Language Log

More fun with VPE

From an Andy Gill article on D.A. Pennebaker's rockumentary Don't Look Back (on Bob Dylan), in The Independent of 4/27/07 (the Joan in question is Joan Baez):


Yes, one of those see-through blouses you don't even want to, with a Verb Phrase Ellipsis (VPE) on the edge.  It takes some work to interpret -- you're likely to think of it as some kind of play on words -- but you can do it, and figuring it out contributes to the humor (a good thing, since the content of the quip is an unpleasant put-down).  The other ingredient in this example is the "anaphoric island" phenomenon, which we seem not to have discussed before on Language Log.


(Hat Tip to Empty Pockets.)

I've talked about playful uses of VPE twice in the last year: John McWhorter's


and Eric Bakovic's


The first of these postings has an introduction to VPE, which I won't repeat here; what's important here is that in this construction a VP is omitted when it can be supplied from the immediate linguistic context.  (There's a huge literature on the details of VPE.)  In the Neuwirth example, there's a relative clause in which the complement of infinitival to is omitted:


What is this complement?  Something like the VP


which itself has a gap in it -- the gap of the relativized NP.

Now, VPE is possible when the ellipted material has a gap of relativization in it, as in this example (from a posting to the OutIL mailing list on 5/17/04):


The omitted complement VP here is


with a gap in it, just as in the Neuwirth sentence. 

A bit of a digression now: what fills the gap in pay attention to ___? If you take a relative-clause gap to be filled by its head, then this relative clause is to be interpreted as (that) they don't want to pay attention to any state marriage laws.  But that's not right.  The object of pay attention to should be understood as something like any state marriage laws they don't want to pay attention to ___.  Whoops.  There's a gap in THAT, and we're in an infinite regress. 

What we have here is a instance of "antecedent-contained deletion", usually illustrated by somewhat simpler examples like John read every book that Mary did ___.  The phenomenon has been studied since 1970; there's even a Wikipedia page.   Finding a satisfactory analysis involves abandoning simple antecedent-substitution analyses for the gaps in VPE and relativization, and/or re-considering the analysis of quantification in NPs (as in any state marriage laws and every book).

These are gripping issues for syntacticians and semanticists.  What's important for us here, though, is the question of whether gap-containing VPE is somewhat harder to process than gapless VPE, like the one in


where the omitted complement VP is the gapless pay attention to all state marriage laws.  My impression is that the relative-clause gap contributes some processing difficulty, but I don't know if there's research bearing on the question.
 
But in any case, the DOMA sentence doesn't present the kind of processing puzzle that the Neuwirth sentence does.  What's the difference?  The nature of the ANTECEDENT for the omitted VP. 

The DOMA sentence has a VP antecedent, pay attention to ___, in the linguistic context, just as VPE requires.  But the Neuwirth sentence has instead the adjectival see-through, modifying the noun blouse.  Admittedly, see-through is derived morphologically from the verb see (plus an accompanying preposition, through), but it is not a VP, or even a V.  There is no VP see through ___ in the linguistic context to satisfy the requirements of VPE.  To understand the sentence, you have to get the interpretation 'see through' by "going inside" the lexical item see-through.

Another topic from roughly forty years ago, when it was first suggested that lexical items are "islands" for anaphora, that parts of lexical items or referents merely evoked by lexical items cannot serve as antecedents for anaphoric elements (of several different kinds).  Here are typical violations of the Anaphoric Island Constraint (AIC):


There's a huge literature on the AIC.  Early on, it was observed that more morphologically transparent lexical items are less problematic than more opaque ones.  Compare the examples above with:


And many examples improve considerably in context.  That is, various factors contribute to easing the task of finding antecedents within islands.  Eventually, some linguists began to argue that the AIC was not a syntactic phenomenon at all, but a pragmatic one, having to do ease of antecedent retrieval (as related to contextual cues and morphological transparency, in particular); for a summary, see Gregory Ward's 1997 "The battle over anaphoric 'islands': syntax vs. pragmatics" (in Directions in Functional Linguistics, ed. by Akio Kamio).  Some examples present no problem, others are extremely hard to interpret, even in context, and many lie in between.

It seems to me that the anaphoric-island violation in the Neuwirth sentence is in this middle territory.  It takes some work to figure out the speaker's intention, but the puzzle isn't insoluble.  The gap within the omitted VP might contribute to the listener's work, though most of the work is, I think, a consequence of the fact that the antecedent for VPE is not a VP in the linguistic context, but is instead evoked by the lexical item see-through.

So you can play with VPE for humorous effect.  You can also stretch VPE without playful intent, as some of the awkward examples in my previous postings on VPE illustrate.  My current favorite of the latter type is from the final episode of the TV series Charmed, in the following (presumably scripted) exchange:


The omitted VP in B's response would seem to be get Christy back, which is not actually in A's statement, but is conveyed by it, if it's understood as meaning 'I just want to get Christy back'.  If I hadn't been a VPE junky and had just been listening like a normal person, I might not even have noticed the off-flavor of B's response.

Posted by Arnold Zwicky at October 20, 2007 03:58 PM