Witnessing this display of emotion, I hadn’t the heart to keep them apart any longer, and opened the connecting door. Kanzi leapt into Matata’s arms, and they screamed and hugged for fully five minutes, and then stepped back to gaze at each other in happiness. They then played like children, laughing all the time as only bonobos can. The laughter of a bonobo sounds like the laughter of someone who has laughed so hard that he has run out of air but can’t stop laughing anyway. Eventually, exhausted, Kanzi and Matata quieted down and began tenderly grooming each other. How could anyone have contemplated not allowing them to be together? If we’ve lost him back to the world of bonobos, so be it, I thought to myself.
Just then, Kanzi gestured for the keyboard and indicated open. I opened the cage door and went in. Kanzi climbed on my shoulders and gestured for me to go through the door. I was shocked. Surely he will change his mind as soon as we leave Matata, I thought. But he didn’t. We got Matata some bananas and juice, she vocalized with contentment, and Kanzi looked at me and said, childside. Off we went, to see who was still there at the end of the day, something Kanzi often liked to do. All our concern had been for nothing. Kanzi had elected to be with Matata and also with Homo sapiens; he was to negotiate two worlds.
For seventeen months we kept a complete record of Kanzi’s utterances, either directly on the computer when he was indoors, or manually while outdoors. By the end of the period, Kanzi had a vocabulary of about fifty symbols. Within a month of separation from Matata, he was already producing combinations of words, and he continued to do so throughout the seventeen-month period of our first report. Other apes had also produced combinations, but Kanzi’s were different in that they reflected a competence born of comprehension, rather than a need to form longer symbol groups to answer questions, as had been the case with other apes. Nearly all (more than 90 percent) of Kanzi’s multiword utterances were spontaneous; that is, they were not responses to teachers’ requests or imitations of teachers’ utterances. By contrast, Terrace found exactly the opposite pattern in his chimpanzee, Nim. Three-quarters of Nim’s multiword utterances followed something a teacher said. Spontaneity of utterances, rather than their frequency, is a critical aspect of language use, for it is the spontaneous utterances that tell us something about what another party is thinking. Utterances that are prodded out of an ape in response to questions such as Who Nim hug—while Nim is already hugging someone named Laura—do not really provide the listener with new information, meaning information that is not already self-evident in the situation.
Kanzi, however, formed spontaneous utterances such as Matata grouproom tickle to ask that his mother be permitted to join in a game of tickle in the group room. This happened on one occasion shortly after Matata had vocalized to him. Since Matata typically did not join us in the grouproom play sessions, this request on Kanzi’s part completely surprised us. Perhaps Kanzi knew from sounds that Matata was making that she wished to join us and thus sought to tell us about her wishes. We cannot know for certain; however, we do know that we did not prod this utterance from Kanzi, nor would we have considered bringing Matata into the group room were it not for Kanzi’s communication. Many of Kanzi’s multiword utterances had this character of novelty and functioned to suggest completely new actions and alternatives to our normal ways of doing things.
It was also the case that as Kanzi added more elements to his utterances, the information content increased. For instance, one of Kanzi’s multiword combinations was ice water go (with go indicated by a gesture), by which he was asking someone to get some ice water for him. A combination such as play me Nim play is typical of Nim’s utterances, which contain a great deal of redundancy. Nim may have been more loquacious than Kanzi, but what Kanzi was producing was more like language.
Of Kanzi’s three-item utterances, the most interesting—and significant—were those in which he indicated someone other than himself as the agent or recipient of an action. Most of his three-item combinations involved the initiation of play, such as grab, chase, and tickle. Some of these games involved Kanzi directly, but others were intended for his teachers. For instance, Kanzi might indicate grab chase at the keyboard, and then take one person’s hand and push it toward a second person: the chaser and the chasee. Statements of this sort were Kanzi’s inventions, as none of us suggested we play with each other, leaving Kanzi as spectator. Compared with food requests or requests to be tickled, where the chimpanzee is always the recipient of the action, statements initiating action between two other individuals is complex. As my colleagues and I observed in a paper describing these utterances: “Clearly, prior to the emergence of syntax must be the emergence of the concept that one can request that A act on B, where the speaker is neither A nor B.”3 The fact that Kanzi mastered this betrays a real sense of self and of others.
We tested Kanzi’s language competence in many ways throughout the seventeen-month period, and one significant feature that emerged was that his comprehension of symbols consistently preceded his use of them. This is the human pattern of language acquisition, and the opposite of how we had concluded from our previous work that apes learned. Moreover, we formed a steadily growing conviction that Kanzi also had a good comprehension of spoken words. This is a difficult issue, one fraught with strongly held beliefs and little scientific data. Many investigators have assumed that apes easily come to understand English, a view held by Charles Darwin. Referring to other animals, too, he wrote in 1871: “[T]hat which distinguishes man from the lower animals is not the understanding of articulate sounds, for, as every one knows, dogs understand many words and sentences.”4
So strong is the impression that Sherman and Austin can comprehend spoken English—and, perhaps too, so strong is the urge to believe it—that visitors to the Language Research Center have often refused to accept our explanation of how they appear to do so. Even when we point out that careful tests, which eliminate contextual information, have demonstrated they do not really have the level of comprehension that appearances suggest, these visitors still choose to believe the chimps can understand the spoken word.
Apes are so aware of, and competent in interpreting nonverbal aspects of, communication that they often infer a speaker’s intent while not truly understanding the words. If you watch a soap opera on television with the sound turned off, you will gain a sense of the kind and breadth of information that can be conveyed without language proper—it is rather extensive. The ability to “read” information into the situation from a variety of sources, including gestures, glances, actions, intonation, and knowledge of similar previous situations, is highly developed in apes. It is this capacity to read the “meaning” inherent in a given situation that often convinces people that apes are understanding language, for we humans focus so exclusively on language, once it is learned, that we are often unaware of the other channels through which we are gaining information.
When we conducted tests on Austin and Sherman that eliminated contextual information, forcing them to respond to the auditory stimulus alone, their answers were clearly random guesses. In these tests, Sherman and Austin were presented with words that they knew well as lexigrams and were asked to select a matching photo whenever they heard the word. They did well whenever they saw a lexigram, but not when they heard the speech. We could therefore be certain that they understood what they were being asked to do. Often, when they heard only a word, they even gestured to the keyboard, asking me to “say” the word in lexigrams so that they could find the correct picture.
These findings are in accord with observations made of chimps reared in home environments earlier this century, by Louise and Winthrop Kellogg in the 1930s and Catherine and Keith Hayes in the 1940s. The chimps, Viki and Gua, displayed limited comprehension under test conditions, and those words they did understand were closely tied to a specific context. The Kelloggs spent many hours teaching Viki the names of various body parts, but with no success, and the Hayeses drilled Gua with the names of familiar objects, again without
success.
More recently, Roger Fouts has suggested that Washoe, Ali, and a few other apes understand at least three or four words. Unfortunately the data are few and the number of overall correct responses, even to a word that is thought to be understood, are extremely low. Even though the data suggest that there is a statistically significant difference between words that are recognizable and those that are not, these words are understood on such a small percentage of occasions that most oral communications made in real time would be missed. Francine Patterson also suggests comprehension of speech for Koko the gorilla, but she lacks the test data to rule out the impact of contextual information. It may be that other apes show some speech comprehension, but none of them seem to have made the leap into language comprehension that Kanzi was on the verge of at this time.
We first detected what seemed like spoken word comprehension when Kanzi was one and a half years old. We began to notice that often, when we talked about lights, Kanzi would run to the switch on the wall and flip it on and off. Later he simply looked at the light switch on hearing the word, apparently visually forming the pairing. How was this different from what we and others had seen in other apes? Was Kanzi not responding to some undetermined contextual cues? At first we thought this must be the case. However, there were differences. Sherman and Austin responded to speech and contextual information when someone was talking to them or saying something of relevance to or about them. They often responded appropriately to queries like, “Do you want to go outdoors?” “Please don’t push the TV set,” and “Stop tickling so hard.” Kanzi also responded in this manner, but in addition, he seemed to be “listening in” on conversations that had nothing to do with him. For example, I once asked another caretaker if “someone had left the lights on last night” and then noted that Kanzi was gazing at the light switch, even though neither I nor anyone else had done so, and in fact we had not even been discussing the particular light that he was noting.
It was what seemed to be this uncanny ability to “listen in” on conversations that were not directed to him, and indeed ones that he was not supposed to pay attention to, that first began to convince me that Kanzi was processing sounds in a manner that I had not experienced in other apes—even in those who had been reared in human homes. I believed, however, that this was not possible. The prevailing theory of the time was that of Phillip Lieberman, who asserted that vocal and auditory capacities co-evolved in all creatures. According to Lieberman, the auditory system, composed of the inner ear and the auditory cortex, was capable of processing sounds in an efficient manner only if they corresponded to the types of signals that a species made with its vocal tract. An animal’s auditory comprehension was, so to speak, tuned to the capacity of its throat by co-evolution of the communicative process. This theory seemed very reasonable and it explained, for me, why Sherman and Austin had difficulty understanding words even though we spoke to them constantly.
As time passed, Kanzi appeared able to understand more and more spoken words. Finally, I could no longer ignore the fact that what Kanzi was doing really did not fit within the standard theory and that it was not like the kind of apparent comprehension or partial comprehension that I had encountered in Sherman and Austin or Lucy and Washoe. Kanzi had become so proficient that not only was he listening in on conversations that were not directed to him, he was beginning to translate some of the words we said on the keyboard. For example, one afternoon someone approached me to tell me about a fight that had occurred between Sherman and Austin. Kanzi listened, then went to his keyboard and said Austin and gestured go, pointing in the direction where Austin and Sherman were housed. Another time, someone mentioned in passing that Kanzi had learned how to turn the light on and off. He immediately went to his keyboard and hit the key for light and then gestured toward the light switch. In response, we had to do what many parents do when they don’t want their children to overhear; we began to spell out some words around Kanzi. Kanzi, like most children, recognized that we were doing this to avoid his listening and simply began to listen all the harder.
At this point we decided to test Kanzi’s speech comprehension carefully and with controls, just as we had previously done with Sherman and Austin. We didn’t train Kanzi for this test, nor did we offer rewards for correct answers. We simply showed him an array of three photographs and lexigrams and then spoke the word to indicate which one he was to give us. We performed three testing sessions, in which our requests alternated between spoken English and lexigrams. The tests included thirty-five different items, used in 180 trials in English and 180 with lexigrams. Kanzi scored 95 percent correct on the lexigram trials and almost as well, 93 percent, on the English trials. We were able to determine that Kanzi understood 150 spoken words at the end of the seventeen-month period. By comparison, Sherman and Austin were correct 98 percent of the time they were shown a symbol and asked to find the corresponding photo from an array of two or three pictures. However, when we used spoken English to ask for a particular photo, they appeared to guess. They hesitated, tried to look for cues, and asked us to turn the keyboard on. When all else failed, they would simply select the nearest or most interesting picture. Their overall score was near 30 percent, as one would expect if they were simply selecting the photos randomly.
Kanzi’s ability to comprehend spoken English is part of the explanation of the larger phenomenon, namely, his humanlike acquisition of language capacity. Comprehension aided the emergence of the productive skill in Kanzi, as it does in humans, central to which is the understanding that words and lexigrams are referential and can be used as a mode of symbolic communication.
This discovery, another first for Kanzi, clearly would force us to rethink our ideas even further about language and about human uniqueness. “If an ape can begin to comprehend spoken English without being so trained, and was able to do more than emit differential motor responses on cue, it would appear that the ape possessed speech and language abilities similar to our own,” I later commented with my colleagues, in a scientific paper. “Even if the ape was unable to speak, an ability to comprehend language would be the cognitive equivalent of having acquired language.”5
The amount of time and effort that had gone into the first four years of rearing and testing Kanzi had exceeded that in any other published study I had undertaken, and the resulting body of data was the most extensive. I felt ready to present what I considered to be a strong case for a reevaluation of the language capacities in apes. Five years had passed since Herbert Terrace had published his famous paper in Science, in 1979, which concluded that, despite many claims, no one had demonstrated language in apes. I therefore selected Science as the most appropriate journal to which to submit a report of the work.
I prepared a manuscript titled “Spontaneous Language Acquisition by a Pygmy Chimpanzee,” and in December 1984 sent it to the editorial offices of Science, in Washington, D.C. I believed that the findings with Kanzi should at least reopen the door to the consideration of language skills in other species. “These results indicate that the propensity of the pygmy chimpanzee for the acquisition of primitive language skills is considerably in advance of that yet reported for other apes,” we wrote. “Language acquisition in the pygmy chimpanzee seems to be accompanied (and facilitated) by the ability to understand spoken English.” We pointed out that this was the first report of an ape with these abilities, and that the results had important implications for the evolution of language capacities.
Three months later, we received a rather curt rejection letter from one of the staff editors at Science. Two researchers had reviewed the manuscript. The first reviewer acknowledged that we were reporting important new information that was potentially significant to progress in the field. However, the review followed this with a blizzard of specific criticisms which, it suggested, must be addressed if the paper was to be published. The second review asserted that our claim for spontaneous language acquisition in a nonhuman species was not new.
How could
anyone who knew anything about the field suggest that spontaneous acquisition was not novel? The reviewer stated that he/she could not envision any way in which a lexigram system could lead to the emergence of linguistic communication, thus there could be no validity to our findings, regardless of the data we presented. I wrote to the editor at Science and pointed out the second reviewer’s apparent bias, and asked that the rejection decision be reconsidered. In mid-April we received a second rejection letter.
In the world of scientific publication, Science represents one of the two most widely circulated and respected general journals. The second one is Nature, which is published in London. We therefore recast our manuscript, to take account of some of the suggestions the first reviewer had made, and mailed it to London in May. In July we received a letter from an editor at Nature, saying “After careful consideration … we have concluded that publication in Nature would not be appropriate.”
We had no better luck with the Journal of Comparative Psychology, whose reviewers displayed the same mix of general negativism and incredulity in their comments with which we were by now becoming quite familiar. Finally, I submitted a much more detailed and extensive revision to the Journal of Experimental Psychology, along with the previous reviews I had received, so that the editor could see the difficulties that lay ahead. Most, if not all, of the other reviews appeared to have been written by people in the field of ape language. It seemed that, for the most part, they simply refused to believe what was being said about Kanzi. They did not raise methodological or philosophical objections; they simply denied the validity of our findings outright.
Kanzi: The Ape at the Brink of the Human Mind Page 18