The Tell-Tale Brain: A Neuroscientist's Quest for What Makes Us Human

I would like to propose a different framework for thinking about language evolution that incorporates some features of both but then goes well beyond them. I call it the “synesthetic bootstrapping theory.” As we shall see, it provides a valuable clue to understanding the origins of not only language, but also a host of other uniquely human traits such as metaphorical thinking and abstraction. In particular, I’ll argue that language and many aspects of abstract thought evolved through exaptations whose fortuitous combination yielded novel solutions. Notice that this is different from saying that language evolved from some general mechanism such as thinking, and it also differs from Pinker’s idea that language evolved as a specialized mechanism exclusively for communication.

NO DISCUSSION OF the evolution of language would be complete without considering the question of nature versus nurture. To what extent are the rules of language innate, and to what extent are they absorbed from the world early in life? Arguments about the evolution of language have been fierce, and the nature-versus-nurture debate has been the most acrimonious of all. I mention it here only briefly because it has already been the subject of a number of recent books. Everyone agrees that words are not hardwired in the brain. The same object can have different names in different languages—“dog” in English, “chien” in French, “kutta” in Hindi, “maaa” in Thai, and “nai” in Tamil—which don’t even sound alike. But with regard to the rules of language, there is no such agreement. Rather, three viewpoints vie for supremacy.

In the first view, the rules themselves are entirely hardwired. Exposure to adult speech is needed only to act as a switch to turn the mechanism on. The second view asserts that the rules of language are extracted statistically through listening. Bolstering this idea, artificial neural networks have been trained to categorize words and infer rules of syntax simply through passive exposure to language.

While these two models certainly capture some aspect of language acquisition, they cannot be the whole story. After all, apes, housecats, and iguanas have neural networks in their skulls, but they do not learn language even when raised in human households. A bonobo ape educated at Eton or Cambridge would still be an ape without language.

According to the third view, the competence to acquire the rules is innate, but exposure is needed to pick up the actual rules. This competence is bestowed by a still-unidentified “language acquisition device,” or LAD. Humans have this LAD. Apes lack it.

I favor this third view because it is the one most compatible with my evolutionary framework, and is supported by two complementary facts. First, apes cannot acquire true language even when they are treated like human children and trained daily in hand signs. They end up being able to sign for something they need right away, but their signing lacks generativity (the ability to generate arbitrarily complex new combinations of words), function words, and recursion. Conversely, it is nearly impossible to prevent human children from acquiring language. In some areas of the world, where people from different language backgrounds must trade or work together, children and adults develop a simplified pseudo-language—one with a limited vocabulary, rudimentary syntax, and little flexibility—called a pidgin. But the first generation of children who grow up surrounded by a pidgin spontaneously turn it into a creole—a full-fledged language, with true syntax and all the flexibility and nuance needed to compose novels, songs, and poetry. The fact that creoles arise time and time again from pidgins is compelling evidence for an LAD.

These are important and obviously difficult issues, and it’s unfortunate that the popular press often oversimplifies them by just asking questions like, Is language mainly innate or mainly acquired? Or similarly, Is IQ determined mainly by one’s genes or mainly by one’s environment? When two processes interact linearly, in ways that can be tracked with arithmetic, such questions can be meaningful. You can ask, for instance, “How much of our profits came from investments and how much from sales?” But if the relationships are complex and nonlinear—as they are for any mental attribute, be it language, IQ, or creativity—the question should be not, Which contributes more? but rather, How do they interact to create the final product? Asking whether language is mainly nurture is as silly as asking whether the saltiness of table salt comes mainly from chlorine or mainly from sodium.