2004. Resource tracking by bumble bees: re- draconis complex (Orchidaceae). Evolution 51: 45Ð53. sponses to plant-level differences in quality. Ecology 85: Kearns, C., and J. Thomson. 2001. The natural history of 2764Ð2771. bumblebees: a sourcebook for investigations. University Carvell, C., P. Rothery, R. F. Pywell, and M. S. Heard. 2008. of Colorado Press, Boulder, CO. Effects of resource availability and social parasite inva- Kearns, C. A., D. W. Inouye, and N. M. Waser. 1998. En- sion on Þeld colonies of Bombus terrestris. Ecol. Entomol. dangered mutualisms: the conservation of plant-pollina- 33: 321Ð327. tor interactions. Annu. Rev. Ecol. Syst. 29: 83Ð112. December 2009 ELLIOTT:BUMBLE BEE NECTAR SURPLUSES 1689

Michener, C. 2007. The bees of the world. The John Hop- Sahli, H. F., and J. K. Conner. 2007. Visitation, effectiveness, kins University Press, Baltimore, MD. and efÞciency of 15 genera of visitors to wild radish, Minckley, R. L., W. T. Weislo, D. Yanega, and S. L. Buch- Raphanus raphanismum (Brassicaceae). Am. J. Botany 94: mann. 1994. Behavior and phenology of a specialist bee 203Ð209. (Dieunomia) and sunßower (Helianthus) pollen avail- SAS Institute. 2001. JMP 4.04 computer program. SAS In- ability. Ecology 75: 1406Ð1419. stitute, Cary, NC. Morales, M. A. 2000. Mechanisms and density dependence Simms, E., and D. Taylor. 2002. Partner choice in nitrogen- of beneÞt in an ant-membracid mutualism. Ecology 81: Þxation of legumes and rhizobia. Integrat. Comp. Biol. 42: 482Ð489. 369Ð380. Morgan, M. T., W. G. Wilson, and T. M. Knight. 2005. Plant Steffan-Dewenter, I., U. Munzenberg, C. Burger, C. Thies, population dynamics, pollinator foraging, and the selec- and T. Tscharntke. 2002. Scale-dependent effects of tion of self-fertilization. Am. Nat. 166: 169Ð183. landscape context on three pollinator guilds. Ecology 83: National Resource Council. 2006. Status of pollinators in 1421Ð1432. North America. National Academy Press, Washington, Stenstrom, M., and P. Bergman. 1998. Bumblebees at an DC. alpine site in northern Sweden: temporal development, Ness, J. H., J. L. Bronstein, A. N. Andersen, and J. N. Holland. population size, and plant utilization. Ecography 21: 306Ð 2004. Ant body size predicts dispersal distance of ant- 316. adapted seeds: implications of small-ant invasions. Ecol- Tepedino, V. J., and N. L. Stanton. 1982. Estimating ßoral ogy 85: 1244Ð1250. Packer, L., A.I.D. Gravel, and G. Lebuhn. 2007. Phenology resources and ßower visitors in studies of pollinator-plant and social organization of Halictus (Seladonia) tripartitus communities. Oikos 38: 384Ð386. (Hymenoptera: Halictidae). J. Hymenoptera Res. 16: Thiele, R. 2005. Phenology and nest site preferences of 281Ð292. wood-nesting bees in a Neotropical lowland rain forest. Pake, C. E., and D. L. Venable. 1996. Seed banks in desert Studies Neotrop. Fauna Environ. 40: 39Ð48. annuals: implications for persistence and coexistance in Thompson, J. D. 2005. The geographic mosaic of coevolu- variable environments. Ecology 77: 1427Ð1435. tion. University of Chicago Press, Chicago, IL. Pelletier, L., and J. N. McNeil. 2003. The effect of food Thomson, D. 2004. Competitive interactions between the supplementation on reproductive success in bumblebee invasive European honey bee and native bumble bees. Þeld colonies. Oikos 103: 688Ð694. Ecology 85: 458Ð470. Pellmyr, O., and J. N. Thompson. 1996. Sources of variation Weinberg, D., and C.M.S. Plowright. 2006. Pollen collec- in pollinator contribution within a build: The effects of tion by bumblebees (Bombus impatiens): the effects of plant and pollinator factors. Oecologia (Berl.) 107: 595Ð resource manipulation, foraging experience and colony 604. size. J. Apicult. Res. 45: 22Ð27. Plowright, R. C., J. D. Thomson, L. P. Lefkovitch, and C.M.S. Westphal, C., I. Steffan-Dewenter, and T. Tscharntke. 2006. Plowright. 1993. An experimental study of the effect of Bumblebees experience landscapes at different spatial colony resource level manipulation on foraging for pollen scales: possible implications for coexistence. Oecologia by worker bumblebees (Hymenoptera, Apidae). Can. J. (Berl.) 149: 289Ð300. Zool. 71: 1393Ð1396. Williams, N. M., and C. Kremen. 2007. Resource distribu- Pyke, G. H. 1982. Local geographic distributions of bum- tions among habitats determine solitary bee offspring blebees near Crested Butte, Colorado: competition and production in a mosaic landscape. Ecol. Applic. 17: 910Ð community structure. Ecology 63: 555Ð573. 921. Rasheed, S. A., and L. D. Harder. 1997. Foraging currencies Young, H. J., and T. P. Young. 1992. Alternative outcomes of for non-energetic resources: pollen collection by bum- natural and experimental high pollen loads. Ecology 73: blebees. Anim. Behav. 54: 911Ð926. 639Ð647. Rudgers, J. A., and S. Y. Strauss. 2004. A selection mosaic in the facultative mutualism between ants and wild cotton. Proc. R. Soc. Lond. Ser. B Biol. Sci. 271: 2481Ð2488. Received 16 January 2009; accepted 6 August 2009.