Foraging bumblebees use social cues more when the task is difficult

Foraging bumblebees use social cues more when the task is difficult Abstract When foraging in their natural environment, many animals readily complement their personal knowledge with additional social information. To balance the costs and benefits of copying others, animals have to discern situations in which it is more advantageous to use social rather than personal information. Here, we used foraging bumblebees (Bombus terrestris) in a controlled laboratory setting and showed that the difficulty of a foraging task affects how the bees weight the 2 types of information. We used artificial flowers to devise easy and difficult discriminatory tasks, and tested the influence of floral and social cues on decision making. When facing an easy discrimination task, foraging bees were likely to rely on personal information and were only marginally affected by social information. By contrast, they prioritized social over personal information when flower discrimination was difficult and therefore the probability of making errors was higher. In summary, bees are able to use social and personal information to optimize foraging decisions in a flexible way. INTRODUCTION Animals are continuously challenged to extract reliable information from their environment, which can be particularly demanding under the unpredictable and ever changing circumstances that typically occur in nature (Danchin et al. 2004; Dall et al. 2005). One way to reduce the cost of obtaining personal information through individual exploration is to copy others (Laland 2004; Hoppitt and Laland 2013). Such social learning is taxonomically widespread (Giraldeau and Caraco 2000; Kendal et al. 2004; Galef 2009; Jones et al. 2015). However, social learning strategies have their own shortcomings (Boyd and Richerson 1988; Giraldeau et al. 2002). For instance, in rapidly changing environments, copying others excessively can lead to the circulation of obsolete information (Giraldeau et al. 2002). To balance the costs and benefits of copying others, animals have to find the optimal combination of social and personal information use (van Bergen et al. 2004; Kendal et al. 2004; Laland 2004; Kendal and Coolen 2011; Rieucau and Giraldeau 2011; Heyes 2012). Pollinators are ideal models for investigating social learning mechanisms. In a natural setting, they encounter many different species of flowers, some of which they may not have seen before. As unfamiliar flowers can be unrewarding (Heinrich 1979), so flower constancy—the tendency to exclusively visit certain flower species, while ignoring others—is advantageous (Chittka et al. 1999). Bees are naturally well equipped for flower discrimination (Menzel 1990) and can use different sensory modalities to rapidly learn to discriminate among flowers according to their reward (Menzel 1985; Chittka and Thomson 2001). Social information is also frequently available to bees, and an increasing body of studies reports that bees readily make use of additional social information to complement their personal knowledge (Dawson et al. 2013; Avarguès-Weber and Chittka 2014a). Experienced foragers are attracted to flowers with conspecifics (Leadbeater and Chittka 2005, 2007, 2009) and can learn simple flower color—reward associations by observing the choices of conspecifics from a distance (Worden and Papaj 2005; Dawson and Chittka 2012; Avarguès-Weber and Chittka 2014a). The mechanistic basis of social learning might be very simple and recent works in cognition have shown that basic principles of associative learning can also account for many aspects of social learning (Avarguès-Weber et al. 2015; Heyes and Pearce 2015, but see Avarguès-Weber and Chittka 2014b). For instance, Dawson et al. (2013) reported that seemingly complex social learning in bumblebees can emerge through the simple Pavlovian ability to integrate 2 learned associations (i.e. second order conditioning). Moreover, in bumblebees, individuals seem not to have an innate sense of the value of a social cue and have to learn it through experience. Consequently, individuals with no previous social foraging experience tend to ignore conspecific foraging choices when making foraging decisions (Dawson et al. 2013; Avarguès-Weber and Chittka 2014a). In bumblebees, the values attributed to social and personal cues change flexibly depending on the specific circumstances. Foraging bees are more likely to rely on social information when personal exploration is costly, for example because of a high predation risk (Abbott 2006; Ings and Chittka 2009). They also weight floral and social cues according their reliability, and prefer to use social cues when flowers offer highly variable rewards (Smolla et al. 2016). In addition, bees prefer to use social information when encountering unfamiliar flower types (Kawaguchi et al. 2007), or when their past foraging experience is with flowers with low rewards (Jones et al. 2015). One recent study (Dunlap et al. 2016) found that bees flexibly use social and personal information depending on the reliability of both types of information. Precisely, while bees account for the reliability of personally acquired information and social information in making foraging decisions, they do no treat social and personal information equally. Bees prefer to use social information if it consistently predicts a reward, but if social information becomes entirely unreliable, personal information is preferred (Dunlap et al. 2016). To date, as far as we know, no study has addressed how bees adjust the use of personal versus social information depending on the difficulty of the foraging tasks, and whether the reliance on social information in such contexts is flexible. We investigated how bumblebees (Bombus terrestris audax) use previously acquired social and personal information when challenged with a novel discrimination task, and whether they flexibly alter their information use according to the difficulty of the task. We used artificial flowers patterned with a varying number of black vertical bars to devise easy and hard discriminatory tasks between familiar and novel flowers. We used naive bumblebees, (i.e., bees that never had foraged with conspecifics before the start of a pretraining) and made sure they gained the same personal and social information by means of fully controlled pretrainings with rewarding flowers and rewarding social information. Then, we tested the bees in novel situations where we manipulated the task type and difficulty to explore variation in social information use. The series of experiments that we conducted were the following. 1) We first tested bumblebees in a task where social cues were not available, to investigate whether foraging performance would be affected by the task difficulty. This would be the case if individuals mistakenly chose unrewarding novel flowers more often when the discrimination task was more difficult. Theoretically, to circumvent the difficulty of a hard task and avoid incorrect choices, individuals that experienced reliable social information earlier should use additional social information when available, particularly when personal information does not yield sufficient rewards. To quantify this possibility, 2) we tested bumblebees in a second experiment with the same artificial flowers, where the rewarding flowers were also indicated by a social cue. We predicted that the additional social cue would improve the performance, as our bees experienced reliable social information previously, and that bees should show a greater decrease in incorrect choices when the task was harder in comparison to the test with no social information. To quantify the extent to which bees relied on social information over personal information, 3) we tested bumblebees in a situation where social and personal information were set in conflict. Here, we set the social cue on the novel, unrewarding flower. We predicted that the number of incorrect choices would increase in the hard task, as individuals should rely more on social information. In comparison, the number of incorrect choices in the easy task should be more similar to the scenario without available social information. MATERIAL AND METHODS Study species and setup Bumblebee foragers from a total of 4 colonies obtained from a continuous rearing program (Biobest, Belgium N.V.) were used. Bees from different colonies were roughly equally allocated to the experimental groups, so that each group was composed of bees from at least 2 different colonies. Each colony was kept in a wooden box (28 × 20 × 10 cm) connected to a flight arena (60 × 50 × 35 cm) by a Plexiglas tunnel (25 cm length; 3.5 × 3.5 cm in cross-section). A transparent Plexiglas lid (60 × 50 cm) covered the flight arena. Shutters along the length of the connecting tunnel helped us to control the flow of bees between the nest-box and the flight arena. Each bumblebee was tagged with a colored numbered disk (Opalithplättchen, Warnholz & Bienenvoigt, Germany) for individual identification. Colonies were supplied with commercial pollen (Koppert B.V., The Netherlands) and 30% (w/w) sucrose solution ad libitum. Pollen was simply added to the nest while sucrose solution was offered in a communal plastic feeder placed into the wooden box in the dark. Bees were not allowed to forage in the experimental arena, neither before nor during the period of the experiments. Each experimental day, a single naive bumblebee was allowed to enter the arena. Only active and motivated foragers were selected. Tests were conducted under standardized light (12 h:12 h) and temperature (23 ± 2 °C) conditions. Experimental procedure The general procedure consisted of 2 consecutive trainings and a single test. In both the trainings and the test, naive bees foraged individually in the arena, where 6 unrewarding and 6 rewarding artificial flowers were available. The rewarding flowers contained 5 μL of 30% sucrose solution (w/w) and the unrewarding flowers contained 5 μL of water. The artificial flowers were made of a vertical laminated green card (5 × 7 cm) fastened to the top of a 3 cm high rubber cylindrical pedestal (~2 cm in diameter), (Figure 1). Each flower was adorned with a floral cue made of a square of white paper (3 × 3 cm) with a black bar vertical pattern (1, 4, or 5 bars according to the scenario, see below), attached to the green card. These patterns were selected for creating a hard discrimination task (4 vs. 5 bar flowers, Hard Task) and an easy discrimination task (1 vs. 4 bar flowers, Easy Task). The respective task difficulties were tested and demonstrated by the fact that when only floral information was available, bumblebees made significantly more errors on the Hard Task (mean percentage ± SD = 33.7 ± 10.1) than the Easy Task (15.2 ± 6.0), (Mann–Whitney test, Z = −3.45, n1 = 12, n2 = 12, P = 0.001). For the social cue, we used freshly freeze-killed conspecifics pinned on the artificial flowers (according to the scenario, see below) (Kawaguchi et al. 2006; Leadbeater and Chittka 2009; Dawson and Chittka 2012). Previous works have already established that pinned dead bees or model clay bees elicit reactions similar to live conspecifics in foraging bees despite the lack of movement (Worden and Papaj 2005; Kawaguchi et al. 2006; Leadbeater and Chittka 2007, 2009). A transparent Plexiglas platform with a small cavity (Ø 0.5 cm, 0.2 cm in depth) contained the sucrose reward. The platform was attached to the card via a Velcro strip, below the white square. In the course of trainings and tests, emptied rewarding flowers were refilled after the bee moved to a different one. Thus, bees never experienced an empty flower, unless they returned to the last visited one. Between tests, we cleaned Plexiglas platforms with hot water and ethanol to remove any scent marks left by previous foragers. To prevent the bees from using the flowers’ spatial arrangement to solve the task, we changed the positions of the flowers randomly between each successive foraging bout. We observed the foragers’ flower choices during the successive trainings and test. A choice was recorded only when the bee landed on the Plexiglas platform. Choices were categorized as correct when the bees obtained the reward and incorrect otherwise. Within 1 h after the end of the training, subjects were tested and then removed from the colony. Training During trainings individual bumblebees were successively trained in 2 separate routines (A and B) to learn simple associations between either the vertical patterns (flower cue) or the demonstrators (social cue) and the reward (Figure 1). In training A, the foragers learned to associate a pattern of 4 vertical bars printed on a white background with the reward. The unrewarded flowers displayed neither the black bar pattern nor the white background. In training B, demonstrators pinned onto patternless flowers indicated the reward. To balance out the possible effect of training order, half of the bees received training B before training A and vice versa. We defined a foraging bout as a bee’s separate visit to the arena, from the moment the bee entered until it exited voluntarily. Training was considered accomplished when the number of mistakes per foraging bout fell to zero, and bees were allowed to complete as many individual foraging bouts as necessary to learn these simple associations. The number of bouts, as well as the total number of visits that bees needed to achieve learning during training A (bars: mean bouts ± SD = 6.08 ± 2.2; mean visits ± SD = 113.6 ± 24.1) and B (bees: bouts ± SD = 6.1 ± 1.5, mean visits ± SD = 108.2 ± 23.6) was similar (Wilcoxon test, bouts: Z = −0.16, P = 0.87; visits: Z = −1.15, P = 0.26), suggesting comparable difficulties of associating the reward with the 4-bar pattern and with the demonstrator. Figure 1 View largeDownload slide Above: Trainings received by all bees (A and B). Below: A total of 6 single tests were grouped into 3 experimental scenarios (Floral, Consistent Information and Conflicting Information). In all scenarios only flowers with 4 bars were rewarded with sucrose (S). One and 5 bar flowers contained water (W). In the first scenario, there were no demonstrators, only floral cues. In the second scenario, the demonstrators were pinned onto the familiar rewarding flowers. In the third scenario, the demonstrators were pinned onto the novel, unrewarding flowers. Figure 1 View largeDownload slide Above: Trainings received by all bees (A and B). Below: A total of 6 single tests were grouped into 3 experimental scenarios (Floral, Consistent Information and Conflicting Information). In all scenarios only flowers with 4 bars were rewarded with sucrose (S). One and 5 bar flowers contained water (W). In the first scenario, there were no demonstrators, only floral cues. In the second scenario, the demonstrators were pinned onto the familiar rewarding flowers. In the third scenario, the demonstrators were pinned onto the novel, unrewarding flowers. Test During the test, bees had to discriminate between rewarding and unrewarding flowers using information previously acquired in the trainings. To investigate whether the difficulty of a flower discrimination task influenced the use of social information, the bees were offered a choice between the familiar 4 bar patterned flowers and novel, nonrewarding flowers (1 out of 5 bars in the Easy/Hard Task tests, respectively), (Figure 1). We tested 3 experimental scenarios, each differing in the position of the social cues. The scenarios were further divided into an Easy Task test and a Hard Task test. 1) In the Floral Scenario, the flowers were without social cues and foragers had to discriminate between familiar rewarding 4 bar patterned flowers and novel nonrewarding flowers displaying 1 bar (Easy Task, n = 12 bees) or 5 bars (Hard Task, n = 12 bees) (Figure 1). 2) In the Consistent Information Scenario, demonstrators were pinned onto the rewarding 4 bar patterned flowers in both the Easy Task (n = 10 bees) and the Hard Task (n = 11 bees) (Figure 1). 3) In the Conflicting Information Scenario, demonstrators were pinned onto the novel and nonrewarding flowers (i.e., onto 1 bar flowers in the Easy Task (n = 12 bees); onto 5 bar patterned flowers in the Hard Task (n = 12 bees); Figure 1). Each bee completed one test that consisted of 50 flower choices and was not tested further. Bees were allowed as many foraging bouts as necessary to complete 50 flower choices. The number of foraging bouts needed to complete flower choices did not differ among treatments (Kruskal–Wallis test, χ2 = 8.16, df = 5, P = 0.15). Data analysis Data were analyzed using ANOVA designs. Bees’ choices (correct/incorrect) were examined using generalized linear mixed models (GLMMs) with a binomial error structure (logit-link function, glmer function of R package lme4 [Bates et al. 2014]) with “Scenario”, “Task Type”, “Training Order”, and their interactions as fixed factors. To compare how bees improved their foraging accuracy (i.e., the speed of learning) over different scenarios and tasks “Visits” (i.e., 50 flower visits) was entered in the model as a covariate and individuals’ identities and colony of origin were set as random factors to account for within-colony similarities and repeated measures. We retained the model with the highest explanatory power (i.e., the lowest AIC value). Models were optimized with the iterative algorithm, BOBYQA (Powell 2009). In the best model, the interactions (Scenario × Task Type) and (Scenario × Visits) were also included. We used Dunnett’s post-hoc tests to detect differences between the different groups (glht function from R package multcomp [Bretz et al. 2011]). In order to compare the bee performance in the different scenarios and tests before the bee had the chance to update its information, we analyzed the first foraging choices using χ2 tests. We also compared the individual percentage of incorrect choices over the entire test (50 choices) across different scenarios using Mann–Whitney U-test and Wilcoxon signed-rank test, or single sample t-test when assumption of normality and homogeneity of variance were met. All statistical analyses were performed with R 3.2.3 (R Development Core Team, 2016). RESULTS Overall, the number of errors made by the bees was significantly affected by the scenario and the difficulty of the task both for the first flower choice (GLMM, scenario: χ22 = 11.58, P = 0.003; task: χ21 =19.85, P < 0.0001, Figure 2) and over the 50 flower choices (GLMM, scenario: χ22 = 27.79, P < 0.0001; task: χ21 = 45.56, P < 0.0001, Figure 3). The interaction between the “Scenario” and the “Task Type” was also significant for the 50 flower choices (GLMM, χ2 = 16.06, df = 2, P < 0.0001), indicating that, as we predicted, forager bees used social cues differently depending on the difficulty of the task. Similarly, a significant interaction between the “Scenario” and the “Visits” (GLMM, χ2 = 14.09, df = 2, P < 0.0001) showed that the learning speed (i.e., slope of curve) differed between scenarios (Figure 4). Figure 2 View largeDownload slide First choices made in the Easy Task and the Hard Task tests by bumblebees in the Floral Scenario (n = 24), and when social and floral information were in agreement (Consistent Information Scenario, n = 21) or in conflict (Conflicting Information Scenario, n = 24) (*P < 0.005). Figure 2 View largeDownload slide First choices made in the Easy Task and the Hard Task tests by bumblebees in the Floral Scenario (n = 24), and when social and floral information were in agreement (Consistent Information Scenario, n = 21) or in conflict (Conflicting Information Scenario, n = 24) (*P < 0.005). Figure 3 View largeDownload slide Percentage of incorrect choices made by subjects over the entire test in the 3 scenarios (Floral (n = 24), Consistent Information (n = 21) and Conflicting Information (n = 24)) when challenged with the Easy and Hard Tasks. Red dots indicate individual values. Box plots show medians (thick black lines), 25th and 75th percentiles (a–b: P < 0.0001, c–d: P = 0.04, b–d: P < 0.0001). Figure 3 View largeDownload slide Percentage of incorrect choices made by subjects over the entire test in the 3 scenarios (Floral (n = 24), Consistent Information (n = 21) and Conflicting Information (n = 24)) when challenged with the Easy and Hard Tasks. Red dots indicate individual values. Box plots show medians (thick black lines), 25th and 75th percentiles (a–b: P < 0.0001, c–d: P = 0.04, b–d: P < 0.0001). Figure 4 View largeDownload slide Plot of individual bee foraging performance over the entire tests each consisting of 50 flower choices (x-axis). Six independent groups of bees were tested in 3 experimental scenarios (Floral [n = 24], Consistent Information [n = 21], and Conflicting Information [n = 24]) when challenged with the Easy and Hard Tasks. Different colors represent different individual bees. Circles represent raw data of bee choices that were pooled to 3 large time-bins for clarity. Curves were fitted to these pooled time-bins as proxy of the 50 visits completed by each individual bee. Figure 4 View largeDownload slide Plot of individual bee foraging performance over the entire tests each consisting of 50 flower choices (x-axis). Six independent groups of bees were tested in 3 experimental scenarios (Floral [n = 24], Consistent Information [n = 21], and Conflicting Information [n = 24]) when challenged with the Easy and Hard Tasks. Different colors represent different individual bees. Circles represent raw data of bee choices that were pooled to 3 large time-bins for clarity. Curves were fitted to these pooled time-bins as proxy of the 50 visits completed by each individual bee. When tested without demonstrators (Floral Scenario), bumblebees made significantly more errors over the entire test when challenged with the Hard Task (mean percentage ± SD = 33.7 ± 10.1) than with the Easy Task (15.2 ± 6.0), (Dunnett contrast, Z = −6.7, P < 0.0001, Figure 3). These results indicate that the bees found these discrimination tasks hard or easy as intended. Importantly, the first flower choice in the Hard Task was random (χ2 test, χ21 = 2.16, P = 0.2, > 50% of bees chose the novel pattern, Figure 2), indicating that the bees did not initially differentiate between the 2 patterns. However, the bees significantly improved over the course of the test (GLMM, χ21 = 20.11, P < 0.0001, Figure 4), so that, over the entire test flower choice was no longer random (one sample t-test, t = −5.6, df = 11, P < 0.001). Similarly, bees improved their foraging performance also when facing the Easy Task, although to a smaller extent, probably because the room for improvement was less (GLMM, χ21 = 4.26, P < 0.04, Figure 4). These results indicate that bumblebees were able to perceive the differences between the flower types and to solve the discrimination task. When social and personal information were in agreement (Consistent Information Scenario), the bees’ first choices in the test did not differ between the Easy and the Hard Task (100% and 90.9% of bees chose the correct flower in the Easy and Hard tasks, respectively, χ2 test, χ21 = 0.95, P = 1, Figure 2). Assuming that, as in the Floral Scenario, bees did not initially differentiate between the 2 test patterns in the Hard Task, the first choice here was between floral vs floral plus social cue. Over the whole test, the presence of a demonstrator on the rewarding flowers made the Hard Task (7.8 ± 3.7% of errors) significantly easier to solve (i.e., 4.3 times less errors) than in the Floral Scenario (Dunnett contrast, Z = 7.76, P < 0.0001; Figure 3). The bumblebees’ accuracy increased to the point where the percentage of errors was negligible (5–10%) and only marginally different between the Hard and Easy Tasks (3.6 ± 3.2), (Dunnett contrast, Z = −2.93, P = 0.04, Figure 3), underlining the usefulness of additional social information in discrimination tasks. As the bees solved the task near perfectly from the beginning, their improvement in performance over the test was small, although statistically significant (GLMM, Easy Task: χ21 = 3.18, P < 0.08; Hard Task: χ21 = 9.65, P < 0.02, Figure 4). When social and personal information were in conflict (Conflicting Information Scenario), the first choice made by the bees was strongly affected by the difficulty of the task (8.3% and 91.7% of test bees erroneously chose the social cue in the Easy and Hard Task respectively, χ2 test, χ21 = 8.11, P = 0.004, Figure 2). Note that the very same percentage of bees chose the social cue in the Hard Task, regardless of whether it was on the familiar 4 bar pattern (Consistent Information Scenario, 90.9%) or on the novel 5 bar pattern (Conflicting Information Scenario, 91.7%). This finding again shows that before bumblebees had a chance to update their information, they made their decisions entirely based on the social cue. On the contrary, foragers confronted with the Easy Task were not influenced by the social cues and showed similar performance as in the Floral Scenario (χ21 = 1.2, P = 0.27). Overall, these results indicate that forager bees relied on floral cues when facing an easy task and on only social cues when facing a difficult task. Next, we turned our attention to flexibility in the use of social information. Interestingly, the overall performance of the bees tested with the Hard Task in the Conflicting Information Scenario was not significantly different from their performance with the same task in the Floral Scenario (Dunnett contrast, Z = −2.04, P = 0.32, Figure 3). This was due to the fact that while bees were attracted to social cues in their first choice (11 out of 12 bees chose a flower with a demonstrator bee), they improved their performance over the 50 visits (χ21 = 58.71, P < 0.0001, Figure 4). In contrast, the overall performance of bumblebees on the Easy Task started the same and followed the same course of improvement in the Floral Scenario and the Conflicting Information Scenario (11.8 ± 2.2; Dunnett contrast, Z = 0.79, P = 0.97, Figure 3). As a consequence, the greatest discrepancy between the performances on the Easy and the Hard Tasks occurred in the Conflicting Information Scenario (Dunnett contrast, Z = −10.95, P < 0.0001, Figure 3). DISCUSSION We found that the value associated with social and personal information was flexible across a variety of experimental conditions. In particular, the level of similarity between previously learnt and novel flowers was a key factor affecting the way bees utilized the available information. When the foraging task was easy to solve, bees relied on personal information and seemed only marginally affected by social information. On the other hand, when the task was hard, the bees utilized social cues. During the training, our bees gradually conferred a positive value to the social cues through experience gained and, as a results, were attracted to the demonstrated flowers. Associative learning theory (Pearce and Bouton 2001) predicts that, when presented together, concordant cues have an additive effect on the strength of the animal’s reward expectation. Consistent with this view, we found that when bees were provided with concordant social and personal information, rewarding flowers had a higher attractiveness and fewer errors in flower choice were made. Moreover, our findings indicate that bumblebees use social information to circumvent the difficulty of solving a hard task, in agreement with evidence showing that social and personal information can act together to improve foraging efficiency (Rendell et al. 2010; Grüter et al. 2011). These results open up the possibility for the selective use of social information, when personal information is uncertain or when the behavior turns out to be unproductive (Laland 2004). When social cues accompany personal information, they are not always in agreement, and then bees have to discern situations where it is more advantageous to prioritize social rather than personal information, or vice versa. Overall, bumblebees resorted to social information when confronted with the Hard Task, but would not abandon known flower types for socially indicated novel types in the Easy Task. This reliance on flower information over social cue in the Easy Task might be in part due to the artificial floral advertisement cue itself. Indeed, artificial flowers were larger and thus probably more conspicuous than conspecific cues. Moreover, unlike on a real flower, where conspecifics would likely be found on the advertising tissue of the flower, these artificial flowers in part spatially separate conspecific and floral advertisement cues. This could have made it more difficult for foragers to devote attention to both cues simultaneously. Having said that, our results demonstrated that foragers facing a Hard Task strongly resorted to social information in their first choices, indicating that the social cue was salient enough even in the presence of a floral cue. In a similar experiment, Dunlap et al. (2016) concluded that when floral and social cues are perceived to be equally reliable, bees preferentially use social cues. They trained bees to yellow versus orange artificial flowers; as bees lack red receptors, these colors appear similar to them (Peitsch et al. 1992; Chittka and Waser 1997). Therefore, their color learning task corresponds to a “hard” task in our framework, for which we predict the preferential use of social cues. Upon leaving the nest, just before making the very first choice, bees considered both social and floral cues as reliable, as they were previously trained in scenarios where both flowers and demonstrator bees were associated with a reward. Hence, the very first choice was solely based on their propensity to rely on social or personal information to solve the discrimination task, but soon after the first choice bees faced uncertainty due to contradictory reward information conveyed by the social cues. We found that bees were not only flexible in the way they used personal and social information, but they updated their foraging preferences and discarded the newly unrewarding social cue quickly, in less than the 50 choices they were allowed in the tests. Overall, our results suggest that in the tests, bees simply learnt to ignore the social cues. However, we cannot exclude the possibility that the bees have learnt to use the demonstrator bees as aversive cues. Either way, this finding points to the ease at which learning of social cues can be reversed. The use of freshly freeze-killed conspecifics as a proxy for social cues raises the question whether dead bees were considered as live conspecifics or just as an additional nonsocial stimulus by foragers. Previous works have already established that pinned dead bees or model clay bees elicit reactions similar to live conspecifics in foraging bees despite the lack of movement (Worden and Papaj 2005; Kawaguchi et al. 2006; Leadbeater and Chittka 2007, 2009). Moreover, immobile pinned bees have been shown to be more efficient in drawing forager attention than arbitrary stimuli, suggesting that bees may have an innate ability to learn and use social-like stimuli (Avarguès-Weber et al. 2013; Avarguès-Weber and Chittka 2014b). While such an innate ability is still to be definitively proven, as there could be some learned components occurring within the colony such as odors, it represents a characteristic shared by many distant taxa including social insects. For instance, social wasps, which have the remarkable ability to recognize nest-mate faces (Tibbetts 2002; Baracchi et al. 2013, 2016), are also predisposed to memorize faces rather than arbitrary stimuli (Sheehan and Tibbetts 2011; Sheehan et al. 2014). Taken as a whole the evidence suggests that the use of dead bees as demonstrators successfully provides observer bees with social information under laboratory conditions. To conclude, our findings confirmed that bees readily take advantage of additional social information while foraging. Hard tasks can become relatively easy to solve when both social and personal information are available. As a result, by using simple associative mechanisms, bees can relatively quickly learn to solve tasks that they would have great difficulty to solve solely with the available personal information. Consistent with this idea, recent studies showed that with the help of concordant social information, bees can learn to solve extraordinary tasks to obtain a reward (Alem et al. 2016; Lokoula et al. 2017). We note, however, that bees never attained 100% correct choices, suggesting that individuals will keep on exploring and looking for more rewarding alternatives even if social information is available. The authors thank Professor Lars Chittka for useful advice and facilities, Oscar Ramos-Rodriguez for help in behavioral experiments and for providing technical support in the laboratory, and Tristan Matthews for comments on the earlier version of the manuscript. At the time of experiments, D.B. was supported by a Marie Curie Intra European Fellowship (IEF) within the 7th European Community Framework Programme. V.V. was funded by Human Frontier Science Program (reference no. LT-001074/2013-L). S.A. was funded by the Fondation Fyssen. D.B. and S.A. conceived the study, D.B., S.J.I., and V.V. performed the experiments. D.B. carried out data analysis. D.B., V.V., and S.A. wrote the paper. Data accessibilty: Analyses reported in this article can be reproduced using the data provided by Baracchi et al. (2017). REFERENCES Abbott K. 2006. Bumblebees avoid flowers containing evidence of past predation events. Can J Zool . 84: 1240– 1247. Google Scholar CrossRef Search ADS   Alem S, Perry CJ, Zhu X, Loukola OJ, Ingraham T, Søvik E, Chittka L. 2016. Associative mechanisms allow for social learning and cultural transmission of string pulling in an insect. PLoS Biol . 14: e1002564. Google Scholar CrossRef Search ADS PubMed  Avarguès-Weber A, Chittka L. 2014a. Observational conditioning in flower choice copying by bumblebees (Bombus terrestris): influence of observer distance and demonstrator movement. PLoS One . 9: e88415. Google Scholar CrossRef Search ADS   Avarguès-Weber A, Chittka L. 2014b. Local enhancement or stimulus enhancement? Bumblebees social learning results in a specific pattern of flower prefernce. Anim Behav . 97: 185– 191. Google Scholar CrossRef Search ADS   Avarguès-Weber A, Lihoreau M, Isabel G, Giurfa M. 2015. Information transfer beyond the waggle dance: observational learning in bees and flies. Front Ecol Evol . 3: 24. Avarguès-Weber A, Dawson E, Chittka L. 2013. Mechanisms of social learning across species boundaries. J Zool . 290: 1– 11. Google Scholar CrossRef Search ADS   Baracchi D, Petrocelli I, Cusseau G, Pizzocaro L, Teseo S, Turillazzi S. 2013. Facial markings in the hover wasps: quality signals and familiar recognition cues in two species of Stenogastrinae. Anim Behav . 85: 203– 212. Google Scholar CrossRef Search ADS   Baracchi D, Turillazzi S, Chittka L. 2016. Facial patterns in a tropical social wasp correlate with colony membership. Sci Nat . 103: 80. Google Scholar CrossRef Search ADS   Baracchi D, Vasas V, Jamshed Iqbal S, Alem S. 2017. Data from: foraging bumblebees use social cues more when the task is difficult. Dryad Digital Repository . http://dx.doi.org/10.5061/dryad.743g3 Bates D, Mächler M, Bolker B, Walker S. 2014. Fitting linear mixed-effects models using lme4. arXiv preprint. arXiv: 1406. 5823. van Bergen Y, Coolen I, Laland KN. 2004. Nine-spined sticklebacks exploit the most reliable source when public and private information conflict. Proc R Soc Lond B Biol Sci . 271: 957– 962. Google Scholar CrossRef Search ADS   Bretz F, Hothorn T, Westfall PH. 2011. Multiple comparisons using R . Boca Raton: CRC Press. Boyd R, Richerson PJ. 1988. An evolutionary model of social learning: the effects of spatial and temporal variation. Social learning: Psychological and Biological Perspectives : 29– 48. Chittka L, Thomson JD, Waser NM. 1999. Flower constancy, insect psychology, and plant evolution. Naturwissenschaften . 86: 361– 377. Google Scholar CrossRef Search ADS   Chittka L, Thomson JD. 2001. Cognitive ecology of pollination: animal behaviour and floral evolution . UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Chittka L, Waser NM. 1997. Why red flowers are not invisible to bees. Isr J Plant Sci . 45: 169– 183. Google Scholar CrossRef Search ADS   Dall SR, Giraldeau LA, Olsson O, McNamara JM, Stephens DW. 2005. Information and its use by animals in evolutionary ecology. Trend Ecol Evol . 20: 187– 193. Google Scholar CrossRef Search ADS   Danchin É, Giraldeau LA, Valone TJ, Wagner RH. 2004. Public information: from nosy neighbors to cultural evolution. Science . 305: 487– 491. Google Scholar CrossRef Search ADS PubMed  Dawson EH, Avargues-Weber A, Chittka L, Leadbeater E. 2013. Learning by observation emerges from simple associations in an insect model. Curr Biol . 23: 727– 730. Google Scholar CrossRef Search ADS PubMed  Dawson EH, Chittka L. 2012. Conspecific and heterospecific information use in bumblebees. PLoSOne . 7: e31444. Google Scholar CrossRef Search ADS   Dunlap AS, Nielsen ME, Dornhaus A, Papaj DR. 2016. Foraging bumble bees weigh the reliability of personal and social information. Curr Biol . 26: 1195– 1199. Google Scholar CrossRef Search ADS PubMed  Galef BG. 2009. Strategies for social learning: testing predictions from formal theory. Adv Stud Behav . 39: 117– 151. Google Scholar CrossRef Search ADS   Giraldeau LA, Caraco T. 2000. Social foraging theory . Princeton, New Jersey: Princeton University Press. Giraldeau LA, Valone TJ, Templeton JJ. 2002. Potential disadvantages of using socially acquired information. Philos Trans R Soc Lond B Biol Sci . 357: 1559– 1566. Google Scholar CrossRef Search ADS PubMed  Grüter C, Czaczkes TJ, Ratnieks FL. 2011. Decision making in ant foragers (Lasius niger) facing conflicting private and social information. Behav Ecol Sociobiol . 65: 141– 148. Google Scholar CrossRef Search ADS   Heinrich B. 1979. Resource heterogeneity and patterns of movement in foraging bumblebees. Oecologia . 40: 235– 245. Google Scholar CrossRef Search ADS PubMed  Heyes C. 2012. What’s social about social learning?. J Comp Psychol . 126: 193. Google Scholar CrossRef Search ADS PubMed  Heyes C, Pearce JM. 2015. Not-so-social learning strategies. Proc R Soc Lond B Biol Sci . 282: 20141709. Google Scholar CrossRef Search ADS   Hoppitt W, Laland KN. 2013. Social learning: an introduction to mechanisms, methods, and models . Princeton, New Jersey: Princeton University Press. Google Scholar CrossRef Search ADS   Ings TC, Chittka L. 2009. Predator crypsis enhances behaviourally mediated indirect effects on plants by altering bumblebee foraging preferences. Proc R Soc Lond B Biol Sci . 276: 2031– 2036. Google Scholar CrossRef Search ADS   Jones PL, Ryan MJ, Chittka L. 2015. The influence of past experience with flower reward quality on social learning in bumblebees. Anim Behav . 101: 11– 18. Google Scholar CrossRef Search ADS   Kawaguchi L, Ohashi K, Toquenaga Y. 2006. Do bumble bees save time when choosing novel flowers by following conspecifics? Funct Ecol . 20: 239– 244. Google Scholar CrossRef Search ADS   Kawaguchi LG, Ohashi K, Toquenaga Y. 2007. Contrasting responses of bumble bees to feeding conspecifics on their familiar and unfamiliar flowers. Proc R Soc Lond B Biol Sci . 274: 2661– 2667. Google Scholar CrossRef Search ADS   Kendal RL, Coolen I. 2011. Adaptive trade-offs in the use of social and personal information. In Cognitive ecology II . Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Kendal RL, Coolen I, Laland KN. 2004. The role of conformity in foraging when personal and social information conflict. Behav Ecol . 15: 269– 277. Google Scholar CrossRef Search ADS   Laland KN. 2004. Social learning strategies. Anim Learn Behav . 32: 4– 14. Google Scholar CrossRef Search ADS   Leadbeater E, Chittka L. 2005. A new mode of information transfer in foraging bumblebees? Curr Biol . 15: R447– R448. Google Scholar CrossRef Search ADS PubMed  Leadbeater E, Chittka L. 2007. The dynamics of social learning in an insect model, the bumblebee (Bombus terrestris). Behav Ecol Sociobiol . 61: 1789– 1796. Google Scholar CrossRef Search ADS   Leadbeater E, Chittka L. 2009. Bumble-bees learn the value of social cues through experience. Biol Lett . 5: 310– 2. Google Scholar CrossRef Search ADS PubMed  Loukola OJ, Perry CJ, Coscos L, Chittka L. 2017. Bumblebees show cognitive flexibility by improving on an observed complex behavior. Science . 355: 833– 836. Google Scholar CrossRef Search ADS PubMed  Menzel R. 1985. Learning in honey bees in an ecological and behavioral context. Exp Behav Ecol . 31: 55– 74. Menzel R. 1990. Learning, memory, and “cognition” in honey bees. In: Kesner RP, Olton DS, editors. Comparative cognition and neuroscience. Neurobiology of comparative cognition . Hillsdale (NJ): Lawrence Erlbaum Associates. p. 237– 292. Pearce JM, Bouton MR. 2001. Theories of associative learning in animals. Ann Rev Psychol . 52: 111– 139. Google Scholar CrossRef Search ADS   Peitsch D, Fietz A, Hertel H, Souza J, Ventura DF, Menzel R. 1992. The spectral input systems of hymenopteran insects and their receptor-based colour vision. J Comp Physiol A . 170: 23– 40. Google Scholar CrossRef Search ADS PubMed  Powell MJD. 2009. The BOBYQA algorithm for bound constrained optimization without derivatives . Cambridge (England): Department of Applied Mathematics and Theoretical Physics, Technical Report NA2009/06. Rendell L, Boyd R, Cownden D, Enquist M, Eriksson K, Feldman MV, Fogarty L, Ghirlanda S, Lillicrap T, Laland KN. 2010. Why copy others? Insights from the social learning strategies tournament. Science . 328: 208– 213. Google Scholar CrossRef Search ADS PubMed  Rieucau G, Giraldeau LA. 2011. Exploring the costs and benefits of social information use: an appraisal of current experimental evidence. Philos Trans R Soc Lond B Biol Sci . 366: 949– 957. Google Scholar CrossRef Search ADS PubMed  Sheehan MJ, Sholler D, Tibbetts EA. 2014. Specialized visual learning of facial signals of quality in the paper wasp, Polistes dominula. Biol J Linnean Soc . 113: 992– 997. Google Scholar CrossRef Search ADS   Sheehan MJ, Tibbetts EA. 2011. Specialized face learning is associated with individual recognition in paper wasps. Science . 334: 1272– 1275. Google Scholar CrossRef Search ADS PubMed  Smolla M, Alem S, Chittka L, Shultz S. 2016. Copy-when-uncertain: bumblebees rely on social information when rewards are highly variable. Biol Lett . 12: 20160188. Google Scholar CrossRef Search ADS PubMed  Tibbetts EA. 2002. Visual signals of individual identity in the wasp Polistes fuscatus. Proc R Soc Lond B Biol Sci . 269: 1423– 1428. Google Scholar CrossRef Search ADS   Worden BD, Papaj DR. 2005. Flower choice copying in bumblebees. Biol Lett . 1: 504– 507. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral Ecology Oxford University Press

Foraging bumblebees use social cues more when the task is difficult

Loading next page...
 
/lp/ou_press/foraging-bumblebees-use-social-cues-more-when-the-task-is-difficult-wf6kkNARNT
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1045-2249
eISSN
1465-7279
D.O.I.
10.1093/beheco/arx143
Publisher site
See Article on Publisher Site

Abstract

Abstract When foraging in their natural environment, many animals readily complement their personal knowledge with additional social information. To balance the costs and benefits of copying others, animals have to discern situations in which it is more advantageous to use social rather than personal information. Here, we used foraging bumblebees (Bombus terrestris) in a controlled laboratory setting and showed that the difficulty of a foraging task affects how the bees weight the 2 types of information. We used artificial flowers to devise easy and difficult discriminatory tasks, and tested the influence of floral and social cues on decision making. When facing an easy discrimination task, foraging bees were likely to rely on personal information and were only marginally affected by social information. By contrast, they prioritized social over personal information when flower discrimination was difficult and therefore the probability of making errors was higher. In summary, bees are able to use social and personal information to optimize foraging decisions in a flexible way. INTRODUCTION Animals are continuously challenged to extract reliable information from their environment, which can be particularly demanding under the unpredictable and ever changing circumstances that typically occur in nature (Danchin et al. 2004; Dall et al. 2005). One way to reduce the cost of obtaining personal information through individual exploration is to copy others (Laland 2004; Hoppitt and Laland 2013). Such social learning is taxonomically widespread (Giraldeau and Caraco 2000; Kendal et al. 2004; Galef 2009; Jones et al. 2015). However, social learning strategies have their own shortcomings (Boyd and Richerson 1988; Giraldeau et al. 2002). For instance, in rapidly changing environments, copying others excessively can lead to the circulation of obsolete information (Giraldeau et al. 2002). To balance the costs and benefits of copying others, animals have to find the optimal combination of social and personal information use (van Bergen et al. 2004; Kendal et al. 2004; Laland 2004; Kendal and Coolen 2011; Rieucau and Giraldeau 2011; Heyes 2012). Pollinators are ideal models for investigating social learning mechanisms. In a natural setting, they encounter many different species of flowers, some of which they may not have seen before. As unfamiliar flowers can be unrewarding (Heinrich 1979), so flower constancy—the tendency to exclusively visit certain flower species, while ignoring others—is advantageous (Chittka et al. 1999). Bees are naturally well equipped for flower discrimination (Menzel 1990) and can use different sensory modalities to rapidly learn to discriminate among flowers according to their reward (Menzel 1985; Chittka and Thomson 2001). Social information is also frequently available to bees, and an increasing body of studies reports that bees readily make use of additional social information to complement their personal knowledge (Dawson et al. 2013; Avarguès-Weber and Chittka 2014a). Experienced foragers are attracted to flowers with conspecifics (Leadbeater and Chittka 2005, 2007, 2009) and can learn simple flower color—reward associations by observing the choices of conspecifics from a distance (Worden and Papaj 2005; Dawson and Chittka 2012; Avarguès-Weber and Chittka 2014a). The mechanistic basis of social learning might be very simple and recent works in cognition have shown that basic principles of associative learning can also account for many aspects of social learning (Avarguès-Weber et al. 2015; Heyes and Pearce 2015, but see Avarguès-Weber and Chittka 2014b). For instance, Dawson et al. (2013) reported that seemingly complex social learning in bumblebees can emerge through the simple Pavlovian ability to integrate 2 learned associations (i.e. second order conditioning). Moreover, in bumblebees, individuals seem not to have an innate sense of the value of a social cue and have to learn it through experience. Consequently, individuals with no previous social foraging experience tend to ignore conspecific foraging choices when making foraging decisions (Dawson et al. 2013; Avarguès-Weber and Chittka 2014a). In bumblebees, the values attributed to social and personal cues change flexibly depending on the specific circumstances. Foraging bees are more likely to rely on social information when personal exploration is costly, for example because of a high predation risk (Abbott 2006; Ings and Chittka 2009). They also weight floral and social cues according their reliability, and prefer to use social cues when flowers offer highly variable rewards (Smolla et al. 2016). In addition, bees prefer to use social information when encountering unfamiliar flower types (Kawaguchi et al. 2007), or when their past foraging experience is with flowers with low rewards (Jones et al. 2015). One recent study (Dunlap et al. 2016) found that bees flexibly use social and personal information depending on the reliability of both types of information. Precisely, while bees account for the reliability of personally acquired information and social information in making foraging decisions, they do no treat social and personal information equally. Bees prefer to use social information if it consistently predicts a reward, but if social information becomes entirely unreliable, personal information is preferred (Dunlap et al. 2016). To date, as far as we know, no study has addressed how bees adjust the use of personal versus social information depending on the difficulty of the foraging tasks, and whether the reliance on social information in such contexts is flexible. We investigated how bumblebees (Bombus terrestris audax) use previously acquired social and personal information when challenged with a novel discrimination task, and whether they flexibly alter their information use according to the difficulty of the task. We used artificial flowers patterned with a varying number of black vertical bars to devise easy and hard discriminatory tasks between familiar and novel flowers. We used naive bumblebees, (i.e., bees that never had foraged with conspecifics before the start of a pretraining) and made sure they gained the same personal and social information by means of fully controlled pretrainings with rewarding flowers and rewarding social information. Then, we tested the bees in novel situations where we manipulated the task type and difficulty to explore variation in social information use. The series of experiments that we conducted were the following. 1) We first tested bumblebees in a task where social cues were not available, to investigate whether foraging performance would be affected by the task difficulty. This would be the case if individuals mistakenly chose unrewarding novel flowers more often when the discrimination task was more difficult. Theoretically, to circumvent the difficulty of a hard task and avoid incorrect choices, individuals that experienced reliable social information earlier should use additional social information when available, particularly when personal information does not yield sufficient rewards. To quantify this possibility, 2) we tested bumblebees in a second experiment with the same artificial flowers, where the rewarding flowers were also indicated by a social cue. We predicted that the additional social cue would improve the performance, as our bees experienced reliable social information previously, and that bees should show a greater decrease in incorrect choices when the task was harder in comparison to the test with no social information. To quantify the extent to which bees relied on social information over personal information, 3) we tested bumblebees in a situation where social and personal information were set in conflict. Here, we set the social cue on the novel, unrewarding flower. We predicted that the number of incorrect choices would increase in the hard task, as individuals should rely more on social information. In comparison, the number of incorrect choices in the easy task should be more similar to the scenario without available social information. MATERIAL AND METHODS Study species and setup Bumblebee foragers from a total of 4 colonies obtained from a continuous rearing program (Biobest, Belgium N.V.) were used. Bees from different colonies were roughly equally allocated to the experimental groups, so that each group was composed of bees from at least 2 different colonies. Each colony was kept in a wooden box (28 × 20 × 10 cm) connected to a flight arena (60 × 50 × 35 cm) by a Plexiglas tunnel (25 cm length; 3.5 × 3.5 cm in cross-section). A transparent Plexiglas lid (60 × 50 cm) covered the flight arena. Shutters along the length of the connecting tunnel helped us to control the flow of bees between the nest-box and the flight arena. Each bumblebee was tagged with a colored numbered disk (Opalithplättchen, Warnholz & Bienenvoigt, Germany) for individual identification. Colonies were supplied with commercial pollen (Koppert B.V., The Netherlands) and 30% (w/w) sucrose solution ad libitum. Pollen was simply added to the nest while sucrose solution was offered in a communal plastic feeder placed into the wooden box in the dark. Bees were not allowed to forage in the experimental arena, neither before nor during the period of the experiments. Each experimental day, a single naive bumblebee was allowed to enter the arena. Only active and motivated foragers were selected. Tests were conducted under standardized light (12 h:12 h) and temperature (23 ± 2 °C) conditions. Experimental procedure The general procedure consisted of 2 consecutive trainings and a single test. In both the trainings and the test, naive bees foraged individually in the arena, where 6 unrewarding and 6 rewarding artificial flowers were available. The rewarding flowers contained 5 μL of 30% sucrose solution (w/w) and the unrewarding flowers contained 5 μL of water. The artificial flowers were made of a vertical laminated green card (5 × 7 cm) fastened to the top of a 3 cm high rubber cylindrical pedestal (~2 cm in diameter), (Figure 1). Each flower was adorned with a floral cue made of a square of white paper (3 × 3 cm) with a black bar vertical pattern (1, 4, or 5 bars according to the scenario, see below), attached to the green card. These patterns were selected for creating a hard discrimination task (4 vs. 5 bar flowers, Hard Task) and an easy discrimination task (1 vs. 4 bar flowers, Easy Task). The respective task difficulties were tested and demonstrated by the fact that when only floral information was available, bumblebees made significantly more errors on the Hard Task (mean percentage ± SD = 33.7 ± 10.1) than the Easy Task (15.2 ± 6.0), (Mann–Whitney test, Z = −3.45, n1 = 12, n2 = 12, P = 0.001). For the social cue, we used freshly freeze-killed conspecifics pinned on the artificial flowers (according to the scenario, see below) (Kawaguchi et al. 2006; Leadbeater and Chittka 2009; Dawson and Chittka 2012). Previous works have already established that pinned dead bees or model clay bees elicit reactions similar to live conspecifics in foraging bees despite the lack of movement (Worden and Papaj 2005; Kawaguchi et al. 2006; Leadbeater and Chittka 2007, 2009). A transparent Plexiglas platform with a small cavity (Ø 0.5 cm, 0.2 cm in depth) contained the sucrose reward. The platform was attached to the card via a Velcro strip, below the white square. In the course of trainings and tests, emptied rewarding flowers were refilled after the bee moved to a different one. Thus, bees never experienced an empty flower, unless they returned to the last visited one. Between tests, we cleaned Plexiglas platforms with hot water and ethanol to remove any scent marks left by previous foragers. To prevent the bees from using the flowers’ spatial arrangement to solve the task, we changed the positions of the flowers randomly between each successive foraging bout. We observed the foragers’ flower choices during the successive trainings and test. A choice was recorded only when the bee landed on the Plexiglas platform. Choices were categorized as correct when the bees obtained the reward and incorrect otherwise. Within 1 h after the end of the training, subjects were tested and then removed from the colony. Training During trainings individual bumblebees were successively trained in 2 separate routines (A and B) to learn simple associations between either the vertical patterns (flower cue) or the demonstrators (social cue) and the reward (Figure 1). In training A, the foragers learned to associate a pattern of 4 vertical bars printed on a white background with the reward. The unrewarded flowers displayed neither the black bar pattern nor the white background. In training B, demonstrators pinned onto patternless flowers indicated the reward. To balance out the possible effect of training order, half of the bees received training B before training A and vice versa. We defined a foraging bout as a bee’s separate visit to the arena, from the moment the bee entered until it exited voluntarily. Training was considered accomplished when the number of mistakes per foraging bout fell to zero, and bees were allowed to complete as many individual foraging bouts as necessary to learn these simple associations. The number of bouts, as well as the total number of visits that bees needed to achieve learning during training A (bars: mean bouts ± SD = 6.08 ± 2.2; mean visits ± SD = 113.6 ± 24.1) and B (bees: bouts ± SD = 6.1 ± 1.5, mean visits ± SD = 108.2 ± 23.6) was similar (Wilcoxon test, bouts: Z = −0.16, P = 0.87; visits: Z = −1.15, P = 0.26), suggesting comparable difficulties of associating the reward with the 4-bar pattern and with the demonstrator. Figure 1 View largeDownload slide Above: Trainings received by all bees (A and B). Below: A total of 6 single tests were grouped into 3 experimental scenarios (Floral, Consistent Information and Conflicting Information). In all scenarios only flowers with 4 bars were rewarded with sucrose (S). One and 5 bar flowers contained water (W). In the first scenario, there were no demonstrators, only floral cues. In the second scenario, the demonstrators were pinned onto the familiar rewarding flowers. In the third scenario, the demonstrators were pinned onto the novel, unrewarding flowers. Figure 1 View largeDownload slide Above: Trainings received by all bees (A and B). Below: A total of 6 single tests were grouped into 3 experimental scenarios (Floral, Consistent Information and Conflicting Information). In all scenarios only flowers with 4 bars were rewarded with sucrose (S). One and 5 bar flowers contained water (W). In the first scenario, there were no demonstrators, only floral cues. In the second scenario, the demonstrators were pinned onto the familiar rewarding flowers. In the third scenario, the demonstrators were pinned onto the novel, unrewarding flowers. Test During the test, bees had to discriminate between rewarding and unrewarding flowers using information previously acquired in the trainings. To investigate whether the difficulty of a flower discrimination task influenced the use of social information, the bees were offered a choice between the familiar 4 bar patterned flowers and novel, nonrewarding flowers (1 out of 5 bars in the Easy/Hard Task tests, respectively), (Figure 1). We tested 3 experimental scenarios, each differing in the position of the social cues. The scenarios were further divided into an Easy Task test and a Hard Task test. 1) In the Floral Scenario, the flowers were without social cues and foragers had to discriminate between familiar rewarding 4 bar patterned flowers and novel nonrewarding flowers displaying 1 bar (Easy Task, n = 12 bees) or 5 bars (Hard Task, n = 12 bees) (Figure 1). 2) In the Consistent Information Scenario, demonstrators were pinned onto the rewarding 4 bar patterned flowers in both the Easy Task (n = 10 bees) and the Hard Task (n = 11 bees) (Figure 1). 3) In the Conflicting Information Scenario, demonstrators were pinned onto the novel and nonrewarding flowers (i.e., onto 1 bar flowers in the Easy Task (n = 12 bees); onto 5 bar patterned flowers in the Hard Task (n = 12 bees); Figure 1). Each bee completed one test that consisted of 50 flower choices and was not tested further. Bees were allowed as many foraging bouts as necessary to complete 50 flower choices. The number of foraging bouts needed to complete flower choices did not differ among treatments (Kruskal–Wallis test, χ2 = 8.16, df = 5, P = 0.15). Data analysis Data were analyzed using ANOVA designs. Bees’ choices (correct/incorrect) were examined using generalized linear mixed models (GLMMs) with a binomial error structure (logit-link function, glmer function of R package lme4 [Bates et al. 2014]) with “Scenario”, “Task Type”, “Training Order”, and their interactions as fixed factors. To compare how bees improved their foraging accuracy (i.e., the speed of learning) over different scenarios and tasks “Visits” (i.e., 50 flower visits) was entered in the model as a covariate and individuals’ identities and colony of origin were set as random factors to account for within-colony similarities and repeated measures. We retained the model with the highest explanatory power (i.e., the lowest AIC value). Models were optimized with the iterative algorithm, BOBYQA (Powell 2009). In the best model, the interactions (Scenario × Task Type) and (Scenario × Visits) were also included. We used Dunnett’s post-hoc tests to detect differences between the different groups (glht function from R package multcomp [Bretz et al. 2011]). In order to compare the bee performance in the different scenarios and tests before the bee had the chance to update its information, we analyzed the first foraging choices using χ2 tests. We also compared the individual percentage of incorrect choices over the entire test (50 choices) across different scenarios using Mann–Whitney U-test and Wilcoxon signed-rank test, or single sample t-test when assumption of normality and homogeneity of variance were met. All statistical analyses were performed with R 3.2.3 (R Development Core Team, 2016). RESULTS Overall, the number of errors made by the bees was significantly affected by the scenario and the difficulty of the task both for the first flower choice (GLMM, scenario: χ22 = 11.58, P = 0.003; task: χ21 =19.85, P < 0.0001, Figure 2) and over the 50 flower choices (GLMM, scenario: χ22 = 27.79, P < 0.0001; task: χ21 = 45.56, P < 0.0001, Figure 3). The interaction between the “Scenario” and the “Task Type” was also significant for the 50 flower choices (GLMM, χ2 = 16.06, df = 2, P < 0.0001), indicating that, as we predicted, forager bees used social cues differently depending on the difficulty of the task. Similarly, a significant interaction between the “Scenario” and the “Visits” (GLMM, χ2 = 14.09, df = 2, P < 0.0001) showed that the learning speed (i.e., slope of curve) differed between scenarios (Figure 4). Figure 2 View largeDownload slide First choices made in the Easy Task and the Hard Task tests by bumblebees in the Floral Scenario (n = 24), and when social and floral information were in agreement (Consistent Information Scenario, n = 21) or in conflict (Conflicting Information Scenario, n = 24) (*P < 0.005). Figure 2 View largeDownload slide First choices made in the Easy Task and the Hard Task tests by bumblebees in the Floral Scenario (n = 24), and when social and floral information were in agreement (Consistent Information Scenario, n = 21) or in conflict (Conflicting Information Scenario, n = 24) (*P < 0.005). Figure 3 View largeDownload slide Percentage of incorrect choices made by subjects over the entire test in the 3 scenarios (Floral (n = 24), Consistent Information (n = 21) and Conflicting Information (n = 24)) when challenged with the Easy and Hard Tasks. Red dots indicate individual values. Box plots show medians (thick black lines), 25th and 75th percentiles (a–b: P < 0.0001, c–d: P = 0.04, b–d: P < 0.0001). Figure 3 View largeDownload slide Percentage of incorrect choices made by subjects over the entire test in the 3 scenarios (Floral (n = 24), Consistent Information (n = 21) and Conflicting Information (n = 24)) when challenged with the Easy and Hard Tasks. Red dots indicate individual values. Box plots show medians (thick black lines), 25th and 75th percentiles (a–b: P < 0.0001, c–d: P = 0.04, b–d: P < 0.0001). Figure 4 View largeDownload slide Plot of individual bee foraging performance over the entire tests each consisting of 50 flower choices (x-axis). Six independent groups of bees were tested in 3 experimental scenarios (Floral [n = 24], Consistent Information [n = 21], and Conflicting Information [n = 24]) when challenged with the Easy and Hard Tasks. Different colors represent different individual bees. Circles represent raw data of bee choices that were pooled to 3 large time-bins for clarity. Curves were fitted to these pooled time-bins as proxy of the 50 visits completed by each individual bee. Figure 4 View largeDownload slide Plot of individual bee foraging performance over the entire tests each consisting of 50 flower choices (x-axis). Six independent groups of bees were tested in 3 experimental scenarios (Floral [n = 24], Consistent Information [n = 21], and Conflicting Information [n = 24]) when challenged with the Easy and Hard Tasks. Different colors represent different individual bees. Circles represent raw data of bee choices that were pooled to 3 large time-bins for clarity. Curves were fitted to these pooled time-bins as proxy of the 50 visits completed by each individual bee. When tested without demonstrators (Floral Scenario), bumblebees made significantly more errors over the entire test when challenged with the Hard Task (mean percentage ± SD = 33.7 ± 10.1) than with the Easy Task (15.2 ± 6.0), (Dunnett contrast, Z = −6.7, P < 0.0001, Figure 3). These results indicate that the bees found these discrimination tasks hard or easy as intended. Importantly, the first flower choice in the Hard Task was random (χ2 test, χ21 = 2.16, P = 0.2, > 50% of bees chose the novel pattern, Figure 2), indicating that the bees did not initially differentiate between the 2 patterns. However, the bees significantly improved over the course of the test (GLMM, χ21 = 20.11, P < 0.0001, Figure 4), so that, over the entire test flower choice was no longer random (one sample t-test, t = −5.6, df = 11, P < 0.001). Similarly, bees improved their foraging performance also when facing the Easy Task, although to a smaller extent, probably because the room for improvement was less (GLMM, χ21 = 4.26, P < 0.04, Figure 4). These results indicate that bumblebees were able to perceive the differences between the flower types and to solve the discrimination task. When social and personal information were in agreement (Consistent Information Scenario), the bees’ first choices in the test did not differ between the Easy and the Hard Task (100% and 90.9% of bees chose the correct flower in the Easy and Hard tasks, respectively, χ2 test, χ21 = 0.95, P = 1, Figure 2). Assuming that, as in the Floral Scenario, bees did not initially differentiate between the 2 test patterns in the Hard Task, the first choice here was between floral vs floral plus social cue. Over the whole test, the presence of a demonstrator on the rewarding flowers made the Hard Task (7.8 ± 3.7% of errors) significantly easier to solve (i.e., 4.3 times less errors) than in the Floral Scenario (Dunnett contrast, Z = 7.76, P < 0.0001; Figure 3). The bumblebees’ accuracy increased to the point where the percentage of errors was negligible (5–10%) and only marginally different between the Hard and Easy Tasks (3.6 ± 3.2), (Dunnett contrast, Z = −2.93, P = 0.04, Figure 3), underlining the usefulness of additional social information in discrimination tasks. As the bees solved the task near perfectly from the beginning, their improvement in performance over the test was small, although statistically significant (GLMM, Easy Task: χ21 = 3.18, P < 0.08; Hard Task: χ21 = 9.65, P < 0.02, Figure 4). When social and personal information were in conflict (Conflicting Information Scenario), the first choice made by the bees was strongly affected by the difficulty of the task (8.3% and 91.7% of test bees erroneously chose the social cue in the Easy and Hard Task respectively, χ2 test, χ21 = 8.11, P = 0.004, Figure 2). Note that the very same percentage of bees chose the social cue in the Hard Task, regardless of whether it was on the familiar 4 bar pattern (Consistent Information Scenario, 90.9%) or on the novel 5 bar pattern (Conflicting Information Scenario, 91.7%). This finding again shows that before bumblebees had a chance to update their information, they made their decisions entirely based on the social cue. On the contrary, foragers confronted with the Easy Task were not influenced by the social cues and showed similar performance as in the Floral Scenario (χ21 = 1.2, P = 0.27). Overall, these results indicate that forager bees relied on floral cues when facing an easy task and on only social cues when facing a difficult task. Next, we turned our attention to flexibility in the use of social information. Interestingly, the overall performance of the bees tested with the Hard Task in the Conflicting Information Scenario was not significantly different from their performance with the same task in the Floral Scenario (Dunnett contrast, Z = −2.04, P = 0.32, Figure 3). This was due to the fact that while bees were attracted to social cues in their first choice (11 out of 12 bees chose a flower with a demonstrator bee), they improved their performance over the 50 visits (χ21 = 58.71, P < 0.0001, Figure 4). In contrast, the overall performance of bumblebees on the Easy Task started the same and followed the same course of improvement in the Floral Scenario and the Conflicting Information Scenario (11.8 ± 2.2; Dunnett contrast, Z = 0.79, P = 0.97, Figure 3). As a consequence, the greatest discrepancy between the performances on the Easy and the Hard Tasks occurred in the Conflicting Information Scenario (Dunnett contrast, Z = −10.95, P < 0.0001, Figure 3). DISCUSSION We found that the value associated with social and personal information was flexible across a variety of experimental conditions. In particular, the level of similarity between previously learnt and novel flowers was a key factor affecting the way bees utilized the available information. When the foraging task was easy to solve, bees relied on personal information and seemed only marginally affected by social information. On the other hand, when the task was hard, the bees utilized social cues. During the training, our bees gradually conferred a positive value to the social cues through experience gained and, as a results, were attracted to the demonstrated flowers. Associative learning theory (Pearce and Bouton 2001) predicts that, when presented together, concordant cues have an additive effect on the strength of the animal’s reward expectation. Consistent with this view, we found that when bees were provided with concordant social and personal information, rewarding flowers had a higher attractiveness and fewer errors in flower choice were made. Moreover, our findings indicate that bumblebees use social information to circumvent the difficulty of solving a hard task, in agreement with evidence showing that social and personal information can act together to improve foraging efficiency (Rendell et al. 2010; Grüter et al. 2011). These results open up the possibility for the selective use of social information, when personal information is uncertain or when the behavior turns out to be unproductive (Laland 2004). When social cues accompany personal information, they are not always in agreement, and then bees have to discern situations where it is more advantageous to prioritize social rather than personal information, or vice versa. Overall, bumblebees resorted to social information when confronted with the Hard Task, but would not abandon known flower types for socially indicated novel types in the Easy Task. This reliance on flower information over social cue in the Easy Task might be in part due to the artificial floral advertisement cue itself. Indeed, artificial flowers were larger and thus probably more conspicuous than conspecific cues. Moreover, unlike on a real flower, where conspecifics would likely be found on the advertising tissue of the flower, these artificial flowers in part spatially separate conspecific and floral advertisement cues. This could have made it more difficult for foragers to devote attention to both cues simultaneously. Having said that, our results demonstrated that foragers facing a Hard Task strongly resorted to social information in their first choices, indicating that the social cue was salient enough even in the presence of a floral cue. In a similar experiment, Dunlap et al. (2016) concluded that when floral and social cues are perceived to be equally reliable, bees preferentially use social cues. They trained bees to yellow versus orange artificial flowers; as bees lack red receptors, these colors appear similar to them (Peitsch et al. 1992; Chittka and Waser 1997). Therefore, their color learning task corresponds to a “hard” task in our framework, for which we predict the preferential use of social cues. Upon leaving the nest, just before making the very first choice, bees considered both social and floral cues as reliable, as they were previously trained in scenarios where both flowers and demonstrator bees were associated with a reward. Hence, the very first choice was solely based on their propensity to rely on social or personal information to solve the discrimination task, but soon after the first choice bees faced uncertainty due to contradictory reward information conveyed by the social cues. We found that bees were not only flexible in the way they used personal and social information, but they updated their foraging preferences and discarded the newly unrewarding social cue quickly, in less than the 50 choices they were allowed in the tests. Overall, our results suggest that in the tests, bees simply learnt to ignore the social cues. However, we cannot exclude the possibility that the bees have learnt to use the demonstrator bees as aversive cues. Either way, this finding points to the ease at which learning of social cues can be reversed. The use of freshly freeze-killed conspecifics as a proxy for social cues raises the question whether dead bees were considered as live conspecifics or just as an additional nonsocial stimulus by foragers. Previous works have already established that pinned dead bees or model clay bees elicit reactions similar to live conspecifics in foraging bees despite the lack of movement (Worden and Papaj 2005; Kawaguchi et al. 2006; Leadbeater and Chittka 2007, 2009). Moreover, immobile pinned bees have been shown to be more efficient in drawing forager attention than arbitrary stimuli, suggesting that bees may have an innate ability to learn and use social-like stimuli (Avarguès-Weber et al. 2013; Avarguès-Weber and Chittka 2014b). While such an innate ability is still to be definitively proven, as there could be some learned components occurring within the colony such as odors, it represents a characteristic shared by many distant taxa including social insects. For instance, social wasps, which have the remarkable ability to recognize nest-mate faces (Tibbetts 2002; Baracchi et al. 2013, 2016), are also predisposed to memorize faces rather than arbitrary stimuli (Sheehan and Tibbetts 2011; Sheehan et al. 2014). Taken as a whole the evidence suggests that the use of dead bees as demonstrators successfully provides observer bees with social information under laboratory conditions. To conclude, our findings confirmed that bees readily take advantage of additional social information while foraging. Hard tasks can become relatively easy to solve when both social and personal information are available. As a result, by using simple associative mechanisms, bees can relatively quickly learn to solve tasks that they would have great difficulty to solve solely with the available personal information. Consistent with this idea, recent studies showed that with the help of concordant social information, bees can learn to solve extraordinary tasks to obtain a reward (Alem et al. 2016; Lokoula et al. 2017). We note, however, that bees never attained 100% correct choices, suggesting that individuals will keep on exploring and looking for more rewarding alternatives even if social information is available. The authors thank Professor Lars Chittka for useful advice and facilities, Oscar Ramos-Rodriguez for help in behavioral experiments and for providing technical support in the laboratory, and Tristan Matthews for comments on the earlier version of the manuscript. At the time of experiments, D.B. was supported by a Marie Curie Intra European Fellowship (IEF) within the 7th European Community Framework Programme. V.V. was funded by Human Frontier Science Program (reference no. LT-001074/2013-L). S.A. was funded by the Fondation Fyssen. D.B. and S.A. conceived the study, D.B., S.J.I., and V.V. performed the experiments. D.B. carried out data analysis. D.B., V.V., and S.A. wrote the paper. Data accessibilty: Analyses reported in this article can be reproduced using the data provided by Baracchi et al. (2017). REFERENCES Abbott K. 2006. Bumblebees avoid flowers containing evidence of past predation events. Can J Zool . 84: 1240– 1247. Google Scholar CrossRef Search ADS   Alem S, Perry CJ, Zhu X, Loukola OJ, Ingraham T, Søvik E, Chittka L. 2016. Associative mechanisms allow for social learning and cultural transmission of string pulling in an insect. PLoS Biol . 14: e1002564. Google Scholar CrossRef Search ADS PubMed  Avarguès-Weber A, Chittka L. 2014a. Observational conditioning in flower choice copying by bumblebees (Bombus terrestris): influence of observer distance and demonstrator movement. PLoS One . 9: e88415. Google Scholar CrossRef Search ADS   Avarguès-Weber A, Chittka L. 2014b. Local enhancement or stimulus enhancement? Bumblebees social learning results in a specific pattern of flower prefernce. Anim Behav . 97: 185– 191. Google Scholar CrossRef Search ADS   Avarguès-Weber A, Lihoreau M, Isabel G, Giurfa M. 2015. Information transfer beyond the waggle dance: observational learning in bees and flies. Front Ecol Evol . 3: 24. Avarguès-Weber A, Dawson E, Chittka L. 2013. Mechanisms of social learning across species boundaries. J Zool . 290: 1– 11. Google Scholar CrossRef Search ADS   Baracchi D, Petrocelli I, Cusseau G, Pizzocaro L, Teseo S, Turillazzi S. 2013. Facial markings in the hover wasps: quality signals and familiar recognition cues in two species of Stenogastrinae. Anim Behav . 85: 203– 212. Google Scholar CrossRef Search ADS   Baracchi D, Turillazzi S, Chittka L. 2016. Facial patterns in a tropical social wasp correlate with colony membership. Sci Nat . 103: 80. Google Scholar CrossRef Search ADS   Baracchi D, Vasas V, Jamshed Iqbal S, Alem S. 2017. Data from: foraging bumblebees use social cues more when the task is difficult. Dryad Digital Repository . http://dx.doi.org/10.5061/dryad.743g3 Bates D, Mächler M, Bolker B, Walker S. 2014. Fitting linear mixed-effects models using lme4. arXiv preprint. arXiv: 1406. 5823. van Bergen Y, Coolen I, Laland KN. 2004. Nine-spined sticklebacks exploit the most reliable source when public and private information conflict. Proc R Soc Lond B Biol Sci . 271: 957– 962. Google Scholar CrossRef Search ADS   Bretz F, Hothorn T, Westfall PH. 2011. Multiple comparisons using R . Boca Raton: CRC Press. Boyd R, Richerson PJ. 1988. An evolutionary model of social learning: the effects of spatial and temporal variation. Social learning: Psychological and Biological Perspectives : 29– 48. Chittka L, Thomson JD, Waser NM. 1999. Flower constancy, insect psychology, and plant evolution. Naturwissenschaften . 86: 361– 377. Google Scholar CrossRef Search ADS   Chittka L, Thomson JD. 2001. Cognitive ecology of pollination: animal behaviour and floral evolution . UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Chittka L, Waser NM. 1997. Why red flowers are not invisible to bees. Isr J Plant Sci . 45: 169– 183. Google Scholar CrossRef Search ADS   Dall SR, Giraldeau LA, Olsson O, McNamara JM, Stephens DW. 2005. Information and its use by animals in evolutionary ecology. Trend Ecol Evol . 20: 187– 193. Google Scholar CrossRef Search ADS   Danchin É, Giraldeau LA, Valone TJ, Wagner RH. 2004. Public information: from nosy neighbors to cultural evolution. Science . 305: 487– 491. Google Scholar CrossRef Search ADS PubMed  Dawson EH, Avargues-Weber A, Chittka L, Leadbeater E. 2013. Learning by observation emerges from simple associations in an insect model. Curr Biol . 23: 727– 730. Google Scholar CrossRef Search ADS PubMed  Dawson EH, Chittka L. 2012. Conspecific and heterospecific information use in bumblebees. PLoSOne . 7: e31444. Google Scholar CrossRef Search ADS   Dunlap AS, Nielsen ME, Dornhaus A, Papaj DR. 2016. Foraging bumble bees weigh the reliability of personal and social information. Curr Biol . 26: 1195– 1199. Google Scholar CrossRef Search ADS PubMed  Galef BG. 2009. Strategies for social learning: testing predictions from formal theory. Adv Stud Behav . 39: 117– 151. Google Scholar CrossRef Search ADS   Giraldeau LA, Caraco T. 2000. Social foraging theory . Princeton, New Jersey: Princeton University Press. Giraldeau LA, Valone TJ, Templeton JJ. 2002. Potential disadvantages of using socially acquired information. Philos Trans R Soc Lond B Biol Sci . 357: 1559– 1566. Google Scholar CrossRef Search ADS PubMed  Grüter C, Czaczkes TJ, Ratnieks FL. 2011. Decision making in ant foragers (Lasius niger) facing conflicting private and social information. Behav Ecol Sociobiol . 65: 141– 148. Google Scholar CrossRef Search ADS   Heinrich B. 1979. Resource heterogeneity and patterns of movement in foraging bumblebees. Oecologia . 40: 235– 245. Google Scholar CrossRef Search ADS PubMed  Heyes C. 2012. What’s social about social learning?. J Comp Psychol . 126: 193. Google Scholar CrossRef Search ADS PubMed  Heyes C, Pearce JM. 2015. Not-so-social learning strategies. Proc R Soc Lond B Biol Sci . 282: 20141709. Google Scholar CrossRef Search ADS   Hoppitt W, Laland KN. 2013. Social learning: an introduction to mechanisms, methods, and models . Princeton, New Jersey: Princeton University Press. Google Scholar CrossRef Search ADS   Ings TC, Chittka L. 2009. Predator crypsis enhances behaviourally mediated indirect effects on plants by altering bumblebee foraging preferences. Proc R Soc Lond B Biol Sci . 276: 2031– 2036. Google Scholar CrossRef Search ADS   Jones PL, Ryan MJ, Chittka L. 2015. The influence of past experience with flower reward quality on social learning in bumblebees. Anim Behav . 101: 11– 18. Google Scholar CrossRef Search ADS   Kawaguchi L, Ohashi K, Toquenaga Y. 2006. Do bumble bees save time when choosing novel flowers by following conspecifics? Funct Ecol . 20: 239– 244. Google Scholar CrossRef Search ADS   Kawaguchi LG, Ohashi K, Toquenaga Y. 2007. Contrasting responses of bumble bees to feeding conspecifics on their familiar and unfamiliar flowers. Proc R Soc Lond B Biol Sci . 274: 2661– 2667. Google Scholar CrossRef Search ADS   Kendal RL, Coolen I. 2011. Adaptive trade-offs in the use of social and personal information. In Cognitive ecology II . Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Kendal RL, Coolen I, Laland KN. 2004. The role of conformity in foraging when personal and social information conflict. Behav Ecol . 15: 269– 277. Google Scholar CrossRef Search ADS   Laland KN. 2004. Social learning strategies. Anim Learn Behav . 32: 4– 14. Google Scholar CrossRef Search ADS   Leadbeater E, Chittka L. 2005. A new mode of information transfer in foraging bumblebees? Curr Biol . 15: R447– R448. Google Scholar CrossRef Search ADS PubMed  Leadbeater E, Chittka L. 2007. The dynamics of social learning in an insect model, the bumblebee (Bombus terrestris). Behav Ecol Sociobiol . 61: 1789– 1796. Google Scholar CrossRef Search ADS   Leadbeater E, Chittka L. 2009. Bumble-bees learn the value of social cues through experience. Biol Lett . 5: 310– 2. Google Scholar CrossRef Search ADS PubMed  Loukola OJ, Perry CJ, Coscos L, Chittka L. 2017. Bumblebees show cognitive flexibility by improving on an observed complex behavior. Science . 355: 833– 836. Google Scholar CrossRef Search ADS PubMed  Menzel R. 1985. Learning in honey bees in an ecological and behavioral context. Exp Behav Ecol . 31: 55– 74. Menzel R. 1990. Learning, memory, and “cognition” in honey bees. In: Kesner RP, Olton DS, editors. Comparative cognition and neuroscience. Neurobiology of comparative cognition . Hillsdale (NJ): Lawrence Erlbaum Associates. p. 237– 292. Pearce JM, Bouton MR. 2001. Theories of associative learning in animals. Ann Rev Psychol . 52: 111– 139. Google Scholar CrossRef Search ADS   Peitsch D, Fietz A, Hertel H, Souza J, Ventura DF, Menzel R. 1992. The spectral input systems of hymenopteran insects and their receptor-based colour vision. J Comp Physiol A . 170: 23– 40. Google Scholar CrossRef Search ADS PubMed  Powell MJD. 2009. The BOBYQA algorithm for bound constrained optimization without derivatives . Cambridge (England): Department of Applied Mathematics and Theoretical Physics, Technical Report NA2009/06. Rendell L, Boyd R, Cownden D, Enquist M, Eriksson K, Feldman MV, Fogarty L, Ghirlanda S, Lillicrap T, Laland KN. 2010. Why copy others? Insights from the social learning strategies tournament. Science . 328: 208– 213. Google Scholar CrossRef Search ADS PubMed  Rieucau G, Giraldeau LA. 2011. Exploring the costs and benefits of social information use: an appraisal of current experimental evidence. Philos Trans R Soc Lond B Biol Sci . 366: 949– 957. Google Scholar CrossRef Search ADS PubMed  Sheehan MJ, Sholler D, Tibbetts EA. 2014. Specialized visual learning of facial signals of quality in the paper wasp, Polistes dominula. Biol J Linnean Soc . 113: 992– 997. Google Scholar CrossRef Search ADS   Sheehan MJ, Tibbetts EA. 2011. Specialized face learning is associated with individual recognition in paper wasps. Science . 334: 1272– 1275. Google Scholar CrossRef Search ADS PubMed  Smolla M, Alem S, Chittka L, Shultz S. 2016. Copy-when-uncertain: bumblebees rely on social information when rewards are highly variable. Biol Lett . 12: 20160188. Google Scholar CrossRef Search ADS PubMed  Tibbetts EA. 2002. Visual signals of individual identity in the wasp Polistes fuscatus. Proc R Soc Lond B Biol Sci . 269: 1423– 1428. Google Scholar CrossRef Search ADS   Worden BD, Papaj DR. 2005. Flower choice copying in bumblebees. Biol Lett . 1: 504– 507. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Behavioral EcologyOxford University Press

Published: Jan 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off