Why a Chimpanzee Can’t Learn Language: 2

They're unable to learn that things have names.

Posted Nov 02, 2019

How can one teach language to a chimpanzee who is unable to produce the basic sounds of spoken language (phonemes)? Extensive research has shown they can’t. Two approaches, each using a visual language, have been tried and failed. In one, discussed in my last blog, a chimpanzee was raised informally in a home environment.

He was unable to learn American Sign Language, a gestural language used by hundreds of thousands of deaf people. Of necessity, communication was informal. Researchers had to rely on home diaries and occasional videos to capture the teacher’s and the chimpanzee's utterances.

 Terrace (2019)
Source: Terrace (2019)

In another approach, carried out in laboratories, researchers hoped that their results would be easier to replicate since they used objective methods. These projects trained chimpanzees to learn languages composed of visual symbols. Examples from two languages are shown on the left. One language was invented by David Premack, the other by Duane Rumbaugh and Sue Savage-Rumbaugh.

 Terrace (2019)
Sarah
Source: Terrace (2019)

Premack’s project trained chimpanzees to use visual symbols in a particular order. On the left, Sarah, his star student, can be seen obtaining an apple after producing the sequence of symbols, please -> give -> Sarah -> apple.

The symbols used by Rumbaugh and Savage-Rumbaugh were referred to as "lexigrams." Examples are shown on the right side of the first figure. Each lexigram is composed of a geometric shape that was presented on a colored background.

 Terrace (2019)
Lana.
Source: Terrace (2019)

Lana, the first chimpanzee who learned to use lexigrams is shown on the left, producing the sequence, please -> machine -> give -> apple. That sequence earned her an apple.

We’ve seen that the results of both projects are similar. Because the automated apparatus Rumbaugh and Savage-Rumbaugh used ruled out the possibility of cuing, and was therefore the most objective, I will only discuss their results in this blog.

Like the projects that used sign language, the hope was to find evidence that chimpanzees could create sentences. However, both approaches made the same error. Neither considered the status of the symbols the chimpanzees learned as words. Instead, the chimpanzees’ utterances functioned as imperatives, at best, -demands for particular rewards. Indeed, many of the lexigrams were meaningless and didn't even qualify as imperatives.

Consider the sequence, please -> machine -> give -> apple, that Lana and other chimpanzees learned. There is no evidence that Lana knew the meanings of the first three lexigrams: please, machine, or give. That would require the ability to contrast please with I insist, dammit, and so on. Likewise, a chimpanzee would have to be able to contrast machine with a person's name or some other agent, and give with take or other relevant verbs. Without such evidence, the meanings glossed by Rumbaugh and Savage Rumbaugh exist only in their imaginations.

There is evidence, however, that these chimpanzees understood the difference between the lexigrams apple and lexigrams for other rewards. If, for example, Lana selected the lexigram bread, a less desirable reward than an apple, she repeated the sequence on her own so that it ended with apple.

But even if all of the lexigrams functioned as imperatives, sequences of lexigrams would not qualify as sentences. That's because they were learned by trial and error, in much the same manner as people who learn to use a password to operate an ATM. The only function of such sequences is to obtain a reward. In the case of chimpanzees, the reward is food or drink. In the case of an ATM, it is cash.

The folly of interpreting sequences of lexigrams as sentences is illustrated by some experiments I performed on monkeys, animals not known for their linguistic ability (Terrace, 2005). The monkeys were trained, by trial and error, to produce rote sequences that were composed of photographs of natural objects.

The figure below shows a monkey responding to seven photographs, all of which appeared simultaneously on a touch-sensitive video monitor: mountain -> birds -> frog ->deer -> person -> dolphin -> flowers. If the monkey touched the phots in that order, he was given a banana pellet.

 Terrace (2019)
Source: Terrace (2019)

Those photographs are as meaningless as the rote sequences of lexigrams that chimpanzees learned to produce or the passwords people learn to obtain cash at an ATM. If we use the logic of Lana's trainers to interpret the meaning of the monkey’s sequence, we might conclude that it was saying, please -> machine -> give -> banana pallet -> been -> good -> today!

The best that can be said of all of the attempts to teach chimpanzees to learn language is that the signs and symbols they learned only functioned as imperatives. If that were the only function of words that human infants learned, they would never learn language. By contrast, the words an infant learns to speak are spontaneous and conversational. Infants and their caretakers take turns as speakers and listeners. No chimpanzee has that ability.

Knowing language, it's all too easy for us to attribute it to animals who were trained to use symbols. In each instance, linguistic knowledge is not needed to explain their behavior.

References

Terrace, H. (2005). The simultaneous chain: A new approach to serial learning. Trends in Cognitive Science, 9, 202-210.

Terrace, H. S. (2019). Why Chimpanzees Can’t Learn Language and Only Humans Can. Columbia University Press. New York, NY.