Portal de Periódicos da CAPES

Animal imitation

2009; Elsevier BV; Volume: 19; Issue: 3 Linguagem: Inglês

10.1016/j.cub.2008.11.027

1879-0445

Richard W. Byrne,

Action Observation and Synchronization

The term 'imitation' has a range of meanings in everyday usage and no single agreed definition in science. In biology, imitation has usually referred to morphological adaptations for camouflage or mimicking the appearance of another species (Figure 1). Only recently has there been intense interest in the imitation of behaviour by animals; animal learning theory has traditionally ignored imitation. One purpose of this Primer is to help explain why researchers do now care about animal imitation; another is to chart the various kinds of action imitation that may be important for biology and illustrate the sometimes confusing array of terms that have been coined to describe them. Among the variety of definitions of 'imitate' found in English dictionaries, three quite distinct senses are generally apparent. All three have biological equivalents in the adaptive functions served by animal imitation, and I shall use these as a structure to understand what cognitive mechanisms are required for the different forms of imitation. One objective of imitation in everyday life is to resemble as closely as possible the individual whose behaviour is copied, usually because their behaviour is judged to be admirable or because the imitator wishes to be seen as like them in some key ways ("imitation is the sincerest form of flattery"). There are several circumstances in which animal imitation of this sort might be adaptive; most are cases where the imitation is immediate and the result is behavioural synchrony. Predators are thought to target individuals that stand out in some way, so behaving just like other members of a flock or herd may reduce risk. A general tendency to match the current actions of others, when in a group, might therefore be expected to evolve in social species. Moreover, by copying the current behaviour of other group members, individuals of social species can potentially gain from others' knowledge of a range of costs and benefits that are associated with place, including local level of predation risk, locations of optimal feeding sites, and what is edible there. Note that these gains come automatically from behavioural matching, not as a result of knowledge transfer. Behavioural synchrony is in fact reported in a wide range of social animals, and extends to a detailed level of activity copying, such as preening together [1Hoppitt W. Blackburn L. Laland K.N. Response facilitation in the domestic fowl.Anim. Behav. 2007; 73: 229-238Crossref Scopus (39) Google Scholar], or the nearly simultaneous turning of each individual in a flying flock of shorebirds that generates such spectacular aerial acrobatics. In some instances, the adaptive function of behaviour matching is less obvious. Contagious yawning is found in chimpanzees as well as humans, and the synchrony appears to involve arousal level as well as overt activity: one possibility is that synchronization of sleeping is adaptive, or was in human ancestry. Human yawning is also contagious to dogs, but it is not yet known whether dogs affect each other in this way: in a long-domesticated species, the function may instead relate to how dogs mesh with human behaviour. Cognitively, imitation that produces immediate behavioural synchrony requires an individual to recognize specific actions in others' behaviour that are already in its own repertoire. This sort of copying may be understood simply as response facilitation, where seeing an action 'primes' the individual to do the same [2Byrne R.W. Imitation of novel complex actions: What does the evidence from animals mean?.Adv. Stud. Behav. 2002; 31: 77-105Crossref Google Scholar, 3Laland K.N. Animal cultures.Curr. Biol. 2008; 18: R366-R370Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar]. No special mechanisms are required to understand performance of the action, as no new behaviour is added to the individual's repertoire. A different kind of imitation is when the copying itself conveys a social signal. Postural mimicry, in which two people who like or love each other quite unconsciously adopt the same or a mirror image of body posture has long been noted by social psychologists. Very young, even new-born infants copy the facial gestures of adults interacting with them, for instance smiling or tongue-protrusion [4Meltzoff A.N. Prinz W. The Imitative Mind: Development, Evolution, and Brain Bases. Cambridge University Press, Cambridge2002Crossref Google Scholar]. This 'neonatal imitation' may increase maternal investment in the child by signalling alert awareness and cognitive competence, and has also been described in chimpanzees. Adult humans use imitation of actions to signal that a social connection has been made, for instance returning a wave. Many greeting signals and ceremonies of animals might be thought of as imitation of this kind, but since these actions are stereotyped they may simply be evoked by the social circumstances. Although monkeys have been described as unable to imitate, when their actions are instantly copied by human experimenters they react strikingly, suggesting signalling social connection by imitating is part of the natural communicative repertoire of monkeys. People also use mimicry to mock or deride others, as in the cruel mimicry of schoolchildren and in sophisticated satire. The capacity to derive pleasure from manipulating the feelings of others implies a rich understanding of mental states [5Frith C. Frith U. Theory of mind.Curr. Biol. 2005; 15 (R655–R645)Google Scholar]. Intriguingly, chimpanzees at the Burger's Zoo, Arnhem, have been seen to imitate the limp of a disabled group member. However, wild chimpanzees at several sites suffer high frequencies of disability from snare injuries, yet no similar mimicry has ever been noted. Rather than enjoying cruelty, it may be that the Arnhem chimpanzees were influenced by a general ape tendency to behavioural synchrony and the suffering of the mimicked individual was incidental; ascribing understanding of complex mental states may be unwarranted. In all these cases, any social message conveyed is an immediate one, but humans also show a sort of imitation when they converge on the behavioural norms of their social group, conveying the message of group membership — and conversely, out-group exclusion. Conformity to group norms of behaviour has long been studied in social psychology. In chimpanzees and rats, a general tendency to conform to the actions of the majority of the group has also been reported and described as conformity. Individuals that have already discovered how to open a puzzle box satisfactorily (chimpanzees), or found a food that is palatable (rats), switched to the actions and choices of the majority. In these experiments conformity was valueless, but the behaviour may be a by-product of a trait that functions in reducing exposure to risk in unpredictable environments. Certainly, both chimpanzees and rats have been noted as conservative in behaviour, for instance being remarkably cautious about trying novel foods. A similar explanation may apply to the short-lasting 'fads' shown by young gorillas, such as the once-popular habit of pushing through the legs of researchers rather than walking around them. Finding behavioural conformity among our closest relatives, the great apes, it is tempting to presume that its function might relate to a sense of group identity, as in humans. Male chimpanzees certainly show violence to members of other communities in a way that is disturbingly human. However, there is no indication that intercommunity violence is predicated on non-conformity, arguing against such a rich interpretation. The cognitive mechanism of imitation when used as a social signal is likely to vary with the actions copied and the motivation to do so [2Byrne R.W. Imitation of novel complex actions: What does the evidence from animals mean?.Adv. Stud. Behav. 2002; 31: 77-105Crossref Google Scholar]. Where the actions are 'transparent', that is, they look much the same from the perspective of mimic and model, then priming or facilitation of specific actions may be all that is required, as in the previous category. With neonatal imitation, however, the action copied is wholly opaque and any explanation will require an innate system that matches observed facial gestures with the motor commands needed to imitate them. It is difficult to determine how extensive a set of correspondences might exist, when the neonate has such a limited motor repertoire; theorists range from those who attribute hard-wired matching of just a few key facial gestures to those who posit in the infant the ability to match 1:1 the complete motor repertoire [4Meltzoff A.N. Prinz W. The Imitative Mind: Development, Evolution, and Brain Bases. Cambridge University Press, Cambridge2002Crossref Google Scholar]. Where lasting consistency is seen, a tendency to conform is additionally required. This can override what has been learnt from individual experience; and the bias to social learning is frequency-dependent, more potent if more individuals are showing the same actions. Learning how to do something from seeing it done might seem the most obvious sense of imitation, yet scientific interpretations of it have varied from a cheap monkey trick to an exalted pinnacle of animal cognition. Even now, it is unclear which species of animal have the ability to learn by imitation. Partly this relates to semantic confusion, between learning by imitation and learning that follows from imitation. As we have noted already, the tendency to copy the actions of nearby conspecifics occurs widely among animals and can be straightforwardly explained in most cases as a result of priming those responses that match actions seen. Where animals are engaged in instrumental activity, this behaviour-matching tendency may result in a naïve individual applying actions that are in fact appropriate to the task sooner than it otherwise might. If that brings success the animal will often learn to use these actions again. A tendency to copy — whether described as imitation, response facilitation or priming — can therefore accelerate learning, and indeed such benefits may have contributed to the evolution of response facilitation in some species. But the learning itself is consequent on getting a favourable result: reinforced trial-and-error learning, in behaviourist jargon. Most experimental tests of animal imitation can be explained in this way, because the experimenters presented the task immediately after the subjects had observed a skilled performer, with rewards contingent on success. Learning by imitation can be shown by introducing a delay before testing, as has been done successfully with both quail and budgerigars. In these cases, the animal evidently learnt to link an action in its repertoire — for example, pulling, pecking or stepping — with a particular task, by seeing it used. This has been described as contextual imitation, because what the animal learns is when and where to apply an existing behavioural tactic in its repertoire [2Byrne R.W. Imitation of novel complex actions: What does the evidence from animals mean?.Adv. Stud. Behav. 2002; 31: 77-105Crossref Google Scholar]. The Holy Grail of animal imitation studies, however, is to discover a species that is capable of learning a new skill by observation. Another species, that is: there is no doubt that humans can learn in this way, even though in western cultures it is now more common to acquire new skills with some form of verbal instruction. Examples of human skills that seem particularly dependent on learning by imitation are sushi-making, blacksmithing, and stone-working by medieval masons. In a more mundane context, most of us have watched an expert take apart a machine — perhaps a mechanic helping us fix our automobile — and come away knowing, however imperfectly, how to do it ourselves next time. This kind of imitation, called production-imitation, involves an individual constructing a behavioural routine new to it, out of components in its pre-existing repertoire, from watching a more expert model [2Byrne R.W. Imitation of novel complex actions: What does the evidence from animals mean?.Adv. Stud. Behav. 2002; 31: 77-105Crossref Google Scholar, 6Byrne R.W. Russon A. Learning by imitation: a hierarchical approach.Behav. Brain Sci. 1998; 21: 667-721PubMed Google Scholar]. Of course, it is unlikely that an entirely novel task will be mastered in a single viewing without practice or relevant prior experience; but crucially, production-imitation allows some part or outline of the task to be put together from observation before ever trying it out. Cognitively, the requirements for imitating a good example are quite different to those cases where the function is to resemble another individual or give out a social message [6Byrne R.W. Russon A. Learning by imitation: a hierarchical approach.Behav. Brain Sci. 1998; 21: 667-721PubMed Google Scholar]. Fine detail is unimportant as long as the right result is obtained, and indeed it is likely that details are better acquired by individual experience. A child, for instance, cannot copy the precise way her mother operates on objects, as she has smaller hands and less strength: precise copying of every action would be maladaptive. Moreover, the potential benefits of learning by imitation vary with task difficulty: where complexity is low, purely individual learning is to be expected, avoiding the costs of finding and watching a skilled model. It is therefore in learning the most technically complex and involved tasks that imitation should pay. Complex skills cannot be assembled in a single attempt, so we should expect acquisition to be hierarchical, progressively building up larger and larger components. That is true whatever learning mechanism is involved, but if the gist of the right approach can be discerned by imitation, learning can progress by organizational leaps rather than by plodding acquisition of actions in sequence. Observational learning of the organizational gist of a task has been termed 'program-level imitation' [6Byrne R.W. Russon A. Learning by imitation: a hierarchical approach.Behav. Brain Sci. 1998; 21: 667-721PubMed Google Scholar]. Program-level imitation is nicely illustrated when a child copies a word she has not heard before. The child's sound pattern is typically quite different to that of the adult model, with much higher-pitched vowels and often systematic simplification of consonant clusters. This shows that what is copied is the program-level gist of the word, a new way to assemble the motor programs for producing vowels and consonants — which are already in the child's repertoire. Crucially, this process of synthesis depends on prior analysis that parses the adult's sound into its component units. Compare this with the same word imitated by a myna bird: the resulting sound pattern is so close to the human model that it can be hard to tell them apart spectrographically. Because the vocal apparatus of the bird is quite different, a double syrinx rather than a single larynx, it is clear that the bird matches the audible result rather than the organization of behaviour. Copying of results rather than actions, termed emulation learning, has sometimes been considered 'simpler' for animals than imitation and assumed to underlie animal behaviour that appears imitative. The evidence from child psychology, however, suggests the reverse when it comes to practical tasks with objects: extensive evidence exists for early child imitation, but none for effective emulation prior to 4–5 years. This perspective makes especially remarkable some of the emulation noted in cetaceans: for instance, a bottlenose dolphin calf in an aquarium, teased by a visitor blowing smoke at the glass, went to its mother to suck milk, then returned to the visitor and blew a cloud of milk towards him, like smoke! Unsurprisingly, the strongest evidence for program-level imitation in animals comes from the technically complex achievements of the great apes (Figure 2). Chimpanzees in northern Congo regularly make two different types of tool in advance of arriving at sites where subterranean termites can be found. A sturdy rod is used to punch a hole deep enough to reach the termites, then a more delicate probe is used to agitate the termites, which bite onto the stem and can thus be fished out [7Sanz C.M. Morgan D.B. Chimpanzee tool technology in the Goualougo Triangle, Republic of Congo.J. Hum. Evol. 2007; 52: 420-433Crossref PubMed Scopus (166) Google Scholar]. Rwandan gorillas regularly process a local species of stinging nettle: stripping up stems to detach the leaves, twisting off the petioles, sometimes repeating both these steps several times to accumulate a larger handful, then pulling up the leaf-bundle and folding it over the thumb, finally re-grasping the package before ingestion. In both chimpanzee and gorilla examples, the actions used at each processing stage are highly specific, and it is very improbable that each single individual could invent the appropriate process without some information from an expert model. Great apes that live with people copy human actions, sometimes remarkably complex ones. One orangutan was noted pushing glowing embers together, fanning the embers with a flat plate, decanting kerosene into a container and putting it on glowing embers — fortunately, not completely efficiently or the jungle camp might have burnt down [8Russon A.E. Galdikas B.M.F. Imitation in free-ranging rehabilitant orangutans (Pongo pygmaeus).J. Comp. Psychol. 1993; 107: 147-161Crossref PubMed Scopus (116) Google Scholar]. In all these cases, the individuals concerned had prolonged opportunity for casual observation of skilled models, and their imitative ability can be explained by a perceptual process, behaviour parsing, which detects the statistical regularities of complex but repeated behaviour [9Byrne R.W. Imitation as behaviour parsing.Phil. Trans. R. Soc. Lond. B. 2003; 358: 529-536Crossref PubMed Scopus (133) Google Scholar]. Imitation by behaviour parsing is presumably part of the human repertoire, too, but human children go beyond this relatively unselective process when they imitate. Eighteen-month-old infants, when shown how to turn on a light by an adult who leans forward to press the switch with their forehead, often copy the whole performance — even though they could use a hand more easily. But when researchers modified the task so that the adult was holding a cloak around their shoulders, infants never copied the use of the forehead [10Gergely G. Bekkering H. Király I. Rational imitation in preverbal infants.Nature. 2002; 415: 755Crossref PubMed Scopus (807) Google Scholar]. Apparently, in some way the infant is able to compute that forehead is only needed because hands are occupied, and can therefore be ignored. Something similar to this 'rational imitation' has been seen in chimpanzees, presented with a demonstration of how to get food from a puzzle box [11Horner V. Whiten A. Causal knowledge and imitation/emulation switching in chimpanzees (Pan troglodytes) and children (Homo sapiens).Anim. Cogn. 2005; 8: 164-181Crossref PubMed Scopus (605) Google Scholar]. When the box is opaque, the apes copy both actions used by the experimenter; but when it is transparent, thus revealing that the first action makes no contact with the food, they ignore it. It is not yet clear whether chimpanzees will indeed prove to show rational cause-and-effect understanding of what they imitate, or whether simpler explanations such as behaviour parsing will suffice. But a recurring theme in this article has been the intriguing behaviour of great apes (and occasionally cetaceans), that has repeatedly led researchers to suspect cognitively complex explanations. In contrast, imitation of actions is widespread and biologically important among other taxa, but the cognitive mechanism of these cases is relatively straightforward to understand.