Sounding out ideas on language, vivid sensory words, and iconicity

But is it grammar?

Finally, some commentaries on the Evans & Levinson paper are trickling down the blogosphere. Nigel Duffield‘s “Roll up for the mystery tour” is one. Unfortunately, the comments on that post are closed. I have a question, so let me just post it here, where the comments are open.

The commentary is entertainingly written. Basically, it agrees with E&L’s rallying cry for the need to describe and recognize diversity; but it argues that, due to certain misconceptions about UG on the part of E&L, “Universal Grammar … walks free from the courtroom.” The first point that E&L get wrong about UG according to Duffield is the status of the notion ‘subject’. He has an interesting quote from McCloskey to support this point; and he also points out that, ironically, the topic/agent/pivot distinction championed by E&L is in fact ‘commonly accepted, if differently formalized’ in (some) UG quarters.

But it is really the second point, about the ontology of UG, that piqued my interest. Duffield argues that “UG is a theory of the initial state, which Chomsky now terms FL (Faculty of Language), not of any particular endstate grammar (LEnglish, LJiwarli, LPiraha, etc.,).” But that is not all; he adds, “The problem is not merely that UG is not claimed to be a property of final state grammars, but that it need not even be definitional of these grammars.” (emphasis mine, MD). And then comes the crucial point, for me as well as for Duffield:

the crucial point here is that facts about attained, endstate grammars bear only tangentially on theories of UG. Baldly stated, the absence of Language Universals—granting for the sake of argument that these are a ‘myth’—does not imply the absence of UG.

To be honest, this baffled me. Not so much because I disagree, but rather because there is so little left to agree or disagree with! I have wondered before (in an admittedly tongue-in-cheek post on the ‘grammar of the gaps‘) about the gradual shift of UG to evermore abstract territories — compare for example the switch-like parameters of the Principles & Parameters approach (some of which clearly refer to concrete (endstate) grammatical phenomena) with the most recent claims that recursion and some form of Merge should be sufficient for the FLN (Hauser, Chomsky & Fitch 2002). Duffield himself quotes Chomsky to the effect that ‘It is a coherent and perhaps correct proposal that the language faculty constructs a grammar only in conjunction with other faculties of mind.’ (Chomsky 1975:41). In my earlier post, I mention the question posed to Adele Goldberg by Jan-Wouter Zwart at the Nijmegen Lectures 2007. Zwart, in search of common ground between generative grammar and construction grammar, asked ‘Is it conceivable that underlying the structure of constructions are abstract principles of a simple kind, rooted in universal properties of human cognition?’ Goldberg’s answer was affirmative — but as I note in my discussion, the statement is sufficiently general to engender agreement from almost everyone.

But is it grammar?

The big question such abstract conceptions of UG raise for me is this: but is it grammar? That is, if it is indeed the case, as Duffield holds, that UG is not a property of endstate grammars; that it is not even definitional of these grammars; and moreover that ‘attained, endstate grammars bear only tangentially on UG’, what exactly is UG supposed to be, how do we go about empirically validating the UG hypothesis, and why are we calling it “universal grammar“?

My worry goes deeper than the apparent misnomer (though I do wonder whether the theory is not in need of a new name, if ‘endstate’ grammars have so little to do with it; but then I’m probably overlooking useful connotations of ‘grammar’ for the initial state). I find it difficult to see (1) how such an abstract concept could be isolated from more general cognitive abilities, and (2) why one would want to isolate it a priori. To isolate it —i.e. to show that UG is the ‘language faculty’ in some relevant sense— would one not need to show that its core business is language? (But how would one go about that if it doesn’t necessarily show in ‘endstate’ grammars?) And would one not need to show, conversely, that more general cognitive abilities cannot take care of language — in other words, that UG is necessary to explain (aspects of) language usage and language structure?

Excavating UG

A final issue, prodded by another statement from Duffield’s commentary:

No matter how deep one digs into mature grammatical systems, there is no logical reason to expect that one will excavate UG in any recognizable form, any more than one should discover universal principles of embryology through an in-depth study of mature organisms.

This does raise the question of what linguistics as a science is looking at. Following the analogy here (which is always a dangerous thing to do, but then, it is perhaps a dangerous analogy), Duffield seems to say that UG is to language structure what universal principles of embryology are to mature organisms. That would imply that UG is not about language structure (since digging into language structure is not going to yield UG ‘in any recognizable form’) but about the early ontogenesis of language in acquisition. I would agree that acquisition is of central importance (though again, I don’t see why we shouldn’t try and see how far we get with (1) domain-general cognitive abilities and (2) a socially grounded approach, before assuming there has to be a non-trivial ‘language organ’). But what of the flood of generativist literature that does dig deep into ‘mature grammatical systems’, purporting to learn things about UG (presumably in recognizable form)? Does this literature bear more than a tangential relation to the notion of UG espoused here?

Now, I expect there are interesting answers to these questions. No doubt I have overlooked some important ramifications, and perhaps I, too, have mistaken the ontology of UG. If so, set me straight! Comments are open.


In a review of the debate, psychologist Michael Corballis puts it well:

In defense of Chomsky, we should note that universal grammar does not apply to the external languages (or E-languages) that we actually use, but is a feature of what he calls internal language, or I-language. In this view, diversity arises not in I-language itself, but in the mapping of I-language onto E-languages. So could there still be common principles within I-language that could accommodate the huge diversity of E-languages? In Chomsky’s most recent formulation, the so-called “minimalist program,” universal grammar reduces to what he calls “unbounded Merge.” That is, unspecified elements are merged in recursive fashion to create structures of any desired level of complexity.

Perhaps the question, then, is not so much whether universal grammar is dead, but whether it has sufficient vitality to be useful. Yes, all languages probably involve the merging of elements of various sorts-that’s what makes language generative. But rather than dwell on this simple idea, we might be better to study languages in all of their diversity, and understand how they are shaped by culture, and by the pressures of the everyday commerce of cooperation and competition.

I find a theory which locates the most important restrictions on language learning or language structure solely in some ill-defined innate organ uninteresting; at best it prejudges the issue and at worst it trivializes what is so interesting about language as a complex evolved system. Restrictions on evolved systems (and hence on our theories of evolved systems) come from a multifarious evolutionary landscape in which cultural transmission and social interaction are just as important as biology and cognition.


  1. Chomsky, Noam. 1975. Reflections on Language. 1st ed. New York: Pantheon Books.
  2. Evans, Nicholas, and Stephen C. Levinson. 2009. The Myth of Language Universals: Language diversity and its importance for cognitive science. Behavioral and Brain Sciences 32: 429-492. DOI: 10.1017/S0140525X0999094X
  3. Hauser, Marc D., Noam Chomsky, and W. Tecumseh Fitch. 2002. The Faculty of Language: What Is It, Who Has It, and How Did It Evolve. Science 298: 1569-1579.

4 responses to “But is it grammar?”

  1. Mark, I agree with your unease about the implications of my assertions about the ontology of UG: I’m not convinced either that radical Minimalism has much to say any more about grammar. My criticism of E&L was not a defence of UG, simply an observation that claims about the validity (or otherwise) of UG are logically independent of claims about Language Universals.

    My own view, having worked with researchers on both sides of the debate —including MPI colleagues— is that UG/FL as an object of inquiry is not terribly interesting, nor is it as well-motivated by language acquisition facts as its practitioners maintain: in that respect I side with more descriptive typologists (and with less rhetorical generativists!) in preferring to explore relatively superficial syntactic variation.

    And yet my own specialist research on languages as apparently diverse as Modern Irish and Vietnamese—the two languages I have focussed most on over the last 20 years—convinces me, contra E&L, of the reality and substance of (formal) Language Universals: there are simply too many quirky, and “dysfunctional” commonalities between grammatical systems not to reasonably suspect the operation of universal principles. (I discuss this further in a chapter of my manuscript on Vietnamese, which—God willing—will be finished in the middle of next year. Meantime, I’ll post the relevant chapter soon.)

  2. Nigel, thanks for responding. I’d be interested to hear more about the Modern Irish/Vietnamese stuff. Many recurrent commonalities could of course be chalked up to the “stable engineering solutions” of cultural evolution that E&L talk about — there seems to be no a priori reason to invoke a universalist ontology for that kind of thing. So I suppose the “quirky” and “dysfunctional” are crucial to your argument.

  3. I would direct you to Robert Hanna’s book – an attempt to square a “proto-logic” with Chomsky’s UG, to construct a “logical cognitivist” view.

    The claim is indeed more general than a grammar as normally understood, and I think Chomsky’s own writing bears this out. The claim that an “I-language” is making is that it is supposed that there is an interfacing between sent and received language.

    If we change the topic from what we imagine language to be “discrete sounds integrated with visual stimulus, traversed by speakers communicating”, and concentrate instead from the frame of “thinking”, then I believe Hanna is onto something in claiming that such a thing as language is really a question about logic.

    That is, in the study of logic it has been shown that, so far, no proto-logic can be found. There are many logics. Classical logic, extended classical logic(s), deviant logic(s) including paraconsistent logics in which maintaining truth and falsity simultaneous is made to function by ejecting something from classical logic.

    This is due to the dilemma that Godel pointed out and others developed, chiefly that a formal system is only coherent if it is incomplete. If the system is complete (having the rules to verify itself) then it is necessarily incomplete. This leads to the frustrating problem of deductive logics seeming to be domain-specific; essentially arbitrary. It also suggests that logic is circular, or recursive – the rules of inference in a logic are “explained” only by showing the domain that the logic exists within (as opposed to being domain-independent).

    The claim of UG, then, is that there is an object in the ontology which sits at the top of the hierarchy, and human languages – human thought perhaps – are transforms from the general state.

    This does make me find it confusing that one author quoted above claims that there should be no trace of UG in any human language. It seems like it should be simply the *only* shared element of all human language, which if removed would eliminate any sense of relation between human languages in one fell swoop. Perhaps they are thinking narrowly of some factor within linguistics as being the determining element, which may no longer be “visible” in human languages within the grammar.

    That is, it could very well be that this element in the ontology is not a grammatical object, but the “circular” process of inference itself, which for instance has the appeal of explaining the creative property of developing infinite language constructs out of a finite set of elements.

Leave a Reply

Your email address will not be published. Required fields are marked *