March 29, 2008

More WTF coordinate questions

Today's find in the world of WTF coordinate questions is
(1) Anyone Here Been Raped and Speaks English?
This is the title of a book by war reporter Edward Behr that was mentioned on NPR yesterday morning.  It's a quote from a British television reporter who Behr observed looking for interviewees in a Congo airport in the 1960s.  Appallingly callous, but my topic today isn't the morals of journalists, but the twists of coordination in English.

We've been here before, though with a slightly less complicated example:
(2) Are you like most Americans, and don't always eat as you should?
The slight complication in (1) is that it's missing an initial auxiliary verb (has) -- but casual variants of English yes-no questions lacking an initial auxiliary are common and have been much studied.  In fact, what the television reporter said is quoted in a number of places, including Brewer's Famous Quotations (Nigel Rees, 2006), as having been
(3) Has anyone here been raped and speaks English?

Examples like (2) and (3) aren't nearly as bad as the failures of superficial parallelism in them -- a clause with subject-auxiliary inversion conjoined with a finite VP -- might lead you to expect, and they present serious problems for a reduction analysis of coordination, in which shared material in parallel positions is "factored out".

[A couple more examples from real life: from Paul Kay on 9/13/06, a television commercial for the over-the-counter sleeping medicine Lunesta, noted 9/11/06: Do you wake up in the night and can't go back to sleep?  And from Elizabeth Daingerfield Zwicky on 2/14/08, a notice at Kepler's book store in Menlo Park CA: Do you like to knit but are looking for a meaningful project?]

Here's the problem: leaving out many important details, examples like (2) and (3) appear to have something like the structure
Aux Su [ Complement-of-Aux Conj VP ]
but Complement-of-Aux here (like most Americans in (2), been raped in (3)) is non-finite (not even necessarily a verb phrase), while VP here (don't always eat as you should in (2), speaks English in (3)) is finite, and these are not parallel.  Worse, though Aux + Su here is interpreted as combining with Complement-of-Aux (are you like most Americans in (2), has anyone here been raped in (3)), only Su (most Americans in (2), anyone here in (3)), and not Aux + Su, is interpreted as combining with VP, since Aux is incompatible with the initial auxiliary (*are don't always eat as you should in (2), *has speaks English in (3)).  Such examples resist ordinary "reduced conjunction" analysis.

Instead, as I observed back in 2005, things like (2) and (3) seem to be simply the yes-no question counterparts to declaratives like
(2') You are like most Americans, and don't eat as you should.
(3') Someone here has been raped and speaks English.
(which have ordinary conjoined VPs).  As I said then,

Semantically, this is just right.  And pragmatically: coordinating the VPs conveys that [VP Conj VP] taken together are to be understood as characterizing a single state.  But there's a syntactic problem: [examples like (2) and (3)] 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.

In a later posting I returned briefly to these WTF coordinate questions and alluded briefly to three alternative analyses for them, all of which treat them as involving a coordination of some constituent with an ELLIPTICAL subpart, rather than as being "reductions" of coordinations of full clauses.  (I owe these ideas to Language Log and ADS-L readers who wrote me about my first posting and to colleagues at Stanford who commented on a presentation I gave in August 2005.)  All would treat simpler coordinations like saw Kim and Sandy as [saw Kim] andSandy] (or, better, [ saw Kim] [andSandy]], but I'll put aside the question of where conjunctions fit into these structures), but via different formal mechanisms:

Idea 1: the Ø is an instance of zero anaphora, as in Verb Phrase Ellipsis (They told me to go, but I didn't want to Ø).

Idea 2: the Ø is an instance of "functional control", as in Kim wants Ø to leave.

Idea 3: the Ø is part of an Initially Reduced Question, as in (1) vs. (3) (cf: Ø Sandy gone yet?).

There are knotty technical questions here.  I bring these three ideas up only to demonstrate that there are alternatives to the reduction idea.  A little more on this below.  But first, a note that standard assumptions about constituency might also be called into question.  The usual assumption about (3') is that it has a coordinate VP:
[ Someone here ] [ [ has been raped ] and [ speaks English ] ]
But once the possibility of ellipsis comes into play, we can entertain the possibility that (3') has instead a structure that fits better with (3), something like:
[ [ Someone here ] [ has been raped ] ] and [ Ø speaks English ] ]
(something along these lines has been proposed by John Beavers and Ivan Sag in a 2004 paper, "Coordinate ellipsis and apparent non-constituent coordination", Proceedings of the HPSG04 Conference).

A final remark on "deletion", "omission", "reduction" and similar turns of phrase.  These terms, which strongly suggest an analysis in which one construction is secondary and is in some way derived from another, primary, construction via various formal operations, can be deeply misleading.  (The terms are useful, maybe necessary, but in the end they're just terms, not claims to truth or any kind of analysis.)  Even when the historical sequence seems clear, my current position is that once the variants are out there, they are just variants; they will share elements of their structures, true, but they'll have their own details and uses and lexical idiosyncrasies and the like.  Each should be described on its own.  We should take seriously the idea that "reduced coordinations" are not just full coordinations with some pieces left out, but might be constructions in their own right.

Posted by Arnold Zwicky at March 29, 2008 11:57 PM