Pirahã recursion - new interpretation of data

[Radagast]

According to this article researchers reviewing Everett's data have found that there may actually be recursion in Pirahã and that it is marked by a tone different in the particle transcribed as "sai" by Everett, but which seems to be either sai or saí depending on whether it works as a nominalizer or as the head of a conditional phrase. Interesting read and interesting to see how even interpretation of basic syntax is completely dependent on correct perception of small phonetic differences.

[alice]

Does this mean you can solve the Towers of Hanoi efficiently in it, then?

[Guitarplayer II]

*instant rimshot*

[Xephyr]

Independently of the specific factor determining the choice between sai and saí, the result
provides evidence for the existence of complex clauses in Pirahã. Namely, if there was no
grammatical relationship between the two nuclear clauses, it would not be expected that the
relationship of the two nuclear clauses should determine the tone of sai/saí.

But,

[a] possible hypothess that seem[s] worth pursuing [is] the following ... A prosodic explanation: Sentence prosody could explain the tone distinction we observed since in the data we considered sai always occurred utterance-medially, while saí occurred utterance-finally. Specifically, such an account would postulate that there is only one morpheme sai/saí but it is pronounced with low tone if it occurs in the middle of a complex clause and with high tone if it occurs on the right edge of a complex clause. In the absence of further evidence we cannot determine which of these two hypotheses is correct.

Does that seem a bit weak to anyone else here? Tone difference depending on whether something is "utterance-medial" or "utterance-final" seems to make sense, but then they restate it as "occur[ring] in the middle/right edge of a <i>complex</i> clause" (emphasis mine). How do non-complex clauses not count as "utterances" here? Where do they support the implicit claim that non-complex-clause-medial and non-complex-clause-final tones <i>don't</i> show the same prosodic tone difference?

[zompist]

What do you expect them to say? They say they don't have the evidence to decide.

[Xephyr]

I expect them not to say that both explanations would provide evidence for complex sentences, which they (meaning the explanations) don't and which they (meaning the authors) do (say so, that is).

[zompist]

OK, I've actually read the article now. I think their logic is as follows:

Possibility A: the sai/saí distinction is lexical— conditional vs. nominalizer. As these are both ways of forming complex sentences, Pirahã would have complex sentences.

Possibility B: the distinction is prosodic— utterance-medial vs. utterance-final. Nonetheless the very fact that the sentences can be linked this way shows that Pirahã has complex sentences. Or to put it another way, if you posit that the two halves are separate sentences, ipso facto there is no way for sai to differ between them. The very point of difference is that the sentences are linked somehow (i.e. complex).

But the data were collected for some other purpose and highly limited (essentially variations on two sentences), thus far from settling the matter.

Or to answer your first question more directly, there's no such thing as "non-complex-clause-medial" vs. "non-complex-clause-final", precisely because if there are no complex clauses, no clause is medial.

Tangentially, not knowing Pirahã at all but reading Everett's varying glosses, it strikes me that he's stretching things a bit to make his point. Concatenation is a form of connection.. The number "forty-two" implies addition using concatenation alone; it would be weird to analyze it as "forty. two." or something as part of a claim that English had no numbers above 19. And even a topicalizer as he describes it is a form of linking sentences.

[merijn]

In the meantime Everett has posted an answer here: http://ling.auf.net/lingBuzz/001119

Another more theoretic article about Pirahã and recursion is here: http://ling.auf.net/lingBuzz/001110

[finlay]

OK, I've actually read the article now. I think their logic is as follows:

Possibility A: the sai/saí distinction is lexical— conditional vs. nominalizer. As these are both ways of forming complex sentences, Pirahã would have complex sentences.

Possibility B: the distinction is prosodic— utterance-medial vs. utterance-final. Nonetheless the very fact that the sentences can be linked this way shows that Pirahã has complex sentences. Or to put it another way, if you posit that the two halves are separate sentences, ipso facto there is no way for sai to differ between them. The very point of difference is that the sentences are linked somehow (i.e. complex).

But the data were collected for some other purpose and highly limited (essentially variations on two sentences), thus far from settling the matter.

Or to answer your first question more directly, there's no such thing as "non-complex-clause-medial" vs. "non-complex-clause-final", precisely because if there are no complex clauses, no clause is medial.

Tangentially, not knowing Pirahã at all but reading Everett's varying glosses, it strikes me that he's stretching things a bit to make his point. Concatenation is a form of connection.. The number "forty-two" implies addition using concatenation alone; it would be weird to analyze it as "forty. two." or something as part of a claim that English had no numbers above 19. And even a topicalizer as he describes it is a form of linking sentences.

I have sat round with a group of friends and skimmed an paper about Pirahã – to our eyes it was full of contradictions. Everett's attitude towards them also struck us as very colonial, as if he was only a couple of steps away from referring to them as noble savages.

Of course, I can't really have a good opinion of my own, because I haven't looked into the matter enough. On the other hand, though, I think Everett possibly uses that to his advantage – if anyone thinks his claims are wrong, he'll just turn around and point out that he was the one who actually lived with them for several years. An unfair advantage if ever there was one.

[psygnisfive]

Having seen Everett discuss these things first hand, and discuss the Piraha as a culture, I can say he's full of contradictory twaddle. He will in one sentence make some big claim about them not having mythology and never talking about things they've never seen, then mention their origin myths. He'll say they don't believe in demons or spirits, then say they've seen them as figments so they do but it's justified, and then later he'll say that they've only actually ever seen them once and it was probably mass hysteria or maybe just a practical joke they were playing on him.

[Rik]

Of course, I can't really have a good opinion of my own, because I haven't looked into the matter enough. On the other hand, though, I think Everett possibly uses that to his advantage – if anyone thinks his claims are wrong, he'll just turn around and point out that he was the one who actually lived with them for several years. An unfair advantage if ever there was one. :P

I'm sure I read somewhere that Everett's wife (Keren) disagreed with a number of his more contentious conclusions, but buggered if I can remember the name of the article, or where it was published.

[Radagast]

Mostly the conclusion that God doesn't exist. She is still a missionary trying to convert the Pirahã.

[TaylorS]

Methinks Everett is degenerating into quackdom of the same form as Lee Whorf.

[Whimemsz]

"Degenerating"? He's been saying the same thing for a number of years now, it's not like this is news.

Anyway, more directly on topic, I really am not that impressed with that paper. There's just...way too little data to actually draw reasonable conclusions. One day someone other than Everett will actually have to learn the friggin' language and maybe then we'll have a better idea of how close to reality Everett's claims are.

[Rik]

Mostly the conclusion that God doesn't exist. She is still a missionary trying to convert the Pirahã.

No ... it was something about her watching mothers playing with their babies and how they were speaking to them - something in the way mothers were teaching their children the language which might show that some of her husband's assumptions were a bit on the shaky side.

[Radagast]

Methinks Everett is degenerating into quackdom of the same form as Lee Whorf.

[oblique rant] ... [/oblique rant]

[Radius Solis]

Methinks Everett is degenerating into quackdom of the same form as Lee Whorf.

There's an important difference between a quack and a scientist who's convinced of something false, and it lies in willingness to seriously consider whether contrary evidence may be correct. (Also linguists wouldn't be wasting grant money on fieldwork to check Everett's claims, nor writing papers about them, if they thought he was a quack. Quacks mostly get ignored, not argued with.)

[merijn]

In the mean time Everett has posted a new article on lingbuzz. I haven't read it yet but I have skimmed through the first pages, which he spends arguing that where Hauser, Finch and Chomsky talk about recursion they are not talking about recursive merge* but about whatever Everett's definition of recursion is. When I was a young promising student in the early 00's we talked about those articles of Hauser, Finch, and Chomsky and I was told by the teacher that their definition of recursion is recursive merge, and that was way before Everett came with the Pirahã data. There is already an answer here. It is not very extensive; if you click on the PDF you get one page without text. The answer is just the title [ ... head ... [ ... head ... ]]. Here is recursion.

*recursive merge means that you can merge the result of an earlier merge. If you have three words A, B and C, you can first merge A and B to form the phrase [A B], and then you can merge the result of that with C resulting in [C [A B]]. Every tree structure that is not flat is the result of recursive merge, and since virtually every minimalist believes in binary branching (or rather a version of merge that can only merge two terms) every sentence with three or more words must be the result of recursive merge.

[Atom]

Mostly the conclusion that God doesn't exist. She is still a missionary trying to convert the Pirahã.

No ... it was something about her watching mothers playing with their babies and how they were speaking to them - something in the way mothers were teaching their children the language which might show that some of her husband's assumptions were a bit on the shaky side.

Everett believes that Pirahã mothers do not use "baby talk".

Really I agree with a lot of the above posters: Pirahã won't be settled until someone else goes and learns their language.

[Jetboy]

Why is it that no one else has gone out to learn it? From a quick look at the Wikipedia article, it doesn't seem to be especially isolated, and with this much controversy, it seems rather strange that no one has tried to verify Everett's claims.

[Yng]

I thought other linguists HAD learnt it? I don't think Everett's the first to study it, even, is he? He's just gained an awful lot of publicity from it.

[finlay]

Why is it that no one else has gone out to learn it? From a quick look at the Wikipedia article, it doesn't seem to be especially isolated, and with this much controversy, it seems rather strange that no one has tried to verify Everett's claims.

I get the impression that it took a lot for them to trust Everett and his wife enough for them to let him learn the language. They might simply not be willing to put in the effort with other linguists. They are, after all, just human beings...

[Radagast]

Everett believes that Pirahã mothers do not use "baby talk".

This is not at all a controversial claim. Loads of cultures have been shown not to have a register of baby talk. Samoan and Kaluli for starters.

[Radius Solis]

I thought other linguists HAD learnt it? I don't think Everett's the first to study it, even, is he? He's just gained an awful lot of publicity from it.

As far as I'm aware no other linguists have anywhere close to Everett's degree of competence in the language, and he isn't the first to make any study of it, just the first to make a thorough study of the sort that can only be accomplished by living with the group for an extended period. Another obstacle, IIRC, is that there is a somewhat pidginized version of Pirahà that they are accustomed to using with outsiders, and getting past that to the full variety of the language requires a degree of acceptance by the tribe that's not easy to come by.

[Yng]

Ahh, yeah, that might well be the case. I wonder if Everett just woke up one day having spent six years gaining the trust of the Piraha with immense difficulty and thought 'well at least now I can say whatever the hell I want about their language, right?'