Spontaneous cross-species imitation in interactions between chimpanzees and zoo visitors

By enabling quick and high-fidelity social learning, imitation is regarded as a key mechanism for mediating the cross-generational transfer of knowledge and skills, (e.g. Nielsen 2009). Besides its ‘cognitive’ or ‘learning’ function, imitation accomplishes important social and communicative functions as well, by facilitating social interaction and promoting prosociality (e.g. Užgiris et al. 1989; Eckerman et al. 1989; Eckerman and Stein 1990; Carpenter et al. 2013; Duffy and Chartrand 2015). In the present study we focus on this social–communicative function of imitation, which, unlike imitation learning is less investigated in nonhuman primates¹. In fact, this has even been claimed to be absent in nonhuman primates based on the observation that, in the context of imitation learning tasks, chimpanzees exhibit lower levels of joint attention, gaze at the experimenter’s face and show ‘over-imitation’ compared to children (Nielsen 2009; Carpenter and Call 2009). The latter captures the fact that children—but not chimpanzees—show a propensity for slavishly copying actions over outcomes, even when the demonstrated actions are irrelevant for achieving a demonstrated outcome. These differences, it has been argued, are to be attributed to the fact that a motivation to be social and promote shared experience lies behind children’s imitative behaviour.

We contend that such claims are grounded on a too narrow definition of the social function of imitation. Evidence from social, developmental and comparative psychology indicates that the latter encompasses both reactive and arguably non-intentional phenomena (such as nonconscious mimicry and imitation-induced prosociality), as well as more proactive and arguably intentional ones (e.g. certain social conformist tendencies or toddlers’ use of imitation for establishing and maintaining social interaction), as detailed below.

Nonconscious mimicry can be defined as the matching of postures, gestures, verbal or facial expressions between interaction partners, without awareness or intent (for a recent review see e.g. Duffy and Chartrand 2015). In nonhuman primates, non-conscious mimicry has been documented in the form of postural congruence (Jazrawi 2000), rapid facial mimicry (as reviewed e.g. by Mancini et al. 2013b), behavioural contagion (e.g. Madsen et al. 2013; Amici et al. 2014 for recent reviews) and interactional synchrony (Yu and Tomonaga 2015). Intriguingly, nonconscious mimicry seems to have a protracted onset in both human and chimpanzee ontogeny, being shown earliest at 3–4 years of age (Carpenter et al. 2013; Madsen et al. 2013; van Schaik and Hunnius 2016).

Nonconscious mimicry, while not being intentionally communicative, still seems to communicate information that leads to prosociality in both human and nonhuman primates. When mimicked, human adults are more helpful and show increased affiliation towards imitating interaction partners (Duffy and Chartrand 2015). Likewise, rapid facial mimicry increases the duration of playful interactions in gelada baboons (Mancini et al. 2013a) and chimpanzees (Davila-Ross et al. 2011). Interestingly, in toddlers and monkeys, similar prosocial effects arise also when they are exposed to overt and systematic imitation. Indeed, toddlers are more helpful after being overtly imitated (Carpenter et al. 2013), while macaque (Macaca nemestrina, Paukner et al. 2005) and capuchin monkeys (Sapajus apella, Paukner et al. 2009) show increased levels of attention, interactivity and affiliation.

While the responses produced by the monkeys suggest that they implicitly discriminate being imitated from other types of interactions, toddlers and great apes show explicit imitation recognition, as they respond with so-called testing behaviours (e.g. Nielsen et al. 2005; Haun and Call 2008; Pope et al. 2015). Often these take the form of ‘behavioural repetitions’, i.e. the subject repeats a behaviour previously imitated by the experimenter. Such responses are regarded as active tests of the contingent correspondence between subjects’ own actions and those of the experimenter, and thus as indicating that apes and toddlers are aware of both action correspondence and the social causality involved in such imitative interactions (e.g. Nielsen et al. 2005; Bates and Byrne 2010). In other words, they are seemingly aware of the impact of their own actions on others’ behaviour (Bates and Byrne 2010) and also grasp the experimenter’s intention to imitate (Whiten and Suddendorf 2001) or, at least, to engage socially. In line with this interpretation, imitation recognition appears to be related to general social communicative skills. For example, chimpanzees that are more responsive to being imitated, also score higher on gesture production and comprehension, joint attention and gaze following (Pope et al. 2015).

Finally, the social conformism found for chimpanzees in some imitation learning tasks can arguably constitute further evidence of the proactive social function of imitation. More specifically, the chimpanzees have been found to copy the solution employed by other group members, even when this is less efficient or it goes against individual preferences. It has been proposed that such tendencies are an expression of conformity and benefit group cohesion (e.g. Hopper et al. 2011 and references therein).

In the present study, we focus on a kind of communicative imitation that has been described as the main mode of peer communication among pre-verbal toddlers (e.g. Eckerman et al. 1989; Eckerman and Stein 1990; Cordonier and Deschenaux 2014). This is characterised by the intentional, overt and immediate copying of predominantly familiar play actions and is accompanied by communicative behaviours, such as gaze contact and smiling (e.g. Cordonier and Deschenaux 2014). By means of such overt imitation, young children initiate and maintain interaction for significantly longer than in the absence of imitation (e.g. Eckerman and Stein 1990). In human ontogeny, this intentional form of communicative imitation peaks at the age of 18–24 months, thus preceding the emergence of nonconscious mimicry.

To our knowledge, this form of communicative imitation has not been previously studied in great apes, presumably because spontaneous imitation is suggested to be rare among apes (e.g. Call 2001), and evidence of familiar-action imitation is currently scarce (e.g. Tennie et al. 2012; Amici et al. 2014). Nevertheless, spontaneous imitation has been attested in great apes, both in captivity and in the wild, in inter- (e.g. Russon and Galdikas 1995; Byrne and Tanner 2006) and intraspecific contexts (e.g. Russon and Galdikas 1995; Hobaiter and Byrne 2010). It is often pointed out that apes’ spontaneous imitation can vary greatly with respect to fidelity to the modelled actions (e.g. Russon and Galdikas 1995; Byrne and Tanner 2006). This is also highlighted in studies in which apes, after extensive training with food reinforcement, learn to imitate on command (e.g. Custance et al. 1995; Call 2001; Carrasco et al. 2009). Nevertheless, such variation in imitation precision is present in toddlers as well, and, in the context of communicative imitation, ‘producing an exact match of an observed action is secondary to the goal of maintaining the interaction’ (Užgiris et al. 1989).

With respect to familiar-action imitation, recent experimental attempts at documenting it have yielded mainly negative results. In an imitation learning study by Tennie et al. (2012), only one non-enculturated chimpanzee showed evidence of familiar-action imitation. In another study involving zoo housed apes (Amici et al. 2014), none of the subjects imitated voluntary gestures (hand closing, wrist shaking), or other involuntary gestures than yawning (scratching, nose wiping). However, the imitation of familiar actions remains an understudied topic in nonhuman primates, thus it is possible that its frequency is underestimated². Thus far, great apes have not been reported to imitate for seemingly social–communicative purposes. When imitated by humans, however, they often repeat the imitated action. It is unclear if this response is merely aimed at testing the ‘social mirror’ or if it qualifies as returned imitation aimed at maintaining the interaction, as argued for toddlers’ communicative imitation (e.g. Eckerman and Stein 1990). Arguing for the latter interpretation is the fact that imitation recognition appears to be related to social–communicative skills in chimpanzees (Pope et al. 2015).

Since imitation outside of learning tasks, i.e. the spontaneous imitation of familiar actions in social interaction, is underrepresented in the animal literature, the present study is exploratory and asks several questions:

Finally, it is important to stress that obtaining data on the visitors’ imitative behaviours was equally important as collecting chimpanzee data. Indeed, studies on human spontaneous imitation in the context of cross-species interaction are lacking, yet such research is highly relevant for comparative purposes.

To our knowledge, the present study is the first to systematically investigate spontaneous imitation in free interactions between zoo visitors and zoo-housed chimpanzees. Previous visitor studies have found, in general, little interaction between H. sapiens and P. troglodytes. Moreover, visitor-directed chimpanzee actions appeared to have been generally unrelated to visitors’ behaviours unless food was involved, which elicited begging behaviours (e.g. Cook and Hosey 1995; Wood 1998). As suggested by reviewers, factors that might account for such different results are exhibit design and animal husbandry practices. The presence of large glass walls might facilitate close proximity between the species and thus higher levels of interaction. With respect to husbandry practices, zoo animals that are engaged by keepers in social interactions subsequently respond more positively to unfamiliar visitors (see Hosey and Melfi 2015).

In this study, during 52 h of observation, we recorded 354 episodes of cross-species reciprocal interactions. Food-related gestures were exclusively performed by AF. Imitative actions by one species or by both occurred in 37% of the 354 reciprocal interactions, and had an effect of prolonging and increasing engagement in the interaction. Indeed, episodes that included imitation contained twice as many actions as those without imitation. For both species, all imitated actions were familiar ones. This supports the view that, for a scientific understanding of the roles of behaviour copying, studies of imitation should not only be limited to the area of social learning.

The imitative responses documented here did not exhibit the characteristics of reflexive responding. On the contrary, several of their features suggest intentional communication as both species produced ‘targeted’ imitations, i.e. while visually and bodily oriented towards the model and while monitoring the interaction. Directedness toward a specific receiver and sensitivity to attentional stances are the most commonly accepted behavioural criteria for intentional communication in apes (e.g. Hobaiter and Byrne 2011). It has been argued, however, that apes’ intentionality in communicative situations lacks a crucial ‘shared’ aspect, which in humans supports the motivation to share experiences with others by means of communication (e.g. Carpenter and Call 2009; Nielsen 2009). As a consequence, these authors have proposed that, in apes, imitative exchanges might lack the communicative component. For example, in imitation learning tasks, apes are found to engage in shorter bouts of communicative behaviours (joint attention, gazing at the model’s face) than children (e.g. Carpenter and Tomasello 1995). However, it has also been found that communicative imitation in children is favoured by non-learning contexts, and that communicative behaviours decrease or even vanish in instrumental imitation learning tasks (Cordonier and Deschenaux 2014). With respect to imitation recognition, previously tested apes have not been reported to show signs of enjoyment and playfulness, e.g. to engage in imitative games (Nielsen 2009). The evidence in our study suggests, however, that apes can show playfulness in imitative contexts and that imitation might serve a communicative purpose. We base this conclusion on the fact that, on several occasions, the chimpanzees engaged in prolonged imitative exchanges that satisfy the criteria to qualify as imitative games. Additionally, we observed that, at least on some occasions, AM exhibited play face expressions in the context of imitative exchanges. We therefore propose that the imitative episodes described here are reminiscent of toddlers’ communicative imitation. In such contexts familiar actions are overtly imitated for initiating and maintaining interaction, resulting in mutually rewarding and sustained engagement in the interaction, occasionally leading to imitation games (Eckerman et al. 1989). This is in agreement with the suggestion that apes exhibit signs of shared intentionality in the context of social games, especially considering that in at least three documented cases—two with bonobos (Pika and Zuberbühler 2008) and one with gorillas (Tanner and Byrne 2010)—the social game involved symmetrical object exchanges, and thus could have been potentially imitative in nature.

Further argumenting for the intentional and communicative character of the imitation documented in this study is evidence of flexibility and selectivity. Interestingly, the most flexible chimpanzee exhibited an imitative repertoire of a similar size to that of the most imitative visitor categories (adults, subadults and juveniles). On the other hand, only few actions were preferentially selected for seemingly imitative–communicative purposes by both species. This selectivity is likely the result of a long interactional history between visitors and chimpanzees that—among other things—allowed the chimpanzees to learn the effect that copying specific actions has on visitors. Such an interpretation is plausible, considering that apes show selectivity for behaviours experienced as communicatively efficient (Hobaiter and Byrne 2011). Even if repeated interaction had a reinforcing role in shaping chimpanzees’ imitative responses, it is worth emphasizing that the only plausible reinforcement was the rewarding nature of (continued) social interaction itself. While zoo staff uses food incentives in many interactions with the chimpanzees, in our experience these interactions do not involve imitation, and are thus not likely to reinforce such responses. They more likely result in begging gestures, which we do see directed at zoo visitors as well. The actions we observed in imitative games, however, were typically not ritualized begging gestures.

Another finding is that, in both species, imitation rate increased in the indoor environment, suggesting that physical proximity plays a role in facilitating imitation. In both species, physical proximity further coincided with increased partial imitation; conversely, exact imitation dominated almost exclusively in distal imitative exchanges. It is unclear, however, whether such results are due to higher requirements for precision in distal communication, or by the different repertoires of imitated actions indoors and outdoors. For example, Call (2001) found an orangutan subject to be more successful at matching actions performed with gross body parts than those involving smaller parts, e.g. a raise-finger action was mimicked with a raise-arm action. Thus the almost exclusive predominance of exact imitation in the outdoor locations could be related to the fact that larger gestures and gross motor patterns were more likely to be performed and imitated in that environment. In the indoor locations, on the other hand, imitated actions involved more detailed matching, such as, for example, knocking on the window with knuckles or palm. In any case, it has to be stressed that this increase in partial imitation was similar in both species. Physical proximity, especially in cases when space can be shared, such as a windowpane, might be a crucial variable and requires further attention in imitation research. It is also possible, however, that glass surfaces (rather than proximity), afforded actions that facilitated imitation.

Besides cross-species commonalities there were also notable differences. For example, humans showed overall higher imitative precision. The data further indicate differences concerning the size of the cumulative repertoire of imitated behaviours, which was larger among visitors than chimpanzees. This can be due to the large difference in the number of individuals that represented each species, but also to the fact that—unlike humans—the chimpanzees did not imitate bodily postures or behaviours with a physiological function. In this respect, we found a potential parallelism with imitation recognition, i.e. how the chimpanzees responded when their behaviours were imitated by the visitors. The chimpanzees responded when visitors imitated their actions, but they were not observed to react when visitors imitated their postures or physiological behaviours (e.g. yawn, scratch). That chimpanzees failed to show a response when visitors copied their postures could be related to the generally non-conscious nature of this type of mirroring behaviour, which, as opposed to overt imitation, does not elicit behaviours indicative of explicit imitation recognition. Although recent studies highlight the similar positive social consequences of both overt imitation and nonconscious mimicry and regard them as related phenomena (e.g. Carpenter et al. 2013; Duffy and Chartrand 2015), further research is needed to specify the relationship between these imitative behaviours.

Further research should moreover be directed towards establishing how representative imitative communication is of chimpanzees’ intra- and cross-specific interactions, and if the present results can be replicated. Our data show that four of the five chimpanzees that were observed produced imitative actions in cross-species interactions. These results contrast with previous experimental studies that found familiar action imitation to be restricted to certain, ‘gifted’ individuals (Tennie et al. 2012), or to specific non-voluntary behaviours such as yawning (Amici et al. 2014). Since human-raised chimpanzees show a higher orientation towards human behaviour and better imitative skills (e.g. Byrne and Tanner 2006; Tennie et al. 2012 for discussions), rearing history is a potential candidate to explain these differences. In the case of this group, however, it is not possible to draw any firm conclusions. The only mother-reared chimpanzee (SF1) in the group interacted only minimally with visitors, while the four imitating chimpanzees, which had mixed rearing histories, exhibited variations in imitation rates. Interestingly, AF, the only chimpanzee in the group that was temporarily (i.e. for 18 months) raised in a human home without simultaneous contact with conspecifics, was also the chimpanzee that imitated the least. One of the chimpanzees (AM) was by far the most interactive individual in the group, as well as the most flexible imitator, which might support the view that certain ‘gifted’ individuals are more prone to imitation (Tennie et al. 2012). Additionally, imitation recognition in chimpanzees appears to be related to sex, as males are more responsive than females when being imitated by a human experimenter (Pope et al. 2015). Yet AM’s overall imitation rate was relatively similar to those of SF2 and JF. It thus remains to be established whether other factors can further account for a chimpanzee’s propensity for copying behaviour. Finally, the results reported could be attributed to the nature of the familiar actions performed in these imitative interactions, as well as to the context in which they were deployed. In our study, we capture the imitation of familiar communicative or play behaviours in the context of relaxed and positive social interactions. Moreover, food acquisition was not involved, unlike in previous studies.

With respect to visitors’ behaviour, subadult males were the most imitating category outdoors, while toddlers almost never imitated in this environment. Indoors, however, toddlers showed an increase in both interactivity and imitation rate, becoming the most imitating category in this proximal setting.

An advantage of using an observational approach is that we were able to simultaneously document a broader range of variables than is possible in an experimental setting, while capturing spontaneous occurrences in a naturalistic setting. Inherent limitations to this method (e.g. the inability to study potentially moderating variables in controlled conditions) limit, however, the conclusions which can be drawn from this study. Overall, the present study highlights the imitation of familiar behaviours in apes as a relevant research topic and provides testable hypotheses concerning its role in promoting and maintaining play and interaction, as well as concerning the potential role of proximity in promoting imitation. Our data add to previous observations of spontaneous imitation by apes and indicate that, unlike previously suggested, the chimpanzees’ imitative actions might serve a socio-communicative function by maintaining social interaction. Moreover, imitation seemed to take on the characteristics of play, giving rise to so-called imitation games. We hope that the results presented here will incite further inquiries into this topic.