Honey bee foraging behaviour can be in ﬂ uenced by preferences for oscillating ﬂ owers

e

A substantial body of research has explored the interactions of flying insects with moving objects under various behavioural contexts including predation, mating and chasing.We can observe these interactions when insects intercept their prey, chase away predators, pursue their mates, and avoid or land on objects.Flying insects employ a range of visual cues which help them identify, track, land on or catch their targets.For example, passing prey can generate high image velocities in the retinae of dragonflies and serve as potential triggers for them to initiate pursuit and interception (Olberg et al., 2000).Dragonflies are believed to use an interception strategy known as proportional navigation.Here, they change course through rotational accelerations which are proportional to the rate of change of the line of sight (LoS), i.e. the angular velocity of the target.This helps them maintain the LoS angle with the target (Chance, 2019;Lohmann et al., 2019;Olberg, 2012).Killer flies, on the other hand, have been reported to use a threshold of the ratio of angular velocity to the subtended size of the passing object in order to decide whether, and if so when, to launch an attack (Wardill et al., 2015).Studies of robber flies have demonstrated in-flight avoidance of incoming objects where the veering angle to avoid collision is dependent on the relative increase in the stimulus size, following which, they too use proportional navigation to control their flight path (Fabian et al., 2022;Knight, 2022).When investigating the detection of incoming predatory objects, studies on Drosophila have shown their ability to detect looming stimuli and revealed escape mechanisms dependent on multiple neural pathways, which allow them to choose between flights that are quick and unstable or slower but more stable (von Reyn et al., 2014;von Reyn et al., 2017).Mosquitoes also demonstrate similar escape behaviour based on looming detection (Cribellier et al., 2022).These are just some of the various examples where flying insects are required to detect and interact with moving objects.While previous studies have mainly focused on interactions between moving stimuli and insects in the contexts of predation and chasing, the effects of stimuli movement in other behavioural contexts, such as foraging, have been less explored.
Central place foragers like honey bees spend substantial time flying in their environments to gather food, develop and maintain the hive, navigate around conspecifics, avoid debris and chase off predators.During many of these behaviours, they would be required to parsimoniously extract relevant visual information and enact efficient flight manoeuvres.Studies so far have shown that honey bees and bumble bees can identify and land on stationary flowers (Von Frisch, 1914) using edge detection, the discrimination of patterns, colour contrast and feature detection (Howard et al., 2021).The detectability of a relevant visual stimulus, such as a flower, depends upon its colour contrast as well as the visual angle that it subtends on the retinae (Giurfa et al., 1996).Additionally, bees can recognize natural flowers on the basis of their symmetry, as well as from the temporal patterns that are generated in the eye while flying past them (Horridge & Zhang, 1995;Von Frisch, 1914).They can combine these with memory information that retains the flowers' locations and characteristics of nearby landmarks, to return reliably and repeatedly to rewarding sources of food (Gould, 1987a;1987b).
In nature, however, flowers are frequently prone to disturbances from wind or nearby fauna, causing them to exhibit characteristic oscillatory motions along a fixed but flexible stem (Sponberg et al., 2015).Previous studies have focused on the static visual features of flowers and the cues used by insects to identify them, but the effects of flower movement remain relatively unexplored.A few studies have sought to investigate this question by examining the direct and indirect ways in which wind affects the kinematics and foraging behaviour of such insects (Baird et al., 2021;Barron & Srinivasan, 2006).Studies on bumble bees found that, rather than avoiding foraging altogether, bees increase their wingbeat frequency and amplitude to compensate for high wind speeds (Crall et al., 2017).Increasing wind speeds can also affect foraging behaviour in honey bees, where flower visits drop with increasing wind speed (Hennessy et al., 2021).This foraging frequency is governed by the flower handling rate of the honey bees which, in turn, depends on the wind speed.Honey bees are also more hesitant to take off from flowers while foraging in windy conditions (Hennessy et al., 2021).In relation to the indirect effects of flower movement, (Hennessy et al., 2020) also conducted some experiments using stimuli that mimicked flowers blowing in the wind.They did not find any significant correlation between the frequency of oscillation of the stimulus and the frequency of visits, but the search time (described as the time a bee spends on a flower after landing on it and before probing for nectar) decreased significantly with increasing oscillation frequency, suggesting a hesitancy to start feeding under high oscillation conditions.The behavioural traits observed in these experiments are more likely to be indicative of the biomechanical challenges faced by the honey bees when taking off in high-wind environments, rather than their visually directed behavioural strategies.There is a general lack of research that explores the behavioural impacts arising from the visual characteristics of flower movement.From a visual processing standpoint, a flower oscillating in the wind, when observed from a sufficiently close distance to be individually identifiable, would generate perceptible optic flow cues, thus increasing the visual signal available for detection.Additionally, if object motion acts as a visual cue in honey bee decision making, it is also possible that honey bees demonstrate a context-dependent approach to moving objects.Peat and Goulson (2005) have suggested that insects like bumble bees can correlate high-humidity weather conditions with higher nectar secretion and lower nectar evaporation, making them more favourable for foraging.Honey bees could similarly recognize the characteristics of motion most commonly associated with healthy, nectar-bearing flowers, to recognize them quickly and differentiate them from other moving objects.A nectar-rich flower could make the flower top-heavy, leading to a greater amplitude of oscillation compared to a depleted flower.Such distinguishing oscillatory behaviour could potentially be used by honey bees to identify favourable flowers.
In this study, we investigated honey bees' foraging behaviour to test whether they can identify object motion and distinguish between stationary and nonstationary stimuli.We hypothesized that bees would be adept at approaching oscillating flowers and even display a preference for such flowers.To test this, we designed choice-based experiments where bees were provided with an option to choose between flower-like stimuli that were either stationary or exhibited movements similar to those of real flowers.

Animals
Experiments were conducted between August and October of 2022 at the University of New South Wales in Canberra, Australia.The experimental setting was a rooftop location that was part of the School of Engineering and Technology, where Langstroth hives were maintained by a commercial breeder and apiarist.Western honey bees, Apis mellifera, were used for the experiments.Honey bees were initially trained to forage on a gravity feeder (feeder-A) set up around 10 m from the hives.The sugar concentration in this feeder ranged between 10 and 20% v/v, depending on the ambient conditions, to maintain a healthy and manageable population of 30e40 visiting bees at any given time.Honey bees were allowed to feed ad libitum during the training stages, but during the experimental stages, they were captured soon after landing to prevent them from attempting multiple landings.

Experimental Tunnel
Corflute was used to construct a 1.44 Â 0.55 Â 0.44 m 3 tunnel for conducting the experiments (Fig. 1).The top of the tunnel was covered by transparent UV-blocking Plexiglas for clear visibility.The floor and walls of the tunnel were lined with random cloud patterns with a (1/f) spatial frequency spectrum, as in other recent studies (Ravi et al., 2019(Ravi et al., , 2020).An entrance hole on one side of the tunnel, equipped with a door, was used to control the entry of individual bees into the tunnel.The other side of the tunnel carried a similarly designed exit hole in the Plexiglas ceiling.

Stimuli
As bees entered the flight tunnel, they would proceed along the midline to the end of the tunnel, where they were presented with visual stimuli (Fig. 1).The stimuli were yellow 3D-printed flowers that were either disc-shaped (58 mm in diameter) or diamond-shaped (squares of diagonal length 58 mm, with the diagonal oriented vertically), carrying a small feeding platform on which the honey bees could land and feed from an artificial nectary.The stimuli were either stationary or moving (oscillating from side to side forming an arc).The stationary stimuli were carried by a thin vertical aluminium rod mounted on a wooden pedestal (Fig. 1) and the moving stimuli were carried by a similar rod attached to a servo motor that was controlled by an Arduino Uno microcontroller board.The tunnel was constructed such that the view from its entrance revealed only the stimulus and the supporting rod, while the motors, wiring and other structures were concealed by a baffle (Fig. 1).Placed 1 m from the entrance of the tunnel, the stimuli would subtend an angle of about 3.32 from the tunnel entrance when centrally placed and 3.26 when placed in the middle of the left or right half of the tunnel and these angular sizes increase as the bee moves along the tunnel towards the stimuli.These sizes are sufficiently higher than the feature detectability thresholds of honey bees identified from previous experiments (Hecht & Wolf, 1929;Rigosi et al., 2017) and should allow the honey bees to visually distinguish the stimuli as individual entities from the time they first enter the tunnel.
Activating the servo motor moved the stimulus in an oscillatory inverted-pendulum motion, with an arc length of 100 mm subtending an amplitude of ±35 .The frequency of oscillation was 1.4 Hz, which is within the range of oscillatory frequencies determined to be most suitable for landing and foraging by honey bees in other studies (Hennessy et al., 2020) and in line with the flower movement frequencies observed in other plants that insects like moths land on and interact with (Sponberg et al., 2015).

Training and Experiments
Prior to each experiment, honey bee foragers were trained to fly in the flight tunnel and allowed to become familiar with the apparatus.This was achieved by recruiting foragers from Feeder-A to move into the tunnel through the entrance, feed from a secondary feeder placed at the end of the tunnel (Feeder-B) and leave the tunnel from the exit hole on the ceiling directly above the stimulus.The population of bees visiting Feeder-B would gradually increase as trained honey bees recruited additional ones.The sugar concentration in Feeder-B was maintained at a higher level (40% v/ v), to motivate the recruited subset from Feeder-A to enter the tunnel, while avoiding an unmanageable influx of new recruits to Feeder-B.Honey bees were considered adequately trained for the experiments when each of them had performed 10 undistracted and straight flights from the entrance to Feeder-B.
A total of five experiments, consisting of two tests and three controls, were performed.Each of these experiments was performed in bouts.Each bout consisted of around 10 honey bees, which helped control the population of honey bees undergoing the experiment.During the experiments, only one bee was allowed into the tunnel at a time to prevent other individuals influencing it.As the bee reached the end of the tunnel, it had to choose between two visual stimuli, placed on either side (left or right).The experiments were designed to eliminate the effects of any possible side (left -right) biases in individual bees by randomly swapping the locations of the two stimuli throughout the experiment.The stimuli and the nectaries were also cleaned with fresh water after every bee visit and deodorized with 99% v/v ethanol every two to three visits, followed by rinsing with water and drying with paper towels to eliminate olfactory cues.
All of the stimuli that were presented (during the training as well as the tests) were reward bearing.The reason was to keep the conditions similar between the training and test phases.Retaining the reward in the tests should not alter the bees' relative preferences for the two test stimuli.
Since each trial in the experiment required a unique bee, honey bees from each experimental bout were temporarily captured in a specialized nonlethal bug-vacuum (Bug-Vac, BioQuip, Rancho Domingues, CA, U.S.A.) after first landing on the experimental stimuli to prevent any revisits.They were then marked using Poska pens on their abdomens before being released after each experiment.This allowed the recognition of revisiting honey bees so that their choices on subsequent revisits could be ignored.

Experiment 1
This experiment, which is central to the question posed in this study, examined whether bees can distinguish between a stationary and a moving disc-shaped stimulus (disc) and whether they display any preferences between them.The test was performed on two sets of honey bees (N ¼ 51 for each set), in five successive bouts of 10, 10, 10, 10 and 11 individuals, respectively.
One set of honey bees was trained on a stationary disc and the other on a moving disc, and the two training stimuli were placed centrally at the end of the tunnel.After sufficient training, the training disc was replaced by two identical rewarding test discs, one stationary and the other moving.The positions of the moving and stationary discs were swapped randomly during the trials in a predetermined manner.The trained honey bees were thus presented with a choice between two stimuli, one of which was moving, while the other remained stationary (Fig. 2).Only their first landing choices on these discs were recorded and their choices were then marked.If a honey bee landed on the moving disc, its choice was recorded as a '1'; if it landed on the stationary disc, the choice was marked as a '0'.Throughout all the experiments, the bees were consistently observed to align their flight direction with their landing choice during the early part of their flight through the tunnel and last-moment decision making was infrequent, occurring in only about 5% of the flights that were analysed.

Experiment 2
This experiment was performed to test whether it was only the state of stimulus motion that governed the honey bees' choices.This question was investigated by training the bees on a stimulus of a certain shape, and subsequently presenting them with stimuli of a different shape in the tests.For this, two sets of honey bees (N ¼ 51 for each set) were trained on a disc-shaped stimulus that was either stationary or moving, and placed in the middle of the tunnel.
Following training, as each bee entered the flight tunnel during the test, it was presented with a choice between two rewarding diamond-shaped stimuli.One diamond was moving, while the other was stationary, and these were swapped randomly during the trials (Fig. 3).Similar to the previous experiment, only the first choice of each bee was recorded.Choices were then scored as '1' or '0' as in experiment 1, depending on whether the bees chose the moving or stationary diamond.These experiments were also performed in bouts of 10, 10, 10, 10 and 11 individuals, respectively, as in experiment 1.
Experiment 3 A control experiment was conducted to test for possible inherent directional biases in the choices of individual bees, or for possible unintentional asymmetries in the experimental apparatus or procedure.For this control test, bees were initially trained with either a stationary or moving disc, placed centrally at the end of the tunnel.
The trained bees were then tested by presenting them with two identical rewarding discs, one on either side of the tunnel (Fig. 4).For each training condition (stationary or moving disc), the discs in the tests had matching motion profiles (stationary or moving).Experiments were performed on two sets of honey bees (N ¼ 39 for each set), in four bouts of 10, 10, 10 and 9 individuals, respectively.Choices were recorded as the side (left or right) on which each bee landed and scored as '1' or '0' depending on whether they landed on the left (scored as 1) or right (scored as 0).Only the first choice of each bee was recorded.

Experiment 4
In experiments 1 and 2, only one of the stimuli was moving, actuated by a servo motor.Therefore, the oscillating stimulus generated some noise whereas the stationary stimulus was silent.Experiment 4 was performed to test whether this asymmetry in the noise profiles of the stimuli could have influenced the bees' choices.
The experiment was performed on one set of honey bees (N ¼ 51), in successive bouts of 10, 10, 10, 10 and 11 bees.For this experiment, the set-up was modified to incorporate two servo motors, one placed under each stimulus.The bees were initially trained to a stationary disc.They were then tested by presenting them with a choice between two identical rewarding stationary discs, where one of the discs was associated with noise from the servo motor beneath it, while the other was silent (Fig. 5).The positions of the noisy and silent discs were swapped randomly during the trials.The bees' choices for the two test stimuli were recorded and marked as a '1' or '0' depending on the chosen side to examine whether there was any aversion (or attraction) to the noisy stimulus.

Experiment 5
This control experiment examined whether the bees were capable of distinguishing between the circular and the diamondshaped stimulus used in experiment 2. This was investigated by training two sets of honey bees (N ¼ 26 for each set), one set to a stationary disc and the other to a diamond.
Experiments were performed in three bouts of 10, 10 and 6, respectively, for each set.Following the training, the bees were presented with a choice between a rewarding disc and a diamond (Fig. 6).Both stimuli were stationary.The bees' choices were recorded as correct (scored as '1') when they matched the trained stimulus, and incorrect (scored as '0') when they did not.

Statistical Analysis of Data
All the statistical analyses were conducted using SPSS version 27 Statistics (IBM Corp., Armonk, NY, U.S.A.) and MATLAB 2023b (MathWorks, Natick, MA, U.S.A.).For all the analyses, a two-tailed binomial test was used to investigate the relationships between stimuli placement and honey bee choices by comparing the recorded choices against a random chance-based distribution of 50%.To test for the dichotomous choices made by honey bees between identical stimuli in experiment 1, a binomial test was performed.

Ethical Note
While working with honey bees does not require ethical reviews, licences and permits, we exercised great caution nevertheless, causing them minimal disruption during our experiments.A total of 385 free-flying honey bees were recruited for the experiments during the training and testing phases.Honey bees were rewarded with sugar solution throughout the experiments, and since they were already foraging when recruited for the experiments, there was negligible disturbance to their routine.During the experiments, bees were captured using a nonlethal modified vacuum after which they were marked and released to continue their regular foraging activities with no further experiments necessary.

Experiment 1: Stationary Versus Moving Stimulus
In experiment 1, when bees were trained on a moving disc and then offered a choice between a stationary and a moving disc, 76.5% chose the moving disc, while 23.5% chose the stationary one (Fig. 7a).There was a highly significant difference between the choices for the two stimuli (Fig. 7a; N ¼ 51; two-tailed binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.14; P ¼ 1.980 Â 10 À4 ).When bees were trained on a stationary disc and then offered a choice between a stationary and a moving disc, 43.1% chose the stationary disc, while 56.9% chose the moving one

Experiment 2: Generalization of Motion Discrimination
Experiment 2 examined whether bees that are trained to distinguish between a stationary and a moving stimulus of a given shape can also distinguish between a stationary and a moving stimulus of a different shape, on which they have not been trained.When bees trained on a moving disc were tested with a choice between a stationary and a moving diamond, they showed a strong and statistically significant preference for the moving diamond (80.4%), as opposed to 19.6% for the stationary diamond (Fig. 8a; N ¼ 51; two-tailed binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.38; P ¼ 1.100 Â 10 À2 ).When bees trained on a stationary disc were tested with a choice between a stationary and a moving diamond, they again showed a strong and statistically significant preference for the moving diamond (68.6%), as opposed to 31.4% for the stationary diamond (Fig. 8b; N ¼ 51; two-tailed binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.6; P ¼ 1.500 Â 10 À5 ).Consistent with the results in experiment 1, the bees preferred the moving stimulus over the stationary stimulus, when trained with the moving stimulus.Additionally, they also showed a similar preference when trained on the stationary stimulus, after each training condition.These results suggest that: (1) trained bees can potentially distinguish between stationary and moving stimuli even when they are tested on novel stimuli and (2) they exhibit a preference for a moving stimulus, regardless of whether the movement is generated by an object with a familiar or an unfamiliar shape.

Experiment 3: Control to Test for Side Bias
When bees were trained with a centrally positioned moving disc and then tested with two moving discs identical to the one they were trained with, the landing choices for the test stimuli were 53.8% (left) and 46.2% (right; Fig. 9a).When the bees were trained with a centrally positioned stationary disc and then tested with two identical stationary discs (one placed on the left and the other on the right), the landing choices for the test stimuli were 51.3% (left) and 48.7% (right; Fig. 9b).
In both cases, the bees' preferences for the left or right disc were not statistically significant from the random-choice level of 50% (Fig. 9; N ¼ 39 for each case; binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.02 when training with the stationary disc, and 1.08 when training with the moving disc; P ¼ 1.000 when training with the stationary disc and P ¼ 0.749 when training with the moving disc).

Experiment 4: Control to Test for Influence of Noise
When bees, trained to a silent, centrally positioned disc, were presented with two identical discs, one silent and the other associated with servo motor noise, the choices for the silent and noisy discs were 47.1% and 52.9%, respectively, which were not significantly different (Fig. 10; N ¼ 51; binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.06; P ¼ 0.779).

Experiment 5: Control to Test for Shape Discrimination
Bees, trained on a stationary disc (N ¼ 26 bees) or a stationary diamond (N ¼ 26 bees), were tested for their choice between a stationary diamond and a stationary disc.In the tests, bees trained on the disc displayed a choice frequency of 84.6% for the disc and 15.4% for the diamond (Fig. 11a).The preference for the disc was highly significant (Fig. 11a; N ¼ 26; two-tailed binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.62; P ¼ 5.340 Â 10 À4 ).
The bees that were trained on the diamond displayed a choice frequency of 80.8% for the diamond in the tests and 19.2% for the disc (Fig. 11b).The preference for the diamond was highly significant (Fig. 11b; N ¼ 26; two-tailed binomial test; expected value ¼ 0.5; a ¼ 0.05; effect size ¼ 1.7; P ¼ 2.494 Â 10 À3 ).These results demonstrate that the bees were clearly able to distinguish between the two shapes (disc and diamond).Therefore, in experiment 2, the bees' consistent preference for the moving diamond shape, even when trained on a disc, reinforces the fact that they predominantly chose the moving diamond stimuli in spite of the noticeable shape differences when compared to the disc, and not because the differences between the two shapes were imperceptible.In fact, experiments 2 and 5, considered together, suggest that bees treat the motion of an object as a separate visual cue that is independent of the shape of the object, even when they can perceive and memorize this shape.

DISCUSSION
A honey bee, in its natural environment, carries out multiple trips over several hundreds of metres or even a few kilometres from the hive in search of nectar and pollen (Beekman & Ratnieks, 2000;Von Frisch, 2013).In these journeys, it is likely to encounter a range of objects, and appropriate interactions with them may require different perceptual and motor capacities.Objects such as trees and natural or artificial features in the environment could be perceived as obstacles to navigate around.On the other hand, the presence of predators or other unfamiliar objects may require evasive manoeuvres.Objects of interest such as moving flowers or approaching nest intruders could elicit tracking, pursuit or landing (and stinging, in the case of an intruder).In their natural environments, flowers are not always stationary; factors such as wind or disturbances caused by other animals in the vicinity may cause them to sway.
There has been considerable research investigating the interactions of flying insects with moving objects, particularly in the context of tracking and interception of moving targets (Collett & Land, 1978;Land & Collett, 1974).Dragonflies, for example, use a predictive interception strategy (Olberg et al., 2000;Olberg, 2012) which involves course corrections by rotating the body in proportion to the change in angle between the target and an exocentric reference frame (Fabian et al., 2018).Some insects, however, use a pursuit strategy that is less computationally intensive in which they simply fly directly towards the target, reorienting the flight direction moment-to-moment, by maintaining the image of the target in the centre of their visual field (Mischiati et al., 2015).Honey bees have demonstrated such behaviour when landing on moving food targets (Zhang et al., 1990).House flies and blow flies also show similar behaviour when chasing conspecifics (Land & Collett, 1974).By giving honey bees a choice between stationary discs and discs that moved along a circular trajectory, (Lehrer & Srinivasan, 1992) demonstrated their ability to spontaneously discriminate between moving and stationary discs.They noted the ability of honey bees to distinguish and spontaneously choose the moving disks over the stationary ones when the disc movement was against a stationary background, leading to the conclusion that the ability to discriminate moving targets depended on the motion contrast that the moving disc provides with respect to the background, rather than the movement of its image on the retina itself.They also showed that honey bees were more attracted to faster-moving objects.However, in these experiments, movement was simulated through black discs which described large circular paths of circumference 104 cm in a horizontal plane at an orbital rate of 14.5 rpm, corresponding to a low oscillation frequency of about 1/4 Hz.This was in contrast to the higher frequency, vertically oriented objects used in our experiments which moved in small arcs to represent a more ecologically relevant scenario reminiscent of a moving flower.More recently, research conducted by Hennessy et al. (2020) also provides preliminary evidence suggesting that foraging honey bees find it easier to land on flowers that oscillate at higher frequencies (ca.1e2 Hz), which is compatible with the frequency of 1.4 Hz used in our experiments.
The results of our study align with Lehrer and Srinivasan's (1992) observations and suggest that, in the context of foraging on flowers, honey bees display a preference for moving/oscillating flowers over stationary ones.In experiment 1, when two sets of honey bees were trained with either a moving or a stationary disc, and subsequently tested for their preference between simultaneously placed moving and stationary discs, those trained on the moving disc demonstrated an affinity for it (see Fig. 7).
After training with a stationary stimulus, however, while more bees chose the moving stimulus in the test, the statistical analysis failed to establish a clear preference for it.The influence of the stationary training stimulus could be a potential reason; however, this does not appear to be the case.If the bees' choices were driven purely by the stimulus on which they were trained, they would have shown a higher preference for the stationary stimulus in the test, when they were trained on a stationary stimulus.However, there was no significant difference between the preferences for the stationary and moving stimuli in this case, suggesting that a potential preference for a moving stimulus was partly overridden by training on the stationary stimulus.
The results from experiment 2, however, provide clearer results by replacing the test stimuli from experiment 1 with differently shaped ones to avoid any influences from the geometric makeup of the training stimulus.Our results from experiment 2 confirm that even when the bees were trained on a stimulus of a particular shape prior to choice testing in all the experiments, their decisions were not contingent on the geometric makeup of the training stimulus since they landed on diamonds in spite of their initial familiarity with disc-shaped stimuli (see Fig. 8).When the test from experiment 1 was repeated using the disc as a training stimulus but the test stimulus was a diamond, the honey bees continued to retain their preference for the moving stimulus, which in this case had a different shape.Thus, despite the change in shape between the training and test stimuli, honey bees demonstrated a statistically significant preference for the moving stimulus in both training conditions.
These experiments reveal a tendency of honey bees to distinguish between motion and stillness, which is independent of the stimulus shape and irrespective of whether the stimuli are familiar or not, suggesting that motion is a visual cue that they potentially use in their decision making.Note that while the relevance of shape as a possible feature cue has been examined, the investigation of honey bee behaviour in the presence of multiple cues remains to be investigated.
One of the reasons that bees display this preference could be the additional visual salience provided by a moving flower.Honey bees are known to use optic flow information for a variety of mechanisms such as course control (Mauss & Borst, 2020), altitude control (Portelli et al., 2010), odometry (Labhart & Meyer, 2002;Srinivasan et al., 2000) and regulating flight speed (Mauss & Borst, 2020).Previous research has also demonstrated that side-to-side motion, which insects like honey bees and bumblebees perform during their flights, can help them glean range information, based on the optic flow generated by various objects in the visual field (Boeddeker & Hemmi, 2010;Dittmar et al., 2010;Lehrer, 1996;Ravi et al., 2019;Srinivasan & Zhang, 2004).In line with these findings, honey bees approaching the stimulus in our experiment were also observed to perform lateral translations.As they approach the stimuli roughly along the midline of the tunnel, both stimuli would generate an expanding optic flow.The stationary stimulus would generate a smooth lateral optic flow of gradually increasing magnitude as it is approached, while the oscillating stimulus would modulate this lateral optic flow.In the case of the current experiments where a bee moved in a straight line with a constant velocity towards identical stationary and moving stimuli placed to its right and left, we observed that the perceived optic flow was predictably altered by the moving stimulus.While the stationary stimulus caused a gradual increase in optic flow that peaked at around 35 degrees/s (assuming bee velocity was 1 m/s) when the stimulus was closest to the bee (distance 1 cm), the optic flow generated by the moving stimulus was variable with multiple peaks throughout the trajectory (due to the motion of the flower).The magnitudes of these peaks were two to three times higher than the corresponding optic flow observed with the stationary stimulus.The more variable and greater optic flow generated by a moving flower could serve as an attractor and make it easier to distinguish from the stationary background than a stationary flower.Further research is essential to identify the parameters of object motion that honey bees are likely to use for tracking and landing on moving flowers.
In the context of natural foraging where honey bees gather most of their nutrition from flowers, a salience-driven preference for moving flowers may be beneficial.Nectar-rich flowers that are topheavy from bearing the weight of nectar would be more likely to oscillate compared to flowers that are dry or depleted of nectar.It is hypothesized that a higher flower mass would contribute to a greater moment of inertia, so that as the wind overcomes the initial resistance to movement, a heavier flower would undergo a larger displacement from equilibrium (inverted pendulum effect), resulting in longer-lasting oscillations of greater amplitude and reduced damping.In this case, honey bees could associate the oscillatory behaviour in flowers with foraging potential, implying that ecological relevance could also be a potential driving factor for this preference.Previous research has demonstrated that bees can store and use ecologically relevant information about bountiful flowers, such as their colour (Brown et al., 1998), shape (Gould, 1987a), scent (Von Frisch, 2013) and positive or negative charge (Clarke et al., 2013) to guide foraging choice.Similarly, it may also be beneficial to guide foraging choices through the movement profiles of flowers.However, further research is required to explore the hypothesis that oscillations in flowers are linked to foraging profitability.
In summary, this study reveals that honey bees display a potential preference to land on oscillating flower-like objects, compared to stationary ones.The flower movements in our experiments likely provided the honey bees with additional visual cues that enabled easier detection.However, establishing the precise nature of this preference is beyond the scope of our current study but would be a logical next step in understanding the interactions of honey bees with moving objects.The current study also does not establish whether this motion-based decision making is exclusive of other visual cues.The results of this investigation prompt further studies identifying innate preferences for object motion and exploring the interplay between different visual cues involved in honey bee decision making.Further research would uncover the various features of motion that influence honey bee behaviour, as well as the corresponding strategies they employ when interacting with moving stimuli.

Data Availability
The data set used for this study is provided in the Supplementary Material.

Declaration of Interest
No conflicts of interest to declare.

Figure 1 .
Figure 1.(a) Schematic of the tunnel design and set-up of the experiment where the location and shape of the reward-carrying stimuli could be varied between experiments.(b) Design of the disc-shaped stimulus used for the choice experiments.

Figure 2 .
Figure2.A cross-sectional view of the design for experiment 1 as seen from the tunnel entrance.Bees were trained with either a stationary or moving disc-shaped flower (disc) and then given a choice between two identical moving and stationary discs.

Figure 3 .Figure 4 .Figure 5 .Figure 6 .
Figure3.A cross-sectional view of the design for experiment 2 as seen from the tunnel entrance.Bees were trained with either a stationary or moving disc.Each bee was then given a choice between identical moving and stationary diamonds.

Figure 7 .Figure 8 .
Figure 7.The results from experiment 1 showing the honey bees' choices when (a) trained with a moving stimulus and (b) trained with a stationary stimulus.Asterisks indicate significant differences (*P < 0.05) between the bees' choices.

Figure 9 .Figure 10 .
Figure 9.The results from experiment 3 showing the honey bees' preferences when tested with two discs that were the same as the training discs in shape and movement profile.(a) Bees' choices when trained with a moving disc.(b) Bees' choices when trained with a stationary disc.

Figure 11 .
Figure11.The results from experiment 5 in which honey bees were (a) trained with a disc and (b) trained with a diamond and given a choice between the two stimuli.Asterisks indicate significant differences (*P < 0.05) between the bees' choices.