Discussing the concept of recursion is like a rite of passage for anyone interested in language evolution: you go through it once, take a position and hope it doesn’t come back to haunt you. As Hannah pointed out last year, there are two definitions of recursion:
(1) embeddedness of phrases within other phrases, which entails keeping track of long-distance dependencies among phrases;
(2) the specification of the computed output string itself, including meta-recursion, where recursion is both the recipe for an utterance and the overarching process that creates and executes the recipe.
The case of grammatical recursion (see definition 1) is perhaps most famously associated with Noam Chomsky. Not only does he claim all human languages are recursive, but also that this ability is biologically hardwired as part of our genetic makeup. Countering Chomsky’s first claim is the debate surrounding a small Amazonian tribe called the Pirahã: even though they show signs of recursion, such as the ability to recursively embed structures within stories, the Pirahã grammar is claimed not to recursively embed phrases within other phrases. If true, then are numerous implications for a wide variety of fields in linguistics, but this is still an unsubstantiated claim: for the most part, we are relying on one specific researcher (Daniel Everett) who, despite having dedicated a large portion of his life to studying the tribe, could very well have been misled. That said, I retain a large amount of respect for Everett, having watched him speak at Edinburgh a few years ago and read his book on the topic: Don’t Sleep, There are Snakes: Life and Language in the Amazonian Jungle.
So, why am I rambling on about recursion? Well, besides its obvious relevance, — and perhaps under-representation on this blog (deserved or not, I’ll let you decide) — Everrett has recently published a series of slides about a corpus study of Pirahã grammar (see below).
[gview file=”http://tedlab.mit.edu/tedlab_website/researchpapers/piantadosi_et_al_piraha_lsa_2012.pdf”]
His tentative conclusion: there is no strong evidence for recursion among relative clauses, complement clauses, possessive structures and conjunctions/disjunctions. However, there is possible evidence of recursive structure in topics/repeated arguments. He also posits cultural pressures for longer or shorter sentences, such as writing systems (as I mentioned way back in 2009).
I’m sure this debate will be brought to the fore at this year’s EvoLang, with Chomsky Berwick Piattelli-Palmarini and many of the Biolinguistic crowd in attendance, and it’s a shame I’ll almost certainly miss it (unless someone wants to pay for my ticket… Just hit the donate button in the left-hand corner 😉 ).
Update: Dan Everett highlighted there are other, perhaps more exciting mediums to explore his ideas (see comments, but I still feel you can’t beat a good slideshow). So, to add some jazz to this post, below is a video of the forthcoming film he mentioned, The Grammar of Happiness:
There are other things coming out on this issue. One is a film, which has just won the Young Europeans Jury Award at the FIPA film festival in Biarritz. Chomsky, Pinker, and others appear in the film.
http://www.fipa.tm.fr/en/programs/2012/the-grammar-of-happiness-23616.htm
The other is a new book, to appear in March, Language: The Cultural Tool:
http://www.amazon.com/Language-Cultural-Daniel-L-Everett/dp/0307378535/ref=sr_1_1?s=books&ie=UTF8&qid=1324403805&sr=1-1
Thanks for pointing this out. I look forward to reading the book and watching the film. It’ll be good to see these ideas reach a wider audience — at the very least to readdress the long shadow cast by Chomsky on the issue.
Also, congrats on the award.
This certainly will come up at EvoLang, I’m sure. Unfortunately, Chomsky won’t be there to hold his ground:
http://kyoto.evolang.org/content/plenary-speaker-program-changes-0
Massimo Piattelli-Palmarini… He who is co-author of What Darwin Got Wrong? I’m sure that talk will be a riveting example of being hoisted by one’s own petard. I wonder if history will come to ask the question: what did Fodor (and Piattelli-Palamarini) actually get right? Followed by the addendum: a classic case-in-point of referential opacity
I was at this presentation at the LSA, and I am still scratching my head about what it is supposed to show.
What they have done is gloss the sentences of a number of Pirahã narratives in keeping with Everett’s recent proposals that what he used to view as subordinate clauses are actually independent sentences (see slide 15). Then they counted the number of subordinate clauses in these glosses, and – surprise, surprise! – that number is zero.
Apparently, to judge from the response in the question period to my question about this, the same tactic was applied to relative clauses. These were analyzed as corelatives in earlier work (e.g. “What Mary ate, I want it too”), but now get analyzed as independent sentences (“Mary ate it. I want it.”). So we look through the corpus for embedded relative clauses, assuming the new analysis under which they don’t exist, and – surprise, surprise once more! – that number too is zero.
Certainly, if I analyze a language as lacking X, and then I start counting X’s on the assumption that my analysis is correct, I will come up with zero instances of X. What, however, is this supposed to teach us, besides the fact that zero means zero?
Right, there seems to be a lot head scratching going on, so let me get this straight: you’re saying that Everett is counting subordinate clauses as independent sentences, and he’s doing the same for relative clauses etc. Put simply: he’s not looking for these, so he obviously won’t find them. Couldn’t you equally say that when we look for subordinate clauses with an a priori assumption, then we’ll find subordinate clauses if we look hard enough?
“Put simply: he’s not looking for these, so he obviously won’t find them.”
No, it’s worse than that. They’ve constructed the data set so there won’t be any subordinate clauses in it and then they present the absence of subordinate clauses in their data set as if it were a finding.
(or at least I think that’s what they’ve done)
Okay, that’s sort of what I was getting at… But it still doesn’t avoid the larger issue: that you could construct the dataset to have subordinate clauses in it!
Really, it comes down to the issue of hypothesis testing: we need a way of testing these competing hypotheses to explain the data, which probably means we also need to incorporate linguistic data outside of just the syntactic realm. This includes any prosodic cues and discourse features (such as pauses) that might shed light on the syntactic structure. Still, having said that, I’m relatively lay about what has and hasn’t been done in studying the Pirahã language.
I think other people have pointed this out, but suppose we take Everett’s no-recursion claim at face value. So what? So Pirahã doesn’t make use of recursion. I’m sure it also doesn’t make use of the phone [θ] or a case system with 90 distinct agglutinating case morphemes. In fact, I think it’s pretty clear that *most* languages don’t use [θ], and that no language has 90 such case morphemes. We don’t find that fact surprising, nor should we: we don’t believe that these things are *necessary* for what constitutes any given language. The same is true of recursion of the relevant sort. I mean, obviously there is the trivial recursion that all languages obviously have — an expression containing pieces that are also expressions of some form. Even Pirahã has this, I think Dan will agree. They just don’t have the fancier sort where sentences are inside sentences, and so forth. Well ok, but no one thinks that *that* is an essential component of a human language. Certainly not Chomsky, who’s notion of recursion is simply “some expression formed from smaller expressions”. Put two words together and you have a recursion of precisely the sort Chomsky talks about. Under old school PS rules this just amounts to saying that Pirahã *happens* to lack a symbol X for which some sequence of rule applications can lead to another X (e.g. S -> NP VP, VP -> V S). But noone ever thought *that* was an inherent property of language. In modern parlance it’s just a matter of what features a word has, (e.g. V[uC], T[uV], C[uT]). But again, no one thinks that your lexicon *must* have entries which, as a whole, have this cyclic nature to them. What we have always taken to be the crucial thing about language is that such a cyclic structure is *possible*. And Chomsky, I think it’s clear, is now saying that this is *all there is* to what makes language special. Whether it’s deployed or not is irrelevant — humans can deploy that capacity if they need to, honey bees cannot. Hell, if you want to be especially true to Chomsky’s word, then it’s obviously not true that this is all you have — he’s said numerous times that the Strong Minimalist Thesis is almost certainly wrong — but you can’t know how it’s wrong until you try to push it far enough for it to fail.
It was a pity about Chomsky’ s evolang cancellation, but I’m sure the topic of recursion will come up anyway (most likely multiple times!). Unfortunately can’t afford to shout anyone a ticket, but I am hopeful we will be able to record some of the discussions and put them online.
@Darryl McAdams: Doesn’t that surely bring the debate back to whether recursion is unique or not to language? In the sense your describing it, then there is no reason why we should expect it to be — and it is instead a domain-general (or domain-independent) heuristic device (something I don’t think Everett would necessarily disagree with).
@Luke McCrohon: That’ll be great if you can get some videos up. I’m sure one or two of the people here will be able to provide some coverage of the conference for the blog.
@Wintz, oh yes, definitely, and I think Chomsky is simply wrong on this matter — I think there’s good reason to think that recursion is necessary for all sorts of complex cognition that animals do, especially in things like planning hunts, or ferreting food out of places that it’s hidden. I mean, just look at crows — they can do some absolutely amazing cognitive feats, looking at a series of tubes with tools in them, and without trying anything, take out the right tools in the right order so that they end up with the tool they need to eventually get the food (e.g. http://www.youtube.com/watch?v=ZE4BT8QSgZk). I don’t know how crows can do this if not by having some recursively structured problem solving techniques.
However, Chomsky might be right in that *recursive mental representations* might be uniquely linguistic — there’s a difference between recursive *thoughts* and recursive *computations*. But I think that this is an empirical matter that is completely distinct from the Pirahã question, because the Pirahã question is about whether language is or isn’t recursive, either necessarily or potentially, which is clearly not the same question as is recursion unique to humans. It could still be unique to humans, and the Pirahã don’t have to make use of it, in the same way that honeybee dance is unique to a handful of species of honeybees but no honeybee will ever use the dance if it’s isolated from birth from any other honeybee.
There’s a lot of work going on about recursion in Brazilian languages and a conference planned
for Rio next year by Marcus Maia. Exactly how recursion is captured across a variety of
structures in language-particular ways becomes a more interesting question than the fact that
some form of recursion is required for repeated Merge for every language.
In particular, in acquisiition, conjunction seems to precede recursive embedding where
a Phase, non-phase alternation occurs as in sentence-embedding, possessives, recursive
compounds (coffee-maker-maker).
This film seems to be from a Swedish TV documentary I saw.
http://www.svtplay.se/video/367509/22-del-2-av-3
You need a Swedish proxy in order to see it.