2An array of behaviours indicative of cultural variation has been identified in chimpanzees and orangutans. For example, chimpanzees from Gombe in Tanzania use 60-centimeter-long sticks to fish driver ants out of their nests. They then swipe their free hand along the stick, obtaining a large ball of edible ants in the process. Adopting a less efficient approach, chimpanzees from the Tai Forest in the Ivory Coast fish for ants using shorter sticks on which only a few ants are gathered. These ants are then transferred directly into the chimpanzee’s mouth. Thirty-nine such group specific traditions have been documented in chimpanzees (Whiten et al., 2001), and 19 in orangutans (van Schaik et al., 2003). This level of behavioral diversity reflects a complexity of cultural variation that has yet to be found in any other nonhuman animal.

3As impressive as it may be, the 39 distinct cultural traditions identified in chimpanzees seem insignificant when compared to an even cursory reflection on the variety of traditions that permeate human social behaviour. Indeed, differences between humans and our closest living relatives in the breadth and complexity of cultural traditions seem profound. What, then, might lie at the heart of these differences? For many authors the development and intergenerational transmission of culture is built around a capacity for imitation (e.g., Heyes, 1993; Tomasello, 1999a; Whiten et al., 2003; Whiten, 2005; Gergely and Csibra, 2005; Meltzoff, 2005). One could therefore reasonably expect differences to be found in the imitative behaviour of chimpanzees and humans. In the following pages I explore some of the evidence in support of this expectation, with a focus on comparisons between human children and chimpanzees, and outline how recent research findings can provide insight into the possible origins of differences in the proliferation of human and chimpanzee cultures.

4Before continuing, it would be customary to define my terms. What do I mean by culture? What do I accept imitation to entail? Unfortunately, doing so is no easy task. With regard to ‘culture’, more than 300 scholarly definitions are available (Baldwin et al., 2006). Not wishing to add to this list, I will adopt a definition that has been previously used for comparative purposes (e.g., Whiten et al., 1999; Whiten et al., 2001; van Schaik et al., 2003). Thus, the term ‘culture’ is herein taken to represent a system of socially transmitted behaviours that are customary or habitual in one community but absent in another community, and where such variation cannot be explained by ecological or genetic differences alone.

5With regard to imitation, during the last decade there has been a proliferation in the number of terms used to define and describe different ways in which individuals can copy others. This terminology has been useful in providing a framework for studies of social learning and for providing a way of characterising the types of social learning mechanisms distinct populations, such as different species or children of different ages, tend to use. Unfortunately, the number of definitions now available and the lack of concurrence over what they mean and how to correctly apply them has led to some confusion and disagreement on how particular studies are to be interpreted. Where some see imitation others see alternative mechanisms of social learning, such as mimicry (copying another’s behaviour without understanding its functional significance) or emulation (where the focus is on reproducing the outcome of another's behaviour but not the process by which it was achieved) (e.g., Russon and Galdikas, 1993; Whiten et al., 1996; Tomasello, 1999b). In an attempt to avoid any such confusion, here I will use ‘imitation’ to refer very broadly to instances where individuals reproduce an action or actions they have witnessed being produced by another. I leave it to the reader to decide whether the copying behaviours I describe would be better defined using social learning descriptors other than imitation.

6A final point regards the two principal populations that are to be contrasted here: children and chimpanzees. The majority of studies investigating imitation in great apes have used adult animals. Comparing adults of one species with the young of another may seem invalid or improper. However, there are several reasons why such comparisons can be defensibly made. If the young of a species exhibit a particular behaviour, then it can be reasonably assumed that adults have the capacity to exhibit this behaviour also. The converse is an obviously less valid assumption. Absence of, or diminution in, a behaviour in the young of a species may represent immature development rather than a true species-typical deficit. Where our aim is to identify uniquely human traits, or the evolutionary origins of them, we want to be sure that we do not draw incorrect conclusions by assessing individuals who might not have had the requisite developmental opportunities. Why, then, not compare adult chimps with adult humans? First, a number of reviews (e.g., Tomasello and Call, 1997; Suddendorf and Whiten, 2001; Suddendorf and Whiten, 2003) have highlighted similarities in the cognitive abilities of young human children and great apes – reviews that are primarily based on research involving adult apes. In this context, it is important that any inter-species comparisons are not conflated by differences in mental capacity. Further, as attested to in the review that follows, documenting ontogenetic changes in children’s behaviour can provide considerable insight into the ways our species may differ from those closely related to us – insights that would not be evident if focus was placed only on adults. Of course, it should be acknowledged that, in contrast to the ever-growing literature on human children, there is scarce cross-sectional or longitudinal data on the development of imitation in nonhuman primates (for some exceptions see Bjorklund and Bering, 2003; Tomasello and Carpenter, 2005a). Research documenting the ontogeny of imitation in each great ape species is certainly needed to compliment the data derived from studies conducted with adult animals.

1 Imitation in Children and Chimpanzees

7Using the broad definition previously outlined, a capacity for imitation has been well-documented in chimpanzees (for reviews see Whiten, 2005; Tomasello and Carpenter, 2005b). Moreover, the results of some controlled experiments indicate that the imitative behaviour of children and chimpanzees may not be all that dissimilar. For example, Horner and Whiten (2005) showed wild born chimpanzees and 3- to 4-year-old children how to obtain a reward from an opaque box. A human model first removed a bolt on the top of the box, revealing a hole into which a stick was inserted. The model then opened a door located on the front of the box and, by inserting the stick, retrieved the reward (food for the chimpanzees, a sticker for the children). Because the box was opaque the subjects could not see the causal relation between the actions and the desired outcome. When given the box to explore, both chimpanzees and children copied what the model had done. They first removed the bolt and placed the stick into the top hole and then opened the door and placed the stick into the hole at the front. The authors interpreted this as evidence of imitation in both species. However, Horner and Whiten also documented some interesting inter-species differences.

8In addition to the opaque box, Horner and Whiten (2005) tested their subjects for imitation using a transparent box. The actions occurring inside the box could now be seen, making it obvious that when the stick was inserted into the top hole it struck a barrier and made no contact with that part of the apparatus from which the reward could be retrieved. The action involving the top hole could be clearly seen to have no causal relation to the desired outcome. When the same actions that had been modeled on the opaque box were modeled on the transparent box, the chimpanzees ignored the first action and instead copied only the model’s insertion of the tool into the front hole. They focused principally on replicating the actions that were causally related to the desired outcome. By contrast, the children imitated the model’s entire sequence of actions, including the obviously irrelevant insertion of the stick into the top hole.

9The findings of Horner and Whiten (2005) reflect those in other studies that have found a far greater fidelity in copying behaviour by children compared to other apes (Tomasello et al., 1993; Whiten et al., 1996; e.g., Call et al., 2005; Tennie et al., 2006). Indeed, unlike chimpanzees, children consistently show a predilection for copying the specific means an adult uses to produce a particular outcome, even if a more efficient method is available. Research indicates that this predilection emerges during the second year.

2 Copying actions over outcomes

10In a cross-sectional study of children aged 12, 18, and 24 months, an adult model demonstrated how to open three novel boxes using an arbitrary object (e.g., a plastic block) to manipulate a switch located on the front of each box (Nielsen, 2006, Experiment 1). Correctly manipulating the switch, which was different for each box, disengaged a hidden latch and released the box’s lid, enabling a desirable toy to be retrieved. The children could have copied the model’s behavioral means of activating the switches (i.e., used the object) or they could have devised their own means (i.e., used their hands – something the model did not do). Two control conditions established that children of all age groups could easily activate the switches by hand if shown how but that they would not spontaneously attempt to do so, either by hand or by object, in the absence of modeling.

11Surprisingly, 12-month-olds opened as many boxes as did 24-month-olds when the model used an object to activate the switches. This counterintuitive finding can be explained by the specific actions the children employed when attempting to get the boxes open. Whereas the 24-month-olds tried to open every box using an object, as was modeled to them, the 12-month-olds only attempted to open the boxes with their hands. Eighteen-month-olds showed reactions that were intermediate between the older and younger age groups. All age groups were equally successful at getting the boxes open, but because the 24-month-olds, and to a lesser extent the 18-month-olds, persisted in using the objects, their ability to activate the switches was diminished. With regard to getting the boxes open, these children would have been better off employing their own behavioral means of activating the switches (i.e., their hands) and ignoring the actions used by the model.

12In a subsequent study (Nielsen, 2006, Experiment 2), 12-month-olds did use an object in an attempt to activate the switches, but only when the model had successfully used an object after first ‘attempting but failing’ to activate the switches by hand. Thus, it appears that 12-month-olds did not fail to copy the model’s object use because they could not use the object, but rather because they did not interpret this action to be the most efficient alternative available (see also Gergely et al., 2002; Gergely, 2003).

13Thus, from around the middle of their second year children begin to show a tendency for copying the specific actions of an adult model, even if by doing so they fail to achieve the modeled outcome. This tendency is not strong at the beginning of the second year. It was rare for the 12-month-olds in Nielsen (2006) to attempt to touch a switch using an object; they used their hands instead. Thus, although these toddlers attempted to reproduce the result of the model’s actions (i.e., opening the boxes), they did not copy the behavioral means used by the model (i.e., using an object). It appears that 12-month-olds, like chimpanzees, are more prone than 18- and 24-month-olds to ignore seemingly irrelevant actions and to devise their own ways of bringing about the modeled outcome. How might we explain this pattern of findings?

14A growing body of research has shown that the generation of action shares common neural and cognitive mechanisms with the recognition of actions performed by others (for reviews see chapters in Hurley and Chater, 2005). The production of matched behaviour – forming an equivalence between what we see and what we do – thus appears to rely on the same capacities as recognising matched behaviour – detecting an equivalence between what we do and what we see (Nadel, 2002; Mitchell, 2002; Meltzoff and Decety, 2003). Perhaps chimpanzees and young human children are not particularly good at matching their own actions to the actions they have seen performed by others. If so, one could reasonably expect them to show deficiencies in recognizing when others are imitating them.

3 Imitation recognition in Children and Chimpanzees

15A number of studies have investigated the ability of young children to detect when an adult is copying their behaviour (Meltzoff, 1990; Eckerman and Stein, 1990; Asendorpf et al., 1996; Meltzoff and Decety, 2003; Agnetta and Rochat, 2004). One way imitation recognition has been assessed is by having infants sit opposite two adult experimenters. One adult replicates everything the infant does while the second adult replicates the actions of a previously tested infant (commonly presented via pre-recorded video). Thus, both adults exhibit typical infant behaviour, but only one is directly imitating the infant. From 9 months of age infants look more, and smile more frequently, to the imitating rather than to the non-imitating adult. Infants also tend to repeat the particular gesture that is being imitated and engage in ‘testing’ behaviour (i.e., systematically varying activity while closely watching the imitating adult) to see if their mismatching gestures will be copied. It is therefore generally accepted that children from a young age can recognise when their own actions are being imitated. What about chimpanzees?

16The first study to directly investigate imitation recognition in any nonhuman animal was published by Nielsen, Collier-Baker, Davis, & Suddendorf (2005; see also Paukner et al., 2005; Haun and Call, 2008) . Nielsen et al. had an adult human imitate the postures and behaviours of Cassie, a male, captive-born chimpanzee. In a series of control conditions the same experimenter exhibited: (a) actions that were contingent on, but different from, Cassie’s actions; (b) actions that were not contingent on, and different from, his actions; or (c) no action at all. Cassie showed more ‘testing’ sequences and more repetitive behaviour when he was being imitated, than when he was not. Moreover, when the experimenter repeated the same actions she displayed in the experimental condition back to Cassie some 4 months later, these actions now failed to elicit any testing sequences or repetitive behaviour. However, a live imitation condition did. Cassie’s responses to being imitated were therefore similar to the responses of young human children.

17The afore-cited studies indicate that chimpanzees and children younger than 12-months can recognize when another is imitating them. It would appear that the imitative behaviour of chimpanzees and young children cannot be explained by a disruption of, or diminution in, their ability to match their own actions to the actions being performed by others. An alternative proposal is that children from the middle of their second year become increasingly motivated to be social and to promote shared experience with others, and this motivation manifests itself in the way children approach tasks of imitation (Nielsen et al., 2006; Carpenter, 2006; Nielsen and Slaughter, 2007).

4 The social function of imitation

18There is a long-standing though often neglected view that imitation can serve two distinct but complementary developmental functions: A cognitive function that promotes learning about events in the world and an interpersonal function that promotes children’s sharing of experience with others (Baldwin, 1894; Wallon, 1934; Uzgiris, 1981; Mitchell, 1987; Meltzoff, 1990; Meltzoff and Gopnik, 1993; Nadel et al., 1999; Tomasello, 1999a). Uzgiris (1981) argued that as children get older their motivation to copy others changes according to these two functions. Young toddlers are primarily motivated to imitate in order to acquire new skills whereas older toddlers may imitate in the same context in order to satisfy social motivations.

19The variant nature of the functions of imitation is highlighted by changes during the second year in the way pre-verbal infants interact with each other. Towards the middle of the second year children begin to coordinate their own actions with the thematic specifics of a social partner’s play, and this helps generate and sustain ongoing interaction (Nadel and Baudonnière, 1980; 1982; Nadel et al., 1983; Eckerman et al., 1989; Eckerman and Didow, 1989; Nadel and Fontaine, 1989). Children show a preference for engaging with objects that are similar to ones chosen by their play partner, and tend to use the common object in a similar way. When this copying behavior is performed in concert with the play partner, and the partners do not solely adopt one role but alternate between model and imitator, it is referred to as synchronic imitation (e.g., Asendorpf and Baudonnière, 1993; Nadel, 2002).

20In controlled studies of synchronic imitation, an adult experimenter continuously models simple actions on a series of objects (e.g., tapping a toy hammer on the ground) to children who have a duplicate of the object available to them (Asendorpf et al., 1996; Nielsen and Dissanayake, 2004). To be classified as synchronic imitation, children must not only reproduce the actions of the experimenter but do so continuously and simultaneously for a certain amount of time. Using this approach, Nielsen and Dissanayake (2004) assessed 86 toddlers for synchronic imitation at intervals of three months from 12 to 24 months of age. Toddlers sat on a play mat opposite an experimenter. The experimenter took an object and offered the toddler a duplicate of the object. The experimenter continuously modeled an action for 15 seconds and then performed a second action with the same object for a further 15 seconds. This procedure was repeated on an additional three objects. Toddlers were classified as imitating synchronically if they took the duplicate object and, while the experimenter was modeling the action, copied him continuously for at least 3 seconds (following Asendorpf et al., 1996). The duration of the sequence was coded for as long as the toddler maintained imitation of the modeled action and continued to look at the experimenter at least once every ten seconds.

21When aged 12 and 15 months, the infants exhibited little to no synchronic imitation. It was not until 18 months of age that they began to exhibit sustained imitative sequences, and by the 24-month session toddlers were spending approximately one third of the 120-second episode engaging in synchronic imitation with the experimenter. Importantly, the actions Nielsen and Dissanayake (2004) used in administration of the synchronic imitation task were simple (e.g., banging a hammer on the ground). It is thus unlikely that the low exhibition of synchronic imitation by 12 and 15 month olds can be attributed to changes in their ability to either produce or copy the modelled actions. Indeed, Nielsen and Dissanayake noted that when the children were younger it was not uncommon for them to reproduce the target actions of the experimenter during administration of the task; they just did not copy him continuously and simultaneously. When younger, the children engaged in imitation but not synchronic imitation. It has thus been argued that when toddlers engage in synchronic imitation they do so primarily because they want to be social and to interact with the experimenter (Nielsen and Dissanayake, 2003; Nielsen et al., 2006; Nielsen and Slaughter, 2007; Slaughter et al., 2008).

22Uzgiris (1981) also speculated that because young toddlers engage in imitation primarily to promote learning about events in the world they will focus more on copying what was done rather than copying the way it was done. Older toddlers, by contrast, will copy a model in order to engage socially and to sustain interaction. This speculation is reflected in the previously discussed findings from Nielsen (2006) where 12-month-olds, unlike 18- to 24-month-olds, tended not to use the same behavioral means as the model to bring about the target outcomes.

5 Putting the social explanation to test

23At the time of writing, only two empirical studies had experimentally manipulated the sociability of a model in order to investigate how differences in the availability of social interaction impacts children’s copying behaviour (Gergely and Király, 2004, May; Nielsen, 2006; for a more recent study see Nielsen et al., 2008). In Nielsen (2006, Experiment 3), 18- and 24-month-old children were shown how a miscellaneous object could be used to open each of the three previously-introduced boxes (see p. 7; section titled “Copying Actions Over Outcomes”). In a ‘social’ condition the model was typically engaging and interactive whereas in an ‘aloof’ condition the model acted in a detached and disinterested manner: She avoided eye contact and did not direct any attention towards the child. The social disposition of the model affected the children differently. Eighteen-month-olds opened the same number of boxes regardless of condition, but were more likely to copy the model’s object use when she acted socially. In contrast, 24-month-olds used an object at equivalent rates across conditions but opened less boxes when the model was aloof and unengaging. Following Uzgiris (1981), it was argued that 18-month-olds copied the specific actions of the social model in order to sustain interaction and convey mutuality with her. In the absence of social reinforcement, they focused instead on copying what was done rather than copying the way in which it was done. For 24-month-olds, the lack of social interaction provided by the aloof model reduced their motivation to produce the modelled outcome. Rather, they copied her specific actions in an attempt to initiate interaction. These arguments are in line with perspectives that assert a social and communicative role in young children’s copying behaviour (for an alternative view see Gergely and Csibra, 2005; Csibra and Gergely, 2006).

24No study has directly manipulated the sociability of the model in any test of great ape imitation. Nevertheless, there are hints in the literature to suggest that apes do not treat imitative exchanges in the same communicative manner as young children. Two species of chimpanzee (Pan troglodytes and Pan paniscus) and 18-month-old human children were observed whilst engaged in an object-directed imitation task involving an adult human model (Carpenteret al., 1995). All three species showed episodes of joint attention involving the model and the objects. However, the bouts of joint attention were longer in the children and they spent more time looking at the model’s face. Tomasello, Carpenter, Call, Behne and Moll (2005) suggest that these inter-species differences might have arisen because the chimpanzees were looking at the model to see what she was doing or was likely to do next, whereas the children looked to share interest. They further suggest that this might indicate that although apes know that others have goals and perceptions, unlike children, they have little desire to share them.

25If chimpanzees are not inclined to use imitation as a mechanism for social interaction, long-lasting interactions about an object through the alternation of imitating and being imitated are unlikely to be built. That is, evidence of synchronic imitation in chimpanzees is likely to prove elusive. Recent evidence that chimpanzees, unlike 18- and 24-month-old children, are uninterested in social games substantiates this point (Warneken et al., 2006). It is also likely that structured comparison of the responses of children and chimpanzees to being imitated will yield communication-based, inter-species differences. For example, as part of the previously introduced longitudinal study (Nielsen and Dissanayake, 2003; 2004), we tested young children for imitation recognition using a procedure similar to the one employed to assess chimpanzee imitation recognition (Nielsen et al., 2005). Anecdotally, the children clearly found it amusing when they understood that they were being imitated. Testing frequently developed into a game, accompanied by giggles and squeals of delight. The children would even go to great lengths to produce sequences of behaviour that were as difficult as possible for the experimenter to copy. Cassie, the captive chimpanzee, showed none of these signs.

6 Conclusion: The social drive to imitate and the transmission of culture

26From the middle of their second year children will persist in reproducing the exact, at times maladaptive, actions of an adult model – even at the expense of producing the outcome of those actions. Comparable evidence of this kind of imitative behaviour in our closest living relatives has not been forthcoming. That chimpanzees will not slavishly copy a model, as do human children, cannot be attributed to competence. Recall that in the study by Horner and Whiten (2005), when the box was opaque the chimpanzees copied the adult model’s full behavioural sequence – the redundant components were omitted only when the causal relation between the actions and their outcomes was made apparent. They were thus capable of performing the modelled actions. Evidence of imitation recognition also suggests that chimpanzees are not in some way deficient in matching across seen and felt modalities. Rather, the argument put forward here is that, unlike chimpanzees, children are motivated to copy others in order to be social and to promote shared experience with the model. This social motivation manifests itself in a pull towards copying how something was done rather than what was done.

27Over 5 million years of distinct evolutionary history has resulted in a seemingly infinite array of cultural variants in humans, and 39 in chimpanzees. Given the well-argued position that imitation is a fundamental component of the transmission of culture (e.g., Heyes, 1993; Tomasello, 1999a; Whiten et al., 2003; Whiten, 2005; Gergely and Csibra, 2005; Meltzoff, 2005), differences should prevail in the imitative behaviour of these two species. Based on the arguments presented here, the principle difference lies with a peculiarly human tendency to imitate actions over outcomes. For us, at least as children, means are more important than ends. With regard to cultural traditions, this makes intuitive sense. If one thinks about distinguishing cultures, it is how things are done that is most relevant, not what is done. That the Japanese and the British drink tea provides far less insight into their respective cultural traditions than the way they do so. Doing things the way those around us do things lies at the heart of the transmission of culture. A tendency to behave in this way is evident in the way human children imitate from the middle of their second year.

28It has been argued here that young children’s focus on copying actions over outcomes is a function of their motivation to be social and to interact with the model. The maturation of domain general cognitive abilities are likely to be important in the emergence of this motivation (Stone, 2005). Other components of development, such as emotional engagement (Hobson, 2004), intention reading (Tomasello, 1999b; Tomasello et al., 2005; Tomasello and Carpenter, 2005b) and receptivity to being taught (Gergely and Csibra, 2005; Csibra and Gergely, 2006; Gergely and Csibra, 2006) may also play a role. Establishing which of these components of development (and others) determine why young children imitate in the way they do will no doubt form the foundation for much ongoing research and debate. The line between the imitative abilities of human children and chimpanzees remains somewhat blurred. The use of imitation as a form of communication may yet prove to be a distinguishing feature between our closest evolutionary relatives and us.