Cognitive aspects of social learning

Thorndike’s (1911) winged (but disputable) words “Apes badly ape” generate a series of questions such as «Do monkeys ape? " (Visalberghi and Fragaszy, 1990); “do rats ape? ” (Byrne and Tomasello, 1995); followed by Tomasello's (1996) revision " Do apes ape? " to which I have added “Do ants ape? ”. All these questions are derived from a discussion about which, if any, form of social learning is more intelligent. In particular, Whiten (2000) asks a question: Which is more intelligent? Imitation or emulation?

There is a growing body of evidence in literature that observational learning, irrespective of whether it includes imitation or “only” emulation or stimulus enhancement has cognitive implications. It is generally assumed that imitation is a more sophisticated cognitive process. It first of all concerns imitative translation process which includes cognitive implications of how organisms view the behaviour of others, relative to their own behaviour. It implies the ability to take the perspective of another. For this reason, researchers have tried to distinguish imitation from other kinds of social learning and influence. Recent reviews (e.g. Tomasello and Call, 1997) have concluded that only humans, or in some cases chimpanzees, can truly imitate. However, some researchers have claimed that some species of birds can imitate (Zentall, 1996). In fact, researchers are still not certain what mechanisms underlie this ability (Zentall 2003).

Emulation also demands feats of intelligence as it implies that the learner can select from the model’s performance just the new information it needs, and then efficiently combine this information with its own practical knowledge to deal with the task in its own way. I agree with Whiten (2000) in that it may be appropriate to think of imitation versus emulation as a useful but qualified distinction, that makes sense only in relative terms in a particular context, rather than in absolute terms. So in some situations it may have sense to consider “observational learning” without clarifications of types of social learning.

There are several experimental paradigms for comparative studying cognitive aspects of social learning. A paradigm that is known as “do-as-I-do” test allows testing imitation as “kinesthetic visual matching” (Mitchell, 1993), or as a process “especially demanding variety of visual cross-modal performance” (Heyes, 1993). One can then ask if an animal can learn to match any behaviour of another " on cue", in other words, can an animal learn the general concept of imitation and then apply it when asked to do so in a " do-as-I-do" test. Virginia and Keith Hayes (1952) gave intensive training to Viki, their young chimpanzee. They taught her by using an imitation set: whenever Vicki responded to the order " Do this, Vicki" by imitating the experimenter's actions, she was rewarded. Viki learned to respond correctly to the command " Do this! " over a broad class of behaviour. More recently, Custance et al. (1995) have replicated this result under more highly controlled conditions. Using the “do as I do” procedure, the experimenters found that actions on parts of the body that cannot be seen by the performer were just as readily copied as those that could be seen. Both chimpanzees and young children were able to imitate acts that involved parts of their body they simply could not see (facial expressions, head movements, touching parts of the body out-of-sight) just as well as those in sight. The importance of behaviour that cannot be seen by the performer (e.g., touching the back of one’s head) is that it rules out the possibility that some form of visual stimulus matching might account for the behavioural match. The establishment of a “do as I do” concept not only verifies that chimpanzees can imitate, but it also demonstrates that they are capable of forming a generalized behavioural-matching concept (i.e., the chimpanzees have acquired an imitation concept (Zentall, 2003).

An instrumental method that gives wider possibilities for comparative studies of social learning is known as the “ two action method ”, or “ two-ways action/one outcome ” as there are two possible actions which can be performed on one object. Imitation can therefore be tested by finding out whether subjects tend to perform whichever of the two actions they have seen. This can control for displays of other types of social learning such as stimulus enhancement and emulation. This method was first applied by Thorndike (1911) in his studies on chicks. Thorndike noted that those chicks which had a possibility to observe how their companions escape from a puzzle-box, coped with this task faster. He then divided demonstrating chicks into two groups, and trained each of two groups to escape by two different ways. Both ways were available for the observers. They, however, chose that way which they had seen from their demonstrator. This method has been developed in many studies. For example, Dowson and Foss (1965) trained budgerigars, Melopsittacus undulatus, to remove a lid from a cup using either their beak or their foot. When naive budgerigars were allowed to observe one of these techniques they showed a significant tendency to use the same method as their demonstrator.

More recently, the two-ways action/one outcome paradigm was successfully used to show evidence of observational learning in common marmosets, Callithrix jacchus (Voelkl, Huber, 2000), rats, Rattus norvegicus (Heyes et al., 1994), ravens, Corvus corax (Fritz and Kortschal, 1999), and some other species. The majority of studies have used very simple manipulations of simple objects. For example, European starlings, Sturnus vulgaris, were requested to use either pushing or pulling actions manipulating a plug for access to a food reward (Fawcett et al., 2002), whereas pigeons, Columba livia (Zentall, 1996), and Japanise quails, Coturnix coturnix japonica, (Akins and Zentall, 1998) pecked or stepped on a treadle.

In order to examine cognitive aspects of more complex, sequence imitation, experimenters combine the two-ways action/one outcome paradigm with the use of “ artificial fruit ”, that is, a device that should be opened for food reward. To open the fruit, several defences have to be removed, as happens in many natural foods used by many species, especially, by primates and parrots. The artificial fruit may be of different levels of complexities, from a simple plastic container that can be easily opened, say, either by teeth or by extremities, to complex devices equipped with bolts, latches and so on. This combined method offers the possibility of a “gold standard” within comparative imitation research.

Even the use of relatively simple “fruits” have enabled researchers to pose the question of whether an all-or-none test for imitation is appropriate, as successful imitation may depend on the integration of many different sources of information. For instance, common marmosets were displaying some fairly complex social learning, but were not imitating the entire structure of the observed action (see Fig. VIII-4). In experiments of Caldwell et al. (1999) one animal was trained to open an artificial fruit to get a food reward. Another animal was allowed to lick a food reward from the outside of the fruit. Each of these models was observed by four subject animals, all of which were subsequently given an opportunity to open the fruit. Although none of the subjects was able to open the fruit, marmosets appeared to be capable of some degree of imitative matching, and the movements made, though futile, were clearly directed by what they had seen from the demonstrator. Analysis of the subjects’ actions towards the apparatus showed distinct differences between the groups: In their exploratory behaviours, those that had observed the manipulation model used their hands more, and those that had observed the licking model used their mouths more. The first group also devoted attention to the components manipulated, whereas the second focused on the parts they had seen licked.

The use of the artificial fruit paradigm across different species led to the conclusion that some primates are more skilled imitators than others. In experiments of Whiten et al. (1996) chimpanzees and young children were presented with adult human models opening an artificial fruit in one of two alternative ways. In one experiments the defence consisted of a pair of bolts that had to be either poked out through the back, or pulled out at the front with a twisting motion to open the lid and to gain the edible treat inside. In another experiment a pin was spun round and removed using one of two different methods, after which a handle could be disabled by either pulling it out or turning it to one side, allowing the lid to be opened. Chimpanzees were found to copy the method they witnessed being used to remove the bolts, as did children. However, while the children also imitated the method of handle removal, the chimpanzees did not - all tended to use the same method of pulling out. Thus, in a situation when young children learned a technique with quite high fidelity, chimpanzees did not copy all they witnessed so faithfully. Further studies enabled researchers to suggest that, as it has been already noted before, the capacity for true imitation is restricted to humans and apes only, more precisely, to children and chimpanzees rose in a human environment.

It is still not completely clear to what extend our close relatives use their brains when copying behaviour of companions. Several authors consider chimpanzees and gorillas being able to copy hierarchically organised patterns of food processing in nature and in laboratory (Byrne and Byrne, 1993; Byrne and Russon, 1998) whereas others argue for a kind of sequential emulation in apes (Whiten, 1998; Povinelli, 1999), all demanding abilities to extract and remember the basic plan of the action sequence.

It has been recently shown that autistic children display widely ranging imitation deficit whereas they do not differ from normal children in performance in emulation tests (for a review see: Heyes, 2001). This enables to consider imitation a part of the normal development in our species which includes predisposition for copying actions of close company. It is interesting to note that cross fostering experiments revealed shifts to foster parents’ behaviour just to that extend to what members of adopted species are predisposed for mimicking and imitation. As we have seen in Part VII, this was clearly demonstrated on birds, and we can also recall an experiment in which a young fox performed species specific behaviour of his own species with only minor shifts to the behaviour of his foster mother, the dog (Mainardi, 1976). Similar results were obtained with ants raised by members of other species. It is, however, noteworthy that minor but distinct changes in motor patterns were revealed in adopted ants similar to those of their “mentors” (Reznikova, 1996, 2003). As far as human infants are concerned, nobody could think about cross fostering experiments but there are some documentary evidences from abandoned children who grew up together with animals (in most cases, with dogs). These infants copied a number of behavioural acts from their companions: eating by licking, walking on four feet and using sounds similar to those of the dogs (Mikló si, 1999). Meltzoff (1988) argues that humans are genetically predisposed to imitate others, and this predisposition allows us to become imitative generalists.

Some insight that helps to clear the grey area between imitation and other types of social learning and to enlighten cognitive aspects of these came from studies of exploratory or curiosity behaviour, for which social learning may be particularly adaptive (Mikló si, 1999; Huber et al., 2001). As we have seen in Chapter 17, there is a high level of specificity in displays of exploratory behaviour and the tendency of animals to explore details of their environment correlates with coping with intelligence tests (Reznikova, 1982).

Huber et al. (2001) have investigated how social learning affects object exploration and manipulation in keas, Nestor notabilis. This New Zeland parrot, as the authors note, has been used as an example of curiosity in birds for a century, and its natural habitat is thought to have led to the evolution of extreme behavioural flexibility (see Fig. VIII-5).

Researchers adapted the experimental paradigm used by Whiten at al. (1996) to investigate imitation of foraging technique in chimpanzees (see above in this paragraph). Five young keas were allowed to observe a trained conspecific that iteratively demonstrated several techniques to open a large steel box. The lid of the box could be opened only after several locking devices had been dismantled: a bolt had to be pocked out, a split pin had to be pulled, and a screw had to be twisted out (Fig. VIII-6). The observers’ initial manipulative actions were compared with those of five naive control birds (non-observers). Although the kea observers failed to open the box completely in their first attempts, they explored more, approached the locking device sooner and were more successful at opening them. These results provide evidence for effects of social facilitation and both generalised and local stimulus enhancement on object exploration in this species. The obtained data also suggest that the keas definitely learned something during observation. Although their initial attempts did not match the response topography or the sequence of model’s actions, the birds’ efficiency at unlocking the device seemed to reflect the acquisition of some functional understanding of the task through observation, that is, emulation learning.

There was no evidence of true imitation in keas, and a salient explanation given by the authors is that the kea’s propensity for exploration, object play and demolition runs counter to the exact reproduction of movements demonstrated by others. Keas are justly mentioned “chimpanzees among birds”. Their dynamic and playful style of life does not coincide with close watching and imitating actions of others. Being attracted by a conspecific to explore a novel object does not necessarily lead to slavish copying but may lead to learning what parts of the object are worth exploring (Huber, 1998). Together with data on anthropoids, the cited results on parrots enable us to regard emulation learning as being cognitively quite demanding.

It is important for comparative imitation studies that many factors should be taken into account in order not to place a species into a list of “backwards”. Among these factors, motivation is of great importance as well as ranking and “self-confidence” of individuals that play roles of demonstrators. For example, von Holst (1937) observed that in wrasses, fishes which live in shoals, it is enough to extract a forebrain in an individual and it will become a recognized leader in its shoal. The fact is that, if surgically deprived of its forebrain, a wrasse loses its shoal-forming reactions. The lobotomised fish braves to swim where it pleases and the shoal tags along behind it.

In wrasse shoals, as in many other fish species, no fish has individual knowledge of another. For members of personalised animal societies it is important to copy actions of highly ranked individuals. It was taken into account in the Huber at al.’s work on keas cited above: highly ranked birds were appointed as demonstrators. Another example comes from Van… atov< ’s (1984) study in which imitation behaviour in capuchins Cebus apella was clearly demonstrated while in other studies this species were unable to learn to use efficiently a tool they had repeatedly observed being used by others. In the cited experiment the monkeys were highly motivated by a nature of a reward, that is, a little mouse to eat instead of items of vegetarian diet, and, what is not less important, it was the dominating individual who was used as a demonstrator.

In cases of interspecies social learning naive animals benefit from watching skilled demonstrator mastering detour tasks in order to obtain reward. Let us consider two examples.

The first example concerns Pongrá cz et al.’s (2001) study of the role of social learning in interactions between pet dogs and humans. Dogs had to solve several variants of a detour problem. The ability of dogs to solve detour problems was described in Chapter 12. Here we are interested in the analysis of dogs’ ability to improve their results by observing a human demonstrator. In series of tasks, an object or food was placed behind a transparent, V-shaped wire mesh fence. In certain groups of subjects tested, two opened doors were offered as an alternative way to get inside the fence. In Experiment 1 the doors were opened only in trial 1, then they were closed for both trial 2 and 3. Other dogs (Experiment 2) first were taught to detour the fence with closed doors after having observed a detouring human demonstrator, then the doors were opened for three subsequent trials. In Experiment 1 all dogs got the object through the doors in trial 1, but their detouring performance was very poor after the doors had been closed, if they had to solve the task on their own. However, in another experimental group dogs were allowed to watch a detouring human demonstrator after the doors had been closed. These dogs showed increased detouring ability, in comparison to the first group. In Experiment 2 the dogs tended to keep on detouring along the fence even if the doors were opened, giving a chance of getting behind the fence on a shorter route. These results show that dogs use information gained by observing a human demonstrator to overcome their own mistakenly preferred solution in a problem situation. In a reversed situation social learning can also contribute to the emergence of preference for a less adaptive behaviour. However, only repeated individual and social experience leads to a durable manifestation of maladaptive behaviour.

The second example concerns the use of public information heterospecifically by two ant species connected by interspecies cleptoparasitic relations. What is of special interest here is that a subdominating species serves as a scout for dominating one when it is necessary to obtain food in complex situations (Reznikova, 1982). In the Siberian steppes Formica pratensis is a territorial dominant over Formica cunicularia, driving its workers away from favourite nest sites and food finds. Occasionally, F. cunicularia foragers can steal food items that have been temporarily laid on the ground by pratensis foragers, but in general they have to rely on quickness and luck to gather food before pratensis arrive on the scene. For their part the pratensis use the cunicularia as scouts. Reznikova (1982, 2001)

tested this form of inter-species transferring of information in field experiments presenting the ants with food baits hidden in a “sectorial maze”. Meat baits were offered in a central area, the approach to which was divided into ten equal sectors (Fig. VIII-7).This apparatus was presented in plots visited by only one of two species, and in plots visited by both. Improvements in the ease with which the bait was found over periods of exposition of the maze during three hours at each experimental plot were recorded. The improvements of ants’ actions were assessed by assigning error points which were the sum of the sector numbers entered before the bait was reached. Individual marking showed that on each plot where the species were active a constant group of about 10 F.pratensis ans 2-3 F. cunicularia were working.

It turned out that within 1-15 minutes after the beginning of trials the bait was found by F. cunicularia ants and quickly dragged away. In the case when mistakes were made they ran around and entered the correct sector. This species needed not more than 3 or 4 visits to remember the right sector. F. pratensis, on the contrary, roamed about the sectors for an hour, or even two, and found the bait only by chance. Later on their searching became orderly, but the sum of errors remained high. If F. pratensis had the possibility to observe F. cunicularia then they did not touch the bait at all for the first 30-40 minutes, allowing the subdominant to drag it away. During this time F. pratensis foragers stood near the maze watching the actions of the scout-species. Then they drove F. cunicularia away and collected the food pellets by themselves. Judging by analogous data obtained from the example of F. uralensis and F. picea, and also of F. pratensis and F. rufibarbis (Stebaev and Reznikova, 1972, Reznikova, 1994), it may be assumed that relations in ant communication of this type are in general characteristic of ecologically similar pairs of species playing the role of dominants and subdominants.

These examples of interspecies social learning can be attributed to the generalised notion of observational learning which includes elements of stimulus enhancement and emulation and can be considered cognitively demanding.

26. THE SPREAD OF INNOVATION WITHIN POPULATIONS

As we have seen from the previous chapter, in laboratory studies experimenters create “innovators” by themselves. They choose active and exploratory animals that have high ranks in their groups, train them to solve a problem and after training let them to “inculcate” new knowledge among naï ve members of social groups. This way of experimental investigation helps to enlighten a process of social learning and to estimate potentials of different species. However, such an approach does not give a possibility to learn how innovations spread within populations in the wild.

It is an intriguing question whether a single prodigy individual or may be several advanced individuals can propagate a new tradition in animal community. To catch sight of the transmission of novel behaviour in groups of animals, detailed observations in natural populations are needed, supplemented by experiments in captivity. Sometimes researchers are lucky to witness the gradual establishment of a new tradition. In the majority of cases described in literature new traditions concern vital situations such as feeding techniques or fear of predators. However, exquisite patterns of social behaviour such as specific modes of grooming or mating rituals can also serve as subjects for discussion.

In this chapter we will consider possible ways of establishing new behavioural traditions within groups of animals as well as characteristics of potential innovators.