The Apes Who Learned Sign Language

by Brian Dunning
July 3, 2018

In June of 2018, the famous gorilla Koko died of natural causes. Koko had become famous around the world for her incredible ability to speak in sign language. Koko wrote books. She raised kittens. Koko developed friendships with visiting celebrities. Koko would tell jokes and be silly. Koko would hear things on the news, feel sad, and express her sadness, all using sign language. But it turns out that the field of ape linguistics has virtually evaporated. What evidence there is that apes have been able to sign comes to us not from the academic literature, but only from the mass media — feel-good stories written for the people, apparently devoid of much scientific validity. Today we're going to point our skeptical eye at the field of apes who speak in sign language.

Here is the first thing that suggests not all is as it seems in the world of signing apes. If it works, and the ability to communicate with us can be instilled into apes, then why aren't there signing apes everywhere? Why don't we teach it to all of them — indeed, why don't ape parents teach it to their own young? If apes had any meaningful ability to communicate at even a rudimentary level using sign language, you'd think many of those who interact with humans would do so. They would tell their zookeepers what they want (probably to be set free). They would accompany primatologists on expeditions and translate between the scientists and the native apes. Every animal research institution would have signing apes, and everyone in the world with a degree in primatology would have experience communicating with them. After all these decades, we'd have a deep understanding of ape communication and their thought processes. Signing with apes would be commonplace and widespread.

But none of these things has happened. Not one of them. Not only did signing apes never become common, the number of research programs studying ape signing has gone from a few to even fewer. At its peak in the 1970s, the field of teaching apes to communicate with humans never had more active research programs than you could count on your fingers and toes; today, there is not even a single program anywhere in the world making publishable claims. Backwards is not where promising directions in research tend to go. In every field of science, when we see researchers abandoning projects, the reason nearly always tends to be that the project was a dead end.

Most histories of human-ape communication cite the work of Herbert Terrace as the moment the field became a dead end. Terrace was inspired by the case of a chimpanzee named Washoe, raised from infancy as a human child by Allen and Beatrix Gardner, a married couple of University of Nevada researchers. Accounts vary depending on whose you read, but Washoe is said to have learned anywhere from 100 to 350 signs. Terrace wanted to do better, so his lab at Columbia University acquired a chimp and named him Nim Chimpsky, a play on the name of prominent linguist Noam Chomsky.

For a decade, Terrace raised Nim like a human child in a home environment, though unlike the Gardners with Washoe, Nim was raised by multiple researchers in addition to Terrace. In 1979, Terrace and his associates published a seminal paper in the journal Science which became the trigger point for the subsequent decades of controversy over whether apes can actually use language. Terrace's finding was that they could not. He concluded:

For the moment, our detailed investigation suggests that an ape's language learning is severely restricted. Apes can learn many isolated symbols (as can dogs, horses, and other nonhuman species), but they show no unequivocal evidence of mastering the conversational, semantic, or syntactic organization of language.

The apes were using their symbol knowledge really only as a type of begging, when they wanted to play, eat, drink, or be tickled — little different than a dog pushing its bowl over to you with its nose when it is hungry. Any claims grander than this — for example, of sentence structure or evidence of emotions or how the ape was feeling — were likely nothing more than self-deception on the part of researchers who would prompt their apes and over-interpret their responses.

The Gardners disagreed vehemently with Terrace. Some ape researchers took their side, but the vast majority of scientists who viewed films of Washoe, Nim, and other apes in similar programs, and compared them to the linguistic data recorded by the researchers, found that Terrace was right. Apes weren't using the signs in any way that could be described as language, they were using them to beg for whatever it was they wanted using the communication method they were raised with, just like other domesticated animals.

More than a dozen similar research projects around the world fizzled out, none producing any better evidence that apes could communicate linguistically. By the mid 1980s, there basically weren't any research projects left anywhere that were publishing data. As a science, apes communicating via sign language was essentially dead.

Some programs used symbolic keyboards instead of sign language. The most notable example was a bonobo name Kanzi who was studied by psychobiologists Duane and Sue Rumbaugh. They published that Kanzi understood sentences, including semantics and syntax, and could carry out complex commands. They filmed a demonstration in which Kanzi was given 660 complex commands, including some which depended on word order to truly test Kanzi's comprehension. It sounded truly impressive.

But animal behaviorist Dr. Clive Wynne from Arizona State University took his own look at the film, and compared what actually took place to the scores that were recorded, and his report was abysmal. Commands often had to be repeated multiple times, and if Kanzi got even part of it right — for example, picking up the correct item but taking the wrong action with it — it was always scored as correct. "It became clear to me," he wrote, "that their method of coding Kanzi's responses was unreasonably generous." When he excluded all the false positives, Wynne found that Kanzi was correct no more than 30% of the time. Like all the published accounts of apes using language, Kanzi's showed little promise that apes could do anything more than recognize symbols, in same way other animals do.

However, one ape's ability remained undiminished in the public's eye: that of Koko. Koko, as practically any person on the street could tell you, was the gorilla who spoke in sign language. Koko, so far as anyone believed, was living proof that apes have complex emotions, can think abstractly, and can hold intimate conversations with anyone using sign language. Koko was the media darling of signing apes, the only one whose name was widely known, the one all the visiting celebrities would have their pictures taken with. Robin Williams and Fred Rogers were two who notably spent time with Koko — videos of Williams playing with Koko are available online. Koko was clearly gentle, lovable, and welcoming, truly a living teddy bear you couldn't help but love. When she died of natural causes in June 2018, the worldwide media lamented her passing with tributes and obituaries — always referring to her as the gorilla who spoke in sign language.

But what's always been missing from Koko's story is experimental validity. Fans of Koko's might note that they've never seen her on TV (or in any of scores of YouTube videos) speaking with anyone other than her handler, Dr. Penny Patterson, who does not publish research. In the videos, it is not at all apparent that Koko pays very much attention at all to Patterson's commands, though Patterson is constantly laughing and commending Koko's successes. As researchers discovered in the other ape research programs, Koko never initiated signing, but only responded to cuing from Patterson, and even then, it typically only made sense when Patterson would interpret and explain what she said Koko actually meant. Terrace's article also pointed out that 100% of the signs Koko made in a NOVA television special on PBS were made immediately after Patterson made the sign first. Other apes fared little better; analysis of a longer film of a chimp named Ally showed that 92% of Ally's signs were repeats of signs shown to him first by the teacher.

The most prominent demonstration of Koko's language ability was in 1998 when she went on a live AOL chat with several thousand people. Patterson signed for Koko and interpreted over the phone to the AOL chat editors. Of about 25 responses that Koko made, only one seemed to have any relevance to the question: she signed dog after Patterson asked if she'd like to have a kitty, dog, or gorilla as a friend. All the other answers made no sense at all; though in most cases, Patterson provided an explanation stating that it was what Koko meant. Several times, Koko begged for candy or juice, though in none of those instances was it relevant to the question. Rather than being an impressive demonstration of interspecies communication, the AOL chat was a fairly stark example of unsuccessful efforts to cue and prompt the gorilla.

The probable reason for the dramatic failure of the AOL chat was described in a major investigative article by two science writers for the San Francisco Examiner, Keay Davidson and Janet Hopson titled "Gorilla Business". They wrote:

The consensus of numerous animal researchers the reporters contacted for an assessment of Penny Patterson's work is that overinterpretation of data and subtle cueing of the animal still plague the efforts and that the picture is clouded further by underreporting in the scientific literature. These negatives offset much of the good accomplished by the greater sensitivity toward apes Patterson's publicity engenders.

It is fair to point out that as disappointing as the published research has been finding that the apes' language abilities are severely limited, opposing claims made by many of the ape language researchers are just as passionate. It is easy to find articles and videos — usually in consumer publications — arguing that apes like Koko experience and express complex emotions, and have thorough understanding of the sign language they engage in. By no means is the debate over ape language ability one-sided; however, the science of that ability regrettably is.

When we take the example of an animal as humanlike and genial as Koko, one who has been known around the world and loved by so many people, it's hard not to over-anthropomorphize them. We want to assign them humanlike qualities, humanlike abilities. We want to form organic bonds with them. We want Patterson's interpretation of random actions to mean that Koko is indeed communicating with us about something deeply moving. We want to see ourselves in these wonderful animals; and often we don't stop to think whether what the animals actually are — stripped of our imposed anthropomorphism — might, itself, be sufficiently wonderful.

Correction: An earlier version of this erroneously misplaced Dr. Clive Wynne at the University of Arizona rather than Arizona State University. —BD

By Brian Dunning
Follow @BrianDunning