Personality predicts the propensity for social learning in a wild primate

Introduction

Individuals can acquire information either through interacting with their environment directly—personal information—or through observing the behaviour of other individuals—social information (Dall et al., 2005). Personal information can be costly and time consuming to collect (Laland, 2004). For social animals, social information provides an alternative to gathering costly personal information, potentially facilitating the acquisition of information about when, where and how to get food, where to travel and remain, whom to choose as a mate, and which predators to avoid (Giraldeau, Valone & Templeton, 2002; Reader & Biro, 2010). Social learning—where an individual changes its behaviour after observing another—allows the rapid dissemination of novel information among group members, and has also been widely implicated in the formation of tradition and cultures within species (Whiten, 2000; Castro & Toro, 2004). On the other hand, social learning can be conceived as a form of ‘information parasitism’ and can be maladaptive if information rapidly becomes outdated (Laland, 2004). Nonetheless individual differences in innovation and social learning abilities remain largely overlooked (Thornton & Lukas, 2012). This is surprising, as an understanding of individual variation in social learning is likely to be necessary for predicting the formation and persistence of traditions and culture within groups, and the individual fitness advantages of personal versus social information use (Thornton & Lukas, 2012).

Animal personality refers to between-individual differences in behaviour that persist through time (Sih, Bell & Johnson, 2004). Evidence for personality-related individual differences in information use (Kurvers et al., 2010; Carter et al., 2013) suggests personality may also explain individual differences in social learning (Marchetti & Drent, 2000). This may occur at two different steps in the social learning process, when (1) a subset of individuals initially gathers personal information about a novel situation and (2) this subset—the ‘demonstrators’—provide other individuals with the opportunity to then gather and use social information about the novel situation from them. In the first step, for individuals to exploit a novel food or solve a novel task they must interact with it for long enough to obtain sufficient personal information about it (Seferta et al., 2001; Thornton & Samson, 2012). Individual variation in responses to novel stimuli are frequently used to differentiate individuals by two personality traits, namely boldness and exploration (Carter et al., 2012b). Those animals that are less bold, i.e., shyer, or less explorative of novelty, i.e., ‘neophobic’, individuals (Réale et al., 2007; Carter et al., 2012c) tend to be less willing to interact with a novel food or task and are thus less able to acquire the personal information necessary to solve the problem (but see Cole & Quinn, 2011). For example, neophobic myna (Acridotheres tristis) are less likely to solve innovation tasks than neophilic myna (see also Coleman, Tully & McMillan, 2005; Toxopeus et al., 2005; Sol et al., 2011). In the second step, those animals that are less willing or able to gather personal information may be more likely to use social information. For instance, where individuals seek to exploit a novel food or solve a novel task, bolder individuals may do so directly because they are willing to gain personal information about it, and subsequently become demonstrators for shyer individuals who may be more likely to attend to, and act upon, social information (Kurvers et al., 2010).

We investigated the relationship between animal personality and social learning in wild chacma baboons (Papio ursinus). Baboons present an ideal system to explore such individual differences, since they are highly social, live in stable groups, and social learning has previously been documented among infants and juveniles (Whiten, 2000). In addition, individual baboons in our study population vary consistently in two unrelated personality traits, boldness and anxiety, that are repeatable over a three-year period and for which all individuals in the study groups have been assayed up to six times (Carter et al., 2012c). We use the term ‘boldness’ to refer to an individual’s response to a novel food, where bolder individuals spend more time inspecting a novel food; and ‘anxiety’ for an individual’s response to a threatening stimulus, where more anxious individuals show a stronger fear response towards, and spend more time investigating, a threat (Carter et al., 2012c). Though threat responses can also be used to describe boldness, we have previously found no evidence that our threat presentation test measures boldness in this species (for further details, see Carter et al., 2012c).

To test our hypothesis that personality traits (boldness, anxiety) influence individual propensity for social learning, we performed two types of field experiments in which social learning to solve a task could occur: exploitation of a novel food (experiment NF) and a hidden food (HF). We investigated whether (Q1) propensity to solve the tasks using personal information alone, (Q2) time spent collecting social information (watching a demonstrator solve the tasks), and (Q3) social learning, i.e., changes in task-solving behaviour after exposure to a demonstrator, were related to the baboons’ boldness or anxiety. Thus we addressed three questions with two experiments for two personality traits that could influence social learning.

We had several predictions in line with our questions. First, we predicted that bolder baboons would be more likely to ‘solve’ the hidden food task as bolder individuals are more willing to interact with a novel food (Carter et al., 2012a) (prediction 1 for boldness: P1_HFb). We further predicted that there would be no relationship between anxiety and the propensity to solve the hidden food task as anxiety is unrelated to boldness in this species (Carter et al., 2012c) and anxious individuals are otherwise no more likely to interact with a novel food than calm individuals (P1_HFa). Second, regarding whether an individual would collect social information, we predicted that more anxious baboons would spend more time watching the demonstrator than calmer baboons, in both the novel food and hidden food tasks, as anxious baboons attend to a threatening stimulus for longer than calm baboons (Carter et al., 2012c) (P2_NF,HFa). We further predicted that shyer baboons would spend more time watching a demonstrator than bolder baboons in both tasks as shy geese attend to social information more than bold geese (Kurvers et al., 2010) (P2_NF,HFb). Finally, regarding whether individuals differed in their propensity for social learning in the novel food and hidden food tasks, we predicted that more anxious individuals (P3_NF,HFa) would show greater social learning than calmer baboons, as anxious individuals were predicted to attend to social information more than calm baboons (P2_NF,HFa above). Similarly, we predicted that shyer individuals (P3_NF,HFb) would show greater social learning than bold baboons in both tasks, as shy individuals were predicted to attend to social information more than bold baboons (P2_NF,HFb above).

Materials and Method

Results

The minimal models for all analyses described below are provided in Table 1, and a summary of our findings with regard to our predictions is provided in Table 2. In addition to the results we present below regarding personality, it is notable that age class had an effect on many of the variables that we investigated (Table 1). In all cases, there was a positive effect of being juvenile on our measured variables: juveniles solved the hidden task more often (m1_HF), spent longer watching a demonstrator (m2_NF), and showed evidence for greater social learning than adults: juveniles handled the novel food for longer after watching a demonstrator (m3_NFa), and showed greater responses to (and improvement in, see File S1) the hidden food task (m3_HF). Few other variables had such a consistent effect on our measurements (Table 1).

Novel food experiments

We assessed individual variation in the time spent observing demonstrators manipulate the novel food (Q2). We found no evidence that shyer baboons were more attentive observers, contrary to P2_NFb. Further, in contrast to P2_NFa, calmer baboons spent more time watching demonstrators interacting with a novel food, after accounting for the subject’s age and rank (m2_NF: β ± s.e. = −2.79 ± 0.51, t = −2.01, Fig. 1).

Next, we assessed the evidence for social learning, and individual variation in social learning arising from personality differences. In the case of novel foods, we achieved this by analysing the food handling time in the ‘re-test’ conditions both for those baboons that either had no exposure to a demonstrator after the initial test (the control group) and for those that did (the experimental group). In both the models for anxiety (a) and boldness (b) (models 3_NFa and b), the three-way interaction was significant (3_NFa: β ± s.e. = 7.66 ± 3.04, z = 2.52, p = 0.01; 3_NFb: β ± s.e. = −0.68 ± 0.08, z = −8.67, p < 0.001). Overall, in the control groups, individuals spent less time handling the novel food when re-tested, and both bolder and more anxious individuals showed a greater decline in handling time when re-tested (Table 1; Figs. 2A and 2C). In the treatment groups, individuals showed the opposite response—handling times tended to increase when re-tested—and bolder and more anxious individuals showed higher handling times than shyer and calmer individuals, respectively (models 3_NFa and b, Table 1; Figs. 2B and 2D). Although calmer individuals also showed higher handling times in the second presentation after watching a demonstrator, this reflected a non-significant 1.4-fold increase in handling time in comparison to the 2.7-fold increase in more anxious individuals. Thus, bolder and more anxious animals showed social learning after exposure to a demonstrator (contrary to P3_NFb, but in support of P3_NFa, respectively).

Hidden food experiments

In our assessment of individual propensity for task-solving (Q1), we found that bolder baboons were more likely to obtain food from the hidden food experiment (β ± s.e. = −0.85 ± 0.32, z = −2.63, p = 0.008), after accounting for differences in age (model 1, Table 1; Fig. 3), in support of P1_HFb. Anxiety, however, had no effect on task-solving, in support of P1_HFa (model 1: β ± s.e. = −0.28 ± 0.72, z = −0.40, p = 0.69).

We then assessed individual variation in the probability of observing the demonstrators (Q2). We found no evidence that shyer or more anxious baboons were more likely to watch an individual manipulate the hidden food task, contrary to P2_HFb and a respectively (Table 1, m2_HF).

Finally, we assessed the evidence for social learning by analysing the intensity of the response to the stimulus in relation to the number of prior total, social or personal presentations. We found no evidence for an interaction between any of the presentation numbers and either personality trait; in all cases the minimal model was the same (model 3_HF). Thus, while bolder animals continued to be better at solving the hidden food task (as observed in m1), we found no evidence that either shyer or more anxious individuals showed improved responses with more prior experience, contrary to predictions 3_HFb and 3_HFa respectively.

Discussion

Our results strongly suggest that personality predicts social learning propensity in wild baboons. In a field experiment in which individuals had the opportunity to learn about a novel food from an experienced demonstrator, both bolder and more anxious individuals increased their handling time of the novel food if they had had an opportunity for social learning. Below we address those predictions that were not supported before focussing on three important issues emerging from our study that have implications not only for understanding social learning in the wild (van de Waal & Bshary, 2011), but also for individual differences in cognitive abilities (Thornton & Lukas, 2012) and the formation of traditions and culture in animal groups (Whiten, 2000).

Contrary to expectation, we found no effect of boldness on time spent watching a demonstrator. These results differ from those of Kurvers et al. (2010) which suggested that shy individuals used social information more than bold individuals, but this discrepancy may be explained by differences between studies in the familiarity of the study subjects with the experimental challenge (see below). However, our finding that bold individuals were more likely to solve the hidden food task confirms previous research showing an increase in task-solving success when individuals are less neophobic (Seferta et al., 2001; Webster & Lefebvre, 2001). Our failure to find a comparable improvement in task solving after watching a demonstrator for the hidden food experiment is unexpected, but may reflect the fact that the task contained a highly preferred food, and individuals may be limited by their rank in their access to the task. That is, individuals may have acquired the social information necessary to solve the task, but were unable or unwilling to access the task due to their lower rank. Though we do not have the resolution of data to test this explicitly, we discuss this possibility further below.

Three important issues emerged from our study. First, an individual’s ability or interest in collecting social information does not necessarily correlate with its ability or interest in using social information. This observation arises from our findings that bolder and more anxious individuals showed a greater improvement than shyer and calmer individuals in their response to a novel food after watching a demonstrator, despite the fact that there was no effect of boldness on the time spent watching a demonstrator, and calmer individuals were more attentive than more anxious individuals. Three possible explanations could be considered for these results: that (i) a mechanism other than watching a demonstrator is responsible for social learning in baboons, (ii) baboons need little time to acquire sufficient social information to subsequently change their behaviour, or (iii) as mentioned above, the acquisition of social information is not always correlated with its use. The first explanation is unlikely, since the task is a visual one and baboons are strongly visual animals. It therefore seems probable that the second or third explanations are most likely, and these need not be mutually exclusive.

Social learning mechanisms such as stimulus enhancement, where the behaviour or presence of a demonstrator attracts the attention of a naïve individual (Brown & Laland, 2003), may quickly result in a behavioural change in naïve individuals without the detailed and lengthy observation assumed necessary in processes such as imitation. As such, the baboons could learn that an object is worth investigating in more detail in the future in the time it takes to identify the object that a demonstrator is manipulating, as may be the case in our study. Further, it may be that bold individuals show greater social learning not necessarily because they are ‘better’ at collecting or processing social information, but because they are generally more willing to interact with a novel task. This, coupled with the rapid social learning through stimulus enhancement we posit to be possible in this study, may result in bolder individuals showing greater social learning. In support of this interpretation, in the novel food control trials, bold baboons decreased their handling time of the novel food but this handling time still exceeded that of shy individuals on their first presentations. This may also explain why our findings differ to those of Kurvers et al. (2010), as the authors of that study habituated the study subjects to the experimental arena prior to testing. In both cases (our study, and Kurvers et al.’s study), individual variation in response to novel tasks is related to personality, but not necessarily an individual’s ability to collect and process social information. The fact that some individuals may be too shy or anxious to interact with a novel task, despite having the information or ability to do so, may have wide implications. For example, an individual’s cognitive performance may be systematically underestimated if it is unwilling, but still able, to solve a task, which may hinder studies interested in the fitness advantages and evolution of cognitive performance. Future studies could assess this effect experimentally by testing study animals of different boldness with varying degrees of habituation to a test apparatus. Our interpretation would lead to the prediction that bolder individuals should outperform shyer individuals under low habituation regimes, but that performance may be more equal when the task is no longer threatening.

Second, the identity of demonstrators may be important. In both our experiments, many of our initial demonstrators were bold. In the novel food experiment, this was inevitable, as our definition of boldness precluded individuals that did not eat the novel food. However, it also transpired that those individuals that solved the hidden food task were more likely to be bold, though we note that the boldness and anxiety of the demonstrators was, overall, not particularly different to the boldness and anxiety of the population (see summary statistics reported in the methods for these groups). As these experiments were not designed to address this particular question and thus the sample of demonstrators is limited, we have not analysed these data here. However, previous research has suggested that the characteristics of a demonstrator may affect whether an observer gathers or uses social information from them or not (termed ‘directed social learning’: (Coussi-Korbel & Fragaszy, 1995; Laland, 2004)). For instance, wild vervet monkeys (Chlorocebus aethiops) are more likely to attend to the social information provided by the philopatric sex (females) (van de Waal et al., 2010). It remains to be established whether the personality of the demonstrator might be important, but it seems possible that it may play a role. For example, slow explorers are more thorough at exploring novel environments (Careau et al., 2009) and may therefore be better informed about the location of resources. Although it is difficult to anticipate how demonstrator boldness might affect audience attentiveness and social learning in baboons, we can consider its correlates. Specifically, while sex is unrelated to boldness in our population, there is an effect of age: juvenile baboons are bolder than adults (A Carter, H Marshall, G Cowlishaw, unpublished data, 2009–2011). Since this led to a strong representation of juvenile animals among the demonstrators (10 juvenile, 4 adult demonstrators), and adults may attend less closely to juveniles (only vertical transmission of social information has been documented in this species: Camberfort, 1981), it is possible that the adults in our sample might have shown lower levels of attentiveness and social learning than would have been observed with adult-only demonstrators.

Finally, an important aspect of our study that we did not anticipate was the difficulty of testing certain individuals in the presence of a demonstrator. While we endeavoured to provide all individuals with the opportunity to watch a demonstrator for both tasks, it was not possible to find a situation in which certain individuals were foraging in sufficiently close proximity to potential demonstrators for this to occur (Sih & Del Giudice, 2012). In great tits (Parus major), slow exploring birds showed higher attraction to unknown conspecifics than fast explorers (Carere et al., 2001) suggesting that personality may influence sociability and thus opportunities to learn from others in some species. Further, in guppies (Poecilia reticulata), an individual’s social network is related to its personality: individuals associated assortatively according to behavioural type, and shy fish had more and stronger network connections than bold fish (Croft et al., 2009). An individual’s social network is thus likely to affect which individuals can supply social information, and thus the flow of information through a group (Voelkl & Noë, 2008; Sih, Hanser & McHugh, 2009; Claidiére et al., 2013). If similar personality-related patterning among the social networks of baboons occurs, it may be impossible for some individuals to obtain social information about a novel food or task as they may have only weak associations with demonstrators. Together with the preceding points, it thus appears possible that personality could affect social learning, and hence the formation of culture in animal groups, through a combination of mechanisms. These include not only the differential collection of social information, as we have shown in this study, but also the differential application of social information, directed social learning, and assortative social bonds.

Supplemental information