The Guardian interviewed Dan Everett while he was in the UK recently for lectures in Edinburgh and London, and has published a piece about Dan and the Pirahã. The Language Log fan who was the first to point it out to us (thanks, Rachele) asks about its example of recursion. It says:
Chomsky … recently refined his theory to argue that recursion — the linguistic practice of inserting phrases inside others – was the cornerstone of all languages. (An example of recursion is extending the sentence "Daniel Everett talked about the story of his life" to read, "Daniel Everett flew to London and talked about the story of his life".)
Is that recursion? Well, unfortunately the matter isn't clear. Let me explain.
The fact is that the paper alluded to, the one by Hauser, Chomsky and Fitch in Science Vol 298, Issue 5598, 1569-157 (22 November 2002; see this Language Log post to catch up on the history), doesn't give any clear definition. Four different kinds of syntactic construction have occasionally been mentioned as relevant.
1. Some people seem to use the term "recursion" to refer to what I would call self-embedding: embedding a phrase inside another larger phrase of the same type in such a way that some of the larger phrase is to the left of the smaller phrase and some is to the right:
[The rat that [the cat that [the dog chased] dragged in] is still there behind the couch.]
That's a relative clause (that the dog chased) in the middle of a relative clause (that the cat that the dog chased dragged in) which is itself in the middle of another clause (The rat that the cat that the dog chased dragged in is still there behind the couch). It's hard to understand even for us, and our language is supposed to have recursive embedding of clauses. (Technical note: the existence of an unbounded set of configurations of this sort is sufficient to show that a context-free set of strings over a vocabulary of two or more symbols is non-regular, but not necessary.)
2. Other people call it recursion when you embed a phrase inside any other larger phrase, whether it is the same type or not, so that some of the larger phrase is to the left and some is to the right:
[The rat that [your [astonishingly aggressive] cat] dragged in] is still there behind the couch.]
That's an adjective phrase (astonishingly aggressive) in the middle of a relative clause (that your astonishingly aggressive cat dragged in) in the middle of a main clause (The rat that your astonishingly aggressive cat dragged in is still there behind the couch). Much easier to understand. (Technical note: Notice that one could have this kind of thing without any explicit depth restriction and yet have a finite set of strings overall. It would all depend on whether there was a chain of phrase types that could lead eventually to a phrase of some type X being included somewhere within a larger phrase of type X, no matter how far down. The existence of an unbounded set of configurations in which this does happen is necessary and sufficient to show that a context-free set of strings over a vocabulary of two or more symbols is non-regular.)
3. Still other people call it recursion when you attach a phrase as the first or last part of a larger phrase. In this example the additions dangle on the end:
[the rat that [was dragged in by [the cat that [was chased by the dog]]]]…
That's result of adding that was dragged in by the cat to the noun phrase the rat, but at the end of the phrase, and then adding that was chased by the dog to the nound phrase the cat, but again at the end. "Recursion" of this kind, sometimes called tail recursion, can be found in regular languages. It is quite easy for us to parse. So is addition of subordinate parts of a phrase at the beginning, which English allows in genitive noun phrase determiners of noun phrases:
[somebody's [brother's [wife's [aunt's [best friend]]]]]
4. Finally, some people might call it recursion when you add phrases to phrases and attach them with coordinators like and:
[[the dog chased the cat] [and the cat dragged in a rat] [and the rat is still there behind the couch]].
This is a coordination, with clauses as the coordinates. You could say that the coordinate clauses are embedded inside the unit made up by the whole coordination, or you could argue that the coordinates are just loosely strung together in a sequence. But you can add as many of them as you like, and they do seem to form a sentential unit. This is the construction the Guardian writer uses for exemplification (and he didn't get it from Everett, who never cites coordination to illustrate recursion).
According to Everett, the Pirahã language doesn't seem to have any of these types of construction, not even clause coordination. So although the Guardian writer used the fourth and weakest kind of example, and many linguists would say it does not illustrate recursion at all, that turns out not to matter very much. In a more serious piece of work it would be crucial to keep these concepts separate and be clearer about what is claimed; but here, in an article for ordinary newspaper readers, the term "recursion" has been thoroughly attenuated: nothing much more seems to be meant by it than "making sentences longer by repeatedly adding stuff of the same sort".
[Written on the evening of November 10; revised slightly on the morning of November 11.]