Could Language Modules Have Evolved?

Unsynchronized clocks were spotted by a New York Times photographer (story here) this week, providing a handy visual metaphor for the way biological and linguistic evolution are out of synch.

Language changes too rapidly to permit the biological evolution of the kind of language module in the brain proposed by Steven Pinker and perhaps the co-evolution of language and brain as proposed by Terrence Deacon, according to a paper published this week in Publications of the National Academy of Sciences (PNAS) by Nick Chater, Florencia Reali, and Morton H. Christiansen. (See "Restrictions on biological adaptation in language evolution," abstract here.) I did notice, however, that their calculations do not exclude Noam Chomsky’s anti-communication theory of language origins.

Pinker’s theory was set down in his popular book, The Language Instinct, It proposes that people learn to speak because they are born with language-processing abilities (modules) in their brains. I admit to having always thought the idea was ridiculous because I doubted there was time to evolve and fine tune a specialized module like that. Pinker would reply that there is evidence of specialized speech regions from stroke. Many strokes result in speech deficits

Stroke evidence, however, is ambiguous. First, strokes in different parts of the brain produce different deficits, suggesting there are either multiple modules or that language is put together through the interaction of more general brain structures. Second, strokes that cause language-only damage are rare, suggesting that language emerges from more general features of the brain. Imaging work has further lent support to the view that language rests on more general features. (See: Brain Circuitry Challenges Linguistic Models)

Nevertheless, Noam Chomsky found an underlying logic that argues the empirical evidence must be wrong and modules must exist:

• Universal constraints on behavior must be built in rather than learned.
• Syntactic statements contain a unique structure of their own, revealing constraints that are universal to all speakers, no matter what their environment.
• Therefore, the logic of syntactic structures must be built in rather than learned.

Pinker’s theory of modular instinct was an attempt to give Darwinian heft to Chomsky’s work. It is ironic, therefore, that this week’s PNAS paper challenges Pinker’s theory of language origins without challenging Chomsky’s. The reason is simple: Chomsky is no Darwinian and never worried about giving his ideas biological respectability; Pinker is less radical and did seek to make Darwin and Chomsky compatible by saying that the brain’s innate capacity to generate syntactically correct sentences is an evolutionary adaptation.

While I was scoffing at the notion of a speedy evolution of properly functioning language modules, others were trying to think of some way the process could be made to work. The only good idea found was Baldwinian evolution, a Darwinian process by which a change in behavior leads to a genetic change. One example offered is the built-in calluses on ostrich knees. Ostriches kneel and, therefore, develop calluses. Natural selection favors the more able callus developer. The competition persists for generations until finally selection favors the ostrich born with calluses in place.

A similar process might produce syntactic modules:

• People begin speaking, for whatever reason.
• Natural selection favors the better speakers.
• Over time, the genetic factors that make someone a better speaker spread throughout the population.
• The above process repeats itself many times, turning cultural peculiarities into genetic features.

This sequence has the strength of being at least plausible, and if Pinker had argued this way in his book I probably would have taken his case more seriously. But Chater & Christiansen present calculations that work against the idea. The problem is one we know from history: language evolves more quickly than natural selection.

The authors ran simulations to compute the answers to a series of questions.

Can the Baldwin Effect produce language modules? Yes. A lineage will genetically improve its ability to learn a language, if the language is stable through the generations.

Will the module appear if the language itself changes? Not unless the genetic mutation rate is greater than the rate of language change. The Baldwin Effect is “maladaptive” in cases where language changes more quickly than the module. These simulations reject the Pinker model.

Can the module co-evolve with the language’s own evolution? Only if the process begins with a very strong module already in place. Terrence Deacon proposed that while language was shaping the brain, the brain was also shaping language. The simulations show, however, that co-evolution will only work if the genetic component is already at least 50% of the story. In that case, the Baldwin Effect will sharpen the existing genetic bias, but this situation relies on what we were trying to explain in the first place, the presence of a linguistic module in the brain.

Calculations, however, carry only so much weight in argumentation. They do not confront the logic that kicked off the original search for a way to evolve a language module. The proposition that universal constraints on behavior must be built in rather than learned rests on an old philosophical distinction between innate ideas and learning from experience. The debate was hot and heavy back in the 1600s and then faded as the experience side of the argument carried the day. Fifty years ago, however, Chomsky picked up the innate idea position more or less as René Descartes had left it 300 years earlier.

Many things happened between Descartes’ demise and Chomsky’s rise, especially the introduction of a biological theory that says humans are descended, with modification, from apes. Even more recently it has been established that although apes in the wild do not use words, they are smart enough to do so in situations where the other party cares to pay attention. Determining just what that ape capacity means is one of the major tasks of this blog. For now, I’m limiting my comment to saying that the second part of the logical argument—syntactic statements contain a unique structure of their own—is probably overstated.

Note: This blog has previously reported on a Christiansen & Chater paper that presented their idea of where language came from. See: The Richness of the Stimulus and Language Adapted to Us.

Comments

[JoseAngel]

Every time I read this blog I learn something new. Congratulations!