Simon Smith said

 

<at> Mike: Can you say any more about the theory that words do not have a pre-determined POS? I have to confess I had always thought the interpretation of "church" in "church tower" as an adjective to be nothing more than a non-linguist's misunderstanding. 

 

It seems more economical to say that any noun can (in principle) modify any other noun, as part of a noun compound, than to record an additional POS, namely adjective, for every single noun in the lexicon. I would say, too, that there is no more reason to make out a special case for substance-item compounds, such as "gold watch" and "lead balloon", than there is for "radium watch" (G Pullum's example) or "iridium balloon".

 

However, theories which assign POS elsewhere than in the lexicon would cause problems for that explanation (as well as for lexicographers, I would have thought).

 

That assumes there is a "true" POS which each word possesses. A rose is a noun is a noun, so to speak . Wouldn't it be more economical still, though, to say there are roles, and that almost any word can take on almost any role?  So that in the case of
(1) That "through" I saw in your essay should have been a "throughout"
both through and throughout were playing noun roles and we know that by their being preceded by that or a.