Barbara Wallraff was kind enough to respond at some length to my post about the metaphorical 'legal authority' of prescriptive rules (see the comment on the last post). She gives the fuller entry from OED which I cited only in part, as well as the American Heritage Dictionary. The former, as I noted earlier, lists the semi-auxiliary best/better (or 'defective auxiliary', as some would call it) as 'colloquial'. The latter -- a source I occasionally use, but didn't happen to have at hand -- is a little more assertive, very much along the lines of the argument of the Word Court column: "In informal speech, people tend to omit had, especially with had better, as in You better do it. In formal contexts and in writing, however, had or its contraction must be preserved: You had better do it or You'd better do it."
I think I now see the issue: Barbara Wallraff's advice seems to involve teaching the kid formal, written style at the age of three. I would assume that the style of language we'd find appropriate and, yes, 'correct' from a little kid can differ somewhat from what we expect in formal contexts and writing.
That mismatch is pretty salient to linguists: Correctness, we would argue, doesn't inhere only in the language of formal situations or certain kinds of written texts. In fact, people who speak standard and a dramatically different variety of the same language generally seem to have a sense of 'correct' language in particular situations. A friend of this blog (if blogs can have friends) is from the South and grew up speaking a dramatically different kind of English from what he uses daily now as a professor. He likes to say that if he went to his family reunion and spoke the way he does around us on campus, he'd be tossed out on his ear. That is, what's socially appropriate at his family reunion includes multiple negation, double modals (might could), and so on. To use formal English there would be ostentatious, at best. The situation is even more dramatic for many African-Americans, of course. This isn't some weird relativistic notion of 'correctness', but just an open acknowledgment that much of the richness of language resides in the range of styles and registers we have access to. I fully appreciate the value of usage guides (and cited Merriam-Webster's yesterday), and I think pretty much all linguists do (Steve Pinker had praise for Wallraff's last book, in fact), but kids can pick that up later, as they develop their range of social activities and contacts, and broaden their stylistic range.
One more little note, while we'll on the topic: Anything involving deletion (you had better > you better) instantly strikes most people as bad, wrong, lazy, etc. That's just a perfectly normal historical process, of course, and our language is filled with fully accepted products of such processes. More interesting here is the syntactic behavior here: semi-auxiliary better/best not only doesn't inflect (like beware, ought, etc.) but it also doesn't allow inversion in questions:
we do ~ do we?
we better ~ *better we?
This is a little piece of the bigger shallow-time-depth history of English, I think, but that's probably a topic for another time ...