Generalization of Foraging Experience Biases Bees Toward Flowers With Complex Morphologies

Abstract

The importance of pollinators as selective agents for many floral traits is well established, but understanding their role in the evolution of complex floral shapes remains challenging. This is because pollinators often need much practice to efficiently handle morphologically complex flowers and extract their food rewards. What induces foragers to persistently visit and pollinate complex flowers despite their initial low profitability? We previously found that naive bumblebees, and unsuccessful feeding attempts of experienced ones, contribute to the pollination of complex flowers. Here we tested a complementary hypothesis, positing that successful foraging on flowers of one complex shape prepares pollinators to visit other species of different complex morphologies. We trained bumblebees to computer-controlled artificial flowers that were either simple, complex or both simple and complex. We then recorded their feeding choices and handling times on a second array of simple and complex flowers that had different shapes and required another handling technique. Bees trained on a single flower type (whether simple or complex) preferred flowers of the same type in the testing array. The foragers’ preferences after training on both flower types depended on the reward schedule during training: when both flower types rewarded equally, simple flowers were preferred at the test phase; when complex flowers provided higher reward during training, they became the preferred flower type during testing. These results suggest that successful foraging on complex flowers, especially when highly rewarding, can indeed induce insect pollinators to attempt additional flower species with other complex shapes.

Full text

fevo-09-655086 May 8, 2021 Time: 17:31 # 1
ORIGINAL RESEARCH
published: 13 May 2021
doi: 10.3389/fevo.2021.655086
Edited by:
Eirik Søvik,
Volda University College, Norway
Reviewed by:
Aurore Avargues-Weber,
UMR 5169 Centre de Recherches sur
la Cognition Animale (CRCA), France
Natalie Hempel de Ibarra,
University of Exeter, United Kingdom
*Correspondence:
Tamar Keasar
tkeasar@research.haifa.ac.il
Specialty section:
This article was submitted to
Behavioral and Evolutionary Ecology,
a section of the journal
Frontiers in Ecology and Evolution
Received: 20 January 2021
Accepted: 21 April 2021
Published: 13 May 2021
Citation:
Krishna S and Keasar T (2021)
Generalization of Foraging Experience
Biases Bees Toward Flowers With
Complex Morphologies.
Front. Ecol. Evol. 9:655086.
doi: 10.3389/fevo.2021.655086
Generalization of Foraging
Experience Biases Bees Toward
Flowers With Complex Morphologies
Shivani Krishna1,2 and Tamar Keasar3*
1Department of Evolutionary and Environmental Biology, University of Haifa, Haifa, Israel, 2Department of Biology, Ashoka
University, Sonipat, India, 3Department of Biology and Environment, University of Haifa—Oranim, Haifa, Israel
The importance of pollinators as selective agents for many floral traits is well established,
but understanding their role in the evolution of complex floral shapes remains
challenging. This is because pollinators often need much practice to efficiently handle
morphologically complex flowers and extract their food rewards. What induces foragers
to persistently visit and pollinate complex flowers despite their initial low profitability?
We previously found that naive bumblebees, and unsuccessful feeding attempts of
experienced ones, contribute to the pollination of complex flowers. Here we tested a
complementary hypothesis, positing that successful foraging on flowers of one complex
shape prepares pollinators to visit other species of different complex morphologies. We
trained bumblebees to computer-controlled artificial flowers that were either simple,
complex or both simple and complex. We then recorded their feeding choices and
handling times on a second array of simple and complex flowers that had different
shapes and required another handling technique. Bees trained on a single flower
type (whether simple or complex) preferred flowers of the same type in the testing
array. The foragers’ preferences after training on both flower types depended on the
reward schedule during training: when both flower types rewarded equally, simple
flowers were preferred at the test phase; when complex flowers provided higher reward
during training, they became the preferred flower type during testing. These results
suggest that successful foraging on complex flowers, especially when highly rewarding,
can indeed induce insect pollinators to attempt additional flower species with other
complex shapes.
Keywords: Bombus terrestris, flower handling, flower morphology, generalization, learning, pollination
INTRODUCTION
Animal pollinators act as important selective agents that shape many floral traits, including color,
scent, symmetry and blooming phenology (Chittka and Raine, 2006;Schiestl and Johnson, 2013).
Pollinators are generally assumed to associate specific floral displays with the food rewards (nectar
and/or pollen) that they obtain from the flowers. High-rewarding flowers, advertised by preferred
display traits, receive better pollination service than flowers that lack them, causing such preferred
displays to increase in frequency in plant populations (Caruso et al., 2019).
This straightforward scenario seems insufficient to explain the evolution of flowers with complex
morphologies, in which food rewards are hidden in long constrained tubes, tunnels or spurs.
Frontiers in Ecology and Evolution | www.frontiersin.org 1May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 2
Krishna and Keasar Handling of Complex Flowers
Such flowers are common in numerous plant families, for
example Fabaceae, Orchidaceae, and Lamiaceae (Krishna and
Keasar, 2018). In other plants, the complexity of flower handling
involves finding the high-rewarding nectaries out of several
available ones (as in some mustards) or sonicating the anthers
to access their pollen (as in several Solanaceae species). Some of
the existing studies suggest that morphological complexity and
pollinator specialization are related. For example, simple dish-
shaped flowers are visited by a wider range of pollinator taxa
than complex flag-shaped flowers (reviewed in Keasar, 2020).
Nectar production often increases with flower complexity, while
competition with generalist foragers decreases. Specialization
on complex flowers is thus considered a profitable foraging
strategy (reviewed in Krishna and Keasar, 2018). Nevertheless,
“complexity” is difficult to measure experimentally, therefore
empirical assessment of the function relating floral complexity
and rewards available to pollinators is scarce.
Inexperienced pollinators often need much time to get at
the rewards of complex flowers, and sometimes fail to feed
from them altogether. Many repeated attempts are needed to
learn efficient motor routines for pollen (Raine and Chittka,
2007) and nectar (Laverty, 1994) extraction. Thus, complex
flowers can be highly rewarding for the subset of pollinators
that have mastered their handling technique and experience
less competition on them than on simple flowers. However,
flowers that are difficult to handle are often poor resources for
inexperienced individuals. For complex flowers to reproduce and
maintain stable populations in plant communities, pollinators
need to persist on them despite their initial low profitability.
Current models of operant conditioning in invertebrates focus
on incremental reinforcement of behaviors that are rewarding in
the short-term (Menzel and Benjamin, 2013). However, standard
learning theory does not sufficiently address learning of motor
tasks that are initially low-rewarding, but become profitable
after long practice, such as handling of complex flowers. What
behavioral mechanisms might explain the persistence of bees on
such tasks?
In a previous study, we presented bumblebees with flowers of
three species that varied in morphological complexity and were
either intact or experimentally simplified (Krishna and Keasar,
2019). We identified two foraging patterns used by the bees,
which could serve the evolutionary interests of complex flowers.
First, flower-naive bees mildly but consistently preferred complex
flowers to simplified ones. Second, unsuccessful feeding attempts
by experienced bees were common, especially to the intact flowers
of the most complex species. Such exploratory attempts, although
non-rewarding for bees, can nevertheless provide pollination
services to plant species with complex flowers.
A complementary mechanism that could potentially further
enhance pollinator visits to complex flowers involves previous
successful experience with other complex species. Previous
studies showed that bees tend to repeat flower-handling motor
routines that led them to successful foraging, and that this
preference affects their foraging choices (Laverty, 1994;Dukas,
1995). Bumblebees that foraged in mixed patches of two or
three flower species were usually flower-constant, i.e., visited a
single flower species for many consecutive visits. When switches
between species did occur, they were mostly between flowers that
required similar handling techniques (Ishii and Kadoya, 2016).
Bumblebees that were trained to artificial flowers, accessible
either from the bottom or from the top of their corollas,
retained the acquired handling method when they could access
those flowers by both techniques (Barker et al., 2018). Such
“tactic-constancy” was proposed to improve foraging efficiency,
since the time required for flower handling decreases as its
handling technique becomes routine (Bronstein et al., 2017).
Most of these earlier studies have focused on either flower
handling behavior and their associated costs or on transfer of
motor skills. Here, placing both these together and expanding
on these findings, we hypothesized that learning to handle
complex flowers may involve motor skills that can be later
transferred and applied to other species with different shapes.
Such a mechanism would make complex flowers attractive to
foragers, after they had previously foraged on other complex
species. To test the hypothesis, we trained bumblebees to
artificial flowers that were either simple, complex or both
simple and complex. We next tested the foragers’ choices in
a flower patch with equal numbers of simple and complex
morphologies that required a different handling technique from
the training flowers. We addressed the following questions:
(a) Are the bees’ overall foraging choices, feeding success and
foraging costs (handling time) in the experiment’s test phase
affected by the training that they had received? (b) How do
these change with experience of the bees along the experiment’s
test phase?
We hypothesized that the foraging choices would be strongly
influenced by the training type. Specifically, we predicted
that training to simple flowers would increase the bees’ visit
frequencies and feeding success on other simple flowers in the
test phase, whereas training to complex flowers would induce
more test-phase visits to complex flowers. When trained to both
the flower types with equal rewards, we predicted that bees
would equally visit both flowers. In the above scenarios, based
on previous findings with real (Krishna and Keasar, 2019) and
artificial (Barker et al., 2018) flowers, we also expected bees to
visit simple artificial flowers increasingly as they gain experience
in the testing array, regardless of their training. To explore
the interactive effect of food rewards and foraging experience
on choices of complex morphologies, we also trained foragers
to complex flowers that were paired with higher rewards than
simple flowers. We expected that more of these bees would
specialize on complex flowers in the test phase than after training
on equally rewarding complex and simple flowers.
MATERIALS AND METHODS
Four Bombus terrestris colonies were obtained from BioBee Ltd.
(Sde Eliyahu, Israel). Colonies were housed in a 2.90 ×3.60 m
flight room, which was maintained at a temperature range of 26–
29◦C and a relative humidity range of 40–60%. The room was
illuminated by D-65 lights between 0600 and 1,800 h. The colony
was placed at one end of the room with a plexiglass corridor
attached to its entrance, allowing us to control the exit of foragers
Frontiers in Ecology and Evolution | www.frontiersin.org 2May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 3
Krishna and Keasar Handling of Complex Flowers
from the colony. Pollen was provided directly to the colony
every day. Flower-naive bees were pre-trained to a transparent
feeder (Petri dish) filled with 30% sucrose solution (w/w) and
placed on the experimental table. Foragers were marked with
numbered tags after they learnt to visit the feeder regularly. For
experiments, we used robotic flowers that were developed in the
lab. The automated flower system gave us the flexibility to track
the bees’ visit sequences and durations to numerous flowers, and
to precisely control the floral rewards.
Artificial Flower System
Thirty-two artificial flowers were used in the experiments. The
dimensions and basic design of the flowers are described in detail
in Keasar (2000). They contained an infrared (IR) sensor, an
electromagnet, a plexiglass reservoir filled with sugar solution,
and a plexiglass float with a magnet glued to it. The flowers
were computer-controlled by an Arduino Mega 2560, single-
board microcontroller platform1, which is based on ATmega2560
processor chip. An external power supply was used to provide
direct current to all flowers in the system. The program
running on the microcontroller handled the electronics in the
system: it read infrared sensors and generated signals to push
the electromagnet down and up for refilling (Supplementary
Movie 1). The display color of all the flowers used in the
experiments was blue, to eliminate potential effects of the bees’
color preferences on their choices. For the same reason, the sugar
solution was spiked with the same scent in all flowers. The flowers
were placed at least 10 cm apart in both the training and test
phases of the experiment (see below). To rule out social learning
during the experiment, a single bee foraged on the artificial
flowers at a time. After a foraging session, flowers were wiped
with ethanol to remove any potential scent marks that might
influence the next bee.
Experimental Design
We conducted the experiment in two phases: a training and a
test phase. Simple flowers in the training phase had one feeding
hole, and complex flowers had three holes, of which only the
central one led to reward. This design of the flowers is inspired
from some of the flower species (such as mustards, Sinapis spp.)
which provide differential rewards in their nectaries and share
a common floral morphology. In the test phase, the simple
morphology was represented by a flower with single feeding hole
accessible through a transparent tunnel. The tunnel was placed
such that the morphology was visible through the sides of the
tunnel as well as from the top if the bee hovered close to the
flowers (Supplementary Figure 1). The complex morphology
was represented by a flower with three holes, one of which
was a feeding hole, accessible along a similar transparent tunnel
as the simple flower. Thus, in the test phase, the bees had to
learn a new technique for accessing the flowers that required
crawling in the tunnel, and which was different from the handling
technique of the training phase. During the training phase, the
time taken to access the reward was higher on the complex
flowers (28.0 ±3.7 s) than the simple flowers (15.9 ±1.8
1http://www.arduino.cc
s, Wilcoxon test: W =202.5, p=0.004), suggesting that the
complex morphology was indeed the difficult option for the bees
to begin with (also see Supplementary Figure 2).
During the training phase, we randomly allocated bees from
each of the four colonies to one of four floral arrays: (a) “Simple”:
16 simple flowers (17 bees); (b) “Complex”: 16 complex flowers
(15 bees); (c) “SCequal”: equally rewarding simple and complex
flowers (50 visits on simple and 50 visits on complex, 20 bees); or
(d) “SChigh”: simple and higher-rewarding complex flowers (50
visits on simple and 50 visits on complex, 20 bees). These four
floral arrays are referred to as “training types.” Supplementary
Table 1 contains further details on the assignment of bees to
training types. Each bee was exposed to one training array only.
To provide higher reward in the complex flowers in “training
type (d),” they were filled with a 50% (w/w) sucrose solution.
The simple flowers in this, and all flowers of training types (a–c),
dispensed a 40% (w/w) sucrose solution. The training comprised
100 successful (fed) visits on the arrays mentioned above. One
half of the bees allocated to (c) and (d) were first trained to
simple, then to complex flowers. This was done by concealing
the flowers of complex morphology and allowing the bee to
complete their 50 visits on simple flowers, and then exposing
the complex flowers and concealing the simple ones for 50 more
visits. We reversed the training order for the remaining bees.
This sequential way of exposing the flower array allowed us to
accurately control the amount of experience they have received
on each flower type.
In the test phase, bees individually foraged on an array of 16
simple and 16 complex flowers for up to 100 visits (4–5 bouts).
All of these flowers rewarded the bees with 1 µl of a 40% sucrose
solution on each visit. Bees were allowed to forage freely till they
returned to the colony by themselves. Some bees returned to the
colony after their first bout and did not attempt to exit again on
the same day, thus they completed fewer visits. A scheme of the
experimental design is provided in Figure 1.
In the training phase, we counted only successful visits (i.e.,
visits that involved sucrose feeding). In the test phase, any landing
by a bee on a flower tunnel (where the legs have contacted the
flower) was regarded as a choice (or visit). Those visits that
resulted in successfully accessing the reward were considered as
fed visits. Unsuccessful visits included failures to either locate
the rewarding hole or to feed from it. The experiments were
video recorded, and the amount of time taken to access the
reward (henceforth referred to as “handling time”) was extracted
from the videos.
Ethical Note
Bees were kept in the colony that they were brought in and
maintained under 12 h L: 12 h D with ambient temperature
and humidity conditions. Pollen and sucrose were provided
ad libitum to the colony without causing any disturbance. Bees
were captured while feeding with a sponge-lined cage and
handled with care. Identification tags were attached to the thorax
with a drop of non-toxic glue and the bees were carefully released
after being marked. The colonies were euthanized by placing in
the freezer after all the foragers have died naturally and only
drones remained in the colony.
Frontiers in Ecology and Evolution | www.frontiersin.org 3May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 4
Krishna and Keasar Handling of Complex Flowers
FIGURE 1 | Experimental design and scheme of the artificial flowers. Bees were allocated to one of four training types on artificial flowers with a flat landing surface.
In the test phase, all foragers were exposed to a mixture of simple and complex flowers, which could only be accessed by passing through a short tunnel. Floral
choices, feeding success and handling times during the test phase were recorded.
Statistical Analysis
All the analyses were conducted using R statistical software (R
version 3.3.2). The test-phase choices of bees that had been
trained on both flower types (“SCequal” and “SChigh ”), were
not affected by the flower presentation order during training
(Wilcoxon test: “SCequal”: W =39, p=0.43 and “SChigh”: W =54,
p=0.76), therefore the data of all bees from each training type
were pooled for analysis.
We tested whether the training type affected the frequency
of visits to the complex flowers using generalized linear mixed
effects models (binomial GLMM, lme4 package, Bates et al.,
2014). The response variable was flower type (complex/simple)
and the fixed effects were training type and visit number during
the test phase (experience). Bee and colony identities were used as
random effects. The significance of each fixed effect in the model
was determined using the function “Anova” (car package, Fox
et al., 2012). A similar model was run considering only successful
visits. Mean run lengths (consecutive visits to a flower type) were
computed for each of the bees. We examined, using a linear mixed
effects model, whether the training type and the type of flowers
visited in the test array (simple and complex) influenced run
lengths. We computed a constancy index (Bateman, 1951) for
each forager, which is based on the number of transitions between
the same flower type and different flower types (complex and
simple). The value of this index ranges from −1 (switching flower
type after each visit) to 1 (complete constancy). We used a one-
way ANOVA to analyse if the constancy index differed between
the training types.
To determine whether the time taken to successfully access
the reward (log-transformed handling time) from complex and
simple flower types depended on training type, experience (visit
number) in the test phase, or their interactive effect, we used
a linear mixed effects model with bee and colony identity
as random effects.
In all the models, when a significant effect of the predictor
variable was found, we ran Tukey’s post hoc tests (R package
lsmeans, Lenth, 2016) to determine pairwise differences.
RESULTS
Flower-Type Choices
First Visits
0.35 of the bees that were trained on simple flowers, and 0.73
of those trained on complex flowers, made their first visit to a
complex flower in the test phase. Complex flowers were chosen
on the first visit of the test phase by 0.45 of the foragers that
were trained to equally rewarding simple and complex flowers
(“SCequal”), and 0.60 of those trained to both flower types with
higher reward in complex flowers (“SChigh”). While none of
these proportions deviated significantly from 0.5 (binomial tests,
p>0.05 in all tests), a significantly larger proportion of bees
trained on complex flowers made their first visit to a complex
flower, compared to bees trained on simple flowers (Complex vs.
Simple, χ2=4.63, df =1, p=0.03).
All Visits
The type of training experienced by the bees affected their choices
over the whole test phase (GLMM, training: χ2=56.94, df =3,
p<0.0001). In the “Simple” and “Complex” training types, there
was a bias toward visiting the flower type that the bees had been
trained on. Training on equally rewarding simple and complex
flowers (“SCequal”) resulted in choices that were intermediate
between “Simple” and “Complex” training types. After training
on both flower types with higher rewards in complex flowers
(“SChigh”), bees chose complex flowers as often as individuals
from the “Complex” training type (Figure 2A). Most of the visits
to both simple and complex flowers resulted in sucrose feeding
(Figure 2B). The foragers’ experience during the test phase, alone
and in interaction with the training type, did not influence their
flower-type choices (Training∗Experience: χ2=3.26, df =3,
p=0.35, Experience: χ2=0.03, df =1, p=0.84, Figure 3).
The bees’ preference for a flower morphology (whether simple
or complex) was manifested in longer sequences of consecutive
visits (run lengths) to their favored flower type than to the
less-preferred type (Figure 4). Thus, training type significantly
Frontiers in Ecology and Evolution | www.frontiersin.org 4May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 5
Krishna and Keasar Handling of Complex Flowers
FIGURE 2 | Mean ±SE percent of test-phase visits in complex flowers.
(A) Both successful and unsuccessful visits, (B) only successful (fed) visits.
Different letters indicate significant differences in Tukey post hoc tests, which
are detailed in Supplementary Table 2.
FIGURE 3 | The effects of experience during the test phase on the mean ±SE
percentages of complex flowers chosen by the bees. Note that experience
(visit number) was analyzed as a continuous variable in the statistical models.
In the figure, however, visits are grouped into bins of 10 for graphical clarity.
affected run lengths in the test phase (χ2=16.55, df =3,
p<0.0001). Bees trained to “Simple” and to “SCequal” flowers
made shorter runs of visits to complex flowers than bees trained
to “Complex” and to “SChigh” flowers (Figure 4).
We defined a forager as specialized if it made at least 60%
of its visits to one of the flower types (the 60% threshold was
set arbitrarily). The proportion of specialized bees varied among
experimental treatments, being higher in bees trained on simple
flowers than in the other experimental treatments (Chi-square
test of independence, χ2=16.55, df =6, p<0.0001). In
this training type, all of the specialized bees foraged mostly on
simple flowers. After training on complex flowers only, or on
flower types with higher reward in complex (“SChigh”), most of
the specialization was on the complex type. Training on equally
rewarding simple and complex flowers (“SCequal”) resulted in
most specializations on the simple morphology (Figure 5).
Barring the bees that were trained on only simple flowers, 0.40
of the bees in all the three training types (“Complex” =0.46,
SCequal =0.65, SChigh =0.45) did not specialize on either flower
morphology, namely directed 0.41–0.59 of their visits to complex
flowers.
The mean constancy index of bees trained to better rewards in
complex flower types was higher than in the other training types
(“SChigh”: 0.31 >“SCequal ”: 0.26 >“Simple”: 0.25 >“Complex”:
0.22, Supplementary Figure 3). However, this difference was not
statistically significant (F(3,68)=1.05, p=0.37), as bees within
each training type varied widely in the frequencies of transitions
between flower types.
Feeding Success
The type of training experienced by the bees influenced their
success at accessing the reward during the test phase (GLMM,
training: χ2=45.83, df =3, p<0.0001). Bees that were trained
on complex flowers alone, and bees that were trained with higher
rewards in complex flowers, had similar success rates, and were
more successful at obtaining reward from complex than from
simple flowers. The foragers’ experience during the test phase,
alone and in interaction with the training type, did not affect
their success at flowers (Training∗Experience: χ2=4.05, df =3,
p=0.25, Experience: χ2=2.21, df =1, p=0.13, Figure 6 and
Supplementary Table 4).
Handling Time
The time needed to handle the flowers and access the reward
successfully declined with experience during the test phase
(Experience: χ2=577.23.15, df =1, p<0.0001, Figure 7).
There were no significant differences between handling times of
complex and simple flowers and they were similar across the four
training types (Supplementary Table 5).
DISCUSSION
Training to a single flower-handling method (using either a single
hole, or one of three holes, to access a nectar reward) induced
bees to apply it to a new foraging task during the test phase of our
experiment. Thus, bees trained to complex flowers chose other
complex flowers in the test phase more often than bees trained
to simple flowers. In bees that were trained to both flower types,
the food rewards during training affected the bee’s later choices.
Increasing the relative profitability of complex flowers in the
training phase caused bees to choose complex flowers more often
when tested. These results are in line with our first prediction,
that a successful foraging experience on one complex flower
species would increase the bees’ preference for other complex
species. Moreover, the significance of such foraging experience
is also evident from the absence of difference in the proportion of
unsuccessful and successful visits to the flowers in the test phase.
We further predicted that the bees’ training-phase experience
would affect their choices primarily at the start of the test phase.
This prediction derived from our previous experiment with real
flowers, where bees gradually reduced their attempts to feed from
Frontiers in Ecology and Evolution | www.frontiersin.org 5May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 6
Krishna and Keasar Handling of Complex Flowers
FIGURE 4 | Mean ±SE run lengths (numbers of consecutive visits) to simple and complex flowers during the test phase. Different letters indicate significant
differences in Tukey post hoc tests, which are detailed in Supplementary Table 3.
FIGURE 5 | The proportions of bees that specialized on simple flowers, specialized on complex flowers, or showed no flower type specialization during the test
phase. Specialized foragers directed >60% of their visits to one of the flower types.
complex flowers when simplified and equally rewarding complex
flowers were available (Krishna and Keasar, 2019). Contrary to
this expectation, the bees’ experience in the test phase did not
affect their overall flower choices (Figure 3) or feeding success
(Figure 6) in the present study. We also found that bees handled
simple and complex flowers equally quickly during the test phase
(Supplementary Table 5). It thus seems that our complex and
simple artificial flowers became similarly profitable to foragers
by the end of the training phase, since they provided identical
rewards and required similar handling times. Hence no benefit, in
terms of foraging success, would have been gained from changing
flower choices over the course of the test phase. Furthermore,
in our test phase, we considered a bee to visit or make a choice
only if it landed. Previous studies have demonstrated that bees
sample and discriminate floral features from close distances,
thereby making some of the foraging decisions without probing
the flowers (Dukas, 2001;Ma et al., 2016). Such rejections without
landing possibly explain how bees that had been trained to
complex flowers, and that received higher rewards on them, chose
complex flowers from the very beginning of the test phase. The
similar profitability of both flower types in the test phase differs
from our earlier observations on complex real flowers, which
required longer handling times than simple ones and offered
lower feeding success, even for experienced bees (Krishna and
Keasar, 2019). On the other hand, in a previous experiment with
simple and complex artificial flowers, experienced bees attained
similar handling times in both flower types (Muth et al., 2015).
Our experiment also allows unplanned comparisons between
bees trained to a single motor task (“Simple” and “Complex”)
and identically rewarded bees exposed to two tasks at training
(“SCequal”). Bumblebees readily learnt to perform one of two
motor routines (turning either right or left in a Y-maze) in
response to a color cue, and retained such learnt associations
for several days (Chittka and Thomson, 1997;Chittka, 1998).
Frontiers in Ecology and Evolution | www.frontiersin.org 6May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 7
Krishna and Keasar Handling of Complex Flowers
FIGURE 6 | The effects of experience during the test phase on the mean ±SE percentages of successful (fed) visits to complex flowers. As in Figure 3, experience
(visit number) was analyzed as a continuous variable in the statistical models but is plotted here in groups of 10 visits for visual clarity.
FIGURE 7 | Mean ±SE handling times (in seconds) of complex (black) and simple (gray) flowers throughout the experiment’s test phase. Although aggregated here
into 10-visit bins for visual clarity, the handling time of each visit was treated as a data point in the statistical analysis.
However, they were slower and more error-prone than bees
trained to one of these tasks only. In our experiment, on the
other hand, handling times were similar across all training types
(Figure 7). The bees’ feeding success in “SCequal” also resembled
the average feeding success of “Simple” and “Complex” training
types (Figure 2B). Thus, we found no reduction in foraging
efficiency in bees that learnt two motor tasks. Possibly, our
learning tasks were easier than those posed by Chittka and
Thomson (1997) and Chittka (1998), allowing the bees in our
study to reach high foraging proficiency even after training
on two routines. Note also that the design of our experiment
confounds the number of training tasks (one vs. two) with the
length of training (100 to a single flower type vs. 50 visits to each
of the two flower types).
Barker et al. (2018) trained bumblebees to either a simple
(“secondary robbing”) or a more complex (“legitimate”) nectar
extraction technique from artificial flowers, and then exposed
them to similar flowers that could be exploited by both
techniques. As in our experiment, the bees initially preferred
the technique to which they had been trained. Our study
extends these findings by showing that practice with the
complex technique induces preference for a complex flower
type that the foragers had not encountered before. These
findings are in line with extensive previous demonstrations of
bees’ ability to generalize a learnt preference to perceptually
similar stimuli (Gerber et al., 1996;Vieira et al., 2018). Most
studies on generalization and transfer in bees have focused on
learning of olfactory and visual signals in classical conditioning
experiments. Our work adds evidence for transfer learning of a
complex motor task.
Natural plant communities in bloom comprise multiple
species, which vary in flower shape, handling difficulty and food
Frontiers in Ecology and Evolution | www.frontiersin.org 7May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 8
Krishna and Keasar Handling of Complex Flowers
rewards. The frequency of species with complex flowers correlates
with the composition of the pollinator community, increasing
in habitats that are rich in specialized pollinators such as birds,
large bees and lepidopterans. Such habitats include islands, alpine
areas and forest canopies (Keasar, 2020). Pollinating insects
distribute among the available flower species proportionately
to their detectability, accessibility, abundance and profitability
(Junker et al., 2013). Yet, how this distribution is achieved
in practice, and which behavioral mechanisms shape pairwise
species interactions in pollination networks, is not well known
(Olito and Fox, 2015). Our experiment suggests that foragers
that had learnt to handle a complex flower species may remain
specialized to complex floral morphologies in new foraging
situations. Similarly, bumblebees showed high constancy to their
flower-handling tactics (legitimate visits in some plant species,
nectar-robbing in others) in a natural montane plant community
(Lichtenberg et al., 2020). Such transfer of preferences could
help explain how plants with complex flowers persist in diverse
plant communities, even though they have low profitability
for naïve pollinators. Likewise, in plant species with poricidal
anthers bees have to learn to efficiently buzz in order to extract
pollen and they extend this buzzing to species with non-
poricidal anthers, such as Pedicularis spp. (Buchmann, 1985).
Such a “carry over” phenomenon (Buchmann, 1983) is likely to
benefit flowers with non-poricidal anthers that release pollen in
response to the buzzing, if they co-occur with species containing
poricidal anthers. In our study, the training arrays resembled
some common field conditions, such as in a seasonally flowering
community where complex flowered species are the first to
bloom and communities where complex flowered species provide
greater rewards compared to co-flowering simple flowered
species (bumblebees are known to frequently sample other
species as part of their minoring). Such design allowed us to better
understand possible ways by which repeated visits and thereby
successful pollination is achieved in complex flowered species.
Information on the frequencies of complex flowered species and
on their spatial and temporal distributions in communities will
help us comprehend handling strategies used by bees at the
individual as well as the colony level.
From the pollinators’ point of view, extrapolating from prior
experience in complex floral environments is a key skill that can
be put to use to improve foraging efficiency. The transfer of
preferences documented in our experiment may result in some
within-colony foraging specialization under field conditions, with
part of the workers visiting mainly simple flowers and others
preferring complex ones. This may be an efficient colony-level
foraging strategy, as it allows the colony to utilize a wide range
of flower species, while enabling some workers to specialize
on the high-risk-high-reward complex species. How naïve and
experienced workers from a colony evaluate and utilize social
cues while handling such complex flowers is intriguing, but
this question was beyond the scope of the present experiment.
Whether and how colony-level specializations in fact occur in
social pollinators is a promising direction for further study.
DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online
repositories. The names of the repository/repositories and
accession number(s) can be found below: https://tamarkeasarlab.
weebly.com/data-sets.html.
AUTHOR CONTRIBUTIONS
TK conceived the study. TK and SK designed the experiments. SK
conducted the experiments and statistical analysis. Both authors
were involved in writing and editing the manuscript drafts.
FUNDING
This work was supported by grant #250/16 from the Israel
Science Foundation.
ACKNOWLEDGMENTS
We are extremely thankful to Yehuda Agus for the development
of artificial flower system used in the study.
SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found
online at: https://www.frontiersin.org/articles/10.3389/fevo.2021.
655086/full#supplementary-material
REFERENCES
Barker, J. L., Dornhaus, A., Bronstein, J. L., and Muth, F. (2018). Learning about
larceny: experience can bias bumble bees to rob nectar. Behav. Ecol. Sociobiol.
72:68. doi: 10.1007/s00265-018-2478-6
Bateman, A. J. (1951). The taxonomic discrimination of bees. Heredity 5, 271–278.
doi: 10.1038/hdy.1951.24
Bates, D., Mächler,M., Bolker, B., and Walker, S. (2014). Fitting linear mixed-effects
models using lme4. J. Statist. Softw. arXiv 1406:5823.
Bronstein, J. L., Barker, J. L., Lichtenberg, E. M., Richardson, L. L., and Irwin, R. E.
(2017). The behavioral ecology of nectar robbing: why be tactic constant? Curr.
Opin. Insect Sci. 21, 14–18. doi: 10.1016/j.cois.2017.05.013
Buchmann, S. L. (1983). Buzz pollination in angiosperms. In C. E. Jones and R. J.
Little (eds.), Handbook of Experimental Pollination Biology. (New York: Van
Nostrand Reinhold).
Buchmann, S. L. (1985). Bees use vibration to aid pollen collection from non-
poricidal flowers. J. Kansas Entomol. Soc. 58, 517–525.
Caruso, C. M., Eisen, K. E., Martin, R. A., and Sletvold, N. (2019). A meta-analysis
of the agents of selection on floral traits. Evolution 73, 4–14. doi: 10.1111/evo.
13639
Chittka, L. (1998). Sensorimotor learning in bumblebees: long-term retention and
reversal training. J. Exp. Biol. 201, 515–524.
Chittka, L., and Raine, N. E. (2006). Recognition of flowers by pollinators. Curr.
Opin. Plant Biol. 9, 428–435. doi: 10.1016/j.pbi.2006.05.002
Frontiers in Ecology and Evolution | www.frontiersin.org 8May 2021 | Volume 9 | Article 655086

fevo-09-655086 May 8, 2021 Time: 17:31 # 9
Krishna and Keasar Handling of Complex Flowers
Chittka, L., and Thomson, J. D. (1997). Sensori-motor learning and its relevance
for task specialization in bumble bees. Behav. Ecol. Sociobiol. 41, 385–398.
doi: 10.1007/s002650050400
Dukas, R. (1995). Transfer and interference in bumblebee learning. Anim. Behav.
49, 1481–1490. doi: 10.1016/0003-3472(95)90069-1
Dukas, R. (2001). Effects of perceived danger on flower choice by bees. Ecology Lett.
4, 327–333. doi: 10.1046/j.1461-0248.2001.00228.x
Fox, J., Weisberg, S., Adler, D., Bates, D., Baud-Bovy, G., Ellison, S., et al. (2012).
Package ‘car’. (Vienna: R Foundation for Statistical Computing).
Gerber, B., Geberzahn, N., Hellstern, F., Klein, J., Kowalksy, O., Wüstenberg, D.,
et al., (1996). Honey bees transfer olfactory memories established during flower
visits to a proboscis extension paradigm in the laboratory. Anim. Behav. 52,
1079–1085. doi: 10.1006/anbe.1996.0255
Ishii, H. S., and Kadoya, E. Z. (2016). Legitimate visitors and nectar robbers on
Trifolium pratense showed contrasting flower fidelity versus co-flowering plant
species: could motor learning be a major determinant of flower constancy by
bumble bees? Behav. Ecol. Sociobiol. 70, 377–386. doi: 10.1007/s00265-016-
2057-7
Junker, R. R., Blüthgen, N., Brehm, T., Binkenstein, J., Paulus, J., Martin Schaefer,
H., et al. (2013). Specialization on traits as basis for the niche-breadth of flower
visitors and as structuring mechanism of ecological networks. Funct. Ecol. 27,
329–341. doi: 10.1111/1365-2435.12005
Keasar, T. (2000). The spatial distribution of nonrewarding artificial flowers
affects pollinator attraction. Anim. Behav. 60, 639–646. doi: 10.1006/anbe.2000.
1484
Keasar, T. (2020). Patterns of flower complexity in plant communities. Annu. Plant
Rev. 3, 643–660. doi: 10.1002/9781119312994.apr0757
Krishna, S., and Keasar, T. (2018). Morphological complexity as a floral signal: from
perception by insect pollinators to co-evolutionary implications. Int. J. Mol. Sci.
19:1681. doi: 10.3390/ijms19061681
Krishna, S., and Keasar, T. (2019). Bumblebees forage on flowers of increasingly
complex morphologies despite low success. Anim. Behav. 155, 119–130. doi:
10.1016/j.anbehav.2019.06.028
Laverty, T. M. (1994). Bumble bee learning and flower morphology. Anim. Behav.
47, 531–545. doi: 10.1006/anbe.1994.1077
Lenth, R.V. (2016). Least-squares means: the R Package lsmeans. J. Stat. Softw. 69,
1–33.
Lichtenberg, E. M., Richman, S. K., Irwin, R. E., and Bronstein, J. L. (2020).
Competition for nectar resources does not affect bee foraging tactic constancy.
Ecol. Entomol. 45, 904–909. doi: 10.1111/een.12866
Ma, X., Shi, J., Bänziger, H., Sun, Y., Guo, Y., Liu, Z., et al., (2016). The functional
significance of complex floral colour pattern in a food-deceptive orchid. Funct.
Ecol. 30, 721–732. doi: 10.1111/1365-2435.12571
Menzel, R, Benjamin P. (Eds) (2013). Invertebrate Learning and Memory. (London:
Elsevier).
Muth, F., Keasar, T., and Dornhaus, A. (2015). Trading off short-term costs
for long-term gains: how do bumblebees decide to learn morphologically
complex flowers? Anim. Behav. 101, 191–199. doi: 10.1016/j.anbehav.2014.
12.024
Olito, C., and Fox, J. W. (2015). Species traits and abundances predict metrics of
plant–pollinator network structure, but not pairwise interactions. Oikos 124,
428–436. doi: 10.1111/oik.01439
Raine, N. E., and Chittka, L. (2007). Pollen foraging: learning a complex motor
skill by bumblebees (Bombus terrestris). Naturwissenschaften 94, 459–464. doi:
10.1007/s00114-006- 0184-0
Schiestl, F. P., and Johnson, S. D. (2013). Pollinator-mediated evolution of floral
signals. Trends Ecol. Evol. 28, 307–315. doi: 10.1016/j.tree.2013.01.019
Vieira, A. R., Salles, N., Borges, M., and Mota, T. (2018). Visual discrimination
transfer and modulation by biogenic amines in honeybees. J. Exp. Biol.
221:jeb178830. doi: 10.1242/jeb.178830
Conflict of Interest: The authors declare that the research was conducted in the
absence of any commercial or financial relationships that could be construed as a
potential conflict of interest.
The reviewer NI declared a past co-authorship with one of the authors SK to the
handling editor.
Copyright © 2021 Krishna and Keasar. This is an open-access article distributed
under the terms of the Creative Commons Attribution License (CC BY). The use,
distribution or reproduction in other forums is permitted, provided the original
author(s) and the copyright owner(s) are credited and that the original publication
in this journal is cited, in accordance with accepted academic practice. No use,
distribution or reproduction is permitted which does not comply with these terms.
Frontiers in Ecology and Evolution | www.frontiersin.org 9May 2021 | Volume 9 | Article 655086