[Corpora-List] Chomsky and computationnel linguistics

[Dominic Widdows]

On Jul 11, 2007, at 2:40 PM, Oliver Mason wrote:

> Just a quick remark, which I feel summarises my main issue that I have
> with non-empirical linguistics:
>
>> ...  An example is the dialects of English that have
>> subjectless for-to constructions, e.g. "I want for to go."  There  
>> have
>> been several papers on this construction in this dialect, and  
>> IIRC, all
>> the work was done on the basis of constructed examples, because  
>> there is
>> very little corpus on this.
>
> So you study a phenomenon based purely on made-up examples?  In other
> words, I can invent my own dialect, make up some odd sentences, and
> write papers about it (and presumably get them published in journals
> on theoretical linguistics)?
>
> This might sound a bit exaggerated, but I cannot really see how that
> is different from studying language purely based on intuition.
> There's a very fitting quote by John Sinclair about how botanists
> don't study artificial flowers.  If we want to find out anything about
> the true nature of language then we will need to study authentic
> examples, not invented borderline sentences about farmers beating
> donkeys and colourless ideas.

But the nature of much laboratory science is precisely that you study  
made up examples. We don't know the speed of light because someone  
summarized thousands of natural examples of light rays and deduced a  
commonality: we know if because Michelson and Morley invented an  
extremely artificial light ray and measured that one.
Botanists don't study artificial flowers, but I'm sure molecular  
biologists do. So the position "artificial examples have no place in  
science" is hard to credit. To defend your point, you would therefore  
need to argue that they have no place in linguistics: and so you'd  
also need to demonstrate why linguistics is different from other  
sciences.

My guess would be that artificial examples / laboratory experiments  
probably do have a place in linguistics as they do in other sciences.  
The hard things are i. picking examples that really tell you  
something important, and ii. providing a clear argument why these  
examples generalize. So for example, part of the value of the  
Michelson and Morley experiment is that we have good reason to  
believe (I think!) that they measured the speed of not one light ray,  
but all light rays - at least, the experiment has been repeated with  
the same results. This doesn't appear to have been done to the same  
extent with theoretical linguistics, and I certainly share your  
frustration that a fixation on quantifier scope in mammal beating  
scenarios doesn't really cut it.

Yours,
Dominic