June 10, 2005

WTF coordinate questions

wtf Over on the newsgroup sci.lang, Ron Hardin has posted (3 June 2005) another one of his finds, the radio ad:

(1)  Are you like most Americans, and don't always eat as you should?

Hardin thought of this as just an error, but I find it surprisingly good, despite the failure of parallelism between the two conjuncts.  Even more interesting, the obvious way of "fixing" the failure of parallelism makes things worse, rather than better.  Example (1) looks like an instance of a special construction, the details of which I don't yet understand.


Hardin is exquisitely sensitive to failures of parallelism and almost always attributes the examples he collects to failures of attention, inappropriate revising for conciseness, or, more generally, "deafness" to the conditions on coordination.  This literalism can be annoying, but there's no denying it leads him to notice a variety of interesting examples, especially of WTF coordination.

Some invented examples like (1):

(2a)  Do you like Spam, but can't stand Spam casseroles?
(2b)  Have you written a thesis, but have no idea what to do next?
(2c)  Was Kim a typical American, and couldn't speak a word of another language?

Hardin's first reaction was that the problem was just the coordination of an interrogative with a declarative.  But coordination of unlike sentence types is not in itself fatal; as is well known, English has a variety of constructions of this sort:

(3a)  Eat that sushi, and you're a dead man!
(3b)  Eat that sushi, or you're a dead man!
(3c)  Eat that sushi, or are you too much of a wimp to do it?
(3d)  You have to eat that sushi, or are you too much of a wimp to do it?

In the right circumstances, you can even coordinate an NP with a declarative clause, in the "one more beer" construction made famous by Peter Culicover (Linguistic Inquiry 1.366-9 (1970)):

(4)  One click of the mouse, and I'm on my way to untold, Trump-level riches, only without the bankruptcies.  ("Doonesbury" cartoon by Garry Trudeau, 9 June 2005)

As Greg Lee pointed out in sci.lang, the problem is that the missing subject in the second clause (which is not inverted) is not in the same position as the explicit subject in the first clause (which is inverted).  If they weren't questions, there would be no puzzle about these examples; the corresponding declaratives have unproblematic conjoined VPs, marked by square brackets below:

(1')  You [are like most Americans], and [don't always eat as you should].
(2a')  You [like Spam], but [can't stand Spam casseroles].
(2b')  You [have written a thesis], but [have no idea what to do next].
(2c')  Kim [was a typical American], and [couldn't speak a word of another language].

Now, examples (1')-(2c') have semantically equivalent versions in which full clauses are conjoined, as in:

(5)  [You are like most Americans], and [you don't always eat as you should].

But supplying an explicit subject for the second VP in (1)-(2c) produces odd results, as in:

(6a)  ??Are you like most Americans, and you don't always eat as you should?
(6b)  ??Do you like Spam, but you can't stand Spam casseroles?

What seems to be going on with (1)-(2c) is that they are simply the yes-no question versions of (1')-(2c').  Semantically, this is just right.  And pragmatically: coordinating the VPs conveys that the VPs taken together are to be understood as characterizing a single state.  But there's a syntactic problem: (1)-(2c) have the inversion associated with yes-no questions only in the first clause.  It's as if the syntax follows the semantics/pragmatics in treating [VP Conj VP] as a unitary constituent, with the first V as its head.  Now to work out the details.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at June 10, 2005 02:00 PM