Don't Sleep, There Are Snakes (31 page)

If I am right about Pirahã lacking recursion, Chomsky and other researchers have some head scratching to do. They need to suggest how a language without recursion can fit into a theory in which recursion is the crucial component of language.

One answer that Chomsky and others have given to my claim that Pirahã lacks recursion is that recursion is a tool that’s made available by the brain, but it doesn’t have to be used. But then that’s very difficult to reconcile with the idea that it’s an essential property of human language, because if recursion doesn’t have to appear in one given language, then, in principle, it doesn’t have to appear in any language. This places them in the unenviable position of claiming that the unique property of human language does not actually have to be found in any human language.

It is not that hard, really, to tell whether recursion does any work in figuring out the grammar of a specific language. Quite simply, the question is twofold. First, can the grammar you write without recursion handle the language you are studying more simply than a grammar with recursion? Second, what kinds of phrases would you expect to find if the grammar did in fact have recursion? A language without recursion will look different from a language that has recursion. The main way is that it will not have phrases inside other phrases. If you find a phrase within a phrase, the language has recursion, period. If you don’t, it might not, though more data will be needed. The first question, then, is whether there are phrases within phrases in Pirahã. The answer is that there are not, following the standard argumentation used in theoretical linguistics to establish this: it lacks the pitch marking, words, or sentence size of a language with recursion.

The grammars of the world’s languages employ various markers to indicate that a given structure is recursive, that is, that one phrase is inside another. Such marking is not required, but it is very common. Some of these markers are independent words. In English, we say things like
I said that he was coming.
In this sentence, the phrase
he was coming
is located inside the phrase
I said . . . ; He was coming
is the content of what was said. In English,
that
is a frequently used “complementizer” for marking recursion. If we look at the relative clause complex that Kóhoi gave me, we see three independent sentences, interpreted jointly, without a shred of evidence that one sentence is inside another.

Another common marker of recursion is intonation, the use of pitch to mark different meaning and structural relations between sentences and their parts. The verb phrases of main clauses, for example, often get a higher pitch in English than the verb phrases of subordinate clauses. For example, in the most common pronunciation of the sentence
The man that you saw yesterday is here, is here
gets a higher pitch than
saw yesterday.
This is because
saw yesterday
is the subordinate, or embedded verb, phrase and
is here
is the main verb phrase. But Robert Van Valin and I, in a three-year National Science Foundation project dedicated to the study of intonation and its relation to syntax in five Amazonian languages, found no evidence that Pirahã uses intonation as an alternative marker of recursion. Now, Pirahã intonation does group sets of sentences together in paragraphs and stories, but this is not recursion in the grammar proper, at least not according to the entire history of Chomskyan grammar (though many linguists disagree with Chomsky and do place stories in the grammar—I have no quarrel at all with these other schools of linguistics in this sense). It is recursion in reasoning. In fact, many specialists on the role of intonation in human speech believe that it would be naive to try to link intonation directly to the structure of sentences rather than to the meanings of sentences and how they are used in stories. If this is correct, then intonation has nothing conclusive to say about whether a language has recursion or not.

Confusing language and reasoning is something that we have already seen to be a serious mistake. It is easy to confuse the two because reasoning involves many of the cognitive operations that some linguists associate with language, including recursion. Herbert Simon’s classic 1962 article, “The Architecture of Complexity,” gives a fascinating example of recursion outside of language. Simon’s example even shows how recursion can help your business! His example is worth citing in full:

This watchmaking example has nothing to do with language. So by this example, and many others, we know that human reasoning is recursive. In fact we know that many things in the world apart from humans are recursive (even atoms manifest recursivelike hierarchies in their construction from subatomic particles). Familiar Russian matrioshka dolls illustrate another type of recursion, known as nesting, where one doll is placed inside another of the same type, and that pair into another of the same type, and so on.

An important inference from the presence of recursion is this: if a language has recursion then there should be no longest sentence in the language. For example, in English any sentence that someone utters can be made longer.
The cat that ate the rat is well
can be extended to
The cat that ate the rat that ate the cheese is well,
and so on.

Crucially,
none of these diverse types of evidence for recursion is found in Pirahã.
The story about the panther that Kaaboogí told me is typical. No evidence along any of these dimensions is found in that or other Pirahã texts for recursion in the grammar.

Most interesting, perhaps, for illustrating my point against recursion, is a sentence like the following, because there is no obvious way to make it longer in Pirahã:
Xahoapióxio xigihí toioxaagá hi kabatií xogií xi mahaháíhiigí xiboítopí piohoaó, hoíhio
(Another day an old man slowly butchered big tapirs by the side of the water, two of them). Anything else added to this, like the word
brown
in
big brown tapirs,
would render the sentence ungrammatical. Phrases can have a single modifier (phrases that are found in natural stories—I do have some artificial examples where I was able to get some Pirahãs to place more modifiers in the phrase, but they didn’t like it and never use more than one in a phrase in natural stories). A second one can occasionally be inserted at the end of the sentence as an afterthought—like the
two of them
at the end of this sentence. If this is correct, then Pirahã is finite and cannot be recursive.

I should rule out one final bit of potential evidence for recursion in Pirahã that has been suggested to me by several linguists. The first linguist to do so was Professor Ian Roberts, the head of linguistics at Cambridge University, during a debate with me on the BBC’s radio program
Material World.
He claimed that Pirahã must have recursion if it can add or repeat words or phrases after sentences, because, as he put it, “Iteration is a form of recursion.” Logically this is correct. Putting one phrase inside another at the end of a sentence is mathematically identical to repeating elements after a phrase or sentence. If I say, “John says that he is coming,” the sentence
that he is coming
is placed inside the sentence
John says . . .
at the end. This is known as “tail recursion.”Mathematically or logically this is equivalent to saying, “John runs, he does,” where the sentence
he does
is just a sentence repeated after another sentence. Pirahã can, indeed must, have one sentence follow another sentence, as in “
Kóxoí soxóá kahapii. Hi xaoxai hiaba
” (Kóxoí already left. He is not here). But if mere repetition, iteration, of one sentence after another satisfies Hauser, Chomsky, and Fitch’s definition of recursion (as some of their followers tell me it would), then it is found in species other than
Homo sapiens.

Our Rhodesian ridgeback, Bentley, is an emotional soul. Among the things he gets emotional about is other dogs passing our house—he wants to eat them or otherwise harm them. He always barks when they go by. I for one don’t think his barks are devoid of content. I think he is communicating with his barks something like “Get the hell out of my yard.” But it doesn’t matter what exactly he is communicating—he is communicating
something
with his barks. Now sometimes Bentley barks once or twice and then stops. This is because the dog he is barking at has left the lawn. Other times he barks repeatedly, that is, he iterates his barks, and this indicates his rising anger/desire that the dog leave our yard (or whatever it means to him). What does his repeated barking mean? Well, if iteration just is a form of recursion, it means Bentley has recursive barking. But Bentley is not a human. So recursion isn’t limited to humans. Or, more sensibly, iteration shouldn’t be considered recursion.

Yet the reasons that lead me to claim that Pirahã lacks recursion are not merely negative. Saying that a language lacks recursion makes claims about what the language’s grammar will look like. We want to look at those predictions and see how they fare with respect to Pirahã.

The pervasive immediacy of experience principle (IEP) could explain why Pirahã lacks embedded sentences. Consider relative clauses again, as in
The man who is tall is on the path.
This English sentence is composed of two smaller sentences: the main sentence,
The man is on the path,
and the embedded, or subordinate, sentence,
who is tall.
The new information, or what linguists call the assertion, is found in the main sentence,
The man is on the path.
The embedded sentence merely adds some old information shared by the hearer and the speaker—there is a tall man that we both know—and draws attention to a particular man in order to help the hearer know who the man on the path is. This is not an assertion. Embedded sentences rarely, if ever, are used to make assertions. So the IEP predicts that Pirahã will lack embedded sentences because it says that declarative utterances may contain only assertions. Containing an embedded clause would be to contain a nonassertion, in violation of the IEP.

Another example comes from sentences like
The dog’s tail’s tip is broken.
This is something the Pirahãs would say regularly, since a high proportion of their dogs have damaged tails. One evening I noticed a dog in the village with the end of its tail missing. I said,
“Giopaí xígatoi xaóxio baábikoi,”
which is how I thought I should say, “The dog’s tail’s tip is malformed.” It literally means “Dog tail at the end is bad.” The Pirahãs responded, “
Xígatoi xaóxio baábikoi
” (The tip of the tail is bad).I didn’t think anything about the omission initially because omissions are common in any language when speakers share information in common—no need to restate that we’re talking about a dog; we all knew that already.

But as I investigated further, the only way to get something like
The dog’s tail’s tip is broken
is
“Giopaí xígatoi baábikoi, xaóxio”
(The dog’s tail is bad, on the tip). What I discovered is that no more than one possessor can occur in a given phrase (
dog
is the possessor of
tail,
for example). If there were no recursion in the language, this would make sense. You can get one possessor without recursion by simply having a cultural or linguistic understanding shared by speakers that when two nouns are next to each other, the first one is interpreted as the possessor. But if you have two possessors in the clause, one of them has to be in a phrase that is within another phrase.

Pirahã lacks these structures. It is hard for many linguists to see how culture could be responsible for this. And I agree that the route from a cultural constraint to complex noun phrases can seem a bit circuitous.

Starting with subordinate clauses, the first thing to remember is that according to the IEP, the embedded clause is not allowed because it is not an assertion. The question this raises is how the grammar of Pirahã could eliminate the unwanted embedded clauses in order to obey a cultural taboo.

There are three ways that it could do this. First, the grammar could prohibit the emergence of rules that create recursive structures—rules that are technically expressed like
A
→
AB.
If the grammar does not contain this rule, it cannot place one phrase or sentence immediately inside another phrase or sentence of the same type.

Second, the grammar could fail to have evolved recursion. There is a growing consensus among linguists that grammars without recursion precede grammars with recursion evolutionarily and that even in grammars with recursion, nonrecursive structures are used in most environments.

A final possibility is that Pirahã grammar simply fails to provide for structure in sentences. There would be no recursion because in effect there would be no phrases, only words placed side by side and interpreted as a sentence.