The clearest instance I know of a discovery in English grammar that should have called for revision of certain traditional doctrines—though instead it was just ignored—is the radically improved understanding of prepositions offered by the great Otto Jespersen in The Philosophy of Grammar (1924), subsequently argued for in great detail by late-20th-century theoretical linguists like Edward Klima, Bruce Fraser, Joseph Emonds, and Ray Jackendoff.
The traditional definition of “preposition” says (and I quote the excellent Concise Oxford Dictionary): “a word governing (and usu. preceding) a noun or pronoun and expressing a relation to another word or element, as in: ‘the man on the platform’, ‘came after dinner’, ‘what did you do it for?’” This definition does have the virtue of being syntactic, rather than being mushily based on intuitive meaning as with traditional definitions of noun and verb; but unfortunately it’s completely inadequate.
The category of words that includes on, after, for, along with many other words (far more than most people realize—about a hundred) is not characterized by government of a noun or pronoun.
To paraphrase what is essentially Jespersen’s point, we really shouldn’t posit three different words spelled before. But, shocking though it is to report, the Concise Oxford does exactly that. It says that before is (i) a “conjunction” (because of sentences like I never saw him before that happened), and (ii) a preposition (because of sentences like I never saw him before that event), and (iii) an adverb (because of sentences like I never saw him before). This is more than just inelegant; it is positively irrational. Not only are the spellings identical for the three parts of speech, the syntax is also the same (before heads a phrase that functions as a temporal adjunct), and the meaning is unitary as well.
The right way to describe such words is to say that prepositions have varying degrees of versatility: Some are as limited as at (for which the traditional definition works); but a preposition like before can also occur with a clause as its complement, or with no complement.
It’s exactly the same word in each case, and has the same meaning, roughly paraphrasable as “earlier than the time indicated,” where the indicating is done either by the denotation of the noun phrase complement (before this event) or by the event time of the clause complement (before this happened) or by the time implied in the context. In Men had never reached the moon before we understand before as before X where X is the time referred to in the reach clause; and in I’ve never done this before we understand before as before X where X is the moment of utterance.
Before doesn’t behave anything like an adverb syntactically. For example, I had undressed quickly can be paraphrased as I had quickly undressed, but I had undressed before cannot be paraphrased as the ungrammatical *I had before undressed.
Before likewise has little in common with either “coordinating conjunctions” or “subordinating conjunctions.”
The modifier word right, as in right in the middle, is entirely specialized to the role of modifying prepositions in contemporary standard English. It cannot modify adverbs, so *Please do it right immediately is not grammatical; yet it occurs happily with before, regardless of what the complement is: He had called right before bedtime or He had called right before she went to bed or He had called right before.
Similar evidence points to a number of other words being prepositions. Some of them never take complements (or at least never take noun-phrase complements): away, back, home, now, here, and there, for example, are all best treated as prepositions that don’t take complements. They can all be modified by right, and they all take the positions in sentences that preposition phrases take.
The full statement of the case against what the standard dictionaries say (I cannot develop it in detail here) is overwhelming. Yet no English grammar book during the 20th century pointed this out. (As far as I know, The Cambridge Grammar of the English Language was the first reference grammar to adopt the Jespersen-style view of prepositions; see especially Chapter 7.)
In case you think this is new stuff, and dictionaries couldn’t be expected to have caught up yet, let me mention that Jespersen’s remarks were not without precedent: The basic insight can be tracked back as far as the 18th century (see this paper for references). There has been plenty of time for the idea to be accepted.
But instead (as I said in an earlier post), although linguists produce empirical discoveries and theoretical developments relating to the grammar of English (and of course other languages), they are ignored by the wider world. Linguists are just not breaking through to the point of influencing what is taught to the general public.