I got a letter the other day that presented a theory of language my correspondent wanted me to consider. Regrettably, I doubted his idea could be accomplished by plain old evolution and I told him that I cannot take seriously any account of language that requires a miraculous beginning.
Years ago I read Stephen Pinker’s famous book The Language Instinct. Pinker is a gifted writer and his book is filled with interesting and entertaining facts, but I could never persuade myself to even consider his account of how language works. His system requires a set of modules in the brain for generating sentences. If language had been evolving for 15 or 20 million years I might have said OK, but we have had nowhere near enough time to evolve and perfect the linguistic modules Pinker talks about.
Back in 1970 it struck me that evolution had played an important part in language origins and ever since that time I have required ideas about language origins to make evolutionary sense. So naturally I was delighted when reading Daniel Dor’s chapter, “The Evolution of Language and Its Speakers,” in The Instruction of Imagination, I came across this passage:
The question of the evolution of language… [is] the most crucial bottleneck that any theory of language should be able to squeeze through…. For every theory of language we should thus ask: how is this evolvable? [p. 184]
What makes a language evolvable?
First, language should be adaptive; it should solve some problem. This may seem like an easy one, as language is so generally useful, but we always run into the fact that only humans have language. Why don’t many other species converse, if language is so darned useful? The most obvious answer is that it solves a problem faced only by humans. In writing this blog, I have found the problem of adaptation peculiarly vexing. The work of Michael Tomasello has been particularly helpful. His thesis is that humans alone live in dependent communities where we share things or die.
Many thinkers try to finesse the issue by saying some primary support for language (e.g., recursive capacity, control of the lips, the ability to breathe irregularly while talking, etc.) evolved for some reason other than language and was co-opted by language. But this only pushes the question back a step. Why did that first step evolve? Many scholars seem willing to say the reason is unknown. Dor closely examines the adaptive issue in the notorious Hauser, Chomsky, Fitch paper of 2002 in which the authors propose that language depends on a recursive capacity that may have evolved originally for some task other than generating syntax. Dor groans:
There is nothing here but a weary and desperate attempt to keep the essence of language … away from … evolutionary expression. Of course, capacities may evolve for one function and then be adapted for others, … but in such processes the capacities evolve and change to fit their new functional contexts: they do not stay the same. 
Many attempts at evolutionary accounts begin with a genetic mutation that produces some novel benefit. But Dor is strong on the point that mutation need not be the trigger. The environment to which an organism is well adapted may change, tossing the creature into a sink or swim setting. New patterns of behavior will arise and, if they prove helpful, the genes will bring up the rear and evolve to support the new behavior, turning something burdensome into something easy.
This explanation may seem scandalously Lamarckian, urging evolution based on traits developed behaviorally rather than genetically. I wondered if Dor was going to say the giraffe got its neck by stretching. But there is a sound, Darwinian process in which behavior can precede genetic change. Known as the Baldwin Effect, it allows for selection based on behavior. Suppose a species is confined to a single mountain valley and the valley suddenly becomes uninhabitable. Some of the species members die, and some are able to find a mountain pass that lets them escape to a new environment where they (with difficulty) survive. The members that died are out of the story. The descendants come from those who for whatever reason escaped and over time we would expect a series of genetic changes that allow them to do well enough in the new setting.
There have been many attempts to explain language origins in Baldwinian terms, but the work of Morten Christiansen and Nick Chater has established an important limitation. The genetic support for the new behavior comes slowly, but cannot come so slowly that the behavior often changes before the new genetic structure can be fixed. I do not think Dor’s account is clear enough on the limits of flexible (or plastic) behavior.
The qualities that made Latin Latin, for example, could not be genetic because well before the Latin genes could be fixed the language moved on to become French, Spanish, Romanian, etc. This limitation puts the rules of vocabulary, syntax, and phonetics outside the range of the Baldwin effect, and since those rules are the main elements of any language, we have to wonder what is left to evolve.
The answer is surprisingly rich. The psychological supports for language are universal and they include our compulsion to learn to speak (when young) and our compulsion to speak when we witness something amazing, our ability to remember what we are talking about so that we can understand pronouns, our ability to organize the directing of another’s attention, and so on. All of these features could trace to Baldwinian plasticity rather than a mutation.
One complaint I do have about this critically important chapter is that it does not devote enough attention to the issue of selection. Can it really be true that a way with words is important to an individual’s survival capacity or ability to leave descendants?
My mother once told me that she fell in love with my father partly because of his unusual way with words, yet I still have a hard time believing that glibness is a trait that increases descendants. Many inarticulate adults have large families.
On the other hand, I can readily imagine a group of people out-surviving their neighbors because their linguistic skills make them better able to act cooperatively. Ergo, I lean toward group explanations of linguistic selection.
Dor does not have to do the same, but I wish he had more fully addressed the problem of how and why language-supporting traits were selected. He brought such diligence to the rest of this chapter.