The Reality Club: RECURSION AND HUMAN THOUGHT


STEVEN PINKER [6.13.07]

I have favorably cited Daniel Everett's work with the Pirahã, both in a scholarly article and in my forthcoming book, and believe that linguists should take his criticisms of the field seriously. But I have become increasingly skeptical of the strong version of his claims, and of the importance that has been attached to his work by the media.

1. Everett claims that Pirahã violates Charles Hockett's famous list of language universals in that it provides no means to discuss events remote from experience. That claim is belied by many of Everett's own observations.

To take just the most obvious example, he writes that "spirits and the spirit world play a very large role in their lives."  Assuming that they don't literally see ghosts and spirits (which would be a discovery far more radical than any of the linguistic or psychological claims at issue) they must have a richly developed, culturally transmitted set of beliefs about entities and events that lie outside the realm of their immediate experience.

Everett's claim that Pirahã lacks the mechanism of recursive embedding (in which a word or phrase can be inserted inside a word or phrase of the same type) must be qualified as well. Pirahã allows for a degree of semantic embedding using verb suffixes and conversions of nouns to verbs (so one can express the thought "I said that Kó'oí intends to leave," with two levels of semantic embedding), and one can conjoin propositions within a sentence, as in  "We ate a lot of the fish, but there was some fish we did not eat." Nevins, Pesetsky, and Rodrigues give further examples. It is questionable of Everett to disavow his own data on the grounds that at the time he was in the grip of ideology—his current stance is as polemical and tendentious as anything in Chomskyan linguistics.

2. Whatever grain of truth there may be to the observation that the Pirahã are more concerned with the here-and-now than we are, it is by no means unique to the Pirahã. On the contrary, the observation has been by numerous scholars about numerous foraging and nonstate peoples. For example, in the 19th century, Alfred Russel Wallace observed of the Indonesian natives he met during his fieldwork:

Compare this (European culture) with the savage languages, which contain no words for abstract conceptions; the utter want of foresight of the savage man beyond his simplest necessities; his inability to combine, or to compare, or to reason on any general subject that does not immediately appeal to his senses. ...

The great linguist Otto Jespersen made similar observations about native Hawaiians, and an anthropologist I know had the same impression (expressed in private) of the !Kung San he worked with in the Kalahari. I suspect that this is simply the default impression that modern Europeans or Americans have of many native peoples, but with the rise of politically correct anthropology in the 20th century, one wasn't allowed to say such things in public directly. In this background, Everett could claim that he was making a discovery about a trait that was unique to the culture he studied, whereas it was only the prior taboo against saying these things about other peoples that made the observation seem novel. By framing his observations as an anti-Chomsky discovery rather than as un-PC Eurocentric condescension, Everett was able to get away with claims that would have aroused the fury of anthropologists in any other context. This is not to say that there is no difference in the amount of abstract thinking between foraging and postindustrial societies, just that Everett (and the journalists that have reproduced his claims) are almost certainly wrong in writing that this is unique to the Pirahã, or even unusual among nonliterate peoples.

(The same is true, incidentally, of their counting system. "One, two, many" systems are widespread among foraging peoples, and may be the default counting system among nonliterate peoples.)

3. Everett's truly radical linguistic claim is not about Universal Grammar, and his main opponent is not Chomsky. His radical claim is about variation—the non­-universal aspects of language—and his opponents in this debate are probably 99% of linguists, including most non-Chomskyans.

One of the great findings of linguistics, vastly underappreciated by the rest of the intellectual world (and probably not highlighted enough by linguists themselves) is that the non-universal, learned, variable aspects of language don't fit into any meaningful, purposive narrative about the surrounding culture. Linguists have documented vast amounts of variation, and have a good handle on many of its causes, but the causes are internal to language (such as phonological assimilation and enhancement, semantic drift, and syntactic reanalysis) and aren't part of any symbolic or teleological plan of the culture. There are Subject-Object-Verb and Subject-Verb-Object languages, and tone and non-tone language, and null-subject and non-null-subject languages, but there are no SOV or SVO cultures, null-subject and non-null-subject cultures, and so on. The variation is just as autonomous as the universals. And this is the discovery that Everett is trying to overturn in his claim that the linguistic properties of Pirahã are meaningfully explained by an overarching theme in their culture, namely their alleged unwillingness to think about concepts that lie outside their immediate experience. As I mentioned, numerous observations by Everett himself are inconsistent with this remarkable claim, and Nevins, Pesetsky, and Rodrigues show that the connection between his claims about their culture and the details of their language is tenuous at best.


DANIEL EVERETT [7.14.07]

I appreciate Steve Pinker's response to my work, even though it is negative, because it provides me yet another opportunity to clarify my statements on Pirahã. It is always interesting to me to see how people read into what I am saying about Pirahã based on their own theoretical backgrounds. The influence of our theoretical and other biases on the way that we interpret the world around us is further illustration of my general point that language and grammar can be deeply affected by culture. But let me be more specific. I will respond to each section of his criticism.

1. Self-contradictory: Pinker is persuaded by his own reading of my work and the criticism of my work in the paper by Nevins, Pesetsky, and Rodrigues, that my recent work contradicts my previous work in ways that cannot be attributed merely to my previous theoretical baggage, i.e. that I worked within Chomskyan theory. The first example he gives is that the Pirahãs rich and textured beliefs about spirits cannot be fit into my principle of immediacy of experience.

But there is nothing in the immediacy of experience principle — which, by the way, is merely a very small part of the intricate web of values of Pirahã culture — that claims that the Pirahãs cannot have detailed explanations of the world they live in. This principle claims simply that they have to have direct evidence of the things they talk about. And there is no claim that this principle is unique to the Pirahãs. The claim is, rather, that this principle constrains Pirahã language in ways we don't see in other cultures.

Not that other cultures don't have similar values. It is the interaction of cultural values and grammar that makes each language unique in a profound sense. Pirahãs do see spirits. I have seen an entire village yelling at a spirit on a beach on which they claim to see a spirit but where I can see nothing. But they also believe that spirits manifest themselves as different animals and people. And seeing such an animal counts as seeing a spirit, literally. This is not at all uncommon of course, cross-culturally. So a Pirahã can see a jaguar and say that they have seen a spirit and believe it, depending on their spiritual state at the time. I have almost lost my life by going out to 'scare off' a spirit at their request at 3AM, only to realize that the spirit was a jaguar!

I claim that the Pirahãs lack recursion in the syntax. I made no such claim about their semantics or their discourse, for example. In fact, I have given many examples of recursion in discourse as different ideas are contained in others within the main and subordinate story lines. My claim is about syntax. And the examples that Pinker uses to show self-contradiction in my account were mistranslated (by me) initially.

The word translated 'but' in his example does not mean that exactly and can be used in isolated clauses as in 'But I give this to you' or 'But I am watching the river'. In these cases, what the speaker means is that what I am doing violates your expectations (since giving normally involves an expectation of receiving something in return, in this case there is nothing expected in return; and in the second case it means that the Pirahã is idly watching the water and has no other purpose.

Pinker's (and others') reaction to the idea that my present account could be violating my previous accounts shows a profound lack of familiarity with the nature of linguistic fieldwork, something that none of the major critics of my work so far have had any experience with.

2.I think that it probably is correct that hunters and gatherers generally attribute, by necessity, more importance to the here and now than many of us in Western societies with huge surpluses of resources. I am not claiming that this is unique among the Pirahãs. Nor do I see why such a banal statement should outrage anthropologists or anyone else.

What is unusual, perhaps unique, about the Pirahãs (though, again, nothing at all hinges on them being unique in any of these matters) is the way in which they have codified a principle of immediacy of experience and the way in which it constrains their grammar. My comments are not those of a tourist, such as Otto Jespersen on Hawaiian. They are based on years of trying to figure out what made the Pirahãs different from other Amazonian groups — a difference which everyone who has seen the Pirahãs and other Amazonian groups notices almost immediately.

And I have done field research on 23 other Amazonian groups in addition to the Pirahãs. Based on this cross-cultural research, I have made a proposal. This proposal is not based on the Pirahãs' intelligence, genetics, or any perceived inferiority on their part whatsoever. It is based on what seems to be a cultural taboo on certain ways of speaking. Nothing more.

With regard to their counting system, Pinker has it wrong again. Pirahãs do not have 'one', 'two', and 'many'. That indeed is a common system. Rather, the Pirahãs have no numbers whatsoever and no counting, not even tallying, whatsoever. This claim has been tested in recent work by researchers from MIT's Brain and Cognitive Science Department and a paper is underway to report the results of those tests.

3. Pinker is right that my quarrel is not merely with Chomsky (by now the whole idea of a language instinct or universal grammar is so vacuous and untestable as to hardly warrant a criticism from me or anyone else in any case), but with the field of linguistics more generally. As Sapir warned, it can be very misguided and unscientific to attempt to correlate broad features of culture, e.g. cattle-breeding, with specific linguistic properties, e.g. whether the language mainly has the order Subject Verb Object or not. But there is absolutely nothing similar in that to what I am proposing.

I identify a specific cultural value, needed independently, and unusual syntactic properties, independently recognized, and propose a connection between them that can be tested. Linguistics needs to look harder at such culture-grammar connections. It has been misguided, in my opinion, for not doing so. In a way this is similar to the resurgence of work on language evolution. For some time the speculations on the origins of language were so unscientific and spurious that serious scientists spurned them and said that concern for language evolution was unscientific. In the same way, earlier speculations on culture's influence on language and grammar were so unscientific as to merit strong criticism and the avoidance of this issue altogether. But I am asking that we reconsider this and proceed to a more scientific approach to possible language-culture pairings.

My claim is that there is no such thing as 'just a language' and that the homogenizing efforts of Pinker and others, focusing principally on theories that stretch and chop grammars to fit preconceived notions of what a language should look like do the science of linguistics a serious disservice. Each language in this sense, while sharing cognitive and communicative principles in common with all other languages spoken by Homo sapiens, is unique. This is why it is such a tragedy when a language dies — we don't just lose a grammar. We lose an entire way of thinking and talking about the world; we lose a set of solutions to the problems that beset us all as humans.

The Pirahãs enrich the world through the brilliance and uniqueness of the interaction of their culture and language. Just like all languages and cultures do.


ROBERT D. VAN VALIN, JR. [6.14.07]
Heinrich Heine University, Düsseldorf

There are a number of points worth emphasizing with respect to Dan Everett's claims about Pirahã. First, and most important, he is not claiming that Pirahã speakers are in any way limited in what they can say by the lack of recursion in the syntax. Saying 'John has a brother. His brother has a house' communicates the same content as 'John's brother's house', albeit with less perspicuous packaging. The fact that Pirahã speakers can formulate such utterances supports Everett's claim that they can form recursive semantic propositions, which are then expressed in this non-recursive way in the syntax. There are analogues in other languages. I worked for many years with speakers of Lakhota, the language of the Sioux, which definitely has recursive structures in its syntax. If one asked a Lakhota speaker if the Lakhota equivalent of 'I know that Bill stole the money', with 'that Bill stole the money' as an embedded clause, is a possible Lakhota sentence, he or she would say that it is. If, on the other hand, one asked a Lakhota speaker how he or she would say that sentence, they would respond 'Bill stole the money, and I know it', which is exactly the same kind of non-recursive structure found in Pirahã. Given a choice, the Lakhota speakers I have worked with always chose the non-recursive structure. There are good reasons why they would want to avoid such embedded clauses, given certain features of Lakhota syntax, but the point is that speakers find it to be communicatively equivalent to the recursive structure.

John Searle long ago proposed a principle of effability, which states that all languages are capable of expressing the same content. Despite the lack of recursion, Pirahã speakers are indeed able to express complex propositions. This is relevant to Chomsky's claim that recursion is the key feature of human language. Chomsky's approach treats syntax as the main backbone of language, to which other aspects of language are secondary. Because speakers are capable of formulating complex recursive propositions, this must, given Chomsky's view of the centrality and primacy of syntax to language, be realized in terms of recursion in the syntax.

Chomsky has long maintained that the purpose of human language is to permit the free, creative expression of human thought, and it follows that there must be recursion in the syntax in order for the expression of complex propositions to be possible. He has also long denied that the communicative function of language is in any way relevant to an understanding of the structure of language, maintaining in fact the the structure of language is dysfunctional with respect to communication. Now, suppose one took the opposite view from Chomsky and claimed that the function of communication is relevant to the understanding of the structure of language and that in analyzing language one should treat it as a system exhibiting an complex interaction between syntax, semantics and pragmatics (the principles governing the use of language in context).

From this perspective, the formulation of complex propositions in the semantics, reflecting complex ideas and concepts, need not be reflected in only one property in the linguistic system, namely recursive syntax. If one of the functions of language is the conveying of complex propositional information, then one should take the whole system into account in evaluating whether the principle of effability is satisfied in Pirahã, and on Everett's account, it is.

This leads to a second point. Because the principle of effability is satisfied with respect to complex propositions (the expression of number concepts is another matter, but this issue is easier to resolve than the recursion one, with independent work confirming Everett's claim), it is misleading and inaccurate to accuse Everett of denigrating the Pirahã language or its speakers in any way. While the idea of cultural constraints on the grammar of a language is anathema to many linguists, as the reaction to Everett's work clearly shows, it is difficult to see what other explanation there could be for the lacunae in the system. It cannot be that there is anything genetically different about the Pirahã.

If a Pirahã child were taken at birth from the tribe and raised by a Brazilian family, he or she would learn Portuguese like any other child, with all of its features. There is in fact such a case approximating this situation, and interestingly, when the child as a teenager moved back to live among the Pirahã, she stopped speaking Portuguese, even refusing to speak it, did not use recursive structures, did not count, etc. This can only be explained in terms of cultural constraints and social conventions, since she clearly had those concepts from her learning of Portuguese.


DAVID PESETSKY [6.23.07]
Ferrari P. Ward Professor of Modern Languages and Linguistics at MIT

1. When Andrew Nevins, Cilene Rodrigues and I began discussing Everett's 2005 paper in Current Anthropology, our goals were simple and limited. Everett had described a set of "gaps" in Pirahã language and culture, and suggested that these gaps must be attributed to the effects of a cultural Immediate Experience Principle. In our lengthy reply to Everett's paper, posted to the LingBuzz website (and submitted for publication), we focused on two main questions:

• Had the gaps had really been shown to exist in the first place? We concluded that in several key instances, the answer was no. Evidence for many of these gaps was weak or nonexistent.

• Do we need the cultural Immediate Experience Principle to account for the gaps that did seem to be real? Here too our answer was negative.

Though Everett had called the Pirahã "very surprising from just about any grammarian's perspective", that is not what we found. In almost every case, Pirahã looked just like many other languages of the world—languages spoken in many different kinds of societies.

For example, Pirahã appears to position its subordinate clauses after the verb—despite the tendency elsewhere in the language for the verb to come last in the sentence. Everett used this fact as an argument that what look like subordinate clauses in Pirahã are really independent sentences. This claim formed part of an argument that Pirahã lacks "recursion", which in turn was supposed to reflect Pirahã's special cultural restrictions. In fact, however, the grammatical property that he had identified in Pirahã is extraordinarily common in the languages of the world—and has no detectable correlation with recursion or culture. We offered German, Hindi, and Wappo (a language of California) as examples of languages with this property, and could have given many more examples. No cultural thread links the speakers of these various languages. (Who has ever accused German speakers, for example, of living exclusively in the "here and now"?) Consequently, there is no reason to attribute the word order of Pirahã embedded clauses to any particularly remarkable grammatical property or to any property of culture.

This was typical of our initial findings. There were other issues as well. Much of the Pirahã data included in the Current Anthropology paper itself seemed too sparse to support any conclusions. Other arguments were left incomplete. Often the discussion seemed "personalized", presenting assertions in lieu of argument. Nonetheless, we could imagine that we were dealing here with a flaw in presentation, not a flaw in the actual research. Perhaps a broader examination of other available material on Pirahã would fill in the missing pieces and strengthen the arguments.

Consequently, we attempted to integrate the data from Current Anthropology with other published data available from the language. When we did so, however, we reached even more strongly negative conclusions. Where we previously found certain arguments weak, we now began to find actual counterevidence to them. For example, though Everett had claimed that apparent "subordinate clauses" in Pirahã were actually independent sentences, the literature turned out to offer crucial examples that simply could not be analyzed that way. One might imagine that what looks like the Pirahã counterpart of the single sentence "He watched the foreigner catching fish" is actually two separate sentences: "He watched the foreigner. He was catching fish." But one could hardly give the same analysis for the Pirahã counterpart to the English "He does not want me to go", which cannot be said to mean the same as, say, "He does not want me. I go."—or any similar two-sentence counterpart. Consequently, it looked like we truly were dealing with a subordinate clause.

By the time we concluded our research, we could find no remaining arguments for Everett's claims about subordinate clauses, and several good arguments against them. We reached much the same conclusions in other areas of Pirahã grammar crucial to Everett's general claims.

Our findings also touched on the cultural claims. The Brazilian anthropologist Marco Antonio Gonçalves turned out to have written two lengthy scholarly books in Portuguese, in which he recounts and discusses Pirahã mythic narratives, including a (re)creation myth. These reports seemed to contradict Everett's claims about the Pirahã's lack of mythology and their lack of interest in creations and origins.

Here then were our conclusions. The real grammatical gaps of Pirahã seemed to have no connection to culture. Other claimed gaps did not seem to be real. The cultural principle invoked to explain these gaps was not needed, and there were open questions about the cultural description as well. Finally, Everett had also asserted that his findings about Pirahã had broad implications for all sorts of deep issues in linguistics and related fields. If we were correct, no broader discussion was called for.


2.
Everett has replied to us formally in an article also posted on the LingBuzz website, and less formally in interviews and articles like the one here. Unfortunately, much of this discussion strikes me as beside the point.

Everett often writes, for example, as if the institutional and intellectual allegiances of his intellectual opponents should influence our evaluation of their arguments—and even the evaluation of their data. Among these intellectual opponents he numbers not only us, but also his own former self. It is in this light that he attempts to discredit his own detailed sketch of Pirahã syntax and morphology from the late 1980s, still the most important source for any researcher seeking hard published data about the language. "The 2005 article supersedes my previous work", he writes. The only reasons offered for ignoring the earlier work seem to be the following: (1) the earlier work presents data that appear to contradict Everett's current claims; and (2) the earlier work was written when the author held beliefs about language that he no longer does (described as the "theoretical baggage" of "Chomskyan theory"). "I did my very best to make Pirahã come out and look like a 'normal' language," Everett writes. If our paper is correct, of course, Everett succeeded because Pirahã actually is a "normal" language.

The criticism of us that Everett offers rather politely on in his article here has not been expressed so politely elsewhere. In Everett's reply, for example, we are called, for example, "armchair linguists", our work is denigrated as "armchair speculation" colored by "a vested interest in the Chomskyan framework", and our attempts to suggest alternative analyses of Everett's data are repeatedly parodied as instances of "eyeballing".

Even our professional resumes are apparently a topic for discussion—and this is probably the place to correct some errors. Everett in his article here on Edge describes my coauthor Andrew Nevins as a former "student at MIT who now holds a temporary appointment in Linguistics at Harvard". In fact, he is a regular member of the Harvard faculty. Likewise, Everett describes Cilene Rodrigues as "a Brazilian linguist who I think is doing her Ph.D. at MIT". In fact, she received her PhD in 2004 from the University of Maryland, and is currently teaching at the University of Campinas (Everett's own doctoral alma mater). In other words, we are all experienced researchers, and we are not all from MIT.

Finally, Everett claims that our paper contains insinuations of racism. This unfortunate charge once again gives a personal twist to a discussion that we carefully kept non-personal. In the final section of our paper, we wrote of a "more general discomfort with the overall presentation of Pirahã language and culture" in the Current Anthropology article, and noted that we shared this discomfort with a previous commentator on the paper who had called for a "more balanced picture". In contrast to this commentator, however, we pointed out that Everett's picture of Pirahã culture might actually be the balanced picture. But we also noted that the shakiness of the grammatical evidence for Pirahã's "gaps" might be grounds to take special care with sensitive cultural characterizations. As an example of what we had in mind, we quoted a paragraph about the Pirahã from Everett's own archived Pirahã website (along with several captions from the same webpage). To judge from the version of Everett's complaint presented in his reply, it is to these quotations that he most strongly objects. He calls them "contrived and decontextualized".

I wish I could invite readers of Edge to visit the Pirahã website and judge the charge of "decontextualization" for themselves. Unfortunately, I cannot. For some time now, any reader who attempts to access the Pirahã site sees a page with the text "Blocked Site Error". According to information provided by the archivists, this message signals the blocking of the site "at the request of the site owner". The blocking of these pages has also eliminated free access to all but one of the previously available Pirahã texts, and has prevented access to what (I believe) was the only publicly available Pirahã-English word list—two important research tools.


3.
I still hope it is possible to look forward to a calmer, more reasonable discussion of Pirahã grammar and culture, rather than what has unfolded. If the discussion improves, who knows—perhaps even the press will get the hint and report more carefully. Perhaps we will begin to see fewer articles with nonsense like the Chicago Tribune's claim that that Hauser, Chomsky and Fitch's (2002) article was a reply to Everett's (2005) article, and their statement that the Pirahã "don't use recursives". (There's no such thing as "a recursive"!) And perhaps we will even see a healthier, more relevant discussion of opposing views such as ours.


DANIEL EVERETT [6.25.07]

David Pesetsky's reply on the Edge corrects some errors I made with regard to the professional background of Andrew Nevins and Cilene Rodrigues. As he says, they are both established and experienced researchers. I in no way wished to indicate otherwise. They are all deservedly respected scientists.

I also do not want to give the impression, as I seem to have given to David, that I believe that their long criticism of my work is completely blinded by their own theoretical commitments. I do think that the history of science shows that our theoretical commitments deeply affect our conclusions, but certainly not entirely.

Now, to get to the point, in the article that Nevins, Pesetsky, and Rodrigues wrote and posted on the LingBuzz site, they purport to show, as David repeats in his Edge posting, that my arguments for the lack of recursion in Pirahã do not go through, based on my earlier data, i.e. from my PhD dissertation.

But data, as I discuss in my reply and as Chomsky has said in many different publications, data 'do not wear their analyses on their sleeves'. The way we analyze a sentence from any language depends to a large degree on the meaning that we assign to that sentence. Meaning and form are carefully linked in all linguistic theories. So if the linguist assigns the wrong meaning to an utterance, even a meaning slightly off from the speakers' intended meaning, the structure that is assigned to that utterance will be wrong as well.

The sentences that I record in my PhD dissertation are all fairly accurate as recorded utterances, but often their meanings were too subtle for my understanding of the language at the time I was writing my PhD dissertation, based on 14 months of field research and a beginning knowledge of the language. Long before I wrote my Current Anthropology article I had corrected the mistranslations of examples, but had not written a paper to discuss the implications of these more accurate translations. The Current Anthropology article did discuss these and I discussed them further in my reply to Pesetsky, Nevins, and Rodrigues. But, as Pesetsky points out, it is true, trivially so in fact, that when one makes claims of the kind I have made, no one article provides enough space to thoroughly document and argue for all the conclusions in it. This is why I, along with colleagues from MIT, University of Manchester, and University of Edinburgh, plan to write a much more detailed grammar of Pirahã over the next couple of years, based partially on experimentation to test all the more controversial claims about Pirahã.

Ultimately, though, I am puzzled by David's reply. First, I have replied to every single one of the criticisms he raises in his Edge reply in my paper on LingBuzz. There is not a single argument that he raises on his Edge posting that hasn't been answered. The Brazilian anthropologist he cites has indeed studied Piraha myths, but only in the sense of stories based on immediate experience that bind the Pirahas together. That anthropologist does not have a single text in Piraha on creation. I have stated this many times and Pesetsky has failed to note it each time.

Moreover, if you do not do fieldwork, you are an armchair linguist. That is simply the way the world works. You study languages in the field or from your armchair. This is not a moral judgment, though it becomes relevant when non-fieldworkers opine about the nature of fieldwork. David has never done fieldwork on any language, nor has his co-authors, that remotely approaches the Piraha situation in being monolingual, extremely difficult access, and so on. This is not to say that there is not a lot of important work that can be done without field research. But it does mean that Nevins, Pesetsky, and Rodrigues are much more sanguine about interpreting examples, such as those from my PhD dissertation, at face-value, apparently not realizing how hard it is to translate even simply examples in a field work situation and how being even just a little bit off in those examples' translations can dramatically affect the picture of the grammar that emerges ultimately.

Ultimately, once again, Nevins, Pesetsky, and Rodrigues have not added anything to our knowledge of Piraha or my work on Piraha except to show that I currently disagree with my PhD dissertation.

It is true that I am one of the few people that have worked on Piraha. But for just about any grammar of any smaller, difficult to access language we have in the literature, there is only one or a small handful of people that have studied the languages in question. It is as likely that we have many 'false positives' in these grammars (like my dissertation) that support the idea of recursion as it is that we have false negatives (as Pesetsky would have readers think of my 2005 paper) against recursion. More experiments are needed on all fronts. I have made my claims. What is needed are new experiments, not an expression of dismay that my dissertation is different from my current work.

Finally, I should say that the great, now deceased, MIT linguist, Kenneth Hale, claimed more than thirty years ago that it looked like several Australian languages lacked syntactic embedding (and a fortiori recursion), though he never investigated the implications or the nature of this claim as thoroughly as has been done for Pirahã.

Several Australian linguists believe that Hale was correct, and papers are beginning to emerge which support the idea that some Australian languages lack recursion too. Pirahã is unlikely, therefore, to be the only counterexample to the claims of Chomsky and others on the centrality of recursion.

I only hope that the debate, as David urges in his posting, will become more civil and less emotionally charged. This is just science, after all. Life is enriched by science, but is a lot more than that.


ROBERT D. VAN VALIN, JR. [6.27.07]
Heinrich Heine University, Düsseldorf

The dispute over Pirahã is curious in many respects, not least with regard to the fact that Everett is not the first linguist to claim that a language lacks embedded clauses and therewith recursion. In a series of important papers published in the late 70's, the late MIT linguist Kenneth Hale argued that certain Australian Aboriginal languages lack embedding of the type found in Indo-European languages in their complex sentences and furthermore that one of them, Warlpiri, has a completely 'flat' syntactic structure. The latter claim was amended somewhat in the published version of the paper, but the point about the complex sentences remained valid. In the mid-1980's, William Foley, a linguist at the University of Sydney, described Iatmul, a language of Papua New Guinea, as having non-hierarchical clause combining, i.e. no embedded of clauses in complex sentences, hence no recursion in the syntax. So the question arises, given that such claims go back a good thirty years, and the most important of them was from a former colleague of Chomsky's, why has Everett's claim engendered such controversy?

There are two reasons. The first is that it is only recently that Chomsky elevated the concept of recursion in the syntax to being the defining feature of human language. Prior to this, claims such as Hale's and Foley's would not have generated such interest, although Hale's claims did have theoretical consequences, albeit with respect to different issues. Chomsky stands at the end of a long and very distinguished career, and his current theoretical work and the claims about language evolution and specifically about syntactic recursion being the key property of human language are the capstones of that career. For them to be called into question now is a serious challenge to Chomsky, his work, and his intellectual legacy. Everett has raised important scientific questions, and they deserve serious and dispassionate investigation. Pesetsky et al.'s reply to Everett and his response to their criticisms are important steps in that direction, as are the other investigations of Pirahã that have been carried out recently and are planned.

One of the striking features of this controversy has been how little dispassionate discussion there actually has been. Indeed, both Pesetsky in his commentary and Everett in his response call for a calmer tone to the discussion. The non-linguist might well wonder, why so much sound and fury over a claim about syntax, of all things? The reason for this is summarized by Pinker in a comment quoted in the article in The New Yorker (p. 131):

There's a lot of strange stuff going on in the Chomskyan paradigm He's a guru, he makes pronouncements that his disciples accept on faith and that he doesn't feel compelled to defend in the conventional scientific manner. Some of them become accepted within his circle as God's truth without really being properly evaluated, and, surprisingly for someone who talks about universal grammar, he hasn't actually done the spadework of seeing how it works in some weird little language that they speak in New Guinea.

So, first of all, Everett's claim is a direct challenge to the "guru's" teachings, and second, Everett himself was once one of Chomsky "disciples" (he was once a leading figure in Chomskyan linguistics in Brazil and spent time at MIT) and is now apostate. A former true believer is always attacked in a way that those who were never believers are not. There is a further factor. Chomsky is perhaps best known to non-linguists for his prolific political writings, and while he has always taken great pains to keep his political work separate from his linguistic work, many people take both of them as two parts of a single oeuvre, so that an attack on his linguistic work is also perceived as an attack on his political work. This is particularly true in other countries, especially in those countries in which Chomsky's trenchant criticisms of the US government resonate with local attitudes toward the US.

I would like to revisit one of the points I made in my earlier contribution in light of one of the arguments made in Pesetsky's contribution. He writes:

For example, Pirahã appears to position its subordinate clauses after the verb—despite the tendency elsewhere in the language for the verb to come last in the sentence. Everett used this fact as an argument that what look like subordinate clauses in Pirahã are really independent sentences. This claim formed part of an argument that Pirahã lacks "recursion", which in turn was supposed to reflect Pirahã's special cultural restrictions. In fact, however, the grammatical property that he had identified in Pirahã is extraordinarily common in the languages of the world—and has no detectable correlation with recursion or culture. We offered German, Hindi, and Wappo (a language of California) as examples of languages with this property, and could have given many more examples. No cultural thread links the speakers of these various languages. (Who has ever accused German speakers, for example, of living exclusively in the "here and now"?)

There are three significant issues in this passage.

First, Pesetsky does not refer to Everett's counter-arguments to his claim that there are subordinate clauses in Pirahã given in the Pesetsky, et al. paper; he simply assumes that there are, without confronting Everett's arguments in his response.

Second, in comparing languages like German, Hindi, and Wappo with Pirahã, the crucial point is not that all four of these languages permit a clause to occur after the verb, which is true, but rather that in Pirahã it is the only option. Furthermore, as I noted with respect to Lakhota, the two possible placements of a linked clause, either before the main verb in an embedded structure, or in a flat, conjoined structure, are considered to be communicatively equivalent by native speakers. Thus, the lack of a particular structural option in Pirahã is not evidence of any kind of deficiency with respect to communication or with regard to semantic complexity.

Third, Everett does not claim that the fact that subordinate clauses follow the verb in German or other languages has anything to do with the kind of cultural constraint he proposes for Pirahã. The reason why this phenomenon is so widespread in verb-final languages is that it allows speakers to avoid what are called 'center-embedded' structures, which are difficult to process. An example of center-embedding from English is 'The cat [that the mouse [that ran away] saw] purred', which is composed of 'the cat purred' and 'the mouse saw the cat' and 'the mouse ran away.' This sentence is technically grammatical but very difficult to understand. Similar structures are created in verb-final languages if, for example, a subordinate clause follows the subject and precedes the verb. This point is, for languages which allow both options for the placement of subordinate clauses, there is a good processing reason why they prefer the post-verbal option. (Hence the comment, "Who has ever accused German speakers, for example, of living exclusively in the "here and now"? is an attempt to ridicule a claim which they imply follows from Everett's analysis, but which does not.)

What is so striking about Pirahã is not only that the language does not have the preverbal option, which languages which have true embedding do, but also that there is no evidence that in the two-clause structures that the second clause is in any way grammatically dependent on the first. The language is definitely exceptional in this regard.


DAVID PESETSKY [7.9.07]

When I expressed my hopes for a "calmer, more reasonable discussion of Pirahã grammar and culture", discussions like Van Valin's talk of disciples, apostates and gurus were definitely not what I had in mind. On the other hand, several of Van Valin's other remarks are important and productive. In these remarks, I would like to highlight and develop just one of them as an example of what calm, reasonable discussion of Pirahã might look like. I will not attempt to address everything.

At several points in our response to Everett's Current Anthropology paper, we suggested that various allegedly unusual properties of Pirahã grammar are actually found in other languages as well -- languages spoken in a variety of cultures. This was important to the argument because if we are in fact dealing with the same grammatical phenomenon in each language, the phenomenon is unlikely to have anything to do with culture. Nonetheless, we took pains to express ourselves cautiously, repeatedly stressing the "if" that I italicized above. It is always possible that two phenomena that look the same on the surface — for example, the position of (alleged) subordinate clauses in German and Pirahã — will turn out to be distinct "under the hood".

This is in effect what Van Valin suggests in the second of the "significant issues" that he raises. He calls attention to the possibility that languages like German differ from Pirahã in that German does not require its (alleged) subordinate clauses to follow the verb, but Pirahã does. If this is true, it might mean that the relevant word order laws of German are not the same as those of Pirahã (i.e. that our claim was wrong) -- or it could merely show that some independent factor intervenes in Pirahã to block the placement of a subordinate clause in one of the positions where German allows it. The second of these possibilities, of course, would require us to identify and understand the independent factor. That would be an important secondary task.

Before we get this far, however, we might start by trying to pin down the facts a bit better. For example, the tendency in German to put a that-clause after the verb is extremely strong -- though indeed not absolute. Is the situation really so different in Pirahã? At present, I think, we do not know. In our paper, we cited a Pirahã example of the form "He [me to-go] doesn't-want". This looks on the face of it like a counterexample to Van Valin's claim, since it appears to show an embedded clause ("me to-go") preceding the main verb. This example might suggest that the two languages are not so different after all. On the other hand, single examples do not decide arguments. One needs to develop carefully investigated paradigms of data (and, ideally, careful examination of texts as well). So there are plenty of investigations to carry out and much thinking to be done before we can settle the matter.


ROBERT D. VAN VALIN, JR. [7.10.07]

I'd like to make just a couple of comments with respect to David Pesetsky's reply to my posting. First, the reference to gurus and disciples was taken from Pinker's comments in the New Yorker article; that's why there are double quotation marks around those terms in my comments. The term 'apostate' I added, and it seemed not only to fit with Pinker's remarks, but it also alludes to Everett's past as a missionary and his later rejection of religion. Second, without getting too technical, there are two problems with the alleged counterexample to my claim, namely the 'he [me to-go] doesn't want' example, with reference to which Pesetsky says "it appears to show an embedded clause ("me to-go") preceding the main verb." The first problem is the status of 'me to-go' as a clause. In most theories, but not the one he works in, such things are analyzed not as clauses but as a unit smaller than a clause. Second, and more significantly, there is no evidence, despite their position before the verb, that they are in fact embedded. I have argued for many years that such constructions, even in English, do not involve embedding. So there are good reasons to believe that [me to-go] is neither embedded nor a clause.

I would also like to add that I completely agree with his final comments: "On the other hand, single examples do not decide arguments. One needs to develop carefully investigated paradigms of data (and, ideally, careful examination of texts as well). So there are plenty of investigations to carry out and much thinking to be done before we can settle the matter."


Return to "RECURSION AND HUMAN THOUGHT: WHY THE PIRAHÃ DON'T HAVE NUMBERS" A Talk With Daniel L. Everett


John Brockman, Editor and Publisher
Russell Weinberger, Associate Publisher

contact: [email protected]
Copyright © 2007 By
Edge Foundation, Inc
All Rights Reserved.

|Top|