Gaps inside adjunct phrases
« previous post | next post »
Linguists have often assumed that the principles of English syntax do not allow a dependency between the head noun and the "gap" in a relative clause to span the boundaries of an adjunct such as a conditional if phrase. They will invent pairs of this sort to illustrate the ungrammatical results:
- I'm working with a man that I think you would absolutely hate.
- *I'm working with a man that if you saw you would throw up.
In the first, the meaning of the relative clause is "I think you would absolutely hate him", and syntactically there is a gap where the object of hate (underlined) would have been. But in the second, the meaning of the relative clause is if you saw him you would throw up, and the underlined pronoun is inside the conditional adjunct if you saw [him]. Having the gap inside the adjunct is not permitted, they say.
And they mean that descriptively: the claim is not that you ought to avoid sentences like 2 above; the claim is that all speakers have a natural instinctive aversion to syntactic structures of this sort.
But is that true?
Michael Hamiel wrote a letter to the San Francisco Chronicle (published April 19, 2014, p. A10), commenting on a story about an executive who was fired from Yahoo but got a $58 million severance package. "Yahoo is to be commended for raising the bar," he wrote. And he continued:
These are the kind of exit packages that if all fired employees got, state and federal unemployment bureaus could be closed down saving the government millions.
The gap is in the conditional adjunct, which has the meaning "if all fired employees got it". The noun kind in the phrase kind of exit packages is modified by the relative clause that if all fired employees got __, state and federal unemployment bureaus could be closed down saving the government millions.
This is exactly the sort of structure that syntax specialists have regularly said the syntax of English (and perhaps all languages) disallows.
Now, the situation is more subtle than you might think. Nobody doubts that sometimes people who write letters to the newspaper get their letters printed despite the fact that they contain mistakes. (The previous letter in the Chronicle had a typo in it.) So this could be an inadvertent syntactic slip that the editors didn't catch or decided not to fix.
Or it could be crucial evidence that ordinary, unreflective language use is not governed by any inbuilt mental constraint forbidding dependencies between head nouns and gaps that span the boundaries of adjunct phrases.
There's no simple or immediate way of deciding this. Yet the validity of a general syntactic theory will hang on deciding indefinitely many such quandaries correctly. Linguistics ain't easy.
However, Jerry Friedman found some more examples with the same structure in COCA (the Corpus of Contemporary American). This is a quote from Trent Lott:
We're going to look for things that if he doesn't sign, he will catch hell for it.
Here is one from Geraldo:
Nothing that could hurt me, but things that if I had, it would help with my fight against him.
So now it's looking a bit more like the sentences are deep-down grammatically OK, rather than slips, isn't it?