While most models of decision-making assume that individuals assign options absolute values, animals often assess options comparatively, violating principles of economic rationality. Such ‘irrational’ preferences are especially common when two rewards vary along multiple dimensions of quality and a third, ‘decoy’ option is available. Bumblebees are models of decision-making, yet whether they are subject to decoy effects is unknown. We addressed this question using bumblebees (Bombus impatiens) choosing between flowers that varied in their nectar concentration and reward rate. We first gave bees a choice between two flower types, one higher in concentration and the other higher in reward rate. Bees were then given a choice between these flowers and either a ‘concentration’ or ‘rate’ decoy, designed to be asymmetrically dominated on each axis. The rate decoy increased bees’ preference in the expected direction, while the concentration decoy did not. In a second experiment, we manipulated choices along two single reward dimensions to test whether this discrepancy was explained by differences in how concentration vs. reward rate were evaluated. We found that low-concentration decoys increased bees’ preference for the medium option as predicted, whereas low-rate decoys had no effect. Our results suggest that both low- and high-value flowers can influence pollinator preferences in ways previously unconsidered.

a)  Study system

We conducted experiments with bumblebees (Bombus impatiens) (n=103) from commercially-reared queenright colonies (n = 8) (Koppert, USA); sample sizes per colony are shown in Table S1. Colonies were maintained in small boxes (~40cm³) on 30% (w/w) sucrose solution and honeybee-collected pollen (~0.5g/day, Koppert Biological Systems, USA). We connected colonies sequentially to a flight arena (l × w × h: 122 × 61 × 61 cm), using a clear plastic tube with sliding doors to control the entry and exit of bees into the arena. The floor of the flight arena was lined with green laminate and the sides and top consisted of black mesh screens. The arena was lit from above by a 40-watt LED light placed atop the arena and the room was illuminated with fluorescent light on a 12/12-hour light/dark schedule.

b)  Experimental arrays and floral stimuli

The vertical arrays used for experiments consisted of black corrugated plastic sheets with 24 holes for ‘flowers’ arranged in a 6 × 4 grid. Twelve artificial flowers were pseudorandomly arranged to be equally represented across the array, with unoccupied holes sealed using black rubber bungs. We constructed artificial flowers from 1.5 ml plastic Eppendorf tubes with small holes cut in the base of the tubes to pipette sucrose solution. The ‘corollas’ of these artificial flowers were made from laminated, coloured circular disks (5 cm in diameter) placed around the opening of the tube. We used three colour stimuli to correspond to our three flower types: blue, yellow, and white.

To accurately monitor and manipulate the reward rate of flowers (see section d below), we attached rotating paper disks to the back of our artificial flowers, which corresponded to the four rates of reinforcement used across both experiments (Figure S1). There were eight positions on the disks that were either marked with an “X” or marked with the sucrose concentration used for that flower type (e.g., 50%). The spot in the upright position denoted whether a flower was empty or full, and whether to refill the flower following a bee visit. The experimenter rotated the disk a single position each time a bee visited the flower, only refilling it on positions not marked by “X”. Disks were only visible to the experimenter and not to the foraging bee.

c)  Initial training

During the initial training phase, we first allowed bees to forage for sucrose in the foraging arena by giving a whole colony access to a white-wicked feeder containing 30% (w/w) sucrose solution. This was a single plastic container filled with sucrose solution that had an immersed and saturated wick (a braided dental cotton roll, Richmond Dental, USA) protruding vertically from the lid, on which bees could land and drink. We initially placed the feeder close to where foragers exited their colony. Over 1-2 days, we incrementally moved the feeder towards the end of the arena until bees readily foraged on it and returned to the colony. We then trained bees to visit artificial flowers on a 24-flower array similar to the experimental array with the exception that no colour stimuli were used (Eppendorf tubes alone). All flowers were provisioned with 30% (w/w) sucrose solution. During this initial training, flowers contained an additional hole on top which we used to mark individual bees’ thoraces using non-toxic, water-based paint markers (POSCA, USA). Bees were each given a unique colour combination for identification. Foragers consistently leaving the colony to forage from this array (three or more visits) were selected for use in experiments.

d)  Experimental protocol

General Protocol

Each bee was assigned to one of either three (Experiment 1) or four (Experiment 2) treatments. All treatments were represented across all colonies within each experiment. We systematically varied treatment order across bees within a given colony to avoid any temporal confounds. All bees experienced a total of six consecutive training and testing trials. Trials lasted around 10 minutes each with bees making ~ 20-30 flower choices per trial. Twelve flowers were used on the array during the training and testing trials, and individual flowers could be re-visited since they were either re-filled or left empty. For all experiments, a flower ‘choice’ was defined as a bee entering a flower and extending her proboscis. Flowers varied in concentration (20%, 30%, 40%, and 50% sucrose (w/w)) and reward rate (25%, 50%, 75%, and 100% rewarding). Reward volumes were kept constant at 4 μl per flower. We used three different flower colors: blue, yellow, and white to indicate the different flower types. Blue and yellow flowers were always used as the two main flower types for each experiment, while decoy flowers were always white. We chose this design since a previous study indicated that bees did not show a strong preference between this particular blue and yellow, and because counter-balancing colors was not feasible in terms of the sample sizes required. Reward values were chosen such that bees had roughly equal preferences between the two main flower types and were consistent with previous studies demonstrating robust preferences in bees for higher nectar concentrations and rates of reinforcement.

In both experiments, each bee was given the following trials: Trial 1: Binary training: Bees were given a choice between six blue and six yellow flowers; this trial allowed bees to gain experience with the two main flower types. Trials 2 and 3: Binary preference tests: These trials allowed us to assess bees’ preference between the two main flower types. Trial 4: Decoy training: Bees in trinary treatments received a trial with only decoy flowers to ensure that all bees gained experience with the decoy before the trinary preference trials; bees had to visit a minimum of 10 decoy flowers to move on to the next trial. In place of the decoy training trial, bees in binary treatment received another trial with the initial two flower types. We did this rather than a ‘blank’ trial to ensure that foragers did not lose foraging motivation and to give bees across treatments the same number of trials overall. Trials 5 and 6: Trinary preference test: Bees in binary groups were given the same choice again, while bees in trinary groups had the original two flowers alongside the decoy option. These final two trials allowed us to compare bees’ binary vs. trinary preferences both within and across treatments. Bees were not trained to the decoy until trial 4, as prior experience with decoy flowers may have influenced bees’ initial preferences for R vs. C, even if those flower types were absent during later choice tests. All trials were recorded using a Canon camcorder on a tripod placed near the nest box and facing the array.

Experiment 1: Are bumblebees subject to decoy effects?

We tested 55 bees from five colonies. Bees were trained and tested individually in one of three experimental treatments: a binary choice between two flowers, one higher in reward rate (flower R, yellow) and the other higher in concentration (flower C, blue), or one of two trinary choice treatments where they also had a decoy flower (either DR or DC). Flowers R (30% sucrose, 100% rewarding) and C (50% sucrose, 50% rewarding) were designed to be roughly equally preferred, with each option being superior to the other on one dimension. Flowers DR (20% sucrose, 75% rewarding) and DC (40% sucrose, 25% rewarding) were designed to be asymmetrically dominated by flowers R and C, respectively.

Experiment 2: How do bumblebees evaluate reward rate and concentration dimensions separately?

After finding in Experiment 1 that DR induced the predicted effect, but that DC did not affect bees’ relative preferences between flowers, we conducted a second experiment to better understand this result. In Experiment 2, we tested how bees evaluate each reward dimension (concentration and reward rate) separately by determining how the relative preference between a medium and high option was affected by the presence of a lower (decoy) option. We tested 48 bees from three colonies. Using a similar design to Experiment 1, bees were trained and tested individually across one of two experimental treatments (binary or trinary) for each reward dimension. Bees in binary treatments were given a choice between a medium-quality flower (flower M) and a high-quality flower (flower H). Bees in trinary treatments were also given experience with a low-quality decoy flower (flower L) and subsequently given a three-choice test between all three flower types. In this experiment, because there was a clear ranking of flowers within each choice set, any initial color preference would likely be enhanced by the preference for the higher-quality flower. As such, we counter-balanced color (blue or yellow) across the two flower types (flowers M and H) to control for any color bias associated with the best option.