One of my recurring frustrations with the way grammar is taught in the K-12 world is that by clinging so tenaciously to books that have not seen any real innovations in syntactic theory since the nineteenth century, teachers wind up with no explanation for many phenomena that occur all the time. A teacher sent me an email with the following question:

I've tried looking this up in every grammar guide I can find, but I haven't found the answer. Consider these two sentences: I bought a car to drive myself to work. I bought an alarm clock to wake me in the morning. I'm interested in the myself/me issue here. To me, both sentences seem correct, but I'm not sure I can justify the difference. Can you help?

I suspect that many teachers, when confronted with this pair, would give a semantic explanation along these lines: Well, if you said, "I bought a car to drive me to work," that would make it sound as if the car was driving you rather than the other way around. Notice, however, that such an explanation doesn't really tell us why the change of pronouns switches things around. It just appeals to our intuitive sense of how the grammar works to create meaning without actually identifying the grammatical constructions involved. Traditional grammar books, as this teacher said, give little help here. Warriner's, for example, merely says that personal pronouns with self are used "reflexively," without even explaining what that means. We can start by noting that reflexive pronouns are often used to refer to subjects: I hurt myself. *I hurt himself. On the surface, that would seem to explain the first sentence, but what about the second one? The subject of the sentence is still I. Why don't we use a reflexive pronoun there? It turns out that reflexive use is more complicated than just "refer to the subject." It has to refer to an immediate subject [examples from Haegeman and Guéron (1999)]: John considers that Mary should not contradict herself. *John considers that Mary should not contradict himself. The pronouns here are objects of contradict, and so the reflexivity restriction applies to the subject Mary, and not the subject of the outer clause. Given that background, what are we to make of the infinitives? It won't work to look for multiple subjects. The only word in either sentence that functions as a subject is I. But what if we think about the infinitive verb as having an implied subject? Infinitives, after all, can have explicit subjects (introduced by for), and when they do, reflexive pronouns obey the same rules we've seen above: I bought a car for my son to drive himself to work. *I bought a car for my son to drive myself to work. In this case, the reflexive pronoun refers to "my son," which is clearly the subject of to drive, even though that verb is non-finite. If we try to rewrite the two original sentences to contain overt subjects for the infinitive verbs, we get something slightly odd, but still comprehensible: I bought a car for me to drive myself to work. I bought an alarm clock for it to wake me in the morning. This isn't the normal way we would phrase either sentence, but it suffices to show that the subject of drive in the first sentence refers to I/me, and the subject of wake in the second to alarm clock. These two elements don't play the same role in the matrix clause: I is the subject and alarm clock is the direct object. So if we assume that myself and me refer directly to those phrase in the matrix clause, we can't formulate a consistent rule for using reflexive pronouns. On the other hand, if we assume that there is, in effect, a silent pronoun as the subject for each infinitive, then our rule for the use of reflexive pronouns remains the same. The direct object of drive refers to drive's subject, let's label it PRO (for an abstract pronoun) so it takes the form of a reflexive pronoun. The direct object of wake does not refer to wake's subject (also PRO), so it takes the form of an ordinary personal pronoun. In turn, each PRO refers to a different noun phrase in the matrix clause. Postulating an empty category (in essence, an implied noun/pronoun) may seem strange, and it is admittedly an abstraction, but it allows us to give a much more parsimonious account of English grammar than we would otherwise have. Also, empty categories are not completely alien to traditional grammar, which admits one such element in the case of the subject of imperative sentences (e.g., "Shut the window."), which is conventionally said to be an implied "you." One consequence of this analysis of infinitives is that they resemble clauses much more than traditional grammar (which analyzes them as phrases) admits. Indeed, contemporary analyses of English don't typically draw a sharp distinction between clause and phrase the way that traditional grammar does.