Honest signals to maintain a long‐lasting relationship: floral colour change prevents plant‐level avoidance by experienced pollinators
Summary
- Honest signals prevail when those signals manipulate receivers to benefit the senders more than deceptive signals do. Floral colour change (FCC), reported in hundreds of species, is a well‐known example of honest signalling in plant–pollinator interactions. It occurs in fully turgid flowers, and usually correlates with a cessation of reward production in individual flowers. This trait has been considered a plant strategy that enhances distant pollinator attraction by extended displays, while minimizing visits to non‐reproductive flowers by honest colour signals.
- Here, we propose an additional, overlooked benefit of FCC, which emerges when we consider the spatial learning ability of pollinators to avoid unprofitable plants. If a plant retains rewardless flowers without FCC, it is difficult for pollinators to visually locate the rewarding flowers. Although the enhanced display initially attracts more pollinators, its low profitability for foraging may cause plant‐level avoidance by them. The avoidance resulting from rewardless flower retention may be prevented by FCC because it helps pollinators to find rewarding flowers. To test this hypothesis, we observed the behavioural changes of bumblebees foraging in an array of artificial plants.
- We found that rewardless flowers without FCC could initially attract bees by increasing the plant's display size, but their lack of reward resulted in plant‐level avoidance by those bees with spatial memories. The FCC in rewardless flowers, in contrast, encouraged bees to return by helping them to find rewards on plants. Consequently, honest plants with FCC received more visits than those dishonest plants that did not display colour change.
- Floral colour change thus can prevent plant‐level avoidance by pollinators that use spatial memory when choosing plants. The spatial learning ability of pollinators may, therefore, be one of the keys to understanding why both colour‐changing and non‐colour‐changing plant species occur among angiosperms.
A lay summary is available for this article.
Introduction
Honest signals are thought to prevail when they control receivers in a more beneficial way for senders than deceptive signals do (Davies, Krebs & West 2012). One of the keys to understanding their evolution is the cognitive ability of receivers to detect deception. That is, if receivers are able to detect deception and, thereby punish cheaters, deceptive signals will be purged, while honest signals dominate sender–receiver interactions. Here, we apply this view to a well‐known example of honest signals in plant–pollinator interactions.
Floral colour change (FCC) has been reported for hundreds of species of plants across many taxa (Weiss 1995; Weiss & Lamont 1997). It occurs in fully turgid flowers in various combinations of pre‐ and post‐change colours, and usually correlates well with a loss of reproductive potential and a cessation of pollen and nectar production by individual flowers, i.e. flowers that have changed colour are often rewardless and have completed their reproductive tasks (Gori 1989; Niesenbaum, Patselas & Weiner 1999; Ida & Kudo 2003; Yan et al. 2016). To explain why those plants honestly advertise the location of unrewarding flowers to pollinators, previous studies have demonstrated that (i) the retention of old flowers enhances pollinator attraction from a distance (Fig. S1a, Supporting Information), and (ii) at close range, honest colour signals guide pollinators to rewarding flowers, thereby improving the chance of young flowers being pollinated and preventing pollen deposition at old flowers (Fig. S1b) (Gori 1989; Weiss 1991; Oberrath & Böhning‐Gaese 1999). It should be noted that the enhanced attraction can be achieved without FCC (Fig. S1a), but the retention of old flowers without FCC increases risk of pollen loss and decreases the likelihood of young flowers being pollinated. These problems caused by retaining old flowers can be solved by FCC.
Another different, previously overlooked problem resulting from flower retention was recently highlighted in reports on pollinator cognition. Several studies have suggested that pollinators, such as bumblebees and hummingbirds, learn to avoid feeding locations with few rewards (Hurly 1996; Cartar 2004; Burns & Thomson 2006; Makino & Sakai 2007). This was usually seen in pollinators with hours of experience, and repeatedly foraging in the same, small areas or territories. Makino & Sakai (2007), for example, allowed bumblebees to forage for 8 h in an area with rewarding and rewardless plants, and found, at first, that the bees visited both types of plants at similar rates. Gradually, however, they learned to avoid the locations of rewardless plants (Fig. S1c). Such pollinators may also avoid plants that retain rewardless flowers without FCC because it is difficult for them to locate rewarding flowers on those plants (Fig. S1d). As a result, flower retention without FCC may result in insufficient pollination. This possible punishment for dishonest plants, in the form of plant‐level avoidance, has not previously been considered in the context of FCC plant–pollinator interactions.
In addition to plant‐level avoidance, we hypothesized that FCC prevents pollinators from avoiding plants that retain rewardless flowers because FCC facilitates pollinator foraging by highlighting which flowers are rewarding (Fig. S1d). As a result, an honest plant that retains rewardless flowers with FCC (hereafter known as Honest plants) is expected to not only attract more naive pollinators by its extended display, but also receive frequent visits by experienced pollinators due to the honest signals (Fig. S1e). Conversely, a dishonest plant that retains rewardless flowers without FCC (hereafter, Dishonest plants) can attract more naive visitors, but may be avoided by experienced visitors. Plant‐level avoidance can be averted if a plant does not retain rewardless flowers, thus facilitating pollinator foraging (Fig. S1d). A plant without flower retention (hereafter, Mini plants, which are characterized to smaller floral displays) will, however, receive fewer visits by naive pollinators because of its small display. It was thus hypothesized that, due to the adaptation to both naive and experienced pollinators, Honest type – plants retaining old flowers coupled with FCC – were likely to achieve the most visits among the three plant types.
To test our hypotheses, and demonstrate the additional benefit conferred by FCC, we observed the behavioural changes of bumblebees foraging in an array of artificial plants in a laboratory (Fig. 1). The array consisted of all three plant types (Honest, Dishonest and Mini), in which each bee made eight foraging trips. We then counted the number of visits to each plant during each trip. Specifically, we tested whether or not (i) a bee made more visits to Honest and Dishonest plants than Mini plants in the earlier foraging trips; (ii) the bee learned to avoid Dishonest plants while preferentially visiting Honest and Mini plants; and (iii) Honest plants drew the most visits from a single bee over the long term. We also tested whether or not (iv) FCC facilitated pollinator foraging; (v) bees used spatial memory; and (vi) the results were consistent irrespective of the combination of pre‐ and post‐change flower colours. It is difficult to determine and control, in field conditions, prior experience of pollinator individuals. In addition, it is technically difficult to create Dishonest plants from real plants. That is, it is not easy to terminate the reward production of individual flowers. The artificial nature of our setups, which solved these problems, was an appropriate as the first step to test of our hypotheses and determine whether or not they are worth testing under field conditions.
Materials and methods
We used, in total, four artificially reared colonies of bumblebees, Bombus ignitus, provided by the Firefly Breeding Institute of Itabashi‐Ward, Tokyo, Japan. In a laboratory, a colony was connected to a flight cage with a gated tunnel. The cage was 5·5 m long × 4·0 m wide × 2·0 m high and was made of transparent vinyl film (Fig. S2). The floor was covered with a green‐coloured mat, on which we placed eight cardboard landmarks along the edges of the cage (Fig. S2). The landmarks differed in size, shape and colour. They comprised, for example, of a 0·3‐m‐high T‐shaped light‐blue column, a 0·4‐m‐high pink rectangular prism and 0·2‐m‐high blue pyramid. We also set out four flat 0·4‐m‐long cardboard landmarks (Fig. S1). Two video cameras (HDR‐SR8; Sony, Tokyo, Japan) set on tripods were used to record the bees in action, and also served as landmarks. The floor was illuminated by 22 fluorescent and six black‐light tubes (FL20SBLB‐A and FL20SS EX‐D18‐Z; Toshiba, Tokyo, Japan) set in 14 fluorescent light fixtures, included inverters (LXE‐14000; Maruzendenki Co., Ltd., Osaka, Japan) mounted on the ceiling. The flicker frequency of these lights was too high for bees to perceive the flicker. The illumination was not an exact match to sunlight, but more closely resembled it than ordinary fluorescent lighting.
Each plant was created by arranging artificial flowers on a square plate (Fig. 1b). An individual flower was constructed using a 1·5 mL micro tube, which contained a feeder that was gradually refilled with nectar by capillary action [see Makino & Sakai (2007) for details]. All three plant types had two rewarding flowers (20% w/w sucrose). Both Dishonest and Honest plants had an additional 10 non‐rewarding flowers (plain water). Three colours [yellow (Y), purple (P) and white (W); Fig. S3] were used to make six colour pairs: WP, PW, WY, YW, YP and PY. The first letter represents the pre‐change colour, while the second was for post‐change colour. The colours used are distinctive for bees, as indicated by the colour‐discriminating behaviour of bees (see results), and also by colour distance evaluated in bee colour space (Chittka & Kevan 2005). Colour distance between each pair of three colours ranged from 0·11 to 0·28, which was above the criteria reported by Dyer (2006). The positions of the rewarding flowers were randomly assigned to each plate based on random numbers generated by a personal computer.
Training and testing
Bees were trained to collect 10% w/w sucrose from a plant at the centre of the experimental array (Fig. 1c). This plant had three red flowers on a square plate. Each flower was connected to a vial behind the plate, with cotton inside it that quickly transferred nectar to the vial, allowing bees to freely drink it. The high flow rates of nectar meant that bees did not need to move from flower to flower to feed during the training period. The flowers were painted opaque red. The bees had no previous experience with other floral colours prior to the tests. We glued numbered tags to the thoraxes of those bees that learned the flowers.
Just before testing, we allowed the bees to collect nectar from capillary feeders by setting two pre‐test plants at the training site. Each plant consisted of six red flowers with 20% w/w sucrose solution, and six red flowers with water. This procedure was intended to get the bees accustomed to slow‐filling capillary feeders. To collect nectar efficiently, bees had to move from flower to flower. A focal bee was selected from those making regular foraging trips.
Once a focal bee completed more than three foraging trips, we replaced the pre‐test plants with 24 new plants in a regular grid, with eight plants per plant type (Fig. 1c). Their positions in the grid were randomly assigned, but in such a manner that the central four plants included all three types. This was intended to reduce the effect of the bees’ initial tendency to search for plants around the centre of the array.
Each target bee was allowed to make at least eight foraging trips (approximately 3–4 h). After the last trip, we interchanged the locations of Dishonest and Honest plants, and let the bee make an additional trip to confirm the use of spatial memory. It was hypothesized that if the bee chose Honest plants according to their spatial position, it would return to Dishonest plants more often than Honest plants after the exchange.
An observer recorded the sequence of plants visited, and then counted the number of visits to each plant. Two video cameras (HDR‐SR8; Sony) also recorded the bee's flower‐to‐flower movements to count the number of rewarding (or non‐rewarding) flowers probed on each plant during the 1st, 3rd, 5th and 7th trips. The 2nd, 4th, 6th and 8th trips were skipped to save time and effort. The target bee was then removed and not used again. The positions of the plants were regenerated for each new target bee.
Twenty‐seven bees were used in total. The number of bees for each colour pair was 5, 5, 3, 5, 5 and 4, for WP, PW, WY, YW, YP and PY respectively. Before the plant exchange treatment, we allowed 12 bees to complete 16 trips, while the remaining 15 bees made eight trips. We decreased the number of trips from 16 to eight because the time‐dependent changes in plant choice by bees became clear before the eighth trip. The plant exchange treatment was not conducted for one bee, which stopped foraging during its ninth trip.
Analysis
To test whether the performance of pollinator attraction differed among plant types for each colour pair, a generalized linear mixed model (GLMM) with a logarithmic link function and a Poisson error distribution was applied to the number of visits to a plant. Plant type and individual bee were treated as fixed and random factors respectively. The Wald chi‐square test was used to assess the significance of plant type. Tukey's test was used for a posteriori comparisons. This analysis was conducted for each trip, and also for the total number of visits in the first eight trips. To test whether foraging performance was lower on Dishonest plants, the same analysis was applied to the numbers of rewarding flowers probed in a single visit to a plant. We treated the total number of probes on a plant as a dependent variable and the log of the number of visits to the plant as an offset term. We also applied the same analysis to the number of non‐rewarding flowers probed in a single visit to a plant, to examine whether FCC helped bees to avoid non‐rewarding flowers. We used r (ver. 3.2.4) with function glmer in package lme4 (ver. 1.1–11), Anova in car (ver. 2.1–2) and glht in multcomp (ver. 1.4–1).
Results
In total, Honest plants received significantly more visits than Dishonest plants for every colour pair (Fig. 2a, Table S1 and S2). Furthermore, in WP, WY and YP, Honest plants significantly surpassed Mini plants as the most visited type. At the first foraging trip in these colour pairs, Honest plants were significantly preferred over Mini and Dishonest plants. Visitation to Dishonest plants decreased in all pairs.
During single visits to plants, bees probed significantly more rewarding flowers, and fewer unrewarding flowers, on Honest plants compared to Dishonest plants (Fig. 2b).
When the locations of Dishonest and Honest plants were exchanged, experienced bees returned to previously honest locations significantly more, resulting in preferential visits to Dishonest over Honest plants (Fig. 2c).
Discussion
Although the composition of rewarding and unrewarding flowers was identical to both Dishonest and Honest plants, bees made fewer visits to Dishonest plants during later foraging trips (Fig. 2). This was for every colour pair. The different performance in pollinator attraction suggests that honest colour signals prevent experienced bees from avoiding plants that retained unrewarding flowers (Fig. S1d). As a result of pollinator avoidance, Dishonest plants became the least visited type of all (see “total” in Fig. 2a), even when Dishonest plants were the most visited type during the first foraging trip (YP, PW, YW and PY). The retention of unrewarding flowers without FCC may thus attract many naive visitors, but is not a good strategy for plants to establish a long‐lasting relationship with those visitors.
In contrast to the lowest total performance of Dishonest plants, Honest plants were the most visited type in total for WP, WY and YP colours (Fig. 2a). For those colour pairs, Honest plants were more attractive from the beginning than Mini plants, and kept receiving frequent visits. Although the advertising effect was ephemeral, as evidenced by the similar visitation frequencies to Mini and Honest plants at the end, the initial advantage was enough to enable Honest plants to draw more visits overall from a single bee than Mini plants with WP, WY and YP colour combinations. Thus, with appropriate colour combinations, the retention of rewardless flowers, coupled with FCC, can be the best strategy for plants to maximize pollinator visitation.
Floral colour change prevented plant‐level avoidance from occurring, probably because it helped bees probe more rewarding flowers and fewer empty flowers (Fig. 2b), thereby facilitating their foraging (Fig. S1e). In contrast, encounters with unrewarding flowers on Dishonest plants may explain the avoidance of those plants by experienced bees.
Persistent returns to previously honest locations after exchange treatments (Fig. 2c, right) indicated that the experienced bees decided on which plants to revisit based on spatial memory, instead of plant appearance or coloration. Since display size and the proportion of old flowers could vary greatly among plants in field conditions, it is reasonable for a forager to learn the individual locations of honest plants.
Our data also indicate that there is a cost associated with FCC. The higher number of probes to rewarding, sexually viable flowers of Honest plants, when compared to Dishonest plants (Fig. 2b), suggests there is more geitonogamous pollen transfer (de Jong, Waser & Klinkhamer 1993) occurring within a colour‐changing plant than within a non‐changing plant. Although FCC has been assumed to save pollen loss in old flowers, and improve the opportunity for young flowers to be pollinated (Gori 1989; Weiss 1991; Oberrath & Böhning‐Gaese 1999), such benefits may be diluted by increased levels of geitonogamy.
One possible strategy by which FCC plants could avoid geitonogamy is to reduce the number of young flowers open at a given time (e.g. from 2 to 1 in our case). In addition, increasing the number of post‐change flowers on a plant might be effective in reducing geitonogamy, as indicated by some field studies (Ida & Kudo 2003; Ollerton, Grace & Smith 2007). Ida & Kudo (2003) reported that the number of probes to pre‐change flowers during a single visit to a plant (i.e. geitonogamous movements of pollinators) decreased with the number of post‐change flowers retained on a Weigela middendorffiana plant. The proportion of pre‐change flowers among the total number of flowers on a plant varies greatly among species [e.g. it varies from 6 to 56% among eight shrub species (Lamont 1985)], which might reflect interspecific variations in self‐compatibility, or the degree of inbreeding depression. The total display size of whole plants might also affect the likelihood of geitonogamy in FCC plants.
The fact that Honest plants never surpassed Mini plants in terms of bumblebee attraction for some colour pairs (Fig. 2a) indicates that an inappropriate colour combination may render flower retention a waste of resources. In this case, Mini plants (no flower retention) is the best strategy for plants to maximize pollinator visitation with minimum resources. To make flower retention more effective, a flower has to change into an appropriate colour that matches pollinator preferences. The suitability of a post‐change colour depends not only on the colour itself, but also on its combined attractiveness with the pre‐change colour. Although post‐change yellow was effective when paired with pre‐change white (WY), the same colour was ineffective with pre‐change purple (PY), indicating that the interactions among colours and pollinators are complex. It remains unclear whether or not the appropriate colour pairs in this study (WP, WY and YP) are frequent in FCC species commonly visited by bumblebees. Bumblebees are known to have innate preferences for colours with higher spectral purity (Lunau 1990), and the rewarding colours of six bee‐pollinated FCC species match such preferences better than the alternative colours (Lunau 1996). Although this finding is not consistent with the successful colour pairs in this study (the spectral purity is higher in the unrewarding colour in WP, WY and YP), pollinator preference seems to be a key to understanding the colour pairs exhibited by FCC species. We should also keep in mind that prior experience with other flower colours can affect subsequent colour choice by pollinators (Rhode, Papiorek & Lunau 2013), which makes the situation more complex. The actual diversity of paired colours in colour‐changing species (Weiss & Lamont 1997) may reflect such interactions with different pollinators that have different colour preferences (Lunau & Maier 1995; Weiss 1997; Gumbert 2000).
It should be noted that there are two types of FCC in the angiosperms (Weiss 1995): FCC in whole parts (whole change, with which we have dealt here), and FCC in localized parts such as nectar guides or nectaries (partial change). The restriction on colour combinations found in this study may be relieved in species exhibiting partial change because the non‐changing parts could retain the appeal to distant pollinators. This may explain the dominance of partial change over whole change in FCC species (Weiss 1995; Weiss & Lamont 1997).
Our data may explain a discrepancy in previous studies, some of which reported decreased pollinator visits after picking off post‐change flowers (Cruzan, Neal & Willson 1988; Gori 1989; Weiss 1991; Oberrath & Böhning‐Gaese 1999), while others detected no decrease at all (Casper & La Pine 1984; Delph & Lively 1989; Ida & Kudo 2003). The latter reports are consistent with our observation that experienced bees visited Honest and Mini plants at similar frequencies (Fig. 2a), suggesting perhaps that those researchers focused on plants frequented by experienced visitors. In other words, there may have been too few naive pollinators for the advertising effect to be detected. Before any conclusions about an increased attractiveness not always being the primary force driving the evolution of flower retention in colour‐changing species can be drawn, a careful re‐examination of these systems, paying attention to visitors’ experience, is needed (Thomson & Chittka 2001).
Other than FCC, there may be certain alternative or complementary ways whereby plants send honest signals to pollinators. For example, structural changes in old flowers, such as the relatively closed corollas of Gentiana triflora, which guide pollinators to young unpollinated flowers (Fusato et al. 2015), can serve as honest signals. The position of flowers in an inflorescence with sequential flowering (e.g. a vertical inflorescence with acropetally flowering) can also be a reliable indicator of rewarding flowers. Indeed, in an experimental study using artificial flowers (Makino & Sakai 2007), bumblebees showed no avoidance of plants bearing both rewarding and unrewarding flowers at predictable positions. Changes in floral scents (Schiestl & Ayasse 2001; Yan et al. 2016) could also replace or complement FCC. Applying the concept of honest signals to plant–pollinator interactions may improve our understanding of various floral traits.
The degree of plant‐level avoidance might be minimized by the use of olfactory footprints by bees. When visiting Dishonest plants, our bees were often observed to reject unrewarding flowers before landing them, suggesting the presence of some remotely detected cues for selecting rewarding flowers. At a single visit to a Dishonest plant, bees found one rewarding flower on average (Fig. 2b). If the bees chose flowers at random, they would on average have to probe five unrewarding flowers in order to find one rewarding flower. However, the number of unrewarding flowers probed was approximately one, which was far fewer than expected (Fig. 2b). Under laboratory conditions, bumblebees are found to utilize their olfactory footprints that they deposited on rewarding flowers (Cameron 1981; Schmitt & Bertsch 1990; Schmitt, Lübke & Francke 1991). Our bees were also likely to use such scent marks to efficiently find rewards on Dishonest plants, which might minimize the degree of plant‐level avoidance. However, it remains controversial as to whether bees use scent marks as attractants under field conditions (Witjes & Eltz 2007). If they do not use scent marks and have more difficulty in locating rewards, plant‐level avoidance in the field might be more intense than we found in this study.
Conclusion
By revealing experience‐dependent changes in plant choice by captive bees foraging on artificial plants, we demonstrated that FCC can function as an honest signal to prevent plant‐level avoidance by pollinators with spatial memory. Flower colour change thereby enables plants with old flower retention to attract both inexperienced and experienced visitors (Fig. S1d). Note that our data support the benefits of FCC proposed in the previous studies (saving pollen loss at old flowers, and increasing the likelihood of young flowers to be pollinated; e.g. Gori 1989; Weiss 1991; Oberrath & Böhning‐Gaese 1999; Fig. S1b and Fig. 2b), suggesting that the newly discovered benefit (preventing pollinator avoidance; Fig. S1d) can function together with those benefits to improve plant reproduction. The contribution of each of those benefits (and also the costs of FCC) is likely to vary depending on a number of factors. The presence of closely related pairs of colour‐changing and non‐colour‐changing species (e.g. Weigela coraeensis and Weigela hortensis; Suzuki & Ohashi 2014) may reflect the different outcomes of such cost–benefit trade‐offs associated with FCC. Flowers of both species secrete nectar for an initial 2 to 3 days and persist for a few more days, but only W. coraeensis exhibits FCC (Suzuki & Ohashi 2014), providing a suitable system for comparative studies. In such studies, whether pollinating partners possess spatial learning ability is definitely one the factors affecting the cost–benefit trade‐offs associated with FCC, and may be an important key to understanding why both colour‐changing and non‐colour‐changing plant species occur among the angiosperms (Weiss & Lamont 1997; Suzuki & Ohashi 2014). Indeed, Suzuki & Ohashi (2014) reported that a colour‐changing species (W. coraeensis) was more preferred by bees. In addition, the relevance of memory‐based spatial foraging is also indicated by a phylogenetic analysis revealing a possible evolutionary association between FCC and bee pollination (Ohashi, Makino & Arikawa 2015). Additional studies are required to confirm whether or not FCC prevents plant‐level avoidance in field conditions.
Acknowledgements
We thank J. Ollerton, K. Lunau, J.D. Thomson, J. Yokoyama and S. Sasaki for comments, and N. Abe and T. Ayabe for providing bee colonies. This work was supported by Grant‐in‐aid for JSPS fellow (18.3737) to TTM; KAKENHI to KO (JP19770011) and TTM (JP26840139). The experiments comply with current Japanese law.
Data accessibility
All data are uploaded as supporting information.
