The whole prescription/description argument about usage strikes me as embodying a category mistake. Prescription and description are properly theories of lexicography: A dictionary can either try to reflect the language as it is actually used, or it can try to reflect the language as those who are its most skilled users think it ought to be used. But we all have to be prescriptive in deciding what to say ourselves and in evaluating (when we must) things said by others, such as our students or writers whose word choice we decide to make fun of. (Not, I would argue prescriptively, “of whose word choice we decide to make fun,” because this is an informal rather than a formal setting and because “make fun of” is an idiom that can properly be treated as a “virtual verb” in such constructions.)
Now in prescribing to ourselves or others how to use the language, we can take either an “elitist” view, which will tend toward conservatism in accepting change, or a “democratic” view, which will tend toward ready acceptance of innovation. An elitist will want to use dictionaries and usage manuals prepared according to prescriptivist theories, while a democrat will find descriptivist manuals more useful. I tend to elitism, while Eugene Volokh has democratic leanings, and we’ve gone back and forth on issues such as (not “like”) the “nucular” pronunciation.
But today he boldly stands up for the grossly slandered split infinitive, and I proudly stand shoulder-to-shoulder with him against the hordes of offended pedants who assail it. The prejudice against the split infinitive stems, I have been told, from the tendency to try to fit English grammar into the categories of Latin. Where Latin makes its verbs do different jobs mostly by inflecting their endings, English does the same thing mainly by adding auxiliary words. In Latin, the infinitive is a single word, which is, as we all know, unbe-f*cking-leivably hard to split. In English, though, it’s two words, and there’s no earthly reason not to put an adverb between them when the construction makes it natural to do so. No one would think twice about “splitting” a phrase including a verb and its other helpers: “I was happily going to the store” is both impeccable grammatically and different in meaning from “I happily was going to the store.” Why treat the infinitive differently, just because it was a single word in a different language?
Stubbornly to insist on unsplit infinitives is pointlessly to annoy the reader.
UPDATE
A reader points out that the final sentence above is needlessly tortured, and proposes the following unsplit construction: “To insist on unsplit infinitives stubbornly is to annoy the reader pointlessly.” He’s right about the final phrase, which is much better with “pointlessly” at the end. I’d still put “stubbornly” in the split position, though, to avoid having the reader try to parse “stubbornly is to annoy” until it becomes obvious that the adverb refers back and not forward.
[In a follow-up post, Eugene identifies the Latinate origins of the taboo — with Dryden as the villain of the piece — and claims a similar origin for the preposition-at-the-end taboo, up with which I will not put.]