William Safire often strikes me as being akin to the 19th century grammarians who told us we simply must not split our infinitives.
Why?
Well, it began with a faulty premise. . .
That language had devolved from some perfect state.
For them, of course, this perfect state was Latin.
Now, in Latin, you don't split infinitives; therefore, in order to have English be nearer this "more perfect" language, they decreed that we should avoid doing so as well.
Leaving aside the obvious absurdity of believing one language to be syntactically superior to another, the punchline of this exercise is rather obvious. . .
There is a very simple reason why Latin speakers never split their infinitives. . .they can't! It's affixed to the root--unlike in English.
Where was I? Oh, William Safire. . .
He makes a habit of confusing
prescriptive or learned grammar (the don't split your infinitives nonsense) with the
descriptive (those rules that naturally follow from the underlying (acquired)structure of a language).
A structure such as "attorney general" would not naturally arise in modern English; and if the archaic form and its equally archaic pluralization "attorneys general" were not learned, it would properly be analyzed in accord with the "rules" of English and realized as "attorney generals". In other words, it would be analyzed as a noun (such as book-shelf), and not as an otherwise (but for the learned and accepted)
"ungrammatical" noun-adjective construction (i.e. adjectives prececde the nouns they modify in English).
A problem with many of these learned forms is that they can not be readily applied to seemingly similar cases--and when they are, it is often done incorrectly. They really aren't "English", if by English you mean those constructions that are in accord with the underlying "rules".
Ironically, I think the use of these types of constructs could well serve to speed language change--by creating more opportunities for for "mis-analysis" at the acquisition stage (as happened with the shift in English from a verb second language such as German)--thereby increasing the very "ungrammatical" utterances that people such as Safire constantly complain of.
I'd better stop--this could go on indefinitely.
Here's another interesting tidbit. . .
American: "Manchester United is. . ."
Brit: "Manchester United are. . ."
Leaving aside learned behavior, it really depends what you're conceptualizing, a unified or abstract whole, or a collection of individuals and interests (or something along those lines).
BTW, for the lay person interested in linguistics, I can't recommend Steven Pinker's book "The Language Instinct" highly enough.