Kanzi: The Ape at the Brink of the Human Mind

by Sue Savage-Rumbaugh

  I met Duane Rumbaugh, of Georgia State University, at the 1974 meeting of the International Primatological Society, in Kyoto, Japan. By that time, Duane was two years into his ape-language project with the chimpanzee Lana, at the Yerkes Regional Primate Research Center of Emory University in Adanta. In Kyoto I gave a paper on some work I had done earlier with Lucy, so Duane knew I’d had an interest in ape-language research. He also knew I was skeptical of some of the rich interpretation that was being made of apes’ language competence. Nevertheless, he invited me to a symposium he was organizing on ape language, to be held at the Southeastern Psychological Association meeting, which was to take place in Atlanta the following year. The paper I gave in Atlanta compared my experiences with Lucy and Pancho, and stated clearly that I saw no communicative advantage being bestowed on Lucy by her ability to sign. I was therefore clearly stating my position as an “unbeliever.”

  Six months later Duane called me in Oklahoma with an offer of a postdoctoral position at Georgia State University. This position would permit me to study at the Yerkes Center. I accepted at once, because Yerkes is internationally recognized as being preeminent in primate studies—and it had bonobos. To Duane’s disappointment, I refused to work on the Lana project, and concentrated instead on studying bonobo behavior and comparing it with that of common chimps. Gradually, however, I was drawn into the language project, pardy because of various problems that had developed when Lana was moved to a new, improved facility; Duane asked me to help sort them out.

  Initiated in 1971 with the explicit aim of developing systems with which to teach language competence to severely mentally retarded children, the project had by far the most sophisticated means of symbol manipulation of all ape-language projects. The symbols were arbitrary geometric forms, which were displayed on a computerized keyboard. Lana activated a symbol by touching a key, which then lit up and was projected on a screen. As with all the ape-language projects at this point, Lana’s communication was principally about food. Unlike what happened in other projects, however, she was required to build a sentence in order to obtain food or some kind of play activity: for example, please machine give piece of orange and please machine make music. After four years Lana had an extensive vocabulary with the system, some one hundred symbols, and she had produced a small but significant number of novel sentences. Most of her utterances, however, were of the sort just mentioned. My first impression of Lana was that she knew what she was talking about. I’m sure that the sentence structure she had been taught to produce on the keyboard encouraged that belief. After all, linguists insisted that word order—syntax—was the key to language, and Lana was producing such order. However, I soon formed the same unease with Lana that had surfaced at Oklahoma concerning the chimps’ ability to comprehend the words they used. Lana was producing sentences that were comprehensible to me as English construction, but, I wondered, did Lana understand the meaning of the words she was using? When I asked her to do things (such as give me an object), using the same vocabulary of words she employed in her sentences, she could not respond reliably. In an attempt to respond, she would produce inappropriate stock sentences and sometimes nonsentences. She seemed to be searching around for something that would satisfy me, but she wasn’t sure what it was. Lana, like Washoe and the other chimps, appeared to be productively competent (using words to request things) but not receptively competent (comprehending what was said to her).

  I began to wonder whether there might be several aspects to what we call a word. We assume with human children that the learning of a word includes its comprehension. It seemed to me that productive competence and receptive competence might be discrete cognitive abilities and, in apes at least, might have to be taught separately. Perhaps ape-language researchers had made the mistake of assuming that, as with human infants, once an ape learned a word it also understood its meaning. Perhaps making the leap to search for signs of syntax was not only premature, but also irrelevant to the core of language.

  With these kinds of questions still somewhat inchoate in my mind, I began voicing my concerns to Duane. At first he was unconvinced by my suggestions that he and other researchers were making assumptions about the language competence of apes, and were falling victim to a rich interpretation of the apes’ utterances. But he listened. We agreed that I would assist him with a new project, with two young male chimpanzees, Sherman and Austin, with a slightly different focus. Beginning in 1976, I established a much closer physical proximity with the apes, interacting with them in a social, preschool-like setting. This would emphasize communicative needs rather than promoting teaching efficiency. The distinction, I believed, was fundamentally important. Further, unlike all previous ape-language projects, this one would not have as its goal the production of word combinations or sentences. I wasn’t in search of the linguists’ holy grail. I was going to focus on words: What does a word mean to a chimpanzee, and how can we find out?

  Just as I was embarking on the Sherman and Austin project, in 1976, a storm was gathering over ape-language research as a whole. The storm clouds rolled in from two directions: From one, linguists poured scorn on the validity of the research and questioned the competence of the researchers. From the other, a prominent ape-language researcher—Herbert Terrace—declared that he and his fellow researchers had been mistaken in believing that apes had acquired language. By the end of the decade ape-language research was completely engulfed in the resulting turbulence, and as a field of study was all but destroyed.

  For some time during the mid-1970s, Thomas Sebeok, a linguist at Indiana University, had been expressing strongly negative views on ape-language research. And in May 1980 he organized a conference under the auspices of the New York Academy of Sciences, which made his position brutally clear. The conference was called “The Clever Hans Phenomenon: Communication with Horses, Whales, Apes, and People.” Clever Hans was a horse that performed apparently amazing arithmetical feats in nineteenth-century music halls. The putative equine genius would tap its hoof the appropriate number of times in response to a puzzle asked of him by his owner, Wilhelm Von Osten. Unbeknownst to the innocent Von Osten, however, a barely perceptible movement of his head when Hans reached the right number cued the horse to stop tapping. The phenomenon, besides revealing how extremely sensitive animals can be to body language, showed how easy it is for a human to cue an animal unwittingly, thus shaping its behavior. Useful in the music hall, cuing can disrupt the objectivity of research on animal behavior.

  Ape-language research is prey to the Clever Hans effect, or worse, opined Sebeok. The results of the research are to be explained as a result of “unconscious bias, self-deception, magic, and circus performance,”3 he said six months before the New York conference. Conference participants included experts on the tricks of animal performers and a magician, the Amazing Randi. It was “a celebration of deception in all its varieties,” a reporter for Science noted. “It was amazing that any ape-language researchers should even have considered stepping into such a lions’ den.”4 Many such researchers, having accepted a preliminary invitation, dropped out when the tenor of the gathering became obvious. Step into it Duane and I did, however, mainly at his insistence. He wanted to demonstrate that not only did we have courage, but we also had good data. We would stand on our data, our research methods, and our convictions.

  It didn’t matter how good our data were, however. The atmosphere was so very negative that the issue was effectively prejudged. For instance, the first speaker, Heini Hediger, of the University of Zurich, declared it “amazing” that anyone would seriously consider that nonhuman animals might display elements of human language. In his presentation, Sebeok suggested that funding for the work should be halted, and perhaps dispersed to more worthy causes, like cancer research. There was even a move, fortunately thwarted, to have the conference vote for a ban on the research. This was reminiscent of a ban on the study of the origin of language, instigated by the Linguistic Society of Paris, in 1866.

&
nbsp; Sebeok, with Jean Umiker-Sebeok, had circulated a manuscript that was highly critical of all ape-language research and contained much that was inflammatory. “Thus we find the ape ‘language’ researchers replete with personalities who believe themselves to be acting according to the most exalted motivations and sophisticated manners, but in reality have involved themselves in the most rudimentary circus-like performances,” they wrote, a remark that Science noted was “hardly best calculated to brush up their colleagues the right way.” Science was right. We were further brushed the wrong way with the following: “The principal investigators themselves, of course, require success in order to obtain continued financial support for the project, as well as personal recognition and career advancement… .”5

  At a press conference at the end of the meeting, Sebeok expressed his view most stridently of all: “In my opinion, the alleged language experiments with apes divide into three groups: one, outright fraud; two, self-deception; three, those conducted by [Herbert] Terrace.” Pressed by reporters, Sebeok declined to present any evidence to substantiate his accusation of fraud. Whatever motives lay behind such remarks, and behind the initiative for a vote to ban the research, they hardly seemed to be based in science. A “Talk of the Town” column in the 26 May 1980 issue of The New Yorker began a commentary on the conference this way: “Can scientists speak to apes? Can apes speak to scientists? Can scientists speak to scientists?” Its answer was that “the jury is still out on all three questions.” The answer to the third question, of course, is: Not when the discourse has clearly gone beyond the realms of science.

  In my presentation I responded directly to the Sebeoks’ lengthy critique, embodied in their circulated manuscript. For instance, they suggested that because ape-language researchers are critical of one another’s work, all the work should be dismissed as inconclusive. This struck me then, and still does, as a curious way of looking at science. There are always disagreements among scientists in new, fast-moving fields. And if the conflicts in my field seemed unusual in any way, one can see that this stems from the perhaps inevitable way in which results were sensationalized in the popular press. The Sebeoks asserted that because the apes in the studies work better with some humans than others, this indicates that some people are better at cuing than others. We pointed out that anyone who has ever worked with animals, particularly ones that are sensitive to social interaction, knows that the best work flows from a trusting relationship. To invoke cuing as the sole reason for such differences is to display an ignorance of human/animal interaction.

  The Sebeoks suggested that circus trainers, magicians, and others practiced in the art of purposefully manipulating animal behavior to create the illusion of humanlike activities should be allowed to investigate ape-language projects. “Circus performers and gurus are not qualified to evaluate the serious efforts of scientists,” we said in our reply. “To turn to psychics and magicians as a cure for unintentional cues by ape researchers is anti-science and anti-intellectual.” I would be the last to deny that cuing has happened in some of the ape-language projects. Chimpanzees are extremely smart animals, and can pick up on the slightest traces of approval or disapproval in a human’s facial expression or posture. But it is facile at best to dismiss all ape-language research as beset by cuing, or to assume that scientists are unable to eliminate cuing possibilities by scientific means.

  Perhaps Sebeok’s position is best illustrated by his remark to a reporter for the British science magazine New Scientist. Asked what kind of evidence would convince him of the validity of some kind of cognitive substrate for language in apes, he replied: “Facts do not convince me. Theories do.” Sebeok apparently overlooked the fact that ape language is helping us formulate new theories of language acquisition that have an evolutionary basis.

  Besides Duane and myself, the only other ape-language researcher to attend the Clever Hans Conference was Herbert Terrace. Unlike us, however, he was there to concur with Sebeok: The languagelike utterances of apes were a form of Clever Hans effect, he said. “Nim had fooled me,” he later wrote.6

  Project Nim started in late 1973, shortly after Terrace collected the infant chimp named Nim from the Institute for Primate Studies, in Oklahoma. Terrace’s initial plan was to demonstrate conclusively what he considered other ape-language researchers had concluded only anecdotally. For instance, the Gardners stated: “The most significant results of Project Washoe were those based on comparisons between Washoe and children, as in the use of order in early sentences.”7 The demonstration of syntax, the holy grail of linguists and, by default, ape-language researchers, was also Terrace’s goal: “knowing a human language entails knowing a grammar,” he said.8 Terrace’s hope was to demonstrate that Nim knew grammar, by videotaping and analyzing as large a proportion of the chimp’s utterances as possible over a period of years, as he learned and used sign language.

  More than sixty sign-language teachers worked in Terrace’s project in its four years, and Nim learned just as other chimps had. He eventually had a vocabulary of more than 125 signs, and produced many double-word combinations. As the project proceeded, Terrace and his colleagues grew more and more confident that Nim’s utterances displayed syntax. For instance, when Nim used more in a two-word combination, it appeared in the first position 85 percent of the time, as in more banana and more drink. Similar appropriate word placement occurred with give (as in give apple) and with transitive verbs (such as hug, tickle, give) when combined with me and Nim. “The more I analyzed Nim’s combinations, the more certain I felt we were on solid ground in concluding that they were grammatical and that they were comparable to the first sentences of a child,” said Terrace.9

  In 1977, after Nim returned to Oklahoma, Terrace and his colleagues had time to scrutinize the hundreds of hours of videotape they had amassed. The tape had captured twenty thousand of Nim’s utterances, half of which were combinations of two words or more. The more Terrace studied these utterances, however, the more his previous certainty eroded. What had seemed like grammatical conversational utterances in the intimacy of the teaching room dissolved into zombielike imitations on the video screen. Terrace had been troubled early on by the lack of increase in the length of Nim’s utterances, something that is a natural part of language-learning in children. Nim did produce some long utterances (such as give orange me give eat orange me eat orange give me eat orange give me you), but they were usually nonsense, and contained no more information than short utterances. However, the insight that to a great extent Nim was imitating his teachers, rather than spontaneously conversing, was the most telling blow. By the time he was four years old, Nim fully or partially repeated what had just been said to him. And for more than three-quarters of the time, Nim’s utterances were preceded by a teacher’s utterance.

  Terrace came to realize that Nim’s use of signs focused on requests for food or play activities. He also realized that Nim frequently threw in “free words,” such as Nim, me, hurry, give, and more, which often added an air of “sentence structure” to an otherwise simple “food” or “tickle” request.

  As a result of the analysis, Terrace published a paper in the 23 November 1979 issue of Science, titled, “Can an Ape Create a Sentence?” with Laura Perito, Richard Saunders, and Tom Bever as co-authors. The answer to the title’s question was a resounding No! “Objective analyses of our data, as well as those obtained from other studies, yielded no evidence of an ape’s ability to use a grammar,” they concluded.10 “Sequences of signs, produced by Nim and by other apes, may resemble the first multiword sequences produced by children. But unless alternative explanations of an ape’s combination of signs are eliminated, in particular the habit of partially imitating teachers’ recent utterances, there is no reason to regard an ape’s utterance as a sentence.”11

  Terrace clearly did a thorough piece of work in scrutinizing the reality of Nim’s utterances, recognizing that what was inferred as syntax was in fact an illusion of the teaching method and the assumptions
of the research. (However, Terrace was incorrect to say that Nim fooled him; in reality, Terrace fooled himself, because the system effectively taught Nim to imitate.)

  The combined effect of Sebeok’s Clever Hans Conference and Terrace’s Science paper was, however, to instigate an extremely rapid and violent swing of the pendulum. Ape-language research went from being a field of perceived intellectual excitement and public acclaim to one that, at best, should be viewed askance. Suddenly, it became extremely difficult to have research papers reviewed, let alone published. And funding for most of the major projects virtually dried up. Fortunately, our work had just received a five-year renewal of funding as the storm broke, which allowed us to ride out the storm.

  As I watched the credibility of the field crumble, I worried about a complete loss of objectivity about the research. With Duane I wrote a note to The Psychological Record, pointing out the danger of overreaction to Terrace’s work: “If … the general scientific community now concludes (a) that apes are just like other animals, (b) that all animals are very different from men, and (c) that problems only arise when we attempt to study animal cognitive processes, Terrace’s personal step forward will become a large step backward in man’s struggle to understand the emergence of human uniqueness.”12

  As important as Terrace’s work had been, I recognized that it was not fundamental to real progress in ape-language research. It was, of course, important to demonstrate that multiword utterances were often the result of imitation. But this still left the field firmly in the hands of linguists, to whom syntax is sacred. By the time of the Clever Hans Conference and the publication of Terrace’s Science paper, I knew that this was misguided, particularly as applied to ape-language research. As a result of my earlier concerns over apes’ comprehension, and my more recent project with Sherman and Austin, I knew that we had to shift our attention away from sentences and turn it to words.