More on Chomsky’s Binding Theory. It’s a good example of how generative rules are constantly formulated and re-formulated in light of new evidence—languages are infinite, there’s always new evidence—a seemingly endless process that to my mind undermines the entire concept of Universal Grammar (though not the fact of linguistic structure).
To undermine Binding Theory in particular, here’s a piece of evidence that complicates Binding Principle A. Of course, many linguists have presented reams of evidence to complicate Principle A as traditionally construed, but I’ve never seen this particular data-point, which, I think, complicates not only Principle A but also the centrality of c-command to anaphor distribution, which is what Principle A is supposed to account for.
Principle A states that one copy of a reflexive in a chain must be bound within the smallest CP or DP containing it and a potential antecedent. A reflexive is bound if it is co-indexed with and c-commanded by its antecedent Determiner Phrase (DP). Co-indexation simply means that both DPs refer to the same entity (e.g. , John and himself). C-command is a structural relation. In a syntax tree, a node c-commands its sister node and all the nodes dominated by its sister. In practical terms, a phrase in English will usually but not always c-command all the other words and phrases to the right of it (e.g., all the words spoken after the phrase):
According to Principle A, a reflexive pronoun (also called an anaphor in generative linguistics) must be bound in its domain. It must be co-indexed with and c-commanded by another DP:
In the sentence The girl loves herself, the anaphor is co-indexed with and c-commanded by its antecedent DP. Thus, the sentence is grammatical. The anaphor cannot refer to anyone but the girl. If you wanted the anaphor to refer to everything but the girl—that is, if you added a different index to the anaphor—then you would need to change the anaphor to a pronoun, it or her, to make the sentence grammatical: The girl loves it.
The sentence *Herself loves the girl is ungrammatical, according to Principle A, because herself c-commands the girl. But it’s supposed to be the other way around: the anaphor needs to be c-commanded. It’s not, so the sentence doesn’t work.
The notion of c-command is a vital component of nearly all theories of pronoun and anaphor distribution, even the ones that have completely overhauled Chomsky’s original Binding Principles. But look at the grammatical examples in (1) and (2) below:
(1) There was a man in an attic searching through an old photo album. Surprisingly, the man’s search turned up images of himself and not his son, like he had expected.
(2) The photographer thought his lab was developing pictures of his girlfriend. Surprisingly, the photographer’s lab developed pictures of both his girlfriend and himself.
The man’s search and The photographer’s lab are possessor DPs. They have the following structure:
With possessor DPs, the possessor is actually a second DP embedded within the DP that expresses the possessor-possessee relationship. In other words, the photographer is embedded lower in the tree than the photographer’s lab. I said a moment ago that a phrase in English will usually but not always c-command all the words and phrases to the right of it. The two examples above fall under “but not always”:
In (2), the photographer only c-commands lab; it is embedded too deep to c-command anything else. In (1), the man c-commands search; it is embedded too deep to c-command anything else. Neither DP c-commands into the Verb Phrase, which means that neither DP c-commands the anaphor embedded within the Verb Phrase. The anaphors in (1) and (2) are not c-commanded and thus not bound. This should trigger a Principle A violation, but according to my judgment and the judgment of several informants, (1) and (2) sound just fine.
If anaphor distribution truly relied on c-command, then (1) and (2) above should sound just as awful as *Herself loves the girl.
I said at the beginning that Chomsky’s Binding Theory has been called into question for many years now, but as far as I know, most attempts to re-theorize it continue to rely on c-command as an important structural element for describing constraints on anaphor distribution. However, the data presented here demonstrate that anaphors can still sound grammatical even when they are not c-commanded. This indicates that discursive contexts can override the constraints of c-command on anaphor distribution.