Theory of Mind and Insight in Chimpanzees, Elephants, and Other Animals?

 

Source:  Chapter 7 (pp. 227-261) of Comparative Vertebrate Cognition: Are Primates Superior to Non-Primates? Edited by Lesley J. Rogers and Gisela Kaplan (a featured volume in Developments in Primatology: Progress and Prospects).  New York: Kluwer Academic/Plenum Publishers, January 2004.

 

Moti Nissani, Department of Interdisciplinary Studies, 5700 Cass Ave., Wayne State University, Detroit, MI 48202 (USA)

E-mail:  aa1674@wayne.edu

 

Note:  For a more recent summary of our elephant research, please click here.

 

Clickable Table of Contents:

Elephant Cognition

Do Elephants and Chimpanzees Know that People See? 

EXPERIMENT 1: DO ELEPHANTS KNOW THAT PEOPLE SEE?  

EXPERIMENT 2: DO CHIMPANZEES KNOW THAT PEOPLE SEE? 

Insight in animals? 

RETRACTABLE CORD-PULLING IN ELEPHANTS 

DO ELEPHANTS KNOW WHEN TO SUCK OR BLOW? 

Conclusion 

Acknowledgments 

References

  

This chapter attempts to provide an accessible review of a fundamental scientific and philosophical question: Are animals conscious?   Instead of trying to address the issue of consciousness as a whole, this partial review will for the most part touch upon just two facets of consciousness— theory of mind and insight. 

This review takes it for granted that we need to place a question mark on the understandable but as yet unfounded assumption that our evolutionary next of kin—the great apes—are also our closest cognitive relatives.  At the moment, we cannot rule out the view that “a classification of the animal kingdom based on intelligence would probably cut right across the classifications based on structure” (Hobhouse, 1915).  As a matter of fact, at the moment we cannot even refute the counterintuitive claim that there are no differences in intelligence between one nonhuman vertebrate and another (Macphail, 1982; Thomas, 1986). 

On first sight, the case for consciousness in animals seems overwhelming.  Given the evolutionary notion of continuity, and given moreover the remarkable physiological and genetic similarity between apes and humans, it seems scarcely credible to argue that apes are devoid of any trace of consciousness.  Furthermore, it seems reasonable to suppose that consciousness confers an enormous evolutionary advantage, for it allows animals to try out possible actions in their head without actually performing them through costly trial and error (Griffin, 2001).  Common intuition seems to point in the same direction, as the following passage suggests. ”When I looked into Washoe’s eyes she caught my gaze and regarded me thoughtfully, just like my own son did.  There was a person inside that ape ‘costume.’  And in those moments of steady eye contact I knew that Washoe was a child” (Fouts, 1997).

Regretfully, the question of animal consciousness cannot be resolved by either intuition or theory. Although intuition often serves as valuable breeding ground for research ideas, it is notoriously fallible.  And as far as theory is concerned, it could be just as well supposed that consciousness exerts its own evolutionary price (e.g., hesitation when swift action is called for), or that it is not readily achieved, even when favored by natural selection. 

 

Elephant Cognition 

Much of the original work reported in this chapter deals with elephants, so it may be worthwhile to briefly review their cognition.  Because we know so little about the minds of elephants, and because the three known species of elephants are similar, this review will focus on the cognition of elephants in general terms, often leaving aside the terra incognita of possible cognitive and behavioral differences among the three species.

Although elephants have been in close association with humankind for thousands of years, and although anecdotes about their wisdom or witlessness are many, their mentality has only been subject to a mere handful of controlled studies.  This curious gap in the research literature provides fertile grounds for speculations and controversies.  Indeed, views on the subject range from the assertion that “elephants are exceedingly intelligent; that they have a form of intelligence which manifests itself in many ways that are very like our own; and that, in these respects, they stand as far apart as we do from all other living things—the great apes not excluded” (I. T. Sanderson, 1962), to the assertion that the “elephant is a stupid animal; and I can assert with confidence that all the stories I have heard of it, except those relating to feats of strength or docility performed under its keeper’s direction, are beyond its intellectual power, and are mere pleasant fictions” (G. P. Sanderson, 1912; see also Carrington, 1958).

At 6 kg, the elephant brain is the largest brain of all land mammals (Rensch, 1957).  Sheer size however means little, for a good part of the brain must support the bulk of an animal and some of its special, non-cognitive, functions.  Seen this way, the elephant is a comparative lightweight: an elephant weighing 75 times as much as a human has a brain weighing only 2-3 times as large (Spinage, 1994).  Likewise, the elephant’s cerebral hemispheres do not cover the cerebellum as they do in humans and apes (Spinage, 1994).  But, in common with the brain of dolphins, great apes, and humans, the brain of elephants is highly convoluted—a possible indicator of high intelligence and of an advanced capacity for learning (Alexander, 2000).  Elephants  also have a large neopallium (the area associated with memory).  Above all, most mammals are born with brains that are around 90 percent of their weight as adults, but humans are born with about 27 percent, elephants with 35 percent, and chimpanzees with about 54 percent (Eltringham, 1982; Poole, 1997, p. 38).  There is thus considerable post-natal brain growth in these three species, a fact which is believed to be linked to the exceptional learning ability of their young.

The only complete study of elephant cognition was conducted on a 5-year-old Asian female at the Munster Zoo.  She faced two small wooden boxes whose lids were painted with two different patterns, for example., circle and cross, and had to remove the correct lid to get a food reward.  It took her about 330 trials, over a period of several days, to consistently choose the reinforced pattern, e.g., the cross (Rensch, 1956, 1957).  By the fourth single-pair discrimination task, she reached criterion on the 10th trial.  After learning to discriminate between 20 symbol pairs, she performed superbly on a test that combined all 20.  The test lasted several hours, yet her performance actually improved toward the end.  A year later, her scores ranged from 63 to 100 percent—a scientific demonstration, Rensch says, of the adage that elephants never forget.

In a partial study, Leslie Squier (reported in Stevens, 1978) confirmed the elephant’s ability to learn a simple discrimination task and to remember it 8 years later, “a remarkable performance and one that many Homo sapiens might have difficulty emulating” (Markowitz, 1982).  In another study, Povinelli (1989) reported that elephants do not recognize themselves in the mirror (an observation that we have confirmed with our two Asian elephants—Nissani and Hoefler-Nissani, 2002, unpublished observations).  Thus, the few controlled  experiments on record seem to lend credence to Carrington’s conclusion that “elephants are not sufficiently intelligent to grasp an idea easily or quickly in the early stages, but once it has penetrated to their somewhat slow brains it is virtually ineradicable.”

In contrast to the picture which emerges from these zoo studies, field observers of elephants in the wild are often more impressed by their minds (e.g., “the more we learn about elephants, the fainter the line we have drawn between man and other animals will become”—Poole, 1997).  In one field study, McComb et al. (2000) found that an average female can recognize the contact calls of about 100 other females.  Equally remarkable, the “legend” of coming to the aid of a wounded comrade is true.  In one confirmed instance, when a bull was shot, two younger elephants came to his rescue and tried to lead him away (Denis, 1963; cf. Carrington, 1958).  Romanes’ (1882) seemingly tall tale that, during an operation, “elephants behave like human beings, as if conscious that the operation was for their good, and the pain unavoidable,” is likewise probably true (Blashford-Snell and Lenska, 1997; Groning and Saller, 1999; Shand, 1995).  Aelian’s fantastic 3rd century recount of an Ethiopian tale that “if one elephant sees another lying dead, it will not pass by without drawing up some earth with its trunk and casting it upon the corpse, as though it were performing some sacred and mysterious rite on behalf of their common nature” has been likewise confirmed by contemporary researchers.  During drought, the African elephant is the only animal known to dig for water, and there is moreover an account of an elephant chewing bark to fashion a plug, using it to plug a water hole, and hiding the hole from other animals by covering it with sand (Gordon, 1966).  In Burma (=Myanmar), according to one observer (Williams, 1950), raiding young elephants often stuff their bell with mud in order to avoid detection.    

Elephants are proficient tool users (Chevalier-Skolnikoff and Liska, 1993; Hart and Hart, 1994).  They often scratch themselves with sticks and break fences by piling logs—or even young elephants—over them.  There are likewise many reliable accounts of elephants hurling objects (Chevalier-Skolnikoff and Liska, 1993; Wickler and Seibt, 1997).  A 27-year-old bull, for instance, routinely concealed a rock in his trunk and hurled it at his keeper (Linden, 1999).  Once the training of elephants in playing cricket or soccer is over, they “play the game with the enthusiasm of boys having a knock-up on the village green” (Carrington, 1958; I. T. Sanderson, 1962).

I have tried to confine my very partial account of the elephant’s mind to the more trustworthy, or to independently-corroborated, observations, omitting the more anecdotal, and even more fantastic, tales.  Even so, the picture that emerges is of an animal whose mentality deserves closer scientific attention than it has so far received.  

 

Do Elephants and Chimpanzees Know that People See? 

An organism can be said to possess a theory of mind when it is capable of attributing mental states to others—when it understands that others see, feel, and know.  An organism’s theory of mind might be inextricably linked to its capacity for consciousness, insight, concept of self, and such emotions as pride, shame, empathy, and compassion.  The knowing component of this complex concept can perhaps be illustrated by imagining the following sequence (following Zentall, 2,000).  In the first scene, little Susan observes Wong Tsu hiding a candy under a pillow and then going outside to play.  As the next scene unfolds, Susan sees an adult entering the room, removing the candy without disturbing the pillow, and hiding the candy in a desk drawer.  Next, Susan sees Wong Tsu coming back.  If Susan expects him to look for the candy under the pillow and not in the drawer (i.e., if she implicitly theorizes that Wong Tsu has a mind and that in his mind the candy is still under the pillow), we can say that Susan possesses the knowing component of a theory of mind. 

As we shall see, attempts to prove the existence of a theory of mind are often exceedingly complex when we divert our gaze from young children to animals.  According to Lorenz (cited in Thorpe, 1956), if a raven is observed hiding food, it often retrieves the food, scolds the observer, and hides the food out of human reach.  On first sight, this behavior appears to involve the same conceptual understanding as Susan’s, but here the cause is slippery—it could be traceable to understanding the perspective of another, or to genetic programming, or to prior trial and error learning.

Probably owing to such conceptual roadblocks, the question of whether any animal besides ourselves possesses a theory of mind has never been scientifically resolved.  Hunter-gatherers, ancient civilizations, and Buddhism seem to have taken it for granted that animals are conscious.    But with the advent of Christianity and, later, Cartesian philosophy, most educated westerners came to view animals—including the ones closest to us in appearance and behavior—as mindless, unfeeling, automatons.  A couple of centuries later came yet another paradigmatic shift, for evolutionary theory seemed to require a continuity of  physical and mental characteristics among humans and their closest relatives.   

Darwin’s views prevailed until the mid-1990s or so.  On the experimental side, numerous studies and field observations seemed to independently converge on the belief that some animals, at least, possess a theory of mind.  Chimpanzees, for example, were said to be capable of insightful problem solving, acquiring a rudimentary form of language, self-referential behavior in front of a mirror, and deception (cf. Russon et al., 1996).  Moreover, chimpanzees split from the human line only a few million years ago, share close to 99 percent of the human genome, and show remarkable behavioral similarities to human beings.  All these observations seemed to imply that chimpanzees, and perhaps other species, possess a theory of mind.

Many researchers, to be sure, rejected this seemingly irresistible avalanche, but theirs appeared to be the kind of rearguard action that is often seen in science.  In the 1990s, however, the pendulum began swinging again in a neo-Cartesian direction, and here one set of elegant experiments stands out, for it appeared to provide the clearest experimental evidence to date that chimpanzees—and perhaps all other non-human animals—possess “clever brains but blank minds” (N. Humphrey, cited in Gallup, 1998).

In these experiments, Daniel J. Povinelli and his colleagues capitalized on captive chimpanzees’ tendency to beg from their keepers to ask:  Do chimpanzees understand the seemingly elementary fact that people see? (Povinelli and Eddy,1996; Povinelli et al., 2000; Reaux et al., 1999).  Six or seven young chimpanzees took part in a complex series of longitudinal studies.  In a typical pre-training phase, a youngster entered a room where it faced, across a clear Plexiglas barrier, a familiar experimenter.  The experimenter sat on either the right or left side of the barrier, in each case directly across a hole in the Plexiglas.  In an average of 48 consecutive sessions consisting of 10 trials each, the chimpanzee learned to insert a hand through the hole which faced the experimenter, as opposed to the hole which did not.  Once this skill was acquired, the chimpanzee had to choose between one experimenter who had appealing food in her hand and one who held a neutral block of wood.  In this case, the chimpanzee received a food reward only when it inserted its hand through the hole facing the food-carrying experimenter.   

In the experiment itself, the chimpanzees faced variations along a single theme: in all variations they had to spontaneously decide whether to insert their hand toward an experimenter who could see them or toward an experimenter who could not.  Only a few of these variations need to be described here (cf. Povinelli and Eddy, 1996).  In all these variations, one of the experimenters faced the right hole of the barrier while the other experimenter faced the left hole (cf. Figure 1).  Thus, in a typical blindfold trial, both experimenters stood motionlessly in front of the right or left hole, one with a blindfold covering her mouth and the other with an identical blindfold covering her eyes.  Thus, the former could see the chimpanzee, while the latter could not.  In the buckets condition, the experimenters stood or sat in the same place, but one held a large bucket on her shoulder and could see the animal, while the other experimenter covered her head with a bucket and could not see the animal.  In the similar screen condition, one experimenter held a cardboard screen on her shoulder while the other held it so that it completely obscured her face from the chimpanzee.  In the Back/Front condition, one experimenter faced the Plexiglas partition and the chimpanzee, while the other had her back turned to both.  In the looking-over-shoulder condition, both experimenters assumed the same position and both had their backs turned to the chimpanzee, but one experimenter had her face turned toward the animal, while the other experimenter turned her face away. 

Startlingly, in the first few trials, in all but one variation, the chimpanzees consistently performed at chance level.  That is, and in contrast to the performance of young children (3 years of age or older) in a similar task, Povinelli’s chimpanzees were just as likely to beg from a person who could see the begging gesture as from a person who could not.  The one exception in the original set of experiments involved the back/front combination (Figure 2, frame B3), but further experiments suggested to the authors that the chimpanzees’ proficiency in this task was merely “a consequence of their reinforcement history in the training phase” (Povinelli and Eddy, 1996).  Thus, chimpanzees “do not seem to appreciate that others ‘see’ things” (Povinelli and Giambrone, 2001).

Figure 1. Four probe conditions in chimpanzee seeing experiments. Top left:  blindfolds.  Top right:  buckets.  Bottom Left:  sideways.  Bottom right: screens.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

These results, along with similarly discouraging studies of pointing and gaze-following, led Povinelli to the neo-Cartesian view that only humans can represent explicitly their “own psychological states and those of others” (Povinelli, 1998).  Later, Carruthers (1998) based his view that sentience is exclusively human in part on Povinelli’s results.  

The question of animal consciousness in turn has profound implications for our very conception of the world around us, for evolutionary biology, for comparative psychology, and perhaps also for the treatment of human autism (Baron-Cohen et al., 2000; Carruthers and Smith, 1996).  It has also practical and controversial implications, ranging from a call for putting into place “more rigorous standards for conducting and reporting empirical research” in comparative psychology (Povinelli and Giambone, 2001), to endorsing intrusive research on apes (Wynne, 1999).

Thus, if the neo-Cartesians are right, only one animal possesses a measure of understanding of its own and others’ actions and mental states—and that animal is us!  It follows that the ongoing, extensive search for culture, imitation, deception, or genuine self-awareness in animals is misguided in principle, somewhat akin to the quest for the  philosopher’s stone.  Looking closely at any great ape, we cannot, indeed, help feeling that it thinks and feels, at least in some rudimentary fashion, as we do.  But this, the neo-Cartesians insist, is an illusion.

Povinelli’s work is consistent with some well-known observations.  For one thing, his work throws light on our failure to come up with a single unequivocal demonstration of a theory of mind in animals; we may have failed to come up with a clearcut proof, neo-Cartesians may say, for the very good reason that such proof does not exist.  Computer models likewise raise the possibility that complex behaviors, of the kind that are often shown by chimpanzees, can emerge mechanically, through mere learning and the pressure of natural selection, from the interaction of simple behavioral building blocks (cf. van Hooff, 2000; see also Morgan, 1920).

Povinelli’s experiments, and the neo-Cartesian hypothesis they led to, have however been questioned (Byrne 2001; Byrne and Whiten, 1988; Call, cited in Tuma, 2000; Gallup, 1998; Hare et al., 2000, Heyes, 1998; Russon et al., 1996).  Before describing collaborative research efforts with elephants and chimpanzees at the Detroit Zoological Institute, and in an effort to provide a rationale for their design features, I wish to highlight a few additional uncertainties in Povinelli’s experiments. 

Povinelli tacitly assumes that, cognitively speaking, the chimpanzee is our closest relative.  This assumption is not necessarily true, for reasons that emerge elsewhere in this chapter.  Thus, the neo-Cartesian hypothesis might be wrong, even if it can be shown that chimpanzees lack a theory of mind.

Although inconclusive, there is a body of circumstantial and experimental evidence suggesting that apes do have a theory of mind (e.g., Povinelli et al., 1990; Premack, 1988; Uller and Nichols, 2000).  There is, in particular, experimental evidence that “chimpanzees are capable of modeling the visual perspectives of others” (Povinelli et al., 1990) and that, at least in some situations, they know “what conspecifics do and do not see” (Hare et al., 2000; but cf. Povinelli and Giambrone, 2001 for a dismissal of this latter claim). The tentative confirmatory evidence extends to other species as well; for instance, marmosets are claimed to be capable of true imitation (Voelkl and Huber, 2000), bottlenose dolphins seem capable of mirror self-recognition (Reiss and Marino, 2001), and even the green bee-eaters of India seem capable of distinguishing a person who can see them from a person who cannot (Smitha et al., 1999, cited in Griffin 2001).  Given these observations, more than one set of experiments, one basic design, and six or seven subjects of one species (Povinelli and Eddy, 1996) are needed to resolve the theory of mind controversy.

Chimpanzees are not only physiologically our closest kin (de Waal, 2001), but they often behave similarly to human beings, for example, like us, they too follow the gaze of another, use tools,  and engage in tribal warfare.  Occam’s Razor would seem to imply similarity of the cognitive processes which underlie such analogous behavioral patterns.  But, if cognitive similarity is rejected, one must resort to an auxiliary evolutionary hypothesis.  Thus Povinelli and Giambrone (2001) are forced to put forward a reinterpretation hypothesis, which counterintuitively suggests, for example, that what appears to us as deception in animals is in fact traceable to the evolution of “brain systems dedicated to processing information about the regularities of the behaviors of others.”

One unresolved issue of Povinelli’s work involves the age of his chimpanzees.  Since the initial work was carried out with young chimpanzees roughly between the ages of 5 and 6 years, the experiments left open the possibility that theory of mind arises later in chimpanzee development, as Povinelli and Eddy (1996) noted.  To meet this objection, Povinelli and colleagues re-tested the same chimpanzees twice, up to the age of 9 years.  In the first retest, a year or so after the original work was completed, the chimpanzees forgot the little they had learned, and again seemed unable to tell the difference between a human being who could see them and a human being who could not.  But in the second test, when the same chimpanzees approached 9 years of age, they performed extremely well from the start in the buckets, screens, and looking-over-shoulder tests (see Figure 1), and only did poorly in tests that involved a direct understanding of the role of human eyes in visual perception (Reaux et al., 1999).  These striking improvements at close to age 9 could then be interpreted in two radically different ways.  One could assume, with Povinelli et al., “that there was no development between 5 and 9 years of age in the animals’ understanding of visual perception as an internal state of attention,” and that throughout this longitudinal study, they merely learned stimulus-based rule structures.  Alternatively, one could assume that the older chimpanzees of this third series of experiments have by now matured enough to understand the attentional aspects of seeing (or at least, have acquired on their own the broad rule of begging from the person whose face is visible), and, in turn, explain their random performance in subtle tests involving covering the eyes by assuming that chimpanzees, like some of Povinelli’s children, may understand that there is indeed someone behind those eyes—without understanding the working of eyes (Hare et al., 2000).  That is, the first hypothesis is congruent with the radical assertion that chimpanzees lack a theory of mind; the second, with the trivial conclusion that chimpanzees do not understand how human eyes work.  Needless to say, to decide which of these two hypotheses is more nearly correct, one must test adults who have not taken part in such tests before. 

It is probable that the seeing experiments of Povinelli and colleagues lack ecological validity, a related set of problems that could only have been dealt with through a radical alteration of some features of their experiments. 

To begin with, the experiments are claimed to have relied on the species-typical begging gesture.  In point of fact, however, the chimpanzees had to learn the unnatural gesture of inserting a hand through one of two holes of a transparent plastic barrier.  As might be expected, it took them an average of 479 trials, and dozens of sessions, to learn this allegedly natural response.  Human observers are tempted to interpret such insertions as begging, but to the chimpanzees themselves they may have had no meaning whatsoever, thus accounting for their chance performance in the critical tests.  In a comparable context, Gomez (1998) argues that the prolonged training of subjects in artificial tasks is a mistaken approach.

It is not clear either what the natural begging response of captive chimpanzees is.  Our chimpanzees (but not our elephants) seemed to employ a variety of idiosyncratic begging postures, instead of just one.  Thus, when begging services, toys, or food from their keepers, one of our captive chimpanzees often nodded her head while another whimpered, banged on the wire mesh, or clapped his hands.  Indeed, often throughout our own experiments the chimpanzees employed one of these more natural responses first, and only then resorted to the required learned gesture of inserting their fingers through or under the wire mesh. 

Moreover, Povinelli’s chimpanzees begged from people who often assumed unnatural, seemingly indifferent, postures.  For instance, throughout any buckets trial, even the seeing experimenter remained motionless and made no eye contact with the animal.

Povinelli’s overworked subjects have been experimented with virtually all their lives.  They were all born in captivity.  Four were reared in a nursery from birth and two were reared by their mother for 1 year and then transferred to the nursery.  This background raises serious questions about their natural curiosity and their emotional and cognitive development (Harlow and Harlow 1965; Thompson and Melzack, 1956).  Needless to say, few if any of the children in Povinelli and Eddy’s initial experiment—which provided the backdrop, controls, and confirmation of the results—suffered the traumas of lifelong experimentation and maternal deprivation.

A related problem involves the over-reliance in most trials on the food versus block of wood configuration.  This “baseline probe” was frequently used during extensive pre-training sessions, and it often outnumbered the “treatment probe” in experimental sessions by a ratio of 2:1.  This might have led the chimpanzees to believe that their task in some of the critical trials involved guessing who had the food, not begging from the person who could see them.  Indeed, Povinelli and Eddy’s data lend a measure of support to this interpretation.  Thus, in their Experiment 7, when the majority of trials involved begging from a single person only, and only comparatively few trials involved food versus block of wood before and during critical sessions, the chimps performed significantly above chance in the screens test, leading the authors to suggest that some learning had occurred in this case.  But, in their Experiment 8, when only food versus block of wood were used as spacer trials (and no singles), the chimpanzees’ performances in the screen test returned to chance level.  In this context, Povinelli and Eddy are puzzled: “It is difficult to know why the subjects performed so poorly on the screen-over-face probe trials in this experiment as compared to Experiment 7.”  Such unaccountable results can in turn be readily explained by the assumption that the chimpanzees were confused about the nature of the task they were to perform in these experiments.  And again, when working with children, the equivalent of food versus wood was used in training, but only once during the trials themselves, thus raising doubts about the validity of the key comparison between the young chimpanzees and the younger children.   

Another possible design flaw involves inadequate payoffs.  In a typical experiment (for instance, their Experiment 1), 80 percent of the trials presented no choice at all, 8.9 percent  presented the easy, immediate choice between a block of wood and a favorite food item, 4.4 percent the easy choice between an experimenter’s back and front, and only 6.7 percent presented the more difficult, experimental choices (e.g., looking over the shoulder).  Thus, the subjects were assured of being rewarded in 168 of 180 trials (93.3 percent) by acting mechanically, and were likely to be rewarded in 174 of every 180 trials (96.7 percent).  This high reward ratio seems to beg mechanical, behavioristic, action, rather than thought and contemplation—if you wish to encourage thinking, you must create situations where the payoffs for thinking before acting involve is a bit more than a 3.3 percent increase in the reward schedule.  Thus, we may conclude that the choice between a seeing and non-seeing person was not as automatic as the choice between back and front, or food vs. wood, but not that it could not be made at all.  This difficulty is compounded by the fact that the relevant fractions are entirely different for the human subjects, who had to make the more difficult choice in 30 percent of the trials.  Acting mechanically, for the toddlers, would have involved a 15 percent loss of rewards vs. a 3.3 percent for the chimpanzees.

To the best of our knowledge, Povinelli’s ingenious experiments, despite their procedural simplicity, serious methodological and conceptual flaws, counterintuitive character, wide acclaim, and profound  implications, have not been replicated with chimpanzees (although other researchers addressed similar questions by relying on different approaches, e.g., Hare et al., 2000), nor have they been extended to other species.  It seemed worthwhile therefore to reproduce Povinelli’s essential procedure while safeguarding against some of its design flaws.  Thus, to avoid reliance on just one species before rejecting Darwin’s cognitive continuity hypothesis, the experiments described here were carried out with seven chimpanzees and two elephants.  To meet the age objection, six of the seven chimpanzees we worked with, and both elephants, were adults.  To avoid extensive pre-training of the subjects, we tested the animals in their own compounds and, instead of requiring them to insert their fingers or trunks through one or another hole of an artificial barrier, we monitored the almost as reliable criteria of insertion and direction of hand (and trunk) gesture (see below for further details).  To bypass the problem of excessive experimentation, our work was carried out in a zoo setting, with subjects that were thought to have never taken part in psychological experiments.  To partially meet the objection of emotional deprivation, at least some of the subjects we chose were raised by their mothers.  To maximize the chance that the chimpanzees understood that their task was not guessing who had the food, we drastically curtailed the use of the food versus block of wood condition, relying instead on the back/front and on the new lying-down (Figure 2, B4) conditions.  To avoid the inadequate payoffs difficulty, the fraction of critical trials in each session was raised, this fraction was varied from one session to another, and the number of trials in each session was varied as well. 

Thus, if we too observe chance performance, despite deliberate attempts to alter Povinelli’s design, his work and its far-reaching conclusions would receive much-needed confirmation.  On the other hand, a partial failure to replicate his results would require additional work to tease apart the variable(s) that account for the difference, and would require new experiments to decide whether some chimpanzees “understand the attentional aspect of gaze” (Povinelli and O’Neill, 2000).  A partial failure, in other words, will return the pivotal question of theory of mind in chimpanzees and other animals to the status quo ante, prior to Povinelli and Eddy’s 1996 experiments.

 

Figure 2. Four background conditions in chimpanzee seeing experiments. Top left: One experimenter with food tray. Top right:  Food versus flashlight. Bottom left:  Back versus front. Bottom right: Lying toward or away from subject.

 

 EXPERIMENT 1: DO ELEPHANTS KNOW THAT PEOPLE SEE?

Two Asian elephants (Elephas maximus) of the Detroit Zoological Institute served as  subjects. Winky, the older of the two, was wild-born in Cambodia, probably in 1952, acquired by the Sacramento Zoo in 1955, and moved to Detroit in 1991.  Wanda was wild-born in India in 1958, brought to the United States around 1960, and moved to Detroit in 1994. During their tenure in  Detroit, these elephants have never served as experimental subjects.  Little is known about their history prior to arrival in Detroit.  

In the experiments described here, which took place from May to December of 2001, a group of 3-5 researchers and keepers entered an enclosed courtyard inside the elephants’ one-acre compound, typically around 12:45 p.m., three times a week.  If not already by the cables separating the courtyard from their open-air enclosure, both elephants typically approached the cables and stayed nearby for the duration of the experiment.

In each experimental trial, the elephant was first brought in line by one keeper (the commander of that trial), so that she faced the center of the courtyard.  In each trial, one or two other keepers served as givers, standing approximately 4m away from the elephant with food in their pockets or on a stand behind them, and with their backs to the elephants.  The givers then turned, assumed the experimental position, stood motionless in that position for a few seconds, and then approached the elephant in that position (e.g., one walking forward toward the elephant while the other waking backward at the same pace), coming to a stop some 2m apart from each other and about 1m away from the elephant, avoiding eye contact with the elephant but looking instead about 2m below the eyes, at the center of the elephant’s trunk. 

In over 90 percent of the trials, the elephant waited for this scene to unfold before begging.  If her trunk crossed the cables before the givers came to a standstill, she was given the familiar verbal command “trunk down” by the keeper/commander (but not by the givers).  Following this command, she readily withdrew her trunk beyond the cables and waited for all movements to cease before begging again.  The roles of commander and givers were changed in consecutive trials, with the keeper who last gave the food either not taking part or serving as the commander.  In each trial, positions and roles were predetermined in a quasi-random fashion (i.e., never assigning a plus value to the same side in more than two consecutive trials). 

When the elephant withdrew from the experiment (especially at the beginning, where the novelty of the situation seemed to distress them), the experiment continued on the following session.

At least two observers independently recorded their observations of each session, while a third person photographed each session on a digital video camera.  Interobserver reliability was calculated from the independent records of the two observers and from the computer log of a third person who independently observed the video record. 

The single condition, back/front, buckets and screen variations were almost identical to those of Povinelli and Eddy’s (1996).  The rock/food condition was comparable to their block of wood/food condition.  The sideways posture (Figure 3, Frame P1), which is comparable to their looking-over-the-shoulder posture, was chosen to provide sharper visual definition.  We have, as well, introduced a lying-down posture (Figure 4, B4), which involved one keeper lying down on her side and facing the elephant, and the other lying on her side and looking away from the elephant.

 

Figure 3. Three experimental conditions in the elephant seeing experiments. Left frame:  Sideways, right choice. Center frame:  Bucket over head versus bucket over shoulder, wrong choice. Right frame:  Screen over face versus screen over shoulder, right choice.

 

 

Figure 4. Four background conditions in the elephant seeing experiment.  Top left: A captive elephant’s natural begging pose. Top right:  Food versus rock, wrong choice.  Bottom Left:  Back versus front, equivocal choice. Bottom right: Lying down, wrong choice.

 

At the start, we had to conduct preliminary vision tests.  Wild elephants are often active at night, they are more dependent on olfaction than vision, and there are even reports of a blind matriarch competently leading her troop (Groning and Saller, 1999).  Moreover, their vision often deteriorates with age (Stevens, 1979).  When food was placed inside a  hollow PVC tube (15cm in diameter), both our elephants seemed to grope for the opening with their trunk, not by relying on vision.  Fortunately, other lines of evidence suggested that our two elephants had adequate vision.  In routine veterinary eye examinations, Wanda’s eyes were judged normal while Winky appeared to have limited vision in the right eye and normal vision in the left.  Similarly, the two elephants showed little interest in the back side of a 86X193 cm² mirror, but became agitated, manifesting behavior never seen before by their keepers, when the mirror was turned toward them.  When a turkey’s feather was taped to Wanda’s forehead on three different occasions, Winky deftly removed it, suggesting that she saw it clearly enough. Both elephants immediately retrieved distant food items attached to nearby ropes and chains with diameters as small as 1 cm.  When 1.4cm³ (4g) white sugar cubes were tossed near them, the elephants appeared to notice the movement, looked in the direction of the rolling cubes, and retrieved them with ease.  Finally, although Wanda and Winky daily respond to a variety of commands, each consisting of both arm movements and words, in 10 trials with each (5 in one session and 5 in another about one month later) they obeyed immediately 19 of 20 commands when only the gestural component of each was used, again suggesting that their vision sufficed to detect the experimental conditions depicted in Figure 3.  To be on the safe side, however, our elephant design avoided fine discrimination tasks (e.g., involving blindfolds or closed eyes).

Another critical assumption of the seeing experiment with the elephants (but not with the chimpanzees) involved the question of imitation or social learning, for the elephants could not be physically separated from each other during the experiment.  While we cannot go into this question here, a variety of control procedures, and our observations of the two elephants in many other tasks, suggest absence of imitation.  Our results can therefore be safely viewed as an indication of individual, not collective, responses.

For reasons given earlier, and because the elephants performed poorly on the rock/food begging task (Table 1), we chose, as spacer trials, the more reliable single (Figure 4, B1) and back/front (Figure 4, B3 ) variations.  On any given session, 6-8 such trials were interspersed with 2-4 experimental conditions.

We have also attempted to rule out the possibility that the elephants were given unconscious cues by their keepers, experimenters, and onlookers.  To begin with, the elephants performed comparatively poorly on the rock/food task (Table 1), and thus failed not only to distinguish the presence of food, but also to receive any subtle cues from the experimenters.  Moreover, at some tasks (e.g., screens; Figure 3, P3), the two elephants performed at chance level, again suggesting that they were not responding to subtle, unintentional cues.  Finally, in preliminary experiments aimed in part at inquiring into this question of unintentional cues, we presented each elephant with a choice between garlic- and vanilla-soaked identical blocks of wood, and rewarded them only when they tapped the vanilla block.  In over 100 trials, although the experimenters knew what the correct choice was, each elephant performed at chance level.  This again suggests that our elephants were not responding to unintentional cues. 

In the experiment itself, when an elephant begged correctly, holding her trunk steady for at least 2s directly in front of the appropriate keeper, she was praised and given a portion of a choice morsel.  When she begged from the wrong person, or when her choice was not clear, the keepers withdrew.

Probably because the experimental setup mimicked so closely the elephants’ normal routine and natural begging gesture, the elephants reached criterion of begging correctly (9 of 10 times) from a single person by the first pre-training session.

The results are given in Table 1.  While the small sample, and the limited number of trials (to minimize learning effects), do not allow us to carry out meaningful statistical tests, the elephants seemed to have performed in the back/front condition as well as Povinelli and Eddy’s (1996) chimpanzees, and to outperform them in the three experimental conditions.  On average, in the three conditions on the right, our elephants were correct in 69 percent of the trials (as opposed to Povinelli and Eddy’s 54 percent). Moreover, in the lying-down condition—a configuration they have most likely never experienced before in their life—they again did exceptionally well.

 

Table 1.  Do elephants know that people see?

*In chimps, the actual position was the very similar one of looking over the shoulder.

†† Povinelli and Eddy, 1996.

 

To throw additional light on our data, in subsequent natural trials we capitalized on Winky’s (but not Wanda’s) proclivity to beg from people some distance away.  In ten consecutive trials spaced 1-5 minutes apart, a keeper stood just outside her trunk’s reach, holding a 1-liter bottle of carbonated soda drink.  In odd-numbered trials, the keeper faced Winky; in even-numbered, the keeper had her back turned to Winky.  When the keeper faced her, Winky begged silently by extending her trunk toward that person.  When the keeper had her back turned to Winky, in four trials Winky forcefully blew air on the keeper; in the last trial she begged and then blew.  In another set of similar trials, the keeper was just within trunk’s reach.  In the five front trials, Winky begged and retained the begging posture without being told to, for at least 5s, never touching the keeper.  In all five back trials, Winky never begged; instead, she immediately touched the keeper.

Our preliminary results could of course be ascribable to chance—we had only two elephants, and we could only test them a few times (to minimize trial-and-error learning).  It is likewise possible that elephants are better at this task than chimpanzees.  But, we felt, the difference could also stem from the comparatively older age of our two elephants and from our modified design.  It seemed worthwhile therefore to apply our design to adult chimpanzees. 

 

EXPERIMENT 2: DO CHIMPANZEES KNOW THAT PEOPLE SEE?

Six adult (range 15-31 years; mean age 23.8 years) and one subadult (age 11) chimpanzees, all members of the single social group of ten of the Detroit Zoological Institute, took part in this study (Table 2).  The study started on November 2001 and ended on May 2002.   All sessions were conducted while subjects were in their ordinary 3x3x3 m³ feeding and sleeping morning cages.

 

Table 2. Name, age (at time of experiment), sex, birthplace, and extant rearing history of chimpanzee subjects

Name	Age (years)	Sex	Birthplace	Extant Rearing History	Known Family Relationship to other subjects
Bubbles	31	F	Wild	Unknown
Joe Joe	30	M	Wild	Unknown	Father of Tanya
Beauty	29	F	Wild	Unknown
Barbara	19	F	Captivity	Unknown
Abby	19	F	Captivity	Mother raised
Akati	15	F	Captivity	Hand raised
Tanya	11	F	Captivity	Mother raised	Daughter of Joe Joe

 

At any given trial, some non-participating chimpanzees could hear their fellows and the experimenters, see the experimenters and follow their preparations for a coming session, but could see neither chimpanzee nor the experimenters at the critical point when the subject inserted her fingers through the mesh.

For the most part, we adhered to the design features described earlier in connection with the elephants, so we only need to highlight some differences between the two. For the chimpanzees, begging was defined as the insertion of at least a portion of one finger through or under the mesh toward the correct experimenter, on that experimenter’s side of the cage, beyond the imaginary vertical line which separated the right and left sides of the cage. 

Pretraining Phase I. Finger Insertion to Beg Food.  This phase began with the head keeper approaching the right or left side (pre-selected at random) of the wire mesh with a trayful of tidbits, kneeling down (to be on the subject’s eye level), facing the subject about 0.5 m away from the wire mesh, modeling for her the desired finger-insertion behavior, and only rewarding her (by placing the food in her outstretched fingers, not through the zoo’s customary procedure of mouth feeding) when she correctly inserted her fingers.  At this phase of the experiment, the head keeper used her customary training procedures, which included direct eye contact and soothing, encouraging words.  Criterion was defined as correct insertion in four out of five successive approaches. 

Pretraining Phase II. Finger InsertionToward Experimenter Carrying Food.  The second phase of pretraining involved a choice between two givers.  At the beginning of each trial, the givers consulted that day’s written plan.  They then stood 2m apart from each other, with their backs turned to the subject, about 3m away from the wire mesh.  Following a predetermined schedule, one experimenter held a tray of desired food items on her left or right shoulder, while the other held a .4m long, red flashlight on her same (right or left) shoulder.  Both experimenters simultaneously turned to face the subject and moved toward her in tandem while fixing their gaze on the center of the wire mesh.  Both Es halted at the same time, when they were within 0.5m of the mesh, and knelt.  They now waited for the subject to move toward the food side if necessary, and to insert one or more fingers through the mesh on that side.  If the subject failed to respond within 30 seconds, responded with her own earlier idiosyncratic begging gesture, or inserted her fingers toward the experimenter carrying a flashlight, the two experimenters withdrew.  Such failures were followed by pretraining phase I (single experimenter).  When the S now begged correctly from the single experimenter, Phase II was resumed.  Following the completion of each trial, each giver withdrew and independently recorded the subject’s behavior in her individual scoresheet.  Criterion again was defined as correct insertion toward the food-carrying giver in 4 of 5 successive trials. 

On any given session, a third experimenter operated a video camera.  The roles of givers and photographer were kept constant in a single session but rotated on successive sessions among the six keepers.  Within sessions, experimenters’ position (right or left) and their task (e.g., carrying a food tray or a flashlight; having a bucket over their shoulder or head), and tray-carrying shoulder (right or left) were pre-selected and written down in that day’s experimental plan.  All background and probe trials involving two experimenters followed the general procedure described earlier in relation to Pretraining Phase 2.

As with the elephants, a typical session consisted of 2 to 4 probe trials, interspersed among 6-8 background trials, with inter-trial intervals lasting 30 seconds or more.

Background trials consisted of the single person variation of Pretraining Phase 1 (Figure 3, B1), choice between food and flashlight (B2), and two distinct choices (B3, B4) between one giver who could see them and one who could not.

 

As was the case with the elephants, the chimpanzees were thoroughly familiar with the givers and photographer, but not with the two main experimenters who stood at all times some at least 3 m away and choreographed the actions, addressed problems as they arose, and independently recorded key events.  The chimpanzees were likewise familiar by sight with most of the objects used in this experiment, including the flashlight, trays of food, buckets, and screens; they had not however previously handled these objects.  The blindfolds were unfamiliar.

By the end of the third pre-training session (learning to insert one more finger through the wire mesh toward a single human), all our chimpanzees achieved the pre-determined criterion of inserting one or more fingers through or under the wire mesh in 4 of 5 consecutive trials.  By the end of four additional sessions, all subjects achieved the 80 percent criterion in the food tray vs. flashlight choice.

 

Table 3: Do chimpanzees know that people see?

Name	Food/fla-shlight	Back/ front	Lying		Side-ways	Screens	blindfolds	buckets	Total (4 probe trials)
Bubbles	9/10	8/8	4/4		3/4	3/4	3/4	2/4	11/16  (.69)
Joe Joe	15/15	9/11	3/4		4/4	2/4	2/4	2/4	10/16  (.63)
Beauty	9/10	6/7	4/4		4/4	3/4	4/4	2/4	13/16 (.81)
Barbara	15/16	7/8	4/4		3/4	3/4	3/4	3/4	12/16 (.75)
Abby	11/13	9/11	3/4		4/4	3/4	2/4	2/4	11/16  (.69)
Akati	6/6	8/8	4/4		2/4	4/4	3/4	1/4	10/16 (.63)
Tanya†	3/4	4/4	2/4		2/4	3/4	2/4	1/4	8/16 (.50)
Total	68/74 (.92)*	51/57 (.89)*	24/28 (.86)*		22/28 (.79)*	21/28 (.75)*	19/28 (.68)**	13/28 (.46)	75/112 (.67)
Young chimps††	.98	.84	NA		.50†††	.54	.49	.58
33-month- old humans††		.76				.52
42-month-old humans††		.94				.83

* p < .01; ** p < .05  (t-test, one-tailed, hypothetical mean: .5); † Subadult; †† Source: Povinelli and Eddy, 1996; ††† Data for the similar looking-over-the-shoulder condition

 

To our surprise, in some of the experimental tasks, our six adults (but not the one subadult) performed significantly above chance (Table 3). Our results cannot therefore rule out the notion that both chimpanzees and elephants partially understand that it makes more sense to request food from a human being whose face was directed toward them than from a human being whose face was not.  The results are likewise compatible with the alternative notion that the animals acquired this partial knowledge mechanically, through long association with human beings.  One thing does however emerge unequivocally from these data: Povinelli’s key, striking, implication—that adult chimpanzees perform at chance level in such tasks—is mistaken.  Thus, much more empirical work with chimpanzees, elephants, and other animals will have to be carried out before we can answer the age-old question:  Cognitively speaking, is Homo sapiens one species among many, a species apart, or somewhere in-between?

 

Insight in animals? 

Insight may perhaps be defined as either an understanding, in one’s mind, of the key elements of a situation, or the ability to mentally restructure such elements to solve a problem.    

To qualify as insight, the observed behavior must not be directly attributable to genetic programming, trial and error, or a combination of both.  Given the confusion that surrounds this subject, it may be worthwhile to illustrate here both sources of error. 

One example of genetically-determined behavior comes from Drosophila melanogaster. Even when raised in total isolation, upon first encountering a conspecific female, most mature males manifest the species-specific courtship behavior.  Moreover, these flies court even when the behavior is dysfunctional, for instance, they futilely and tirelessly court sex mosaics with male genitalia, provided only that these mosaics emit the species-specific sex-pheromone (Nissani, 1975).  Obviously, insight can only be attributed to behavioral sequences which sharply differ from such innate, inflexible courtship.

The task of ruling out trial and error learning is far more complicated.  One classical example of such learning is Morgan’s dog (1920).  The iron gate leading to Morgan’s house was held shut by a latch, but swung open by its own weight when the latch was lifted.  Morgan’s fox terrier could smoothly raise the latch with the back of his head, release the gate, and escape. But because the origins of this deceptively insightful sequence were carefully observed, it could be ascribed to “continued trial and failure, until a happy effect is reached,” not to “methodical planning.” 

Captive hamadryas baboon can similarly learn to use tools through trial and error.  When a pan with favorite treats was placed out of their reach, and an L-shaped rod was placed within reach, the baboons tried to reach the food directly but eventually gave up; occasionally directing their attention to the rod.  After about 11hr, one baboon chanced to have flipped the rod over the pan; upon retrieving the rod, the pan was brought within reach.  Gradually, and without an understanding of the task, he learned to use the rod to retrieve the pan (Beck, 1986).

Both dog and baboon examples forcefully recall Hobhouse’s  (1915) admonition that “what appears when complete to be an act of great sagacity, may, when its genesis is carefully traced, reveal itself as the outcome of accident, or perhaps of training.”

It may help to clarify the meaning of insight by relying on a fanciful exemplar.  In Aesop’s days, many people took it for granted that a crow could indeed solve this puzzle: dying of thirst, a crow figured out how to get water from the bottom of a pitcher by sequentially dropping pebbles into it.  What then are the key elements of this performance?

 

1.      The solution has neither been genetically wired nor acquired by trial and error. 

2.      Either this kind of problem is new to the animal, or the animal came up with a new solution to an old problem.

3.      The correct behavioral sequence has been arrived at suddenly and completely, not in a fumbling, piecemeal fashion.

4.      The behavior gives every indication of having been arrived at through a representation of its constituent elements in one’s mind.

5.      The action is carried out relatively smoothly, with all its constituent elements  purposefully subserving a single goal.

6.      The solution is retained after a single performance (Koestler, 1964).

7.      The behavior is flexible, the animal able to transfer its abstract rules across a range of physical stimuli (Mackintosh et al., 1985; e.g., the crow should be able to solve the problem with different pitchers, pebbles, or liquids).  

 

Anecdotal accounts of insight are many.  Romanes (1882), for instance, relates the story of a rogue elephant chasing a few hapless people up a tree.  Unable to topple the tree, and seeing a freshly-cut pile of timber nearby, the elephant removed all 36 pieces one at a time to the root of the tree, and piled them up in a regular, business-like manner.  He then placed his hind feet on the pile, raised the fore part of his body, and lifted his trunk (fortunately still failing to reach the tree sitters).

Similarly, Gary Alt describes many instances of tracking a black bear in fresh snow and inexplicably coming to the end of its trail.  Apparently, to give the slip to its pursuer, the bear backtracked precisely its own footsteps a short distance before leaping off the trail (cited in Shedd, 2000).

Although such stories raise the possibility of animal insight, appear to satisfy the seven criteria above, and may serve as the starting point for careful scientific studies, they cannot by themselves prove the existence of insight in animals—Morgan’s dog and Thorndike’s cats tell us that much.  Yet, we have little more to go on than these anecdotes.

In chimpanzees, one of the earliest insight experiments was carried out by Kohler (1925).  For instance, at least one of his chimpanzees could stack up three or four crates and climbed on them to obtain a banana that could not otherwise be reached.  But such behaviors do not meet the smoothness criterion:  years later, the conduct of his star performers still appeared “laughably inept” (Gould and Gould, 1994). It also appears that chimpanzees stack and climb crates even when no food is in sight, suggesting that, in the experimental conditions too, the behavior has been acquired unintentionally, not through a representational scheme.  Moreover, when a single crate was out of sight in a readily accessible corridor, the chimpanzees found it almost impossible to rely on memory, retrieve the crate, and get the fruit (Kohler, 1925), suggesting absence of mental representation of the key elements of the problem.  Yet another question mark comes from the work of Epstein and Medalie (1983) who observed similar behavior in a pigeon, again raising the possibility that Kohler’s chimpanzees stumbled upon the correct solution by unthinkingly trying everything, not by grasping the basic features of the situation.   

Surprisingly, the most carefully documented example of insight in animals comes from the common raven, not from the great apes.  When food is tied to a string which is then suspended from a branch, many species of birds reach down, grasp the string with their bill, pull it up, hold it with their foot, release it from their bill, reach down again to grasp the string again, and repeat this sequence until they reach the food.  Ravens have been shown to do this too, but in this case this behavior could involve insight (Heinrich, 1999, 2000).  Ravens, it should be noted, do not draw strings in the wild, thus suggesting the absence of genetic programming  (Hertz, cited in Duncker, 1945).  Similarly, Heinrich raised ravens in his own aviary, and could thus rule out prior trial-and-error learning.  Only some ravens solved this problem, they took sometimes hours to reach a solution, and successful ravens resorted to two different techniques of string drawing, thus again arguing against prior learning or direct genetic causation.  These experiments led Heinrich to the conclusion that string drawing “was a new behavior that was acquired without any learning trials.  They acted as though they had already done the trials.  The simplest hypothesis is that they had—in their heads” (Heinrich, 1999, p. 319).  But even in this carefully documented case, insight is just one possibility among many. While it is true, for example, that the entire sequence is extremely unlikely to have arisen by chance, its first two key elements—pulling a string and stepping on it—could very well be.  Once this two-step sequence is accidentally stumbled upon, the raven is partially rewarded by the closer proximity of the food source, and, like Morgan’s dog, is likely to repeat the behavior that brought about such a fortuitous outcome.     

 

RETRACTABLE CORD-PULLING IN ELEPHANTS

This and the next section describe the application of two new experimental procedures to elephants.  In both procedures, the two elephants and the setting were identical to the ones described earlier in this chapter. 

To apply the string-drawing paradigm to our elephants, we first had to determine whether elephants can spontaneously pull a rope when that rope is attached to a desirable object. We have seen that ravens and many other birds can do it, and we know that monkeys (Harlow and Settlage, 1934) and chimpanzees (Finch, 1941; Kohler, 1925) can pull rope too.  On the other hand, rats (Tolman, 1937) and probably cats and dogs, must be specially trained to do so (Hobhouse, 1915).  We subjected our two elephants to hundreds of string-pulling trials.  From the very start, and regardless of the kind of string, reward, and rope, the elephants invariably pulled in the reward immediately and unhesitatingly.

We next had to adapt the string-drawing task to elephants.  This was achieved by means of a stout retractable (“bungee”) cord securely tied to a post on one side and to a heavy rope on the other, with the retractable cord now spanning 2m, the stout rope spanning additional 5.5m, and with the non-attached side of the rope placed within reach of the elephant.  Along the rope, about .5 meter from the point of attachment to the retractable cord and 5m from the edge of the rope on the elephant’s side, a bagel or a whole corn cob was securely tied.  To prevent tearing the retractable cord (an easy task for an elephant!), parallel to this cord and attached to the same post and rope, a second, longer rope was used which only allowed the retractable rope to grow in length by approximately 2m. 

In this setup, the elephants could obtain the treat by pulling the rope with their trunk, anchoring it with another part of their body, pulling it again, and repeating this sequence, if necessary.  Once they reached the food, they had to go on holding the rope with mouth or foot until they could comfortably disentangle the food from the rope. 

In all, 10 daily sessions, ranging from 1 to 8min, were conducted with Winky, and 9 sessions, ranging from 2 to 15min, with Wanda.  In both cases, the sessions spanned a period of 6 weeks. 

Both elephants mastered the pulling-anchoring drawing sequence, with Winky settling on her mouth as an anchor and Wanda on her feet.  As in the case of Heinrich’s ravens, Winky tended to remain in one place while drawing the rope, while Wanda either walked on the rope or remained in one place while using her trunk and left foot.  

To our disappointment, our elephants’ performance fell short of the ravens’.  To begin with, preliminary and extensive rope-pulling observations showed that Wanda tended to pull on any rope (and not only retractable ones) by using both her trunk and feet, so using both trunk and foot with the retractable cord task would fail to satisfy the novelty criterion (Figure 5).  As we expected, she mastered the trunk-foot sequence on the first session.  Winky, on the other hand, only used her trunk to pull the rope in preliminary trials.  She thus satisfied the novelty criterion (barring the unlikely possibility that she faced such tasks in the past) and only obtained the treat on the fourth session. 

 

Figure 5. Wanda drawing a bagel attached to a heavy rope which in turn is attached to a retractable rope (not seen in photo), through a coordinated use of her trunk and feet.

 

Given Winky’s initial failures, we can surmise that this behavior is not genetically wired.  But in both cases, the solution was not sudden, complete, or smooth, nor was it retained after just one performance. Instead, it was fumblingly acquired over time.  For example, on the session immediately following the session at which she successfully used the trunk/mouth sequence, Winky, while clearly motivated, failed to obtain the food and resorted instead to a variety of unproductive strategies.  Likewise, the elephants’ retrieving strategy improved from session to session.  Moreover, both elephants failed a variety of transferability tests.  For instance, when a second rope, not tied to any treat, was attached to a post alongside the first rope in a more readily accessible location, Wanda spent at least 10min futilely stepping on that rope and trying to draw it in.  Neither elephant demonstrated flexibility—for example, when Wanda was faced with a situation where a mouth anchor would have been superior, she continued to rely on her feet. 

It is worth noting perhaps that the ten or so participants in this experiment were convinced that the elephants genuinely grasped the basic elements of this problem.  It was only later, while analyzing the video record, that I was led to reject their unanimous view.  At the moment, retractable cord pulling in our two elephants seems to have arisen through trial and error learning, not through insight.

 

DO ELEPHANTS KNOW WHEN TO SUCK OR BLOW?

The next experiment is partially based on the premise that each of our two captive elephants always maximizes her food intake at the expense of her companion.  Reports of altruism among wild elephants make a brief defense of this premise necessary.  Yet, months of observations categorically convinced us that, when it comes to food, their behavior is self-seeking.  For instance, before they enter the indoor compound for the night, their keepers distribute favorite treats such as coconuts and carrots throughout their four-room indoor enclosure.  In such cases, the dominant Winky successfully takes hold of all the treats.  Likewise, we have never seen them willingly offer food to each other.

The experiment itself capitalized on our two elephants’ ability to obtain objects from an inflexible tube by expertly placing their trunk tightly over one of the tube’s openings and either sucking the object toward them or blowing it out and retrieving it when it falls to the ground from the other side of the tube.

During the first session, a short, 5cm in diameter, PVC tube (in Figure 6, the short tube below the elephant’s left ear) was horizontally clamped to the cable which separated the elephants’ outdoor quarters from the keepers’ courtyard.  A small piece of bagel, or a 1.4 cm³ (4g) white sugar cube, was placed on either side of the tube, at least 2 cm away from the edge so that it could not be directly retrieved by the trunk.  One elephant and then the other served as a subject, while a keeper enticed the other away.  The results for the first 10 trials are given in Table 4, suggesting a random retrieving strategy. 

 

Table 4.  Initial results of tube experiments

Elephant	Sucked to obtain morsel	Blew to obtain morsel	Total
Wanda	5	5	10
Winky	4	6	10

 

Figure 6. A blowing contraption. A treat is placed at the open edge of the short, horizontal lower tube. If sucked, the treat is blocked by the researcher’s fingers and does not reach the elephant, causing the trial to end unsuccessfully. On the other hand, blowing readily sends the morsel toward the elephant.

 

A few days later, both elephants were allowed to remain near the tube, with the dominant Winky standing directly in front of it, while Wanda stood some 2m away, along the cables.  In this setup, Winky could invariably obtain the treat by sucking it from either side of the tube.  She could also get it by blowing it toward the side farther away from Wanda, but she could lose it by blowing it toward the other side.  In 12 out of 13 trials, and in contrast to her behavior on the preceding session, she sucked the morsel from one side of the tube or the other.  Only in the seventh trial did she blow, in Wanda’s direction, lightly; the morsel thus fell close to Winky and she was able to obtain it too.  This test could not be carried out with the subordinate Wanda.

Next, a researcher held a long tube horizontally, with one side directly facing the elephant and the other facing the courtyard (to which the elephant had no access).  In this setup, only sucking  availed them of the morsel.  Both elephants adjusted immediately to this new situation, grabbing the open side available to them and tightening their grip. Winky sucked in six out of six trials and Wanda in seven out of eight.

A similar performance was repeated on the next session, with similar results: Winky sucked in all five trials and Wanda sucked in four out of five.

To overcome the dominance effect and to make competition possible, we now clamped a 1.5m tube horizontally to the cables.  The elephants were directed to stand on either side of the tube, facing the courtyard.  In this case, blowing would give the treat to one’s companion while sucking would land it in one’s trunk.  Wanda was more adept at this tusk; not once in her 24 attempts, nor Winky in her 6, did either elephant resort to blowing.  In each and every case, both elephants sucked the morsel.

The elephants were now separated, and placed successively near the short horizontal tube, alone.  Now, Wanda blew in two out of five trials and sucked in three, while Winky blew in three and sucked in two, again suggesting a random retrieving strategy when alone.

In the next session, the animals again stood on either side of the 1.5m tube.  And again, in over 15min and 35 trials of rather heated (and delightful) competition, they sucked the treat. 

Although it seemed at this point that both elephants understood when to blow the treats and when to suck them, we wanted to reassure ourselves that they did not adopt the correct strategy by trial and error.  To that end, in the next session we presented the animals with a contraption which only rewarded blowing behavior (Figure 6).   

When first presented with this contraption, both elephants resorted to both sucking and blowing. In her last 10 trials, however, Wanda only resorted to blowing and applied just enough pressure to have the treat land right by her front feet.  Winky’s record was not as impressive; she used a combination of sucking and blowing for trials 1-6, and only resorted to blowing on trials 7-10.  Nor was she as adept as Wanda at controlling the landing location. 

Having just faced a situation where only blowing was rewarded, the two subjects were now brought together to compete along the two sides of the 1.5m tube, where only suction will do. In 15 minutes of steady, intense, competition, and over 40 trials, neither elephant blew once. 

The next session began with the blowing contraption above (Figure 6).  Again, this task proved more difficult than all the others.  In trials 1-13, Wanda used a combination of sucking and blowing (of which only blowing was effective), while in trials 14-20 she only blew.  In 10 trials, Winky sucked on the first and then switched to blowing-only strategy.  With each elephant, these trials were immediately followed by exposure to a hand-held long tube (in which sucking yielded the reward while blowing landed it out of reach).  In six consecutive trials, Wanda only sucked while Winky sucked in 13 out of 16 trials.

Although the results appear consistent with the supposition “that elephants must be credited with the ability to anticipate what will come of certain actions (= ideation)” (Rensch, 1957), they cannot  rule out prior exposure to similar tasks, sampling errors, and other explanations. 

 

Conclusion

All past and present claims to the contrary, science cannot yet resolve the question of animal consciousness.  For over a century, scientists believed that evolutionary theory required at least some traces of consciousness in animals.  Empirical studies (e.g., on insight, language acquisition, mirror self-recognition, and deception) lent further support to this viewpoint.  In retrospect, however, neither theory nor experiments proved decisive.

A similar historical pattern seems to apply to the more recent experiments of Povinelli and his colleagues.  For a few years, these experiments provided strong support to the view that animals lacked a theory of mind.  But the similar experiments recounted in this chapter seem to have once again restored greater uncertainty to this field.  Given these recent negative results, and given moreover the picture which emerges from the field as a whole, we ought perhaps to suspend judgment at the moment and instead find new ways of tackling this old problem.

The search for insight in animals likewise confronts us with more questions than answers.  As we have seen, the few controlled experiments in this area lend themselves to alternative interpretations.  Likewise, our own search for insight in elephants failed to resolve this question. Like ravens and other avian species, our two elephants mastered the string-drawing task—but in a manner that appeared more consistent with trial and error learning than with insight.  In contrast, our elephants seemed to have possessed from the start a basic grasp of the tube task; but here too, trial and error interpretations cannot be ruled out.

Decades of innovative experimentation with many species of animals besides the great apes may be needed before we know whether or not we are the only conscious beings on earth.  One way or the other, such knowledge is sure to alter our conception of what it is like to be an animal—or a human being.

 

 

Acknowledgments 

This work was made possible through the generous support of Wayne State University and the Detroit Zoological Institute.  I likewise thank Scott Carter, Michelle Seldon-Koch, and Ron Kagan for their interest in this project.  For camaraderie, advice, cooperation, and friendship, I am deeply indebted to Donna Hoefler-Nissani, Bettie McIntire, Maria Manuguerra-Crews, Rick Wendt, Chris O'Donnell, Erin McEntee, Erin Porth, Jennifer Goode, Kelly Wilson, Kim Van Spronsen, Marilynn Crowley, Mary Mutty, Megan Brunelle, Melanie Hiam, Patrick Smyth, and Patti Rowe.  For patience and first-class editing, Gisela Kaplan and Lesley J. Rogers have my heartfelt thanks.

  

References 

Aelian (3rd century, BC) On the Characteristics of Animals, Harvard University Press, Cambridge (English translation by A. F. Scholfield, 1958).

Alexander, S., 2000, The Astonishing Elephant, Random House, New York.

Baron-Cohen, S., Tager-Flusberg, H., and Cohen, D. J., eds., 2000, Understanding Other Minds, second edition, Oxford University Press, Oxford.

Beck, B. B., 1986, Tradition and social learning in animals, in: Animal Intelligence:Insights into the Animal Mind, R. J. Hoage and L. Goldman, eds., Smithsonian Institution Press, Washington, DC.

Blashford-Snell, J. and Lenska, R., 1997, Mammoth Hunt, HarperCollins, London.

Byrne, R. W. 2001, Social and technical forms of primate intelligence, in: Tree of Origin, F. B. M. de Waal, ed., Harvard University Press, Cambridge, MA.

Byrne, R. W. and Whiten, A., eds., 1988, Machiavellian Intelligence, Clarendon Press, Oxford.

Carrington, R., 1958, Elephants, Chatto and Windus, London.

Carruthers, P., 1998, Natural theories of consciousness, Eur. J. Phil. 6(2):203–222. _http://psyche.cs.monash.edu.au/v4/psyche-4-03-carruthers.html_.

Carruthers, P. and Smith, P. K., 1996, Theories of Theories of Mind, Cambridge University Press, Cambridge.

Chevalier-Skolnikoff, S. and Liska, J., 1993, Tool use by wild and captive elephants, Anim. Behav. 46:209–219.

Denis, A., 1963, On Safari, Collins, London.

Duncker, K., 1945, On problem solving, Psychol. Monog. 58:1–113.

Eltringham, K., 1982, Elephants, Blandford, Dorset.

Epstein, R. and Medalie, S. D., 1983, The spontaneous use of a tool by a pigeon, Behav. Anal. Lett. 3:241–247.

Finch, G., 1941, The solution of patterned string problems by chimpanzees, J. Comp.

Psychol. 32:83–90.

Fouts, Roger., 1997, Next of Kin, William Morrow, New York.

Gallup, G. G., Jr., 1998, Can Animals Empathize? Yes, Sci. Am. http://www.sciam.com/ 1998/1198intelligence/1198gallup.html.

Gomez, J. C., 1998, Assessing theory of mind with nonverbal procedures: Problems with training methods and an alternative “key” procedure, Behav. Brain Sci. 21: 119–120.

Gordon, J. A., 1966, Elephants do think, Af. Wild Life 20:75–79.

Gould, J. and Gould, C. G., 1994, The Animal Mind, Scientific American Library, New York.

Griffin, D. R., 2001, Animal Minds, second edition, University of Chicago Press, Chicago, IL.

Groning, K. and Saller, M., 1999, Elephants: A Cultural and Natural History, Konemann, Cologne.

Hare, B., Call, J., Agnetta, B., and Tomasello, M., 2000, Chimpanzees know what conspecifics do and do not see, Anim. Behav. 59:771–785.

Harlow, H. F. and Harlow, M. K., 1965, The affectional systems, in: Behavior of Non-

Human Primates, A. M. Schrier, H. F. Harlow, and F. Stollnitz, eds., Academic Press, New York.

Harlow, F. and Settlage, P. H., 1934, Comparative behavior of primates. VII. Capacity of monkeys to solve patterned strings tests, J. Comp. Psychol. 18:423–435.

Hart, B. L. and Hart, L. A., 1994, Fly switching by Asian elephants: Tool use to control parasites, Anim. Behav. 48:35–45.

Heinrich, B., 1999,The Mind of the Raven, Harper Collins, New York.

Heinrich, B., 2000, Testing insight in ravens, in: The Evolution of Cognition, C. Heyes, and L. Huber, eds., MIT Press, Cambridge, MA.

Heyes, C. M., 1998, Theory of mind in nonhuman primates, Behav. Brain Sci. 21:101–148.

Hobhouse, L. T., 1915, Mind in Evolution, Macmillan, London.

Hooff, J. A. R. A. M., van., 2000, Primate ethology and socioecology in the Netherlands, in: Primate Encounters, S. C. Strum, and L. M. Pedigan, eds.,University of Chicago Press, Chicago.

Koestler, A., 1964, Act of Creation, Macmillan, New York.

Kohler, W., 1925, The Mentality of Apes, Harcourt, New York.

Linden, E., 1999, The Parrot’s Lament, Dutton, New York.

Mackintosh, N. J., Wilson, B., and Boakes, R. A., 1985, Differences in mechanisms of intelligence among vertebrates, in: Animal Intelligence, L. Weiskrantz, ed., Clarendon Press, Oxford.

Macphail, E. M., 1982, Brain and Intelligence in Vertebrates, Clarendon Press, Oxford.

Markowitz, H., 1982, Behavioral Enrichment in the Zoo, Van Nostrand, New York.

McComb, K., Moss, C., Sayialel, S., and Baker, L., 2000, Unusually extensive networks of vocal recognition in African elephants, Anim. Behav. 59:1103–1109.

Morgan, C. L., 1920, Ani. Behav. Second edition, Edward Arnold, London.

Nissani, M., 1975, A new behavioral bioassay for an analysis of sexual attraction and pheromones in insects, J. Exp. Zool. 192:271–275.

Poole, J., 1997, Elephants, Voyageur Press, Stillwater, MN.

Povinelli, D. J., 1989, Failure to find self-recognition in Asian elephants (Elephas maximus) in contrast to their use of mirror cues to discover hidden food, J. Comp. Psychol. 103:122–131.

Povinelli, D. J., 1998, Can Animals Empathize? Maybe not. Sci. Am. http://www.sciam. com/1998/1198intelligence/1198povinelli.html.

Povinelli, D. J. and Eddy, T. J., 1996, What young chimpanzees know about seeing, Monogr. Soc. Res. Child Dev. 61(No. 2, Serial No. 247).

Povinelli, D. J. and O’Neill, D. K., 2000, Do chimpanzees use their gestures to instruct each other? in: Understanding Other Minds, Second edition, S. Baron- Cohen, H. Tager-Flusberg, and D. J. Cohen, eds., Oxford University Press, Oxford.

Povinelli, D. J. and Giambrone, S., 2001, Reasoning about beliefs: A human specialization? Child Dev. 72:691–695.

Povinelli, D. J., Nelson, K., and Boysen, S., 1990, Inferences about guessing and knowing by chimpanzees, J. Comp. Psychol. 104:203–210.

Povinelli, D. J., Bering, J., and Giambrone, S., 2000, Towards a science of other minds: Escaping the argument by analogy, Cogn. Sci. 24:167–201.

Premack, D., 1988, “Does the chimpanzee have a theory of mind?” revisited, in: Machiavellian Intelligence, R. W. Byrne and A. Whiten, eds., Clarendon Press, Oxford.

Reaux, J. E., Theall, L. A., and Povinelli, D. J., 1999, A longitudinal investigation of chimpanzees’ understanding of visual perception, Child Dev. 70:275–290.

Reiss, D. and Marino, L. 2001, Mirror self-recognition in the bottlenose dolphin: A case of cognitive convergence, Proc. Natl. Acad. Sci. USA 98:5937–5942.

Rensch, B., 1956, Increase of learning capability with increase of brain-size, Am. Nat. 90:81–95.

Rensch, B., 1957, The intelligence of elephants, Sci. Am. 196:44–49.

Romanes, G.J., 1882, Animal Intelligence, K. Paul, London.

Russon, A. E., Bard, K. A., and Parker, S. T., eds., 1996, Reaching into Thought, Cambridge University Press, Cambridge.

Sanderson, G. P., 1912, Thirteen Years among the Wild Beasts of India, J. Grant, Edinburgh.

Sanderson, I. T., 1962, The Dynasty of Abu, Knopf, New York.

Shand, M., 1995, Queen of the Elephants, Jonathan Cape, London.

Shedd, W., 2000, Owls aren’t Wise and Bats aren’t Blind, Harmony, New York.

Smitha, B., Thakar, J., and Watve, M., 1999, Theory of mind and appreciation of geometry by the small green bee eater (Merops orientalis), Curr. Sci. 76:574–577.

Spinage, C. A., 1994, Elephants, Poyser, London.

Stevens, V. J., 1978, Basic operant research in the zoo, in: Behavior of Captive Wild Animals, H. Markowitz and V. J. Stevens, eds., Nelson-Hall, Chicago, IL.

Thomas, R. K., 1986, Reasoning and language in chimpanzees, in: Animal Intelligence: Insights into the Animal Mind, R. J. Hoage and L. Goldman, eds., Smithsonian Institution Press, Washington, DC.

Thompson, W. R. and Melzack, R., 1956, Early environment, Sci. Am. 194:38–42.

Thorpe, W. H., 1956, Learning and Instinct in Animals, Harvard University Press, Cambridge.

Tolman, E. C., 1937, The acquisition of string-pulling by rats—conditioned response or sign-gestalt? Psychol. Rev. 44:195–211.

Tuma, R. S., 2000, Thinking like a chimp, HMS Beagle, Issue 90, http://news.bmn. com/hmsbeagle/90/notes/feature2-illustrator.

Uller, C. and Nichols, S., 2000, Goal attribution in chimpanzees, Cognition 76:B26–B34.

Voelkl, B. and Huber, L., 2000, True imitation in marmosets, Anim. Behav. 60:195–202.

Waal, F. B. M. de, ed., 2001, Introduction, Tree of Origin, Harvard University Press, Cambridge, MA.

Wickler, W. and Seibt, U., 1997, Aimed object-throwing by a wild African elephant in an interspecific encounter, Ethology 103:365–368.

Williams, J. H., 1950, Elephant Bill, Doubleday, Garden City, NY.

Wynne, C., 1999, Do animals think? The case against the animal mind, Psychol. Today 32(6):50–53.

Zentall, T. R., 2000, Animal Intelligence, in: Handbook of Intelligence, R. J. Sternberg, ed., Cambridge University Press, Cambridge.

Elephant Corner                      Moti Nissani's Homepage