APICULTURE AND SOCIAL INSECTS

The Effect of Feeder Location on Pollen Collection by Bumble Bees

in a Tomato Greenhouse in Ontario, Canada
LEVENTE L. ORBAN,1,2 C.M.S. PLOWRIGHT,1

AND

R. C. PLOWRIGHT3

J. Econ. Entomol. 105(1): 3439 (2012); DOI: http://dx.doi.org/10.1603/EC11198

ABSTRACT The foraging behavior of bumble bees (Bombus impatiens Cresson) was examined as
a function of feeder location containing sugar solution in a commercial tomato greenhouse in
Manotick, Ontario, Canada. The feeders were located within the nest-box (fed-close) or placed 1.5 m
away (fed-far) and the placement of the two types of colonies was counterbalanced over time. No
effect of feeder location was found in colony activity levels or in pollen load size. A foraging trade-off
between sugar solution and pollen collection, however, was found: the proportion of foraging trips
in which pollen was brought back was signicantly reduced for fed-far colonies, which contrasts with
our laboratory study in which the opposite effect was found. We interpret our ndings as possibly
reecting a limitation in pollen supply in the greenhouse: an already possibly strained ability to nd
and bring back pollen to the colony was accentuated by increasing the task demands of collecting sugar
solution.
KEY WORDS pollination, greenhouse, bumble bees, foraging tradeoff, tomatoes

Pollination by bumble bees (Bombus impatiens Cresson) in tomato greenhouses has become a mainstream
practice across the world over the past two decades
(Velthuis and van Doorn 2005). Bumble bee pollination leads to better fruit quality and increased yield at
a lower cost in comparison to mechanical pollination
where individual owers are pollinated by vibrating
bristles (Dogterom et al. 1998, Meisels and Chiasson
1996). In an effort to improve cost-effectiveness of using
commercial colonies, researchers have investigated a variety of factors potentially affecting worker behavior.
One such factor is lighting. Bumble bees forage well in
the absence of ultraviolet (UV) light (Dyer and Chittka
2004), but there may be greater potential for bee loss
through ventilation systems in commercial greenhouses
made of UV-blocking transmitting plastic (Morandin et
al. 2002). Another factor is the visual appearance of the
colonys housing. Bee drifting from one colony to another, which is associated with aggression and resource
robbing, occurs in greenhouses where there are dense
aggregations of colonies, but efforts to minimize drifting
by making colonies more distinctive have not been effective (Birmingham and Winston 2004). Here, based on
laboratory research reviewed below, we explore the role
of a new variable in a commercial greenhouse: the placement of the feeders used to provide sugar solution to the
bumble bees.
Because tomato owers release pollen but do not
secrete nectar, bumble bee colonies are typically pro1 School of Psychology, University of Ottawa, 136 Jean Jacques
Lussier, Ottawa, Ontario, Canada, K1N 6N5.
2 Corresponding author, e-mail: lorban@uottawa.ca.
3 Professor Emeritus, University of Toronto, Toronto, Canada, M5S
1A1.

vided with an ad libitum supply of sugar solution in a

feeder placed directly beneath the colony with an
opening only a few centimeters away from the comb.
This practice is highly convenient to the greenhouse
grower who need not tend to the colonies beyond
opening the feeder reservoir upon the colonys arrival
in the greenhouse. Though it stands to reason that if
the colonies are relieved of the task of collecting sugar
solution then they can devote more of their efforts to
pollen gathering from tomato owers, this remains an
untested assumption in the greenhouse. Indeed, we
recently challenged the assumption in a laboratory
setting (Weinberg and Plowright 2006). Colonies
were not fed pollen directly but the workers foraged
for it in a ight cage. Ground pollen was scattered on
a sheet of cheesecloth set on top of a stand and the
bees scrabbled for it, much as they scrabble for pollen
on thistle owers (Plowright et al. 1999). The placement of the sugar solution was manipulated: it was
either placed directly into the colony or in the middle
of the ight cage. Having to y to collect sugar solution, far from having a detrimental effect, improved
pollen collection performance: the frequency of pollen collection trips was increased. The mechanism
underlying this effect may have been motivational:
foraging may exert positive feedback on foraging
(Plowright and Plowright 1988). Alternatively, it may
be one of associative learning, foraging behavior having been reinforced with not one but two sources of
reward. We suggested that in a greenhouse, pollination might well be improved by having the bees do
more for themselves, not less. Nonetheless, the laboratory is not the greenhouse, and the effect of manip-

0022-0493/12/00340039$04.00/0 2012 Entomological Society of America

ORBA N ET AL.: FEEDER LOCATION AND BUMBLE BEE POLLEN GATHERING

February 2012
Table 1.

35

Individual counts for fed-close and fed-far colonies

Fed

Time

Colony

Male
pupae

Queen
pupae

Worker
pupae

Male or worker
cocoons

Queen
cocoons

Initial
worker count

Final worker
count

Final/
initial

Male
count

Queen
count

Close

1
2
5
6
3
4
7
8

10
14
73
8
11
30
45
24

0
0
0
23
3
0
1
1

8
3
67
11
11
17
62
30

255
102
108
36
238
144
62
84

3
0
1
6
2
1
20
11

20
24
113
134
19
33
100
122

74
51
215
135
119
52
110
58

3.70
2.12
1.90
1.01
6.26
1.57
1.10
0.48

17
41
119
50
91
34
109
43

1
1
20
54
7
1
43
25

II
Far

I
II

ulating placement of a sugar solution feeder remained

to be tested in a greenhouse, as we have done here.
An alternative possible outcome of the manipulation is that there will be a foraging trade-off (Stephens and Krebs 1986). Perhaps, collection of sugar
solution away from the colony will be done at the
expense of pollen collection (Plowright and Plowright
1999, Plowright and Silverman 2000): the more difcult it is to obtain sugar solution, the less effort can be
allocated to pollen collection. A nal possibility is that
pollen and sugar solution collection might be truly
independent given that pollen is a source of protein,
primarily used to feed larvae and to promote ovarian
development, whereas sugar solution (or nectar in
natural owers) is a source of energy (Goulson 2010).
Allocation of individual or colony effort to the task of
collecting pollen may in principle be unaffected by
parameters related to availability of sugar solution:
manipulating the ease of obtaining sugar solution may
have no effect on the need for pollen and no effect on
pollen collection behavior. This last outcome, which
amounts to a conrmation of the null hypothesis,
would be surprising in view of work showing that the
two activities cannot be considered in isolation of each
other (Landry et al. 2000), though Tod (1986) found
no effect of feeder placement on pollen collection in
a eld study.
Methods
Colonies. Eight colonies (Young Class A colonies
for research from Biobest Canada Ltd., Leamington,
Ontario, Canada) of B. impatiens were used. The rst
four colonies were tested from 9 March to 24 April
2010 (Time I) and were removed from the greenhouse
on 24 April 2010. The second four colonies were tested
from 24 April to 14 May 2010 (Time II) and were
removed from the greenhouse on 5 June 2010. Frozen
fresh pollen from a variety of ower types was purchased from an apiarist, ground, formed into a paste
and shaped into a ball 1 cm in diameter. A pollen ball
was provided to each colony 3 d before its placement
in the greenhouse. Once in the greenhouse, colonies
were not supplemented with pollen and their only
source of pollen in the greenhouse was the tomato
owers.
Greenhouse. At the time of the study, SunTech
greenhouse in Manotick, Ontario covered an area of
8,018 m2 (the greenhouse has since expanded). Stems

(24,200) of Beefsteak Heritage Variety tomatoes (Solanum lycopersicum L.) produced 6 owers every 8 d
(145,200 owers). Humidity levels varied between
60 90% (most of the time between 70 85%), and a
24 h temperature average of 19C (high of 22C, prenight low of 17C) was maintained. The greenhouse
material was transparent to UV light. Articial lighting
was not used. Thus, the Time I colonies received an
average of 11.5 h of daylight during the month of
March, and the Time II colonies received an average
of 13.5 h of daylight during the months of April and
May.
In addition to the four colonies that we brought in
to the greenhouse at each time, ve commercial colonies were placed in the greenhouse by the grower
(i.e., nonexperimental colonies) to ensure adequate
pollination and subsequent tomato production. The
nearest nonexperimental colonies were placed 3 m
from the experimental colonies. The nonexperimental
colonies began to produce males and young queens,
which occasionally drifted into our colonies during
the experiment.
Design. The location of the feeder of sugar solution
in relation to the nest box was experimentally manipulated. As shown in Table 1, at each time, two colonies
received sugar solution (i.e., fed-close colonies),
supplied by Biobest with the colonies, inside the nest
box in the usual way that commercial colonies are fed
in greenhouses. The other two colonies were provided
with sugar solution outside of the nest box (i.e., fedfar colonies). The feeders consisted of seven 100 ml
plastic vials bound together with duct tape to form a
hexagonal feeder cluster (Fig. 1). The group of vials
was suspended from a string so that it was level with
the colony. Each vial was lined with a metallic screen
to ensure that workers could reach the sugar solution
without drowning. Initially, the feeder was just in front
of the nest box entrance (20 cm) to ensure that the
bees found the sugar solution. We veried that the
workers found the sugar solution by adding blue food
coloring to the feeder and looking for blue coloration
of the wax honey pots in colonies. This method also
enabled us to verify that fed-close colonies supplemented with sugar solution inside the hive were not
gathering sugar solution from fed-far colonies. Once
the workers found the feeder, the distance of the
feeder from the colony was increased by 35 cm at the
end of each day of observation until the distance to
the colony had reached 1.5 m. This way of training

36

JOURNAL OF ECONOMIC ENTOMOLOGY

Vol. 105, no. 1

Fig. 1. Photo of the feeder of sugar solution placed at 20 cm from the colony. The distance of the feeder was increased
to 1.5 m at the rate of 35 cm per week. (Online gure in color.)

bees to a feeder is standard in the honey bee literature.

In comparison to studies on honey bees, however, our
rate of increasing distances was much more gradual:
Zhang et al. (2005) trained honey bees to a feeder
distance of 7.2 m by moving the feeder 20 cm every 10
min.
Procedure. At Time I, the two fed-close colonies
were placed 15 m apart, as were the two fed-far colonies. The two fed-close colonies were placed as far as
possible (88 m) from the two fed-far colonies (i.e., the
fed-far and fed-close colonies were at opposite ends of
the greenhouse) to minimize the possibility that the
fed-close colonies would feed from the feeders meant
for the fed-far colonies. Fed-close and fed-far colonies
were matched as closely as possible for their initial
worker counts. An observer recorded each time there
was a bee that exited the colony and each time
there was a bee that entered. When a bee entered the
colony, the observer also recorded whether there was
pollen on the hind legs or not. Observations were
taken on days 1, 2, 7, 8, 13, and 14. Each colony was
monitored for a 15 min period during an active period
of the day, between 1100 and 1300 hours.
At Time II, four new colonies were taken to the
greenhouse. Observations took place at the same frequency as in Time I. The placement of colonies in each
condition was counterbalanced between Times I and
II, so that a fed-far colony at Time II was placed where
there was a fed-close colony at Time I and vice versa.
Colony exit and entry events were recorded using a
Panasonic 3MOS High Denition camera at a resolution of 1,920 by 1,080 pixels.
All workers were counted the day before the colonies were placed in the greenhouse. After the colo-

nies were removed from the greenhouse, they were

frozen. All individuals, including pupae, males,
queens, and workers were then counted in the lab.
Estimation of Pollen Load Size. During Time II, we
randomly selected 39 workers over 2 d of observations
(19 from the fed-close colonies and 20 from the fed-far
colonies) entering the colony with a pollen load. Approximately 100 frames of video for each worker, near
the moments of nest-box entry, were exported as a
Portable Network Graphics (PNG) image sequence.
The PNG sequence was exported at maximum resolution, which enabled us to quantify the size of pollen
loads (sample video recordings are available at www.
beelab.ca). Two of the best quality images from each
100-image sequence were selected and imported into
Adobe Photoshop CS5 (Adobe Systems Incorporated,
San Jose, CA). The best images were selected by the
researcher based on the clarity of the image, contrast
of the pollen from the background, and on the orientation of the worker. Photoshops Analyze function
was used to estimate the area of pollen on one of the
hind legs of a worker as an index of pollen load size
(see Fig. 2 for a sample), and the areas for the two
images were averaged. A metric paper ruler (Fig. 2)
glued to the entrance of each colony was used to
standardize the scale of the measurements.
Statistical Analysis. The data were analyzed using
GLIM (Generalized Linear Interactive Modeling;
Francis et al. 1993). To assess activity levels, we examined the frequency of entries and exits within each
15 min recording period separately. The counts were
treated as a Poisson process. As is frequently the case
(McCullagh and Nelder 1989), however, the data
were over-dispersed (i.e., the variance was greater

February 2012

ORBA N ET AL.: FEEDER LOCATION AND BUMBLE BEE POLLEN GATHERING

37

Fig. 2. Example of image used for estimating pollen load size. (Online gure in color.)

than the mean) and so, as in Weinberg and Plowright

(2006), we used the following strategy outlined by
Baker and Nelder (1978): the data were square-root
transformed, which stabilized the variance, and a normal error distribution with log link were specied.
Initial and nal worker counts were rst entered in the
model as covariates and then the effects of time and
feeder location were tested. Because the entries with
and without pollen were binary data, a logistic model
was used to analyze the proportion of entries in which
pollen was brought back.
Results
All colonies had males and male pupae, suggesting
that the colonies were producing no more workers
(Table 1). The increase in worker numbers for Time
I colonies was more pronounced than for the Time II
colonies, but this undoubtedly reected their larger
growth potential given that the Time I colonies were
smaller at the outset. The most striking aspect of the
data are the modest nal sizes for all colonies: even the
largest colony was no bigger than 215 workers. Even
if the number of cocoons is taken as an index of nal
size (and this is a highly inated index because empty
cocoons are produced by both males and workers), in
comparison to a eld study on 10 colonies where the
median count was over 500 (Weinberg 2007), here,
the highest count was 255.
Not surprisingly, we found that the greater were the
initial and nal worker counts, the greater were the
frequencies of exits (F 19.81; df 1, 55; P 0.00005
and F 18.99; df 1, 55; P 0.00006, respectively) and
entries (F 18.61; df 1, 55; P 0.00007 and F 16.25;
df 1, 55; P 0.0002, respectively). Once colony size
was taken into account, the frequencies at Time II
were depressed, both for exits (F 4.32; df 1, 55; P
0.04) and entries (F 12.85; df 1, 55; P 0.0008);
this effect of time may possibly reect a nonlinear
relationship between colony size and activity, with the
larger colonies at Time II being less active than would
be predicted if the relationship between colony size
and activity were linear. The effect of feeder location, however, was not signicant, either for exits
(F 2.71; df 1, 55; P 0.10) or for entries (F

2.57; df 1, 55; P 0.12). There were no interactions of feeder location with time (F 0.18; df 1,
55; P 0.67 for exits and F 0.022; df 1, 55; P
0.88 for entries). The pollen areas in the digital
images covered 0.24 cm2 (SE 0.02) for the workers
from the fed-close colonies, and 0.27 cm2 (SE
0.03) for the workers from the fed-far colonies. The
difference in pollen load size was analyzed using an
independent samples t-test, and was not signicant
either (t 1.01; df 37; P 0.32).
In contrast with the analyses above on activity levels
and pollen load size that revealed no effect of feeder
location, at both times the fed-far colonies showed a
reduction, compared with the fed-close colonies, in
the proportion of foraging trips in which pollen was
brought back (Fig. 3). The effect of feeder location
was signicant (2 14.99; df 1; P 0.0002). No
effect of time was detected (2 0.361; df 1; P
0.55). The interaction of time and feeder location was
nonsignicant as well (2 0; df 1; P 1.0).

Discussion
There was no difference between fed-close and
fed-far colonies in their activity levels or their pollen
load sizes. There was, however, a statistically signicant difference in the proportion of foraging trips in
which bees brought back pollen to the colony. The
difference cannot be attributed to colony placement
within the greenhouse, because the fed-close colonies
were placed at one end of the greenhouse at Time I,
and at the other end at Time II.
Contrary to the results of our laboratory study
(Weinberg and Plowright 2006), colonies fed sugar
solution close to their nest collected proportionally
more, not less, pollen than the fed-far comparison
group. Hence our results do not justify any change in
the current practice of attaching the sugar solution
feeder to the colonyindeed they provide some empirical support for what has heretofore been taken as
an article of faith. One caveat, however, is suggested
by the work of Tod (1986) who used the same manipulation in a eld study. He reported one detrimental effect of feeding close: the early development of

38

JOURNAL OF ECONOMIC ENTOMOLOGY

Vol. 105, no. 1

Fig. 3. Mean proportion of trips to the colony in which pollen was brought back, with standard error bars, for fed-far and
fed-close colonies at Times I and II.

worker ovaries leading to the production of male offspring.

Given many differences between the laboratory
and greenhouses, we can only speculate as to the root
of the marked difference in results between our laboratory and greenhouse studies. In the laboratory, the
foraging distances were comparatively small (the
ight cage was 2 2 m) and the pollen availability
was unnaturally high (an unlimited supply awaited the
bees who ew to the middle of the cage), and so a
foraging tradeoff was not to be expected. In our greenhouse study, however, pollen seemed to be a severe
limitation and so the bees would have been hard
pressed to collect enough pollen, and even more so
when they were forced to work for their sugar solution: foraging for sugar solution would be at the expense of pollen collection. In other greenhouses and
other species of bumble bees, colonies may well obtain
adequate nutrition from tomato owers: Whittington
and Winston (2003) provided pollen supplements to
greenhouse colonies of B. occidentalis Greene in British Columbia and found no effect on worker populations. We have three lines of evidence that our colonies, however, were pollen stressed: (1) Every colony
experienced resource deprivation to the extent that
they expelled larvae multiple times throughout the
colonies development. (2) Employees in the greenhouse gave consistent verbal reports of nding a relatively high frequency of bite marks on the tomato
owers. (3) Most importantly, the nal colony sizes
were small for B. impatiens: the largest colony comprised 215 workers, in contrast with the report of
Cnaani et al. (2002) of colony sizes of 374.5 108
workers.
One possible criticism is that we created a pollen
shortage ourselves by increasing the stocking rate beyond what was usual: the grower already had ve
colonies and we provided an additional four in an area

of 0.8 ha. Even with our additional colonies, however,

these stocking rates are well within the norm: Morandin et al. (2001) report typical values of 7.6 19.8 colonies per hectare. Growers are naturally concerned
about pollination and prefer to err on the side of
caution, especially given worker loss to the outside as
a well-known problem in greenhouses (Whittington
et al. 2004). Nonetheless, a high stocking rate may
have the unintended consequence of suppressing colony growth and longevity, overshadowing any possible benecial effect of changing the feeder placement,
and ultimately increasing the cost of colony replacement. Future research on manipulations that possibly
have even small effects on the rate colony replacement has the potential for payoff in economic dividends: savings in the greenhouse industry might possibly still be achieved by modest increases in operating
efciency with cumulative effects in the long run.
In terms of behavioral ecology, this study represents
a step toward understanding the interplay between
the colony requirements, the oral resources and the
various behaviors that underlie pollination. The
greenhouse environment of our study, however, is still
far removed from the eld even if it is more naturalistic
than the laboratory. Specically, it is not self evident
that a reservoir of sugar solution is treated by the bees
as a ower at all, especially when it is placed directly
in the colony (Plowright et al. 1995). Nonetheless,
there may be a possible comparison to be drawn here
with one of the rare eld studies (Cresswell 1999) in
which both pollen and nectar content were manipulated in individual owers (oil-seed rape, Brassica napus L.): there, pollen transfer on the owers stigma
depended on pollen content but not nectar, whereas
the reverse was true for visit duration by bumble bees
(B. lapidarius L.). Future studies might be aimed at a
systematic investigation that bridges the gaps between
laboratory, the greenhouse, and the eld.

February 2012

ORBA N ET AL.: FEEDER LOCATION AND BUMBLE BEE POLLEN GATHERING

Acknowledgments

We thank Bob and Megan Mitchell for giving us unlimited

access to the greenhouse. We also thank Emily Sibbald and
Emma Thompson for their constructive feedback on the
manuscript, Nelson Pomeroy for helpful suggestions and
Rick Norian for his assistance in data collection. The study
was supported by a research grant from the Natural Sciences
and Engineering Research Council to C.M.S.P.

References Cited
Baker, R. J., and R. A. Nelder. 1978. The GLIM system,
release 3: generalised linear interactive modelling. Numerical Algorithms Group, Oxford, United Kingdom.
Birmingham, A. L., and M. L. Winston. 2004. Orientation
and drifting behavior of bumblebees (Hymenoptera: Apidae) in commercial tomato greenhouses. Can. J. Zool. 82:
5259.
Cnaani, J., R. Schmid-Hempel, and J. O. Schmidt. 2002. Colony development, larval development and worker reproduction in Bombus impatiens Cresson. Insectes Soc. 49:
164170.
Cresswell, J. E. 1999. The inuence of nectar and pollen
availability on pollen transfer by individual owers of
oil-seed rape (Brassica napus) when pollinated by bumblebees (Bombus lapidarius). J. Ecol. 87: 670677.
Dogterom, M. H., J. A. Matteoni, and R. C. Plowright, R. C.
1998. Pollination of greenhouse tomatoes by the North
American Bombus vosnesenskii (Hymenoptera: Apidae).
J. Econ. Entomol. 91: 7175.
Dyer, A. G., and L. Chittka. 2004. Bumblebee search time
without ultraviolet light. J. Exp. Biol. 207: 16831688.
Francis, B., M. Green, and C. Payne. 1993. GLIM: the statistical system for generalized linear interactive modeling, version 4.0. Oxford University Press, New York.
Goulson, D. 2010. Bumblebees: behaviour, ecology, and
conservation, 2nd ed. Oxford University Press, Oxford,
United Kingdom.
Landry, F., C.M.S. Plowright, and R. C. Plowright. 2000.
Behavior of individual bumble bees (Bombus impatiens)
on blueberry (Vaccinium angustifolium) owers: the effects of pollen and nectar deprivation. J. Apic. Res. 39:
29 35.
McCullagh, P. and J. A. Nelder. 1989. Generalized linear
models, 2nd ed. Chapman & Hall, New York.
Meisels, S., and H. Chiasson. 1996. Effectiveness of Bombus
impatiens Cr. as pollinators of greenhouse sweet peppers
(Capsicum annuum L.). Acta Hortic. 437: 425 430.
Morandin, L. A., T. M. Laverty, and P. G. Kevan. 2001.
Bumblebee (Hymenoptera: Apidae) activity and pollination levels in commercial tomato greenhouses. J. Econ.
Entomol. 94: 462 467.

39

Morandin, L. A., T. M. Laverty, R. J. Gegear, and P. G. Kevan.

2002. Effect of greenhouse polyethylene covering on activity level and photo-response of bumble bees. Can.
Entomol. 134: 539 549.
Plowright, C.M.S., D. Cohen-Salmon, F. Landry, and V. Simonds. 1999. Foraging for nectar and pollen on thistle
owers (Cirsium vulgare) and articial owers: how
bumblebees (Bombus impatiens) respond to colony requirements. Behaviour 136: 951963.
Plowright, C.M.S., and R. C. Plowright. 1988. Elitism in social insects: a positive feedback model (pp. 419 431). In
R. L. Jeanne (ed.), Interindividual behavioral variability
in social insects. Westview Press, Boulder, CO.
Plowright, C.M.S., R. C. Plowright, R. C., and S. L. Reid.
1995. Home or away?: the effect of mode of locomotion
on consumption of sugar by bumble bees (Hymenoptera:
Apidae). Can. J. Zool. 73: 801 804.
Plowright, C.M.S., and R. C. Plowright. 1999. Resource
management in bumble bee colonies, pp. 2128. In C. A.
Barlow (ed.), Insect population growth and the environment. Research Signpost, Trivandrum, India.
Plowright, C.M.S., and A. Silverman. 2000. Nectar and pollen foraging by bumble bees (Hymenoptera: Apidae):
choice and tradeoffs. Can. Entomol. 132: 677 679.
Stephens, D. W., and J. R. Krebs. 1986. Foraging theory.
Princeton University Press, Princeton, NJ.
Tod, C. 1986. Socio-economic effects on colony size in the
bumble bee Bombus terrestris (Hymenoptera: Apidae).
Masters dissertation, Massey University, Palmerston
North, New Zealand.
Velthuis, H.H.W., and A. van Doorn. 2005. A century of
advances in bumblebee domestication and the economic
and environmental aspects of its commercialization for
pollination. Apidologie 37: 421 451.
Weinberg, D. 2007. Simultaneous deprivation of pollen and
nectar: pollen collection behaviour and development in
bumblebee colonies (Bombus impatiens Cresson). Ph.D.
dissertation, University of Ottawa, Ottawa, Ontario.
Weinberg, D., and C.M.S. Plowright. 2006. Pollen collection by bumblebees (Bombus impatiens): the effects of
resource manipulation, foraging experience and colony
size. J. Apic. Res. Bee World 45: 2227.
Whittington, R., and M. L. Winston. 2003. Are bumble bee
colonies in tomato greenhouses obtaining adequate nutrition? Can. Entomol. 135: 883392.
Whittington, R., M. L. Winston, C. Tucker, and A. L. Parachnowitsch. 2004. Plant-species identity of pollen collected by bumblebees placed in greenhouses for pollination. Can. J. Plant Sci. 84: 599 602.
Zhang, S., F. Bock, A. Si, J. Tautz, and M. V. Srinivasan. 2005.
Visual working memory in decision making by honey
bees. Proc. Nat. Acad. Sci. 102: 5250 52.
Received 17 June 2011; accepted 1 December 2011.