The Chronicle of Higher Education recently published an account of a divisive controversy in linguistics concerning the features of an obscure Amazonian language, Pirahã. Allegedly absent in Pirahã is ‘recursion’—the embedding of sentences within sentences—a feature that is (supposedly) central to human language according to the dominant Chomskyan account. The Chronicle article is unfortunately focused on the personalities involved and the interpersonal conflicts that have developed; more interesting for our purposes are the questions of evidence, falsification, and confirmation raised by Pirahã.

The 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 worked with the Pirahã, Daniel Everett , cannot return to the tribe, and that other linguists and anthropologists have not been able to interact with the Pirahã to replicate his results. Opinions differ concerning the analysis of his data; and many claim that recursion is not absent so much as subject to unusually strong restrictions. The discussion has left me with the following questions:

(1) Do the findings about Pirahã bear a greater evidentiary burden strictly as a result of their potentially discomfirming role? In other words, since Pirahã is the only discovered language to ever (potentially) exhibit such a startling property, and given the paradigm-shaking role this could play, shouldn’t the prior probability of the accuracy of Everett’s results be considered as rather low? It would make a great deal more direct evidence, replicated by multiple researchers, analyzed in an impartial manner, to shake our confidence in the basic Chomskyan view. Which is not to deny that falsifiability is possible from a single contradictory observation: but, as we must all acknowledge, this is on the assumption that the observation is essentially correct. (In other words—to borrow terminology from epistemology—do stakes matter to evidence?).

(2) Are anti-nativists and nativists talking past each other, to some extent? Take this piece, ‘The Slow Explosion of Speech’ which characterizes language as technology—something that developed in complexity as a result of refinement—rather than innate ability. But the technology view espoused here is focused on higher-order semantic and semiotic aspects. It is entirely compatible with a nativist view (such as ‘universal grammar’

) to accept the ‘refinement’ story and continue to maintain that this presents better use being made of pre-existing innate capacity. There perhaps is a wide space available for a syncretic view of language that has both significant innate elements and capacities together with environmental, learned functions: in other words, it is perfectly reasonable to expect language to express ‘phenotypic plasticity’ of a sort. Evolution can select for complex internal structure; what it cannot do is differentiate between what we term ‘mental’ and ‘physical’ functions. A kidney or a language organ appears the same. The evidence will tell the story here, but there is no prima facie reason to be skeptical of innate mental capacities. Nor is this incompatible with variation and conscious refinement.

(3) One commenter claimed the Chomskyan / nativist theory was a stagnating research project that has not produced fruitful results in a long time. Abstracting from the specifics of the case, this makes one wonder to what extent ‘progress’ —new results, new applications—is an genuine aspect of the confirmation of scientific theories. At first glance, it is not clear that constant, uninterrupted ‘progress’ is a necessary feature of a scientific theory in good epistemic standing. A hypothesis could be essentially correct (justified on the grounds of, say, an abductive inference) but researchers could lack the means for further investigation. A ‘correct’ theory could stagnate indefinitely. This seems particularly likely to happen in the case of hypotheses in cognitive science (broadly construed) where so little is known about the specific workings of the brain that well-supported abductive inferences couched in general terms could lay fallow for years or decades. (The ‘MERGE’ property seems a case in point: the specific neural underpinnings are completely unknown, despite the plausibility of the mechanism proposed: as Peter Ludlow commented here, “On Chomsky’s hypothesis, this did not involve a long period of gradual evolution, but a form of “punctuated equilibrium” and this was possible because wiring solution was driven largely by low-level biophysical processes. The idea is that because it is largely a product of physical and mathematical constraints it is not surprising that we see recursive structures in the product (just as we do throughout nature – in the growth of shells, etc). That is, recursion is evidence for the idea that the evolution of language was driven by the physical channel rather than by gradual evolutionary tinkering.” But of course nothing is known about how the ‘biophysics’ actually work out.)

Thoughts and arguments welcome.

(I note also that Everett’s book Language: The Cultural Tool does not focus on Pirahã, but instead marshals a number of other anti-nativist arguments against universal grammar; I have not read the book, and cannot comment on it.)