Although humans may be unique among species in the extent to which we help others, there can little doubt that non-human primates also act in ways that benefit other individuals in some contexts. Some primate species are known to transfer food to unrelated social partners [1
], engage in instrumental helping [2
], and console victims of aggression [7
]. Such behaviors are often referred to as prosocial behaviors.
Note that unlike the term altruism, the term prosociality does not make any stipulations regarding the cost of the behavior to the actor. The extent to which other species share human-like prosocial motivations is controversial, but understanding interspecies variation in prosocial tendencies is critical to furthering our understanding of the evolutionary origins of human altruism.
Recently, there has been renewed interest in the use of experimental paradigms to test whether different primate species exhibit prosocial behaviors (for a review, see Silk and House [8
]). In a widely used paradigm, one individual (the “actor”) must choose between two different options. Although both options yield the same food reward for the actor, one option delivers a highly desired food to a partner (the “recipient”) while the other option does not. Consequently, the actor must choose whether or not to deliver food to the recipient at no (or very minimal) personal cost. The rationale of the design is that if actors are indifferent
to the welfare of the recipient then they should choose randomly between the two options. If actors are antisocial
., they have a preference for decreasing other individuals’ welfare), then they should choose the less good option more often when the recipient is present than in a control condition in which no recipient is present to retrieve the food reward. Finally, if actors are prosocial
, then they should choose the generous option more often when a recipient is present to retrieve the food reward than in control conditions in which no recipient is present.
To date, many researchers have capitalized on this simple design to test primates’ prosocial preferences [9
]. Unfortunately, these experiments together have produced a rather puzzling set of results. For example, species may demonstrate prosocial behavior in some studies but not others (e.g., chimpanzees [9
]; tamarins [10
]). Additionally, even when there is an overall difference in performance between recipient present and recipient absent conditions, prosocial behavior does not necessarily manifest in all actor-recipient dyads (e.g., marmosets [22
]; capuchins [12
]). Whether such differences are the product of variation in individuals’ motivation to engage in prosocial behaviors per se or instead reflect differences in task understanding, sensitivity to inequity, attentional biases, or quality of relationship with the recipient is currently unclear. Finally, the overall percent differences between the recipient present and recipient absent conditions are often fairly modest [19
]. These results raise questions regarding the degree to which prosociality is affected by particular features of the testing environment, as well as the extent to which prosocial behaviors, when observed, can be described as species typical or consistent with a robust preference to help others. Given these open questions, it has been difficult to explain within and across species performance within a single theoretical framework.
To help address these issues it may be useful to further explore the stability of prosocial behaviors, both at the species and individual level. To do this, we decided to test brown capuchin monkeys using a novel testing method. Capuchin monkeys are appropriate for assessing the stability of prosocial preferences for a number of reasons. This species exhibits high social tolerance and is able to successfully cooperate in a variety of tasks [23
]. Moreover, the results of previous tests suggest that capuchins do act prosocially in a number of contexts [12
]. Nevertheless, this same set of studies also provides evidence that prosociality in capuchins may ultimately be quite fragile and subject to many of the inconsistencies listed above. That is, prosocial behaviors do not necessarily characterize all individuals or dyads and mean differences between the test and control conditions are often quite small (e.g., approximately 8% in Lakshminarayanan and Santos [15
]). Capuchins’ prosocial preferences may also disappear when the proposer has limited visual access to the recipient [12
]. Indeed, in such cases capuchins sometimes actually behave antisocially, preferring the antisocial option to a prosocial one [12
]. Finally, at least one study [20
] has reported that prosocial choices were directed primarily down the dominance hierarchy, but this pattern has not been confirmed in other studies [12
In order to further our understanding of the conditions under which capuchins exhibit prosocial behavior, we tested subjects using a novel touchscreen task. Computer-based tasks have proven useful in exploring primate cognition across a number of domains (e.g., memory [26
]; metacognition [27
]; gaze following [28
]; social decision-making [29
]). As in previous studies, monkeys could choose between a prosocial option (food for the recipient) and a selfish option (no food for the recipient). Three of the actors in this study had previously participated in another study exploring prosocial preferences [15
]. Note that in this previous study, capuchins as a group demonstrated prosocial behavior. That is, they selected the prosocial option more often when a recipient monkey was in the adjacent testing chamber compared to when no recipient was present. Because we were particularly interested in whether these three subjects would exhibit a similar pattern of preferences in two different contexts, we analyzed each actor’s performance in the two studies using the same statistical tests, thus allowing us to begin to explore performance across different contexts and different testing methods. Importantly, in addition to including a control condition in which no recipient was present to receive food rewards, we also included an additional control condition designed to probe subjects’ understanding of the contingencies of their responses. This is critical because the difficulties inherent to understanding inter and intra-species inconsistencies in performance across various studies are exacerbated by the fact that it is not always clear what subjects understand about the reward contingencies of their responses on specific tasks. Specifically, if subjects show an indifferent pattern of responding (that is, selecting the prosocial option equally often when a recipient is present compared to when the recipient is absent), it can be difficult to know whether this is because subjects are in fact indifferent to their partner’s payoff or whether subjects are simply not attending to all the relevant features of the task.
The goal of this study was to examine whether capuchin monkeys would exhibit prosocial behavior using a novel touchscreen task. Although previous studies have documented prosocial preferences in capuchins [12
], the same species showed little evidence of such preferences in our task. When presented with the choice between 1/1 and 1/0, capuchin actors did not choose the prosocial option at levels significantly above chance. More importantly, actors did not deliver a food reward to the blue cup more often when either the recipient or the actor had access to the blue cup compared to when no one did. When presented with the choice between 0/1 and 0/0, actors chose the prosocial option at levels significantly above chance in all conditions but again selected the prosocial option equally often across the three conditions.
Our results might at first glance seem to contradict those of previous studies reporting that capuchins demonstrate prosocial preferences, but it is important to consider exactly what subjects understood about our task before making such a conclusion. Although it is possible that capuchins did not understand the specific stimuli used in this study, their responses to the two pairs of stimuli in which one option was always preferable regardless of condition type show that this was not the case. For both pairs of stimuli in which one option was always superior to the other (i.e., 1/1 vs. 0/1 and 1/0 vs. 0/0), capuchins chose the optimal (i.e., food maximizing) option at levels significantly greater than chance across all conditions. Indeed, proposers were quite adept at choosing among these pairs, as a group selecting the choice that maximized their own reward outcome on over 87% of all trials.
These results present a puzzle in that proposers clearly did understand something about the reward outcomes of the stimuli, but failed to behave differently when their prosociality was tested in the three conditions. This was expected for two of the four pairs of stimuli, but we anticipated that when actors were presented with the two pairs in which there was a prosocial option and a selfish option, they would at least choose the prosocial option more often when they themselves would benefit by doing so (that is, in the selfish control condition
). That capuchins did not do so suggests that they were not necessarily attending to the contingencies of the different conditions. Although our training procedure was specifically designed to familiarize actors with the fact that reward outcomes varied across the different conditions and mirrored training previously done in other capuchin prosociality studies, it is possible that actors simply learned to associate each stimulus with, on average, a certain number of food rewards. Consequently, when given a choice between two pairs of stimuli, the strength of actors’ preference for one stimulus over another may merely reflect that probability of receiving a certain number of food rewards generally and may be insensitive to the contingencies specific to a given condition. The group data therefore highlight the importance of including a condition designed to probe subjects’ understanding of the task. In this case, subjects’ behavior in the selfish control condition
cautions against concluding that capuchins do not have prosocial preferences. In addition, our results highlight the benefit of assessing subjects’ understanding of the testing paradigm prior to, as well as parallel with, the testing stage. It’s unclear precisely why subjects failed to learn the contingencies of our task, but it is possible that the complexity of the touchscreen apparatus played some role. Other studies have used (arguably) less complex methods, such as token exchange paradigms [12
] or pulling a physical plank or drawer containing food rewards [15
]. In such cases, the actor can often see the causal sequences between his own choice and the resulting food distribution, which may facilitate task understanding.
Despite the failure of the group to display prosocial behavior, an examination of each individual actor’s data revealed that when given the choice between 1/1 and 1/0 one of the four subjects, NN, did choose the prosocial option more often when a recipient was able to retrieve food rewards delivered to the blue cup compared to when no one had access to the blue cup. NN also showed a trend towards doing so when presented with 0/1 and 0/0. Although it is tempting to speculate as to why only NN demonstrated prosociality in this particular context, the fact that none of the subjects clearly understood the task makes such speculation difficult since other subjects may have chosen prosocially had they understood all the reward contingencies. NN was also the only subjects to be presented with the prosocial test session last in each of the three testing sets, and so we cannot completely rule out that possibility that order effects played some role in subjects’ performance. Furthermore, although NN’s behavior may appear to be prosocially motivated, it is unclear why he did not choose the prosocial option more often in the selfish control condition. Thus, while we think it is possible that NN’s behavior reflects true prosocial motivation, it is unclear as to whether this is the most parsimonious explanation of the data.
Similarly, although an initial goal of this study was to explore whether some individuals might be consistently more prosocial than others by comparing subjects’ performance in the current study and Lakshminarayanan and Santos [15
], variation in subjects’ performance across the two studies seems likely to reflect differences in task understanding rather than prosocial motivation per se. Nevertheless, we hope that future studies will continue to explore individual variation in prosociality by testing the same individuals across a range of tasks. Although group data is obviously informative, looking at prosocial preferences at the level of the individual can also provide important insights. For example, examining each individual’s performance reveals what proportion of subjects actually demonstrated an understanding of the task. Obviously if subjects do not understand the task, we cannot draw any conclusions about whether they do or do not show prosocial preferences. Individual level analyses are necessary to determine whether significant group preferences for prosocial outcomes reflect the behavior of most individuals within that group or whether a few subjects may be largely driving the effect. Such data can help inform debates about the degree to which prosocial behaviors, when observed, are species typical and truly indicative of a robust preference to help others. It is also possible that some subjects that exhibit prosocial behavior do not necessarily seem to understand all aspects of the task (as in our study). In these cases, researchers will want to treat results with caution.
Finally, we add that when exploring inter- and intra-individual consistency in performance, it may be particularly valuable to include both tasks involving donating food to other group members as well as tasks involving other measures of prosociality as there is some doubt as to whether food donation tasks are the best way to assess prosociality in other species. For example, although studies testing prosociality in chimpanzees using food donation tasks have produced mostly negative results [9
], other studies have found that chimpanzees are willing to help others in instrumental helping tasks [3
]. This has led to the suggestion that visible food rewards may in some cases inhibit prosociality, perhaps because they elicit competitive responses from subjects [13
]. Consistent with this, in the one study that found strong evidence of prosociality in a chimpanzees study using a food donation task, food was concealed in paper during the choice phase of the experiment [13
]. Understanding how individuals behave across a range of differently structured prosocial tasks will therefore allow researchers to not only learn more about inter-individual variability in prosociality, but also allow us to better assess how different species may construe these tasks and ultimately lead to the development of better testing paradigms.