Kanzi: The Ape at the Brink of the Human Mind

by Sue Savage-Rumbaugh

  Because researchers saw that apes built vocabularies so easily by associative learning of individual symbols, there was little thought about what the symbols might mean to the apes. Researchers simply leapt ahead to look for signs of syntax in word combinations. In doing so, they were assuming that when an ape learns a word, it attaches the same linguistic representational aspect to it that children do when they learn a word. The central issue is this: Does the ape know that the symbol can stand for an object that may be absent?

  There are four components to linguistic representation: (1) an arbitrary symbol that stands for, and can take the place of, a real object, event, person, action, or relationship; (2) stored knowledge regarding the actions, objects, and relationships relating to that symbol; (3) the intentional use of that symbol to convey stored knowledge to another individual who has similar real-world experiences and has related them to the same symbol system; and (4) the appropriate decoding of and response to the symbol of the recipient. A word comes to represent an object, separate from it in place and time.

  It was not clear that Washoe or Lana, or any of the other signing apes I had encountered, possessed full linguistic representation, or referential, abilities. They could use a symbol, in what was often a contextually cued situation, to request an object or activity. But they were usually unable to decode the symbol when it was used by a human in a simple request. As a result, the complexity of communication achieved by these apes was no greater than the basic communicative level of chimpanzees who were not language trained. This absence of full comprehension in language-trained apes was, I felt strongly, a more fundamental criticism of ape-language research than the absence of syntax, as demonstrated by Terrace. Cooperative comprehension is fundamental to language, and two-way communication that does not reflect comprehension is not language, no matter what other attributes it may possess.

  I had reached these conclusions as the storm over ape-language research was about to break, and assembled them in a paper titled “Do Apes Use Language?” in American Scientist, with. Sarah Boysen and Duane Rumbaugh as co-authors. Because of extraordinary delay in publication, the paper came out two months after Terrace’s Science paper, but four months prior to the Clever Hans Conference. We concluded the paper with the following: “Experimenters must stop looking for superficial similarities between apes and children and must instead investigate the cognitive competencies that underlie symbolic processes.”13 That was precisely where the project with Sherman and Austin would take us.

  3

  Talking to Each Other

  Sherman and Austin were two and a half and one and a half years old, respectively, when the Animal Model Project commenced, in June of 1975. The primary goal of the project was to elucidate the processes of language acquisition in apes and compare them with the phenomenon of spontaneous language acquisition in human children. This goal encompassed practical and theoretical issues. First, it continued and extended the effort to develop language-training techniques that might help severely mentally retarded children. Duane had initiated that endeavor at the Yerkes Regional Primate Research Center in 1970, with the Lana project. Second, it asked, in what sense can a species other than Homo sapiens develop language?

  The unease I had experienced over the strong claims for language competence in the chimpanzees I was most familiar with—Washoe, Lucy, and Lana—encouraged me to guide the investigation with Sherman and Austin in a very different direction. The goal of most researchers in the field had been to determine whether apes have language, in much the same way as you might determine whether they have a thumb or a stomach. As we saw earlier, it was expected that if apes do have language, its presence would be revealed by the animals’ innate syntactical competence, a putatively genetically determined ability to order the symbols in multiword utterances.

  My goal was at once to be more modest and more ambitious than seeking signs of syntax: I planned to focus on words, not sentences. More specifically, I was interested in the animals’ ability to comprehend and communicate. I had no ambitions to instill in Sherman and Austin an impressively extensive vocabulary. Nor would I spend time encouraging multisymbol utterances. Instead, I was reaching beyond these staples of ape-language research, seeking to touch the essence of language: the ability to tell another individual something he or she did not already know. I wanted Sherman and Austin to use symbols referentially with each other, in true, humanlike communication.

  The journey toward that goal turned out to be longer and more arduous than I had expected, and at every step of the way I encountered problems, primarily because of unsubstantiated assumptions I made about what Sherman and Austin would be able to do as we progressed. I came away from the experience—one I felt I had jointly shared with Sherman and Austin—with a better understanding of the nature of language and its acquisition. The results of the project also advanced ape-language research in a fundamentally conceptual manner.

  The unfolding of the Animal Model Project happened to parallel in time the rising fomentation over the validity of ape-language research, which I described in Chapter 2. By the time Herbert Terrace published his influential Science paper, in November 1979, and the Clever Hans Conference of the New York Academy of Sciences had taken place, in May 1980, Sherman and Austin had achieved a level of language competence—in the sense of true symbolic communication—that far surpassed that of any of the apes so frequently cited by both proponents and opponents of ape-language research. I had been working with them for five years at that point. Ironically, because I had eschewed syntax as a goal in my project, Sherman and Austin received not even as much as a footnote in the debate.

  From the very beginning, the two chimps were very different from each other. Sherman was, and has remained, the physically bigger of the two. Partly as a result of this, Sherman has the ability to be the dominant individual whenever he chooses. Just as striking, however, are their personality differences. Sarah Boysen once aptly described Sherman as the football player and Austin as the stamp collector. Sherman has always been extremely active and reactive and always in the middle of things when some commotion is going on. When he was young he was constantly hurling himself into my lap and leaping onto my head—I suffered a perpetual sore neck as a result. And when he was older I had to tape thumbtacks onto the back of my hands to discourage his rough, albeit friendly, play-bites. When Sherman gets mad he erupts volcanically, with impressive displays, rushing around, puffing his hair out, banging things, and even slappings or pushing me. But it is all over very quickly. You know where you are with him.

  Austin is much quieter, gentler, and slower to react. But when he does get angry he is much more dangerous than Sherman, because he often seems to slip beyond his own control. On occasion he has pulled me off my feet and thrown me to the floor so forcefully that it is difficult to get up afterward. I often feel sorry for Austin, because at two months of age he apparently couldn’t digest his mother’s milk and was failing to thrive. He was therefore separated from her and as a result displayed a behavior common to young chimps who suffer early separation: Whenever he was distressed he rocked rhythmically from side to side while holding his blanket. Fortunately, as an adult he has outgrown this behavior. I have no idea why, but when Austin sees a picture or a doll-like figure of a human or chimpanzee infant, he tries to destroy it, and becomes agitated if he is thwarted.

  In the years I’ve known the two chimps, I’ve never seen anything that would indicate a significant difference in their basic intelligences. Nevertheless, it is interesting to see how their personalities are reflected in the way they approach tasks. For instance, Sherman always prefers to communicate with gestures and other nonsymbolic means if he can. Austin is a keyboard individual. And Sherman is much better at tasks that require participation, while Austin excels at tasks requiring close observation and attention. These differences, and differences in the mistakes each makes with symbol-use from time to time, have persuaded me that we shouldn’t talk facilely about the ch
impanzee mind. Surely, Sherman and Austin share something of a basic chimpanzee view of the world, but their individual experience of it is undoubtedly very different.

  All of the lexigrams used by Sherman and Austin, just as they were arranged on their original keyboard.

  The communication system we used with Sherman and Austin was a modified version of the computerized keyboard that Duane had developed for the Lana project. Symbols that represent objects or activities are arbitrary geometric forms, based on the Yerkish language that was invented by Ernst von Glaserfeld. Their keyboard eventually grew to 92 symbols.

  Early on, we activated only a few symbols on the keyboard; as their vocabulary slowly grew we added more and more. When pressed, the key lit up, and was very obvious. To avoid the problem of the chimps simply learning the position of a key rather than the symbol on it, the symbols were randomly reassigned new positions after each keyboard use. The chimps had to search carefully for the symbols they wanted to use. Even though Sherman and Austin have never shown any extensive comprehension of human speech, we decided from the beginning to use speech as well as symbols to get them to do things. They were evidently very sensitive to affect in the voice, and responded appropriately to emotions it conveyed.

  The beginning of the journey toward true communication started with what everyone had assumed was a simple task—teaching the names of objects by association. For Washoe, Sarah, and Lana, this teaching of “names” had shared the important common procedural elements of holding up an object, an apple for example, and then encouraging and/or helping the ape to make the sign or to select the correct symbol. Once the ape could do this on its own, additional signs or symbols were introduced to build vocabulary. It was assumed that teaching words in this way was a rather simple process, if somewhat time-consuming. It was also assumed that such training basically depended on forming conditioned stimulus response associations between items and symbols. The real test of language, it was said, would come when one looked to see if apes could put together sentences from the words they had learned.

  While this sort of word learning may lead to sentence production in adults who are learning a second language, it is not at all the way children go about learning their first language. However, at that time, very little research on how children actually learned words had been done. It seemed reasonable, therefore, to start by teaching apes word-symbol associations. It had also seemed to work.

  Unfortunately, the process did not work with Sherman and Austin. With only one symbol available to be selected, Sherman and Austin learned relatively easily, as would be expected. If we held up a banana, they selected the banana symbol, as that was the only one available. With two, a lot more practice was required, but eventually they succeeded. Beyond two, however, the chimps became hesitant and failed to improve, no matter how much practice we gave them. The chimps were clearly puzzled as to how to proceed. I was puzzled, too, because Washoe and Lana had encountered little difficulty in learning associations between objects and symbols or signs.

  I watched videotapes of training sessions to see if I could figure out what we were doing wrong. The problem was that I had not stopped to ask why a chimpanzee who had had no previous language training would know that symbols encoded anything. I expected the chimp to make an association between the object and the symbol—and thus know the object’s name. Instead, as I learned from close scrutiny of the videotapes, Sherman and Austin were paying attention to the symbol and my subsequent action—that is, whether or not I gave them a reward.

  No wonder we were all confused—humans and chimps alike. We all had different views about what was going on. As teachers holding up objects, we assumed that the object we were showing the chimp would serve as the “stimulus object” since it preceded the response. The chimps, however, were assuming that the symbol-key they selected served as the “stimulus” for us to give them food. They looked for a link between the symbol they depressed and whether or not we elected to give them food—and if so, what sort of food. They paid no attention to the “stimulus item” we displayed. Indeed, they paid no attention even when we attempted to cue them by pointing back and forth repeatedly between the stimulus item and the correct lexigram.

  In hindsight, this seemed obvious. Why should they care what we were showing them? They cared more about whether we were tickling them, chasing them, feeding them, and so forth. Once it was apparent that the chimps were attentive to the consequences that followed their symbol production, rather than to the stimulus conditions that preceded it, it suddenly dawned on me that the Gardners had essentially taught Washoe “names” by giving her the object or action after she made the sign. Thus, if she signed tickle, they tickled her, if she signed banana, they gave her a banana. Similarly, when Lana depressed Please machine give piece of banana, she got a banana; when she said Tim tickle Lana, Tim did so—at least he did so at first while Lana was initially learning. However, this fact had not been pointed out in articles. Instead, it had been emphasized that Washoe and Lana had learned the names of things by being shown the objects they were to name.

  After it became clear that the contingencies which followed symbol use made all the difference, we switched our teaching approach. Now we showed the chimp a banana and when he selected the banana symbol, he was permitted to eat the banana.

  This sort of transaction took the form, “I show you X, you select the lexigram that goes with X, and then I give you X,” and has been characterized as a request task. It has the superficial appearance of language, in that an ape who can make many different signs or select many different symbols when shown various objects seems to be able to ask for a wide variety of things it wants, such as foods, tickling, and so forth. What the chimp does, however, should not be confused with knowing names. Even though the ape appears to “name” something that it is being shown, it is really selecting the symbol because it anticipates that it will receive the object. This is true whether the teacher is holding up a banana, preparing to groom the ape, or getting ready to open up a door or box. In each case, the chimp is selecting the symbol or making the sign on the basis of what it anticipates will happen afterward.

  The switch in training procedure produced rapid learning. Within a few days Sherman and Austin were correctly requesting a number of different foods. Now symbol-object pairings were coming as easily to them as had been described to be the case for Washoe and Lana. However, if Sherman and Austin were asked to name something when all expectancy of being able to receive that item was removed, they began to evidence confusion once again.

  This was the first realization on my part that holding an object up and having a chimp produce the right sign or symbol did not mean that the ape knew the name of the object. It only meant that under certain conditions, the chimp knew what to do in order to obtain the object. In contrast, once a child knows words it seems able to do quite a bit more with them, even before it begins to form sentences. Certainly a child does not limit its word use to occasions set by the parent who is holding up items and asking for their names, or waiting to open doors or tickle the child until it says the appropriate word.

  From these early experiences I learned three important lessons. First, chimpanzees do not necessarily learn object-symbol associations easily through practice and repetition, as had been supposed. Second, I did not have to worry about cuing Sherman and Austin into giving erroneously positive results; cuing, if it was happening, was clearly ineffective as a means of teaching. Third, the simple relationship that was assumed in the object-symbol association was not simple at all; such associations could be of several types, each with a different implication for inferred language competence.

  Also from these experiences flowed a means of communicating wishes, which Sherman and Austin unexpectedly developed on their own. Our technique required either me or another teacher to put some item of food in a dispenser linked to the keyboard. The chimp’s task was to hit the key that corresponded to the food, and thus receive it. Very soon the chimps
began to pay close attention to the food I was about to select, and would become impatient if they thought I was too slow. Before long they began to hit a key before I selected the food. Were they trying to control my behavior, getting me to select foods they liked? It seemed so. They were very attentive, and if they hit a symbol for a food they knew was in the refrigerator, they would stare at the refrigerator after hitting the key, apparently waiting for me to respond appropriately. As they always chose favorite foods, like M&M’s or juice, and never water or chow, the behavior did have the appearance of intention about it.

  Initially I discouraged this behavior, but then I realized it was indeed communicative and as such was important in our overall goal. We therefore incorporated it into our teaching regime. Once the chimps learned to use specific symbols for specific foods in this way, their progress improved dramatically. We were pleased with the chimps’ headway, and added more and more symbols, including some nonfood items, such as “tickle” and “out.” The value of communication as a motivator to the chimps’ learning was plain, and I often reflect on the benefit of abandoning strict experimental procedure when an opportunity for an unexpected breakthrough offers itself, as it did in this case.

  The ability to request is just one of three elements that must combine to produce true communication between individuals. The other two are (1) the ability to name objects and (2) a comprehension of symbols as referents of objects. We were unsure of the precise learning path we needed to take to have Sherman and Austin communicate symbolically with each other, but we knew they would need to be able to use their symbols without expecting some particularly beneficial contingency to follow each time. After all, if Sherman was to do something as simple as to ask Austin to give him a banana, Austin would have to understand that “banana” meant the specific fruit, even though he was giving one rather than getting one. Separating “names” of things from the contingencies associated with the learning of those names proved to be more difficult than we had anticipated.