Kanzi: The Ape at the Brink of the Human Mind

by Sue Savage-Rumbaugh

  Initially, both chimps were extremely reluctant to share, as we had expected. One of us would sit between the chimps with a bowl of food, from which we would take a piece, break it in two, and give half to Austin and half to Sherman. They immediately turned their backs and ate the morsel in private, the way chimps do. Little by little—through a great deal of cajoling and patience on the part of the teachers—we broke down the basic chimpanzee tendency to avoid the sharing of plant food and replaced it with distinctly humanlike food dispersal behaviors. Although Austin was initially nervous about taking food out of Sherman’s hand, because of the difference in dominance, eventually both he and Sherman shared food as humans would. When food was available, Austin would approach Sherman and expectantly request his share, even when no humans seemed to be present (we were watching out of view). Sherman would sometimes look away as if trying to ignore Austin, but Austin would persist and Sherman’s face would begin to assume a guilty expression, as though he were aware of breaking a pact between them. Sherman would then hand over almost half of the food to Austin. The sharing was nearly always done quietly and calmly, and Sherman and Austin even began to look at each other while they ate.

  It was an extraordinary sight, and many primatologists who visited the center were quite shocked to see such behavior. In a review article of our work, Carolyn Ristau and Donald Robbins, of Rockefeller University, suggested that we must somehow have coerced Sherman and Austin into food sharing, because it was so unnatural a behavior for chimpanzees. There was no coercion at all. Indeed, the chimps came to enjoy it as one of their most favored activities.

  We were ready—we thought—to set up food sharing in a communicative context. After all, Sherman and Austin had the communicative skills of requesting and comprehending, and they were willing to share. We therefore set up a table with two trays of food, one for Austin and the other for Sherman. We then encouraged one chimp to request food from the other. They didn’t. When Austin was making a request, Sherman was just as likely to be running playfully around the room as attending to Austin. Even when one of the chimps responded by picking up the requested food, he had to be encouraged by much gesturing on the teacher’s part to fulfill the transaction. We realized that, left to themselves, Sherman and Austin were unlikely to use symbols for an orderly exercise of food sharing.

  One of the key attributes of communication is simply paying attention to the other individual’s actions—the role of being a good listener. This hadn’t happened as Sherman and Austin sat around the food-sharing table (or rather ran around it), and so we decided to have them in separate rooms, joined by an observation window. Each chimp had a keyboard on which to make a request, which would show up on the responder’s keyboard. With surprising ease, Sherman and Austin learned to attend to each other, attend to the requests made, and proffer the specified food. Being spatially separated in a rather formal way had facilitated the learning procedure for them. Having learned the benefits of sharing in this way, they later were able to transfer the process to the less formal arrangement of sitting around a table. They used their keyboards to make requests, and they responded with alacrity. Sherman and Austin frequently brought their food-sharing trays and table to us, looking expectantly. If we said, okay guys, we’ll do some food sharing, they became wildly excited and ran around the room hooting, hugging, and tumbling in a general pandemonium of delight.

  The food-sharing ritual began to be so much a part of their lives that they happily began also to take part in the preparation of the food table itself. This entailed carefully preparing two portions of each type of food and placing each portion group side by side in one of the compartments of the table. They eagerly vied to be the one who got to help prepare the table before each bout. Soon, food-sharing sessions could take place with no human present. Sherman and Austin simply sat themselves down and proceeded to have a meal, talking about which food they desired next and sharing all the portions. Occasionally they made innovations on the method of sharing that we had attempted to enculturate. For example, we had encouraged them to share each distinct food and made this easy by placing two portions of each food side by side on the food table. However, one chimp would sometimes eat both portions of a single food in response to a request. Then, given the next request, he would give both portions of that food to his partner. Of course, we cannot know what they thought as they innovated such alterations in the food-sharing concept, but we could not help but feel at times as if they were trying to say, “I know that I slighted you last time, but that was a food I could not resist … please accept two portions of this food in return.” So much for coercion.

  Sherman and Austin’s ability spontaneously to communicate requests to each other with symbols advanced ape-language research in important ways. Not only was this the first documented interindividual communication, but it also incorporated degrees of comprehension not seen in other projects. Nevertheless, Sherman and Austin’s vocabulary was small, simply because we had concentrated on other skills. In order to broaden communicative possibilities, we needed to build bigger vocabularies, including more food items and extending significantly into nonfood items.

  We were delighted by the ease with which the chimps incorporated symbols for new food items, because it suggested that they were finally recognizing that each food had a unique correspondence to a particular symbol. To determine if this was so, I took a new food into the chimps’ room along with their other foods as we were preparing to do food sharing. I carefully made no attempt to show them a new symbol for this new food. A number of unassigned symbols (or “unknown words”) were present on the keyboard, though Sherman and Austin did not utilize them. Upon spying the new food, Sherman reached out for it and I suggested that he should ask for it. He quickly turned, scrutinized the keyboard, and deliberately, but with no fanfare, indicated one of the unassigned lexigrams as the symbol for the food. Austin watched attentively and subsequently used the symbol Sherman had selected to indicate that he would like to try the new food also. This unexpected skill encouraged us to believe that Sherman and Austin understood that a unique correspondence existed between each food item and a specific symbol. Moreover, they sometimes pointed back and forth between the symbol and the food item. They had named new foods and apparently agreed on coordinated use of the selected symbols.

  When we moved onto nonfood items, however, we seemed to be back to square one. They learned to request a few objects such as blankets and keys, but when we asked the chimps to name these same objects they were unable to do so reliably—just as they had once been unable to name foods. Repeated practice produced little improvement, so we sought another way. When we attempted to introduce tools such as a wrench and lever, they began to confuse the tools and to string their tool symbols together, or to use the symbols interchangeably when requesting a specific tool. The challenge was to make object names more salient to the chimps—which essentially meant linking them to some activity that would make manifest the particular functional properties of each tool.

  The procedure we developed was as follows. We built six food sites in the chimps’ rooms, each of which required a specific tool to gain access to the site—key, money, stick, straw, sponge, and wrench. We put food in one site at a time and introduced one tool at a time. Ultimately, we randomly selected a tool site to be baited, then displayed the tray of tools and encouraged the chimps to survey the situation and decide which tool they needed and how to request it. Both chimps seemed fascinated by this task and were eager to work with the tools, even long after they were tired of using them to obtain food from the tool sites.

  Although the chimps learned to request the tools with relative ease, two problems emerged as we proceeded. While constraining at the time, each problem provided important glimpses into the animals’ minds. One problem appeared when we moved from the request task to the naming task. Although Austin very quickly began to learn to name each of the tools when they were held up to view, Sherman initially had great difficulty. He
was able to request all six tools accurately, but for some reason he simply could not name them unless he intended to use them. It was a very frustrating time, both for Sherman and for me, as I would hold up a tool and he would select the wrong symbol. I would correct him, No, this stick, for example, and he would then respond stick. But even if I held up the stick the very next time he would answer incorrectly again. Sherman simply didn’t seem to understand what was required of him.

  This incomprehension continued for eighty-eight trials. I was then distracted by having to answer the telephone. On the eighty-ninth trial, with no change in procedure on my part, Sherman got it right. And out of the next 104 trials he made just four errors. It was apparent that he suddenly understood what I was asking him to do, and then did it. I’ve witnessed many times this kind of sudden recognition in chimps, and it has an uncanny human aspect to it—like the figurative light bulb going on in your head.

  The second interesting problem we encountered involved functional errors with requesting and naming tools. Both Sherman and Austin experienced these problems, but in different ways. For instance, Austin frequently made wrench/key errors, using the symbol for one when the task required the other. Sherman had a syringe/key problem, often mixing up his request for them. In neither case was this because the symbols were similar for the pairs of tools. It seemed that the chimps were organizing their experience with tools in terms of their function. In other words, for Austin, the turning action of both key and wrench was most salient, while for Sherman, the insertion action of syringe and key was overriding. It was as if the chimps used the symbols as verbs, not nouns as we had intended. In any case, the different problems experienced by Sherman and Austin highlighted for me the individual nature of the chimpanzee mind.

  Once Sherman and Austin had learned to request tools from a teacher, and to name tools reliably, we wanted to switch back again to communication between the chimps. We asked: What if only one chimp had access to tools, but none of the tool sites in his room had food in them? And what if the other chimp saw the sites in his room filled with food, but he had no access to tools? Would the chimps recognize that they could gain access to the food if they cooperated and communicated with each other? Would the chimp who had access to the tools watch the actions of the other at the keyboard, and willingly hand objects back and forth? Would they understand that they could make symbolic requests of one another for objects?

  Sherman and Austin would need to do five things to achieve this: (1) attend to one another; (2) coordinate their communication; (3) exchange roles of tool-requester and tool-provider; (4) comprehend the function and intentionality of their communications; and (5) share their access to tools and the food obtained through tool use. If they achieved this, Sherman and Austin would have taken an important step toward using symbols much as humans do.

  We approached this challenge using two rooms separated by a large window, as we had initially with the food-sharing task. On the first trial, all the tools were located in Austin’s room. I then placed a food item in the wrench site in Sherman’s room. Sherman immediately went to his keyboard and indicated wrench, but he gazed at me, not at Austin. I indicated I hadn’t any tools or access to them, so Sherman then looked directly at Austin, who produced the requested tool. The chimps alternated roles as tool-requester and tool-provider, and very quickly seemed to understand what was required for the task: specifically, who had the tools and who had the information needed to select the appropriate one. The close attention to one another and to the series of actions that occurred was clear indication that the chimps comprehended what was going on and were not merely engaged in a mechanical chain of events.

  As is often the case, we learned almost as much from Sherman and Austin’s mistakes as from their successes. For example, on one trial Sherman mistakenly requested a key when a wrench was appropriate for the task, and he watched as Austin began to look over the toolkit in response to the request. Austin picked up the key, and Sherman looked surprised, turned to look at the keyboard, which still showed the key request he’d made, and realized his mistake. He rushed to the keyboard and corrected himself by tapping on the wrench symbol to draw Austin’s attention to the changed request. Austin looked up, saw what Sherman was doing, dropped the key, and took the wrench to the window to give to Sherman. Such a sequence is indicative of intentionality and comprehension on the chimps’ part, and cannot be dismissed as the rote result of conditioned response training.

  Although these results were extremely pertinent to the ape-language debate, most researchers in the field paid little attention; the supposed primacy of syntax still held them in its thrall. There was, however, an assault from the behaviorist camp, specifically from Robert Epstein, Robert Lanza, and B. F. Skinner, the last being the most prominent figure in behaviorism at the time. In a 1980 issue of Science, they published a description of a choreographed sequence of behaviors in two pigeons, Jack and Jill. This sequence simulated communication that, superficially, looked similar to that achieved by Sherman and Austin.

  Jack and Jill (both males) were housed in adjoining boxes with a transparent wall between them. Jack pecked a key labeled “what color?” Jill then looked through a curtain to the back of his box to see which of three colored lights was lit. He would then peck the appropriate one of three buttons, labeled R (for red), G (green), and Y (yellow); the pecking action illuminated the button. Jack then pecked a key labeled “thank you,” which resulted in food being dispensed to Jill. Finally, after seeing which key Jill had illuminated, Jack pecked the equivalent one in his box, and was rewarded with a food item. The authors explained that, although the sequence of events might look like “sustained and natural conversation,” it was in fact the result of strict conditioning procedures. They then went on to say that “A similar account may be given of the Rumbaugh procedure.”1

  It was at first humorous, but later somewhat frightening, to see an experiment of this nature given so much prominence. In reply, Duane and I wrote: “The description of this work seeks to parody ours and deceptively leads all but the most well-informed reader to conclude that, in fact, identical concepts were learned by chimpanzee and pigeon.”2 Whereas the pigeons had to be taught each step in the sequence and were not required to be aware of what the other pigeon was doing, the communicative behavior between Sherman and Austin emerged from components of true communication, and they achieved it on the first trial, with no training; they also had to be aware of their partner’s actions and of the fact that they were indeed communicating with one another.

  One issue that undoubtedly had provoked the behaviorists’ attack was my conclusion that Sherman and Austin were exhibiting conscious intentionality during their communication—a clear red flag to those who believe behavior should simply be viewed as responses to external stimuli. As a result I became labeled a cognitive psychologist—one who believes that there is more to behavior than stimulus-response, and that animals may actually think about what they are doing. Sherman and Austin’s behavior makes a strong argument for this view. I came to see clearly that intentionality must be accepted if we are to understand language in a functional sense, and tracing the evolutionary emergence of conscious intentionality is crucial to the appearance and understanding of language as it is used by Homo sapiens.

  We had taught Sherman and Austin key elements of communication—request, naming, and comprehension. Once these were in place, other aspects of communication emerged spontaneously. The chimps began to pay close attention to each other’s communications; they engaged each other before delivering their message; they gestured to emphasize or clarify messages; they took turns. None of these behaviors, all of which enhance communication, was taught by us. Sherman and Austin developed them spontaneously.

  Most important of all the communicative behaviors that emerged spontaneously was that of indication, or announcement of intended future action. At the start of one trial session in which I was going to ask Sherman to give me objects I requested, I ap
parently took too long to begin. To my surprise, Sherman said straw and handed it to me. Then he said blanket and handed that, too. Next he fingered the wrench, as if he were thinking of it, then said wrench and pointed to the tool while looking back and forth between me and the wrench, checking to see if I had noticed the tool to which he was pointing. Initially, my impulse was to discourage Sherman, but soon I realized the importance of the behavior. Announcing an intended action is one of the earliest symbol-use skills to appear in human children, and here it was emerging spontaneously with Sherman. The same skill appeared in Austin, too, and it wasn’t always in trial situations. For instance, on one occasion Austin announced at the keyboard he was going to make a funny face, and then stuck his tongue out and pulled his lower lip down over his chin. Comical though the context might have been, it was linguistically important.

  The more I observed such behaviors, as Sherman and Austin sat at their tables independent of teachers, communicating their needs and sharing their food, the more their general behavior took on a human countenance. This was not the result of training but of use and development of communicative skills in a social setting. It is true to say that we really didn’t see language competence in Sherman and Austin until food sharing fully developed, with its exchange of information and subsequent coordination of behavior. It wasn’t a complex language, not a language with syntax. It was more a culture language, a complex set of behaviors that was the way the chimps’ lives were lived in the laboratory. It made one think of Homo sapiens without sophisticated spoken language—intelligent, sensitive creatures, able to communicate and coordinate their behavior in a collective subsistence effort. I am not suggesting that what we have with Sherman and Austin is a precise model for a stage in human evolution. But it is easy to see how elements of social/communicative abilities, which are evidently present in our closest relative, the chimpanzee, could have brought about a new social/subsistence regime in our early human ancestors.