I, meaning me, say there.

There has been a flurry of publicity in the New York Times and The Chronicle of Higher Education about Daniel Everett's new book, Language: the cultural tool. The hype has to do with an old controversy (see a post from 5 years ago: Just the Facts) concerning Everett and Noam Chomsky over recursion. Chomsky (along with Marc Hauser and Tecumseh Fitch) published a paper claiming that the only feature of language that might be limited to humans was a recursive structure. The exact nature of recursion is open to some confusion but is generally taken to embed words or phrases into sentences. Everett reported on a people living in the upper Amazon, the Piraha, whom—Everett said—did not have recursive language, and therefore challenged Chomsky's view of what makes language language.

I can hardly believe that Everett's whole argument against the "language instinct," using Steven Pinker's term, rests on the presence or absence of recursion in Piraha language. I've always thought that making recursion language's sine qua non was absurd. The difference between chimpanzee and human communication is not that chimps converse without recursion. At the same time, recursion is a feature of linguistic structure and it would be astonishing to discover a people who don't take advantage of the capacity. After all, recursion is a way of clarifying what you're saying: That man, the one with the red hat, made a rude gesture at me. Am I to accept the claim that Piraha cannot make themselves more clear?

Actually, I don't have to. Apparently, the corpus of known Piraha sentences includes His mother, Itaha, spoke which is simple recursion. But can Everett's argument really be so simple that this small bit of speech really falsifies his claims? A check of the Amazon page shows many blurbs from serious people, so I'm hoping that Everett's argument is more profound than the hype suggests.

I haven't gotten my hands on the book yet and am opening up this post to comments from readers who have seen Everett's new book. If I manage to snag a copy, I'll pass along what I learn. Meanwhile, I'm returning to my long commute.

Comments

[Trapped-by-the-box.blogspot.com]

I often feel that the great “recursion” debate misses the point. Just ask how might an animal best classify the world, using a parallel network of identical units to make decisions about food, welfare, other members of the social group, etc. The approach I use in The Evolution of Intelligence – From Neural Nets to Langauge suggests that recursive storage structures in a parallel processing framework is the natural way which animal brains work.

The real problem with recursion and human language is that language is a serial rather than a parallel process. To communicate using language it is necessary to convert recursive parallel thoughts into a compact sequential sequence of noises. Such a transformation is a non-trivial task and will require much more “temporary working memory” that normal internal recursive processing. Similar factors apply to converting a serial message back into an internal recursive/parallel form.

This means that the recursive features of language merely reflects the fact that the ideas being transmitted are already stored in a recursive manner internally.

If you think of the human brain being a stretched animal brain the relevant difference between the human and a chimpanzee brain may only be the effective size of the working memory areas – rather than any special genetic difference in the inherent functions the neurons carry out. Give the chimpanzee a brain with more "working space" and perhaps it could communicated in a serial mode using recursion. Without more capacity it is probably limited to communication in "nouns".

[John Roth]

Frankly, I think this is a tempest in a teapot. I agree with the previous commenter that the brain represents objects, actions, attributes and so forth in some kind of a linked network representation. Linearizing this kind of network is a standard computer science task, as is recreating such a network from a linearized version. Thus it's no surprise that human language has the ability to embed; it's also no surprise that human beings tend to have difficulty if there's too much embedding, or if it's the wrong type.

It's also possible to dump such a network in a form that doesn't include recursion or embedding; that is, recursion is not a necessary feature of externalizing such a network; it's simply highly convenient.

To see this, consider the difference between a = b + c*d and x = c * d, a = b + x. One includes recursion, the other doesn't.

The fact that Piraha uses little to no embedding while all other known languages use a lot of embedding is interesting, and cries out for some kind of an explanation. I don't, unfortunately, expect that explanation until appropriate fMRI and similar studies are done, and I don't expect those studies any time soon.

[Raymond Weitzman]

The notion of recursion developed as a formal way of representing the ancient and hoary grammatical concepts of modification and coordination. If element B modifies element A and element C modifies element B, then the whole structure with elements A,B,and C is a hierarchical one. Tree structures don't adequately indicate this hierarchical relation without some additional conditions. Parenthesis do. For example: (A(B(C)) or (((C)B)A). Originally modification was intended to be interpreted as a semantic relationship, where the modifying element added some kind of meaning to the element modified.

On the other hand, coordinated elements were hierarchically on the same level as might be represented with parenthesis as ((A)(B)(C)), which semantically meant they contributed equally to the meaning of the sentence or in some interpretations, functioned in the same way.

But the stipulation that recursion is some kind of innate feature of language embedded in the human mind requires having independent warrants beyond the phenomenal occurrence of modifying and coordinating elements in speech utterances.

Parsimony would suggest that a recursive mental device that somehow evolved in the human species and no other species is not very likely.