Flower movement increases pollinator preference for flowers with better grip


Correspondence author. E-mail: bjg26@cam.ac.uk


1.Conical cells in the petal epidermis are common across many diverse flowering plant species, and it was recently shown that in difficult-to-handle flowers, pollinators prefer conical cells because they increase grip. However, this does not explain the prevalence of conical cells amongst other, simpler, flowers.

2.The movement of objects is an integral part of the world and is of particular importance to bees because the relative motion of objects is essential to a bee's 3D vision. The motion of flowers can increase pollinator attraction; however, it also makes flowers more difficult for a bee to handle. This makes foraging more metabolically expensive. To explore whether conical petal cells make handling moving flowers easier, we tested bumblebee (Bombus terrestris) preference for conical- or flat-celled Petunia (Petunia hybrida) flowers under different conditions of motion. We also used differently coloured Petunia flowers to test how colour and visibility interact with tactile cues to form a pollinator's preferences.

3.Bees preferred to visit conical-celled Petunia flowers except when the conical-celled flowers were harder to detect visually. The bees then favoured flowers that were easier to detect. But when flowers were moving and more difficult to handle, bees always learned to favour conical-celled flowers, irrespective of visual difficulty.

4.By providing easier handling through better grip from conical cells, the plant can benefit from the natural visual attractant of flowers moving in the wind without losing pollinator preference for easier-to-handle flowers. Bee preference for conical-celled flowers when flowers are moving shows how plants can use conical petal cells to take advantage of an attractant that would otherwise decrease pollinator preference by making handling difficult (movement). The selective pressure from pollinators choosing conical-celled flowers when flowers are moving in the wind provides an explanation for the persistence of conical cells in so many diverse angiosperm species across evolutionary time.


Conical (or papillate) cells in the petal epidermis are a common trait, present in the majority of extant flowering plants (Kay, Daoud, & Stirton 1981; Christensen & Hansen 1998). Conical cells provide tactile properties that benefit the plant through influencing pollinator grip and therefore preference (Whitney et al. 2009, 2011a). This study aims to investigate how this tactile benefit relates to flower movement and how visual cues interact with handling properties to form pollinator preferences.

The formation of the conical shape of the epidermal petal cells is controlled by an R2R3 MYB transcription factor called MIXTA (Noda et al. 1994). Without a functional MIXTA protein, epidermal cells remain flat. The most conspicuous effect of this loss of conical cells is in the colour of the flower, as mixta flowers have (to the human eye) a much duller colour than wild-type flowers (Fig. 1). Light incident on a conical-celled petal epidermis is focused into the central pigment-containing vacuole and more light is absorbed by the pigment, whereas on a flat-celled epidermis light passes through the epidermal cells and into the unpigmented mesophyll underneath (Gorton & Vogelmann 1996). More white light is also scattered back towards the observer from flat cells, which dilutes the colour signal (Noda et al. 1994).

Figure 1.

(a) Flowers from the three lines of Petunia used in this experiment, showing similarity of colour to the human eye. C, conical celled; F, flat celled. (b) Spectrophotometer readings of light transmittance of the three flower types. (c) Colour hexagon positions of each flower, showing that all three flower types are very close in bee colour space.

Conical petal cells benefit plant fecundity through their interactions with pollinators. Field trials of Antirrhinum majus showed that when pollinator number limits seed set, plants with conical petal cells set more seed than flat-celled plants (Glover & Martin 1998). There may be many reasons for this, as conical petal cells have diverse effects on a flower, including enhancing colour, slightly increasing floral temperature (Comba et al. 2000; Whitney et al. 2011a), reducing petal wettability (Whitney et al. 2011b) and altering overall petal shape by changing the direction of cell expansion in the epidermis (Baumann et al. 2007). Bees are able to use most of these traits as learned indicators of rewards, and some (such as temperature) are traits that a bee will seek out as a reward in itself (Dyer et al. 2006; Rands & Whitney 2008). While scent is an important cue in pollination by bumblebees, the loss of conical cells as a result of mutation of MIXTA-like genes does not alter the type, proportion or total amount of volatiles produced (Whitney et al. 2009). For Antirrhinum, although bumblebees can easily distinguish the different colours of the conical- and flat-celled flowers, they have no innate preference towards either colour, and there is no difference in salience as measured by the distance at which a bee can distinguish the flower from a neutral background (Dyer et al. 2007). Instead for Antirrhinum, it is a tactile benefit that increases the attractiveness of conical-celled flowers to their pollinators. Its flowers are particularly difficult for a pollinator to handle, having a small landing platform and a hinge mechanism the bee must operate before it can reach the nectar. The textured surface of a conical-celled flower is easy for bees to grip and allows bees to come to rest while they drink, whereas flat-celled flowers are slippery and bees must expend energy to maintain their grip on the flower (Whitney et al. 2009). Being able to rest while drinking may provide a metabolic advantage and cause bees to favour the conical-celled flowers, and the difference in colour enables bees to use colour as a learned cue to discriminate between flowers before landing.

Despite the obvious tactile advantage to pollinator handling of complicated flowers, there is no one flower form that correlates with the presence of conical cells in the petal epidermis, nor is there evidence across the angiosperms of a correlation between the presence of conical cells and flower presentation angle (Rands, Glover, & Whitney 2011). In trying to determine why conical petal cells are so prevalent, it may not be possible to usefully generalise from studies using Antirrhinum. The very specific method of flower handling in Antirrhinum may create equally specific pollinator preferences, so the selective pressures affecting Antirrhinum are very likely to differ from those of plants with simpler flowers.

Flowers that initially appear relatively easy for a pollinator to handle may still benefit from producing petal surfaces that enhance pollinator grip under certain conditions, such as when flowers are moving. Motion is an integral part of a bee's perception of their environment. Bees have limited use of binocular stereopsis (Brünnert, Kelber, & Zeil 1994), and their ability to perceive depth and position in three-dimensional space is strongly dependent on motion parallax and the perceived change in the position of stationary objects because of movement of the observer (Lehrer et al. 1988; Srinivasan, Lehrer, & Horridge 1990). Bee target detection is significantly improved in 3D tasks when bees use motion parallax in conjunction with colour cues (Kapustjansky, Chittka, & Spaethe 2010), and bees are also innately attracted to moving objects (Lehrer & Srinivasan 1992). Recently, plants have been shown to exploit movement as a cue for pollinator attraction, with a change in stalk length as a measure of flower ‘waviness’ correlating to a change in attractiveness to pollinators (Warren & James 2008). Flowers that move in the breeze may be able to take advantage of a bee's normal use of motion as part of their vision system, but the motion of flowers under natural variable wind conditions could make flowers more difficult to handle.

By presenting bees with Petunia flowers, which have a large, flat, open corolla that is easy for bees to land on and drink from (Fig. 1a), we investigated whether the tactile benefit of conical cells as observed for snapdragon flowers could be extrapolated to other, simpler flowers. In the phmyb1 mutant of Petunia hybrida, the PhMYB1 gene, orthologous to MIXTA, has been disrupted by the insertion of the dTph1 transposon. This results in petal epidermal cells that are much flatter than the tall cones of the wild type (van Houwelingen et al. 1998; Baumann et al. 2007). Using the phmyb1 mutant and a revertant wild-type line, in combination with an established and ecologically relevant behavioural system, we were able to test bee preferences. Additionally, as efficiency in foraging time is an important factor for bees when choosing which flowers to visit, using flowers of the V26 line of Petunia, which have similarly hued but much darker flowers, we were able to test the effect of visual difficulty on bee preferences for conical cells. We then imposed an added tactile difficulty by presenting bees with moving flowers to mimic the motion of flowers under variable natural wind conditions.

We find that when flowers are stationary, bees prefer conical-celled flowers unless the conical-celled flowers are harder to see. When flowers are moving, however, bees always visit conical-celled flowers more than flat-celled flowers, regardless of visual difficulty.

Materials and methods

Bee Care

Colonies of bumblebees (Bombus terrestris) were supplied by Syngenta Bioline (Weert, The Netherlands). Housing and lighting conditions as well as bee marking procedures were as described in Dyer et al. 2007, with bees fed 30% sucrose solution daily. We used clear plastic feeders so that bees were kept colour- and flower naïve prior to experiments. Colonies were attached via a plastic tube to a 300 × 750 × 1120 mm plywood flight arena with a UV-penetrable Plexiglas lid. The inside of the box was painted in a bee-neutral shade of green (Fig. 1b).

Bee Experimental Methods

In each experiment 10 flowers, five conical celled and five flat celled, were presented to individual bees in a semi-randomised array in the flight arena. Each flower was fitted with a 50-μL tube in the centre to allow refilling of the 20 μL 30% sucrose reward. During experiments all bees were excluded from the flight arena while flowers were placed, and then, individual marked bees were released into the arena and allowed to forage until satiation. Flowers were refilled with 20 μL 30% sucrose after each visit by a bee. We recorded each type of flower the bee landed on, and whether the bee then drank or aborted the landing without drinking. Each bee's foraging was recorded to 100 choices. After each foraging bout the visited flowers were replaced with fresh ones to minimise any interference from scent markings, and the flowers repositioned into a new semi-randomised array to prevent positional learning.

Bee Experimental Methods Using Orbital Shaker

To simulate the wind movement that a bee might encounter when foraging in the wild, we used a small laboratory orbital shaker (Stuart SSM1 Lab-Scale Orbital Shaker; Bibby Scientific Limited, Staffordshire, UK; see Fig. 2). Flower movement in the wild depends on stalk length, strength and orientation as well as wind conditions and flower weight. This produces flower movement that varies in both speed and direction. This experiment chose to test one simple form of motion, rotation around a fixed point. The shaker produces a circular pattern of movement on a flat platform, with an orbit of diameter 16 mm. The shaker was covered with a bee-neutral shade of green tissue paper similar in colour to the paint in the flight arena, and bees were trained to visit the shaker by placing their usual feeders on and around it. Bees were acclimatised to its noise and movement by running the shaker at its lowest setting of 30 rpm (0·50 Hz) for several days. The lowest setting allowed bees to acclimatise to the shaker while still remaining naive of the experimental conditions of the faster shaking, preventing interference in later experiments. Warren and James (2008) found that hoverfly pollinators had greatest difficulty in handling flowers moving with around 53–67 oscillations per min. Optimal shaking speeds for the experiment were determined by preliminarily testing the foraging ability of several bees across different speeds. Of 100 rpm was chosen as an appropriate upper limit for maximum landing difficulty, because at speeds >100 rpm (1·67 Hz), most bees were unable to land on the flowers. Of 70 rpm (1·17 Hz) was the highest speed at which all bees were able to land and so tested a moderate difficulty of landing, and the easiest conditions tested the bees with flowers stationary but still on the shaker to maximise the similarity to the rest of the experiment.

Figure 2.

Diagram of the shaker showing its green tissue paper camouflage and positioning of the flowers used in experiments.

Statistical Methods

Paired t-tests of bee choices were performed using Microsoft Excel to compare the overall preference for conical- or flat-celled flowers and to compare the first 10 choices with the last 10 choices in each test.

Plant Growth

Flowers were chosen from plants of the flat-celled phmyb1 mutant line. Plants arising from a germinal reversion to the wild type provided an isogenic conical-celled control, and for a darkly coloured conical-celled flower, the V26 line of Petunia was used, which has flowers with a close match in pigment colour but much darker appearance owing to the lower total amount of light reflected from the flowers (Fig. 1b). All lines were kindly supplied by Professor C. Martin, John Innes Centre.

Plants were grown from seed under greenhouse conditions at 23 °C in 4-inch pots in Levingtons (UK) M3 compost. During the growth period, plants received supplemental lighting from Osram 400 W high-pressure sodium lamps (Osram, München, Germany) on a 16-h light/8-h dark photoperiod.

Reflectance Spectra of Flowers

The spectral reflectance of Petunia petals was measured with an Ocean Optics S2000 spectrophotometer (Dundedin, FL, USA), relative to a white reflection standard. Spectrophotometer readings were averaged over three measurements at different positions on the flower.


Spectral Analysis of Petunia Flowers

The reflectance spectra for all three flowers across wavelengths visible to bees are shown in Fig. 1b. The reflectance spectra for flowers of the isogenic lines show a very similar relationship with those of conical- and flat-celled A. majus flowers (Dyer et al. 2007), in that while the position of the peaks of reflectance (and hence perceived colour) was very similar, more light was reflected overall from the flat-celled mutant flowers. Of the three lines, the V26 flowers reflected the lowest amount of light, owing to the increased pigment content. The percentage of light reflected by both the revertant and mutant flowers at their peak wavelengths are in the normal range for flowers, but the V26 flowers reflect <10% of overall incident light, which is extremely low for a flower (Chittka et al. 1994).

The three flowers are very close to each other in bee perceptual colour space, as shown by their relative positions on the colour hexagon in Fig. 1c. The colour difference from revertant to mutant flowers was 0·08 units, and V26 to mutant was 0·05 units. Bees find colours <0·1 units apart very difficult to distinguish (Dyer and Chittka 2004a,b).

Bumblebees Prefer Conical-Celled Petunia Flowers When Plants are Isogenic

To test bumblebee preference for Petunia flowers differing only in the presence or absence of conical cells, we used flowers of the flat-celled phmyb1 mutant line alongside flowers of a conical-celled isogenic revertant arising from this line. Bees presented with stationary arrays of flat-celled and conical-celled flowers in equal numbers chose conical-celled flowers 57% of the time (16 bees to 100 choices each, P < 0·001, Fig. 3).

Figure 3.

Overall bee preference for conical- and flat-celled flowers across different speeds. Preference is shown as percentage of total choices averaged across all bees tested in each experiment. Mutant and revertant flowers are of the same pigment colour, while V26 flowers are darker owing to an increase in the amount of pigment. Bees prefer conical-celled revertant flowers when presented alongside mutant flat-celled flowers regardless of flower movement. However, bees prefer mutant flowers against darker, conical-celled V26 flowers when the flowers are stationary, but when flowers are moving the bees prefer the conical-celled V26 flowers.

Bumblebees Prefer Flat-Celled Petunia Flowers When Conical-Celled Flowers are More Darkly Pigmented

To test bee preference for conical or flat cells when the conical-celled flowers were at a disadvantage visually, bees were tested with the darker and harder to see V26 line conical-celled flowers presented alongside the mutant flat-celled flowers. This time the bees showed a very different preference, favouring the flat-celled mutant flowers 65% of the time and the V26 only 35% while the flowers were stationary (P < 0·001, Fig. 3).

Bumblebees Prefer All Conical-Celled Flowers When Flowers are Moving

When the V26 and mutant flowers were moving at 70 rpm, initially bees showed a similar preference to when these flowers were stationary. However, in this moving array bee preference for conical-celled V26 flowers over mutant flat-celled flowers increased markedly, bringing the overall average percentage of drinks on V26 up to 56% (P = 0·026, Fig. 3). This preference emerged as a product of learning across the hundred choices each bee made. During the first 10 visits to flowers moving at 70 rpm, bees visited conical-celled V26 flowers only 43% of the time, a preference level similar to that observed for stationary flowers (Fig. 4b). As the bees grew more experienced, however, the proportion of visits to conical-celled V26 flowers rapidly increased. At 50 visits the average preference for V26 flowers was 52%, and by the end of their hundred choices the bees visited V26 flowers 62% of the time (P < 0·001). This shows that the initial primarily visual preference was overcome by a learned tactile advantage that the conical-celled flowers gave under the new, moving conditions. Additionally, each flower choice was made before the bee landed on the flower, showing that a visual cue was allowing them to quickly discriminate in favour of this tactile advantage.

Figure 4.

Learning curves showing percentage of bee preference for conical-celled flowers per 10 choices while flowers were moving (not shown: reciprocal preference for flat-celled flowers). (a) Revertant flowers. Bees show no change in preference over their hundred choices. (b) V26 flowers. Bees show a gradual increase in preference over their hundred choices. The average of the final 50 choices is higher than for their first 50 choices because the bees are learning to favour the conical-celled flowers. Error bars represent standard error.

When the speed of the shaker was increased to 100 rpm, only a small number of bees were able to complete the task. The preference in favour of the conical-celled flowers dropped to around 47%, although this is not statistically significantly different from the preference at 70 rpm (t = 0·746, P = 0·46).

In contrast to their increase in preference for V26 conical-celled flowers over the flat-celled flowers when the flowers were moving, bees presented with flowers of the isogenic lines on a moving platform did not show any significant increase or decrease in preference compared to the original stationary experiment (Fig. 3). Bees still chose the conical-celled flowers 57% of the time, and similar to the stationary experiment, this preference did not change over time (Fig. 4a).


The colours of the flowers used in these experiments are very similar to those of bees, and bees will usually only discriminate between such similar colours after differential conditioning (Dyer & Chittka 2004a). Under these conditions bees are capable of a maximum discrimination of around 70% of visits favouring the rewarding colour. We would therefore not expect to see a preference >70% for any flower type when bees are presented with such similar colours, even if foraging from one flower type greatly disadvantaged the bee.

Trials of isogenic mutant and revertant flowers showed that bees prefer conical-celled Petunia flowers when conical-celled flowers present no additional visual disadvantage (Fig. 3). This is as expected when compared to field trials of isogenic lines of Antirrhinum which showed bees to favour conical-celled Antirrhinum flowers over flat-celled mutants (Glover & Martin 1998). The preference for the conical-celled Petunia flowers was evident within the first 10 choices each bee made, and the preference did not change throughout the experiment (Fig. 4a). As conical- and flat-celled flowers with the same pigment colour are equally salient for bees (Dyer et al. 2007) and their preference is evident before any tactile benefits could have been learned, bee preference for conical-celled rather than for flat-celled flowers of isogenic lines suggests an innate preference based on visual cues. This preference did not change when flowers were moving, showing that the bees did not gain any additional advantage from the revertant flowers under conditions of increased handling difficulty (Fig. 4a).

This contrasts to the experiment with non-isogenic flowers, when the darker V26 flowers were only preferred by the bees when the flowers were moving. Bumblebee vision is not usually affected by brightness, but at extremes of high or low light bees become unable to distinguish colour. This is because when too much or too little light is incident on an ommatidium in a bee's eye, all three photoreceptors excite equally and the bee perceives the colour as closer to white or black (Kien & Menzel 1977; Backhaus 1991, 1992; Chittka 1992). V26 flowers are unusually strongly pigmented and reflect little light, so will be difficult for a bee to distinguish from the background. This is supported by the results of the stationary mutant vs. V26 preference experiment when bees preferred the mutant flowers that were easiest to see. This preference did not change as the bees grew more experienced throughout the 100 visits of the stationary trial, showing that the bees were not learning any advantage or disadvantage to change this initial visual preference.

In Silene maritima the length of a flower's stalk determines the flower's attractiveness to hoverfly pollinators, but flowers with the longest stalks experienced shorter pollinator visits and decreased pollination success (Warren & James 2008). This may be due to the increased difficulty in handling. In our experiments, when the V26 and mutant flowers were moving, bees initially discriminated against the V26 flowers but their preference increased over the course of the trial, ending at a preference for conical-celled V26 flowers at levels similar to those observed when isogenic flowers were presented. The change in preference we observe when the V26 and mutant flowers were moving shows that the bees’ preference for the conical-celled V26 flowers is a product of learning, in which the bees overcome their initial visual preference to take advantage of the tactile benefit associated with the darker flowers.

Additionally, the movement of flowers may increase their visibility. As motion parallax (a product of the bee's own movement) is so important to bee vision and target detection (Lehrer et al. 1988), additional motion could further contribute to the ease with which a bee can detect the flower and contribute to the flower's attractiveness. While the Petunia flowers in our experimental setup would be stationary with respect to the shaker platform, the bees would see movement against the body of the shaker and the walls of the flight arena. The movement could contribute to the bee's normal use of motion parallax and create greater distinction between the flower and the background, helping to remove some of the visual disadvantage of the V26 flowers.

As conical cells convey a fecundity benefit to flowers that are difficult to handle (such as Antirrhinum) by giving their pollinators a tactile benefit (Whitney et al. 2009), a similar tactile benefit under motion may explain why conical cells have persisted throughout evolution in flowers that are easy for pollinators to handle, where movement in the wind has produced additional difficulty. The metabolic costs of flight and foraging are large (Kammer & Heinrich 1974) and the less energy a bee has to expend foraging from each flower, the more sugar is left for the colony. As pollinators quickly learn the metabolic benefit in expending less energy ‘hanging on’ to conical-celled flowers, even a small deviation from equal preference would be enough to convey a large difference in fitness to conical-celled plants over evolutionary time.

These fitness benefits of conical cells for flowers that are difficult for pollinators to grip begin to explain the prevalence of this structure on the petal epidermal surface. Conical cells have a range of functions, each of which may be important in specific habitats or with different pollinators. The diversity of size and shape of conical cells supports the hypothesis that petal micromorphology can be optimised for both pollinator and habitat (Whitney et al. 2011a). Here, we have shown that there are interactions between the different properties of conical cells, in this case influencing flower–pollinator interactions through both visual and tactile effects. To establish the extent to which petal micromorphology is optimised, greater knowledge about the specific interactions between plants and their pollinators is required.


Conical cells have multifunctional properties, and we find that these properties interact to enhance pollination success. As bees prefer moving targets (Lehrer & Srinivasan 1992), and conical cells help bees to grip flowers as well as enhancing visually appealing floral properties such as colour, the presence of conical cells in conjunction with flowers that can move in the breeze might be the best way to attract pollinators. Conical cells alone are not particularly visually attractive and moving flowers alone are not easy to handle, but the combination of these two properties might act synergistically to maximise pollinator attraction and foraging efficiency.


We thank Matthew Dorling for excellent care of plants, Professor Cathie Martin for supplying Petunia seed, and Professor Lars Chittka for the use of the spectrophotometer and for helpful discussions. We thank Syngenta Bioline for supplying B. terrestris colonies. This work was funded by the Cambridge Overseas Trust and Cambridge Commonwealth Trust.