« previous post | next post »

Recently, a disagreement about the syntactic analysis of certain aspects of an obscure language has achieved an unusual degree of public interest: Tom Bartlett, "Angry words", The Chronicle of Higher Education, 3/20/2012; Jenny Schuessler, "How do you say 'disagreement' in Pirahã?", NYT, 3/21/2012; etc.  Of course, as those articles explain, this is all part of a broader controversy about the nature of language, whose latest round was kicked off by  the publication of Dan Everett's new book, Language: The Cultural Tool.

Geoff Pullum's latest Lingua Franca column, "The Rise and Fall of a Venomous Dispute", puts this dispute into historical and intellectual perspective. If what you've learned of the squabble's linguistic, philosophical, or political aspects interests you at all, Geoff's essay is the thing to read.  In case you want more, I've collected a list of links below.

There are many fascinating questions about the Pirahã language and culture, which unfortunately we won't learn much more about as long as Dan Everett is prohibited from working with the Pirahã. (His exclusion also prevents him from bringing other scientists to work with them, as he has done with Michael Frank, Ted Gibson, and Peter Gordon, among others.) But as Geoff observes, there's little intellectual (as opposed to political) substance left in the analytic controversy described in the recent articles.  In particular, the controversy about the syntax of Pirahã has effectively been conceded by Everett's opponents, since they've apparently clarified their position in such a way that his arguments no longer apply to it. [Update — as Dave Pesetsky notes in the comments, he and his colleagues no longer believe that if Pirahã lacked subordinate clauses, it would be a problem for Chomsky's current theory of "universal grammar"; but they haven't given up the argument that Pirahã does have subordinate clauses all the same.] In any event, here are the manuscripts and published papers involved in that dispute:

Daniel Everett, "Cultural constraints on grammar and cognition in Pirahã",  Current Anthropology 2005.
Andrew Nevins, David Pesetsky, and Cilene Rodriguez, "Pirahã Exceptionality: a Reassessment", ms. 2007.
Daniel Everett, "Cultural Constraints on Grammar in Pirahã: A Reply to Nevins, Pesetsky, and Rodrigues (2007)", ms. 2007
Andrew Nevins, David Pesetsky, and Cilene Rodriguez, "Pirahã Exceptionality: a Reassessment", version as published in Language 85.2 2009
Daniel Everett,  "Pirahã culture and grammar: A response to some criticisms", Language 85.2 2009.
Andrew Nevins, David Pesetsky, and Cilene Rodriguez, "Evidence and Argumentation: A Reply to Everett (2009)", Language 85.3 2009

Here are the slides from some as-yet unpublished work taking a broader empirical look at the key question (that used to be) in dispute:

Steven Piantadosi, Laura Stearns, Daniel Everett, and Edward Gibson, "A corpus analysis of  Pirahã grammar: An investigation of recursion", LSA presentation 2012.

Here are some Language Log posts on the general topic:

One, two, many — or 'small size', 'large size', 'cause to come together'? (8/20/2004)
Life without counting throwing (8/22/2004)
The Straight Ones: Dan Everett on the Pirahã (8/26/2004)
On counting and throwing (8/27/2004)
No abstract concepts for them (9/7/2004)
Pica on the Mundurucú (11/1/2004)
Cultural constraints on grammar (3/10/2005)
JP versus FHC+CHF versus PJ versus HCF
Good story, bad headline (5/11/2006)
Parataxis in Pirahã (5/19/2006)
Pirahã channels (5/21/2006)
Fear and loathing on Massachusetts Avenue (11/29/2006)
Dan Everett and the Pirahã in the New Yorker (4/9/2007)
Pirahã color terms (4/13/2007)
Comments on 'The Interpreter' (4/23/2007)
The enveloping Pirahã brouhaha (6/11/2007)
The Pirahã and us (10/6/2007)
Ontological promiscuity v. recursion (2/10/2008)
Typological progress (5/11/2008)
The cognitive technology of number (6/11/2008)
Everett on the Pirahã in The Guardian (11/10/2008)
Kaioá (11/25/2009)


  1. David Pesetsky said,

    March 28, 2012 @ 8:00 am

    I have no idea what you mean by "the controversy about the syntax of Pirahã has effectively been conceded by Everett's opponents, since they've apparently clarified their position in such a way that his arguments no longer apply to it", nor why you've linked to someone's private posting of the first draft of our paper, rather than linking to http://ling.auf.net/lingBuzz/000411, which contains the published version (as well as a link to the earlier version, and the later papers).

    [(myl) With respect to the first issue, I'm relying on the quotations in the Chronicle article:

    And what if the Pirahã don't have recursion? Rather than ferreting out flaws in Everett's work as Pesetsky did, Chomsky's preferred response is to say that it doesn't matter. In a lecture he gave last October at University College London, he referred to Everett's work without mentioning his name, talking about those who believed that "exceptions to the generalizations are considered lethal." He went on to say that a "rational reaction" to finding such exceptions "isn't to say 'Let's throw out the field.'" Universal Grammar permits such exceptions. There is no problem. As Pesetsky puts it: "There's nothing that says languages without subordinate clauses can't exist."

    Were you misquoted? If not, it seems that the broader issue about "universal grammar" evaporates, and all that remains is the question of whether Pirahã in particular lacks subordinate clauses, as Ken Hale thought was the case for Warlpiri and indeed for many Australian languages. But I should not have suggested that you're giving up the argument on the syntax of Pirahã, as opposed to the argument about what it might mean for the nature of language in general; so I'll amend the post to avoid giving the wrong impression.

    As for the citations and links, I started with the 2007 papers because they were widely circulated and discussed, and thus form a meaningful part of the intellectual history, for those who want to re-trace it; I got the link from Google Scholar, which put it ahead of the lingbuzz link for whatever query I used; I've now substituted the lingbuzz link.]

  2. David Pesetsky said,

    March 28, 2012 @ 9:27 am

    Thanks for fixing the link, Mark.

    No, I wasn't misquoted. But I was repeating exactly what we'd written in our paper all these years ago — nothing new, no change of heart, no concession. At the time *we* were the ones claiming that even if Everett were right about the absence of certain types of constructions, there's still no implication for any proposal about UG known to us, and Everett was the one insisting on the opposite. As we wrote:

    "At the same time, although Merge [the rule that builds phrase structure] may IN PRINCIPLE combine any two lexical items or phrases an unbounded number of times, not every imaginable instance of Merge is acceptable in actual languages. There are many restrictions on Merge that constrain the repertoire of structures that individual languages allow. Consider a simple example from English. [examples omitted] … Some languages share these particular restrictions, but many others do not. Those languages that differ from English may, however, impose other requirements […] These facts all represent constraints on Merge, instances in which application of a general rule is blocked by independent properties of the language. These restrictions, and the laws that underlie them, form a continuing topic of syntactic research and debate."

    Consequently, *we* were the ones claiming that "all that remains is the question of whether Pirahã in particular lacks subordinate clauses" and similar issues of syntactic analysis — which is why we left that topic quickly and devoted most of our paper to the question of how Pirahã should actually be analyzed, and what we really learn about language in the process.

  3. Dan Everett said,

    March 28, 2012 @ 10:06 am

    Since my new book is the occasion for bringing back up the Piraha recursion wars, I should say that Language: The Cultural Tool has almost nothing to do with this acrimony (though it will likely cause a different set of discussions). It considers and rejects the gamut of arguments adduced in favor of Universal Grammar/Language Instinct independently of anything going on in Piraha, though Piraha is one of many examples in the book. On the other hand, the new film, Grammar of Happiness, which debuts at 9PM on May 12 on the Smithsonian Channel, does focus a lot on the recursion controversy. I had nothing to do with the script of the film. I answered thousands of questions in the 300+ hours of filming, of which 46 minutes made the final cut. That will bring this up again. I am not shying away from it, but my new work has little to do with Piraha recursion – aside from work that I am continuing with Ted Gibson, Steve Piantadosi, and others at MIT. For me the issue is largely settled. As I suppose it is for David Pesetsky, from a very different perspective of what was settled. (Ray Jackendoff and Eva Wittenburg reach a similar conclusion in a new book that they are writing.)

  4. That's what SHE said,

    March 28, 2012 @ 11:46 am

    Sadly this latest flare-up reinforces the widely-held opinions by outsiders in nearby fields that large chunks of linguistics consist of acrimonious debate about inconsequential minutiae. Due to the attention paid to anything involving C. (who generally seems to be drawn to controversy), a squabble in a tiny area of theoretical syntax without a shred of real-world relevance is giving the entire field a bad name. It's bad enough when I have to explain to computer scientists and mathematicians that the term "recursion", as used by (some) linguists, doesn't have its usual meaning. (Fair enough, neither does "root" or "radix" etc.) But any teachable moment brought about by momentary confusion goes out the window when they discover that the substance-to-nastiness ratio (SNR) is practically zero.

  5. Charles Yang said,

    March 28, 2012 @ 1:16 pm

    On Warlpiri: There *is* evidence for clausal recursion, in the corpus data that Ken Hale and others have compiled. Julie Legate has an article in the journal Syntax (2011) discussing it:


    And an excerpt just on the this notion has been put on Lingbuzz for non-subscriber consumption:


    Unlike Warlpiri, a publicly available corpus of Piraha has not been made available.

    [(myl) This may very well be the correct analysis. My point is just that Ken Hale's view — that Warlpiri and other Australian languages lacked clausal embedding — was (as far as I know) uncontroversial for more than 30 years. It was re-examined and challenged, perhaps correctly, in the context of the argument about Dan Everett's analysis of Pirahã. But if (as Dave Pesetsky observes) it doesn't actually matter, from a theoretical point of view, whether such languages have clausal embedding or not, why scrutinize the question at such length, in precisely that context?

    It's certainly worth asking Piantodosi et al. whether the corpus of stories that they analyze can be published — it certainly should be, in my opinion.]

  6. Rubrick said,

    March 28, 2012 @ 1:42 pm

    I'm curious about one thing: Have any of Prof. Everett's adversaries yet leapt on the opportunity to publish an essay entitled "Dan Everett: The Cultural Tool"?

    I'm not a linguist, and don't even have the scent of a distant puppy in this fight, but I can recognize an opportunity for viciously clever wordplay when I see it.

  7. Fritz said,

    March 28, 2012 @ 2:32 pm

    Let me see if I understand what is going on.

    1. Hauser et al. and Nevins et al. say that recursion is fundamental to human language, by which they mean the 'Merge' operation.

    2. The above scholars do not deny that Merge is subject to restrictions, in particular, different restrictions in different languages.

    3. Everett argues that the restrictions on Merge in Piraha are quite severe — to be specific, there is no self-embedding (S within S, NP within NP, etc.).

    4. So what is the debate about, exactly? If Everett is right about Piraha, then we do not have a language without recursion/Merge, but simply a language in which this operation is subject to unusually strong restrictions.

    What am I missing? Why the polemics?

    [(myl) I'm puzzled for roughly the same reasons. If there's no problem in principle restricting recursive compositionality so as not to permit embedded finite clauses, then the logical response to (this aspect of) Everett 2005 was simply to say "No, look, you've misunderstood what our theory means by 'recursion' — we just mean that all human languages construct more complex messages by hierarchical composition of simpler ones, as happens when a modified noun is used as the subject of an inflected verb." You could go on to add, "… and by the way, we're not convinced by your arguments about the syntax of Pirahã, …" but that would be beside the point.

    That's certainly not the argument made in Nevins et al. 2007. There's a form of that argument (that language-specific restrictions on hierarchical composition are theoretically unproblematic) made in the 2009 version of the paper, as Dave Pesetsky notes in his comment above — but it's in respect to the restriction in German against nested possessives. The (plausible) corollary about possible restrictions on clausal embedding is not expressed, as far as I tell by reading the paper.

    So indeed, the vehemence of the pro-clausal-embedding argument from the anti-Everett side is puzzling. The fact that Dan Everett was excited about the no-clausal-embedding analysis makes sense, since he thought (right or wrong) that it constituted a refutation of an important theory. If in fact he was wrong about the relevance of the analysis — leaving aside whether the analysis is correct — I don't see why Nevins et al. argued so heatedly against the analysis, rather than simply noting its irrelevance.]

  8. David Pesetsky said,

    March 28, 2012 @ 2:51 pm

    Reply to Fritz:

    This is a really strange moment in the discussion, because if you read our paper, your points 1and 2 are exactly what we said in our paper. (See my second comment above.) Everett and his followers kept replying "no, your theory says all languages should have embedded clauses and recursive possessors – no matter what you say". We kept saying, "not true — and therefore there's nothing to debate about this", but we kept getting debated anyway.

    Now, by some weird alchemy of intellectual politics, Language Log and Pullum's blog seem to be accusing *us* of prolonging a debate we long ago said was fictitious, for exactly the reasons you give. It's a strange world we live in, isn't it.

    [(myl) No one here is accusing you of prolonging the debate. It's obvious that the issue flared up again in the popular press because of the publication of Everett's book; and there'll be another round, I guess, when the movie comes out. In this post, I've just tried to give a chronological set of links for those interested in more background.]

  9. Rod Johnson said,

    March 28, 2012 @ 3:18 pm

    David, I think the (perceived) vituperativeness of the anti-Everett side is what is making this controversial. (I make no claim as to whether this perception is correct.)

  10. AntC said,

    March 28, 2012 @ 3:23 pm

    It seems a great brouhaha to make about a lone instance. I'm failing to understand why the Chomsky camp feels Everett's analysis is such a threat, if (nearly) all other languages do exhibit the structure of UG.

    Is Pirahã the only language which is claimed to lack recursion (embedding)?

    [(myl) See the discussion in the links above, e.g. here, with respect to clausal embedding in Australian languages.]

    Is Pirahã related to other languages? Do they exhibit embedding, or have they died out before they could be properly recorded?

    [(myl) The Ethnologue report is here. Basically, it's an isolate.]

    Are there other aspects of Pirahã or the culture of its speakers that would make it a one-off in some way? (How many other cultures lack a foundation mythology, for example?)

    [(myl) The lack of a system of counting numbers seems to be found elsewhere — again see the links in the post above, especially e.g. this one. There are other languages with additional "channels" (as described here), such as whistle speech or drum speech, but the particular design and use of "channels" in Pirahã seems unusual. For other issues, see the links int he post above, and especially Dan's descriptive claims in e.g. his 2005 paper.]

    Has Pirahã society suffered some catastrophic collapse in the recent past (perhaps due to European contact?) such that there was a loss of transmission of its cultural artefacts?

    [(myl) According to the Ethnologue page, "In 1960s the population dwindled to 80 due to high infant mortality, death of mothers giving birth, and disease. Modern medicine helped population growth." I haven't heard an argument that this has anything to do with the state of the language.]

    I guess with the effective ban on research, we're never going to find out.

    [(myl) There are some other linguists who continue to have access, though as far as I know, Dan has been the one who has brought in outside researchers.]

  11. David Pesetsky said,

    March 28, 2012 @ 3:35 pm

    Replying to Mark's comments added to Fritz's: "So indeed, the vehemence of the pro-clausal-embedding argument from the anti-Everett side is puzzling."

    I don't understand what's puzzling. We're syntacticians. We care about what's true and what's false about the structure of human language and specific languages. (That counts as minutiae to some Language Log readers, evidently, but not to us.) We were taken aback at what we found when we compared Everett's new claims to his published data, when we examined the overall strength of his arguments, and when we investigated his repeated claim that Pirahã was a radical outlier in the world of languages.

    In light of the evident public and professional interest in Everett's assertions, we believed that we had a duty to report what we had discovered. Our paper was peer reviewed by Language and accepted upon revision. Though our reviewers were anonymous, they appeared to come from diverse corners of linguistics, and they wrote very thorough and constructive comments. Evidently somebody out there understood what we were doing, and considered it not only publishable but of interest to the broad community of linguists that the journal aims to serve. To put it in a nutshell, our goal was not to weigh in on the topic of recursion, but on the question of whether Pirahã was a bizarre language with properties so extreme that they force us to rethink linguistics — or an interesting but non-shocking recombination of properties already spotted in other languages. Our answer was no, but to reach that answer, we wanted to go through each and every argument to the contrary that Everett had made. Maybe that level of thoroughness comes across as "vehemence", but if so, that's a kind of vehemence the field needs more of — not less.

  12. SK said,

    March 28, 2012 @ 3:44 pm

    @ Fritz: I think part of the reason there has been such a lot of argument over Piraha is that various related issues are being argued over at once, and it's hard to separate them out. And that's just on the question of subordination, before you get into hot-button issues like whether the Piraha have a counting system, a creation myth, etc., and what that says about their culture/outlook on life as a whole. Here's my take on things – others who know this area better may be able to tell if it's along the right lines or not.

    Hauser, Chomsky and Fitch identified something called ‘recursion’ as a feature of the human language faculty in its narrow sense (FLN) – in fact, they take recursion to be the *only* thing in FLN. Now, if you understand recursion to mean something along the lines of syntactic embedding, such as NP inside NP (as in possessives like ‘the boy’s dog’) or S inside S (as in subordinate clauses), then Piraha – as described by Everett – is a language which does not possess recursion. The existence of a human language like that obviously has the potential to damage the claim that human language is in some sense defined by the capacity for recursion.

    So, various questions arise. First of all: is Everett describing the facts correctly? Does Piraha really not have subordination, or is he just misanalysing the material he has collected? Lots of tension has arisen on this initial point, especially because Everett’s earliest descriptions of Piraha do say it has subordination, and he has since changed his mind. This is grist to the mill of the linguists arguing against him, but he can justifiably reply that being able to change your mind on the basis of greater familiarity with the evidence is obviously a good thing.

    Let’s assume that he has analysed the data correctly. The next question is: given that Piraha doesn’t have subordination, how much of a problem is that for the idea of Universal Grammar (whatever you take that to mean)? As far as I can tell, one side says that it’s not a problem – because even if subordination is taken as a part of UG, not every language needs to make use of all the features which UG provides for it. But the reply to that is: in that case, what does it mean to say that something is in UG at all? And how would you prove that something *isn’t* in UG, if its absence from actual spoken languages isn’t sufficient? What kind of evidence would be good enough?

    Finally, there is the point that apparently Hauser, Chomsky and Fitch weren’t talking about anything like subordination at all, but were using ‘recursion’ to mean Merge, which is something along the lines of ‘combine words together to make phrases’. Now, in that case, the Piraha evidence isn’t especially relevant here and the problem evaporates. But it comes as a disappointment to hear that all that Hauser, Chomsky and Fitch were claiming about human language is that it combines words together to make phrases. It would be nice to think they were claiming something more interesting than that. But whatever their claim is, the big issue for lots of interested readers like me is: what kind of evidence could prove it wrong?

  13. Charles Yang said,

    March 28, 2012 @ 3:48 pm

    Re: Warlpiri, Piraha, and recursion. Indeed, as Mark noted, Ken's original position had been around for a long while, apparently uncontroversial. Contrast that with the Piraha mess. The only difference I see is that (1) Ken, who had unparalleled knowledge about "exotic" languages, didn't issue a press release trumpeting the demise of anything, and (2) Ken made the data available so further investigation would be possible.

    [(myl) That's all true (though the Warlpiri data was released over a period of time through effects such the Warlpiri Dictionary Project, mostly after the 1976 publication of the original paper on correlatives etc.; and Dan Everett says that he's in the process of doing something similar with Pirahã materials, and he did facilitate the recent corpus analysis by Piantadosi et al.)

    And it's also plausible that Andrew Nevins et al. were annoyed by Dan's (arguably over-strong) argument against their position, and responded in kind — in speaking with Andrew and others during that period, I certainly detected a good deal of negative affect, which can also be seen in things like the email quoted here.

    But there's one other difference that's quite important: Ken's work preceded by almost 30 years the 2002-2005 exchange among Chomsky/Hauser/Fitch on one side, and Pinker/Jackendoff on the other, over the idea that "the uniquely human, language-specific part of the language faculty … consists only of recursion, and that this part cannot be considered an adaptation to communication" (See "JP versus FHC+CHF versus PJ versus HCF", 8/25/2005, for details and links).

    So even if Ken's personality had inclined him towards press releases, there wouldn't have been anything to release, since there was no special theoretical emphasis on the presence or absence of "recursion" in general (whatever that is taken to mean), and clausal embedding in particular.]

  14. Avery Andrews said,

    March 28, 2012 @ 3:50 pm

    My recollection of what Ken said about Warlpiri is that it lacked 'central' embedding – one clause properly inside another, with overt material on both sides. kutja-clauses (relative) look subordinate, but appear either at the beginning or end of the main clause. Then there are the infinitival clauses (kurra, karra, rlarni; Bresnan & Simpson 1983) which are obviously subordinate in nature, tho I don't know if they've been observed recursing (John watched Mary watching the dog watching the cat ).

    [(myl) Hi Avery — in addition to my (admittedly imperfect) memory, I'm relying on e.g. the discussion in Rachel Nordlinger's paper "Spearing the Emu Drinking: Subordination and the Adjoined Relative Clause in Wambaya", Australian Journal of Linguistics, 26(1) 5-29, April 2006:

    Studies of subordination in Australian Aboriginal languages have been heavily influenced by Hale's foundational paper on the 'adjoined relative clause' — a non-embedded, multifunctional subordinate clause type found in Warlpiri and a 'large number of Australian languages' [Hale K 1976 'The adjoined relative clause in Australia' in RMW Dixon (ed.) Grammatical Categories in Australian Languages AIAS Canberra: 78–105 at 78]. Since this paper, almost every Australian grammar makes some reference to this clause type, presenting a general picture of structural homogeneity across subordination structures in Australian languages, and leading to the general perception that Australian languages typically don't have syntactic embedding. In this paper I present an analysis of subordinate clauses in Wambaya, arguing that these share many features of Hale's 'adjoined relative clause' while still being clearly subordinate. The differences between subordinate clauses in Warlpiri and Wambaya show that complex constructions in Australian languages can be structurally dissimilar while sharing many of the properties of the 'adjoined relative clause' type. I argue, therefore, that clause-combining in Australian languages may be more structurally heterogeneous than is traditionally assumed, and that a single analysis for complex sentences across a majority of Australian languages is quite likely inappropriate. This has implications for both the analysis and description of subordination in Australian Aboriginal languages, and for their relationship to the typological literature on subordination more generally.


  15. J.W. Brewer said,

    March 28, 2012 @ 4:10 pm

    This reminded me (when it came up over on languagehat a day or so earlier) of a somewhat similar situation back in the '70's with a seemingly happier outcome. As told here http://www.lel.ed.ac.uk/~gpullum/derbyshire.html Geoff Pullum made the claim (consistent with the literature at the time) that no attested natural language had a particular feature (O-initial default/unmarked word order for shorthand, follow the link if you want gkp's more carefully nuanced statement of the claim), a thitherto obscure researcher (the late Desmond Derbyshire) raised his hand and politely contended that the language he'd been doing fieldwork on in the Amazon (Hixkaryana) in fact had the property Pullum said didn't exist, and eventually Pullum (and the profession as a whole) was convinced. But when Everett came out of the Amazon a few decades later with a similarly novel claim, more strife ensued.

    Why the differences? Possibly in the earlier situation: 1) the personalities involved were simply less contentious; 2) no theoretical edifice had yet been erected around Pullum's mistaken empirical belief which drew (or was widely if incorrectly taken to be drawing) sweeping conclusions from the seeming non-existence of O-initial default word order; 3) at the time he was doing his fieldwork Derbyshire didn't know the significance of what he'd discovered, i.e. didn't know that the typological literature thought that what he'd found didn't exist anywhere, thus removing one ground of suspicion of a motive to overhype ambiguous data; or 4) some combination of the above or some other factors I haven't thought of.

    On the other hand, his earlier LL post gkp links to at the end of his Lingua Franca piece hypothesizes the secularist/anti-SIL bias which is widely said to exist to varying degrees in some parts of the academic linguistics world as one possible source of anti-Everett negativity, but Everett had apparently lapsed from Christianity and the motive to translate the Scriptures whereas Derbyshire never did (and in fact produced an entire New Testament full of O-initial clauses), yet had his findings accepted.

  16. AntC said,

    March 28, 2012 @ 6:53 pm

    Mark, thank you for your earlier replies, I found the discussion in 'Ontological promiscuity v. recursion' particularly helpful.

    As 'That's what SHE said' said, as a non-linguist the word 'recursion' is particularly throwing me off understanding the issue (and that's why I prefer to use 'embedding').

    In computer languages we distinguish embedding (or 'nesting', in which there's two parts to the construction with some other construction in between) vs iteration (or 'repeating' some construction arbitrarily at the end of a construction). They both produce arbitrarily long programs out of finite elements. They can both be described using recursive formal grammars. But we prefer an iteration structure if possible, because it avoids needing to 'remember' the outer structure.

    (Some early programming languages — such as COBOL — banned arbitrary nesting, and allowed only pseudo-nesting to some finite depth. Then their formal grammars have only iteration.)

    Co-ordination is typically processed as a separate pass after syntactic analysis. (This would be the Thing1 etc approach per Hobbs.) And the consequent semantics essentially takes a lambda-calculus approach (which I think corresponds to Montague grammars(?)/Barbara Partee, and New Zealand's own Max Cresswell.). A formalisation would be van Wijngarden grammar developed for ALGOL 68, which proved just too complex for practical compilers.

    Sorry for the long ramble, but the methodological question is this:
    No corpus in fact exhibits 'infinitely' long sentences. (In fact people get quickly confused by more than a few levels of nesting: "the rat the cat the dog chased bit died.") So for any finite corpus, we could choose to describe it using only 'iterative' constructs.

    How do either Everett or Chomsky 'prove' that a language does or does not exhibit embedding, as opposed to iteration? Isn't that just a matter of taste or economy of description on the part of the linguist?

    We're not in 'no word for X' territory: I don't think any of the parties are claiming that non-embedding languages are incapable of expressing certain ideas or connections amongst elements of the discourse?

    (I'm trying to understand why there's been so much (alleged) vituperativeness over what seems such an arid dispute.)

  17. Avery Andrews said,

    March 29, 2012 @ 12:57 am

    "In a large number of Australian languages, the principal responsibility for productive recursion in syntax is shouldered by a structure which I will refer to as the adjoined relative clause . It is typically marked as subordinate in some way, but its surface position with the main clause is marginal rather than embedded–hence the locution 'adjoined'" (Hale 1976, 'The Adjoined Relative Clause in Australia".)

    iirc in those days a big deal was made of the difference between 'central' recursion a.k.a embedding, which can't be computationally optimized, and left and right edge recursion, which can be.

    I am as puzzled as anybody about where the sense of urgency and 'vituperation' comes from in the NPR response, since it seems to me that what David says about it is true (although I don't yet find the arguments for clausal recursion very compelling), but somehow the sense remains, mysteriously.

  18. AntC said,

    March 29, 2012 @ 3:05 am

    From GKP's 'Fear and loathing on Massachusetts Avenue':
    According to Everett, the Pirahã … have essentially no time for considerations of matters historical, …
    … their complete absence of interest in the hypothetical and the abstract has remained stable for some 200 years of contented and almost totally monolingual and monocultural life.

    How did the Pirahã manage to talk to Everett about the past 200 years, and explain abstract notions like stable, contented, monocultural? Or did Everett ask those questions (how?) and simply get non-commital replies?

    Or was there contact 200 years ago with earlier Europeans or other tribes? Anybody that recorded their language at the time? (Even if there's archaeological evidence of occupation, is it enough to reconstruct continuity of culture?)

    Is this a culture that due to dwindling to tiny numbers in the 1960's have lost their oral history? Is their "complete absence of interest" in fact a sad and embarrassing (for them) absence of memory?

    Is what was happening in the 60's something close to the death of the language, which is now brought back from the brink? Are there studies of how languages die, just as there are studies of contact languages being born?

  19. Sudha Lakshmi said,

    March 29, 2012 @ 4:56 am

    I'm not a linguist, just a philosopher – and a half-baked one at that – so please bear with me.

    One question I have is whether there's more than one notion of recursion at play here. I think of recursion as a method that allows us to generate sentences of greater and greater complexity from a finite set of "building blocks," as it were; and it does so by allowing a sentence, in some sense, to operate on (an earlier instance of) itself. It is in this way that we get from a very basic sentence formed in the simplest possible way, an atomic sentence as it might be called in a formal language, to more and more complex sentences.

    Since this is how I understand recursion, when I first read the piece in the Chronicle, I took Everett to be claiming that Piraha had no complex sentences at all: no conjunctions or conditionals or even negation, let alone embedding in other, non-truth-functional contexts. (And how could one possibly draw inferences in this language?!) But perhaps linguists understand recursion in a different way, as some of the comments above suggest. What exactly is recursion in this context, i.e. the context in which Everett says Piraha has no recursion and Chomsky and others contend there must be?

    Someone pointed me to Everett's "You drink. You drive. You go to jail. Where's recursion?," where his point seems to be not that there's no recursion at all, but just that it can't always be found in the grammar. His idea, as far as I could tell anyway, seems to be that in Piraha at least, recursive interpretations are "imposed" by the mind, and not by the grammar. Obviously this is a very different claim from what the article in the Chronicle attributes to Everett. And I simply have no idea to what extent this runs counter to Chomsky's view; does it?

    In any case, it's not clear to me why if there are indeed recursive interpretations, as Everett says, they must be imposed by the mind. Even if, as Everett contends, Piraha doesn't allow complex sentences to be formed via recursion, this simply shows that this procedure doesn't happen at the level of the sentence in Piraha. Everett himself gives several examples of complex thoughts conveyed in Piraha; it's just that in Piraha it seems to take several sentences to put that thought together rather than one nicely nested one in, say, English. Where we might say in English, "If you drink and drive, you will go to jail," in Piraha we might say, "You drink. You drive. You go to jail." I exaggerate, but it almost seems as if the difference could simply be a matter of where we put the period!

  20. Rod Johnson said,

    March 29, 2012 @ 7:01 am

    There have been many studies of language death. Nancy Dorian's work on East Sutherland Scottish Gaelic is one of the most celebrated, but there are many others. The Wikipedia article has a surprisingly good list of references.

  21. alex said,

    March 29, 2012 @ 3:22 pm

    Of course, one possible reason for the vituperation of this dispute is because that style of argument may be favoured by some of the participants, quite independently of the substantive content of the debate. Evidence in support of this conclusion may be found in a paper by leading AI researcher Margaret Boden, writing in response to three critics of a book she had written, in the following paper:

    M. Boden [2008]: "Odd man out: Reply to reviewers." Artificial Intelligence Journal, 172 (8): 1944–-1964.

    One of the critics to whom Boden responded in this paper was Chomsky.

  22. leoboiko said,

    March 29, 2012 @ 5:20 pm

    I couldn't find an online version of Boden’s reply (mentioned by alex) here. It’s behind an evil Elsevier paywall, so I guess one needs to be a privileged member of a paying academic institution to read it.

    [(myl) Here.]

  23. alex said,

    March 29, 2012 @ 5:54 pm


    The paper's opening paragraph:

    "Three very different reviews . . . to which I’ll reply in turn [29,47,123]. Two of them, written by Paul Thagard and Jerry Feldman, engage with my book seriously. The third, by Noam Chomsky, does not. It’s a sadly unscholarly piece, guaranteed to mislead its readers about both the tone and the content of my text. It’s also defamatory. But that’s par for the course: I’m not the first, and I surely shan’t be the last, to be intemperately vilified by Chomsky. However, more on that later."

  24. Anonymous said,

    March 29, 2012 @ 9:05 pm

    These reactions call to mind the well-known "god of the gaps" phenomenon. Linguistic nativists are fighting ever more bitterly for ever smaller territory.

    What is this jiujitsu deflection of decades of powerful anti-nativist arguments from the likes of Bates, Tomasello, Deacon, Croft, Nick Ellis — to set up the findings of one field linguist about one language as if everything was riding on this?

  25. Avery Andrews said,

    March 29, 2012 @ 9:58 pm

    The problem with the 'powerful anti-nativist arguments' is that none of them have yet led to a substantive specification of anything that learns a reasonable approximation of a human language from a reasonable approximation of the kind of data that humans appear to learn the language from. I think it has actually become more interesting to *try* to show that UG is not necessary than to continue arguing that it is, but the word *try* here covers a very large gap between promise and actual performance. And part of the interest is that this kind of project has an interesting failure mode: if your model doesn't work as is, but does with some assumptions about UG added, then you've got an actual argument for UG! (better than the ones we have now).

  26. Anonymous said,

    March 29, 2012 @ 11:46 pm

    @Avery Andrews:

    I guess you'd have to explain "substantive specification" a little better. Construction grammars describe languages (note the plural) in a way that is friendly to psychological principles of learning, automatization and recall. Usage-based acquisition theories point to lots of evidence that the development of language patterns reflects accumulation of and generalization over tokens in the context of social communication. The theoretical advantages are strong parallels with learning in other domains, no need to posit brain functions very different from those recognized elsewhere in psychology, greater plausibility with respect to primate evolution, and a relatively unified account of L1/L2 learning that focuses on input and interference.. We can analyze constructions in languages, we can compare them, we can form lots of empirical generalizations a la Greenberg, and we can do all this while applying general principles from the psychology of learning. Beyond this, what level of "specification" are you looking for?

  27. Rod Johnson said,

    March 30, 2012 @ 8:42 am

    Cite please? My knowledge of CG is pretty out of date.

  28. Avery Andrews said,

    March 31, 2012 @ 7:28 pm

    @Anonymous: I like CG as a project, and actually think that trying to show that UG is not necessary is a more interesting goal than trying to show that it is (40 years after George Lakoff first said this), but 'worth trying to show' != 'has been shown'. In CG, what Johanna Bardhdal calls 'regularity' still seems quite underdeveloped, this being fundamental to solving the projection problem that motivates most generative research.

    I think I'd probably accept UG is decisively shown unnecessary if someone came up with a UG-free learning method that could learn case-stacking in Kayardild and the basic facts about empty subjects (EQUI and Raising) in Icelandic from a reasonable approximation to primary linguistic data. Other people would probably have their own hurdles of adequacy for a fully empiricist language acquisition theory to clamber over.

  29. AMM said,

    April 3, 2012 @ 2:48 pm

    alex's comments fit with what little contact I have had with Chomsky and his students.

    Back when I was in college (~40 years ago), I took a linguistic course, taught by Chomsky disciples, and I attended one lecture by Chomsky. Chomsky's lecture was devoted entirely to castigating various people over issues I couldn't really understand (I do remember the phrase "totally wrongheaded," because it seemed to constitute something like 10% of his lecture.) The teacher of the course I took was a Chomskyite and Chomsky's theories were not open to discussion.

    My contact with Chomsky and his school of linguistics were one factor in why I never went any further in linguistics.

    BTW, how many languages does Chomsky know (well), besides English? I had the distinct impression that my Chomskyite teacher knew few, if any, and the theories he was teaching seemed to fit English better than any other language I knew.

  30. Pirahã and Progress – Nic McGinnis | Rotman Institute of Philosophy said,

    February 23, 2013 @ 4:52 am

    […] Language Log blog has an excellent summary of the situation in a recent post. (So does Lingua Franca). Of particular importance is the fact that the only researcher who has […]

RSS feed for comments on this post