Basic artificial flower construction
The artificial flowers were constructed from upturned petri dish bases (100 mm × 20 mm, Sarstedt, Nϋmbrecht, Germany) covered with aqua blue sticky back plastic (d-c-fix® adhesive film, Hornschuch group, Weissbach, Germany). The thermal emissivity of the plastic was measured, using electrical tape as a standard emissivity reference. The emissivity of the plastic was 0.95 (the same emissivity as the electrical tape), and this value was used for all measurements conducted, following the recommendations given by Harrap et al. (2018). Feeding wells were made from upturned Eppendorf tube lids (Multiply-pro cup 0.2 mlPP, Sarstedt, Germany) stuck on a sheet of thermally insulating 1 mm thick white plastic foam. Three feeding wells were stuck on each flower at 5 mm from the edge, orientated so that the lids’ cap pointed outwards at 120° angles from each other (Fig. 1a).
Each flower top was supported on a 42 mm tall, 85 mm diameter card cylinder wrapped in black electrical tape. Three 3 × 1 cm card rectangles were stuck at 120° angles to the top of the lid of the petri dish, which was used as the underside of the bottom of the artificial flower (Fig. 1), such that 2 cm tabs extended from the dish’s edge. A circular red sticker with a two-digit number written in black ink was stuck to each of the card tabs. Numbering allowed the researcher to identify rewarding and unrewarding feeders without giving meaningful cues to the bees, and followed the protocol described by Harrap et al. (2017).
The bulbs of 3 ml plastic pipettes (Pastettes, Alphalaboratories, UK) were cut down to create a 16 mm plastic hood-shaped tunnel, as used by Pearce et al. (2017) and Lawson et al. (2017a). These bulbs were placed over the feeding wells so that the open end of the tunnel faced into the flower centre (Fig. 1), limiting the bees to only approach the well from the centre of the flower. These were taped down at the beginning of testing each day with a fresh section of clear tape (Scotch Easy Tear, St. Pauls, USA).
Artificial flower patterns
Colour or temperature patterns were added to the basic construction of the experimental artificial flowers described above. Two types of temperature patterns and two types of visual patterns were used in this experiment.
Colour patterns were created by sticking down a 1 × 2 cm adhesive d-c-fix® plastic panel in front of feeder tunnels. Two types of contrasting visual guide panels were used: ‘Blue panels’, which were a darker shade of blue than the flower top (Fig. 1b), and pale pink ‘Pink panels’ (Fig. 1c). To control for tactile effects caused by these coloured portions being stuck on the flower and therefore being slightly raised with a possibly tactile edge, we added ‘Control panels’ in front of feeder tunnels that did not have a blue or pink panel stuck in front of them. These control panels were made of the same aqua blue plastic as the flower, and therefore presented a similar tactile effect to the coloured panels (controlling for the bees being able to feel the raised edges of the plastic when standing on the flower), but without any contrasting colour to the background. An additional control was also created where all three tunnels had the control panel in front of them (Fig. 1d). The reflectance spectra of the coloured panels are described in Appendix S1 (Supplementary Material).
Temperature patterns corresponding with reward location were generated using heating elements placed on the underside of the flower, following a similar design to Harrap et al. (2017, 2019). Two kinds of temperature patterns were used, identified here as ‘Hot’ or ‘Warm’ (Fig. 1e–f). Construction is described in Appendix S2 (Supplementary Material). These two patterns differed in the level of temperature generated, however both patterns show within-flower temperature contrasts comparable to natural flowers (Harrap et al. 2017). In ‘Hot’ artificial flowers, surface temperature settled at approximately 30 °C in front of the ‘Hot’ feeder and 26 °C in front of the cooler ones, giving a temperature difference of 5–6 °C across the flower. In ‘Warm’ artificial flowers, surface temperature settled at approximately 28 °C in front of the ‘Warm’ feeder and 24 °C in front of the cooler ones, giving a temperature difference of 4–5 °C across the flower.
Using combinations of the guides described above, eight different variants of experimental artificial flowers were constructed, as described in Table 1. These included two controls, two unimodal colour guides, two unimodal temperature guides, and two multimodal patterns with both a temperature and colour guide present.
Bee lab experimental techniques were used to investigate bumblebee guide responses and handling of artificial flowers. Flower naïve bumblebees, Bombus terrestris audax, were supplied by Biobest (Westerlo, Belgium) via Agralan (Swindon, UK). General husbandry, marking procedures and flight arena design is described by Lawson et al (2017a, b).
Outside of the testing period, bees were fed sucrose solution daily from PCR racks, gravity feeders and a selection of ‘generic’ artificial flowers placed within their flight arena, to ensure that bees learnt to handle feeding wells on arbitrary artificial flowers. Most generic artificial flowers were constructed from a 44 mm wide specimen jar (Sterilin PS 60 ml, with white plastic lids, Thermo Fisher Scientific, Newport UK), or resin disks of a similar size, with single feeding wells stuck to them. At least a week prior to bee trials, some of these generic artificial flowers were substituted with flowers to prepare bees for this experiment. These flowers showed one or two of the following: a different size (being either made from a larger specimen jar or a petri dish); multiple feeders; feeders not in the centre of the flower; or tunnels over the feeder (constructed as described above). These ‘new’ generic artificial flowers allowed bees to get used to feeding from flowers showing aspects of those used in this experiment. However, none of these new generic artificial flowers showed all these aspects together. Furthermore, these new generic artificial flowers never showed feeders at fixed angles about the flower edge, and never showed any visual or temperature patterns, or any colours associated with visual patterns. Additionally, other generic artificial flowers were still present and made up the majority (four out of seven or eight) of artificial flowers presented outside of trials.
Preliminary trials found that the experimental artificial flowers were too complicated for bees to learn to use in a single visit. Naïve bumblebees did not land on the experimental flowers described above, and so a pre-training phase with simplified flowers ‘pre-training artificial flowers’ was therefore included before the training period and test itself. A pre-training phase allowed bees to learn how to feed from artificial flowers similar to experimental flowers without gaining direct experience on the experimental flower or patterned signals. The top of these pre-training flowers was a petri dish lid covered with the same aqua blue sticky back plastic as the test flowers. Three feeding wells with foam bottoms were stuck to the top of the artificial flower, as in the test flowers (described above), but the sides of the Eppendorf tube lids were painted black. This lid was then supported on a 55 mm tall card cylinder, wrapped in black electrical tape and taped to the outside of the petri dish lid. Pre-training flowers did not possess tunnels, visual or temperature patterns, a base, or feeder labels.
During pre-training, marked forager bees were released into a foraging arena containing a clean pair of pre-training flowers with a droplet of 30% sucrose solutions in all their feeders. These pre-training flowers were placed in the centre of the foraging arena about 30 cm apart from each other in line with the bee’s entrance to the arena. A camcorder (Legria HF r36; Canon, Tokyo, Japan) was placed above each of the pre-training flowers. Each camcorder had a wide-angle lens attachment (XIT pro series 0.43X HD wide-angle lens 52 mm, Xit Group, Brooklyn, USA), placed facing down to view the artificial flower. Though pre-training was not recorded, the camcorder was present in pre-training so bees acclimated to it.
Multiple bees could be released into the flight arena together during pre-training. Bees were allowed to feed freely and return to the nest at will, with feeders being refilled when empty. If a bee completed two foraging bouts feeding on the pre-training flowers (departing and returning to the nest being one bout), it was deemed to have completed the pre-training phase. On a given sampling day, bees that had completed pre-training that day could then be used in the test phase. Most bees (69%) that began the pre-training phase went on to complete it. The bees that failed pre-training will include forgers unable to manipulate these artificial flowers, but also erroneously marked non-forager bees. If a bee completed pre-training but was not used for the test phase that day (for example because other bees took too long to complete testing), it could be used another day but would need to recomplete the pre-training phase. If a bee began the test phase it had to complete it in a single day. Bees that began the test phase were never reused in this experiment, even if they did not complete the test phase.
Following completion of the pre-training phase, bees were assigned artificial flower variants as described in Table 1. Individual bees were presented one variant of experimental artificial flower throughout the test phase. Consequently, the variant of artificial flower that a bee was presented with also describes its experimental test group. Throughout the experiment, temperature patterns were monitored by a FLIR e60bx thermal camera (FLIR systems, Inc., Wilsonville, USA). Before the bee began a foraging bout, the rewarding and unrewarding feeders (described in Table 1) of 8 experimental flowers were filled with 25 µl of 30% (volume-to-volume) sucrose solution or water respectively, using an electronic pipette (HandyStep® Electronic, Wartheim, Germany).
During the test phase, bees that had completed the pre-training phase that day were allowed to make successive foraging bouts on their assigned test flowers. Each test bee foraged alone in the arena during the test phase, and other bees would not be released into the arena during testing. When a bee began a foraging bout, a single artificial flower of the variant assigned to that bee was present in the arena. On the first foraging bout, this first artificial flower was placed in the same position as one of the pre-training flowers had been. The bee was allowed to land and forage on the artificial flower. Once a bee had extended its proboscis into any of the test flower’s feeding wells (recorded as a ‘drinking event’), a fresh artificial flower was placed in the arena at least 15 cm away from the bee. If the bee had drunk from any of the feeding wells on a flower, that flower was removed from the arena after the bee had departed. Bees were acclimated to the removal and insertion of artificial flowers into the arena, as this was done outside to testing for feeding. Thus, it is unlikely they were disturbed by flower removal and placement.
A ‘visit’ began when bees first made physical contact with artificial flowers. As artificial flowers were quite large, bees often flew from one part of the flower to another when searching. Thus, classing departure from the flower simply as the moment a bee broke contact with a flower after landing would not be representative of the bee’s searching effort and would often result in many aborted landings occurring before the first feeding. For this reason, a bee was classed as ‘departing’ if it broke contact with any part of the flower, then either flew away from the flower and did not return within 5 s, or flew over 30 cm away from the flower, or landed on another. These criteria allowed for bees to fly from one part of the flower or hover about the flower after landing without being classed as departing when they were still apparently searching the flower. Additionally, these criteria allowed for bees to climb about and search the lower parts and sides of the flower without being classed as departing. A ‘visit’ was assumed to end when the bee met one of these departure criteria.
Upon the bee’s departure, the flower that had just been drunk from was immediately removed in order to avoid the bee becoming satiated or distracted. This exchange of flowers (placing a new one inside the arena once the bee fed and removal upon the bee’s departure) continued until the bee returned to the nestbox on its own accord or had fed from all eight experimental artificial flowers in a single foraging bout. Once a bee had departed from the eighth flower in a bout, the eighth flower was removed and no more flowers were placed in the arena in that bout. The bee then returned (or was returned) to the nestbox.
Artificial flowers were reused in subsequent foraging bouts. Once a bee had completed a foraging bout, all flowers were removed from the flight arena. Any water and sucrose solution left in the feeding wells of flowers visited in the previous bout were emptied using paper towel. Flowers were then wiped down with ethanol, removing scent marks that may conflate bee decisions (Stout and Goulson 2001; Pearce et al. 2017). The feeding wells of these flowers were then refilled. The cycle of removal and replenishing of water and sucrose between bouts reduced the chance of differences in the temperature of the feeding well contents developing. Flower temperature was checked with the thermal camera and any flower that had overheated or ceased to produce a temperature pattern was replaced. After cleaning, thermal signals were allowed to re-settle before re-use, as ethanol evaporation cools flowers.
Before the bee was let back into the arena an artificial flower was placed inside. Bee foraging was then allowed to continue as described above. The first artificial flower placed in the arena in bouts after the initial foraging bout were placed anywhere in the arena, rather than the same positions of pre-training flowers. Individual bees were allowed to carry out successive bouts of foraging until the bee completed the bout where the number of flowers visited across all bouts was at least 30. At this point the bee was deemed to have completed the test phase.
Occasionally bees were reluctant to visit the flower in the arena. To encourage the bee, another artificial flower would be placed into the arena. In this scenario, if a bee drank from and departed from either flower, that flower would be removed but a new flower would not be placed inside the arena, as one was already present. Otherwise the experiment carried on as normal. If there were already two artificial flowers within the arena and bees still seemed reluctant to visit an artificial flower, an artificial flower would be moved to a new position. Moving flowers would not be carried out if the bee had already visited a flower but not fed. At any one time there were never more than two artificial flowers in the arena.
Video cameras were used to record bee flower visits and flower handling in the test phase. Whenever an artificial flower was placed in the arena, either at the start or during a foraging bout, a camera that was not already viewing a flower would be moved into position above it. This was done immediately after a flower was placed in the arena. Viewing the artificial flowers from above meant that the entire flower top and at least two of the numbered tags at the bottom of the flower were visible.
96 bees from 14 nestboxes completed the test phase, with 12 in each of the 8 test groups (Table 1). Bees completed the test phase in 31.59 ± 0.01 visits (mean ± SEM). However, due to camera recording error the visits after focal visit 29 for bee 5 (in Plain Control group) and 84 (in Multimodal Pink group) were not recorded, similarly visit 30 (of 33) of bee 81 (in Multimodal Pink group) was not recorded due to an error in camera placement. This meant the number of filmed visits for each bee ranged from 29 to 35. The test phases of all bees were carried out between 10:30 and 18:30, the normal time period of bumblebee foraging activity.
Data on flower handling was collected by reviewing the recordings of a bee’s behaviour during visits to experimental artificial flowers. Flower handling data was only collected with reference to a ‘focal flower visit’, defined as a flower visit during which a bee first drinks from any feeding well on the flower, or the last visit a bee makes on a flower where feeding wells are never drank from. This means that we ignored visits prior to the focal visit where the bee does not feed, and also ignored return visits to the flower after the focal visit (in instances where the flower could not be removed in time).
Three metrics were collected for each bee. All metrics were collected by a single observer (MJMH) for internal consistency.
Proportion of failed visits Floral guides have previously been shown to reduce the proportion of failed visits made by the pollinator (Goyret et al. 2007; Hansen et al. 2012). For each bee, we calculated this as the proportion of focal visits where the bee failed to find rewards (‘failed visits’) in the previous 10 focal flower visits at 10, 20 and 30 focal visits (so calculating the proportions for visits 1–10, 11–20 and 21–30). In instances were camera errors meant focal visit 30 was not recorded, noted above, the proportions of failed visits at 30 focal visits for these bees were calculated using the previous 9 visits recorded, but were otherwise treated the same. For these proportions, focal visits made after visit 30 were ignored for analyses.
Proportion of first-feeder visits Floral guides have been identified to indicate reward locations and draw pollinators to them, which can lead to an increased incidence of visits to a rewarding feeder immediately upon sampling a flower (Johnson and Dafni 1998; Lunau et al. 2006; Goodale et al. 2014). We measured these in an identical manner to the proportion of failed visits, except considering the proportion of focal visits where the rewarding feeder was the first on a flower that a visiting bee chose to drink from.
Reward search time Floral guides have previously been shown to reduce the length of reward search time (Leonard and Papaj 2011; Goyret and Kelber 2011; Lawson et al. 2017a). During each focal visit, it was recorded whether bees found the rewarding feeders on the first, second, or third feeder they drank from (ignoring revisits to drink again from the same feeder), or whether they departed after having failed to feed on the rewarding feeder. On focal flower visits where rewards were successfully found (a ‘successful visit’), the reward search time was recorded as the time between the start of the focal visit and the start of the first drinking event from the rewarding feeder. This was measured with a stopwatch while replaying video in real time, as done in Lawson et al. (2017a). All data on reward search times was used in analyses, including recorded focal visits that occurred after 30 visits.
All data were analysed using R 3.4.1 (R Development Core Team 2017). Proportions of failed visits and first-feeder visits were arcsine square root transformed. How proportions of failed visits, proportions of first-feeder visits and reward search time were altered by the floral patterns presented to each bee, and how these changed with experience (the number of focal visits) were analysed independently using a generalised linear models and AIC model simplification techniques. This involves a sequential process of paired comparisons between a standing ‘best model’ and a simpler model fitted to the data using AIC. Simpler models were constructed by removal of parameters from the standing best model (forcing parameters to equal zero). If removal of parameters resulted in a significant increase in AIC, based on Richards (2008), the standing best (more complex) model would remain the best for the next comparison. If otherwise, simpler models would become the standing best model for the next comparison. The best fitting model being the one remaining at the end of the sequence of comparisons. Each metric was assessed independently, through the same sequence as described in detail in Appendix S3 (Supplementary Material).