I recently co-wrote a comment (see here) on a piece by Berlinski and Uriagereka on Vergnaud's theory of case (see here and here), of which I am a fan, much to one of my colleagues continual dismay. The replies were interesting, especially the one by Berlinski. He excoriated the Jacobian position I tried to defend. His rebuttal should warm the hearts of many who took me to task for my thinking that it was legit to study FL/UG without doing much cross G inquiry. He argues (and he knows more about this than I do) that the Jacob idea that the same fundamental bio mechanisms extend from bacteria to butterflies is little more than myth. The rest of his comment is worth reading too for it rightly points out that the bio-ling perspective is, to date, more perspective and less biology.
Reaction? Well, I really like the reply's vigorous pushback. However, I don't agree. In fact, I think that what he admires about Vergnaud's effort is precisely what makes the study of UG in a single language so productive. Here's what I mean.
Vergnaud's theory aimed to rationalize facts about the distribution of nominals in a very case weak language (viz. English). It did this elegantly and without much surface morphology to back it up. Interestingly, as readers of FoL have noted, the cross G morpho evidence is actually quite messy and would not obviously support Vergnaud's thesis (though I am still skeptical that it fails here). So, the argument style that Vergnaud used and that Berlinski really admires supports the idea that intensive study of the properties of a single G is a legit way to study the general properties of FL/UG. In fact, it suggests that precisely because this method of study brackets overt surface features it cannot be driven by the distributions of these features which is what much cross G inquiry studies. Given this logic, intensive study of a single G, especially if it sets aside surface reflexes, should generalize. And GG has, IMO, demonstrated that this logic is largely correct. So, many of the basic features of UG, though discovered studying a small number of languages have generalized quite well cross linguistically. There have, of course, been modifications and changes. But, overall, I think that the basic story has remained the same. And I suspect that not a few biologists would make the same point about their inquiries. If results from model systems did not largely generalize then they would be of little use. Again, maybe they aren't, but this is not my impression.
Ok, take a look at Berlinski's comments. And if you like this, take a look at his, IMO, most readable book Black Mischief.