On the eve of the Iowa caucuses, with the state’s Republican voters set to pick which member of the assholocracy they’d like to see take on Barack Obama, the New York Times runs the following headline:

Iowa, the Early Decider, Still Hasn’t

The English syntactic process that leaves a gap after the negative auxiliary hasn’t in the headline above is called verb phrase ellipsis (VPE). The idea behind VPE is that a sentence like the one in the headline above is derived from a structure in which the verb phrase in question is present: Iowa, the early decider, still hasn’t decided. The details of how one gets from this fully articulated structure to the VPE’d headline above differ among frameworks and implementations, but everyone more or less agrees on this basic premise.

VPE is a relatively well-understood process. We have known at least since the classic treatment of Hankamer and Sag (1976) that there are syntactic (as opposed to merely pragmatic) constraints on where the missing verb phrase can find its antecedent. One such syntactic constraint prohibits the antecedent (of, e.g., decided) from being found in a morphologically related noun (such as decider): this is, in the terminology of Postal (1969), an anaphoric island. Moreover, even Ward et al. (1991), in their thorough debunking of the idea that anaphoric islandhood is anything other than a pragmatic phenomenon, note that do so anaphora, a close cousin of VPE, is systematically/grammatically unable to find its antecedent in an anaphoric island.

All of this is to say that the headline above is not merely awkward, or stilted, or too-clever-by-half. It is ungrammatical. It is ungrammatical for the same reason that it is ungrammatical to call those who produce the aforementioned paper printers of all the news that’s fit to; and for the same reason that, if the Super Bowl gets cancelled, it is ungrammatical to describe that circumstance by saying that This highly anticipated happening didn’t; and for the same reason that it is ungrammatical to say, of the author and editor responsible for the headline above, These presumably-native English speakers nonetheless can’t. Forget about split infinitives: this is honest-to-goodness ungrammaticality we’re talking about here. Perhaps if this campaign without end ever does, the NYT can turn its attention to English syntax.

Update, Jan. 3: Perhaps I have been a bit hasty. It seems that there is a difference in acceptability for some speakers between examples like the headline above, where the anaphoric island for VPE is contained in an appositive/parenthetical phrase, and the ones I made up, where the anaphoric island is more fully integrated into the clausal syntax. Consider the following pairs:

1a. ? Iowa, the early decider, still hasn’t.
1b. * The early decider still hasn’t.

2a. ? Presumably native English speakers, these people nonetheless can’t.
2b. * These presumably-native English speakers nonetheless can’t.

It thus seems that being in an appositive/parenthetical phrase mitigates the anaphoric island effect for VPE somewhat, providing a linguistic antecedent that might be analyzed as outside of the clausal syntax proper. Put differently, these data suggest that clausemate anaphoric islandhood is truly deadly for VPE, but intersentential anaphoric islandhood is a bit less bad; cf. also the following, which seem to me to be equal in acceptability to the (a) examples immediately above, and not as bad as the (b) examples:

3a. ? Iowa is the early decider. Its voters, however, still haven’t.
3b. ? These guys are presumably native English speakers. All the more surprising, then, to find that they can’t.

There are undoubtedly further details to be unraveled here; for example, I would not be at all surprised to find that the choice of auxiliary affects the acceptability of anaphoric-island VPE (aspectual haven’t sounds better to me than modal can’t in the examples above). The syntactic complexity or argument structure of the anaphoric island and ellipted VP may also affect acceptability (i.e., the relative simplicity of decide vs. the greater complexity of speak English).

Thanks to Andrew Garrett for bringing the discrepancy in judgments discussed here to my attention.

References:
Hankamer, Jorge, and Ivan Sag. 1976. Deep and surface anaphora. Linguistic Inquiry 7:391–428.
Postal, Paul. 1969. Anaphoric islands. Chicago Linguistic Society 5:205–239.
Ward, Gregory, Richard Sproat, and Gail McKoon. 1991. A pragmatic analysis of so-called anaphoric islands. Language 67:439–474.