Social Cognition: Social Learning

Question 1

A Social Learning Glossary

Answer 1

Social learning embraces such a potentially bewildering (C. M. Heyes 1993a) variety of different terms that a glossary is useful for keeping track. This one includes those most often encountered in contemporary discussions (Zentall 2006; Hoppitt et al. 2008). Historically, there have been many more, often with overlapping meanings (Galef 1988; Whiten and Ham 1992). To begin with, any form of social learning requires an observer (or actor) and a demonstrator, who performs the behavior later reproduced in whole or part by the observer.

To qualify as learning rather than socially elicited or facilitated behavior, the observer’s performance must take place at a later time, away from direct influence of the demonstrator.

Copying - A generic term for doing the same thing as a demonstrator, mechanisms unspecified; for example copying another’s choice of mate or foraging patch (Section 13.1).

Social facilitation - Individuals are more likely to perform a behavior when in the company of others performing it. For example, yawning is socially facilitated in people (Provine 2005).

Local enhancement/Stimulus enhancement - Increased likelihood of visiting a place (local enhancement) or contacting a type of stimulus (stimulus enhancement) by virtue of observing others doing it. The enhanced attractiveness of the location or stimulus may or may not be confined to times when demonstrators are present.

Observational conditioning - Associating a cue or object with an affective state or behavior(s) by virtue of watching demonstrators respond to it. For example (Section 13.2), having seen other birds mob an owl, an observer later responds to an owl by mobbing. Sometimes extended to cases in which the observer is directly reinforced following a cue or signal by the demonstrator as when parent babblers ‘‘purr’’ before feeding their young (Section 13.4). However, this seems to be direct conditioning of the observer, that is, CS ¼ purr, US ¼ food, CR ¼ approach.

Imitation. Performing the same action as a demonstrator by virtue of having seen the action performed. The action must be novel, thus ruling out such phenomena as ‘‘mate choice copying.’’

Emulation - Copying only elements of a complex action. For example, having seen a demonstrator skillfully use a rake to pull food toward itself, an observer picks up the rake backwards and waves it in the general direction of the food. May be qualified by reference to the element of the sequence apparently emulated, as in goal emulation (Section 13.3).

Learning affordances - Learning what can be done with objects or parts of the environment, not
necessarily through observing the actions of another animal. For examp

Question 2

13.1 Social learning in context

Answer 2

13.1.1 Social transmission of food preferences in rats

One advantage of group living is that individuals foraging together may help each other find food. They may be attracted to feeding conspecifics, or they may follow others, as ants follow each other along chemical trails. Colonies or roosting places may serve as information centers where individuals inform each other about good foraging opportunities in the neighborhood. At one time information exchange was hypothesized to be a major factor in the evolution of sociality, but this information center hypothesis is now considered to be without broad empirical support (see Galef and Laland 2005).

Nevertheless, information exchange is a potential benefit of sociality, and there are some good examples of animals using information about food sources provided by others in their colonies. Bees communicate the locations of nectar (Chapter 14), and as we see next, rats provide other rats with information about the flavors of edible foods. Norway rats (Rattus norvegicus, the common laboratory rat) are colonial omnivores. They can and will eat almost anything that does not poison them. This means that young rats have a lot of potential foods to learn about, and they start learning
before they are born.

The flavors of foods eaten by a mother rat late in pregnancy influence the food preferences of her offspring when they begin to feed on solid food
(review in Galef 1996b). The pups continue to learn from their mother when they are suckling because the flavors of foods she ingests are present in her milk. In addition when the weanling rats begin to leave the nest to forage, they prefer to forage where other rats are or recently have been feeding. Thus the young rat has at least three ways to become familiar with the flavors of foods being eaten safely by its mother and others in its colony. Combined with a preference for familiar over novel flavors, they almost guarantee that a young rat will eat things that are good for it, or at least not
harmful.

In addition to choosing familiar flavors, both young and adult rats choose foods being eaten by their companions over alternatives. This was discovered in experiments designed as depicted in Figure 13.2 (Galef and Wigmore 1983). Pairs of rats lived together for a few days, eating normal laboratory rat chow. Then one rat in each pair, the demonstrator, was removed to another cage and deprived of food for 24 hours before being fed cinnamon or cocoa flavored chow. Next, each demonstrator was returned to its familiar companion, the observer rat, and demonstrators and observers interacted in the absence of food for 15 minutes.

For the following 24 hours the observer, alone once again, had two bowls of food, one flavored with cinnamon and one with cocoa. As shown in Figure 13.2, during this time observers whose demonstrators ate cinnamon consumed more cinnamon-flavored food relative to cocoa-flavored food than those whose demonstrators ate cocoa. A large number of related experiments has shown, among other things, that observers can be socially induced to choose a familiar food that a demonstrator has eaten recently and to seek out a place where that food is available (Galef 1996b).

Thus rats apparently can, in effect, exchange information about what foods are currently available nearby, although the role of these processes in directing food choice in wild colonies is unknown. Such learning is also found in other rodent species (Galef
2007). How do demonstrators communicate about food? A rat that has just been eating might carry bits of food on its fur and whiskers, but that is not all the observers detect. Observers need to smell the flavor on another rat’s breath, more specifically in association with carbon disulfide, a prominent component of rat breath (Galef 1996b).

Rats behave in a way that facilitates this learning: when they encounter one another they engage in mouth to mouth contact and sniffing. To borrow a term from embryology (Waddington 1966), development of food preferences in rats is canalized: in a kind of fail-safe system often found in development, several separate mechanisms independently and redundantly ensure that young rats will eat what others in their colony are eating. The social learning mechanisms available to adults are sufficient to transmit colony members’ acquired food preferences to succeeding generations (Galef and Allen 1995).

In one example colonies of four rats were induced to prefer either Japanese horseradish or cayenne pepper flavored food by making them ill after they ate the alternative diet. The rats in these ‘‘founder’’ colonies were gradually replaced with naive rats until the colonies were made up entirely of rats that had never been poisoned after eating either of the diets. Nevertheless, rats in each colony were still preferring their colony’s ‘‘traditional’’ diet. In one experiment, the tradition was maintained over four generations of replacement rats. Preference was still transmitted
even when the new colony members never fed in the presence of the older members but just interacted with them in the hours between daily feedings.

13.1.2 Producing and scrounging: Social transmission of feeding techniques in pigeons

Baby rats represent a special case in which social learning is undeniably useful. Without influences from their mother and other adults, they would have to choose foods randomly once they were weaned. But in a group of adult animals encountering unfamiliar resources, not everyone should be engaging in social learning. Indeed, this would be an impossible situation: for there to be anything to learn socially, someone has to be acquiring information for himself, that is, engaging in individual learning. Thus when there is new information to be acquired, some should learn for themselves while others copy (Giraldeau, Valone, and Templeton 2002).

And if a given individual already has an effective behavior for the situation, copying may not be his best policy. This informal functional notion suggests that
animals might not always acquire a novel behavior being exhibited by another group member. The research of Louis Lefebvre, Luc-Alain Giraldeau, and their colleagues with captive and free-ranging pigeons (Columba livia) provides some of the best evidence for this suggestion. Like rats, pigeons are highly social opportunistic foragers that are widely associated with humans because of their ability to flourish in a variety of
conditions.

Pigeons in the laboratory learn some novel feeding techniques more readily if they have seen them used by another pigeon than otherwise. One such technique is pecking through a paper cover on a food dish. Pigeons
that watch demonstrators both pierce the paper and eat grain perform the task themselves sooner than pigeons given partial demonstrations or no demonstrations (Palameta and Lefebvre 1985). However, when a skilled paper piercer was placed in a laboratory flock of ten birds, only four learned the skill. The others scrounged food uncovered by the birds that pierced (Lefebvre 1986).

In contrast, when a trained demonstrator was introduced into a free-flying flock in Montreal, 24 birds learned to pierce on their own and only four specialized in scrounging. The sample sizes here are just one captive and one free-living flock, but Lefebvre and Palameta (1988) suggest that one reason for the great difference in proportion of learners is that because individuals could come and go from the urban flock, scroungers sometimes found themselves without anyone to scrounge from and had to learn for themselves to produce food from the apparatus.

In free-ranging flocks different individuals may specialize in different food-finding skills and change roles from producer to scrounger as the situation changes (Giraldeau and Lefebvre 1986). Opportunity to scrounge may reduce performance of a task that has already been learned, but it can also interfere with learning from producers in the first place. When pigeons learned to remove a stopper from an
inverted test tube, causing grain to fall out (Figure 13.3; Giraldeau and Lefebvre 1987), eight out of eight observers that watched another pigeon remove the stopper and eat the grain did the same themselves when given the opportunity.

If the observers could scrounge some of the demonstrator’s grain, however, only two out of eight
birds learned in the same number of trials. Just as with paper piercing, when a trained observer was introduced into a laboratory flock, a few birds learned the tube-opening task and became consistent producers, whereas the majority scrounged as long as the
producers were present. Taken together, these observations indicate that scrounging influences learning, perhaps because pigeons cannot divide attention between looking for food to scrounge and watching what a demonstrator is doing. Some other species of birds, however, may be able to scrounge and learn at the same time (Lefebvre and Bouchard 2003).

13.1.3 Public information, cues, and signals

A young rat approaching a food site frequented by other rats is already familiar with the flavors of some safe foods. This is private information. In contrast, the rat excrement and odors of other rats around the site constitute public information that quantities of edible food are present, and indeed these cues attract rats (Laland and Plotkin 1993). Similarly, by trial and error a Montreal street pigeon might acquire private information about how to open one of Giraldeau and Lefebvre’s feeders, but to find a good foraging patch it could use public information like the sight of a flock of pigeons feeding.

At the end of the twentieth century, uses of public versus private information became a lively topic in behavioral ecology (Danchin et al. 2004; Valone 2007), in parallel with interest in eavesdropping in animal communication (Chapter 12). When animals respond to the behavior of other animals or a byproduct of it to find food or other resources, they are said to be using public information. Using public information does not necessarily require or result in social learning of any kind, but it might.

For instance, the young rat feeding in the presence of other rats or their excrement becomes familiar with the flavor of whatever it is eating there. Eavesdropping is
reserved for cases in which the public information consists of communicative signals, but from a mechanistic point of view there is not necessarily any distinction (Bonnie and Earley 2007; Valone 2007). And like other kinds of public information, signals may arouse specific behaviors or affective states in eavesdroppers without anything necessarily being learned.

For example, seeing conspecifics fighting raises testosterone levels in cichlid fish (Oliveira et al. 2001).
In Section 13.2 we analyze how learning from public information might take place. Here a series of studies with stickleback fish will illustrate some potentially cognitively interesting questions about how public and private information interact. In all of them, the fish acquired information about the value of feeding patches in a setup like that illustrated in Figure 13.4. An observer fish confined to a central compartment could see fish feeding in each end of a tank.

It could not see the worms that were being delivered on different schedules in the two patches, but it could see the demonstrators feeding and attempting to feed. Both nine-spined and three-spined sticklebacks used public information, in that shortly after demonstrations they chose a patch where fish had been feeding over one where no food had been delivered. However, when both patches had had food, only the nine-spined species chose the one that had delivered food at the higher rate (Coolen et al. 2003; see also Webster
and Hart 2006).

Public and private information were opposed in a further study with nine-spined sticklebacks by first letting observers learn for themselves that the richer of two patches was always at a given end of the tank. Between 1 and 7 days later they were exposed to conflicting public information (demonstrators feeding more frequently at the observer’s formerly poor patch than at the rich one) and immediately tested. Fish whose private information training had ended the day before behaved as if ignoring the public information, whereas those trained a week before strongly
preferred the patch that had just been seen to be better (Fig 13.4).

Fish tested at intermediate delays showed no preference (Experiment 2 in van Bergen, Coolen, and
Laland 2004). The authors concluded that the fish ‘‘will weight public and private information appropriately depending on circumstances.’’ This implies that fish tested at the longest delay still remembered what they learned individually a week before but were reweighting this information. However, because the experiment did not include control fish not exposed to conflicting public information just before testing, the
results could as well reflect forgetting of the original private information.

Of course the findings can still be described functionally as showing that recent information is treated as more reliable, but the absence of a forgetting control illustrates how a focus on a functional account can overlook interesting and even important mechanistic questions. (Which is not to say that a focus on mechanism cannot be similarly narrow.) Similarly, the contrast between public information and ‘‘social cues’’
in this context (Coolen et al. 2003) is perhaps not meaningful mechanistically if the former refers to the feeding rate of demonstrators and the latter to their numbers.

There seems to be little other than precedent to justify such distinctions among sources of social information (Bonnie and Earley 2007; Valone 2007) nor much reason to think they affect behavior through fundamentally different mechanisms. Indeed, once an
animal has learned the value of a site, it may not retain any information about whether it learned from seeing conspecifics there or being there itself.

13.1.4 Copying others’ choice of mate

If females are actively choosing mates, some males will be popular simply because they have more of whatever females are basing their choices on: more intense colors, more complex songs, or whatever. But if assessing a potential mate’s characteristics takes time, entails a risk of predation, or is otherwise costly, females could reduce their assessment costs by choosing males they see other females choosing (Dugatkin 1996; White 2004). Of course functional copying would result from females using cues that a male has been chosen before, such as the presence of
eggs in species where males guard a nest.

But remembering the identity of males chosen by other females and later preferring those males would be an example of social learning comparable to that involved when client fish learn about good cleaners or fighting fish and songbirds learn about winners and losers by eavesdropping on their fights (Chapter 12). Indeed, the first examples of mate choice copying involved fish, guppies (review in Dugatkin 1996), but although other fish show mate copying, this example has proven difficult to replicate, perhaps because in guppies mate copying is confined to certain populations (Galef and Laland 2005).

In birds, female black grouse visiting a lek (a communal mating ground) apparently prefer males seen copulating. Stuffed females were placed in males’
territories, either on the ground where males could mount and copulate, or on sticks as if sitting in a bush, where males could not copulate. Subject females
spent more time in the former than in the latter territories (Hoglund et al. 1995). Both females and males of another bird species, Japanese quail, have been the subjects of perhaps the most extensive investigation of mate copying in any species (White 2004).

In the basic demonstration of this phenomenon, a female is first confined equidistant from two males, one of whom is courting a female while the other is alone (Figure 13.5). Later—usually immediately afterward—the subject female is released and the time she spends in defined areas near each male is recorded. ‘‘Mate choice’’ here consists of spending more time
near one male than the other, but this measure does predict partner choice when the birds are free to interact.

Clearly there are a number of potential confounds in this simple test. For example, the female might be choosing a male that had been seen courting or a male that had courted recently or the place where such events had occurred. It turns out that what matters is not seeing mating per se but seeing a female near the male (see White 2004). Given that male quail are quite aggressive, a close approach by a female is enough to
indicate that a male is willing to mate. Male quail also learn which members of the opposite sex have been chosen by others, but experiments analogous to those with female quail subjects show they prefer a female that has not been courted by another male. This sex difference in behavior resulting from essentially similar learning (i.e., in performance rules) means that males do not invest in courting females that are already inseminated.

Question 3

13.2 Mechanisms: Social learning without imitation

Answer 3

13.2.1 Another example and some distinctions

Not so long ago, milk was delivered to the doorsteps of homes in Great Britain and elsewhere in glass bottles sealed with foil or paper. The milk was not homogenized, so it had a thick layer of cream at the top. In the 1920s and 1930s blue tits began
puncturing the bottle tops and stealing the cream (Figure 13.6). Milk bottle opening became relatively common in a few isolated areas, suggesting that it was being transmitted socially within them (Fisher and Hinde 1949; Hinde and Fisher 1951; Lefebvre 1995a).

Pecking or tearing open a bottle top is clearly not imitation, since pecking and tearing at bark and seeds are prominent components of tits’ foraging behavior, but the birds could have learned from one another where to direct these behaviors. This learning need not have been inherently social, however. Rather, the
products of one individual’s behavior—opened bottles—could have provided the conditions under which another individual learned for itself.

The naive tit drinking from an already-opened bottle would associate bottles with food and then approach
closed bottles and engage in food-related behaviors like pecking and tearing, which would be reinforced (Hinde and Fisher 1951). Sherry and Galef (1984, 1990) showed that indeed milk-bottle opening can develop through this process. They taught captive black-capped chickadees, a North American tit species, to open small cream tubs like those served in restaurants. Then some experimentally naive chickadees watched demonstrators opening cream tubs while chickadees in another group simply learned to feed from opened tubs.

Birds in both groups were subsequently more likely to open sealed tubs on their own than chickadees that had observed an empty cage containing a closed cream tub, but the proportion of opening individuals in the two groups did not differ. The products of a conspecific’s behavior may facilitate learning by naive individuals in a number of ways.

Adult black rats of the Israeli pine forests, described at the beginning of the chapter, do not directly teach or demonstrate efficient pine cone stripping to their young. Rather, cones partially stripped by experienced rats have their scales exposed in such a way that a young rat gnawing at the cone can easily remove them in an efficient, spiral, pattern, and get at the seeds underneath. Naive rats encountering completely unopened cones gnaw them all over in an inefficient way (Aisner and Terkel 1992; Zohar and Terkel 1996).

Thus efficient stripping of scales from pine cones, which the rats must develop in order to access their only food in the forest, is socially transmitted when the young rats follow adults around, steal partially
opened cones, and continue the stripping themselves (Terkel 1995). This is ‘‘social learning’’ only because the adults create the conditions necessary for it to occur;
successful actions emerge through trial and error learning by individuals.

In other cases, the products of one individual’s behavior attract others to the same sites, allowing those individuals to learn something there. For instance, rats’ preference for sites surrounded with fresh rat excrement leads them to become familiar with food eaten by other rats (Laland and Plotkin 1991).
This kind of social influence is referred to as local enhancement or stimulus enhancement (Box 13.1; Galef 1988; Whiten and Ham 1992; Heyes 1994a). The demonstrator’s behavior or some product of it attracts
the observer to a location or stimulus which it then learns about on its own.

13.2.2 Observational conditioning

In observational conditioning the demonstrator’s actions or the affective state and behavior they arouse in the observer are associated with stimuli present at the time. As a result, the observer performs similar species-typical behavior when it encounters those stimuli again by itself. One striking example is provided by the mobbing that small birds direct toward predators. In mobbing, as the name suggests, birds
approach a predator in a group, calling in a distinctive way. This behavior functions to alert potential victims in the area to the location of the predator and may also
drive the predator away. Some common predators like owls may be mobbed even by naive birds, but mobbing can depend on social learning (Curio 1988).

Social transmission of enemy recognition has been studied in European blackbirds in the apparatus depicted in Figure 13.7. A ‘‘teacher’’ sees a stuffed owl in the central compartment. The ‘‘pupil’’ sees and hears the teacher mobbing the owl and is stimulated to engage in mobbing behavior itself. However, in its side of the central compartment the pupil sees not the owl but a harmless bird like a honeyeater or an
owl-sized plastic bottle. When the pupil later encounters the training object by itself, it will mob it. The pupil can now ‘‘teach’’ naive blackbirds to mob bottles or honeyeaters. Such mobbing can be socially transmitted across chains of up to six birds (Curio, Ernst, and Vieth 1978).

This is a straightforward case of Pavlovian conditioning. Because the mobbing demonstrator elicits mobbing by the pupil, the pupil acquires an association between
the bottle or honeyeater and its own mobbing behavior system (Figure 13.8). Nonassociative controls are necessary to be sure that mobbing is indeed associated specifically with the training object. For example, birds that have mobbed the honeyeater should not mob bottles as strongly, and vice versa (for review see Curio 1988; A. Griffin 2004).

Experiments on acquired mobbing have generally begun with a phase in which the subjects are habituated to the bottle or the honeyeater, so later
mobbing is clearly the result of having seen the teacher mob. Robust learning to such objects after habituation to them suggests that latent inhibition is not very strong in this system; naive individuals can thus learn about predators even after encountering them while alone. Another possibly specialized feature of this system is that a more predator-like object, a stuffed honeyeater, supports stronger acquired mobbing than a bottle (Curio 1988).

Meerkats, monkeys, and some other social mammals also mob predators, but how mobbing develops in these species has not been studied to the same extent as in birds (A. Griffin 2004; Graw and Manser 2007). Social transmission of predator recognition makes functional sense because individuals that must experience predators for themselves to learn they are dangerous may not survive those experiences. The best-analyzed example involves monkeys’ fear of snakes (Mineka and Cook 1988).

Monkeys reared in captivity do not exhibit fear the first time they encounter live or toy snakes. If they watch another monkey behaving fearfully toward a snake, they later do the same themselves. As with mobbing, during the learning trial the naive observer exhibits behavior like the model’s (in this case responses such as withdrawal, vocalization, and piloerection). If naive monkeys observe a model behaving fearfully toward a snake and neutrally toward another object like a flower, they acquire the same discrimination.

For example, if they are later offered raisins that are out of reach beyond a flower or a snake, they reach quickly over the flower but refuse to reach over the snake.
Selective acquisition of fear shows that the animals are not simply sensitized to behave fearfully to any and all relatively novel objects in the experimental situation.
However, even though naive monkeys do not show fear to snakes or flowers, they acquire fear much more quickly to snakes than to flowers (Cook and Mineka 1990). Subject monkeys that saw videotapes of demonstrators apparently reacting fearfully to snakes and nonfearfully to flowers acquired fear of snakes, just as if they had seen live demonstrators.

However, subjects exposed to tapes edited to depict a monkey fearing flowers but not snakes did not learn to fear either stimulus. This comparison shows simultaneously that snake fear is acquired associatively (it depends on the specific pairing of demonstrator’s behavior with a snake) and that the associative
process is selective (not any initially neutral object will be feared). Selective learning about snakes seems to be specific to fear. Monkeys trained with video images of either snakes or flowers paired with food learned equally quickly in both conditions (Cook and Mineka 1990, Experiment 3).

However, the stimuli used and the discriminations
to be learned were not exactly the same in this experiment as in those involving socially transmitted fear, so this conclusion must be somewhat tentative (Heyes 1994a). Social learning about aversive events seems to be phylogenetically fairly general (see A. Griffin 2004), as functional considerations suggest it should be. At the same time, the events learned about are species-specific. Several species of birds learn to avoid aversive foods by watching others (Mason 1988; but see Avery 1994).

Curio’s paradigm (Figure 13.7) has been used to train New Zealand robins to recognize stoats, an introduced predator (Maloney and McLean 1995). Similar training has been used to prepare captive-raised young of endangered species for release in the wild (A. Griffin 2004). Suboski (1990) termed the form of learning
here releaser-induced recognition learning because in ethological terminology a sign stimulus present at T1 elicits behavior via an innate releasing mechanism. At
T2 the animal reveals its recognition of a neutral stimulus that accompanied the release.

However because the interevent relationships necessary for learning seem to be the same as in simultaneous Pavlovian conditioning (Figure 13.8), it is not clear that any term other than observational conditioning (Heyes 1994a) is needed. Observational conditioning is not confined to aversive USs. Young chicks peck at items they see another chicken or even a motor-driven model beak pecking at, behavior that would normally direct them to food being eaten by a mother hen. If a young chick watches a beak-like object selectively ‘‘pecking’’ dots of one color on the other side of a barrier, it pecks at that same color on its side and retains this discrimination when later tested alone (Suboski and Bartashunas 1984).

Similarly, when young junglefowl watch others pecking for food in a distinctively decorated bowl they later peck more in bowls decorated in the same way (McQuoid and Galef 1992). This socially acquired preference was weak and transitory if the bowls were
empty in the test, but it was robust and long-lasting if the birds got food in the test. This might be typical of socially acquired preferences (Galef 1995). Because positive reinforcement can perpetuate the behavior once the animal makes the socially induced choice, social learning about positive stimuli need have only a small initial effect to have important consequences.

Most examples of social learning described earlier in this chapter could be described as observational conditioning. The social experiences that influence choice of mates, food patches, flavors, or opponents in a fight are in fact simultaneous pairings of particular individuals, places, or other cues with motivationally significant stimuli. However, an associative account has implications that need to be tested.

At the most basic level, what is the role of contingency between the putative CS and US? Contingency apparently plays a role in socially influenced patch choice in ninespined (but not three-spined) sticklebacks in that out of two patches where they had seen other fish feeding they preferred the one where food deliveries had been more frequent (Coolen et al. 2003). Cue competition effects might be expected, too, but
overshadowing and blocking failed to appear in a study of socially transmitted food preferences in rats (Galef and Durlach 1993).

In socially learned mate choice, female zebra finches learn about both a male’s identity and an artificial ornament, the color of the band on a male’s leg (Swaddle et al. 2005), but whether these cues compete for learning was not tested, for example by manipulating their relative validity. In candidate examples of observational conditioning other than mobbing and snake fear it is unclear how the demonstrators’ behavior acts as a US.

However, progress in identifying the effective US has been made not only with rats’ flavor preferences, but
also with mate choice in quail (Ko?ksal and Domjan 1998; White 2004), feeding techniques in pigeons (Palameta and Lefebvre 1985), and feeding patch choice in sticklebacks (Coolen et al. 2005). In summary, there is plenty of scope for more detailed analyses of what and how animals learn from observing conspecifics engaged in species-typical behavior. In the
past such research has been discouraged by confusion over terminology and a tendency to dismiss aspects of social transmission other than imitation as both uninteresting and well understood.

As a result there are few such phenomena for which the conditions for learning, the content of learning, and/or the effects of learning on behavior have been clearly delineated. Typical of terms in social learning, and notwithstanding attempts at clarification by Heyes (1994a) and others, observational conditioning continues to refer to a confusingly large number of phenomena (cf. Hoppitt et al. 2008). Some of them seem to involve a special learning mechanism, others do not.

Learning from watching another animal perform an arbitrary behavior B and receive outcome O could be described as S-S (stimulus-stimulus) learning. It
might even follow associative principles, but no performance rule for normal conditioning seems able to explain how knowledge that someone else gets O for performing B leads the observer to perform B when it desires O (Papineau and Heyes 2006).

13.2.3 Species differences

Social learning might be expected to vary across species with the conditions of social life. So far, however, we have encountered no evidence for any qualitatively special kind of representations or computations. Socially transmitted behavior such as black rats’ pine cone stripping and tits’ milk bottle opening is best described as socially influenced learning in that conspecifics provide the conditions under which the given behaviors are learned by normal associative means.

And mate choice copying, socially transmitted food and patch preferences, or enemy recognition all seem to be instances of observational conditioning, as broadly defined. But smelling food together with carbon disulphide or seeing a hen pecking red food results in learning only in species with the appropriate specializations of perception, attention, or motivation. Rats’ breath is presumably not interesting to chickens, nor is the sight of chickens pecking interesting to rats.

Specializations of learning per se could also play a role, as illustrated by the predisposition of monkeys to acquire fear to snakes but not flowers, stimuli which are apparently equally easy to associate with food (Mineka and Cook 1988). However, with the exception of demonstrations that learning about sexual partners is
expressed through different performance rules in male versus female quail (White 2004), there are virtually no thorough comparative studies of any of the sorts of social learning reviewed in this section.

But there are a few tantalizing suggestions. For example, the notion that animals which do not spend much time in family groups should not learn very well from adults was tested by exposing young brush
turkey chicks to a brush turkey robot pecking at corn in a red as opposed to a blue dish (Go?th and Evans 2005). In studies like those already mentioned in this section (see Go?th and Evans 2005), young chickens acquire the same discrimination as a model.

However, brush turkeys bury their eggs in a mound of rotting vegetation. The young hatch without adults around and have little opportunity for social learning about food. Indeed the young brush turkeys in Goth and Evans’s experiment did not prefer the color
pecked by the model when tested the next day even though they had approached that color more during the demonstration.

Of course data from a single set of conditions are
seldom enough to infer a species difference, let alone show what it consists of, but the effect is quite robust in the comparison species, junglefowl and their domestic descendents. The young brush turkeys seem to attend to and copy the choice of the robot so perhaps they forget more quickly than chickens do. One possible source of species differences in social learning is attentiveness to the activities of other animals, which could perhaps be acquired.

Such differences in attention could be responsible for the differences in social learning about the locations of food caches among corvids described in Chapter 8. Indeed, there is some indication of differences in social attention in two other corvids, ravens and jackdaws. This was demonstrated with the setup shown in Figure 13.9 (Scheid, Range, and Bugnyar 2007), one adopted from experiments with primates (e.g., Range and
Huber 2007).

Both ravens and jackdaws are quite social, but the nature of their sociality differs in a way that the authors argue favors ravens paying more attention to the activities of others, especially their feeding. As predicted, raven subjects spent more time observing a conspecific than did jackdaws. Again, this is only a single set of conditions, and it is important to know whether the results hold up when conditions such as size and ease of access to the viewing ports are varied. Still, this seems a promising method for measuring social attention.

Possible specializations for social learning can be tested very elegantly if the relevant task can be acquired under both social and nonsocial conditions, as in studies of feeding skills in social versus nonsocial birds by Lefebvre and his colleagues (Lefebvre and Giraldeau 1996). They tested the notion that opportunistic animals such as rats and pigeons, which have fairly generalized food requirements and can
take advantage of a wide range of niches, might be more prone to social influences than more conservative species.

But because opportunism by definition is the ability
to function effectively in a many different environments, opportunistic animals might have generally enhanced learning ability. On this latter hypothesis performance on social learning tasks should correlate positively with performance on nonsocial tasks. Yet another hypothesis is that social learning is most evident in species whose foraging is a matter of scramble competition, that is, many individuals feeding at
once on limited food sources, as opposed to interference competition, where foragers
aggressively exclude competitors.

Success in scramble competition is a matter of
speed, so slow individuals can benefit by learning the techniques being used by their speedier competitors (Lefebvre and Giraldeau 1996). If all these factors are important, then social, opportunistic animals that encounter scramble competitions for food will be the best social learners, whereas solitary species that compete with others by exclusion and have conservative food habits will be the poorest.

All these predictions were addressed by comparing how pigeons and a close relative, the Zenaida dove (Zenaida aurita) from Barbados, learn various foraging tasks socially and individually. Pigeons are social and opportunistic and encounter scramble competitions while foraging, so they should excel at social learning. Most Zenaida doves are territorial year-round but tolerate and even forage with birds of other species like grackles (Quiscalus lugubris). At first glance, Zenaida doves and pigeons differ in social learning just as the three ecological hypotheses predict.

Naive pigeons and doves were equally unlikely to push the lid off a bowl of grain (Figure 13.10, top row), but after watching a conspecific push off the lid and eat the grain underneath, more pigeons than doves pushed off the lid by themselves (Lefebvre, Palameta, and Hatch 1996). However, the pigeons were also quicker to feed from an open bowl of food in the experimental situation, and pigeons pushed off the lid sooner than doves after simply eating from the bowl with no demonstrator present (Figure 13.10, middle row).

These findings suggest that pigeons and Zenaida doves differ not in social learning ability but in some general learning ability or in responses to contextual variables (see Chapter 2). However, the story is still more complicated: Zenaida doves’ susceptibility to social influence depends on the species of tutor and on the social situation in which they have been living. In two different feeding tasks, territorial Zenaida doves copied grackles rather than other doves whereas subjects from a gregarious population of Zenaida doves learned more quickly from a dove than
from a grackle.

These population differences may reflect differences in experience. Gregarious doves could also be shaped more readily than territorial doves to perform a complex food-finding task, suggesting the populations differ in learning ability generally or in something else that influences speed of learning such as neophobia
(Dolman, Templeton, and Lefebvre 1996; Carlier and Lefebvre 1997). More extensive comparisons of pigeons and doves, as well as data on several tit species, show that performance on social learning tasks is positively correlated with performance on
comparable nonsocial tasks (Lefebvre and Giraldeau 1996).

It is also correlated with innovation, both across species and in comparisons of individuals within one species, pigeons (Bouchard, Goodyer, and Lefebvre 2007). Just as innovativeness reflects a concatenation of more general cognitive abilities (Box 2.2) so may ‘‘social learning,’’ at least when measured as successfully copying others’ behavior in naturalistic conditions. But this analysis begs the question we take up next: whether the narrow but important kind of social learning known as imitation (Box 13.1) is a specialized kind of learning shown by only a few species.

Question 4

13.3 Mechanisms: Imitation

Answer 4

In imitation, the form of a behavior is learned from a demonstrator. Interest in imitation has a long history (cf. Whiten and Ham 1992), but only toward the end of the twentieth century was much progress made in understanding it, as researchers began looking at the development and mechanisms of imitation in humans. As a result candidates for imitation in other species were no longer simply compared to some assumed ideal of human imitation but children and apes were compared directly, often in the same experiments with tasks resembling those chimpanzees are thought to learn socially in the wild. In addition, the discovery of mirror neurons in monkey brains in the late 1990s provided a possible neural mechanism for imitative behavior. All these new findings in turn stimulated new theories about how imitation is possible and its role in the evolution of human culture.

13.3.1 Some history

Imitation is one of the mental faculties Darwin (1871) claimed other species share with humans. Anecdotes about domestic animals apparently imitating complex
actions performed by people featured prominently in the evidence for mental continuity collected by Romanes and others. Many of these anecdotes involved cats and dogs learning to open doors and gates by manipulating latches, handles, and door
knobs. One of the more colorful of these featured a cat belonging to Romanes’s coachman.

Walking up to the door with a most matter-of-course kind of air, she used to spring at the half-hoop handle just below the thumb-latch. Holding on to the bottom of this half-hoop with one fore-paw, she then raised the other to the thumb-piece, and while depressing the
latter, finally with her hind legs scratched and pushed the doorposts so as to open the door… . Of course in all such cases the cats must have previously observed that the doors are opened by persons placing their hands upon the handles, and, having observed this,
the animals forthwith act by what may be strictly termed rational imitation… .

First the animal must have observed that the door is opened by the hand grasping the handle and moving the latch. Next she must reason, by ‘the logic of feelings’—If a hand can do it, why not a paw? (Romanes 1892, 421–422). ‘‘If a hand, why not a paw’’ captures very well what is cognitively distinctive about
imitation. True imitation entails using a representation of a demonstrator’s action to generate an otherwise unlikely action that matches the demonstrator’s. The cat at the gate is at most matching visually perceptible actions of its own to equally perceptible actions of another, that is, grasping the latch, and so forth.

An apparently greater cognitive challenge is reproducing a model’s perceptually opaque actions, that is, those like facial expressions or whole-body movements that one cannot see or hear oneself perform (Heyes and Ray 2000). In either case, performance of a species-specific activity under the direct influence of another animal does not qualify.
Thorndike (1911/1970) made this point with an anecdote about a flock of sheep being driven along a path, each jumping where the one in front of it had jumped, even
when the barrier that originally occasioned jumping had been removed.

‘‘The sheep jumps when he sees other sheep jump, not because of a general ability to do what he sees done, but because he is furnished with the instinct to jump at such a sight, or because his experience of following the flock over boulders has got him into the habit of jumping at the spot where he sees one ahead of him jump.’’ What Thorndike emphasized still bears repeating (Galef 1996a): by themselves, observations like those of cats opening latches or sheep all jumping in the same place cannot reveal how such behavior came about. Field data can be enormously suggestive, but experiments, or at least systematic observations of acquisition, are required to know whether behavior has developed through imitation or in some other way.

Thorndike’s (1911/1970) own experiments were based fairly directly on Romanes’s stories about dogs and cats opening latches. But instead of letting animals open gates, Thorndike confined them in ‘‘puzzle boxes’’ that could be opened in various ways to allow the animal to escape and find food. In his experiments on imitation, a cat or a young chick was allowed to learn by itself, by trial and error, how to escape. Then a second, observer animal watched. If observers learned faster than demonstrators, imitation must have occurred. Thorndike’s experiments with cats and chicks provided no evidence for imitation, but he did leave open the possibility that monkeys would imitate, a possibility that continues to be debated.

13.3.2 Birds and the two-action test

Because one animal’s behavior can come to resemble another’s in so many ways, imitation sometimes seems to be what’s left over when all other conceivable routes for social learning have been ruled out (Zentall 1996, 2006a). An experimental approach that goes a long way to ruling them out was pioneered by Thorndike (1911/1970). A puzzle box had two escape routes, and a chick watched another chick using one of them. If the observer chick imitated it should follow the same route as the demonstrator rather than the alternative, equally easy, one. In the more refined version developed by Dawson and Foss (1965; see also Galef, Manzig, and Field 1986), this design is known as the two-action test (Zentall 1996; Heyes 1996).

Two-action tests typically involve an object, sometimes a tool, that can be operated on with either of two responses such lifting versus pushing or twisting versus pulling. Ideally both responses move the object in the same way. Otherwise, observers may copy the model not because they imitated its behavior but because they emulated or learned the affordances of the object (see Box 13.1). Emulation tends to be invoked when observers copy demonstrators only crudely. For example, observer chimpanzees learned more quickly to use a tool to rake food into the cage than did controls that had not seen the tool being used, but they did not use the same technique as the demonstrator (Tomasello et al. 1987).

Emulation has come to have a confusing variety of meanings referring to different kinds of learning
thought to underlie the behavior (Box 13.1 and Whiten et al. 2004). Observers may have learned that there is a reward to be obtained or that the object is related to
obtaining the reward. An associative analysis (Heyes 2005; Papineau and Heyes 2006) would see the first of these as situation-outcome learning and the second as
object-outcome learning. Apparent emulators may have learned the object’s affordances, that is, that it can be moved in a certain way, although how this learning could translate into behavior causing that same motion is itself mysterious (Zentall 2004).

In any case, some birds as well as primates (Hopper et al. 2008) show affordance learning. For example, when pigeons saw a door move away from a food tray either
to the left or to the right, they later more often pushed it in the direction they saw than in the opposite direction (Klein and Zentall 2003). Finally, an observer with theory of mind might infer the demonstrator’s intentions and copy those, a process also sometimes referred to as goal emulation (see Whiten et al. 2004). However, given the paucity of more direct evidence for theory of mind in nonhuman animals (Chapter 12), there seems to be no good reason to invoke it here whatever the results of tests of imitation.

As Heyes (1993a, 1008) put it, ‘‘What is apparently essential for imitation is that the imitating animal represent what the demonstrator did, not what it
thought.’’ The same can be said of emulation. Some birds imitate in two-action tests (Zentall 2004). Many of these demonstrations involve a treadle that can be depressed by pecking it or stepping on it. Importantly, pecking and stepping are perceptually opaque responses that differ in topography but cause the lever to move in the same way. In one of the first studies
with quail, for example, each subject was trained to eat from the feeder in the demonstrator’s compartment before being placed in a neighboring compartment to
view a demonstrator either peck or step on the treadle and receive food reinforcers for 10 minutes (Akins and Zentall 1996).

When observers were returned to the response
half of the chamber immediately after this experience, every bird’s first response to the treadle matched the responses it had observed. In the first five minutes of the reinforced test, on average about 90% of the responses to the treadle were imitative responses (Figure 13.11). Of course (see Heyes 1996; Whiten et al. 2004) the birds’ behavior does not strictly qualify as imitation because the motor patterns being copied are not novel and unusual behaviors for the species. Nonetheless, considerable progress has been made in analyzing the learning of quail and pigeons in this situation (see Zentall 2004).

Importantly (see Box 13.1), imitative behavior does not depend on being tested immediately; in quail it is also evident in a test delayed 30 minutes, more consistent with learning than some sort of temporary facilitation (Dorrance and Zentall 2001). The robust copying of pecking and stepping sets the stage to discover what the animals actually learn from watching. Because the treadle moves in the same way whether it is pecked or stepped on, the birds must have acquired some representation of the observer’s action. Does it matter if the demonstrator is seen to be rewarded for its efforts? Studies with quail indicate that little imitation occurs if either demonstrators are not hungry or observers are not rewarded (Zentall 2004).

However, this does not necessarily mean that observers are learning response-food associations by
observation; being hungry and seeing the demonstrator getting food might only increase the observer’s attention to the demonstration. Indeed, there is evidence for blind imitation in this kind of situation (i.e., copying the observer regardless of the
outcome it is getting), at least with already-trained responses. Pigeons that have been shaped both to peck and to step on a treadle and then watch a demonstrator pecking or stepping subsequently increase their own tendency to perform the same action, whether or not the demonstrator was being rewarded (McGregor et al. 2006).

In a similar test in which the demonstrator pecks in the presence of one colored light and steps in the presence of another, pigeons acquire the observed stimulus-response associations (Saggerson, George, and Honey 2005). It is not yet clear whether these
findings means that pigeons (and perhaps other birds) always engage in blind imitation or whether imitation is goal-directed under some conditions (McGregor et al. 2006). As mentioned earlier (Section 13.1.2), social learning is more likely to be adaptive if animals do not always do what others are doing. But the fact that blind imitation occurs when all other factors that might be relevant are tightly controlled does not mean it would lead to maladaptive consequences in nature.

For example, in the study of McGregor et al., observers were not rewarded in the test; they might not have
copied the demonstrator for long if reward had been available for some alternative behavior. In any case, this series of studies is an important beginning to understanding the conditions for learning by imitation. Further insights come from recent studies of
primates.

13.3.3 Chimpanzees and children

Until the last decade or so of the twentieth century, most evidence regarding imitation in monkeys and apes consisted of anecdotes from the field or from captive animals reared in close association with humans (Whiten et al. 2004). Because human children seem to be good imitators, our closest living relatives were assumed to be good imitators too. Indeed, in many languages the same word (e.g., ape) refers both to a nonhuman primate and to the act of imitating (Visalberghi and Fragaszy 1990a). The assumption that apes can ape led to skepticism about suggestions that they do not ape very readily or exactly and to a lack of experimental tests of imitation in primates.

The situation has changed dramatically in the last 15 to 20 years. A recent review lists over 30 studies of apes (Whiten et al. 2004), and that does not include a more recent spate of direct comparisons between chimpanzees and children (e.g., Call, Carpenter,
and Tomasello 2005; Horner and Whiten 2005; Herrmann et al. 2007). These studies are important not only for how they illuminate mechanisms of imitation but also for what they imply about human cognitive uniqueness and the abilities that support human culture.

A breakthrough here was an experiment by Whiten et al. (1996). These researchers both gave chimpanzees a two-action test of imitation and tested young children under the same conditions (see also Nagell, Olguin, and Tomasello 1993). Moreover, their task—opening an ‘‘artificial fruit’’—resembled foraging behaviors chimpanzees might learn by imitation in the wild. The artificial fruit was a transparent plastic box containing a food treat which could be opened by manipulating various handles or bolts (Figure 13.12). In one version the lid was closed by two bolts that could be either poked or twisted out.

Captive chimpanzees or 2-, 3-, or 4-year-old children saw a human adult poke or twist the bolts and then were given a similar ‘‘fruit’’ that could be opened using either action. Subjects’ behavior was videotaped and scored independently by two observers ignorant of which action the subjects had witnessed. Subjects of both species were significantly more likely to use the action they had seen than the alternative (Figure 13.12). The tendency to imitate was least in the chimpanzees and greatest in the 4-year-old children. The children were more likely than the chimpanzees to copy slavishly even nonfunctional parts of the demonstrator’s acts, as if taking for granted that an adult’s way of doing things is worth copying.

The chimpanzees did direct their behavior at the correct part of the box even when they did not use the same actions they had seen, that is, emulating or showing they had learned the affordances of the apparatus (the bolts come out; the box opens).
One of the first questions these findings raise is whether the chimpanzees would copy more precisely with a chimpanzee rather than a human demonstrator. The answer to this question seems to be ‘‘no’’ (Whiten et al. 2004). Given that the chimpanzees did show some copying of the demonstrated actions, another question is what determines the extent to which they imitate specific actions as opposed to emulate or learn affordances?

One suggestion is that imitation plays a greater role in
more complex tasks. Conversely, nonsocial processes such as affordance learning appear more important in simple tasks. This latter conclusion is supported by a
comparison of two separate studies in which chimpanzees watched devices move by themselves (as if moved by a ghost, hence ghost conditions). Using a scaled-up version of Klein and Zentall’s (2003) apparatus for pigeons, Hopper and colleagues (2008)
had chimpanzees and children watch the door on a box move to the left or the right to reveal food inside (Figure 13.13).

The effects of this experience on subjects’ subsequent actions on the door were compared to the effects of watching either the door move by itself in the presence of a conspecific who then retrieved the food (‘‘enhanced ghost condition’’) or a conspecific pushing the door (full demonstration). Chimps and children in all conditions were very likely to push the door in the demonstrated direction on their first opportunity. However, all the children continued to prefer
the demonstrated direction, whereas the chimpanzees maintained this preference only if they had seen a chimpanzee doing the pushing.

Still, their initial responses are evidence that they learned the affordance of this simple apparatus in which the part to be moved was very close to the food. These results contrast with those obtained when a more complex task was used in a test of social transmission within chimpanzee groups (see Section 13.5 and Hopper et al. 2007). Here a stick had to be
used to lift a T-shaped bar on top of a box so that food would roll out at the bottom of the box. Only one of 18 chimpanzees exposed to a ghost condition operated the apparatus successfully in a subsequent 1-hour test. A larger proportion of successes followed demonstrations in which a chimpanzee lifted the T bar.

However, this was a difficult task in that there were relatively few successes compared to those in an
alternative version in which the food was released by poking the tool into a hole. Although increased task difficulty (and perhaps remoteness of the reward from the object to be moved) seems to reduce affordance learning or emulation, it seems to enhance learning by imitation. Perhaps the most striking evidence for this conclusion comes from another comparison of chimpanzees and children (see also Call, Carpenter, and Tomasello 2005; Horner and Whiten 2005) involving copying several actions in sequence, a capability for which there was already some evidence from chimpanzees (Whiten 1998) and gorillas (Stoinski et al. 2001).

Here both chimpanzees and 4-year-old children watched a human adult use a stick to perform one of two sequences of actions on the box shown in Figure 13.14. The only functional part of these sequences involved sliding or lifting the door in the front of the box and pulling out a packet of food with the stick. The demonstrator began, however, by tapping the
bolt on top of the box, then moving it aside to reveal a hole and thrusting the stick into the hole. These actions were done in a slightly different way for each of two sub-groups, making this as well as the sliding versus lifting of the door a two-action test.

In either case they were irrelevant to operation of the box because a barrier separated the top half of the box from the food. Their causal irrelevance was evident in a transparent version of the box but not in an opaque one. Subjects of both species frequently copied the sequence of actions they saw, but the most important result of this study is that whereas the children imitated the irrelevant action of inserting the tool into the
top of the box about 80% of the time whether the box was opaque or clear, the chimpanzees did so much more often when the box was opaque (Figure 13.14).

If exposed and tested with the clear box, they most often bypassed this part of the sequence and went straight to operations on the door over the food. The authors interpret this finding to mean that when the causal structure of the task was evident the chimpanzees emulated, that is, primarily relied on learning about the results of actions. It is not clear from this experiment alone, however, whether the animals’ ability to see the effects of the irrelevant actions affected learning or performance. Perhaps they learn about both the actions of the demonstrator and the goal that can be obtained but goal-related cues take precedence in control of behavior when they are very
salient.

Animals trained first with the opaque box could learn from observation and personal experience about the food-containing part of the apparatus that lay behind
the door; once they could actually see it through the transparent box, direct approach evidently took precedence over imitating earlier parts of the sequence (group A/B in Figure 3.14). Animals trained first with the transparent box continued to go directly to operating the door when given trials with the opaque box (group C/D), but of course by then they had a history of immediate reward for these actions. Interestingly,
however, whatever else they did all animals had a significant tendency to move the door in the way they had seen it moved by the demonstrator.

Horner and Whiten (2005) discuss their findings in the spirit of an analysis of human imitation proposed by Wohlschla?ger, Gattis, and Bekkering (2003). This starts from realizing that a demonstration of a complex action on an object has several distinct elements including not only the actions but the object(s), and the outcome of the actions (i.e., the affordances of the object and/or rewards for the demonstrator). Attention to actions may result in imitation, but an observer might instead attend to and learn about the object and/or the outcome.

In any case, when the observer confronts the task alone later, memory of one or more of these features will be activated and this in turn will elicit relevant motor programs (for example, copying the action, interacting with the object, trying to obtain the goal directly). Wohlschla?ger and colleagues (2003) propose that the goal of the action always takes highest priority in controlling the observer’s behavior. However, priorities vary with the direction of attention, as shown by Bird et al. (2007). People were asked to copy the actions of a model who grasped a pen and placed it into one of two nearby cups.

Different elements of this simple demonstration were made more or less distinctive and subjects’ copying errors were measured. For example, when the model’s hands had differently colored gloves and the cups did not differ in color, subjects made fewer errors in copying which hand to use and more in copying the cup than when the reverse was true. Bird and colleagues (2007, 1166) conclude that, ‘‘the mechanisms that mediate imitation are plastic with respect to the processing of ends and means.

Furthermore, the factors influencing which aspects of an action are imitated are task general.’’ Similarly, Horner and Whiten (2005) suggest that chimpanzees attend to different aspects of a demonstration in different circumstances, and imitation, emulation, or
something else predominates accordingly. Children, however, seem to have a consistent bias toward imitation (see also Want and Harris 2002). Whether this
represents a predisposition present from a very young age and how much it is enhanced by the experience of being constantly shown things by adults is a matter
of debate.

Moreover, under some conditions young children do not slavishly copy unusual actions of a demonstrator but do the same thing with a different action, as if
copying the demonstrator’s intention (Gergely and Csibra 2003) or engaging in goal emulation. In any case, an account of variations in chimpanzees’ tendency to imitate in terms of variations in attention or memory explains everything and nothing. Experiments with chimpanzees like those of Bird et al. (2007) in which factors known to influence attention are manipulated without otherwise changing the structure of the task being demonstrated will be required to test it.

13.3.4 Do monkeys ape?

Insofar as they have been tested, the other three great ape species (gorillas, orangutans, and bonobos) behave similarly to chimpanzees: they imitate to some extent but may copy in other ways too (Whiten et al. 2004). In contrast, there is very little evidence that monkeys of any species imitate in the narrow sense of copying specific actions they have witnessed (Fragaszy and Visalberghi 2004). An exception is the performance of marmosets in a two-action test (Voelkl and Huber 2000).

After seeing a demonstrator marmoset use either its mouth or its hand to pull the lid off a film canister and get food inside, observers were more likely to use the action they saw than the alternative. More typical is the finding that capuchin monkeys exposed to conspecifics that were proficient at using a stick to get reward from a tube in the task described in Chapter 11 showed no evidence of imitating them (see Fragaszy and Visalberghi 2004). But notwithstanding their evident failure to copy exactly the actions they have witnessed, many monkey species do show various kinds of social
influences on learning (see Fragaszy and Visalberghi 2004).

This propensity may lead to social transmission of tool use and other behaviors in some wild monkeys
(Section 13.5). In the laboratory, monkeys have provided the only evidence for two novel kinds of imitative learning. In one case, rhesus macaques looked longer at a person who was copying their actions on a novel object than at a second person who handled the object at the same time but did different things with it (Paukner et al. 2005). Thus even though they do not imitate the actions of others very well, monkeys apparently notice when another is imitating them.

In the second novel form of imitation, one of two experienced rhesus macaques watched from an adjoining operant chamber while the other one executed a simultaneous chain that was novel for the observer (Subiaul et al. 2004). In the simultaneous
chaining task (Section 10.3), the animal learns to touch a series of arbitrary images in a fixed order. By what the authors call cognitive imitation, the observer could learn not the required actions, which in any case varied from trial to trial with the positions of the images on the touchscreen, but the correct sequence of images. 

Both monkeys in this study showed cognitive imitation in that they completed the first trial of a new
sequence with fewer errors after watching the knowledgeable partner than under various control conditions, including exposure to a computer replay of the sequence of images and sounds generated by a knowledgeable partner. Two-year-old children
also show cognitive imitation in this task (Subiaul et al. 2007). Learning sequences of actions in this way may have contributed to the performance of subjects in some of the multistage tool-using tasks described earlier.

13.3.5 Other candidates for visual imitation

Although imitation must ultimately be studied with experiments on groups of subjects, it is pretty compelling to see even a single animal do something like put on lipstick or use a tool in a way it has seen humans do. In a sense these are ‘‘multiple action tests’’ because there is a multitude of things the animal might do at the time. The literature on imitation by primates is full of accounts of such behaviors, mostly by
chimpanzees and orangutans that have lived closely with people (see Whiten and Ham 1992; Whiten et al. 1996). One of the first of these was the chimpanzee Viki, raised like a child by the psychologists Keith and Cathy Hayes (Hayes and Hayes 1952).

The Hayeses demonstrated that Viki had a fairly general ability to imitate novel actions by training her to obey the spoken command, ‘‘Do this.’’ Custance,
Whiten, and Bard (1995) trained two laboratory-reared chimpanzees in a similar way to the Hayeses but documented the procedures and results more fully. The animals were reinforced for obeying ‘‘Do this’’ using a set of fifteen actions like raising the arms, stamping, and wiping one hand on the floor. After more than three months of intensive training, they reproduced these actions with 80% accuracy or better.

In a test with 48 other actions, observers who did not know what the model was doing could classify the chimps’ actions at better than chance levels, but the agreement was far from perfect, suggesting that the animals were still not very good generalized imitators. A host of accounts of orangutans reproducing complex human activities like using hammers and paintbrushes, constructing bridges out of logs, and making fires (!)
comes from observations on formerly captive orangutans being rehabilitated for release in the Indonesian jungle (Russon and Galdikas 1993, 1995).

Observations of complex imitations are not confined to primates, either. Alex the parrot learned to talk by watching two people, one of whom played the role of parrot and was rewarded by the other for pronouncing and using words correctly (Pepperberg 1999). This situation is thought to reproduce the social situation in which wild parrots acquire vocalizations. However, once Alex began to vocalize himself, he received attention, food, and/or access to the objects he was naming, and in any case vocal imitation is usually treated as a special case (see Box 13.2).

Explicit reward was scrupulously avoided with another parrot, Okichoro, trained by Moore (1992) to vocalize and imitate associated movements. The bird lived alone
in a large laboratory room and was visited several times a day by a keeper who performed various stereotyped behavior sequences such as waving while saying
‘‘ciao’’ or opening his mouth and saying ‘‘look at my tongue.’’ Gradually Okichoro, observed continuously on closed-circuit TV, began to imitate both the actions and the words of the keeper while he was alone (Figure 13.15).

Because each vocalization in effect labeled a specific movement, possible imitation could be isolated from the stream of nonimitative behavior. And unlike pecking or stepping in quail, behaviors such as waving a foot while saying ‘‘ciao’’ are normally highly unlikely. Eventually many cases of imitation were recorded, including some nonvocal mimicry of sounds. For instance, the parrot imitated someone rapping on the door by rapping its beak on a perch. Moore (1996) claims that this is a special category of imitative learning, one of several that have evolved independently.

Although they are entertaining, such examples are prone to the weaknesses that afflict most anecdotal evidence. First, they are often based on a very special single subject. We may not know the animal’s history. Was it reinforced for approximations to the purportedly imitated behavior or similar actions in the past? This lack of necessary background information very often characterizes isolated observations in the field, but
studying animals in captivity is not always the solution.

As with Alex or the orangutans, lengthy and complex experience often precedes the behaviors of interest. Even if every effort was made to control this experience, we rarely know precisely what it was. It may also be difficult to determine how selective the observers were in recording the subject’s behavior. For instance, in a ‘‘do this’’ test, as opposed to a two-action test, the alternatives to reproducing the model’s behavior may not be clearly specified nor is the time
interval within which the animal must imitate as opposed to doing something else (Zentall 1996).

Like the proverbial band of monkeys who would reproduce the works of Shakespeare if left long enough in a room full of typewriters, primates raised in
homelike environments have many opportunities to perform humanlike actions, and those that are most humanlike and striking are most likely to be the ones reported. For example, how often did the formerly captive orangutans do something inappropriate
like bite a paintbrush, hold it by the bristles, or hit a nail with it?

Finally, the observers— for the very reason they are living closely with the animals in the first place—may be biased like proud parents to anthropomorphize what they see their animals do. Another problem for long-term research with one or a few subjects is the possibility of ‘‘Clever Hans’’ effects (Chapter 10), that is, the possibility that the observer is unintentionally
influencing the subjects to produce the desired behavior. Unfortunately, being aware that such effects can occur is not necessarily enough to prevent them, and if the relevant contingencies are not detected by the investigators themselves, they may be difficult or
impossible for others to detect in published reports.

The reports of imitation summarized here do not necessarily suffer from all, or even any, of these problems. Moore rigorously avoided Clever Hans effects by collecting data only over closed-circuit TV when the parrot was alone and by stopping data collection on any imitation once it had occurred in the presence of the experimenter. The rehabilitant orangutans imitated some elaborate sequences of
behavior that were actively discouraged, like stealing boats and riding down the river (Russon and Galdikas 1993). Nevertheless, when assessing either anecdotes from the field or long-term work with a few subjects in captivity it is important to keep such potential problems in mind.

13.3.6 How is imitation possible?

Mirror neurons:

How is it possible for me to perform the same action I see someone else perform, especially when that action is perceptually opaque? For example, when quail see
other quail step on a treadle, how is it that they themselves later step rather than peck? This is the correspondence (Brass and Heyes 2005) or translation (Rizzolatti and Fogassi 2007) problem. A solution at the neural level is suggested by one of the most
remarkable discoveries of late twentieth-century neuroscience, the mirror neuron system.

This is a network of cells in the premotor cortex, inferior parietal lobule (IPL) and elsewhere in the brains of rhesus macaques that fire both when the monkey performs an action itself and when it sees the action performed by another. These actions include not only perceptually transparent actions such as grasping and tearing, but actions of the mouth such as biting and sucking (review in Rizzolatti and Fogassi 2007). Some mirror cells respond to auditory as well as visual correlates of actions, for instance firing both to the sound and the sight of paper being torn.

In effect cells in the IPL encode not only the surface features of actions but their intent. The same cells that fire most when the monkey or a person reaches toward an object to grasp it also fire when a person reaches toward an object that the monkey has seen placed behind an occluder, making the grasping action invisible, but they do not fire in the absence of an object to be grasped. And in the example in Figure 13.16, grasping an object to eat it is distinguished from grasping to place it in a bowl, whether the monkey itself or a person does the grasping (Fogassi et al. 2005).

The mirror system evidently includes sensory-motor links between the visual and other cues accompanying performance of an action and its motor representation. Brain imaging shows that humans have a mirror system too (Rizzolatti and Fogassi 2007), and experience influences the strength of its sensory-motor links. For example, watching classical ballet is accompanied by greater activation of the mirror system in ballet dancers
than in capoeira dancers, and the reverse (Rizzolatti and Fogassi 2007). Even experience over a relatively short term can have an effect (Catmur, Walsh, and Heyes 2007).

Here then is a remarkably rich neural representation of actions as such, encoding own and others’ actions in a unitary way. Mirror neurons seem to be just what is
needed to generate imitative behavior, but something must be wrong with this idea because, as we have seen, monkeys are not very good imitators. Instead mirror neurons may play some more general role in social cognition by encoding the actions and intentions of others as, in effect, the same as one’s own (Rizzolatti and Fogassi 2007; de Waal 2008; but see Jacob and Jeannerod 2005).

Still, the mirror system does seem to play a role in imitation in humans, for example being more activated during imitative than control tasks (Brass and Heyes 2005; Rizzolatti and Fogassi 2007). One difference between monkey and human mirror systems that may underlie species differences in imitation is that the human system seems to encode specific actions
more precisely (Rizzolatti and Fogassi 2007). A second difference may be in the degree to which motor output of the mirror system can be engaged selectively.

If viewing another’s action generates the same premotor activation as one’s own intent to perform that action, then some further mechanism must prevent continual automatic and perhaps even dangerous mimicry. On one view (see Rizzolatti and Fogassi 2007) the primary function of the mirror system in all primates is to permit action understanding, not imitation. The flexible inhibitory mechanisms of the human prefrontal cortex permit its selective use to generate imitative actions, whereas in species that lack such mechanisms imitation needs to be inhibited in general.

Associative sequence learning:

Consistent with evidence for an influence of experience on representation in the mirror system is a model of imitation developed by Cecilia Heyes (Heyes and Ray 2000, Heyes, 2005). In the associative sequence learning (ASL) model, imitation is the outcome of
general associative mechanisms rather than a specialized ability, and it depends on experience during development (Brass and Heyes 2005). The elements of the model are so-called vertical associations, associations between sensory and motor activity correlated with one’s own perceptible actions, for example the sight of one’s own hand grasping and the motor commands to grasp.

When an individual observes a sequence of
actions performed by a demonstrator, the sequence is encoded as a set of horizontal associations, that is, associations within the sensory side. Now an action that is represented as such a chain of sensory-sensory associations will excite the associated motor
representations, and, hey presto, the observer reproduces the sequence of actions it saw. The ASL model can also explain copying of perceptually opaque actions such as pecking or stepping in quail and pigeons.

The ASL model assumes that these flock-living birds will have been in situations where all the individuals present are engaged in the same behavior, for example pecking at grain. Such experience allows a bird to
associate its own pecking with the sight of others pecking. When it later sees a demonstrator pecking in a certain experimental context it forms a context-other’s
pecking association. The vertical association between other’s and own pecking in turn activates its pecking behavior. This explains why quail and pigeons are good at copying species-specific behaviors.

It also may explain the population differences in
sensitivity to different kinds of demonstrators documented by Lefebvre and colleagues (Section 13.2.3; Heyes and Ray 2000). It also suggests that experimental manipulations of social experience should influence what and from whom such birds copy, a suggestion that does not seem to have been tested. However, although it does a good job with copying of familiar actions, the ASL model does not seem to account for the essence of true imitation, namely the copying of novel actions as was done by Okichoro. Decomposing such actions into simpler actions which have been performed with conspecifics would seem to make this ‘‘simple’’ account of imitation quite a bit more complex, perhaps unacceptably so.

Question 5

13.4 Do nonhuman animals teach?

Answer 5

Animals clearly learn from one another’s activities or the products of those activities, even if not by imitation. So do any nonhuman species engage in behavior that could be called teaching? In humans, teaching seems to involve theory of mind and intentions to modify the pupil’s behavior, but just as with deception, planning, and similar terms, when it comes to other species we need a clear operational definition that captures the essentials of the relevant behavior without mentalistic implications. In recent years, the accepted functional definition of teaching has been that proposed by
Caro and Hauser (1992).

To qualify as teaching, an animal has to meet three requirements. (1) It must modify its behavior specifically in the presence of naive individuals in such a way as to facilitate their learning. (2) The teacher should incur
some immediate cost to itself, or at least no immediate benefit. But of course for teaching to evolve the teacher needs to reap some benefit in the longer term, such as
reduced time feeding young or increased inclusive fitness due to having knowledgeable offspring. (3) As a result of the teacher’s behavior the pupil should learn something earlier in life or more rapidly than it would otherwise or that it would not learn at all.

Discussion of teaching thus shifts the focus from processes in naive individuals in a social group to those in experienced ones. Do the latter respond to correlates of ignorance in others by behaving so as to correct it? How are those responses, if any, tailored to the social learning mechanisms in potential pupils? And even if not theory of mind or intentionality, are any distinctive cognitive processes involved in it? None of the examples of social transmission of information yet reviewed in this chapter meets Caro and Hauser’s first requirement.

A bird mobbing an owl is not teaching naive individuals what to mob because as far as is known it would be mobbing whether or not they were present. Similarly, rats transmit flavor preferences by serving as passive vehicles for stimuli that other colony members encounter during routine mouth-to-mouth contact. But perhaps teaching is more likely to evolve when the behaviors to be acquired are more demanding and complex than these. Caro and Hauser (1992) described a number of candidates involving capturing prey that are difficult to subdue or handle.

For example, domestic cats bring dead birds and mice
back to the nest and present them to their kittens. As the kittens mature, mother cats carry back live prey and allow the kittens to play with it, but if the prey escapes the mother still catches it again. Finally the kittens capture prey by themselves with little intervention from the mother. Cheetahs behave similarly. Osprey, which snatch fish from the water in their talons, have been seen apparently teaching their fledglings to forage.
However, in none of these cases was it demonstrated what or how much the young actually learn as a result of the adults’ behavior. This gap has been filled by a
study of meerkats (Suricata suricatta; Thornton and McAuliffe 2006), a unique model demonstration of animal teaching.

13.4.1 Meerkats

Meerkats (or suricates, Suricata suricatta; Figure 13.17) are small cooperatively breeding mammals found in the dry parts of Southern Africa. They hold group territories in which the young are mostly produced by a dominant male and female but reared by all members of the group. When meerkat pups are about a month old,
they begin to follow foraging groups around, making begging calls which stimulate older animals to bring them prey. These prey include scorpions, which are difficult or even dangerous to handle.

Helpers often kill or disable such prey before presenting them to the youngest pups. Scorpions are killed or disabled for the pups to a greater extent than are other prey, but over the next two months all kinds of items are increasingly presented intact, as if the helpers are sensitive to the pups’ growing competence(Figure 13.17). The pups’ age is reflected in their begging calls, and a playback experiment showed that this proxy for pup competence determines the proportions of prey offered in different states. In groups with young pups, calls of older pups elicited more provisioning of intact prey, whereas in groups with older pups, begging calls of young pups increased the number of dead prey provided.

In addition to spending foraging effort on obtaining prey for pups, helpers stay nearby for a few seconds after delivering a food item. They stay longer with younger pups, and if a pup of any age does not take an item immediately, they may nudge it, as if drawing the pup’s
attention. If the prey escapes, the helper recovers it and presents it again. Thornton and McAuliffe’s (2006) extensive observational data together with the playback experiment demonstrate that the helpers’ behavior fulfills Caro and Hauser’s first two criteria for teaching: it is conditional on the presence (and here, age) of ignorant others and costly in time and effort in the short term.

The results of a further experiment show that it also meets the requirement of aiding pup learning. Thornton and McAuliffe (2006) compared three groups of pups matched for age and litter in their treatment of a live but stingless scorpion after three days of supplementary
experience with either four live scorpions presented daily by the researchers (a much higher number than normal), four dead scorpions, or equivalent amounts of hardboiled egg. Pups in the first group were markedly more successful in handling the test scorpion, consistent with the experience provided by provisioning ‘‘teachers’’ aiding their learning to subdue and process scorpions.

The behavior of the experienced meerkats is therefore comparable to that of mother black rats in that it allows the young to acquire skill in processing a challenging prey item, but the meerkats respond to stimuli indicative of the pups’ age (the begging calls) and experience (e.g., whether the scorpion escapes, is attended to, etc.) whereas the role of the mother rats in their pups’ learning is mainly to tolerate them nearby and to drop partially eaten pinecones. In neither case however, do we need to invoke adults’ understanding of the pups’
mental state.

13.4.2 Pied babblers

Another recently described example illustrates how ‘‘teaching’’ may result from a specialization in particular parts of a more species-general kind of behavioral
sequence. Pied babblers (Turdoides bicolor) are communally breeding birds found at the same study site in South Africa as the meerkats. As in many altricial species, adults feed the young birds in the nest for 2 to 3 weeks, and the family group forages together once the nestlings fledge. Also as in other altricial birds (e.g., Tinbergen and Kuenen 1939/1957), stimuli associated with an adult’s arrival at the nest elicit begging by the otherwise quiescent nestlings.

Raihani and Ridley (2008) observed that when pied babbler nestlings are 10 to 11 days old, arriving adults begin to emit a ‘‘purr’’ call. When they are about 13 days old, nestlings begin begging in response to purr calls. To show that nestlings’ response reflects learning to associate purr calls with food rather than maturation, beginning when the nestlings were 9 days old
Rahini and Ridley played purr calls at six nests whenever an adult arrived with food.

All them begged in respond to recorded purr calls by the age of 11 days, whereas nestlings in unmanipulated broods did not respond to the same test until Day 13, and begging was seen in only one control that had heard purr calls in the absence of food delivery.
So far, pied babbler purr calling fits the first and third criteria for teaching: it occurs specifically in the presence of ‘‘pupils,’’ and they learn something as a result, presumably a Pavlovian association between purr calls and food.

It also meets the criterion of being costly to the ‘‘teacher.’’ Purr calling is accompanied by fluttering of
the wings, and the more that adults display purring and fluttering within a given time, the less weight they gain. But why should the nestlings learn to respond to purring? The adults feed them anyway, or at least they do so without purring for the first 11 days. The likely function of learning that purring signals food becomes apparent after the young leave the nest around the age of 20 days and accompany loose groups of foraging adults around the territory.

Adults in such groups purr call more often than in groups that do not have fledglings (Radford and Ridley 2006). They purr when they have found food, in effect calling the young (as well as other adults) to approach, a response that in fact increases the nestlings’ foraging success. Sighting a predator also elicits purr calling when fledglings are present, in effect calling them
away from danger. Unlike with the meerkats, where the availability of dangerous and hard to handle but large prey items might create an exceptional pressure for evolution of costly teaching, the situation experienced by the babblers seems much the same as that for other birds in which newly fledged young accompany adults while foraging.

What seems special in the babblers, or at least not yet proven for other species, is the context-specific purr call. But food calling in domestic fowl and the ancestral
Burmese red junglefowl has many similar properties (see also Section 14.2). In food calling, both hens and roosters pick up a morsel of food in the beak, lower the breast and spread the tail while uttering a distinctive call. Hens food call in the presence of young chicks (Sherry 1977). Food calls attract the chicks, and because chicks tend to peck where they see another bird pecking, the hen’s food calling functions to cause the
chicks to peck at the food, in effect teaching them what to peck at (see also Nicol and Pope 1996).

Moreover, although they have a preexisting tendency to move faster toward a call given to high than to low quality food, chicks can learn the reverse discrimination (Moffatt and Hogan 1992). Thus although by Caro and Hauser’s criteria the adult babblers are teaching the young that purring means food, much more could be done to understand whether or why this situation differs from that for many other species in which mobile young accompany foraging adults and use cues to food that they provide.

13.4.3 Teaching in ants?

Ants of the species Albipennis bithorax sometime engage in tandem running when going from the nest to food: one ant travels behind the other, the follower frequently touching the leader on her legs and abdomen. When leaders were established in a
laboratory colony by letting them find food, and naive individuals were then allowed to follow them, leaders were observed to pause when a follower lost contact, as if waiting for the follower to catch up (Franks and Richardson 2006). Moreover, when a follower was removed partway through the trip, leaders waited longer before proceeding the more valuable the food source and the longer the trip had already been in progress (Richardson et al. 2007).

These observations have been interpreted as showing not only that ants teach but as suggesting an additional criterion for teaching, namely that the teacher should be
sensitive to feedback from the pupil (Franks and Richardson 2006; Richardson et al. 2007). Be that as it may, leaders clearly meet some of the criteria for teaching in that they behave differently with than without a follower and pay a time cost by doing so.
However, it has not yet been directly shown that anything is learned by followers in a tandem run, although some indirect evidence is available (Franks and Richardson 2006; Richardson et al. 2007).

It remains to be demonstrated that once a follower has
returned to the nest after a tandem run it finds the food again more quickly than a naive individual searching at random. This second trip of ants that have been
‘‘taught’’ the food’s location should also be compared to that of ants that originally found it on their own to see whether the benefit, if any, from following is confined to
the first trip to the food.

13.4.4 But what about primates?

The folk-psychological assumption that teaching requires cognitive complexity implies chimpanzees and other great apes should teach, but although some apes
and monkeys have population-specific behaviors that may be socially transmitted (Section 13.5), there is essentially no evidence than any such behaviors are taught by experienced to inexperienced individuals. For example, in one population in West Africa chimpanzees crack coula nuts with stone hammers and anvils (Figure 13.18). In over 10 years of field work, Boesch (1991) observed hundreds of cases in which chimpanzee mothers ‘‘stimulated’’ or ‘‘facilitated’’ their infants’ nut cracking but only two cases that might have been teaching.

Stimulation consisted of leaving stone hammers near anvils rather than carrying them off. Facilitation meant providing both hammers and nuts to infants at anvils. Both of these behaviors changed with the ages of the infants. In the two cases of apparent teaching, the mother intervened with an infant attempting to crack a nut and positioned the tool or the nut correctly. No indications of teaching or of imitative learning were found in a detailed analysis of the development of nut cracking in another area of West Africa (Inoue-Nakamura and Matsuzawa 1997).

At most, by exposing their infants to nuts and stones, nut-cracking mothers promoted interactions with stones and nuts by providing the conditions for stimulus
enhancement. There is also little or no evidence that chimpanzees teach their offspring how to ‘‘fish’’ for termites with sticks. Indeed, although infants spend a lot of time watching their mothers extract termites and even get some of the insects to eat, as with nut cracking they seem to need a good deal of individual practice to become efficient fishers themselves (Lonsdorf 2005).

And among the nutcracking capuchins described in Chapter 11, the young themselves make a major contribution to supplying interactions that might serve in social transmission of nut-cracking skills. They
prefer to watch the most proficient adult nutcrackers, perhaps because that gives them the most opportunities to scrounge bits of nut (Ottoni and de Resende 2005).

13.4.5 Conclusions

As with deception or planning, demonstrations that candidates for animal teaching meet a clear functional definition are controversial because they seem to lack key components of analogous human competences (Leadbeater 2006; Csibra 2007). Babblers, meerkats, or ants apparently teach others at most one thing. This may not be inconsistent with the functional definition of teaching, but even if further research reveals a species that teaches in several contexts, in human teaching understanding the learner’s state of knowledge or ignorance (i.e., using theory of mind) confers an ability to teach everything from tying shoes to doing physics (Premack 2007).

In any case, the scattered phylogeny of species with behaviors that function to teach makes it unlikely that such behaviors are homologous with human teaching, that is, evolutionary precursors to it (Galef 2009). This distribution instead raises important questions about what kinds of life history and ecology favor selection for costly behaviors that provide learning opportunities for the young or inexperienced. The analysis in this section suggests these will be on a continuum with other responses to such individuals, for example specializations in responses to the stimuli that elicit
provisioning. There is no evidence so far of any cognitive abilities specific to teaching.

And from the learner’s point of view, behaviors of ‘‘teachers’’ provide opportunities for learning by trial and error (as in meerkats), observational conditioning (as in pied babblers), acquiring spatial information (as suggested for ants) or by some other general mechanism. In summary there is still no reason to question the conclusion stated many years ago by the ethologist R. F. Ewer (1969, 698), ‘‘it is preferable to
think in terms of instinctive behavior patterns which produce learning rather than of the ‘instinct to teach,’ which, in any case, has subjective overtones… . The responses of the mother are simply those which provide the correct situation for evoking the developing repertoire of responses of the young who are thus enabled to educate themselves.’’

Question 6

13.5 Animal cultures?

Answer 6

Whatever else it may mean, when applied to humans, culture refers to multifaceted groupwide traditions: population-specific behaviors, beliefs, and attitudes, transmitted from one generation to the next through language, teaching, and in many less explicit ways. The socially transmitted behaviors of nonhuman animals described so far such as food preferences in rats or enemy recognition in birds influence so few aspects of their lives as to be scarcely the rudiments of culture.

But, in contrast to rats, birds, and most other animals, geographically separated groups of chimpanzees and
orangutans show multiple, populationwide differences in acquired behavior that have been suggested to represent evolutionary precursors to human culture.
The most substantial relevant data come from a collaboration among researchers doing long-term studies of chimpanzees at seven sites in Africa (Whiten et al. 1999, 2001). For each population, the local team estimated the frequency of occurrence of 65 behaviors, many of which involved tool use or other interactions with objects such as manipulating sticks in different ways to obtain ants or termites, using leaves to
sponge up water.

When a behavior had not been seen, a judgment was made as to whether there was an ecological explanation for its absence. For instance, termite
fishing is impossible without termites. The most interesting cases are those 39 in which a behavior was judged relatively common in some populations but absent in others even though the ecological conditions for its appearance were judged to be present. Given that genetic differences among the populations can be assumed to be unimportant, such patterns suggest the behavior must have been discovered by one or
more innovators and then acquired by others in the group by some kind of social transmission (i.e., any one or more of the mechanisms in Box 13.1).

Beginning with the titles of the original reports, these population differences in chimpanzees (Whiten et al. 1999, 2001) and orangutans (van Schaik et al. 2003) have been referred to as ‘‘cultural,’’ but this description is much debated (Galef 2004; Laland and Janik 2006; Perry 2006; Whiten and Van Schaik 2007; Galef 2009).
There are two basic sources of controversy. One, discussion of which is beyond the scope of this book, is that culture has a rich web of connotations in anthropology, archaeology, and a whole range of other disciplines, not to mention in folk psychology, and to some writers these are simply incompatible with the possibility of ‘‘animal cultures’’ no matter how apparently inoffensively and objectively defined.

The other is that even if population-specific behaviors in nonhuman species are referred to instead as behavioral traditions, conclusive evidence is needed that the behaviors involved really are transmitted socially as the term tradition implies rather than learned individually or determined by ecological conditions, and field observations alone rarely if ever can provide such evidence (Galef 2004, 2009). An analysis of ‘‘ant dipping’’ by chimpanzees shows how ecological factors favoring one behavior rather than another may not be obvious. In dipping for ants, a chimpanzee uses a stick or grass stalk to capture biting ants.

The tool is moved back and forth to stimulate the ants to climb up on it. They are then removed either by putting the tool directly into the mouth (‘‘direct mouthing’’) or by pulling it through the hand and putting the resulting clump of ants into the mouth (‘‘pull through
technique’’). Ant dipping is a candidate cultural behavior because different techniques as well as different lengths of tools are prevalent in different populations. However, in one population, at Bossou, Guinea, chimpanzees use both techniques as well as both short and long tools. By combining observations of the conditions under which different tool lengths and removal techniques were used with experiments in which the researchers themselves dipped for ants,

Humle and Matsuzawa (2002) showed that there are good functional reasons for these variations in dipping. It turns out that there are more and less aggressive species of ants; the ants are also more belligerent at
the nest than when migrating along the ground. Using longer tools and the pull through technique limits biting by the ants, and it is the preferred technique for
situations where ants are most aggressive. However, an analysis of behavior of ants at two sites with different patterns of ant dipping indicates that some of the population differences in ant dipping are likely to be cultural (Mobius et al. 2008).

A further issue is that no matter how compelling the observations of young animals watching adults using tools or the like (e.g., Figure 13.18), the occurrence of any kind of social learning or social influence on learning in such interactions needs to be tested
experimentally. Given the paucity of convincing evidence for social transmission of wild chimpanzees’ population-specific behaviors, researchers have turned to demonstrations that tool-using skills can be socially transmitted in captive groups (Whiten, Horner, and
De Waal 2005; Horner et al. 2006; Hopper et al. 2007).

These studies typically involve apparatuses like that used in two-action tests of imitation, introducing each technique for operating it into a different group of subjects. A third group may be left on their own
to see whether one technique or the other, if any, is acquired spontaneously. With an ‘‘artificial fruit’’ having a door that could be lifted or pushed, a transmission chain was formed. Observer 1 learned the technique used by trained demonstrator, then Observer 2 learned it from observing Observer 1, and so on up to a chain of five or six chimpanzees. Some controls who saw food put into the box eventually opened it one way, some the other (Horner et al. 2006).

Consistent with these findings, when a single
trained demonstrator was introduced into whole group, most individuals learned the tool use task being demonstrated and used the same technique as the demonstrator (Whiten, Horner, and De Waal 2005). However, the robustness of a technique across a
transmission chain may depend on the type of task (Hopper et al. 2007). Monkeys, too, have what appear to be traditional behaviors (Perry and Manson
2003).

Indeed, one of the oldest candidates for animal culture is potato washing by Japanese macaques (Box 13.3). More recently, wild white-faced capuchin monkeys
have been observed in what are some of the best candidates for socially learned population-specific behaviors. These are rather bizarre and apparently arbitrary ‘‘games,’’ such as monkeys taking turns putting their fingers into each others’ mouth and getting a firm bite (Perry et al. 2003). It has been possible to trace the spread of some of these behaviors within and between groups. Social transmission of a foraging technique along a chain of animals has also been demonstrated among captive capuchins with similar methods to those used for chimpanzees (Dindo, Thierry, and Whiten 2008).

Conclusions

Notwithstanding the need to look more closely at some of their ecological determinants, it seems likely that at least some of the candidates for traditional behaviors of
chimpanzees as well as orangutans are indeed socially transmitted. But does that mean apes have culture in any meaningful way? On one view (e.g., Perry 2006),
‘‘cultural primatology’’ from the early study of Japanese macaques onward reveals a great deal about how culture evolves and what mechanisms maintain it. On another (e.g., Galef 2009), animal traditions are analogous but not homologous to human culture because the processes that perpetuate them do not include the key component of human cultural transmission, namely imitation.

Human culture is indeed unique because across generations it ‘‘ratchets up’’: changes introduced in one generation are adopted and further elaborated in the next in a process of cumulative change. On one compelling account (Richerson and Boyd 2005), ratcheting up is possible because people are capable of exactly copying (i.e., imitating) behaviors of those around them and then improving on them by trial and error, reasoning, or other processes, whereas
emulation and other social learning mechanisms leave each new generation to relearn much of what was learned by the last. On this view, although humans undoubtedly share some simpler social transmission mechanisms with other species, the propensity to imitate sets us apart even from chimpanzees (see Herrmann et al. 2007) and makes genuine culture possible.

Question 7

13.6 Summary and conclusions

Answer 7

‘‘Social learning’’ has a lot in common with ‘‘spatial learning’’ (Chapter 8). Both are essentially functional categories, that is, based on the kind of information acquired rather than on the way in which it is acquired, and both encompass a variety of specific mechanisms. However, individual mechanisms of spatial learning such as path integration, landmark use, and sun compass orientation are relatively well understood, whereas the analysis of separate mechanisms for social learning has been impeded by disproportionate interest in true imitation.

The wave of recent research combining observations of naturalistic examples of social learning with experimental analyses of mechanism has led to an appreciation of how species-specific fine-tuning of simple learning mechanisms can lead to social transmission of adaptive behavior in natural social contexts. For example, rats learn about food by smelling other rats’ breath because the smell of rat breath has motivational significance for them and
because when rats greet each other the nose of one comes close to the mouth of another.

A young black rat need never see another black rat stripping the scales off a pine cone; it needs only to be provided with cones than have been partially stripped in the right way (Terkel 1995). Nonimitative social learning includes stimulus enhancement, observational conditioning, and emulation. None of these is very well understood in terms of the conditions that bring it about, the contents of that learning, and the effects of learning on behavior. Heyes (1994a) suggested that each is roughly analogous to a recognized
category of associative or perceptual learning, but the questions she raised, such as the role of contingency and the possible occurrence of overshadowing and blocking in such learning, have still hardly been asked.

It is necessary to answer them to know whether these kinds of social learning are distinctive in any way other than in the events that are learned about. There has, however, been considerable progress recently in understanding how imitation occurs and in what species, but emulation and affordance learning still need more study. In most circumstances there may be no need for strict imitation. The job can be done by emulation and the other social learning processes that don’t require storing a representation of the demonstrator’s behavior as such.

Indeed, a tendency to blindly imitate what others do regardless of the positive or negative outcomes for oneself would likely be maladaptive. Thus what may need to be explained is not why most species seem incapable of true imitation but why any are capable of it. This explanation may ultimately have to do with the evolution of teaching and human culture.