This morning in Utrecht got off to a heavy start as Morten Christiansen presented a fulsome talk about “Brains, Genes, and Language Evolution.” It was the strongest counter yet to the language faculty arguments presented at the Evolang conference; however, Christiansen did not present a hard-core, old-fashioned blank slate argument. Genes have their contribution to make, although it is not at all like contribution expected by the hard-core, old-fashioned innate-ideas defenders.
Christiansen began by distinguishing between “arbitrary” features and “functional ones.” The terminology was a little confusing but the distinction is important to his account.
- Arbitrary features are those that, given the laws of nature, could have been different. Nodding one’s head to indicate “yes,” for example, is an example of an arbitrary behavior. You could not your head to say “no” without having to rewrite the laws of nature. Why not call them “cultural” features? Probably because of the quarrel over a universal grammar. A universal feature of language might still be arbitrary in Christiansen’s sense even though Chomsky and his school would not call it cultural.
- Functional features are those that must take into account the laws of nature. An eye, for example, must be built in a way that focuses light on the retina.
A critical fact about functional features is that once established, natural selection works to conserve them. Cats eyes are built for the dark; ours are not. Yet both of them need to keep a similar shape, lens, and relative distance in order to focus the input on the retina. Meanwhile, arbitrary features can change without natural selection bothering its pretty head. They are subject to evolutionary drift, so there does have to be some way that members of the group that shares arbitrary signals can keep in step with the drifting code.
The first part of Christiansen’s presentation made this point about the contrasting evolutionary processes of arbitrary and functional features. The arbitrary ones change too fast for the genes to keep pace. “Genes cannot catch a linguistic moving target,” one of the slides said. This point is particularly important for the Baldwin effect, a way that learning can become fixed in the genes.Arbitrary features change too fast for the Baldwin effect to establish a feature in the genes. (This point has been mention several times before on this blog.)
He also made the important point that if a population did fix an arbitrary feature, population divergence would later lead to “multiple, different ‘universal’ grammars.” When Columbus arrived in the New World he and the Caribs were able to communicate. Native Europeans and Americans were even able to learn one another’s languages. If universal grammars really did fix arbitrary features, we would expect some confusion to arise from drift in arbitrary features during the 50,000 years (over 2,500 generations) the two populations were apart.
On the other hand, functional (non-arbitrary) features might have become genetically fixed. These might include the ability to articulate and hear speech sounds. So, says Christiansen, we are not born with a brain that evolved to learn language’s arbitrary features.
Yet, it is a fact that infants learn to speak with the greatest of ease. How can that be? In this regard, Christiansen is a student of Terrence Deacon: cultural transmission has shaped language to be as learnable as possible by human learning mechanisms. (However, Deacon put much emphasis on the Baldwin effect, which Christiansen rejects.)
Christiansen gave an extended example by considering “sequential learning,” e.g., the ability to learn how to put words, one after the other, to form a sentence. Non-human primates are not good at learning hierarchically ordered sequences, even though they are great at learning to do one thing and then another. (Those who want a detailed account of how the learning experiment worked can read here and don’t overlook the methods section)
The language evolution process is the opposite of evolving an eye, which is constrained by the laws of nature concerning light; language is constrained by what language learners can manage. Linguistic forms that fit the biases of the brain are more readily learned, and spread more effectively from speaker to speaker. Instead of the brain adapting to language (as happens with light), language adapts to the brain.
This rule is not a fixed one like e=mc2. The FoxP2 gene for example could allow for some such learning. (I thought Christiansen got himself into a little trouble with this deviation. He was relying on the original very recent date for the gene, when much of language would have already evolved. I noticed that Roebroeks in his Neanderthal presentation used the much older date of Diller & Cann that puts the FoxP2 gene’s change at the start of Homo erectus, when language was likely in its infancy.)
As to language universals, there are none.