Gore's now classic introduction to Webster's Third outlines this type of Descriptivism's five basic edicts:
"1 — Language changes constantly;
2 — Change is normal;
3 — Spoken language is the language;
4 — Correctness rests upon usage;
5 — All usage is relative."
These principles look prima facie OK — commonsensical and couched in the bland simple s.-v.-o, prose of dispassionate Science — but in fact they're vague and muddled and it takes about three seconds to think of reasonable replies to each one of them, viz.:
1 — OK, but how much and how fast?
2 — Same thing. Is Heraclitean flux as normal or desirable as gradual change ? Do some changes actually serve the language's overall pizzazz better than others? And how many people have to deviate from how many conventions before we say the language has actually changed? Fifty percent? Ten percent?
3 — This is an old claim, at least as old as Plato's Phaedrus. And it's specious. If Derrida and the infamous Deconstructionists have done nothing else, they've debunked the idea that speech is language's primary instantiation. Plus consider the weird arrogance of Gove's (3) w/r/t correctness. Only the most mullahlike Prescriptivists care very much about spoken English; most Prescriptive usage guides concern Standard Written English.
4 — Fine, but whose usage? Gove's (4) begs the whole question. What he wants to imply here, I think, is a reversal of the traditional entailment-relation between abstract rules and concrete usage: Instead of usage ideally corresponding to a rigid set of regulations, the regulations ought to correspond to the way real people are actually using the language. Again, fine, but which people? Urban Latinos? Boston Brahmins? Rural Midwesterners? Appalachian Neogaelics?
5 — Huh? If this means what it seems to mean, then it ends up biting Gove's whole argument in the ass. (5) appears to imply that the correct answer to the above "which people?" is: "All of them!" And it's easy to show why this will not stand up as a lexicographical principle. The most obvious problem with it is that not everything can go in The Dictionary. Why not? Because you can't observe every last bit of every last native speaker's "language behavior," and even if you could, the resultant dictionary would weigh 4 million pounds and have to be updated hourly. The fact is that any lexicographer is going to have to make choices about what gets in and what doesn't. And these choices are based on ... what? And now we're right back where we started.