The First Word: The Search for the Origins of Language

by Christine Kenneally

  However, in 2006 Aubrey Gilbert and colleagues announced in the Proceedings of the National Academy of Sciences (PNAS) that the way a speaker’s language distinguishes color does affect the way he or she sees it. The nature of their experiment had the additional benefit of providing a clue as to why previous experimental results have been so contradictory. Apparently, the way you see color depends on what side of the brain you are using.15

  Gilbert and colleagues hypothesized that if language is dominant on the left side of the brain, it should impact the perception of input in the right visual field. (The left side of the brain controls the right visual field, and vice versa.) They showed subjects colors with color words, and found that subjects were able to make faster judgments about colors and color categories in the right visual field when the color and the word matched. If there was a conflict between the color and the word on this side, they were slowed down in their responses. The wrong word interfered with their ability to decide what the color was. When they were asked to make a judgment about colors and words in the left visual field, using the nonlanguage-dominant side of the brain, they weren’t affected at all.

  It’s not clear yet whether the language of color affects the way an individual physically perceives color in the world or whether the influence of language kicks in after some basic perception has taken place. Nevertheless, Gilbert’s experiments show that linguistic categories affect thought. Lupyan said, “What we are now learning is that besides communicating information, language seems to alter how the brain processes it. Individuals, like stroke patients, who suffer from aphasia—a condition characterized by varying degrees of language loss—do not just find it more difficult to communicate; they also find it more difficult to categorize, remember, and organize information. This is evidence that language is playing a role in these cognitive tasks.”

  In his own research Lupyan addresses the question of how language in general, rather than specific languages, changes the way we perform cognitive tasks. He devised an experiment to tease out some of the ways words might affect how we think. Lupyan used a set of odd-looking clay creatures with prominent heads and strange pointy limbs, which he called aliens. His aliens fell naturally into two groups. In one, the creatures’ heads were fairly smooth, and in the other their heads were somewhat lumpy and misshapen. Crucially, the differences were subtle and not easy to articulate. He then told two groups of students that some of the aliens were friendly and some were not. The students’ task was to decide which was which, and then to assign them to separate groups. After they made each choice, students got feedback about whether they guessed right or wrong, meaning that as they went through the task, they basically learned that smooth heads were friendly and lumpy heads were not.

  Lupyan added a little piece of information to one of the test groups. After the members of the group found out whether their choice had been right or wrong, they were also shown a word. Lupyan told them that previous subjects had found it helpful to label the friendly and unfriendly aliens, calling the friendly ones “leebish” and the unfriendly ones “grecious” (or vice versa). He found that even though both groups eventually learned the difference between the aliens with equal success, the group that had words to label them learned to distinguish them much faster than the non-word group. He concluded that language, specifically the act of naming something with a word, helps categorize.

  “Separating language and thought is hard,” Lupyan acknowledged. “But it is precisely because of this that we have to start thinking of them as not separate things, but as a system. As language is learned, it alters how we process information. Just as when we learn to identify a face with a name, it alters how we treat a face—it’s not just a face, it’s my friend Mike—so learning language results in our automatic labeling of objects, actions, sounds, and even more abstract categories like emotions. This labeling categorizes the item and links it to other instances of the category.”

  Language not only boosts cognition but can help or hinder thought, depending on the task in question. In 1990 Jonathan Schooler at the University of Pittsburgh demonstrated that when people were shown a face in a mock crime videotape and asked to write a description of it, they were worse at picking that face out of a subsequent lineup than people who hadn’t written their impressions down. “This makes sense,” said Lupyan, “if we think of linguistic descriptions as forcing us to think in categories. Writing ‘he had brown hair’ can impair later identification because ‘brown’ refers to a category and not a particular color.”

  Other language and thought experiments have looked at how we process number. Because of researchers like Wynn, Hauser, and their colleagues, we know that certain aspects of number ability do not depend on language, as some animals can think numerically, and children and adults use various number and geometric concepts independently of language. Nevertheless, a recent experiment suggests that some numerical concepts are difficult to understand if they don’t exist in the language you speak.

  Peter Gordon of Columbia University has studied the Pirahã tribe, which lives along the Maici River in Brazil. The Pirahã are known to the scholarly community because of the years of fieldwork carried out by the linguists Keren and Daniel Everett, the latter now professor of linguistics and anthropology, and chair of Languages, Literatures, and Cultures, at Illinois State University. There are only about two hundred Pirahã, who live in groups of ten to twenty and maintain a hunter-gatherer lifestyle, resisting assimilation into mainstream culture. They are completely monolingual and only occasionally communicate in a primitive pidgin with outsiders. There is no precise number system in the Pirahã language, which relies instead on a “one-two-many” categorization, distinguishing merely between amounts that are not much and those that are larger. For example, hoi means “roughly one” or “small”; there is no word for a singular amount. Spoken in a different tone, hoi can also mean “two,” as distinct from “one.” Baagi or aibi designates amounts that are a few or larger. (This kind of system is not uncommon in many languages of the world.)

  Gordon carried out a series of experiments on Pirahã speakers that were designed to test their numerical abilities.16 In one, he sat across from his subjects, with a stick dividing his side from theirs. He positioned a line of evenly spaced AA batteries on his side and asked the Pirahã to place a similar array of batteries on theirs, matching each of his with theirs in a one-to-one correspondence. With each successive repetition of the task Gordon made it harder and harder by asking his subjects to match clusters of nuts to the batteries, or match orthogonal lines of batteries, lines that were unevenly spaced, or lines on a drawing. He found that the Pirahã were successful with two to three objects but had much more difficulty with larger numbers from eight to ten, where their success rate dropped to zero. The exception to this result was the test that asked the Pirahã to match unevenly spaced clusters. Although they had trouble with between three and six objects, they were almost perfect in matching seven to ten. Gordon suggests this was because the uneven display essentially allowed the subjects to break the larger amount down into groups of two and three.

  The Pirahã’s numerical abilities were consistent with the way infants and certain animals can make relatively accurate estimations of small numbers of objects—up to three—concluded Gordon. Beyond this, if a person’s language does not contain a number system that labels quantities like four and five, he may not have the ability to identify or use these numbers. The underlying concept is that languages that contain terms for higher numbers basically teach the learner-speaker to count at this level.

  As experimenters become more sophisticated in their methods, it’s reasonable to imagine that the ways that thought is ramified by the complexities of language will become more apparent. In the meantime, the work of Gordon, Lupyan, and others suggests that words are not just convenient labels for things; rather, they are extremely powerful mental devices. And if there is one aspect of language that appears to be a uniquely human and rela
tively recent innovation, it has to be the sheer size of our vocabulary. It’s thought that speakers can have a vocabulary of sixty thousand words. But how old are words, exactly? Do animals have them? And if they do, does that mean that words have been around longer than humans?

  6. You have words

  In the 1980s two researchers from the University of Pennsylvania, Robert Seyfarth and Dorothy Cheney, published some attention-grabbing data about the communication of African vervet monkeys.1 The researchers confirmed a 1967 discovery that the monkeys made specific, wordlike warning calls in response to particular predators. In a vervet group, all the animals are consistently looking around, and in a group of ten to twenty individuals someone usually spots any nearby predators. When it does, it gives an alarm. If the monkey making the alarm call saw an eagle, it would make one kind of cry sound; if it saw a leopard, it would make another; and if it saw a snake, it would make a different sound altogether.

  Not only did the vervets produce different cries, the rest of the group reacted differently to each type of signal. If the lookout monkey made an eagle call, then all the vervets would take up the cry, echoing the sentry, while running beneath the cover of trees. Being under foliage was the best hiding spot in case of an aerial attack. If the lookout’s alarm cry indicated the sighting of a snake, the vervets would do the opposite, climbing up into the trees and repeating the call—Snake! Snake! Snake! Up off the ground, in this case, was the safest place to be. If the sentry monkey spotted a leopard, it would make the leopard cry, and the vervets would likewise leap into the trees, but now they would climb out onto the narrowest, most lightweight branches. These were the perfect place to be if a hungry leopard was prowling, because the lighter branches wouldn’t support the weight of the predator if it followed them up into the tree. In 1967 the vervet behavior was only observed. Seyfarth and Cheney replicated the observation of the vervets’ different responses in experiments using alarm call recordings.

  Each type of warning cry was consistently the same sound. It was as if there were three words that had been agreed upon by the whole monkey community, in the same way humans agree upon the arbitrary words of each human language (“eagle” if you speak English; aigle if you are French).

  There was great excitement at these findings, which suggested that we’d finally found evidence of an animal word that worked the same way a human word does. The last common ancestor of vervets and humans lived around thirty million years ago. Was it possible that all you needed to achieve the complexity of human language was a proliferation of words, some syntactic rules to make them all work together, and thirty million years? And did this mean that words preceded humans?

  For a number of reasons, it turns out, the answer is probably no. But it is a gray kind of “no,” and the reason the vervet cries are not satisfying candidates as animal words is not the most important thing about them.

  Vervet alarm-calls-as-words had such appeal in the scientific community and the popular press in part because these animals are relatively close kin to humans. If you think of chimpanzees and bonobos as our brothers and sisters, and gorillas and orangutans as our aunts and uncles, then the vervets might be third or fourth cousins. Alarm calls from vervets were much easier to imagine as the antecedents of our language than if they had been coming from, say, a chicken.

  But alarms calls are ubiquitous in the animal world. Monkeys have them. Ground squirrels have them. Meerkats have them. As recently as 2005, researchers in the journal Science discussed the complicated and clever alarm calls that chickadees make. And, yes, even chickens make alarm calls, distinguishing between terrestrial and aerial predators.

  “Most birds,” said Tecumseh Fitch, at the University of St. Andrews in Scotland, “have a sort of generalized alarm call and an aerial predator alarm call. It is by no means unusual in the animal kingdom to have at least two different kinds of alarm for two different types of threat. Ground squirrels have about eighteen calls, and meerkats have more alarm calls than vervets simply because they have more predators than vervets.”

  Even though humans are more closely related to vervets than vervets are to chickens, it appears that vervets and chickens have converged upon a common tactic for survival. The forces that led them both to this strategy are powerful, but alarm calls were probably not bequeathed to them from a common ancestor. In fact, the most important thing that they share with all the other alarm-call-making animals is that they are small and delicious. Fitch explained: “The things that have alarm calls are little tiny guys who get eaten by lots of things, and the common ancestor of chimps and humans wasn’t in that category. Humans don’t have alarm calls, and apes don’t have alarm calls. It’s not that they don’t have threats, but they don’t have all these different threats where it pays to be able to refer very rapidly to aerial threat versus ground threat. Whether you’re the Snickers bar of the Sahara or the Snickers bar of South Dakota, you’re going to evolve alarm calls.”

  Fitch discusses the evolution of communication with enormous energy. He was named for his great-great-great-grandfather William Tecumseh Sherman, who ended the Civil War with his famous march from Atlanta to the sea. (His ancestor, in turn, was named for the Shawnee Indian leader Tecumseh, who traveled up and down the East Coast, uniting tribes in opposition to the spread of the United States.) Fitch himself occupies a unique spot in the new field of language evolution. He studied under Lieberman, writing his Ph.D. thesis on the evolution of speech. And in 2002 he collaborated with Noam Chomsky on the first paper Chomsky wrote about the evolution of language.

  “Alarm calls seem to be a prime candidate for language evolution,” Fitch said, “but they are not.” The calls aren’t like human words, because they are genetically preprogrammed: animals will produce them even when raised in isolation. “What vervets have is the ability to communicate a very limited set of meanings,” he explained, “and because that’s genetically determined, there’s no way other than genetic modification to add new units into the system. Each call type has to evolve over Darwinian time, and you can’t evolve limitless meaning, as you have in human language, in Darwinian time.”

  What would English speakers look like if they inherited sixty thousand words genetically? It’s hard to imagine. Babies would presumably be able to talk from birth, and they’d have an enormous memory capacity. Most animals that have a lot of information genetically coded are born looking fairly well developed. Our nine months of pregnancy might be considerably longer. New words—and the ideas or innovations they represent—would have to propagate through the species genetically, so adding a single word or idea like “wheel” or “fire” or “cooked meat” would take a few thousand years. Science, art, and McDonald’s would just never get off the ground.

  If alarm calls aren’t words, then what are they? “They’re not words in the same sense of language,” Fitch explained. “They’re more like laughter and crying, which are also calls that are innate. You don’t need to hear your mother crying to learn how to cry. Deaf children make these sounds, too.” And as you grow, you learn that when you laugh, people nearby can safely assume that something you find amusing has occurred. If you burst into tears, they can likewise guess that something you find upsetting has happened. “No one has to have any recourse to words to make these sounds or to interpret them,” Fitch said.

  We don’t know exactly how these calls evolved, but it’s not hard to imagine that if you were a vervet monkey with a tendency to laugh hysterically and run up a tree every time you saw a leopard heading your way, then you and your troop might end up more likely to survive and to reproduce. When you did reproduce, you’d pass on that genetic predisposition to at least some of your children. The stoic monkey would be a dead monkey.

  Instead of seeing alarm calls as a primitive form of language, we should look at them as a communication device that many animals share. Across a wide swath of life, animals as genetically distant as birds (famously descended from dinosaurs) and mammals have evolved dis
tinct units of sound that act as pointers to things in the real world.

  It could even be argued that human calls—laughter and crying, which certainly intersect closely with language—are a degenerate form of the alarm calls of prey species. When people hear you laugh, they know you are laughing at something, but don’t necessarily know anything else about it. When vervets make the eagle call, other vervets know that something scary and aerial is headed their way and that they should look up, as opposed to around or down on the ground. In this regard, they make more reliable and specific inferences than we do.

  Some researchers still think it’s possible that alarm calls are a kind of protoword—that we somehow broke the link between the vocal token and the DNA, retaining the ability to use freely a sound token to refer to things in the world. There is some interesting neurological evidence for this possibility. Chris Code, a research fellow in the School of Psychology at the University of Exeter, points out that it is possible neurobiologically to separate swearwords from other words in language. Swearing actually uses parts of the brain that support language and also parts of the brain that are used when laughing and crying. Often people with severe brain damage remain able to swear even when they are unable to produce other language. Perhaps swearing is the remnant of an evolutionary step at which cries were some mix of automatic and voluntary articulation. While the possibility cannot be ruled out altogether, the safest conclusion at this stage is that alarm calls are probably not the antecedents of words.