criticisms of grammaticalization

[Suzanne Kemmer]

I take Martin's points in general, but his historical comment below  
makes no sense to me.

If in the generative world view, syntactic patterns are epiphenomena
that fall out of their theories,  then why would generative linguists  
refer to
observations about grammaticalization as epiphenomena, since they
can't see how these observations fall out of any theory of  
grammaticalization?

Perhaps there is already a dual usage in Linguistics : 'epiphenomena'  
is being used to mean 'complex surface patterns originated by deep  
underlying principles (and such patterns are worth studying because  
they reveal the deep principles)'; vs. 'trivial surface patterns that  
cannot reveal anything meaningful about deep underlying causes'.   
It's the latter sense that I am most familiar with in the form of  
remarks at conferences and in reviews of functional work.

I hope this word falls out of fashion in linguistics sooner rather  
than later.
--Suzanne

Martin Haspelmath wrote:
I think the label "epiphenomenon" for grammaticalization theory arose  
when generative linguists looking at what grammaticalization  
theorists said could not find a theory easily comparable to their  
familiar "binding theory" or "X-bar theory". In the generative world  
view, binding facts and phrase structure facts are epiphenomena  
falling out from the respective theories, but grammaticalization  
phenomena do not fall out so easily from grammaticalization theory.