October 08, 2005

Denotation switching


Mark Liberman puzzles over this passage from a story on the Ig Nobels:

The winners are discretely contacted beforehand to give them an opportunity to decline. It is a testament to the growing prestige of the event that very few turn down the offer and agree to attend at their own expense.

One way of characterizing what the problem is here is that turning down the offer is predicated of some small number of winners, denoted by "very few" -- call this small class of downturners T -- but the writer's intention is to convey that agreeing to attend at their own expense is predicated of the COMPLEMENT OF T, not T: it's the people who don't turn down the offer who agree to attend at their own expense.  You just can't do this by conjoining the VPs "turn down the offer" and "agree to attend at their own expense" with the shared subject NP "very few", because "very few" must be understood as simultaneously denoting T and ~T -- a straightforward violation of what I called, in an earlier posting on failures of parallelism in reduced coordination, the Factor Constancy condition: in factorable coordination, the factor must have the same semantics in combination with each of the conjuncts.

That's not quite the end of things, however.  In similar examples involving not coordination but anaphora (zero or overt), it's much easier to get away with this sort of denotation switching.  Here's an instance I brought up on the American Dialect Society mailing list back in May, from the Palo Alto Daily News ("City OKs university land deal" by Jason Green) of 5/4/05, p. 74:

Although touted by university officials and city staff as a historic deal, not everyone was in favor of the agreement, including council member Yoriko Kishimoto.  Of the six council members eligible to vote on the Mayfield agreement, Kishimoto cast the sole "no" vote.


In the first sentence from the PADN, the inclusion of council member Yoriko Kishimoto is predicated of some group not overtly mentioned in the phrase "including council member Yoriko Kishimoto".  This is zero anaphora.  So we search for an appropriate discourse referent for the anaphor, if possible one recently mentioned in the discourse.  Well, the subject of the main clause of that first sentence is "not everyone", which predicates less-than-universality (within the universe of relevant people) of the class of those in favor of the agreement, and so indirectly introduces this class -- call it F -- as a discourse reference.  Could F be the discourse referent that the zero anaphor picks up?   No: the writer's intention is clearly to convey that Kishimoto is included in ~F (the class of people opposed to the agreement), not in F.  Still, the sentence isn't so bad (though it was troublesome enough for me that  I reflected on it).  ADS-L posters ranged from those who found it plainly ungrammatical to those who had little problem with it.

At this point, as so often happens on ADS-L, Larry Horn stepped in to tell us that there's actually some literature about this phenomenon:

If anyone is really interested in this from a theoretical/empirical direction, there have been a number of publications by Linda Moxey and Tony Sanford, including their book Communicating Quantities (1993), on what they call "Comp-set" as opposed to "Ref-set" reference, i.e. cases in which the reference is to the complement of the set specified ("Few of the students passed, because they [= the ones who didn't] hadn't taken the test seriously" vs. "Many of the students passed, because they [=/= the ones who didn't]...).

Horn's example of Comp-set reference involves an overt anaphor, "they", but  the Kishimoto sentence shows that the same thing is possible with zero anaphors, at least for some people.

Still another zero-anaphor example, again from PADN (Health column, "Fall of Atkins diet traced", 9/15/05, p. 45):

Although studies showed the diet helped followers lose weight -- and quickly -- the Atkins dropout rate was high.  Few managed to stay on the diet for an entire year, complaining that, eventually, even an unlimited amount of steak and eggs can become boring.  Others suffered unpleasant side effects...

Here, the complainers are (presumably) the many who didn't manage to stay on the diet, not the few who did.  The same is true of a version with an overt anaphor:

Few managed to stay on the diet for an entire year; they complained that, eventually, even an unlimited amount of steak and eggs can become boring.

To sum up: Comp-set reference is possible, for at least some speakers in some contexts, for both overt and zero anaphors, but not (apparently) in (reduced) coordination.  It's not customary to think of coordination as being in any way like anaphora, so this difference between the two domains is scarcely a surprise.  But you COULD think of reduced coordination as being akin to zero anaphora, with the factor constituent overt with one conjunct but zero with the other, so that "saw Kim and Sandy" would have a structure like:

[saw Kim]  [and ___ Sandy]

(where the underline marks the position of an omitted V, which is interpreted as picking up the reference of the preceding V "saw").  In case you're tempted by this proposal, the unavailability of Comp-set reference in reduced coordination should make you think twice.

But I'm not done yet.  Ok, maybe reduced coordination isn't really much like zero anaphora, but it might be like another class of omitted-constituent constructions, namely those involving "functional control" (rather than anaphoric control), as in "Kim wants to leave", with a structure like:

[Kim]  [wants [to ___ leave] ]

Just such a proposal was made to me, very tentatively, by Paul Postal in e-mail following my Language Log posting of 6/10/05, on "WTF coordinate questions" like:

Are you like most Americans, and don't always eat as you should?
Have you written a thesis, but have no idea what to do next?

There's a lot to work out about this idea, and I'm still playing with it, though I have to say that my thinking about these WTF coordinate questions has led me away from anything resembling functional control and towards still another class of omitted-constituent constructions, namely Initially Reduced Questions like:

Have no idea what to do next? 'Do you have no idea about what to do next?'
Anxious about your exams?  'Are you anxious about your exams?'

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at October 8, 2005 03:00 PM