Abstract Social learning is widespread but the causes for variation in the use of social versus private information are not always clear. Alongside adaptive explanations, suggesting that animals learn socially only when it is indeed adaptive to do so, it is also possible that the use of social learning is limited by mechanistic constraints. A common, but frequently overlooked challenge for social learning mechanisms is the need to allow learners to solve a problem through watching it being solved by others. This requires animals to be able to shift between contexts: from the context of the observed solution, to the context of the unsolved problem. For instance, for the social learning of cues associated with hidden food, an individual that merely sees a conspecific exploiting the food must, in the later absence of demonstrators or visible rewards, also learn to explore the cue for itself. Here, we show that this shift in context can indeed be difficult. In 2 experiments involving sand colors, house sparrows trained with hidden seeds learned to search for hidden seeds (based on food-color association) better than sparrows trained with exposed seeds. However, the latter showed color preference when tested with seeds exposed on both sand colors. These results demonstrate that context-specific learning makes it difficult to generalize reward-cue association from “exposed” to “hidden” conditions, which may explain why social learning is often more effective when it is based on socially facilitated active search (for hidden food), similar to that used in the context of independent foraging. Introduction Social learning, the process of learning through observation or interaction with others (Heyes 1994), can be adaptive in a variety of circumstances as it allows individuals to learn about the environment while avoiding the potential costs of individual exploration (Kendal et al. 2005; Shettleworth 2010). Evidence for social learning has been demonstrated in a large variety of species (reviewed by Galef and Giraldeau 2001; Hoppitt and Laland 2013), and the conditions under which social learning is likely to succeed have been the focus of extensive theoretical and empirical research (e.g. Galef and Giraldeau 2001; Danchin et al. 2004; Rendell et al. 2010; Rieucau and Giraldeau 2011). In particular, it has been suggested that because learning from others may not always be adaptive, social learning should also involve strategies determining when, what, and from whom to learn and, therefore, that information exchange within social groups may be subjected to transmission biases (Laland 2004; Kendal et al. 2005). However, alongside the triumphs of social learning research, it remained difficult to determine whether the occurrence and absence of social learning under different conditions represents adaptive strategies or mechanistic constraints. The variation in the performance of social learning seems especially striking in cases where animals can easily learn to solve a task independently, but fail miserably to do so when they join knowledgeable conspecifics. This situation is quite frequent in studies of social foraging, in which scrounging opportunities often seem to hinder learning of relevant information: scroungers often perform poorly in learning tasks, when compared with individual learners or with individuals that do not have an opportunity to scrounge (Giraldeau and Lefebvre 1987; Fragaszy and Visalberghi 1989; Beauchamp and Kacelnik 1991; Humle and Snowdon 2008; Ilan et al. 2013). It should be noted, however, that such apparent “failures” may not necessarily indicate maladaptive behavior. Considering that social information may be less reliable than personal information (Boyd and Richerson 1985; Giraldeau et al. 2002), it is frequently suggested that animals are simply prioritizing between different sources of information, applying social learning only when it is indeed adaptive to do so (Laland 2004). However, what may seem to be the result of strategic prioritization may also reflect mechanistic constraints that limit animal’s social learning success. Both explanations may be possible and they are also not mutually exclusive, as selection may operate on the mechanisms underlying social learning (Lotem and Halpern 2012; Heyes 2012). Attempts to explain, in mechanistic terms, the paradoxical instances in which individual learning thrives but social learning fails, often focus on the attributes of the social environment: the distracting presence of the demonstrator and its association with the reward (i.e. blocking or overshadowing: Giraldeau and Lefebvre 1987; Beauchamp and Kacelnik 1991), distractions caused by group settings (Lefebvre and Helder 1997), or a lack of motivation to learn due to the ability to rely on demonstrator’s task solution (Giraldeau and Templeton 1991). However, another, less explored possibility stems from the view of social learning as a biasing of individual learning processes by social stimuli (Galef and Giraldeau 2001; Heyes 2012; Leadbeater 2015). According to this view, the demonstrator’s behavior is important for facilitating an effective individual learning process. However, this facilitation may not be directly related to the presence of the demonstrator or to its behavior, but may simply result from the outcomes of its actions. Such outcomes have been usually considered as helpful for social learning, as in the case of “emulation” (Custance et al. 1999; Hoppitt and Laland 2008) and the spread of milk bottle opening in great tits (Sherry and Galef 1984), pine cone stripping by black rats (Aisner and Terkel 1992), or nectar robbing in bumblebees (Leadbeater and Chittka 2008). Alongside these examples where the outcomes of the demonstrator’s action can facilitate social learning, in other cases they may actually impose difficulties on the social learning process. When the demonstrator engages in solving the task, the task’s properties are often modified or come into sight in a way that might make it harder for the observer to learn to solve it for itself. One of the factors that is often modified by a demonstrator’s behavior is the context in which relevant cues appear with respect to rewards. As the demonstrator approaches the cue, it may expose food that was hidden of sight, or modify the cue’s physical properties. For example, it may handle and change the shape of an extractive foraging task (open cones and expose the pine seeds: Aisner and Terkel 1992; modify an artificial fruit, e.g. Caldwell and Whiten 2004), or expose food that was initially covered by soil or vegetation. Social learning in such cases can be analogous to “learning to solve a problem from seeing its solution”: In order to successfully acquire and implement relevant social information, an observer watching the demonstrator exploiting the uncovered food, needs to learn to explore the cue that was associated with it, in its original state and in the absence of conspecifics or visible rewards (i.e. to reverse-engineer from the observed end-state). The observer is thus required to be able to shift between contexts—from the context of the observed solution (e.g. exposed food) to the context of the unsolved problem (e.g. food hidden under cover). This shift may not be trivial as contextual information can affect both valuation and choice of stimuli in various settings (e.g. Shafir et al. 2002; Pompilio and Kacelnik 2010; Freidin and Kacelnik 2011), and even small adjustments of cues’ appearance can have detrimental effects on cognitive performance (e.g. Pretot et al. 2016). Though such modifications of cue presentation may have important implications for social learning, their role in determining social learning success is often overlooked (but see Root-Bernstein 2010; for a notable exception). A possible role for this context-shifting problem in social learning has been suggested by the results of our previous work on socially foraging house sparrows (Passer domesticus), which exhibit variable performance in social learning of food-related cues. First, we found that adult sparrows engaged in social foraging fail to learn food-related cues (the color of sand in which seeds are hidden) when they join conspecifics and scrounge on the food they uncovered, or even when they feed on seeds that were already exposed by their group members (Ilan et al. 2013). In both cases, the sand color was clearly visible and provided the visual background for the seeds that were eaten, but no social learning was exhibited. In a second recent study (Truskanov and Lotem 2015), we found that young sparrows can learn socially to prefer the sand color used by a parent demonstrator, but their learning is more effective when they need to search actively for the food (in the location hinted by the parent) rather than when the food is exposed by the parent. Although the results of these 2 studies may be explained by a general adaptive tendency to give more weight to information learned through active search (Ilan et al. 2013; Truskanov and Lotem 2015), they are also consistent with the idea that social learning may be more difficult when there is a need to shift from an observed context in which the reward is exposed alongside the cue, to independent search, where only the cue is visible. In other words, social learning may require a level of generalization that contradicts context-specific learning. In the present study, we sought to test these 2 alternative explanations by comparing the sparrows’ success in learning food-related cues (sand colors) when training and test conditions were either similar or different in context. We conducted 2 experiments in which we gave young sparrows individual learning tasks that simulate the 2 possible conditions they face during social foraging. One is when joining others involves active search for food that is hidden below the sand’s surface, and the other is when joining provides an opportunity to scrounge on seeds that were already exposed by others. In the first experiment, we tested whether the positive effect of active search on social learning (Truskanov and Lotem 2015) is not restricted to social learning, but rather a general tendency exhibited also during individual learning. In their training, sparrows were faced with 2 equally rewarding cues that were associated with either hidden or visible reward. We then tested whether they have learnt to prefer the cue associated with the hidden reward. In the second experiment, we studied the effect of contextual information. We presented sparrows with a simple learning task in which only one of 2 color cues contained rewards and divided them into 2 treatment groups experiencing different contexts (hidden vs. exposed seeds). We then tested the sparrows’ choices in 2 consecutive tests that were either similar to or different from their training context. We hypothesized that if learning of food-related cues is indeed challenged by the need to shift from the context of training to the context of independent foraging, learned preferences should be expressed in the test resembling the training conditions. That is, sparrows that were trained with exposed seeds would exhibit a preference for the learned cue in a test offering them exposed seeds on the background of the 2 sand colors, whereas sparrows that were trained with hidden seeds would exhibit a preferences in a test offering two sand colors but no visible reward. METHODS The experiments outlined below took place during the spring of 2013 (May–July), and fall of 2014 (September–October), with either hand-reared house sparrow fledglings (Experiment 1) or young yearlings (Experiment 2) originating from the house sparrow breeding colony of I. Meier Segal’s Garden of Zoological Research, Tel-Aviv University. Our study system included a set of individual cages (size: 47 × 38.5 cm) located in an aviary in the outdoor premises of the zoological garden. The cages were visually isolated from one another, and each of them contained a wooden foraging grid with 30 wells (2.5 cm diameter, 1.8 cm deep, and 8.5 cm distance between wells), bowls with an ad libitum supply of food (a mixture of seeds and mashed boiled eggs) and water, artificial foliage, and a sand bowl for bathing. A narrow flap door at the front bottom of the cage was used to insert and eject the foraging grid when necessary. Prior to the beginning of the experiments, the sparrows were gradually trained to search for millet seeds in the wells of the grid, by presenting them with grids containing seeds that are either exposed on top of natural sand, partially buried in the sand, or completely hidden below the sand’s surface. In Experiment 1, in which corrugated plastic sheets were used to cover parts of the grid during the experiment (see Figure 1), the sparrows were also trained to search the wells in the presence of these sheets. During this pre-experimental training, we gradually removed the sheets, unraveling additional parts of the grid, in a manner similar to that used in the experimental training sessions described below. Figure 1 View largeDownload slide Experimental setup. (a) Experiment 1: sparrows were trained on a grid in which wells filled with blue/red sand colors contained an equal payoff of 3 seeds. In the “hidden color” (here in blue; the roles of the colors were randomly switched between individuals), the seeds were buried below the sand’s surface. In the “exposed color,” the seeds were exposed on top of the sand. During training, two-thirds of the grid were initially covered with plastic sheets, that were gradually removed as the sparrows sampled both colors (in order to make sure that sampling is conducted in a mixed manner). In the test phase, the sparrows were tested in 2 consecutive tests on a seedless stripe of the grid, in which wells of both colors were alternately distributed to allow equal access from all locations. (b) Experiment 2: sparrows were trained on a grid in which wells of the target color (blue) always contained three seeds, whereas wells of the alternative color (red) contained no food. In the exposed treatment, the seeds were exposed on top of the blue sand, whereas in the hidden treatment they were hidden below the blue sand’s surface. In the test phase, sparrows of both treatments were tested in tests simulating the 2 different contexts. In the first test, the wells of the grid contained no seeds and the context appeared similar to the context experienced in the training of the hidden treatment. In the second test, both types of wells contained an equal payoff of three exposed seeds. Figure 1 View largeDownload slide Experimental setup. (a) Experiment 1: sparrows were trained on a grid in which wells filled with blue/red sand colors contained an equal payoff of 3 seeds. In the “hidden color” (here in blue; the roles of the colors were randomly switched between individuals), the seeds were buried below the sand’s surface. In the “exposed color,” the seeds were exposed on top of the sand. During training, two-thirds of the grid were initially covered with plastic sheets, that were gradually removed as the sparrows sampled both colors (in order to make sure that sampling is conducted in a mixed manner). In the test phase, the sparrows were tested in 2 consecutive tests on a seedless stripe of the grid, in which wells of both colors were alternately distributed to allow equal access from all locations. (b) Experiment 2: sparrows were trained on a grid in which wells of the target color (blue) always contained three seeds, whereas wells of the alternative color (red) contained no food. In the exposed treatment, the seeds were exposed on top of the blue sand, whereas in the hidden treatment they were hidden below the blue sand’s surface. In the test phase, sparrows of both treatments were tested in tests simulating the 2 different contexts. In the first test, the wells of the grid contained no seeds and the context appeared similar to the context experienced in the training of the hidden treatment. In the second test, both types of wells contained an equal payoff of three exposed seeds. Experiment 1: the effect of active search on cue preference The goal of the first experiment was to investigate the effect of active search (i.e. the need to dig actively in the sand when searching for food) on individual learning of food-related cues. We hypothesized that sparrows will prefer to search for food based on cues that were related to finding hidden seeds over cues that were associated with visible seeds. Such a result can be expected by both context-specific learning and by giving more weight to information learned during active search (see above). A result fitting this hypothesis would suggest that the enhanced effectiveness of social learning that involves active search (Truskanov and Lotem 2015) can be explained, at least in part, as a byproduct of some general characteristics of individual learning mechanisms and may not be related to the distracting presence of the demonstrator. The 21 sparrows that participated in this experiment were hand-reared during the spring of 2013 in 2 cohorts (similarly to the protocol described in: Katsnelson et al. 2008, 2011; Truskanov and Lotem 2015, 2017). Well after fledging, at the age of 37–39 days, we conducted an experiment in which they were trained on a grid containing 2 types of wells, filled with red/blue sand color that were associated with an equal payoff of 3 seeds. The 2 cues differed only in the context in which the seeds appeared: In one color, the seeds were exposed on top of the sand, whereas in the second color the seeds were hidden underneath the sand’s surface. The role attributed to each of the colors was counterbalanced across individuals. Each sparrow was trained in a set of 8 training sessions, followed by a set of 2 test sessions; all conducted over 3 consecutive days. A 1-h food deprivation preceded each set of 2 sessions. In days containing more than two sessions, the food bowl was returned to the cages for half an hour following the first pair of sessions, and then removed again for the next 1-h food deprivation. To prevent a sequential order effect that could have arisen if the sparrows picked up all the exposed seeds first, in each training session, the foraging grid was gradually exposed so that the sampling order of colors associated with hidden and exposed seeds was well mixed. This gradual exposure was facilitated by dividing the grid into 3 strips, each containing 10 foraging wells, in which 5 wells from each type were randomly distributed. At the beginning of the session, 2 of these strips were covered with fitting corrugated plastic sheets (see Figure 1a), that were gradually removed during the session as the sparrow’s exploitation of the wells in each strip reached the following pre-determined criteria: A minute after the beginning of exposure of each strip, the experimenter would approach the cage, and check whether the sparrow pecked at least 2 wells from each type. If this criterion was reached, the next cover was removed. Otherwise, the experimenter would repeat this process for up to 3 consecutive times (per strip), and then terminate the session. Overall, the total amount of pecks in intact wells that were conducted by the sparrows during the training ranged from 21 to 153 (this includes only sparrows that entered our final sample size; see further details in the following). Finally, note that approaching the cages up close during training is unlikely to have caused any stress-related responses, as the sparrows were hand-reared, and well-adjusted to the presence of experimenters and care takers. At the end of the training, each bird was tested in 2 consecutive and identical tests, in which seedless wells of both color types were distributed alternately in the first strip of the grid. This is a test in extinction, as the wells do not contain seeds and the association between the color cue and the food can undergo a gradual extinction process. The fact that the wells do not contain seeds prevents the subjects from learning or strengthening the association between the cue and the reward during the test itself. The locations of the wells were switched between the 2 testing sessions (so that wells that were blue on the first session were red on the second). In both test sessions, after the sparrow pecked 5 times (the maximum number of wells in each color), the grid was retracted. Due to human error, the grid was not retracted on time in some of the tests of the second testing session, allowing some sparrows to peck more than 5 times in the second test. These extra pecks were excluded from the analysis presented, that includes only the first 10 choices of each individual during the 2 test session. An additional analysis that included these extra pecks yielded similar results. In both training and test sessions, in cases in which the sparrows did not approach the grid during the session, additional sessions were added in which the grid was reintroduced again to the cage (for up to 3 consecutive times). Training and test sessions were video-recorded to allow analysis of individual experience and test preferences. Video analysis was conducted by a single trained research assistant that recorded the fledglings’ choices. Each visit to a well that included at least one peck, was defined as a foraging step, and classified according to well type. Repeated pecks at the same well were distinguished as different foraging steps if between them there was a clear shift in body orientation (i.e. the bird’s legs were directed away from the well for at least 3 seconds). The sparrows’ choices in the tests were counted and used to calculate a measure of learned color preference: the proportion of steps in hidden color wells, out of the total number of steps conducted in the 2 test sessions. We also analyzed the sparrows’ experience during the training, in order to account for the effect of potential differences in sampling between well types. In this analysis, we distinguished between foraging steps in wells that were still intact and steps in wells that were already visited in the same training session (termed “open” wells), and thus may no longer contain seeds. We included in our analysis only birds that pecked at least 6 times during testing and could thus provide a sufficiently reliable color preference score. This criterion was chosen a priori, as this is the minimal number of steps that allows a significant preference to be detected using a binomial sign test (see Ilan et al. 2013; Truskanov and Lotem 2015, for similar analyses). Following this criterion, we omitted from our analysis 7 birds: 6 individuals that did not peck at all during the tests, and one individual that pecked in only 3 wells and thus did not reach the 6 peck criterion. An analysis that included this individual yielded similar results. One bird was further omitted from our sample size due to errors in its training regime (wrong grid presentation). Our final sample size includes 13 sparrows. Experiment 2: the role of context in learning reward-cue associations The second experiment tested the ability of sparrows to associate cues with the presence of hidden or visible rewards, and to later search for these cues in conditions that are both similar to, and different from, the context they experienced during the learning phase. Difficulties to switch between these contexts (from visible to hidden rewards, and vice versa) would be consistent with the idea that social learning of food-related cues may be challenged by the need to shift from observation to independent search. Sixteen yearlings, hatched during the previous summer in the house-sparrow breeding colony of the zoological garden, were caught during the fall of 2015 in 4 experimental cohorts (4 individuals in each) and housed in our individual cage system (described above). As these sparrows were not hand-reared, in order to reduce possible stress, each bird was coupled with a companion, a yearling originating from our colony that was similarly caught and housed in an adjacent cage compartment allowing visual contact. These companions did not go through the same experimental training and are not part of our sample size but were also given foraging grids during the pre-experimental training (in which the sparrows were accustomed to searching for seeds in natural sand) and during the first test session (to encourage the focal birds to forage on their own grid). Based on our previous work (Ilan et al. 2013; Truskanov and Lotem 2015), we assumed that the color choice of the companion individuals would not affect the color choice of our focal individuals. This was indeed confirmed by the lack of correlation between the test choices of focal and companion sparrows (proportion of target color choice in the test: Spearman’s ρ = 0.086, P = 0.761). Moreover, even if subtle social effects did occur, they could not bias our experimental results because color preference by the companion individuals did not differ between the treatment groups of our focal individuals (Wilcoxon Rank Sums test: χ2 = 0.150, P = 0.698). Each experimental cohort lasted 4 days. The sparrows were first housed in their cages, familiarized with foraging on the experimental grid and trained to search for seeds in wells containing natural sand (in the manner described at the beginning of the general methods section). Following this early pre-experimental training, the experiment was conducted on the third day of each cohort and, in the fourth day, the sparrows were released back into the shared aviary of the colony. On the day of the experiment, the sparrows faced a simple learning task: they were presented with foraging grids in which only the 15 wells of the blue sand color (hereafter the “target color”) contained a fixed payoff of 3 seeds, whereas the 15 wells of the red sand color were empty of food (we did not alternate the role of colors this time in order to minimize variation and after confirming the lack of initial color bias in Experiment 1; see also Katsnelson et al. 2011). Sparrows were divided into 2 experimental treatments, differing in the context in which the seeds appeared during these sessions; in the “exposed treatment,” the seeds were exposed on top of the blue sand and, in the “hidden treatment,” they were hidden below its surface (Figure 1b). Each sparrow experienced 3 successive training sessions (a presentation of a new foraging grid), following 2 h of food deprivation. In order to reduce the amount of disturbances by the experimenters, the training sessions of the sparrows in each cohort were conducted simultaneously and each session was long, providing the sparrows with ample opportunity to approach the grid (the first session lasted 45 min, whereas the second and third sessions each lasted 30 min). At the end of the training phase, each sparrow was tested in a set of 2 tests in which wells of both colors were alternately distributed (as in a chessboard), allowing equal access from all locations. Each of these tests resembled a different training condition: In the first test, similar to the context of the hidden treatment (in which no seeds were visible during the training), both type of wells contained no seeds. In the second test, resembling the context of the exposed treatment, all the wells of the grid, from both types, contained an equal payoff of 3 seeds, laid visibly on the top of the sand (Figure 1b). The order of these tests was fixed, to avoid complex effects across tests and assuming that the effect of the first test on the second is less likely to bias our results. This is because the extinction process of the first test could only attenuate or weaken learned differences but cannot generate artificial differences that are not related to the training regime. This issue will be discussed further in light of our results (see Results and Discussion sections below). Finally, note that in the exposed context test, there is presumably no good reason to prefer one of the cues, as all the wells contain seeds that can be directly and easily pecked. Nevertheless, a preference, if developed, would be highly informative in terms of the learning mechanism involved. The behavioral data were analyzed separately for each test type and in the same method described in Experiment 1. Video analysis was conducted in a blind manner (i.e. with no knowledge of the treatment group to which each sparrow belonged). In order to be able to compare the sparrows’ performance in the 2 tests, we divided the data into 3 test phase categories. This division of the data was essential, because the difference between the 2 tests quickly disappears; as the sparrows consumed the seeds available in the exposed context test, they were soon faced with seeds left only on the surface of their less preferred sand color and, after consuming them, they faced a seedless grid that is basically similar to that of the hidden context test. To account for these expected changes, the following 3 phases were a priori determined: 1) Initial phase (phase 1—first 15 steps) before food depletion in the exposed seeds condition was likely to bias the color choice of the tested individuals (recall that each grid contained 30 feeding wells, 15 of each color); 2) Intermediate phase (phase 2—steps 16–30), during which the presence of seeds in the exposed seeds condition could bias color preference (after the seeds were removed from the initially preferred color); and 3) End of test (phase 3—peck 31 and above), after most seeds were consumed and the test of the exposed seed condition resembles the hidden seed condition. The sparrows’ color preferences were calculated separately for each test phase. Statistical analysis Statistical analysis was performed using JMP 10. As our data did not meet the assumptions of parametric testing and could not be effectively transformed, we took a nonparametric approach in our analysis. We used One Sample Wilcoxon Signed Rank tests for testing the color preferences of treatment groups, and Wilcoxon Signed Rank tests and Wicoxon Rank Sum tests for within and between groups comparisons, respectively. All statistical tests are 2-tailed unless reported otherwise. Because Wilcoxon Rank tests assume symmetric distribution of the data around the median, we also confirmed our analysis by using permutation tests in R.3.4.2 (Team 2016), which are independent of assumptions regarding the distribution of the data. The results of these analyses were virtually the same and are provided in the electronic supplementary material (Supplementary Table S1). Finally, to test whether the color preference of individual sparrows was significantly different than random, we used 2-tailed binomial sign-tests. In the analysis of Experiment 1, we initially tested whether the 2 cohorts differed in their test preferences (proportion of hidden color choice) and, as no significant differences were found between the cohorts (Wilcoxon Rank Sums test: χ2 = 1.418, P = 0.234, and N = 8 and 5 for the first and second cohorts, respectively), we pooled the data in subsequent analysis. Due to technical problems with video recording, for 3 of the birds in this experiment, 1 training session was not recorded. The result reported in this manuscript is of an analysis that includes these birds. However, a similar analysis of the training data in which these birds were omitted yielded similar results that were still statistically significant. In the analysis of Experiment 2, sparrows that pecked less than 6 times in a test phase did not enter our analysis of this particular phase (see above), but analyzing the same data while including these individuals yielded similar results. Ethical note This study was carried out under animal care permit no. L-12–028 of Tel-Aviv University. RESULTS Experiment 1: the effect of active search on color preference Analysis of the sparrows’ color preferences during the test phase (the proportion of pecks in the hidden color, out of the total number of pecks performed) reveals that despite the equal payoff offered by the 2 colors, the majority of the sparrows (11 out of 13) developed a preference for the color that was associated with hidden seeds (Figure 2a; One Sample Wilcoxon signed rank test; n =13, W = 34.5, P = 0.013). For 7 of the sparrows, the preference of the hidden color was significantly different from random choice (Binomial Sign Tests: P < 0.05). Only 1 sparrow showed a significant preference for the color in which the seeds were exposed during the training (Binomial Sign Test: P = 0.021). Figure 2 View largeDownload slide Sparrows’ choices in the test phase of Experiment 1. (a) The proportion of choice of the hidden color (in gray) and the exposed color (in white) by each of the sparrows that participated in the tests. Asterisks denote individuals with significant preferences (sign test P < 0.05). (b) The sparrows’ preference for the hidden color plotted against their training experience: the proportion of hidden color wells sampled in the training phase (out of all the intact wells sampled). Dots located on the left side of the plot denote individuals that sampled more exposed color wells than hidden color ones in the training. Figure 2 View largeDownload slide Sparrows’ choices in the test phase of Experiment 1. (a) The proportion of choice of the hidden color (in gray) and the exposed color (in white) by each of the sparrows that participated in the tests. Asterisks denote individuals with significant preferences (sign test P < 0.05). (b) The sparrows’ preference for the hidden color plotted against their training experience: the proportion of hidden color wells sampled in the training phase (out of all the intact wells sampled). Dots located on the left side of the plot denote individuals that sampled more exposed color wells than hidden color ones in the training. It is worth noting that the sparrows’ preference for the color of hidden seeds cannot be attributed to differences in the amount of successful experience with the 2 colors during the training. In fact, during their training, the sparrows experienced more successful events (i.e. pecks in intact wells) in the color associated with exposed seeds than in the color associated with hidden seeds (Figure 2b; almost all data points are left of the vertical 0.5 dashed line, Wilcoxon Signed Rank 2-tailed test: n = 13, S = 35.5, P = 0.009). Experiment 2: the role of context in learning reward-cue associations Activity during the test During the hidden context test, sparrows of the hidden treatment group (that were trained with hidden seeds) were much more active than sparrows of the exposed treatment group, as indicated by the number of pecks they performed on the grid (Figure 3a, Wilcoxon Rank Sums test: χ2 = 11.327, P = 0.0008, n = 8 in both treatment groups). In the exposed context test, in which visible seeds were placed on both colors, the difference between groups in activity levels was smaller and no longer significant (Figure 3a, Wilcoxon Rank Sums test: χ2 = 3.595, P = 0.058, n = 8 in both treatment groups). Figure 3 View largeDownload slide Sparrows’ performance in the test phase of Experiment 2. (a) Activity levels in the test phase (the number of steps performed in each test), in the 2 experimental groups. (b) Target color preference (proportion of steps in the target color) of each treatment in the initial phase of each test. In both plots, the exposed treatment appears in white and the hidden treatment in gray, and data are denoted as mean ± SE. N = 8 in all cases, except for the hidden context test of the exposed treatment in (b), in which n = 7. Figure 3 View largeDownload slide Sparrows’ performance in the test phase of Experiment 2. (a) Activity levels in the test phase (the number of steps performed in each test), in the 2 experimental groups. (b) Target color preference (proportion of steps in the target color) of each treatment in the initial phase of each test. In both plots, the exposed treatment appears in white and the hidden treatment in gray, and data are denoted as mean ± SE. N = 8 in all cases, except for the hidden context test of the exposed treatment in (b), in which n = 7. Color preference in the initial test phase In the initial phases of the 2 tests, when intact wells of both colors were still readily available (see Methods), the sparrows’ proportion of pecks in the target color (hereafter: target color preference) was clearly affected by whether the test context fitted the training context. When tested in conditions resembling their training context, both the birds of the hidden treatment and the birds of the exposed treatment exhibited a highly significant preference for the target color. The birds of the hidden treatment exhibited this preference in the hidden context test, and the birds of the exposed treatment exhibited it in the exposed context test (Figure 3b; Table 1A and D). When tested in conditions that did not resemble their training context, the birds of the exposed treatment failed to show a preference for the target color (Figure 3b; Table 1B; exposed treatment in hidden context test), and the birds of the hidden treatment exhibited a weak though significant preference (Figure 3b; Table 1C; hidden treatment in exposed context test). Table 1 Analysis of target color preferences (proportion of pecks in the blue target color) in the initial phases of the tests in Experiment 2 Analysis Test (initial phase) Treatment group N W P A Hidden context test (Test 1) Hidden treatment 8 18 0.008 B Hidden context test (Test 1) Exposed treatment 7 7.5 0.156 C Exposed context test (Test 2) Hidden treatment 8 15.5 0.039 D Exposed context test (Test 2) Exposed treatment 8 18 0.008 Analysis Test (initial phase) Treatment group N W P A Hidden context test (Test 1) Hidden treatment 8 18 0.008 B Hidden context test (Test 1) Exposed treatment 7 7.5 0.156 C Exposed context test (Test 2) Hidden treatment 8 15.5 0.039 D Exposed context test (Test 2) Exposed treatment 8 18 0.008 The rows of the table report the results of Wilcoxon signed rank (2-tailed) tests that were used to test whether the sparrows’ preference for the target color is significantly different from the 0.5 preference score expected by chance. These analyses were performed separately for each test and treatment group. View Large Table 1 Analysis of target color preferences (proportion of pecks in the blue target color) in the initial phases of the tests in Experiment 2 Analysis Test (initial phase) Treatment group N W P A Hidden context test (Test 1) Hidden treatment 8 18 0.008 B Hidden context test (Test 1) Exposed treatment 7 7.5 0.156 C Exposed context test (Test 2) Hidden treatment 8 15.5 0.039 D Exposed context test (Test 2) Exposed treatment 8 18 0.008 Analysis Test (initial phase) Treatment group N W P A Hidden context test (Test 1) Hidden treatment 8 18 0.008 B Hidden context test (Test 1) Exposed treatment 7 7.5 0.156 C Exposed context test (Test 2) Hidden treatment 8 15.5 0.039 D Exposed context test (Test 2) Exposed treatment 8 18 0.008 The rows of the table report the results of Wilcoxon signed rank (2-tailed) tests that were used to test whether the sparrows’ preference for the target color is significantly different from the 0.5 preference score expected by chance. These analyses were performed separately for each test and treatment group. View Large Color preference in the second and third test phases Analysis of later phases of the tests reveals additional interesting patterns (see Figure 4 and Supplementary Figure S1). In the hidden context test, where the wells contained no seeds, the strong preference of the target color initially exhibited by the hidden treatment birds decreased slightly in later phases (see the black dots in Figure 4a; the reduction between phase 1 and 3 is statistically significant: Wilcoxon Signed Rank, n = 8, S = −12, P = 0.023). This reduction in preference is not surprising as the hidden context test was a test in extinction (i.e. the wells of the grid contain no seeds) and is therefore expected to gradually extinguish learned preferences. Nevertheless, the strength of this preference remained rather high (mean = 0.85 ± 0.035) despite the accumulating number of nonrewarded pecks. Figure 4 View largeDownload slide Target color preference of sparrows of the exposed treatment (○) and sparrows of the hidden treatment (●) in different test phases of (a) the hidden context test (Test 1) and (b) the exposed context test (Test 2). Phase 1 includes the first 15 steps of the test (in which most of the wells of the exposed context test contained seeds). Phase 2 includes steps 15–30 (in which the grid of the exposed test still contained seeds, but half of it is assumed to be depleted). Phase 3 includes all the steps conducted after the grid of the exposed context test was likely to be depleted, making the context of this phase similar to the hidden context. Points denote means ± SE. In the exposed treatment, many of the sparrows ceased pecking rather quickly in the hidden context test. The sample size for each of the points of this group in plot (a) is therefore n = 7, n = 5, and n = 2 in the first, second, and third phases, respectively. The rest of the points in both plots are based on the average of n = 8 sparrows. Figure 4 View largeDownload slide Target color preference of sparrows of the exposed treatment (○) and sparrows of the hidden treatment (●) in different test phases of (a) the hidden context test (Test 1) and (b) the exposed context test (Test 2). Phase 1 includes the first 15 steps of the test (in which most of the wells of the exposed context test contained seeds). Phase 2 includes steps 15–30 (in which the grid of the exposed test still contained seeds, but half of it is assumed to be depleted). Phase 3 includes all the steps conducted after the grid of the exposed context test was likely to be depleted, making the context of this phase similar to the hidden context. Points denote means ± SE. In the exposed treatment, many of the sparrows ceased pecking rather quickly in the hidden context test. The sample size for each of the points of this group in plot (a) is therefore n = 7, n = 5, and n = 2 in the first, second, and third phases, respectively. The rest of the points in both plots are based on the average of n = 8 sparrows. In the exposed context test, where seeds were initially exposed on the sand of each well, a different pattern emerged. The sparrows of the exposed treatment group (see the white dots of Figure 4b), significantly preferred the target color in the initial phase of the test (as described earlier) but dropped their preference in the second and third phases when seeds were no longer available on their preferred color. On the other hand, the sparrows of the hidden treatment group (Figure 4b, black dots) that exhibited only a weak preference for the target color during the initial phase of the test, decreased it even further in the second phase, but increased it in the third phase where their preference of the target color became highly significant (One Sample Wilcoxon Signed Rank test: n = 8, W = 18, P = 0.008). Note that at this third stage (after 30 choices in a foraging grid containing 30 feeding wells with exposed seeds) most of the seeds on the grid were already depleted, so the context was similar to the context of hidden seeds, which is the context experienced by these sparrows during the training period. Discussion The results of our first experiment show that young house sparrows prefer to search for food using cues that were associated with active search conditions. When faced with 2 equally rewarding cues (sand colors) that differed only in the context in which the seeds appeared, the sparrows preferred to search the cue that was associated with hidden seeds and active search (the need to dig in order to uncover the food) rather than the cue associated with seeds that were exposed and could be picked directly. This result complements our previous work, in which we showed that social learning is more effective when it is mediated through active search (Truskanov and Lotem 2015). It further suggests that the reason that sparrows sometimes fail to learn socially when scrounging on food exposed by other individuals (Ilan et al. 2013; Truskanov and Lotem 2015) may not be related to social circumstances or to the presence of a demonstrator (e.g. Giraldeau and Lefebvre 1987; Beauchamp and Kacelnik 1991; Giraldeau and Templeton 1991), but rather be related to some general features of associative learning processes. Yet, the results of the first experiment cannot reveal whether this phenomenon is due to a tendency to give more weight to information learned during active search, or rather a result of context-specific learning (i.e. searching for hidden seeds using only cues previously associated with hidden seeds and ignoring cues learned in association with exposed seeds). A role for context-dependent learning is strongly suggested by the results of the second experiment. First, the conditions the sparrows faced during training strongly influenced their pecking persistence and activity levels in the test phase. In both tests sparrows of the hidden treatment pecked persistently despite repeated disappointments, while sparrows of the exposed treatment gave up rather quickly when no seeds were in sight (Figure 3a). These observed differences in activity levels between groups make sense: while sparrows trained with hidden seeds expected that the seeds will be hidden and should be searched for, sparrows of the exposed treatment were trained under conditions in which the seeds appeared only on top of the sand and digging was not rewarded. Thus, the context in which the food appeared during the learning phase affected the motivation to persistently search cues when rewards are not visible. Although this effect is not directly related to the sparrows’ success in learning the color cues, it has important implications. Persistence and perseverance have been suggested to play a key role in innovative problem solving and can determine success in using novel food sources (Tebbich et al. 2010; Morand-Ferron et al. 2011; Thornton and Samson 2012; Benson-Amram and Holekamp 2012; Cauchard et al. 2013). The need to shift between contexts may thus affect performance through a reduction in persistence even when a color-cue association was indeed learned. In other words, in a natural setting, sparrows that learned to associate a certain cue with exposed seeds, may fail to find seeds hidden in the vicinity of the same cue. This failure may simply occur because their expectation to see the seeds reduces their motivation to search persistently. Second, the learning process exhibited by the sparrows was strongly context-specific. The birds in each group showed the strongest target-color preference when the test situation was similar to the context they experienced in the training phase (Figure 3b). Most importantly, birds of the exposed treatment group that clearly learned to prefer the target color when tested in the exposed context test, failed to prefer it just before, when tested in the hidden context test. This result suggests that scrounging on food exposed by others may fail to facilitate social learning simply because learning in such cases may be too context-specific. Additional evidence for a contextual effect lies in the fact that the differences in preference between the 2 groups were not limited to initial test phases. In the exposed context test, after the seeds were depleted and were no longer visible (at the third phase of the test), the context of the test again resembled that of hidden seeds and active search. At this stage, sparrows of the hidden treatment that were again faced with the context with which they have been trained, increased their preference for the target color (Figure 4b), whereas sparrows of the exposed treatment did not. This result cannot be explained by different foraging success in the 2 colors as the sparrows of the 2 groups were equally successful in picking all the exposed seeds from both colors. Furthermore, although during the second phase the hidden treatment birds had a bit more seeds left to pick from the target color (in comparison with the exposed treatment, birds that picked more seeds from the target color already in the initial phase), this recent experience cannot explain their sharp increase in target color preference during the third phase. The reason for that is that if such a small difference under the exposed condition (see phase 2 in Figure 4b) could result in such an effect, then, the exposed training itself (during which success was limited to the target color) should have certainly allowed the exposed treatment birds to learn to prefer the target color in the hidden context test. The fact that this was not the case (see Figure 3b) suggest that this alternative interpretation is clearly improbable. Finally, the increase in target-color preference by the hidden treatment birds at the end of the exposed context test is not only consistent with context-dependent learning, but can also rule out the possibility that their weak preference during the initial phase of this test was a result of extinction. Admittedly, the effect of the context in the second experiment was not symmetrical. The target color preference exhibited by the hidden treatment group in the hidden context test was stronger than that exhibited by the exposed treatment group in the exposed context test (Figure 3b: 0.98 ± 0.01 vs. 0.72 ± 0.05). However, this asymmetry is not surprising as the presence of visible seeds on both colors in the exposed context test was expected to dilute acquired preferences. In fact, given that the birds could see the seeds on both colors it was somewhat surprising for us that they still exhibited such a significant preference for the target color. It clearly demonstrates that they did learn the color cue. It is worth noting that several studies have demonstrated a negative effect of visible rewards on learning (Boysen and Berntson 1995; Bemtson and Mukobi 2001; Vlamings et al. 2006; Salwiczek et al. 2012; but see Murray et al. 2005). Capuchin monkeys, for example, failed to learn an “ephemeral food priority task” in which the food was exposed on top of colored plates (Salwiczek et al. 2012), but changing the properties of the task such that the food was hidden by the cues (colored cups), led to significant improvement in learning (Pretot et al. 2016). The negative effect of visible reward on learning is often attributed to impulsiveness, the difficulty to inhibit prepotent responses when rewards are visible (Boysen and Berntson 1995; Bemtson and Mukobi 2001; Vlamings et al. 2006; Salwiczek et al. 2012; Pretot et al. 2016). Though it is possible that impulsivity played some role in reducing color preference in our exposed seeds condition, it is clear that it did not prevent learning. First, the sparrows of the exposed seed treatment did show a preference for the target color in the exposed seeds test, even when the seeds were available on both colors and indifferently consuming them from both well types would be potentially easier and more consistent with an impulsivity hypothesis (some of these birds even show a significant individual preference for the target color, see Supplementary Figure S2, which means that they skipped and ignored seeds exposed in front of their beak). Second, few of these birds also showed a significant preference for the target color in the hidden seed test (Supplementary Figure S2), suggesting that not only that learning during the exposed condition was possible, but that, in some cases, it could also be generalized to the hidden context. Similar evidence from previous work also suggest that learning cues that appeared in exposed context and using them in a hidden context is feasible as long as conflicting information from the hidden context is not provided (Truskanov and Lotem 2015). Thus, although shifting from the exposed seeds training context to the context of hidden seeds may be difficult for the sparrows, the learning of color-food association with visible reward was clearly feasible for them and was not prevented by impulsivity. The experimental results discussed so far clearly demonstrate that the difficulty in shifting from the context of exposed-seeds to the context of hidden seeds can explain why learning through active search is more effective in facilitating future independent foraging for hidden food. However, these results cannot preclude additional contributing factors. It is possible that information acquired through active search received more weight in memory or was learned with a greater associative strength than information gained incidentally through passive observation (Truskanov and Lotem 2015). Several possible mechanisms may generate such an effect and should be considered. First, active search may be related to the well-known phenomenon in which “cost increases preference” (Kacelnik and Marsh 2002). Although this possibility cannot be ruled out completely, we view it as a less likely alternative because uncovering hidden seeds seemed to require very little effort (and time) in our study and was unlikely to be related to major differences in state. A second possibility, which relies on a fundamental aspect of associative learning, is that active search allows the updating of information based on prediction errors such as in the Rescorla Wagner model (Rescorla and Wagner 1972) in which updating is based on the extent of “surprise,” or deviance from current expectations (also see Shettleworth 2010). According to this explanation, uncovering hidden seeds may indeed involve deviations from previous expectation, whereas when the seeds are exposed, no predictions errors are likely to occur. Finally, a third possibility is that information gained through active search has evolved to receive more weight in memory because it is generally more reliable than incidental observations (Ilan et al. 2013; Truskanov and Lotem 2015). These second and third hypotheses open an avenue for future work—they can be further investigated by providing sparrows with a choice between cues associated with hidden or exposed seeds, and a cue in which the seeds are visible (i.e. no surprise or prediction errors) but investment of effort is still required in order to reach them (for instance, by covering seeds with a transparent cover). Importantly, although the difficulty in shifting from exposed to hidden contexts was demonstrated here in a study of individual learning, it should be clear that this fundamental feature of associative learning can also pose a challenge to social learning. This is because under many circumstances this is precisely the shift that must be made in order to learn to solve a problem through watching it being solved by others. Moreover, though the present work provides a relatively simple example of shifting between simple contexts, in the social learning of more complicated tasks the shift in context may be far more challenging. For example, efficient pine cone opening is a behavior that is vertically transmitted from mothers to offspring in the black rat (Rattus rattus) (reviewed by Terkel 1996). The successful social learning of pups seems to be mediated at least to some extent, by exposure to partially processed pine cones. Naïve adult rats, on the other hand, fail to learn from conspecifics, as well as from being exposed to fully processed cones. They can be trained to open pine cones, but this involves a long shaping process in which they are repeatedly presented with cones at different stages of processing. It seems that maternal behavior in this case, facilitates this contextual shift by the naive pups: it exposes them to different stages of task’s solution, and by that helps them to successfully shift from the end state of the task—a fully processed pine cone, to the initial handling required when an intact pine cone is approached. Finally, although the difficulty in shifting across contexts may first appear as a constraint, it may also be viewed as an adaptation. In theory, if 2 different contexts can be perceived by the animal as sufficiently similar, generalization across them will become natural. However, over-generalization may be costly. In natural environments, numerous stimuli may be present in different contexts, and the degree of generalization employed by learners conveys a tradeoff as generalizing too broadly might be risky and ineffective (e.g. Thawley and Langkilde 2017). Thus, contextual constraints that may seem paradoxical in the lab may in fact be powerful tools that help animals to avoid generalization errors in nature. The extent to which animals are able to generalize can vary between species, environments and contexts (Ghirlanda and Enquist 2003), and is likely to be shaped by natural selection (e.g. Lotem and Halpern 2012; Kolodny et al. 2015a; Kolodny et al. 2015b). In the case of social learning, the evolution of broader generalization parameters may be possible, but restricting them may form a mechanism through which selection can act on social learning strategies, directing them to operate only when they are likely to be adaptive (e.g. Laland 2004). SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online. Funding This work was supported by the Israel Science Foundation (grant numbers 1312/11, 871/15). Data accessibility Analyses reported in this article can be reproduced using the data provided by (Truskanov and Lotem 2018). We thank Julie Morand-Ferron for a fruitful discussion and an inspiring suggestion, Shira Landman, Sigal Avraham, Gal Shalev, and Yafit Brenner for lab and technical assistance, and Alex Slavenko for his help with statistical analysis. References Aisner R, Terkel J. 1992. Ontogeny of pine-cone opening behavior in the black rat, Rattus rattus. Anim Behav . 44: 327– 336. Google Scholar CrossRef Search ADS Beauchamp G, Kacelnik A. 1991. Effects of the knowledge of partners on learning rates in zebra finches Taeniopygia guttata. Anim Behav . 41: 247– 253. Google Scholar CrossRef Search ADS Bemtson GG, Mukobi KL. 2001. Size matters: impact of item size and quantity on array choice by chimpanzees (Pan troglodytes). J Comp Psychol . 115: 106– 110. Google Scholar CrossRef Search ADS PubMed Benson-Amram S, Holekamp KE. 2012. Innovative problem solving by wild spotted hyenas. Proc Biol Sci . 279: 4087– 4095. Google Scholar CrossRef Search ADS PubMed Boyd R, Richerson PJ. 1985. Culture and the evolutionary process . Chicago (IL): University of Chicago Press. Boysen ST, Berntson GG. 1995. Responses to quantity: perceptual versus cognitive mechanisms in chimpanzees (Pan troglodytes). J Exp Psychol Anim Behav Process . 21: 82– 86. Google Scholar CrossRef Search ADS PubMed Caldwell CA, Whiten A. 2004. Testing for social learning and imitation in common marmosets, Callithrix jacchus, using an artificial fruit. Anim Cogn . 7: 77– 85. Google Scholar CrossRef Search ADS PubMed Cauchard L, Boogert NJ, Lefebvre L, Dubois F, Doligez B. 2013. Problem-solving performance is correlated with reproductive success in a wild bird population. Anim Behav . 85: 19– 26. Google Scholar CrossRef Search ADS Custance D, Whiten A, Fredman T. 1999. Social learning of an artificial fruit task in capuchin monkeys (Cebus apella). J Comp Psychol . 113: 13– 23. Google Scholar CrossRef Search ADS Danchin E, 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 Fragaszy DM, Visalberghi E. 1989. Social influences on the acquisition of tool-using behaviors in tufted capuchin monkeys (Cebus apella). J Comp Psychol . 103: 159– 170. Google Scholar CrossRef Search ADS PubMed Freidin E, Kacelnik A. 2011. Rational choice, context dependence, and the value of information in European starlings (Sturnus vulgaris). Science . 334: 1000– 1002. Google Scholar CrossRef Search ADS PubMed Galef BGJr, Giraldeau LA. 2001. Social influences on foraging in vertebrates: causal mechanisms and adaptive functions. Anim Behav . 61: 3– 15. Google Scholar CrossRef Search ADS PubMed Ghirlanda S, Enquist M. 2003. A century of generalization. Anim Behav . 66: 15– 36. Google Scholar CrossRef Search ADS Giraldeau LA, Lefebvre L. 1987. Scrounging prevents cultural transmission of food-finding behavior in pigeons. Anim Behav . 35: 387– 394. Google Scholar CrossRef Search ADS Giraldeau LA, Templeton JJ. 1991. Food scrounging and diffusion of foraging skills in pigeons, Columba livia: the importance of tutor and observer rewards. Ethology . 89: 63– 72. Google Scholar CrossRef Search ADS 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 Heyes CM. 1994. Social learning in animals: categories and mechanisms. Biol Rev Camb Philos Soc . 69: 207– 231. Google Scholar CrossRef Search ADS PubMed Heyes C. 2012. What’s social about social learning? J Comp Psychol . 126: 193– 202. Google Scholar CrossRef Search ADS PubMed Hoppitt W, Laland KN. 2008. Social processes influencing learning in animals: a review of the evidence. Adv Stud Behav . 38: 105– 165. Google Scholar CrossRef Search ADS Hoppitt W, Laland K. 2013. Social learning: an introduction to mechanisms, methods and models . Princeton (NJ): Princeton University Press. Google Scholar CrossRef Search ADS Humle T, Snowdon CT. 2008. Socially biased learning in the acquisition of a complex foraging task in juvenile cottontop tamarins, Saguinus oedipus. Anim Behav . 75: 267– 277. Google Scholar CrossRef Search ADS Ilan T, Katsnelson E, Motro U, Feldman MW, Lotem A. 2013. The role of beginner’s luck in learning to prefer risky patches by socially foraging house sparrows. Behav Ecol . 24: 1398– 1406. Google Scholar CrossRef Search ADS PubMed Kacelnik A, Marsh B. 2002. Cost can increase preference in starlings. Anim Behav . 63: 245– 250. Google Scholar CrossRef Search ADS Katsnelson E, Motro U, Feldman MW, Lotem A. 2008. Early experience affects producer-scrounger foraging tendencies in the house sparrow. Anim Behav . 75: 1465– 1472. Google Scholar CrossRef Search ADS Katsnelson E, Motro U, Feldman MW, Lotem A. 2011. Individual-learning ability predicts social-foraging strategy in house sparrows. Proc Biol Sci . 278: 582– 589. Google Scholar CrossRef Search ADS PubMed Kendal RL, Coolen I, van Bergen Y, Laland KN. 2005. Trade-offs in the adaptive use of social and asocial learning. In: Slater PJB, Snowdon CT, Brockmann HJ, Roper TJ, Naguib M, editors. Advances in the Study of Behavior . Vol 35. San Diego: Elsevier Academic Press Inc. p. 333– 379. Google Scholar CrossRef Search ADS Kolodny O, Creanza N, Feldman MW. 2015a. Evolution in leaps: the punctuated accumulation and loss of cultural innovations. Proc Natl Acad Sci USA . 112: E6762– E6769. Google Scholar CrossRef Search ADS Kolodny O, Edelman S, Lotem A. 2015b. Evolved to adapt: a computational approach to animal innovation and creativity. Curr Zool . 61: 350– U179. Google Scholar CrossRef Search ADS Laland KN. 2004. Social learning strategies. Learn Behav . 32: 4– 14. Google Scholar CrossRef Search ADS PubMed Leadbeater E. 2015. What evolves in the evolution of social learning? J Zool . 295: 4– 11. Google Scholar CrossRef Search ADS Leadbeater E, Chittka L. 2008. Social transmission of nectar-robbing behaviour in bumble-bees. Proc Biol Sci . 275: 1669– 1674. Google Scholar CrossRef Search ADS PubMed Lefebvre L, Helder R. 1997. Scrounger numbers and the inhibition of social learning in pigeons. Behav Processes . 40: 201– 207. Google Scholar CrossRef Search ADS PubMed Lotem A, Halpern JY. 2012. Coevolution of learning and data-acquisition mechanisms: a model for cognitive evolution. Philos Trans R Soc Lond B Biol Sci . 367: 2686– 2694. Google Scholar CrossRef Search ADS PubMed Morand-Ferron J, Cole EF, Rawles JEC, Quinn JL. 2011. Who are the innovators? A field experiment with 2 passerine species. Behav Ecol . 22: 1241– 1248. Google Scholar CrossRef Search ADS Murray EA, Kralik JD, Wise SP. 2005. Learning to inhibit prepotent responses: successful performance by rhesus macaques, Macaca mulatta, on the reversed-contingency task. Anim Behav . 69: 991– 998. Google Scholar CrossRef Search ADS Pompilio L, Kacelnik A. 2010. Context-dependent utility overrides absolute memory as a determinant of choice. Proc Natl Acad Sci USA . 107: 508– 512. Google Scholar CrossRef Search ADS PubMed Pretot L, Bshary R, Brosnan SF. 2016. Factors influencing the different performance of fish and primates on a dichotomous choice task. Anim Behav . 119: 189– 199. Google Scholar CrossRef Search ADS Rendell L, Boyd R, Cownden D, Enquist M, Eriksson K, Feldman MW, 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 Rescorla RA, Wagner AR. 1972. A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement. In: Black AH, Prokasy WF, editors. Classical conditioning II: current research and theory . New York: Appleton-Century-Crofts. 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 Root-Bernstein M. 2010. The role of context, colour and location cues in socially learned novel food source preferences in starlings, Sternus vulgaris. Behav Processes . 84: 608– 616. Google Scholar CrossRef Search ADS PubMed Salwiczek LH, Prétôt L, Demarta L, Proctor D, Essler J, Pinto AI, Wismer S, Stoinski T, Brosnan SF, Bshary R. 2012. Adult cleaner wrasse outperform capuchin monkeys, chimpanzees and orang-utans in a complex foraging task derived from cleaner–client reef fish cooperation. PLoS One . 7: e49068. Google Scholar CrossRef Search ADS PubMed Shafir S, Waite TA, Smith BH. 2002. Context-dependent violations of rational choice in honeybees (Apis mellifera) and gray jays (Perisoreus canadensis). Behav Ecol Sociobiol . 51: 180– 187. Google Scholar CrossRef Search ADS Sherry DF, Galef BG. 1984. Cultural transmission without imitation - milk bottle opening by birds. Anim Behav . 32: 937– 938. Google Scholar CrossRef Search ADS Shettleworth SJ. 2010. Cognition, evolution, and behavior . 2nd ed. New York: Oxford University Press. Team RC. 2016. R: a language and environment for statistical computing . Vienna (Austria): R Foundation for Statistical Computing. Tebbich S, Sterelny K, Teschke I. 2010. The tale of the finch: adaptive radiation and behavioural flexibility. Philos Trans R Soc Lond B Biol Sci . 365: 1099– 1109. Google Scholar CrossRef Search ADS PubMed Terkel J. 1996. Cultural transmission of feeding behavior in the black rat (Rattus rattus). In: Heyes CM, Galef BGJr, editors. Social learning in animals: the roots of culture . Elsevier: London, UK. p. 17– 47. Google Scholar CrossRef Search ADS Thawley CJ, Langkilde T. 2017. Attracting unwanted attention: generalization of behavioural adaptation to an invasive predator carries costs. Anim Behav . 123: 285– 291. Google Scholar CrossRef Search ADS Thornton A, Samson J. 2012. Innovative problem solving in wild meerkats. Anim Behav . 83: 1459– 1468. Google Scholar CrossRef Search ADS Truskanov N, Lotem A. 2015. The importance of active search for effective social learning: an experimental test in young passerines. Anim Behav . 108: 165– 173. Google Scholar CrossRef Search ADS Truskanov N, Lotem A. 2017. Trial-and-error copying of demonstrated actions reveals how fledglings learn to ‘imitate’ their mothers. Proc R Soc B . 284: 7. Google Scholar CrossRef Search ADS Truskanov N, Lotem A. 2018. Data from: context-specific learning and its implications for social learning. Dryad Digital Repository . http://dx.doi.org/10.5061/dryad.1nc7424. Vlamings PH, Uher J, Call J. 2006. How the great apes (Pan troglodytes, Pongo pygmaeus, Pan paniscus, and Gorilla gorilla) perform on the reversed contingency task: the effects of food quantity and food visibility. J Exp Psychol Anim Behav Process . 32: 60– 70. Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: email@example.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
Behavioral Ecology – Oxford University Press
Published: May 31, 2018
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
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
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.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera