Testing bottom-up cuing effects on target detection and discrimination in Bumblebees

T es ting bot t om -up c uing eff ec ts on t a r get det ection and dis c rimina tion in Bumbl ebees Théo Robert 1X* , Marion Callendr e t 1 , Chloe Sowels 1 , Vivek Nityananda 1+ 1 Biosciences Institute, Henry Wellcome Building, Newcastle University , Framlington Place, Newcastle upon T yne, UK, NE2 4HH X Or cid ID: https://orcid.or g /0000-0002-8475-4154 + Or cid ID: https://orcid.or g /0000-0002-2878-2425 * Correspondence: Theo.Robe [email protected], [email protected] Abs tr ac t A ttention in vertebrates helps prioritise the processing of important sensory inf ormation and filt e r out irrelevant signals. The captur e of attention by sudden or salient stimuli typically called bottom-up attention. Little is known about similar attentional process in insects, although they should be advantageous for insects a s well. We therefore adapt ed two paradigms used to inves tigate bottom- up attention in primates to investigate it in bumblebees: a targe t detection task and a t arget discrimination task. For both tasks, we trained bees to cho ose between two locations on each side of a computer screen and collec t a rewa r d bellow a full contrast t a r get displayed on the scr een. During detec tion task tests, the contrast of the target was varied and it could be pr eceded by a cue flashed on the side of the target, the opposi t e side of the screen or not flashed at all. The discrimination t a sk tests were similar but with a full contrast t arget on one side and a variable contr a st distractor on the opposite side of the sc r een. We tested if the pre sence of the fl ash influenc ed the orientati on and choices of the bees as well as th eir contrast sensitivity as has been se en in primates. Our results show no effect of the prior cue, sugges ting that other paradigms might prove more useful to test these proce sses in insects. Key W o r d s Bumblebee, insect cognition, exogenous attention, bottom-up attention In t r oduction Animals ar e constantly exposed to a multitude of sensory stimuli in their envir onment. The limited neural resourc es available t o an animal means that the se stimuli have to compete for processing and stimuli that a r e releva nt t o survival need to be priori tized. V ertebrates have the r efor e evolved attentional processes through which can prioritise the treatment of the relevant sensory stimuli and filter out noise (reviewe d in Carrasco, 2011). One form of attentional prioritiz ation is based on spatial location. Humans are fa ster to detect a visual target at a particula r location, if they are previously cued to that location (Posner , 1980) Conversely , the detection time incr e ases if the subjects r e cued towards a different location. T wo distinct mechanisms underlie spati al attention. The first involves tuning sensory systems t o increase the saliency of specific stimuli previously associated with a reward (Maunsell & T r eue, 2006; Scolari .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint et al., 2014). Primates can willingly increase their sensory sensitivity in specific spatial ar e as based on available information (Fernández et al., 2022). This active process, generally referred as top-down or endogenous attention, can help animals better detect relevant goals in the environment around them. In the se cond mecha nism, subjects’ attention can be captured and directed towards a region in space by salient stimuli such a flash or a loud sound through an involuntary process called bottom- up or ex ogenous attention (Henderson & Macquistan, 1993). Bottom-up attention has been demonstrated in many vertebrate taxa. In addition to humans, other - non-huma n - primates have also been shown to have such an attentional proce ss (Bowman et al. , 1993; Wang et al., 2015). A ttentional processes in primates are thought to be supported by neural pathways in their neocortex (Bowling et al., 2020; Meyer et al., 2018; for a review see Behrmann et al., 2004). Similar att e nti onal proc esses have, however , also been shown in bir ds (Ques t e t al., 2022; Shimp & Friedrich, 1993; Sridharan et al., 2014) and po ssibly in fish (Gabay et al., 2013), despite the lack of a neocort ex. It therefore seems likely that this is an important cognitive feature that is evolutionarily selected for . We could therefore expect similar attentional processe s to also have evolved in invertebrates. One of the most noticeable effects of visual bott om-up attention in primates is a localised inc rease in contrast sensitivity . Cuing a subject ’ s attention towards a location allows them to per ceive a subsequent target at that location at a lower contrast threshold, compared to when their attention is cued to another location (Barbot et al., 2011; Cameron et al., 2002; F ernández et al., 2019; Herrmann et al ., 2010; Jigo & Carrasco, 2020). An increase in contr a st sensitivity i n response to a sudden locali sed change would be useful for an animal to better perceive and ide nti fy the cause of this change. It could, for example, allow the animal to quickly recognise a predator or rapidly identify potential prey . Such adaptations would also be useful for insects. Y et very f ew studies have directly investigated attentional processes in these animals (reviewed in Nityananda, 2016). Weiderman and O’ Ca rroll (2013) demonstrated that a visual neuron of a dragonfly (CSTMD1) responded selectively to one of two tar gets moving vertically at two diff erent places in the visual field. More recent work showed that if one of the two locations was primed before the simultaneous presentation of the ta r gets, the neuron was more likely to respond to the stimulus shown at the primed location (Lancer et al., 2019). Cuing effects on bottom-up att e ntion in insec ts have also been shown behaviourally (Sareen et al 2011) in the fruit fly D r os op h i la m e la nogas t er . In this study , the authors displayed two vertical bars on a circular screen to tethered flies. When the ba rs moved in opposite directions on the scr een, the flies had an equal probability of turning in the direction of ei the r bar . However , i f one of the stimuli flashed repeatedly before it start e d moving , the flie s followed thi s bar and ignor e d the other one, demonstrating that the flashing bar captured the flies’ attention. The study did not however , test if the flashing bar led to an increase of the per ceived contrast of the subsequent sti muli, as ha s been seen in pr imates . In this study , we therefore investigated the possibility that a cue could increase visual contrast sensi tivity in insects. We conducted two different experiments with bumblebees, another model system for insect visual behaviour . In the first one, we tr ained bumblebees to collect a reward below a full contrast target displayed on a computer sc reen and tested their ability t o detect the t a r get displayed at lower contr a sts when it was preceded by a cue flashed at the same location a s the target, a diff erent location or not flashed at all . Our hypothesis was that if the cue led to a spatially localized increase of their contrast sensitivity , they would be able to detect the t arget at lower contrast when the cue wa s fla shed at the same location compared to the other cuing conditions. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint In the second experiment, bees were trained to discriminate a full contrast target fr om a variable contrast distractor displayed on the opposi t e side of the screen. During tests, a cue could be flashed on the side of the target, on the side of the di s tractor or not fla shed in a control condition. The prediction wa s that when the cue was on the side of the target, it would increase its perceived contrast and enable the bees to better discriminate it from the distractor . Conversely , when the cue was flashed on the side of the distrac tor , if it increased its perceived contrast, we predicted that this would hinder the bees’ ability to discriminate the target from the distrac tor . Ma t e rial an d Met hods Animals and setup We carried out the experiments on the buff-tailed bumblebee Bo mbus t er res t r i s . Bumblebee colonies wer e purchased from commerci al pollinator suppliers (Koppert BV , Netherlands and Agralan Lt d, UK) and transferred t o a nest box (L=28 cm, W=16 cm, H=12 cm). The nest box had two chambers, one to house the br ood and the r e st of the colony and the other containing cat litter for the bee s to dispose of thei r waste. The latter chamber was connected to a transparent tunnel leading to a foraging arena (L = 45 cm, W = 60 cm, H = 40 cm). The arena was covered with a UV transparent plexigla ss board letting thro ugh the illumination fr om a daylight spec trum tube (Philips, Master TL5 HE 35W, 6500K) fitted to a high frequency lighting s ystem (Philips, HF-P 1 14-35 TL5 HE III, >42KHz). The floor of the arena was cover ed with a random red and white che ck erboard pattern to provide the bee s with optic flow . The arena wall facing the tunn el exit had a computer screen (Dell S2419HGF , LCD , 1080p, 144 Hz) on which we could display visual stimuli during the experiments. We mount ed a smartphone (Huawei Nexus 6P ) above the arena t o r ecord test tria ls at 120 fps with a 720p r e solution. The colonies had access to pollen ad libitum in a little cup plac ed in the nest box. During evenings and weekends, f eeders filled with a 20% (w/w) sugar solution were plac ed in the foraging arena so the bee s could feed ad libitum. Detection T ask This experiment tested whether flashing a cue could improve or hinder target detection by bees when the cue wa s on the same side or the opposite side of the computer sc reen relative to the target . Pretrai ning Individually marked bees were pretrained to collect a 50% (v/v) sugar solution re ward from a little well at the centre of a transparent square plastic chip (L=2.5 cm, H=0.5 cm) plac ed on t op of an upside-down cup (H=7 cm, D=6 cm) beneath the centre of the screen in the arena. A drop of 100 µL was placed in the well and the bee was placed over the chip with a transparent container . Once the bee started drinking, the container was removed, and the bee was allowed to return freely to her nest. T r a i ni ng After completing the pretraining, bee s proceeded t o a training phase. During this phase, we placed transparent chips on cups on each side of the comput er screen. In each training bout, the scr e en wa s .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint set to display a green background (RGB values: 0, 1, 0) and a full contrast black circular target (Diameter = 5.56 cm; RGB values: 0, 0, 0) was displayed above one of the chips. The side of the target was chosen pseudo-randomly acr oss bouts, with a maximum of 2 consecutive bouts with the target on the same side. The chip below the target contained 100 µL of 50% Sucrose solution, while the other chip had 100 µL of distilled water . Between each trial, the chips were wiped with 70% ethanol to remove any pheromone marking and then cleaned with dis till ed water t o r emove the scent of the ethanol. We deemed a trial correct when the be e first chose the chip below the target, with a choice defined as probing the contents with her antennae or her proboscis. When a bee c hose the other chip, this was deemed a wrong choice, and she was allowed to correct herself and collect the reward on the correct chip before returning to her nest. Bees we r e allowed to proceed to the test phase once they r eached 80% success on the last 20 trial s (N=18). Te s t i n g The test setup was identical to the training phase but both chips cont ained 100 µL of distilled water . During a test, the experime nt e r manually triggered the cuing as soon as the bee’ s head entered the arena. This sequenc e involved one of three conditions. In the first, a blue square c ue (side 8.33 cm) was presented for 200 ms on the side of the screen where a target would later be displayed. In the second, the cue was presented on the opposite side. In the third condition, which served a s a control, no cue wa s presented. The position of the cue centre on the scr een wa s 6. 64 cm fr om the edge of the screen and from the target location (centre to centre) to prevent a masking effect (See Fig. 1). The target then appeared after a pause of 100 ms. The colour of the cue was chosen ba sed on a pr evious paper demonstrating that blue stimuli disturbed shape learning in honeybees, presumably because this colo ur captured thei r attention (Morawetz et al., 2013). Each test trial presented the bees with a target with one of 5 different contrast values (Michelson contrast: 0, 0.355, 0.615, 0.826, 1) and the side of presentation of the target was counterbalanced across trials and contr a sts. Therefore, a complete test phase wa s composed of 30 trials (3 cuing conditions x 5 contrasts x 2 sides) which wer e present ed in a randomised order . The test was conside r ed finished when the bee lande d on one of the chips and probed the well with its proboscis or antennae. Betwe en each test, we conduc ted two refresher bouts identical to the training bouts to k e ep the bee motivated. The side of the target during the first refresher was picked randomly and was counterbalanc ed for the second one. I f a bee made a mist ak e during a refresher , it was repeated until she made a correc t first choice . In total, 14 out of 32 bees did not complet e the training and could not proceed to the te s t phase either because they had died before completing the training , st opped coming out of their nest to forage or because they never reached the learning crit e rion and their training was abandoned. Among the 18 be es who started the test phase, 8 completed all 30 t e sts before dy ing. The sample siz e for each combination of cuing condition and target contrast was between 20 and 26 trials. Discrimination task This experiment tested how the bee’ s ability t o discriminate between t a r gets of different contrasts was a ff ected by a cue presented on the side of the target or on the side of the distractor . .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Pretrai ning The pr etr aining was identical t o the one of the det ection t ask. Once the bees had learned t o dr ink from the transparent chips, they were allowed to start the training phase. T r a i ni ng For the training pha se, the setup was the same as the detection task. The trai ning phase was done in two stage s. In the firs t, we presented a full contr a st black circular target (diameter = 5.56 cm, Michelson contrast = 1) on one side of the screen over one chip with 100 µL of a 50% (v/v) sugar solution. On the opposite side of the screen, we showed a similar circular distrac t o r with a 0.448 contrast above a chip filled with 100 µL of a saturated quinine solution . The bees therefor e had to learn t o approach the higher contrast target regardless of which side it was displayed. After the bees had r eached 80% success on the last 20 trial s in the first training st a ge, they proceeded to the next one. In the second training stage , the target was presented at full contrast, but the distractor had a variable contrast randomly picked without replacement from 5 possible values (Michel son contrasts: 0, 0.448, 0.680, 0.826, 0.909). The list of possible contrasts was reset after the bee had experienced the 5 contrasts. Here again, a trial was mark ed a s correct by the experimenter if the bee first chose the chip below the target with her a ntennae or her pr oboscis. Bees had to mak e 80% of correct choices on their last 20 trials to be selected for the test phase. Therefore, each bee experienc ed each distr actor contrast at least 4 times duri ng this tr aining phase. Te s t i n g During the tests, both chips contained 100 µL of distilled water . As in the detection task, the cue consisted of a blue square (side = 8.33 cm) 6.64 cm from the edge of the screen and from the ce ntre of the target or distrac tor . Three cuing conditions were implemented: the cue could be flashed on the side of the target, on the side of the distr actor or not appear at all. The duration of the cue wa s 200 ms with a 100 ms pause before the target and dis tractor were displayed. The experimenter triggered the cue as soon as the bee’ s he ad crossed the entranc e of the a r ena. In each test trial, the target was always at full contrast while the distr actor had one of 6 contras ts (Michelson contrasts: 0, 0.448, 0.680, 0.826, 0.909, 1). The side of prese ntation of the target and distractor wer e counterbalanc ed across trials. When the distractor had a contr a st equal to the target, they were indistinguishable from one another . Therefore, to reduc e the total number of trials, we pr esented the cue only on the side of the target in this specific di s tract or contrast condition (3 conditions in total, one per cuing condition). Thus, we had 33 test trials (3 cuing conditions x 6 distrac t or contrasts x 2 sides – 3 trials). Here again, the t e st was considered complete when the bee made its final choice by landing on one of the chips and probing the well with its proboscis or antennae. Betwe en tests, we conducted two refresher trainings, identical to the second training stage. We placed a 100 µL of a 50% (w/w ) sugar reward below the target and 100 µL of quinine solution under the distractor . The target contrast was full contras t while the contrast of the distr a ctor wa s one of the 5 used during the training phase. The side of presentation of each stimulus was ra ndom for the first refresher and counterbalanced for the second one. If the bee made a wrong choice, the refresher trial was repeated until she made a co rrect choice. In total, 34 bee s started their training and 12 finished it an d moved to the test phase. Finally , 8 bee s finished all tests. The sample size for eac h combination of cuing condition and distractor contrast varied between 10 and 20 trial s. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Video and trajec tory analyses T es t trial r ecor dings wer e processed with DeepLabCut ™ (Nat h et al., 2019) using a ResNet-50 network trained on a mix of 1377 video fr ames extracted fr om 68 trials from both experiments to analyse videos fr om the detection task. T o analyse videos from the di scrimination task, the network was trained on 1616 frames extracted from 78 flights from both experiments. This pr ocess allowed us to obtain the tr ajectories of bees. Since De epLabCut made a certain number of errors, the se trajec t orie s were then cleaned fr om tr acking anomalies using a custom-made code in R (version 4.2.3). All frames that were excluded as part of an anomaly were rebuilt by interpolation. In order to have the most accur ate tak e -off location for each trajectory , the videos were manually examined and the number of the frame on which the bee took off was noted. For each trajectory , if the position of the bee on the tak e-off frame had been interpolated, the bee coordinat e s were manually extracted fr om the video and fed back in the trajectory data. We used the trajectories to determine which of the two chips the bee first appr oa ched in each test. This was defined a s the first chip the bee approached at a dist ance less than 10 cm. We also measured the duration of this firs t appr oach by counting the number of video frames each bee took to complete it. As bott om-up attention la s ts for a ve ry short time in primates (Busse e t al., 2008; Hein et al., 2006; Ling & Carrasco, 2006), it was possible that the eff ect of the cue might be visible only at the very early s tage of the flight of our bumblebees. We therefore analysed whether bees flew towa r d the target immediately after their take-off depending on the cuing conditions. We computed the trajec t ory direction of each bee rel ative to the target 1 cm and 5 cm away from the take-off point. T o do so, we computed the angle subtended by the line joining the position of the bee when she was 1 cm (or 5 cm) from her tak e-off point t o her tak e-off point and the line joining the chip below the target and the location where the be e took-off . This measure s whether the direc tion of flight after take-off was, overall, in the direc tion of the target. Statistical analyses Analyse s were conducted using R (version 4.2.3). We analysed the results of the detec tion and the discrimination experiments separately but with identical methods. We ran four analyses on our data, analysing the final choices, the first approaches, the duration of the firs t approaches a nd the direction of the early trajec t ories. Fi nal c hoice For both experime nts, the final choice of the bee was recorded as 1 when the bee landed and probed the chip mark ed with the target and 0 when the bee probed the other chip. We then used Generaliz ed Linea r Mix ed Models (GLMM, package lme4, Bates et al., 2015) to analyse this data with the choice as a dependent variable and a binomial family and a logit link function. The independent variables we r e the cuing condition (cue on the side of the target, cue on the opposit e side, no cue), the contra s t (the contrast of the target for the detec tion task, the contrast of the distrac tor for the discrimination task) and their interaction. We used the ide ntity of the bee as a random eff ect. However , for the discrimination task data, the lack of variance in the random eff ect (bee identity) led .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint to a singular fit of our model. We therefore removed the random eff ec t and used a Gener alized Linear Model (GLM) to analyse thi s data. Fi r st ap proac h The same analysis a s above was conducted on the first approach data. If the first chip approached was the one below the target, this variable was ma rked as 1 otherwise it was marked as 0. A GLMM identical t o the one used for the final choices was r an on this data. He r e again, for the discrimination task dat a , the lack of variance in the random eff ect led to a singular fit of our model. So, we removed the random effect and ran a GLM on the dat a. Fi r st ap proac h d ura tion We also analysed the duration of this first approach with GLMMs using the glmmTMB function from the glmmTMB package (Brooks et al., 2017; McGillycuddy et al., 2025). The dependent variable was the time in seconds between the tak e-off and the point at which the bee was less than 10 cm away from one of the chips for the first time. In one model, we te sted the eff ect of the contrast of the target (for the de t ection task) or of the distrac t or (for the discrimination task), the cuing condition and their interaction. While in a second model, the independent variable was the contrast, the chip chos en by the bee (c orr ect or incorrect chip) and their in ter act ions . Thes e models w er e fitt ed with a Gamma family and a log link func tion. T o improve the fit of our models used to analyse the approach duration in the detection task, the dependent variable had to be trans formed by applying a log10 function to it. Finally , for both experiments, the dispersion of the data had to be modelled by providing a dispersion formula in the glmmTMB function. This dispersion was modelled based on the distr actor contrast, the cuing condition and whether the chosen chip was correct or not (Dispersion formula = Contrast*Cuing condition*Cho sen chip). E arly flight di r e ctio n We tested for the effect of the cue on the initial flight direc tion of the bee s by analysing the bee trajec t ory direction r elative to the target at 1 and 5 cm from their take-off point. T o do so, we separated trials with the target on the right and left sides of the sc reen. Then, for each side group, we pooled all trials for each cuing condition and used the Rayleigh test from the “ circular ” R package (Agostinelli & Lund, 2024) to test whether each group’ s flight direction was signifi cantly oriented towa rd s 0 (pe rfe ctly fly ing towards the ta rget ). In addition, we used the Watson-Wheler test from the same package to test whether bee early flight direction diff ered between cuing conditions in each of the side gr oups. Dat a ex clus ion We excluded flights following visual examination of the t e st recor dings based on the following criteria: first, all trials where be es were seen crashing or landing and walking on the ar e na floor wer e excluded from all analyses (N=9 out of 335 (2.69%) for the detection task and N=1 out of 314 (0.32%) for the discrimination task). Additionally , some flights were excluded from the first appr oach durati on and the early flight direction analyses when the bees were seen turning back to look at the tunnel entrance before a pproaching the screen. We believe that the se sequences during which the bee faced the tunnel were learning flights and were not relevant to the targe t detection ta sk (detection task: N=62 out of 326 (19.02%), di scrimination task N=74 out of 313 (23.64%)). However , these flights were k ept for the first approach and final choices analyses presented here. The same analyses .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint of the bees’ choices were also run with these flig hts included and gave nearly identical results (not presented here). Additionally , in the analysis of early flight dir ection, we only included flights in which we could see the target appear before the bee took off . These videos had either the target and/or the distrac tor appear on screen or we could al t ernatively estimate the time of target appearance based on the cue (cue appear ance frame + 72 fr ames). For the detection task, in the uncued condition, it was often impossible to see the tar get if it was not at one of the two highest contrasts. Thus, we had t o exclude numerous flig hts from the analysis for this specific condi tion. Finally , we only included flights in which the crossing of the 1 cm or 5 cm threshold from the take-off point happe ned on a frame not excluded during our cleaning and reconstruction process of the trajectories. In total, we kept 145 out of 326 flights (44.48% ) for the 1 cm threshold and 146 out of 326 flights (44.79%) for the 5 cm threshold on the detection task. We also k ept 92 out of 239 flights (38.49%) for the 1 cm threshold and 108 out of 239 flights (47.16%) for the 5 cm threshold for the discrimination task. R esults Detection t as k Fi nal c hoice analy s i s The probability of a correct final choice during tests increa sed with an increa se in the target contras t in all three cuing conditions (Fig. 2A ; Uncued: Estimate±Standard Error=2.901±0.740, Z=3.920, p<0.001 ; T arget side cued: E stimate±Standard Error=4.457±0.977 , Z=4.561, p<0 .001 ; Opposite side cued: Estimate±S t andard Error=2.728±0.757, Z=3.605, p<0.001). However , compared to the uncued condition, the display of the cue on the target side (main effect: Estimate±Standard Err or=-0.096 ±0.561, Z= -0.171, p=0.864 ; interaction with contr a st: Estimate±S tandard Error=1.556±1.218, Z=1.278, p=0.201) or on the opposite side of the screen (main effect: Estimate±S tandard Error=0.335 ±0.550, Z=0.608, p=0.543 ; interaction with contrast: Estimate±Standard Error=-0.172±1.054, Z=- 0.164, p=0.870) did not affect the chances of correct choic es. There was also no significant difference between the probability of a correct final choice in the two cued condi tions (main eff ec t: Estimate±Standard Error=0.430±0.550, Z=0.782, p=0.434 ; interaction with contrast: Estimate±Standard Error=-1.729±1.232, Z=-1.403, p=0.161 ). Fi r st ap proac h a naly s i s As in the final choice analysis, an increase in target contr a st increased the probability that the bees approached the correct chip in the uncued condition (Fig. 2B ; E stimat e ±Standard Error=1.829±0.627, Z=2.917, p=0.004). Although a similar trend was observed in the two cued conditions, the effect of the target contrast was not significant (T a r get side cued: Estimate±Standard Error=0.856±0.592, Z=1.146, p=0.148 ; Opposite side cued: E stimate±Standard Error=1.099±0.605, Z=1.816, p=0.069). Despite this, when compa r ed to the uncued condition, we did not find a significant effect of the cue flashed on the side of the target (main effect: Estimate±Standard Error=0.611 ±0.530, Z=1.154, p=0.248 ; inter action with contrast: Estimate±Sta ndard Error=-0.973±0.861, Z=-1.130, p=0.258) or on the opposite side (main effect: Estimate±Standard Error=0.532 ±0.521, Z=1.004, p=0.315 ; interaction with contrast: Es timate±Standard Err or=-0.730±0.870, Z=-0.839, p=0.401) on the pr obability of correct first approach. In addition, the side of the cue did not have an effect on the probability of correct first approach (T arget side vs Opposite side, main eff ect: Es timate±Standard Err or=-0.080 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint ±0.520, Z= -0.154, p=0.878 ; interaction with contr a st: Estimate±S tandard Error=0.243±0.846, Z=0.287, p=0.774). Fi r st ap proac h d ura tion an alysis We first r an a model investigating the effect of contrast and cuing condition. We found a non- significant trend that the first approach duration decreased with the inc rease of the target contrast when the cue wa s not displayed (Fig. 3A ; Estimate±Standard Error= -0.182±0.191, Z=-0.950, p=0.342). However , the same effect was significant in both conditions when the c ue appeared (T a r get side cued: Estimate±Standard Error=-0.470±0.151, Z=-3.119, p=0.002 ; Opposite side cued: Estimate±Standard Error=-0.387±0.124, Z=-3.126, p=0.002 ). However , there wa s no significa nt eff ec t of the cue when comparing both cued conditions to the uncued one (T a r get side cued, main eff ect: Estimate±Standard Error=0.136±0.170, Z=0.800, p=0.424 ; interaction with contrast: Estimate±Standard Error=-0.289±0.240, Z=-1.202, p=0.229 ; Opposite side cued, main effect: Estimate±Standard Error=0.013±0.156, Z=0.080, p=0.936 ; interaction with contrast: Estimate±Standard Error=-0.206±0.229, Z=-0.898, p=0.369 ). Her e again, this absence of significant interaction tends to support the fact that overall, the cue did not have an eff ec t on the firs t approach duration but the more the targe t was visible, the faster the bees made their choice. There wa s also no diff erence between the two cued conditions (main effect: Estimate±S t andard Error=-0.124±0.127, Z=-0.972, p=0.331 ; interaction with contr a s t: Estimate±S tandard Error=0.083±0.197, Z=0.423, p=0.672). T o investigate the eff ects of choice accuracy , we ran a second model with co ntrast and choice accuracy (correct or incorrec t) a s independent variabl es. We found a significant interaction between the two variables: bees made fa ster first approaches with the increase of the target contrast when they approached the correct chip (Fig. 3B ; Estimate± Standard Error=-0.477±0.100, Z=-4.781, p<0.001) but not when they approached the wr ong one (E s timate± Standard Error=0.048±0.139, Z=0.344, p=0.731). This interaction indicated that although the first approach duration was similar between chips when the t arg et w as in visible, i.e. with a con tr as t = 0, (Est imat e±Standar d Err or=- 0.182±0.104, Z=-1.745, p=0.081), the effect of the target contrast diff ered depending on whether the animals were approaching the target chip or the other chip (Es timat e±Standard Error=0.525±0.172, Z=3.058, p=0.002). This may indicate that the bees were more hesitant when approaching a chip when the target was not discernible. E arly flight di r e ctio n Bee flights in the first 1 cm from their take-off poi nt were signi ficantly oriented towards the target in all three cuing conditions both when the target was on the right side of the screen (Fig. 4 ; Uncued: Rayleigh test=0.901, p<0 .001 ; T arget side cued: Rayleigh test=0.855, p<0.001 ; Opposite side cued: Rayleigh test=0.912, p<0 .001) and when it was on the left side of the screen (Uncued: Rayleigh test=0.923, p<0.001 ; T ar get side cue d: Rayleigh test=0.953, p<0.001 ; Opposite si de cued: Rayleigh test=0.960, p<0.001). The same was true for bee flights 5 cm from their take-off points when the target was on the right side of the screen (Fig. 5 ; Uncued: Rayleigh test=0.877, p <0.001 ; T arget side cued: Rayleigh test=0.859, p<0.001 ; Opposite side cued: Rayleigh test=0.876, p<0.001) or the left side of the scr een (Uncued: Rayleigh test=0.938, p<0.001 ; T arget side cued: Rayleigh t e st=0.922, p<0.001 ; Opposite side cued: Rayleigh test=0.939, p<0.001). The direction of trajectories at 1 cm from the take-off point did not significantly di ffer between cuing conditions when the target was p r esented on the right side of the sc reen (W = 4.817, df = 4, p-value = 0.307) or when it wa s presented on the other side (W = 5.239, df = 4, p-value = 0.264). The same .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint was true at 5 cm from the bees’ take-off point when the target was displayed on the right side of the screen (W = 2.188, df = 4, p-value = 0.701) or when it was shown on the left side of the screen (W = 0.654, df = 4, p-value = 0.957). Because our exclusion criteria were very conservative, we ran the same analyses for the early flight directions at 1 and 5cm without excluding any flights and the resul ts obtained were identical. Discrimination task Fi nal c hoice In the discrimination task, increa sing distractor contrast significantly decreased the probability that bees chose the target location in all cuing conditions (F ig. 6A ; Uncued: Estimate± Standard Error=- 5.339±1.737, Z=-3.074, p=0.002 ; T arget side cued: E s timate±Standard Error=-7.518±2.393, Z=-3.142, p=0.002 ; Opposite side cued: Estimate±S tandard Error=-7.493±1.989, Z=-3.768, p<0.001). This reflec ts the fact that increasing distractor contrast made the target and distrac tor less di fficult to discriminate. Cuing condition, however , did not influence the final choices of the bees when compared with the uncued condition (T arget side cued, main eff ec t: χ Estimate±Standard Error=1.372 ±2.440, Z=0.562, p=0.574; interaction with contrast: Estimate±Sta ndard Error=-2.179±2.956, Z=-0.737, p=0.461 ; Opposi t e side cue d, main effect: Estimate±Standard Error=1.772±2.288, Z=0.774, p=0.439 ; interaction with contrast: Estimate±Standar d Err or=-2.154±2.640, Z=-0.816, p=0.415). When comparing the two conditions where the cue was displayed, the effect of its positi on (on the side of the target or the side of the distractor) did not diff er either (main effect: Estimate±Standard Error=0.400±2.628, Z=0.152, p=0.879 ; interaction with contrast: Es timate±Standa r d Error=0.025±3.111, Z=0.008, p=0.994). Fi r st ap proac h Similar results were obtained for the probability of a first approach to the target. The bees were le ss likely t o first approach the target side as the distr actor contrast inc r eased in the three cuing conditions (Fig. 6B ; Uncued: Estimate±Standard Error=-2.536±0.909, Z=-2.791, p= 0.005 ; T arge t side cued: Estimate±S t andard Error=-2.327±0.955, Z=-2.437, p=0.015 ; O ppos ite side cued: Estimate±Standard Error=-3.354±1.090, Z=-3.076, p=0.002 ). Her e again, the cuing condition did not influence the probability of a first approach to the target when compared to the uncued condi tion (T arget side cued, main effect: Estimate±Standar d Error=-0.066±0.992, Z=-0.066, p=0.947; interaction with dis tractor contrast: Estimate±Standar d Err or=0 .209±1.318, Z=0.159, p=0.874 ; Opposi t e side cue d, main effect: Estimate±Standard Error=1.064±1.149, Z=0.926, p=0.355; interaction with dis tractor contrast: Es timate±Standar d Error=-0.818±1.419, Z=-0.576, p=0.564). first approach probabilities in the two conditions where the cue was displayed did not differ either (main eff ec t: Estimate±Standard Error=1.130±1.147, Z=0.985, p=0.325; interaction with distractor contrast: Estimate±Standard Error=-1.027±1.449, Z=-0.709, p=0.479 ). Fi r st ap proac h d ura tion .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Surprisingly , bee s took less time to approach a chip as the distrac t or contrast increased in the uncued condition (Fig. 7A; Estimate±Standard Error=-0.206±0.096, Z=-2.152, p=0.031). This is probably explained by the fact that, contrary to what we expec ted, the be es were more willing to approach either chip as the distractor resembled more the target. However , this effect of distractor contra s t was not observed in the two cued conditions (T arget side cued: Es timat e±Standard Error=- 0.062±0.092, Z=-0.669, p=0.504 ; Opposite side cued: Estimate±Standard Error=-0.041±0.059, Z=- 0.698, p=0.485). Despite these different of eff ects of the contrast in the two cued conditions and in the uncued condition, the cue did not have a significa nt effect on the firs t appr oach dur ati on compared to the uncued condition (T arget side cued, main effect: Estimate±S t andard Error=- 0.128±0.093, Z=-1.377, p=0.168 ; interaction with contrast: Estimate±S t andard Error=0.144±0.133, Z=1.090, p=0.276 ; Opposite side cued, main eff ect: Estimate±Standard Error= -0.157±0.084, Z=- 1.867, p=0.062 ; interaction with contrast: Estimate±Standard Error=0.165±0.114, Z=1.449, p=0.147). Finally , the side of the cue when it was displayed did not have an eff ect on the duration of the first approach (main effect: Es timate±Standard Err or=-0.029±0.073, Z=-0.393, p=0.694 ; int er action with contrast: Estimate±S tandard Error=0.020±0.110, Z=0.183, p=0.855). Contrary to our results in the detection task experiment, the distractor contrast did not significantly influence the first approach duration in trials wher e the bee made a correct choice (Fig. 7B ; Estimate±Standard Error=-0.087±0.048, Z=-1.825, p=0.068 ) but it did when the be e made a wrong choice (Es timat e±Standard Error=-0.327±0.084, Z=-3.914, p<0.001) for her first approach. This was due to the fact that when bees first approached the dis tractor ’ s location and if the distractor was not visible (Michelson contrast=0), they took more time than when they appr oached the target (Comparing the main effect of T arget side cued vs Opposite side cued conditions: Estimate±Standard Error=0.254±0.073, Z=3.484, p<0.001). However , an interaction with the contrast indicated that as the distractor became mor e visible, the bees approached i t faster (Estimate±S t andard Error=- 0.240±0.098, Z=-2.441, p=0.015). E arly flight di r e ctio n A t a distance of 1 cm away fr om their take-off point, bees flew significantly towards the target in every cuing condition both when the t a r get was on the right side of the screen (Fig. 8 ; Uncued: Rayleigh test=0.918, p<0 .001 ; T arget side cued: Rayleigh test=0.899, p<0.001 ; Opposite side cued: Rayleigh test=0.960, p<0 .001) and when it was on the left side of the screen (Uncued: Rayleigh test=0.821, p<0.001 ; T ar get side cue d: Rayleigh test=0.743, p<0.001 ; Opposite si de cued: Rayleigh test=0.798, p<0.001). The same was observed with the bees flig hts at 5 cm from their tak e-off points when the target was on the right side of the screen (Fig. 9 ; Uncued: Rayleigh test=0.936, p<0.001 ; T arget side cued: Rayleigh test=0.922, p< 0.001 ; Opposite side cued: Rayleigh test=0.927, p<0.001) and when the target wa s on the left side (Uncued: Rayleigh test=0.925, p<0.001 ; T arget side cued: Rayleigh test=0.925, p<0 .001 ; Opposite side cued: Rayleigh test=0 .900, p<0.001). Finally , early flight direction up t o 1 cm from the tak e off point showed that the distribution of these did not significantly differ across cuing conditions when the target was on the right side of the screen (W = 2.288, df = 4, p-value = 0.683) or when it was on the other side of the scree n (W = 3.473, df = 4, p-value = 0.482). This was also true for the flight direction at 5 cm from the tak e off point (target on the right: W = 5.469, df = 4, p-value = 0.243 ; target on the left: W = 2.446, df = 4, p-value = 0.654). .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint We ran the same analyses on the flight directions at 1 and 5 cm without excluding any flights to make sure that the results presented above were not only due to our strict exclusi on criteria. The analyses with the full dat a set were identi cal. Dis c us sion The main goal of our study was to test whether a bottom-up cue could influence contrast sensitivity in bumblebee s, a s previous s tudie s have shown in primates (Camer on et al., 2002; Carrasco et al., 2000, 2004; Ling & Carrasco, 2006; Z.-L. Lu & Dosher , 2000; Solomon et al., 1997). By briefly presenting a cue we att empted to capture bee attention and induce an increase of bumblebees’ contrast sensitivity in a localised a r ea of their visual field. Our expectation wa s that in a detection task, if the cue was presented on the side of a subsequent target, bees would be able to detect the t arg et at a lower c on tras t t han if the cue was pr es ent ed a t another loca tion or not f las hed a t all. Similarly , in a discrimination task where bees were presented a high contrast target and a variable contrast distractor , we expected the bees to have more difficulty distinguishing the two stimuli if the cue was presented on the side of a subsequent dis tract or . We found that the main variables influencing the choices of the bees (both for their first approaches and final probing) wa s the contrast of the target in the detection t a sk and the contrast of the distr actor in the discrimination ta sk. A previous study has shown that bumblebe es in a Y-maz e te s t can distinguish a sinusoidal grating of 0.09 cycles per degr ee (or 11.11°) at a Michelson contrast above 63.6% and a grating of 0.18 cycles per degree (or 5.56°) at contrasts above 81% (Chakr avarthi et al., 2016). Because our tar get and distractor subtended a visual angle of around 6.8° fr om the tunnel entrance, we expec ted that at the lowest contrasts, the stimuli would not be detectable by the bee s. Therefore, in the detection task , the bee s were more lik ely to make a random choice at these contrasts. Conversely , during the discrimination task, as the contra s t of the distractor increased it became less distingui shable from the full contrast t a r get and thus the di scrimination between the two became more difficult for the bees. T arget contrast also influenced first approach duration during the detection task. When the target was barely visible or al t ogether absent, bees took longer to first approach a chip compared to when the target was clearly visible. This result indicates that as a result of the successful training, bees wer e really looking for the targe t and we r e reluctant to approach a chip when they could not see one, even without a punishment for wrong choice. The contrast of the distractor in the discrimination task, however , had the opposit e effect on first appr oach duration. This wa s surprising because, as the contrast of the distrac tor increased, it was more simila r to the full contr a st target. Discriminating between the two the refore became more di fficult (as confirmed by the final choices and first approach probabilities of the bees). Thus, we could have expec ted bees to show the same increased approach duration as we saw in the detection task when the target was less visible, making the choice more difficult. Instead, the bees appear to simply choo se one chip (right or wrong) and fly directly towards it, even though a wrong choice was punished by the taste of quinine during the discrimination task training. This suggests that our bees did not really compare the target and distr actor contrasts but only e stimated whether the one they fi rst detec t e d was close enough to the full contrast and, if so, approached and landed on the chip bellow it. We expected early flight orientations to be random when the target was invisible in the detection task and to grow mor e oriented towards the tar get as its contrast increased. For the discrimination task, we expected bees to be oriented early on towards the target when the distract or was invi sible .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint and t o get progressively more randomly oriented a s the dis trac tor contrast increased a nd wa s less distinguishabl e from the target. Co ntrary t o these expectations, target contr a st did not influence the early flight orientation of bees in the detection task, and the same wa s true for distractor contr a st in the discrimination task. One theoretical explanation for this could be that the bees could not se e the target from the entr ance of the tunnel. However , our target subt ended a visual angle of approximately 6.8° at the entr ance and bumblebees can resolve achromatic sinusoidal gr ati ngs of 0.21 cycles per degree or 4.76° (Chakravarthi et al., 2016). Other r e sea r ch also shows that bumblebees can detect yellow targets sustaining angles between 3.4° and 7° depending on the size of the bee (Spaethe & Chittka, 2003), with only the smallest individuals needing target la r ger than 6°. Moreover , later studies have demonstrated in Y-maze experiments that bumblebees could de t ect similar yellow targets subtending an angle as small as 2.3° (Dye r et al., 2008) or 1.8° (Wertlen et al., 2008). Thus, it seems unlikely that the bees were unable t o r e solve the target at the tunnel entr ance. It is perhaps more lik ely that they made their decision at a later point when they wer e closer to the screen. More importantly , we didn’t see any effe ct of the cue on the bees’ ability to detec t the target or discriminate it from the distractor . One possibility is that the bees did not perceive the cue either due to its siz e or to its duration. However , this seems unlik ely . The mea sured irradiance of the cue against the backgr ound provide s a s trong chromatic and achromatic contrast (achromatic Michelson contrast=0.83 ). The cue also subtends an angle of around 9.7 ° at the tunnel entr a nce which bees would be able to resolve (Chakravarthi et al., 2016; Dyer et al., 2008; Spaethe & Chittka, 2003; Wertlen et al., 2008). It is also unlikely that the duration of the cue was too short for the bees to perceive it. The i ntegration time of Bombus terrestris’ blue photoreceptors is 9.7 ms (Sk orupski & Chittka, 2010), well below th e duration of our cue. In addition, thi s species of bumblebee was behaviourally shown t o detect blue bars flashed for as short a s 25 ms (Nityananda et al., 2014). Thus, the be es should have been able to per c eive our cue with a pr e sentation duration of 200 ms. Flashing cues have also been shown captur e insect attention (Sareen et al., 2011). Fruit flies were more lik ely to follow a vertical bar on a circular screen if i t flashed multiple times before to moving. Our cuing paradigm does differ from the one used in this experiment. Ther e , the target was identical to the cue and flashed repeatedly at 10 Hz, while in our case, the cue was presented only once for a duration of 200 ms rather than fla shing on and off . Our cue was also distinct from and did not spatially overlap with the target location to avoid a possible masking effect. Simila rly to Sa r een and colleagues, other expe riments have also used cues that are identical to the targe ts and showed attentional capture. Lancer et al . (2019) recorded from the CSTMD1 neuron of dragonflies in response to two targets simultaneously moving upwar d. They showed that the neuron had equal chances to selectively attend to either one of the tar gets. However , if one target appeared earlier than the second one both temporally and spatially on the screen, it was more lik ely to be attended by the neuron. Here again, the cue and the ta r get were identical. Finally , the bee s in our experiment wer e freely flying compared t o the tethered flies and dragonflies in the previous research. We were trying to better r ecreate classic spatial cuing experiments with our paradigm but this appears to not have bee n effective in capturing attention. If attentional processes in insec ts are comparable to those of vertebr ate s, the nature of the cue, its position relative to the target or the time between the cue and t a r get onset would be important to successfully capture the animal’ s attention (Fr anconeri et al., 2005; Franconeri & Simons, 2003; Fuller et al., 2009; S . Lu, 2006; Posne r & Cohen, 1984; Pratt & McAuli ff e, 2001; Steinman et al., 1997; T sal, 1983). Our results sug gest that our cue did not posse ss the required characteris tic s or duration to captur e bumblebees’ attention. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Since work er bumbl ebees search and forage on rewar ding flowers in their envir onment (Heinrich, 1976), it might b e more critical for them t o evolve a cognitive proc ess resembling vertebrate top- down attention. Such a process would allow them to tune their sensory system to enhance the detec tion of stimuli a ssociat ed with the most rewarding flowers in their environment such as their shape or colour (Liu, 2019). On the other hand, bumblebees are predated by birds at the entranc e of their nest (Goulson et al., 2018) or by wasps (Dukas, 2005) and cr ab spiders while f or aging on flowers (Dukas & Morse, 2003; Morse, 1986; Rodríguez-Gironés & Jiménez, 2019). Crab spiders are sit-and- wait predators that are camouflaged on flowers and strike at bumblebees after they land to forage. T o avoid being predated by these animals, it would be be neficial to have the ability to prioritise sudden change s in the environment (such as a pr edator attack). Having a fa ster reaction time a s a result would be esse ntial for the work ers’ survival. As the predation on work ers decrea ses bumblebee colony fitness (Goulson et al., 2018), with an e specially str ong effect of crab spiders’ predation (Cr e sswell, 2017), we should expect some form of bottom-up attention in bumblebees to help them evade attacks. Given our r e sults, investigating this would lik ely need different approaches to th e one we too k. Recently , studies have shown that fruit flies have dedicated neural pathways responding t o visual looming cues and their characteristics such as size and direction to generate fa st and directed non- ster e otypical escape behaviours, overriding other behaviours (Ache et al., 2019; Card & Dickinson, 2008; de Vries & Clandinin, 2012). Such a rapid and spontaneous analysis of a stimulus strongly resembles bottom-up attention. This sug gests that looming cues may be better suited to investigate similar attention-lik e pr ocesses in insects. Author Con t ributions VN obtained the supporting fundings. VN and TR designed the experiment. MC, CS and TR conducted the experiment. TR analysed the videos. TR ran the statistical analyse s and wrote the paper . VN E dit ed the manuscript. Acknowledg e men ts VN and TR are supported by a BBSRC David Phillips fellowship BB/S009760/1 t o VN. Decla r at i on of Int er e s ts The author s declar e no c ompeting int er ests . Re fe re n c e s Ache, J. M., Pol sky , J., Alghailani, S., Parekh, R., Breads, P ., Pe ek, M. Y ., Bock, D . D . , V on Reyn, C. R., & Card, G. M. (2019). Neural Basis for Looming Siz e and V elocity Encoding in the Drosophila Giant Fiber Escape Pathway . Curre nt Biol ogy , 29 (6), 1073-1081.e4. https://doi.org /10.1016/j.cub.2019.01.079 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Agostinelli, C., & Lund, U. (2024). ‘ circul ar ’: Circula r S t ati s tic s (p. 0.5-1) [Dataset]. https://doi.org /10.32614/CRAN.package.circular Barbot, A., Landy , M. S., & Carr a sco, M. (2011). Ex ogenous attention enhance s 2nd-order contrast sensi tivity . Vi s i o n R e search , 51 (9), 1086–1098. https://doi .org /10.1016/j. visres.2011.02.022 Bates, D., Mächler , M. , Bolk er , B., & Walker , S. (2015). Fitting Linear Mixed-Eff ec ts Models Using lme4. J ourn al o f S t a ti s tica l Softw a r e , 67 (1). https://doi.org /10.18637/jss. v067.i01 Behrmann, M., Geng, J. J., & Shomstein, S. (2004). Parie t al cortex and att e nti on. C urrent Opini on in Neuro biol o gy, 14 (2), 212–217. https://doi.org /10.1016/j.conb.2004.03.012 Bowling, J. T ., Friston, K. J., & Hopfinger , J. B. (2020). T op-down versus bottom-up attention differentially modulate frontal–parietal connectivity . Hu ma n Brai n Map p in g , 41 (4), 928–942. https://doi.org /10.1002/hbm.24850 Bowman, E. M., Brown, V . J., Kertzman, C., Schwarz, U ., & Robinson, D. L. (1993). Covert orienting of attention in macaques. I. Eff ects of behavioral context. J o urna l o f Ne urop h y s io l ogy , 70 (1), 431–443. https://doi.org /10.1152/jn.1993.70.1.431 Brooks, M., E., Kristensen, K., Benthem, K., J. ,van, Magnusson, A., Berg, C., W., Nielsen, A., Skaug, H., J., Mächler , M., & Bolk er , B., M. (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. Th e R Jo urn al , 9 (2), 378. https://doi.org /10.32614/RJ-2017-066 Busse, L., Katzner , S., & T r eue, S. (2008). T emporal dynamics of neuronal modulation during exogenous and endogenous shifts of visual attention in macaque area MT . P r oc ee di ng s o f t he Nati onal Ac adem y of Sc ienc es , 105(42), 16380–16385. https://doi.org /10.1073/pnas.0707369105 Cameron, E. L., T ai, J. C., & Carrasco, M. (2002). Covert attention a ff ects the psychometric function of contrast sensitivity . V is i on R es ea r c h , 42 (8), 949–967. https://doi.org /10.1016/S0042- 6989(02)00039-1 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Card, G., & Dickinson, M. H. (2008). Visually Mediated Motor Planning in the Escape Response of Drosophila. Cu r re nt Bio l ogy , 18 (17), 1300–1307. https://doi.org /10.1016/j.cub .2008.07.094 Carrasco, M. (2011). Visual attention: The past 25 years. Vi si on R e se arch , 51 (13), 1484–1525. https://doi.org /10.1016/j. visres.2011.04.012 Carrasco, M., Ling , S., & Read, S. (2004). A ttention alters appearance. Na ture Ne u ros c ienc e , 7 (3), 308–313. https://doi.org /10.1038/nn1194 Carrasco, M., Penpeci-T algar , C., & Ecks tein, M. (2000). Spatial covert att e ntion inc r ease s contrast sensi tivity across the CSF: Support for signal enhancement. V i s ion R es ea r c h , 40 (10–12), 1203–1215. https://doi.or g /10.1016/S0042-6989(00)00024-9 Chakravarthi, A., Baird, E., Dacke, M., & Kelber , A. (2016). Spatial Vision in Bombus terrestris. F rontie rs in Be havi or a l Ne uro s c ienc e , 10 . https://doi.org /10.3389/fnbeh.2016.00017 Cresswell, J. E. (2017). A demographic appr oach to evaluating the impact of stressors on bumble bee colonies. E cologi cal E nt om ology , 42 (2), 2 21–229. https://doi.or g /10.1111/een.12 376 de Vrie s, S . E. J., & Clandinin, T . R. (2012). Loom-Sensitive Neur ons Link Computation to Ac tion in the Dr os ophila Visual Syst em. C urrent Biol og y, 22 (5), 353–362. https://doi.org /10.1016/j.cub.2012.01.007 Dukas, R. (2005). Bumble Bee P r edators Reduce Pollinator Density And Plant Fitness. E cology , 86 (6), 1401–1406. https://doi.or g /10.1890/04-1663 Dukas, R., & Morse, D. H. (2003 ). Crab spiders affect flower visitation by bees. Oik o s , 101(1), 157– 163. https://doi .org /10.1034/j.1600-0706.2003.12143.x Dyer , A. G., Spaethe , J., & Prack, S. (2008). Comparative psychophysic s of bumblebee and honeybee colour discrimination and object detection. Jo urnal of Co mpar ati ve Ph y siolo gy A , 19 4(7), 617–627. https://doi.org /10.1007/s0035 9-008-0335-1 Fernández, A., Li, H.-H., & Carrasco, M. (2019). How ex ogenous spatial attenti on affects visual representation. Jour nal o f Visi on, 19 (11), 4. https://doi.org /10.1167 /19.11.4 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Fernández, A., Okun, S., & Carr a sco, M. (2022). Differential Eff ects of Endogenous and Exogenous A ttention on Sensory T uning. Th e Jour nal of N e u r o sci ence , 42 (7), 1316–1327. https://doi.org /10.1523/JNEUROSCI.0892-21.2021 Franconeri, S. L. , Hollingworth, A., & Simons, D. J. (2005). Do New Objec ts Capture Attention? Ps yc holo gic al S cie nce , 16 (4), 275–281. https://doi.or g /10.1111/j.0956-7976.2005.01528.x Franconeri, S. L. , & Simons, D. J. (2003 ). Moving and looming stimuli capture attention. Pe r c e p t i o n & Ps yc hoph y sic s , 65 (7), 999–1010. https://doi.org /10.3758/BF03194829 Fuller , S., Park, Y ., & Ca rrasco, M. (2009). Cue contrast modulat e s the eff ects of exogenous attention on appearance. Vi sion Re s e ar c h , 49 (14), 1825–1837. https://doi.org /10.1016/j. visres.2009.04.019 Gabay , S., Leibovich, T ., Ben-Simon, A., Henik, A., & Segev , R. (2013). Inhibition of return in the archer fish. N at ur e Comm uni cati on s , 4 (1), 1657. https://doi.org /10.1038/ncomms2644 Goulson, D., O’ Connor , S., & Park, K. J. (2018). The impacts of predators and par a sites on wild bumblebee colonies. E c o l ogi c a l E n t om o lo gy , 43 (2), 168–181. https://doi.org /10.1111/een.12482 Hein, E., Rolke, B., & Ulrich, R. (2006). Vi sual attention and temporal di scrimination: Diff erential eff ec ts of a utomatic and voluntary cueing. V i sual Co gni tio n , 13 (1), 29–50. https://doi.org /10.1080/13506280500143524 Heinrich, B. (1976). The Foraging Specializations of Individual Bumbleb ees. E c o l ogi c al Monogra ph s , 46 (2), 105–128 . https://doi.org /10 .2307/1942246 Henderson, J. M., & Macquistan, A. D. (1993). The spatial distribution of attention following an e x og enous cue. P e r c epti on & P s yc hoph y sic s , 53 (2), 221–230. https://doi.org /10.3758/BF03211732 Herrmann, K., Monta ser-Kouhsari, L., Carrasco, M. , & Heeger , D. J. (2010). When size matters: A ttention aff ects performance by co ntrast or response gain. N at ure Ne uro sci ence , 13 (12), 1554–1559. https://doi.or g /10.1038/nn.2669 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Jigo, M., & Carra sco, M. (2020). Differential impact of ex ogenous and endogenous attention on the contrast sensitivity function across eccentricity . J our nal o f Vi s i o n, 20 (6), 11. https://doi.org /10.1167/jov .20.6.11 Lancer , B. H., Evans, B. J. E., Fabian, J. M., O’ Carroll, D. C., & Wiederman, S. D. (2019). A T a r get- Detecting Visual Neuron in the Dragonfly Locks on to Selectively Attended T argets. Th e Jour nal o f N e u r o sci ence , 39 (43), 8497–8509. https://doi.org /10.1523/ JNEUROSCI.1431- 19.2019 Ling, S., & Carr a sco, M. (2006). Sustained and transient covert attention enhance the signal via different contrast response functions. Vi s ion R e search , 46(8–9), 1210–1220. https://doi.org /10.1016/j. visres.2005.05.008 Liu, T . (2019). Feature-based attention: Effects and control. C urrent Opi nio n i n P s ycholo g y , 29 , 187– 192. https://doi .org /10.1016/j.cops yc.2019.03.013 Lu, S. (2006). Cue Duration and Parvocel lular Guidance of Visual A ttention. P s y cho logic al Sc ienc e , 17 (2), 101–102 . https://doi.org /10 .1111/j.1467-9280.2005.01671.x Lu, Z.-L., & Dosher , B. A. (2000). Spatial attention: Different mechanisms for central and peripheral temporal precues? J ourn al o f Ex perime n tal Ps y chol o gy: Hum an P erce ptio n an d P er f or mance , 26 (5), 1534–1548. https://doi.org /10.1037/0096-1523.26.5.1534 Maunsell, J. H. R., & T reue, S. (2006). Feature-based att e ntion in visual cortex. T r ends in Neuro scien ces , 29 (6), 317–322 . https://doi.org /10.1016/j.tins.2006.04.001 McGillycuddy , M., Warton, D. I., Popovic, G., & Bolk er , B. M. (2025). Parsimoniously Fitting Lar ge Multivari ate Random E ffects in glmmTMB. Jo urn al of Stati stical Software , 112(1), 1–19. https://doi.org /10.18637/jss.v112.i01 Meyer , K. N., Du, F ., Parks, E., & Hopfinger , J. B. (2018). Ex ogenous vs. endogenous attention: Shifting the balance of fronto-parietal activity . Ne urop s ych ologia , 111, 307–316. https://doi.org /10.1016/j.neur ops ychologia.2018.02.006 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Morawetz, L., Svoboda, A., Spaethe , J., & Dyer , A. G. (2013). Blue colour pref erence in honeybees distr acts visual attenti on for learning closed shape s. J o ur n al of C om p ar a t i v e P h ys iol o g y A , 199(10), 817–827. https://doi.org /10.1007/s00359-013-0843-5 Morse, D. H. (1986). Foraging Behavior of Crab Spiders (Mi sumena vati a) Hunting on Inflorescences of Different Quality . Americ an Mi dlan d Na t ur ali s t , 116 (2), 341. https://doi.org /10.2307/2425742 Nath, T ., Mathis, A., Chen, A. C., Patel, A., Bethge, M., & Mathis, M. W. (2019). Using DeepLabCut for 3D markerless pose estimation across species and behaviors. N ature Pro t ocol s , 14 (7), 2152– 2176. https://doi.org /10.1038/s41596-019-0176-0 Nityananda, V . (2016). A ttention-lik e proc esses in insects. Proc eedi n gs of t h e R oy al Soc iety B: Biol ogica l Sc ienc es , 283(1842), 20161986. https://doi.org /10.1098 /rspb.2016.1986 Nityananda, V ., Chittka, L., & Skorupski, P . (2014). Can Bees See at a Glance? Jo urn al of E xperime nt al Biol ogy , jeb.101394. https://doi.org /10.1242/jeb.101394 Posner , M. I. (1980). Orienting of A ttenti on. Qu art e r ly Jour nal o f E xperime nt al P s yc hology , 32 (1), 3– 25. https://doi.or g /10.1080/00335558008248231 Posner , M. I., & Cohen, Y . (1984). Components of visual orienting. A t t e n t i on and P erf orma nce X: Control o f Lang ua ge Pr oc es s es , 32 , 531–556. Pratt, J., & McAuli ffe, J. (2001). The eff ects of onsets and offsets on visual attention. P s yc hologi cal Re s e a r c h , 65 (3), 185–191. https://doi.org /10.1007/s004260100058 Ques t, M., Rinnert, P ., Hahner , L., & Nieder , A. (2022). Exogenous and endogenous spatial attention in crows. Pr o cee di ng s o f the Nat ion al Acad emy of S cie nces , 11 9(49), e2205515119 . https://doi.org /10.1073/pnas.2205515119 Rodríguez-Gironés, M. A., & Jiménez, O. M. (2019). Encounters with predators fail to trigger predator avoidance in bumblebee s, Bombus terrestris (Hymenopt era: Apidae). Biol ogic al Jo urnal o f the L inn ea n Soc iety , 128(4), 901–908. ht tps://doi.o r g /10.1093/biolinnean/blz155 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Sar e en, P ., Wolf , R., & Heisenber g , M. (2011). A ttracting the attention of a fly . Pr oce edin g s of the Nati onal Ac adem y of Sc ienc es , 108(17), 7230–7235. https://doi.org /10.1073/pnas.1102522108 Scolari, M., Ester , E. F ., & Serences, J. T . (2014). Fea ture - a nd Obje ct -Ba se d A tt e ntio nal M o dul atio n i n t h e H um an V is ual S ys t e m (A. C. (Kia) Nobre & S. Kas tner , Eds.; V ol. 1). Oxford University Press. ht tps://doi.org /10.1093/oxfordhb/9780199675111.013.009 Shimp, C. P ., & Friedrich, F . J. (1993). Behavior al and computational models of spatial attention. Jour nal o f Expe r im e nt al P s yc hology : A ni mal Beh av ior Proces se s , 19 (1), 26–37. https://doi.org /10.1037/0097-7403.19.1.26 Skorupski, P ., & Chittka, L. (2010). Differe nces in Photoreceptor Proce ssing Speed for Chromatic and Achromatic Vision in the Bumblebee, B om bus t e r r es t r is. T he J ou r na l of N e ur os ci en c e , 30 (11), 3896–3903. https://doi.or g /10.1523/JNEUROSCI.5700-09.2010 Solomon, J. A., Lavie, N., & Morgan, M. J. (1997). Contrast discrimination function: Spatial cuing eff ec ts. J ourn al o f th e O ptica l Soc iety o f America A , 14 (9), 2443. https://doi.org /10.1364/JOSAA.14.002443 Spaethe, J., & Chittka, L. (2003). Interindividual variation of eye optics and single object resolution in bumblebees. Jo urn al of E xperim e ntal Bio l ogy , 20 6 (19), 3447–3453. https://doi.org /10.1242/jeb.00570 Sridhar an, D., Ramamurthy , D. L., Schwarz, J. S., & Knudsen, E. I. (2014). Visuospatial selective attention in chick ens. Pro cee ding s o f t he Nati onal A c ademy of S cie nces , 111 (19). https://doi.org /10.1073/pnas.1316824111 Steinman, B. A., Steinman, S. B., & Lehmkuhle, S. (1997). Research Note T r ansient Visual Att e ntion is Dominated by the Magnocellular S tream. Visi on R e se arc h , 37 (1), 17–23. https://doi.org /10.1016/S0042-6989(96)00151-4 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint T sal , Y . (1983). Movement of attention across the visual field. Jour nal o f Expe r im e nt al P s yc hology : Huma n Perceptio n an d P e rf o r m an ce , 9 (4), 523–530. https://doi.org /10.1037 /009 6- 1523.9.4.523 Wang, F ., Chen, M., Y an, Y ., Zhaoping, L., & Li, W . (2015). Modulation of Neuronal Responses by Exogenou s Attention in Maca que Primary Visual Cortex. The Jo urn al of N e uro s c ie nc e , 35 (39), 13419–13429. https://doi.org /10.1523/JNEUROSCI.0527-15.2015 Wertlen, A. M., Nig gebrügge, C., V orobyev , M., & Hempel De Ibarr a, N. (2008). Detection of patches of coloured discs by bees. Jo urnal o f Exp e r im ental Bio lo gy, 21 1 (13), 2101–2104 . https://doi.org /10.1242/jeb.014571 Wiederman, S. D., & O’Carroll, D. C. (2013). Selective Attention in an Insect Vi sual Neuron. Curre nt Biol ogy , 23 (2), 156–161. https://doi.org /10.1016/j.cub.2012 .11.048 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 1: Illustration of the Detection and Discrimination experiments. A) Representations of the two possible target locations during the training phase of the detection experiment. The target was always a full contrast black circle placed either on the left or the right of the computer screen. B) Representation of a display sequence on the computer screen during a test trial of the detection task. After the bee entered the experimental arena, a blue square cue could be flashed for 200 ms on the right side or the left side of the screen or not flashed at all. After the cue disappeared, a target was presented on either side of the screen with one of 5 possible contrasts. C) Representations of the two training stages of the discrimination experiment. In stage 1, the target was always showed with a full contrast, either on the left or the right side of the screen, and the distractor was showed with a 0.448 contrast on the opposite side. Once the bee met the learning criterion, it moved to the second training stage during which the target was also showed with a full contrast, but the distractor could take one of 5 lower contrasts. D) Representation of a display sequence on the computer screen during a test trial of the discrimination task. The cuing sequence was identical to that in the detection task. During target presentation, the target was always presented at full contrast on one or the other side of the screen. On the opposite side of the screen, a distractor was displayed with one of 6 possible contrasts equal or lower to the target one. E) Top-down view of an example first approach trajectory recorded during a detection task test trial. The two red filled circles represent the two chips above which the target could appear. The dotted red circles around them represent the zones which we considered as approach zones (10 cm from the chip). The short red line shows the position of the tunnel entrance to the experimental arena. Trajectory sections up to 1 and 5 cm from the take-off point are marked in green and blue respectively. F) An example first approach trajectory recorded during a discrimination task test trial. Details as in E). G) Picture of an individually tagged bee drinking on one of the transparent chips on top of a transparent cup as used in our experiments. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 2: Detection task: Effect of target contrast and cuing condition on the probability of choosing the target side. A) Mean (±S.E.) proportion of trials in which bees chose target chips as a function of the Michelson contrast of the target. B) Mean (±S.E.) proportion of trials in which the bee first approached the target chip at a distance less than 10 cm. Blue curves represent trials where the cue was presented on the same side as the target. Yellow curves represent trials where the cue was presented on the opposite side to the target. Green curves represent trials without a cue. The schematic on the right shows a top view of a transparent chip. A final choice was when the bee probed the well at the centre of the chip with her antennae or her proboscis. The red dashed circle represents the 10 cm radius from the centre of the chip. A first approach was when the bee crossed the 10 cm radius around one of the two chips for the first time. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 3: Detection task: Effect of target contrast on bee first approach duration. A) Mean (±S.E.) time taken by the bees to first approach either chip at less than 10 cm for each cuing condition. The blue curve represents trials where the cue was presented on the same side as the target. The yellow curve represents trials where the cue was presented on the opposite side to the target. The green curve represents trials without a cue. B) Mean (±S.E.) time taken by the bees to first approach either chip at less than 10 cm for trials where bees made a correct (purple) or incorrect (turquoise) first approach. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 4: Detection task: Effect of the cuing condition on flight direction relative to the target (σ) at 1 cm from the take-off point. The top row shows trials with the target displayed on the right side of the screen and the bottom row shows trials with the target displayed on the left. A and D represent the uncued condition, B and E present data for trials with the target side cued and C and F show trials with the opposite side cued. Black dots represent individual trials, and red dots show the mean direction for each condition. 0 indicates the direction of the target. A clockwise rotation shows deviation towards the right of the target and counterclockwise indicates a deviation to the left of the target. The drawing on the right shows a schematic representation of the angle σ. It represents the two chips placed in front of the screen with the target displayed on the right side. The cross shows the bee’s take-off point, and the black line shows its trajectory. The red circle represents the early flight radius (either 1 or 5 cm in our experiment). The dashed blue lines form the angle between the place where the bee crossed the early flight radius and the direction of the correct chip. This angle therefore represents the bee’s early flight direction relative to the chip below the target. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 5: Detection task: Effect of the cuing condition on flight direction relative to the target (σ) at 5 cm from the take-off point. The top row shows trials with the target displayed on the right side of the screen and the bottom row shows trials with the target displayed on the left. A and D represent the uncued condition, B and E present data for trials with the target side cued and C and F show trials with the opposite side cued. Black dots represent individual trials, and red dots show the mean direction for each condition. 0 indicates the direction of the target. A clockwise rotation shows deviation towards the right of the target and counterclockwise indicates a deviation to the left of the target. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 6: Discrimination task: Effect of distractor contrast and cuing condition on bee choices. A) Mean (±S.E.) proportion of trials in which bees chose the target chip as a function of the Michelson contrast of the distractor. B) Mean (±S.E.) proportion of trials in which bees first approached the target chip at less than 10 cm. Blue curves represent trials where the cue was presented on the same side as the target. Yellow curves represent trials where the cue was presented on the opposite side to the target. Green curves represent trials without a cue. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 7: Discrimination task: Effect of distractor contrast on bee first approach duration. A) Mean (±S.E.) time taken by the bees to first approach either chip at less than 10 cm for each cuing condition. The blue curve represents trials where the cue was presented on the same side as the target. The yellow curve represents trials where the cue was presented on the opposite side to the target. The green curve represents trials without a cue. B) Mean (±S.E.) time taken by the bees to first approach either chip at less than 10 cm for trials where bees made a correct (purple) or incorrect (turquoise) first approach. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 8: Discrimination Task: Effect of the cuing condition on flight direction relative to the target (σ) at 1 cm from the take-off point. The top row shows trials with the target displayed on the right side of the screen and the bottom row shows trials with the target displayed on the left. A and D represent the uncued condition, B and E present data for trials with the target side cued and C and F show trials with the distractor side cued. Black dots represent individual trials, and red dots show the mean direction for each condition. 0 indicates the direction of the target. A clockwise rotation shows deviation towards the right of the target and counterclockwise indicates a deviation to the left of the target. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint Figure 9: Discrimination Task: Effect of the cuing condition on flight direction relative to the target (σ) at 5 cm from the take-off point. The top row shows trials with the target displayed on the right side of the screen and the bottom row shows trials with the target displayed on the left. A and D represent the uncued condition, B and E present data for trials with the target side cued and C and F show trials with the distractor side cued. Black dots represent individual trials, and red dots show the mean direction for each condition. 0 indicates the direction of the target. A clockwise rotation shows deviation towards the right of the target and counterclockwise indicates a deviation to the left of the target. .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.23.650250doi: bioRxiv preprint