Distracted foragers: competitors impair foraging ef ﬁ ciency, accuracy and speed for eastern grey squirrels

Social foraging can be bene ﬁ cial, but it can also disrupt optimal foraging strategies. Animals possess limited capacity for attention allocation and distractions may impact their rates of food acquisition. Attention to conspeci ﬁ c food competitors can prevent kleptoparasitism but likely affects food intake in multiple ways. We used two ﬁ eld experiments to assess the impact of nearby competitors on urban eastern grey squirrel, Sciurus carolinensis , foraging in a forest setting on the University of Toronto Scar-borough campus. We used a tube apparatus in the ﬁ rst experiment to examine foraging ef ﬁ ciency and a tipping container apparatus in the second experiment to examine foraging speed (handling time) and accuracy. Competing squirrels within 1 m or having just been involved in chasing or aggressive physical contact with competitors led to decreases in foraging ef ﬁ ciency, speed and accuracy. Competitors within 5 m of focal squirrels did not impact foraging measures. Over the course of the experiments, squirrels adapted to solving foraging apparatuses in the presence of competitor(s) and improved their ef ﬁ ciency and accuracy but not their speed. Vigilance rates were greater for squirrels that had competitors nearby; thus, decreases in foraging ef ﬁ ciency, accuracy and speed seemed to be caused by the need to allocate attention to conspeci ﬁ cs. These experiments show that social foraging impairs optimal food acquisition for individual eastern grey squirrels. © 2024 The Author(s). Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour. This is an open access article under the CC BY license (http://creativecommons.org/licenses/ by/4.0/). Optimal foraging theory proposes that natural selection favours behaviours that

Optimal foraging theory proposes that natural selection favours behaviours that solve foraging problems the 'best' (Stephens & Krebs, 1986).Foraging with conspecifics can lead to substantial increases in foraging efficiency in some species and contexts (Alexander, 1974;Barta & Giraldeau, 2001;Giraldeau & Caraco, 2000).However, optimal foraging behaviours can also be disrupted and altered by conspecifics (Giraldeau & Caraco, 2000).Indeed, competitors are one of several sources of distraction to foragers with a limited capacity for attention allocation (Bushnell, 1998;Dall et al., 1999;Dukas, 2002Dukas, , 2004)).Distraction occurs whenever attention is divided and an individual has difficulty focusing on a stimulus of interest due to alternative stimuli that occupy processing power (Allen et al., 2021;Chan et al., 2010).While optimal foraging requires attending to food acquisition and ignoring other cues, animals must, at minimum, also remain vigilant for predators and competitors (Beauchamp, 2015;Clark & Dukas, 2003).
We used two separate foraging experiments on an urban population of eastern grey squirrels, Sciurus carolinensis, to assess the impact of nearby food competitors on foraging efficiency, accuracy and speed (i.e.handling time).Grey squirrels are excellent subjects to use for foraging experiments, as they are known problem solvers, both in their anecdotal abilities to access food items in challenging garden bird feeders and in their success under other experimental conditions (e.g.Chow et al., 2016Chow et al., , 2019)).Eastern grey squirrels are obligate scatter hoarders, caching acorns and nuts in many locations throughout their home range to retrieve later and sustain them through the winter, when they do not hibernate (Robin & Jacobs, 2022).Scatter hoarding is suggested to be a cognitively challenging activity as individuals require episodic memory to recall where they cached different types of food items (Yi et al., 2016(Yi et al., , 2021)), and avoiding the pilfering of others may select for evasive tactics (Leaver et al., 2007;Lichti et al., 2017;Steele et al., 2014).Importantly, eastern grey squirrels are sensitive to a conspecific audience in the context of caching.Individuals of this species have been observed to modify their cache spacing and body orientation to hide the location of caches from competitors (Leaver et al., 2007), cache preferred foods in areas of greater predation risk to decrease pilfering (Lichti et al., 2017;Steele et al., 2014Steele et al., , 2015)), disguise their caches and caching behaviour when conspecifics are present (Hopewell & Leaver, 2008), increase their number of revisits to a patch to cache more when a greater number of conspecifics are present and cache farther away when food is scarce (Hopewell et al., 2008).
We expand on this previous work by assessing how nearby competitors impact the ability of eastern grey squirrels to focus on food-handling tasks and whether attention allocation to competitors impacts foraging efficiency, speed or accuracy.In the first experiment, we tested how having potential competitors within 1 m or having just been involved in chasing or aggressive physical contact with competitors impacted individuals' foraging efficiency.To assess foraging efficiency, squirrels were provided a tube apparatus baited with a reward and behaviours were captured by video camera traps.We also assessed vigilance with and without competitors present in this experiment.In the second experiment, we examined how competitor number and distance affected squirrel foraging speed (i.e.handling time) and accuracy.In this experiment, squirrels were provided a tipping container apparatus baited with a reward and behaviours were recorded manually to allow greater observer sight lines than were available with camera traps.Crucially, because experiments were carried out at apparatuses where only a single squirrel could forage, they elicited contest competition if nearby squirrels were also trying to acquire the resources.Contest competition occurs when there are direct displacements or aggression over food resources (Nicholson, 1954) and dominance hierarchies determine priority of access (Janson & van Schaik, 1988).Based on social foraging theory (Giraldeau & Caraco, 2000) and the cognitive constraints of limited attention capacity (Clark & Dukas, 2003;Dukas, 2002Dukas, , 2004)), we hypothesized that the need to attend to food competitors instead of focusing on a foraging task would affect the performance of squirrels.We thus predicted that squirrels would be less efficient, slower and less accurate when competitor squirrels were nearby compared to when they were not present.

METHODS
We ran two field experiments at a single location in the forest on the western edge of the University of Toronto Scarborough campus (43 46 0 57 00 N, 79 11 0 14 00 W).In each experiment, urban eastern grey squirrels interacted with a baited apparatus that was secured to a tree.The subjects were continually motivated to interact with our apparatuses because they could feed on the rewards we provided if they were hungry or cache the rewards if they were satiated.The subjects did not have identification marks and all were of the melanistic morph that is commonly seen in this species in urban environments (Cosentino et al., 2023), so we could not reliably identify them.Our analyses, therefore, are necessarily at the population level.However, our large sample size of trials should mediate idiosyncratic variation in foraging abilities.Eastern grey squirrel home ranges overlap considerably (Riege, 1981) and multiple squirrels were often at or near our apparatuses at the same time.Given the number of squirrels that we saw simultaneously in the vicinity of the experiments, we estimate that there were approximately eight adult squirrels that participated in the first experiment and six that participated in the second.Lack of individual recognition also meant that we could not deduce the dominance hierarchy of the subject population.Our experiments occurred in late summereearly autumn, when the seasonal effects of increased foraging expansion by young squirrels should have been complete (Thompson, 1977), and thus foraging competition by new individuals entering the range should have been limited.This should mean that our squirrels were experiencing regular contact with a limited number of conspecifics and were part of a wellestablished dominance hierarchy (Pack et al., 1967;Thompson, 1977).

Experiment 1
Experiment 1 ran from 2 September to 4 November 2021.All squirrel behaviours at the apparatus were recorded with a video camera trap that was triggered by motion.Behaviours were later coded from the videos.

Apparatus
K.F.built an apparatus modelled off the 'Ape-plunk' used in Jacobson and Hopper (2019) (Fig. 1).The device consisted of a clear PVC tube (60.96 cm in length, 5 cm in diameter) covered in wire mesh (0.5 Â 0.5 cm) on the outside to protect the apparatus from chewing squirrels and to allow the subjects to crawl on and interact with the tube with greater ease.Eight bisecting slots were drilled perpendicular to the tube, with four on the top and four on the bottom (spaced 5 cm apart in each section) and a wider space in the middle (10 cm) where several peanuts were placed as a food reward.For each slot, a flat 13 cm long and 2 cm wide, smoothfinish metal slat was crafted to slide through and block the progress of the peanuts moving down the tube.Subjects had to remove these metal slats when the reward was on top of them to allow the nuts to fall out of the tube for consumption or caching.The squirrels almost always used their feet to cling to the tree or the apparatus and pulled the slats with their mouth (Fig. 1b, Videos S1eS2).The apparatus was attached to a tree with two tension cables (2.5 cm width).The tube sat 10 cm above a flat wooden platform (60 Â 15 cm) attached to the tree (Fig. 1c), 138 cm from the ground.The platform usually caught the peanuts that fell while the squirrels were interacting with the apparatus, and it also provided a resting spot for subjects.

Procedure
Before a trial, the experimenter inserted the lower slats and baited the tube with peanuts in shells so that they would rest on top of the bisecting slats.The number of lower slats inserted before baiting depended on the phase of the experiment.The remaining slats were inserted above the peanuts and the tube was capped closed so the subjects would not try to access the food reward from the top.A trial started once the tube was filled and capped and a squirrel contacted the apparatus.The subject could then remove the slats and attempt to get the reward.A trial ended once all the peanuts fell, or in a few trials, when the remaining peanuts were stuck due to their shape/size.Initially, three preliminary training trials were run on the same day where fewer slats were inserted under the peanuts to ensure that the squirrels could learn how to remove the slats and get the food reward.The squirrels figured out how to remove the lower slats on the first trial and did so successfully on each training trial, so we immediately proceeded to our two test phases for this experiment.
In the first phase of experiment 1 (2 September e 8 October 2021), the food reward was placed in the centre of the tube with four slats above it and four below it (Fig. 1a and b; Jacobson & Hopper, 2019).Phase 1 ran until there were 100 completed trials recorded (mean: 5.76/day, range 1e20), then the study shifted to Phase 2, where the food reward was moved up the tube between the first and the second metal slat (Fig. 1c).This required the subjects to remove seven slats (or in a few trials, six slats, due to loss of one of the slats on the forest floor) to access the food reward.Phase 2 ran between 8 October and 4 November 2021.As the placement of the food reward had shifted, we ran two further preliminary training trials before collecting data in this phase.Only two preliminary trials were conducted, as the subjects could solve the new problem with ease.In total, 30 trials were conducted in Phase 2 (mean: 3.44/day, range 1e6).
The squirrels had access to the reward-filled tube no more than 5 days a week.Some trials were done with the experimenter (E.N.V. or J.A.T.) present, who could rebait the tube and run several successive trials.Other trials were done with experimenter(s) absent; when the tube was baited and left for the squirrels.In all trials, squirrel behaviour and manipulation of the apparatus was captured by a Browning Strike Force APEX motion-triggered trail camera attached to a tree approximately 2.5 m away from the apparatus (Appendix, Fig. A1).

Data coding
In all trials for experiment 1, data were coded from the camera trap videos of the squirrels completing the task.All data were transcribed into Microsoft Excel.The camera trap was triggered by motion and kept recording video if there was continuous motion in the visual field.We set the camera trap to keep recording for 10 s after motion stopped.Most trials consisted of several video clips (range 1e11) because the squirrels tended to remove some slats, which let some of the peanuts fall, that they then cached before trying to get the rest of the nuts.During competitive trials, multiple squirrels were present, which meant that there was sometimes a rotation in which subjects were accessing the tube.Thus, we coded each video clip separately and assigned each slat in the tube with an identifying number based on its position.Due to multiple squirrels being present and multiple video clips for each trial, we were not able to accurately assess the exact order that the slats were pulled by a single squirrel to complete each trial.
For each video clip, we recorded the experiment phase, the date and time recorded by the camera trap and the following variables (defined below): slats pulled, slats removed, slats left, success, efficiency and whether or not there was competition.'Slats pulled' were the identifying numbers for the slats that were manipulated, whether they were removed fully from the tube or not.Squirrels often pulled the slats halfway out and then used them as a lever with further manipulation to make the peanuts fall.Each manipulation in these sequences was recorded.'Slats removed' marked the slats that were completely removed from the apparatus and that had fallen to the platform or the ground.'Slats left' indicated whichever slats were still in the tube at the end of the video clip.'Success' was coded as a binary variable (whether or not the squirrel acquired some reward).We coded subjects as 'efficient' when they removed or manipulated only the slats situated below the food reward and coded subjects as 'inefficient' when they pulled or manipulated the slats above the reward or a combination of slats above and below the food reward.'Competition' was scored as absent if no other squirrel was seen on the video and as present if other individuals were observed.If other squirrels were present, we counted the number of competitors observed.However, because the videos gave a limited view around the apparatus, it was often not possible to know the exact number of other squirrels in competition with subjects.So, in our models, competition was scored as present or absent.
For a subset of the video clips (N ¼ 130), we also coded subjects' head turns to determine whether vigilance rates were greater when competing squirrels were present.We recorded a head turn each time a subject completing the experiment turned its head away from the task at an angle greater than approximately 45 .We also timed the number of seconds the subject's head was visible in the clip and noted the presence or absence of competitors.Two observers (E.N.V. and J.A.T.) coded trials from videos.Interobserver Removal of these slats is inefficient and has no causal effect All metal slats could be manipulated, but the subjects only needed to remove the ones below the reward to access the food (Videos S1eS2).(c) A photo of the apparatus in its Phase 2 configuration.In Phase 2, there were six to seven metal slats under the food reward that had to be removed, with only one redundant slat above the food reward.
reliability tests on scoring efficiency were assessed from a total of 33 trials and the interobserver reliability coefficient (R) was calculated at 96%.

Experiment 2
Our second foraging experiment allowed us to better quantify (1) the number of competitors around focal squirrels and their distance, (2) any chasing or aggression that occurred away from the apparatus, (3) the exact timing of squirrels' efforts to extract the food reward and (4) the number of mistakes they made.The camera trap was useful in allowing us to get a high number of trials for experiment 1, but the limited frame of view only allowed us to determine whether competitors were very close (i.e. on the tree near the squirrel working on the tube) or had just been involved in chasing or aggressive physical contact with the squirrel at the apparatus (Video S2).In addition, while we could also assess squirrel efficiency on the first apparatus, the task was timeconsuming and the cutting out of the video clips did not allow us to determine whether there was a time cost to foraging while competitors were nearby.We therefore switched the apparatus for the second experiment to one that the squirrels could solve quickly and where they could only get a single nut as a reward.Additionally, rather than using a camera trap exclusively to record data, we manually recorded the number and distance of competitors and assessed each focal squirrel's handling time at the apparatus.

Apparatus
We modified an interactive dog puzzle (Trixie 32019 Mad Scientist for Dogs, Level 2, White/Blue, purchased on Amazon, https:// www.amazon.ca/Trixie-32019-Scientist-Level-White/dp/B003TOKTEG/ref¼sr_1_50?crid¼4CE22LLNNUED&dib¼eyJ2IjoiMSJ9.bRb7PgNE6ePNWb1aPQhc_iYnuzQRmrWbVxPQwQyxAEiqw-CXO RlSSaarOUEs8gR5ojTLXKcye2xpl3VAMguSBORlxfAClapys4PP wigNNl3iZ4dcHFCtbSbRHEDiS8ewOawto6aHXXLTFuVBm6R1-_E CEcqRZlzE2PnOW-OunOYpzuZ2Q3mvGPBMmfkVk32nwQqiGpO WHbaJwA4d4mzoaryckw0pC9nTvuqPZtd5MQ5CIEwdZyWiAkpc EhiP7xKjWOKdQAgQ5ud8kmmGI7mRWihITNqSfRDIqfUgwxcojfs.8efAoWhen8bi5RaMVqZ1-zjW-r50-zNkqdMa1KVG5ao&dib_tag¼ se&keywords¼trixie%2Bdog%2Bfeeding%2Btoy&qid¼1720121 289&sprefix¼trixie%2Bdog%2Bfeeding%2Btoy%2Caps%2C94&sr¼8-50&th¼1) for this experiment as a feeder (Fig. 2a).The feeder had three rotatory containers, side by side, each of which could be filled with treats, where animals must learn to rotate the filled containers to allow rewards to fall out.We removed the caps for our purposes and painted the middle container white, to distinguish it from the other two blue containers (Fig. 2b).We continually used this middle, white container as rewarding and baited it with a single hazelnut for each trial, as we wanted squirrels to have clear cue to which container was rewarding.Like most mammals, eastern grey squirrels are dichromats, and can easily differentiate blue from white (Silver, 1976).We attached the apparatus to the same tree as experiment 1 by resting it on the same wooden platform and securing it to the tree with a rope.We ran eight preliminary trials to habituate the squirrels to the apparatus and to practise our data collection techniques.

Procedure
Data were collected over 11 observation days from 22 September to 9 November 2022.During experiments, the observer (H.G.) recorded data manually and the apparatus was brought into the laboratory each afternoon after data collection.In each trial, the observer loaded the middle container with one hazelnut and backed off to about 3 m, standing slightly behind the tree that had the camera trap on it (Appendix, Fig. A1), to allow the squirrels to approach and attempt to collect the nut.At this distance and location, the presence of the observer did not seem to have any effect on the squirrel's behaviour.After loading the nut, H.G. made a sound cue (a whistle) to indicate to the squirrels that they could approach for a trial.Squirrels typically crawled around or straight down the tree and attempted to flip the container(s) to get the nut to fall out (Videos S3eS4); once the nut fell, they collected it from either the platform (Fig. 2c) or the ground and ate or cached it.Once the squirrel had moved away, the observer could approach the platform again and bait the apparatus for another trial, waiting for the same squirrel or a different individual to approach and participate.We recorded data from a total of 140 trials for this experiment (mean: 12.73/day, range 3e20).On every trial, the observer recorded the time it took for the squirrel to collect the nut, mistakes made by the squirrel and any competition.Handling time was recorded using a stopwatch.The watch was started as soon as a squirrel touched the apparatus and was stopped as soon as the nut fell from the container.A 'mistake' was recorded if the squirrel attempted to tip either of the two side containers, which were not the baited, before focusing on the correct (middle) container, which was baited.Competition was recorded as the number of other squirrels within 1 m and/or within 5 m of the platform.We also noted whether the focal squirrel had been involved in a chasing bout with a competitor squirrel just prior to the trial, even if that competitor was no longer within 5 m of the apparatus when the squirrel completed the task.Chasing sometimes led to aggressive physical contact in both experiments (Video S2), but this was minor and no injuries resulted.The camera trap used in experiment 1 was also set up to record trials in experiment 2, and videos were used to double check handling time and the number of mistakes whenever the experimenter was unsure of the accuracy of what was recorded in the field.Vigilance rates could not be accurately determined for videos from experiment 2 because the squirrels completed the task quickly and video clips were too short.

Data Analyses
We ran all analyses in R version 4.0.3(R Core Team, 2020) and an alpha value equal to or less than 0.05 was considered significant.We assessed all models for zero inflation and under-or overdispersion with the 'DHARMa' package (Hartig, 2021).We calculated vigilance rates during experiment 1 as the number of head turns/s squirrels made away from the apparatus that were visible in the video clips.We used a linear mixed effects model (LMM) in the 'lme4' package (Bates et al., 2015) to determine whether vigilance rates differed between clips where competitors were present versus absent (Appendix, Table A2).To model the binary outcome variable of efficient/inefficient for experiment 1, we ran mixed effects logistic regressions.We ran the first model to determine whether the presence or absence of competition or experience (trial number) impacted efficiency.We ran identical separate models for Phase 1 and Phase 2. Based on these results, we split the data set from Phase 1 into 'competition' and 'no competition' trials and ran additional mixed effects logistic regressions to examine how experience in these two situations affected efficiency (Appendix, Table A2).In all these models, we also included trial number as a random effect because squirrel identity (ID) was not known, and we used the 'drop1' function in R to determine significance.
For experiment 2, we first used the 'car' package (Fox & Weisberg, 2019) to assess the variance inflation factors (VIF) for our competition variables.VIFs were all <2, so all three measures of competition (presence of competitors within 5 m and within 1 m and having just been involved in chasing/aggression) could have been included in the same models.However, visual inspection of the data showed that when competitors were at 1 m, they had usually been involved in chasing or aggression as well (3 of 4 trials, 75%).Indeed, when chasing occurred, competitors were always still nearby, although proportionately less at 5 m than at 1 m (Appendix, Table A1).Thus, for our models for experiment 2, we collapsed the categories of competitors at 1 m and having just been involved in chasing into one variable 'all competition within 1 m'.
We modelled the effect of competitors and trial number as our measure of experience on foraging speed using an LMM.We set foraging speed as the dependent variable, testing the fixed effects of all competition at 1 m, competitors present at 5 m and trial number, with experiment day set as a random effect.We tried the same model set-up for the dependent variable of number of mistakes made, but this model was zero-inflated.So, to model foraging mistakes, we used a Hurdle Poisson model (HPM) in the package 'glmmTMB' (Brooks et al., 2017).We set the number of mistakes as the dependent variable and set the fixed effects as all competition at 1 m, competitors present at 5 m and trial number, with experiment day set as a random effect (Appendix, Table A2).Finally, we also calculated the odds ratio of squirrels making mistakes in competitive versus noncompetitive scenarios.

Ethical Note
The experimental methods used in this study were approved by the University of Toronto Animal Care Committee (Protocol: 20012732), the City of Toronto, and the Ontario Ministry of Natural Resources.Experiments met ASAB/ABS and ARRIVE guidelines for the ethical treatment of animals.The wild subjects were free to choose to participate or not in all our experiments.No squirrels were ever captured or touched.Observers stayed at least 2 m from subjects at all times.Competition between squirrels at our apparatuses took the form of displacements and chasing, minor physical aggression was sometimes observed but did not lead to injury.

Experiment 1
Squirrels were successfully able to access the food reward in all 130 trials of this experiment (100% success rate) and vigilance rates were greater when competitors were present compared to when they were absent (LMM: estimate ¼ 0.126, N ¼ 130, P < 0.001; Fig. 3, Appendix, Table A2).

Effect of experience and competition on foraging efficiency
The squirrels improved in efficiency over the course of Phase 1.In the first half of Phase 1, they were efficient in 40% of the trials (manipulating the redundant, topmost slats 60% of the time, 30 of 50 trials).However, in the second half of Phase 1, they were efficient in 70% of trials, decreasing their manipulation of the redundant slats to 30% of trials (15 of 50 trials).Our first linear mixed effects logistic regression showed that this improvement with experience was significant (estimate ¼ 0.029, N ¼ 100, P < 0.001) and that having competitors near the apparatus at the same time had a negative effect on the efficiency of the subjects (estimate ¼ À1.344, N ¼ 100, P ¼ 0.003; Appendix, Table A2).Overall, in Phase 1, in competition trials (N ¼ 56), the squirrels were efficient 44.6% of the time, while without competitors (N ¼ 44), they were efficient 68.2% of the time (Fig. 4).
Across the 56 trials in Phase 1 with competition, the squirrel's interaction with the redundant slats decreased from 75% in the first half of these trials to 36% in the second half (a 39% increase in efficiency).This was a larger increase in efficiency than was seen in Phase 1 trials with no competition (N ¼ 44, increase in efficiency of 18% from the first half of trials to the second half).This was partly due to the trials without competition starting out more efficient than those with competition (Fig. 4).The linear mixed effects logistic regressions on these two data sets showed that the improvement with experience that was found in our first model was actually due to the squirrels becoming more efficient over time at dealing with competitors (estimate ¼ 0.071, N ¼ 56, P < 0.001) and not a significant improvement over time when there were no competitors (estimate ¼ 0.033, N ¼ 44, P ¼ 0.199; Fig. 4).
In the 30 trials of Phase 2, when the squirrels were already familiar with the task, they were efficient in 77% of trials (manipulating the one redundant slat in 23% of trials, 7/30).The linear mixed effects logistic regression showed that neither experience (estimate ¼ 0.028, N ¼ 30, P ¼ 0.591) nor having competitors (estimate ¼ À0.389, N ¼ 30, P ¼ 0.691) had significant effects on efficiency in this phase (Appendix, Table A2).

Experiment 2
Squirrels were successfully able to access the food reward in all 140 trials of this experiment (100% success rate).

Effect of experience and competition on foraging speed and accuracy
Foraging speed was significantly impacted by all competition within 1 m (LMM: estimate ¼ 31.83,SE ¼ 3.85, P < 0.001), but not by having competitors within 5 m (estimate ¼ 3.72, SE ¼ 2.68, P ¼ 0.149) or by experience (estimate ¼ 0.006, SE ¼ 0.029, P ¼ 0.794; Appendix, Table A2).Specifically, having competitors at 1 m or having just been involved in chasing and/or aggressive physical contact with competitors increased the time it took for squirrels to extract the nut from our apparatus (Fig. 5).
The model examining mistakes made by the squirrels showed that these were more frequent when competitors were within 1 m and/or the focal squirrel had just been involved in chasing/aggressive physical contact with competitors (HPM: estimate ¼ 1.321, SE ¼ 0.39, P ¼ 0.001).Mistakes also became less frequent with experience (estimate ¼ À0.016, SE ¼ 0.004, P ¼ 0.002), but there was no effect of competitors within 5 m (estimate ¼ À0.516, SE ¼ 0.36, P ¼ 0.152; Appendix, Table A2).Plots of the conditional probability of mistakes throughout the experiment (Fig. 6) showed that mistakes were more frequent initially when squirrels were foraging alone and they were figuring out the apparatus and learning that the middle (white) container was always the rewarding one.However, their probability of making a mistake when foraging alone quickly fell off and levelled off at about 0.38.After this initial period, the squirrel's probability of making a mistake when foraging in competition was greater than when alone throughout the duration of the experiment but slowly dropped as they adapted to the presence of competitors.Overall, there was an odds ratio of 1.86 (CI: 0.91e3.74)that a squirrel would make a mistake when in competition versus when not in competition.Speed did not show the same improvement over time in either competitive or noncompetitive scenarios (Fig. 7), although the squirrels may have been somewhat affected by competitors within 5 m at the beginning of the experiment, they quickly adapted.

DISCUSSION
We used two foraging apparatuses to assess the effect of nearby competitors on eastern grey squirrel foraging efficiency, speed and accuracy.Our results show that having competitors within 1 m and/ or having just been involved in chasing or contact aggression with another squirrel had negative impacts on foraging efficiency, accuracy and speed (handling time).Competitors slightly farther away, within 5 m, did not significantly impact foraging speed or accuracy.Although speed was initially slower at the beginning of the experiment when conspecifics were within 5 m, the squirrels quickly adapted and acquired food at speeds that were equivalent to when there were no competitors present.Vigilance rates were greater for squirrels solving experiment 1 when other squirrels were present in our videos.Thus, these results support our hypothesis that having nearby food competitors is distracting and that allocating attention to them rather than to foraging tasks decreases performance.
Our results demonstrate that, not only are eastern grey squirrels sensitive to the presence of a conspecific audience when caching (Hopewell et al., 2008;Hopewell & Leaver, 2008;Leaver et al., 2007;Lichti et al., 2017;Steele et al., 2014), the presence of nearby conspecifics also affects their ability to forage efficiently.Nearby competitors and those actively trying to displace an individual obviously impose a great risk of losing feeding opportunities that must be monitored (e.g.Arseneau-Robar et al., 2022;Goss-Custard et al., 1999;Jones, 1998;Knight & Knight, 1986;Pascual & Senar, 2013;Voellmy et al., 2014;Zhao et al., 2020).In addition, the aggression and exertion of a chasing interaction likely raise the squirrel's stress level and heart rate, impairing their focus and affecting their efficiency and accuracy, as seen in human biathletes (Hoffman et al., 1992) and archers (Clemente et al., 2011).
This research builds upon other work supporting that animal's limited capacity for attention allocation (Bushnell, 1998;Dukas, 2002Dukas, , 2004) can lead to distraction by alternative stimuli and to negative impacts on important fitness-maintaining behaviours (Allen et al., 2021;Chan et al., 2010;Clark & Dukas, 2003).For example, Purser and Radford (2011) found that, when distracted by anthropogenic noise, threespine sticklebacks, Gasterosteus aculeatus, made more food-handling mistakes and were not as accurate in discriminating between food and nonfood items, needing more attacks to consume the same amount of prey.In other animal species, anthropogenic noise has also been found to impair attention to predator detection (Caribbean hermit crabs, Coenobita clypeatus:  (Beauchamp, 2015), this research highlights that this need to allocate attention to conspecifics can have costly effects on food acquisition.
In both of our experiments, eastern grey squirrels seemed to adapt to the presence of competitors at these apparatuses.In experiment 1, squirrel efficiency improved significantly over time when facing competition, and in experiment 2, the number of mistakes made in a competitive context slowly decreased, approaching the mistake rate seen when foraging alone.Indeed, if we allowed more trials, squirrel efficiency and mistake rates when in a competitive setting may have reached the equivalent of when they were foraging alone (Figs 4 and 6).Adaptation and habituation to distractions has been reported previously.In studies on the effects of anthropogenic noise on animals, responses have been found to weaken with repeated exposure (reviewed in Rojas et al., 2021).Note, however, that we were giving this population of squirrels novel foraging problems that evolution had not prepared them to solve.Thus, we forced these animals to focus their attention on the apparatuses initially, as they learned solutions, but less attention was probably needed with experience (Teichroeb et al., 2023).In addition, social foraging interactions are modelled using game theory (Giraldeau & Caraco, 2000), so squirrels interacting at these new apparatuses needed not only to learn how to get the reward, but also to establish strategies that would allow them to keep it, given the presence of competing squirrels.
This research brings up many exciting avenues of future research.If squirrels were tagged to be individually recognizable, it would be interesting to see whether there is individual variation in the ability to focus attention on foraging in the presence of distracting conspecifics.There is likely also individual variation in learning and thus efficiency, speed and accuracy over time that could be investigated.Social dominance hierarchies in this species are based on age, with older individuals being dominant.A wellestablished hierarchy is formed and reinforced every winter, and aggression is more focused on socially less familiar individuals (Pack et al., 1967;Thompson, 1977).The presence of a dominance hierarchy in our study population explains why we often observed presumed subordinates waiting nearby our experiments for the higher-ranking subject to retrieve food and leave to cache it before  they would approach to acquire the remaining food reward or begin a new trial.Knowledge of individual identity and dominance rank would allow us to determine whether dominants and subordinates use different strategies, whether these change with experience and handling skill and whether individuals of different rank vary in their motivation to learn at the apparatuses, all of which have been found for vervet monkeys (Arseneau-Robar et al., 2022, 2023).In addition, because the squirrels so easily solved our apparatuses, much future research could be done by presenting them with more difficult foraging tasks.For example, we did not change the location of the rewarding container in experiment 2 because, as there were only three containers, we did not want the rewarding one on the outside of the apparatus where squirrels could approach only that side.An apparatus with more tipping containers would allow changes to the difficulty of the problem presented to the squirrels and would presumably show changes in attention allocation and the effect of nearby competitors.
In conclusion, we found that having competing squirrels within 1 m or having just been involved in chasing or in aggressive interaction with conspecifics decreased the foraging efficiency, speed and accuracy of a population of eastern grey squirrels in an urban forest setting.We attribute these effects to the distraction caused by competitors and difficulties in attention allocation for our subjects (Dukas, 2002(Dukas, , 2004)).Nevertheless, the squirrels seemed to be adapting to the presence of competitors, at least with respect to foraging efficiency and accuracy, as these improved with experience.These results have important implications for how social foraging can impact optimal foraging strategies and highlight many new possible avenues of research.

Figure 1 .
Figure1.Clear-tube apparatus used in experiment 1 for eastern grey squirrels, which was attached to a tree on the University of Toronto Scarborough campus with two tension bands above a wooden platform to allow the squirrels to rest and to catch the reward.(a) Schematic of the tube task set-up in Phase 1, where the food reward was placed in the centre, with four metal slats above and below.(b) A squirrel solving the task in the Phase 1 configuration.All metal slats could be manipulated, but the subjects only needed to remove the ones below the reward to access the food (Videos S1eS2).(c) A photo of the apparatus in its Phase 2 configuration.In Phase 2, there were six to seven metal slats under the food reward that had to be removed, with only one redundant slat above the food reward.

Figure 3 .
Figure 3. Vigilance rates of eastern grey squirrels in video clips with and without competitors present in experiment 1.The lines show the medians, boxes represent interquartile ranges, whiskers show the minimum and maximum values excluding outliers, which are represented by dots.

Figure 4 .Figure 5 .Figure 6 .
Figure 4. Bayesian updating framework showing the conditional probability of eastern grey squirrel efficiency given their past experience over the course of Phase 1 of experiment 1 in competitive trials versus noncompetitive trials.Shaded areas show standard error.

Figure 7 .
Figure 7. Speed (i.e.handling time, s) for eastern grey squirrels to acquire the reward over the duration of experiment 2 with different levels of competition.Shaded areas show standard error.

Figure A1 .
Figure A1.The position of the camera trap relative to the apparatus platform for experiment 1 and experiment 2 on the University of Toronto Scarborough campus.Pictured above the platform is the experiment 1 apparatus.