Tuesday, October 23, 2012

Empiricism, Rationalism and Generative Grammar

 Here is my first rule of research:

 Those things not worth doing are not worth doing well. 

So, qua generative linguist, what’s worth doing? We get a handle on this by asking what are the central questions driving the Generative Enterprise? Two immediately come to mind: (1) What’s in UG and (2) Why is it there?  Taking these two queries as dispositive identifies the object of inquiry as the structure of UG.  The point of generative research is (or should be) to limn its fine structure and research is (or should be) evaluated by whether it helps us achieve this end.

This is a relatively focused conception of the goal(s) of Generative Grammar and it relegates many kinds of inquiry (e.g. what’s the structure of the Japanese DP?, how do kids use language to order chicken fingers in restaurants? Does language L allow multiple case checking? Are relative clauses islands in Swedish?, Are bound pronouns spelled out traces, etc.) to (at most) a subordinate status.  So why do I focus on (1) and (2)? Here’s one reason why.

I grew up in a philosophy department and learned about Chomsky’s work in linguistics through a series of philo debates in the early to mid 70s, most especially  “the innateness controversy.” In those days, linguistics was cutting edge as it was the major battle-ground on which two great philosophical traditions -Empiricism and Rationalism- met and disputed.  For me and my friends, Chomsky stood toe to toe with Plato, Descartes, Leibinz, and Kant and led the charge against the forces of darkness viz. Skinner and Quine, heading the party of Aristotle, Locke and Hume. Unlike Vegas, what happened in linguistics did not stay there, it leached out into the wider intellectual world and had big consequences, or so it felt to us (especially over beers, on Friday nights, in downtown Montreal, in our early 20s). The stakes were high, nothing less than the nature of mind and its relation to the external world.  Linguistics was the leading edge of the cognitive revolution, the best case against the empiricist conception of mind and a model for the emerging cognitive sciences.  How did linguistics manage this?  Here’s a potted reconstruction.

Empiricism, a species of environmentalism (natural selection being another), holds that minds are structured by the environments in which they are situated. The leading metaphor is the mind as soft perfectly receptive wax tablet (or empty cupboard) which the external world shapes (or fills) via sensory input.  The leading slogan, borrowed from the medievals, is “nothing in the intellect that is not first in the senses.”  The mind, at its best, faithfully records the external world’s patterns through the windows of sensation.

Rationalists have a different animating picture. Leibniz, for example, opposed the wax tablet metaphor with another: ideas are in the mind in the way that a figure is implicit in the veins of a piece of marble.  The sculptor cuts along the marble’s grain to reveal the figures that are inchoately there.  In this picture, the environment is the sculptor, the veined marble the mind.  The image highlights two main differences with the empiricist picture. First, minds come to environments structured. They have a natural grain, allowing some figures (ideas) to easily emerge while preventing or slowing the realization of others. Second, whereas a hot wax imprint of an object mirrors the contours of the imprinting object, there is no resemblance between the whacks of the chisel and the forms that such whackings bring to life.  Rationalists allow minds to represent external reality but deny that they do so in virtue of some sort of similarity obtaining between the sensory perceptions and the ideas they prompt. Thus, whereas Rationalists postulated causal connections between mental content and environmental input they denied that environments shape those contents.  The distinction between triggering and shaping was an important one.

Associationism is the modern avatar of empiricism.  The technology is more sophisticated, neural nets and stimulus-response schedules replacing wax tablets and empty cupboards, but the guiding intuition is the same. Minds are pattern matchers able with sufficient exposure to the patterns around them to tune themselves to the patterns impinging on them. What made Chomsky’s ideas about Generative Grammar so exciting was that they showed that this empiricist picture could not be right.  To account for a native speaker’s linguistic competence requires that humans come equipped with special purpose mental procedures and this is inconsistent with empiricisms associationist psychology.  Two features of linguistic competence were of particular importance: first that the competence emerges relatively rapidly, without the learning being guided and despite data that is far from perfect. Second, much of what speakers know about their language is not attested at all in the data they have access to and use.  No data, no possible associationist route to the mind. Ergo: the mind must be structured. Point to the Rationalists. 

In retrospect, it is hard to see why we were so surprised and animated by Chomsky’s arguments.  Indeed, considered naively the idea that humans come equipped with a species specific dedicated linguistic capacity is the ‘duh’-position.  Based on simple observations it’s clear that nothing else learns language as we do. Indeed, nothing else comes anywhere close. Only a sophisticate could conclude otherwise.[1] However, so widespread was the empiricist perspective that the cognoscenti took it to be simple common sense. Hence, it required a sustained frontal attack to displace it.  That’s why watching Chomsky topple empiricism was so exciting and why work in generative grammar reached beyond linguistics to influence thinking in cognitive science, philosophy and computer science as well.

In sum what made linguistics exciting (and still makes it exciting) is that it provides an easy way to plug into a very great long-lived debate about the structure of the mind. All you need to do to participate is the following: next time you read or hear a paper, ask yourself (or the lecturer) “what does this tell us about the structure of UG?”

[1] Come to think of it, maybe this is what made empiricism so attractive to the learned; it challenges the obvious and so has the sheen of scientific sophistication? 


  1. The problem I see with the attitude expressed in the first 2 paragraphs is that people who don't care about the details of how languages work will never be able to make a truly convincing case for any kind of UG, for the reason that among the bright undergraduates there will always be a substantial fraction who would rather pay attention to people who seem to care about the details of the evidence they are advancing for their positions than to people who don't. & if the people who do seem to care are philosophical empiricists, that will assure a steady supply of recruits to that position.

  2. I am not sure that I see Avery's point. I did not mean to imply (though the trouble with tongue in cheek prose is how it shields your intentions) that detail work was unimportant. It is. But, and I do mean this, it is not important in itself, at least if one is a classical generativist. It is important because of what it can tell us about UG. If UG cares about grammars then it can only be investigated by studying grammars. Now grammars must be built by linguists and this is a very complex process, and requires addressing and adequately answering many issues of detail. Some people care about these details in themselves (Chomsky dubbed this finding grammars that are descriptively adequate). I don't actually. But I care about them if they are stepping stones to understanding UG (to allow descriptions of UG that are explanatorily adequate in Chomsky's parlance). UG is where the buck stops and that's what makes these filigree concerns interesting. So should we care about the details of how language work. Yes, but mainly those details that bear on UG and to know which these are it is always worth asking, as it seems to me that too few do: what does this work tell me about UG assuming it is correct?

  3. This comment has been removed by the author.

  4. Good clarification. But isn't it ironic that you can only make a really good case for UG by relying on the work of people who don't care about that so much, but do care about getting the details of their language right.

  5. Here I believe that we would disagree. We have learned a great deal about UG by thinking though the logic of very simple cases (think aux inversion and its limits, or the boundedness of movement and anaphoric domains). Oftentimes the real filigree work has, in my view, not yielded insights proportional to the effort expended. But that's my view, others may (and surely will) disagree. However, even were you right, the insights would not arise unless you asked the question and this I fear is no longer pro forma, as it was when you were starting out in the field.

  6. This comment has been removed by the author.

  7. This comment is not worth commenting on. Sorry.