Wednesday, 11 March 2015

Why chomsky might be right about the evolution of language... but probably isn't. Part 2

This is the second part of a two part blog post. The first part can be found here.

I think I have shown that Chomsky does not deny the evolution of language. In fact, Chomsky has a very specific idea of how language evolved based on what he thinks the function of language is. For Chomsky, the function of language is thought. As with Fodor, Chomsky thinks that thought involves the interaction of symbolic representations according to certain rules. The two are thus inseparable. Chomsky is very clear on this:

"Without merge, there would be no way to assemble the arbitrarily large, hierarchically structured objects, with their specific interpretations in the language of thought that distinguish human language from other animal cognitive systems." (Bolhuis, Tattersall, Chomsky & Berwick, 2014)

Note that language is considered to be a cognitive, not a communication system... 

Gorillas and chimpanzees might have some version of a broad language faculty (FLB), but they don't have the narrow language faculty (FLN), because they don't have merge. 

This theory strikes me as elegant and flexible enough to weather significant criticism. Take for example, the Norman Wiley (2014) paper criticising Chomsky on the nature of inner speech. Wiley criticises Chomsky’s assessment of inner speech as having essentially the same nature as interpersonal speech. Wiley makes the point that inner speech does not have the same syntactic structure in the literature on children or in adults and that the lack of such structure is a major problem for Chomsky's linguistic paradigm.

Leaving aside the subjective nature of inner speech (who knows what it is like for a venerable Professor of linguistics), it seems to me that the introduction of merge allows Chomsky to avoid this criticism. Inner speech does not have to be as syntactically complex as interpersonal speech, any more than any instance of interpersonal speech has to be as syntactically complex as another. Running into a room and shouting "Fire" is no less communicative than shouting "Quickly, the building is on fire, everybody get out." Although their form is different, what both utterances have in common is they merge multiple concepts. Likewise, lack of grammatical complexity on much of inner speech is not a problem, because merge still operates even if it is not always expressed in fully grammatical form. 

Elegant Chomsky's Minimalist Thesis of merge may be, but I also believe that, even on its own terms, it has at least one major flaw that makes it evolutionarily implausible.

One of Coolidge's criticisms of Chomsky is that he does not take into account the comparative literature (i.e. literature comparing humans and other animals). I think Coolidge is spot on with this criticism, because even a passing familiarity with literature on Ape Language Research (ALR) demonstrates that apes at least do have something like merge. For example, the female chimpanzee Washoe was taught to use sign language and when she encountered something new she combined known signs into new sign combinations: water-bird when first viewing a swan; "metal-cup-drink" when introduced to a thermos flask. Although this research has been criticised, similar context-appropriate sign combinations have been shown for several other apes including bonobos such as kanzi, as discussed in Christine Kenneally’s excellent book "The First Word: The Search for the origins of language."

So if chimpanzees have merge why can’t they have or learn full human language and thought? Chomsky might say this is because of a lack of a similar enough FLB. But if this is true, then why argue that merge and the FLN are at the core of human language and not the FLB? 

Chomsky's actual argument is a little different but no less problematic. He simply states that "chimpanzees taught sign language demonstrably lack this combinatorial ability." For evidence he cites Yang (2013) “The Ontogeny and Phylogeny of language.” However Yang’s paper does not argue that chimpanzees lack the capability for combinatorial ability in sign language per se, but rather that they lack complex grammatical rules, as defined by a rather complex statistical test (Zipfs law), And only in one chimpanzee - the troubled case of Nim Chimpsky.

Chomsky seems to want to have it both ways. On the one hand arguing that the core of human language, what really makes thought and language human and provides the basis for its syntax, grammar and abstract symbolism, is a simple combinatorial process called merge. However when faced with this possibility in non-humans he denies it by referring to the lack of rule based grammar!

Which is it? Merge? Or rule based grammar? If it’s the former, then it’s impossible to deny other great apes have it and thus it cannot be the core of what is different about human thought and language. If it’s the latter then we're back to complex innate mechanisms, which are evolutionarily implausible, because nobody, not even Lewontin or Gould, claim that such complex mechanisms can spring into existence so quickly and by chance.

More generally it’s difficult to imagine how any cognitively complex animal could encounter something new and actually learn anything, without combining concepts. Of course this assumes cognition operates via concepts and representations. If it doesn't and a more radical embodied cognitive framework is better suited to describing cognition, then Chomsky is far from alone in being wrong.


  1. Michael, I thought Chomsky considered recursion as at the core of human language. This seems more complex than Merge and I don't believe any animals have shown this ability. What's your take on recursion versus Merge? Regards, Gerry

  2. Hi Gerry,

    Recursion is more complex than merge as it’s the nesting of a phrase within a phrase. Merge is how a phrase comes into existence in the first place. It’s a reduction of recursion to a simpler unit of analysis. This has the benefit of underlying recursion and other aspects of grammar, while not requiring it be used for recursion in all languages. As I mentioned in my first blog post, this may have to do with Dan Everetts documentation of the lack of recursion among the Piraha peoples language. If Everett is correct, then recursion cannot be the “proper unit of analysis”, the “smallest element”, the “cypher” that would crack the “code” of language, if you want to think about language in those terms (which Chomsky does).

    However nobody is going to deny the lack of merge in human language and thought. The simplicity of the merge mechanism solves some problems: lack of recursion among some languages; plausibility as a genetic mutation. But it also creates a major problem because it is difficult to deny such a simple mechanism exists in other animals, in which case merge is not the characteristic of human language he was looking for. There also doesn’t seem to be anywhere else to go – you can’t go further down than merge, and Chomsky’s theory demonstrates to my mind the futility of a reductionist approach in approaching such a complex system.

    I haven’t really said what I think about language but I agree with many of Chomsky’s critics that considering language as a behaviour that supports communication, social interaction and thought is a more useful way to think of it. Such an approach means we can look at “languaging” in the same way we might look at behaviours in any animal and we can deploy similar methodologies, such as studying differences between human and chimpanzee cognitive and social skills and predispositions, an approach Chomsky dismisses.

    In my talk on the intersubjectivity symposium on 28 march, I’ll argue that this approach makes the evolution of language less mysterious and that it may be far older phylogenetically than researchers have traditionally assumed, possibly predating Neanderthals.