“‘Project Nim,’ a new documentary by James Marsh, is a probing, unsettling study of primate behavior, focusing on the complex dynamics of power, sex and group bonding in a species whose startling capacity for selfishness and aggression is offset by occasional displays of intelligence and compassion.
The movie also features a chimpanzee.”

So begins A.O. Scott’s review of “Project” Nim, 2011, in The New York Times.

 

“No chimpanzee has learned sign language. They’ve certainly learned some gestures, but sign language is not just a system of gestures. It’s a full, grammatical language with its own systematic grammar, like Latin.”

Psychologist Steven Pinker, Author of The Language Instinct

This schizophrenic attitude persists today. No reputable scientist disputes Darwin’s assertion of physical continuity from the simplest animals to humans, and the great apes clearly share much with humans. Their anatomy and their genome resembles ours more than any other organism, and even their brains have similar—though smaller—parts. If researchers could emulate the fictional Dr. Doolittle and converse with an animal, surely that animal would be a great ape. On the other hand, some scientists insist that the resemblance ends at language. Despite a continuity of other traits, they say, language stands alone, not merely the most complicated kind of communication, but a unique one, unrelated to that of any non-human animal.

Early attempts, from the 1900s through the 1930s, to teach chimps to speak met with dismal failure, vindicating the critics. The animals just couldn’t wrap their otherwise expressive lips around words. In the most successful cases, they made sounds charitably interpreted as short words, such as mama, papa, cup and up, after years of training.

Following La Mettrie’s suggestion that a gifted teacher of the deaf could succeed with chimpanzees, a 1925 scientific article suggested sign language as an alternative. But serious efforts to teach non-vocal communication to apes only began in the 1960s. Researchers attempted to teach individual signs derived from American Sign Language (ASL) to Washoe, a chimpanzee; Koko, a gorilla; and Chantek, an orangutan. Sarah, a chimpanzee, learned to manipulate arbitrary plastic symbols standing for words, and another chimpanzee, named Lana, used an early computer keyboard, with arbitrary symbols the researchers called lexigrams.

All these projects succeeded where the early speech projects had failed. The apes learned to use hand gestures, plastic symbols or keyboards to communicate with their trainers. The 1960s and 1970s became the golden age of ape language-learning. Researchers claimed (and some continue to claim) that the apes had learned tens or even hundreds of signs.

But popular accounts went farther. They had it that the apes held conversations, and had “learned sign language.” To this day, assertions that apes can converse with humans using symbols or sign language abound in popular magazines and books and even college textbooks. But although the trained apes often used two or three signs or symbols in a sequence, and could clearly get a message across—most often a request for food or attention—researchers wondered if the apes had learned Language with a capital L. Some researchers working in the field feel justified in using the word “language” to describe the results of these experiments, but psychologist Steven Pinker, author of The Language Instinct, disagrees.

Pinker has declared on public radio “No chimpanzee has learned sign language. ... They’ve certainly learned some gestures, but sign language is not just a system of gestures. It’s a full, grammatical language with its own systematic grammar, like Latin.”

Setting the idea of a full language aside, however, did the apes’ hand gestures constitute words? Did they truly understand that signs or lexigrams stood for objects or actions? Were their strings of two or three signs sentences?

Nim Chimpsky

Herbert Terrace, of Columbia University, attempted to solve the sentence problem with a chimpanzee called Nim Chimpsky, named in humorous honor of Noam Chomsky, the renowned linguist. Chomsky had asserted that if apes could use language, they would do so in the wild. They don’t, he said, so they can’t.

Terrace taught Nim gestures based on ASL and succeeded, just as others had succeeded in teaching Washoe, Sarah, Koko, Chantek and Lana. With near round–the–clock help from a platoon of students, Terrace signed with Nim, recording long strings of signs, and accumulated a huge body of data on audio tape and film.

Terrace’s teaching schedule and what he called “the baby-sitting problem” brought Nim’s tenure at Columbia to a close. Nim moved on to the Institute for Primate Studies in Norman, Oklahoma and Terrace turned to an extensive analysis of the data gathered during Project Nim. In his 1979 book Nim, Terrace wrote: “The regularities in our corpus that were noted before Nim returned to Oklahoma gave me reason to believe that Nim was creating primitive sentences. Our intensive post-Oklahoma effort at data analysis had hardly begun, however, when I began to doubt that Nim’s combinations were legitimate sentences.”

Terrace concluded that despite long strings of signs such as “give orange me give eat orange me eat orange give me eat orange give me you,” Nim’s actual “sentences” averaged 1.5 signs. Had Nim’s learning ground to a halt, or had he just gotten what he wanted without longer strings of signs? Terrace concluded that Nim had never signed a true sentence, and that many of Nim’s individual signs immediately followed similar signs by his trainers. Close examination of his films convinced Terrace that Nim mostly imitated trainers, often after prompting. Furthermore, Terrace analyzed films of other ape projects, including two about Washoe, and concluded that in those projects too, trainers repeatedly prompted and then interpreted separate responses as sentences.

In one example, Washoe is supposed to have signed “Baby in my drink” when shown a doll and a cup. Terrace describes the film: “Washoe is with her teacher Susan Nichols, who has a cup and a doll. Ms. Nichols points to a cup and signs ‘that.’ Washoe signs ‘baby.’ Ms. Nichols brings the cup and the doll closer to Washoe, allowing her to touch them, then slowly pulls them away, signing ‘that’ and pointing to the cup. Washoe signs ‘in’ and looks away. Ms. Nichols brings the cup and doll closer to Washoe again, who looks at the two objects once more and signs ‘baby.’ Then, as Ms. Nichols brings the cup still closer, Washoe signs ‘in.’ ‘That,’ signs Ms. Nichols, and points to the cup. ‘My drink,’ signs Washoe. Now the question is, is this utterance by Washoe—baby in baby in my drink—either spontaneous or a significant, creative use of words?”

Terrace reluctantly concludes “Until it is possible to defeat all plausible explanations short of the intellectual capacity to arrange words according to a grammatical rule, it would be premature to conclude that a chimpanzee’s combinations show the same structure evident in the sentences of a child.” Terrace softened this stern proclamation with “This is not to say that a chimpanzee is simply not capable of creating a sentence.”

Kanzi

This statement notwithstanding, other researchers in ape language-learning saw Terrace’s conclusions as an attack. Because of Terrace’s well-reported change of mind, funding dwindled for the projects, some researchers left the field, others distanced themselves from the scientific community, and a sense of blame persists. In the 1994 book Animal Learning and Cognition, Duane M. Rumbaugh and E. Sue Savage-Rumbaugh write “But the view of Terrace and his associates prevailed to the extent that it soon became widely accepted that (1) because Terrace’s Nim did not have language and (2) because analyses at Terrace’s laboratory of taped materials from the laboratories of other signing projects indicated that others’ apes were also imitating, there was the strong implication that (3) no ape had demonstrated any language competence whatsoever and that (4) language was beyond the competence of apes!”

Rumbaugh and Savage-Rumbaugh run one of the few animal-language research projects to continue full steam after Terrace’s change of mind, at Georgia State University’s Language Research Center. Following up on the Lana project, Rumbaugh, Savage-Rumbaugh, Rose A. Sevcik and others continued teaching language to great apes, with some remarkable results and some disappointments. Two chimps, Sherman and Austin, learned, with extensive training, to communicate by way of keyboards in a cooperative effort to use simple tools to get food. In the process, they had to learn lexigrams for foods and for tools they needed to obtain the food. With much less training, they learned to categorize 20 lexigrams for foods and 20 for tools using a lexigram for tool and another for food, a significant linguistic feat in itself. But another task hinted at things to come. Without specific training, both chimps were able to look at a lexigram, then reach into a box they couldn’t see into and pull out the named object. These feats appeared to show at least that Sherman and Austin grasped the concept of naming. They seem to use symbols as words.

But a separate project with an adult ape called Matata turned out to be a bad-news-good-news situation. Matata belongs to a species called bonobos (Pan paniscus), a close relative to the chimpanzee (Pan troglodytes). Often called—incorrectly—pygmy chimpanzees, bonobos differ from chimpanzees in several respects. They stand a head shorter, weigh less and have a more gracile body shape. They stand upright more often, have different faces and even communicate in the wild differently. Furthermore, their social structures in the wild differ from those of chimpanzees. Some see these differences are seen as making bonobos more like humans. Could they learn language more easily? The bad news was Matata failed to reliably use even a few lexigrams after years of training.

Matata was a working mom, however, and brought her adopted baby Kanzi to work. And therein lies the good news part of the tale. While Matata sat bemused by the keyboard, Kanzi crawled in her lap and on her back or played nearby. The researchers tolerated Kanzi, but never trained him. He grew up for two years in an environment where humans continuously made sounds to his mother and tapped at a keyboard, trying to teach her individual signs. When Savage-Rumbaugh finally gave up on Matata—not only hadn’t she learned to use lexigrams to ask for what she wanted, but she had a dangerous tendency to roughly grab from researchers—they allowed Kanzi to remain at the Language Research Center and sent Matata back to the nearby Yerkes Primate Center where she could use her native communication mode to find a mate. Kanzi was two and a half years old.

Like the child of an immigrant, Kanzi soon showed he had absorbed just what Matata had resisted. Within a week he spontaneously began to use the keyboard to make his desires known, But he also appeared to name objects even when he did not want the object. Savage-Rumbaugh and Sevcik decided not to train Kanzi at all, but to see if he could continue to soak up the keyboard “language” during daily interactions with researchers, who talked to him, using both lexigrams and speech, as if he understood. In other words, they treated Kanzi like parents treat a preverbal child constantly hearing language. Kanzi’s keyboard helped in this effort by generating synthesized speech to sound out the English word for each lexigram.

Made wary by the Terrace incident, Savage-Rumbaugh and her colleagues attempt to meet the many legitimate objections raised by Terrace and others. They try to avoid prompting, and to construct experiments that will stand up to close scrutiny. Furthermore, the researchers are careful with their claims. They use the phrase “non-random lexigram combinations” instead of “sentences,” for example. But Kanzi works for attention, not food, and the team can’t eliminate people from the experiments. And Kanzi’s non-random lexigram combinations rarely exceed three lexigrams. Watching Kanzi in casual “conversation” with Sue Savage-Rumbaugh, an observer is struck by the intense give and take, reminiscent of Terrace’s description of the “baby in my drink” film clip.

Kanzi’s two-and-three-word sentences on the keyboard may seem less than impressive. But a set of experiments comparing Kanzi’s understanding of spoken English to that of Alia, the two-and-one-half-year-old daughter of a Language Center researcher appears to show a very different level of understanding. Kanzi and Alia were presented with sentence-understanding tasks as similar as the researchers could make them. Archival videotape of Kanzi’s performance sets the scene.

Kanzi sits in a room with two researchers (one is Rose Sevcik). A third (Sue Savage-Rumbaugh) stands outside the room with a microphone. The two inside researchers wear earphones playing loud music to reduce the chance they can give Kanzi any clues. The room has a “kitchen,” and a large playroom with a number of objects Kanzi has never seen. A child’s toilet, a pitcher of water, a rubber snake, a stuffed dog, a 25-pound bag of carrots, a hand puppet vaguely resembling a rabbit. The voice from outside says “Kanzi, make the dog bite the snake.” Kanzi immediately picks up the rubber snake and the plush toy dog. He carefully puts the snake’s head into the dog’s mouth and gently squeezes the dog’s jaws shut. An impressive show of understanding made more impressive by the fact that Kanzi has generalized the spoken words dog and snake to toys he’s never seen.

“Kanzi, tickle Rose with the bunny,” says Savage-Rumbaugh. Kanzi picks up a bunny hand puppet, carries it to Sevcik and tickles her. Sevcik says in explaining the videotape that Kanzi’s only previous knowledge of “bunny” was a videotape of a Language Research Center worker dressed in a bunny suit. The researchers had never drilled Kanzi (or Alia) on the requests, and all of the objects were new, purchased just for the experiment.

Duane Rumbaugh summarizes the results: “Kanzi’s comprehension of 500 novel sentences of request were very comparable to Alia’s. Both complied with the requests without assistance on about 70% of the sentences.” He emphasizes that Kanzi learned by observation alone very early in life, and further that the researchers only discovered this fact by the lucky decision to keep Kanzi around after Matata was sent home. “The apes can come to understand even the syntax of human speech at a level that compares favorably with that of a two-to-three-year-old child—if they are reared from shortly after birth in a language-structured environment. Reared in this manner, the infant ape’s brain develops in a manner that enables it to acquire language. First through its comprehension and then through its expression, a pattern that characterizes the course of language acquisition in the normal child. We had no intention of studying language-observational learning in [Kanzi]. But it happened and we’ve replicated it with other [bonobos and chimpanzees],” Rumbaugh says.

Duane Rumbaugh and Sue Savage Rumbaugh summarize—fairly, it seems—the current state of research in their chapter of the 1994 Animal Learning and Cognition. “Though none will argue that any animal has the full capacity of humans for language, none should deny that at least some animals have quite impressive competencies for language skills, including speech comprehension.”

Next: Communication Among Apes

Minor copy editing 04 29 11