Sunday, April 01, 2007

Rarely missing a chance to overanalyze (Crosspost from LL)

A few days ago, I wrote the following sentence:

1. I rarely miss a Daily Show, though sometimes not for several days after the fact.

As I read it over at the time, I had a sensation of overnegation bizarrity, but after thinking about it for awhile, I decided that it did say what I wanted it to say, although in a fairly obscure way, which I thought might be amusing, so I left it. Almost immediately after I posted it, though, alert reader Sridhar Ramesh wrote in with a description of having gone through the same boggle, pause, click of comprehension, pause, but it's still weird! that I did, and I though I'd go back and think about it some more.

I don't think there's an overnegation in the sentence—it has just the right amount of negation—but there's a weird ellipsis-reconstruction problem: the understood verb phrase inside the though-clause has to be understood as containing something that does not correspond to a syntactic constituent in the antecedent mail clause.

Here's the way the problem goes, I think.

"I rarely miss a Daily Show" is interpretively equivalent to "I usually watch the Daily Show." Informally, one could say that "miss" = "not watch"1, and "rarely" = "not usually", and the two negations cancel each other out, interpretively, like this:

2. a. I [rarely [miss a Daily Show]] -->

2. b. I do[n't [usually [not [watch the Daily Show]]]] -->

2. c. I [usually [watch the Daily Show]].

Now, consider what happens if I replace "rarely miss" with "usually watch" in my weird sentence above:

3. I usually watch a Daily Show, though sometimes not for several days after the fact.

This has a totally unobjectionable interpretation, derived by filling in an elided "I watch it" in the though clause, modified by the not in the though clause.2 This is shown in (4) below. I have colored the filled-in material red and put it in brackets so you don't lose track of it. The antecedent of the filled-in material is underlined.3

4. I usually watch a Daily Showi, though sometimes (I) don't (watch iti) for several days after the fact.

That is, as Sridhar notes, my not watching it can extend for several days before ending with my watching it. Note that it's not my usually not watching it that extends for several days -- the usually is not part of the filled-in material, just the VP watch the Daily Show.

In order to get this reading out of my actual sentence, one has to dissect "miss" into "not watch". If you don't do this, and reconstruct the elision as involving the actually present VP miss a Daily Show, you get the overnegation problem:

5. I rarely miss a Daily Showi, though sometimes (I) don't(miss iti) for several days after the fact.

The though-clause here is saying there are several days of my not missing a given Daily Show followed by my missing it. That is, it sounds like I watch it repeatedly for several days and then begin to miss it.

The 'good' reading for the sentence that I and Sridhar were getting is based on decomposing 'miss a Daily Show' into 'not watch a Daily Show', as in (2b) above:

6. a. I rarely [miss a Daily Show]

6. b. I rarely do[n't [watch a Daily Show]]

After filling in the elided bit, this would give the right interpretation, namely, one in which [watch it] is the reconstructed constituent:4

7. I rarely don't watch a Daily Showi, but sometimes (I) don't (watch iti) for several days after the fact.

The key thing about (7), of course, is that it involves reconstructing a subpart of the decomposed "miss a Daily Show". This is what's making the sentence in (1) feel so bizarre. That subpart is not present in the syntax of the antecedent clause; if it's present at all, it's only in the sentence's 'logical form'.

One approach to ellipsis phenomena says that the whole understood sentence, repeated material and all, is actually present in the speaker's mental syntax, but because some bit of it is identical to some other bit of it, the second instance of the identical bit is simpy not pronounced. If this is how ellipsis works, we can understand the bizarre feeling: Ellipsis involves non-pronunciation of material that is syntactically identical with its antecedent, so the deleted phrase would have to be the whole verb phrase miss a Daily Show, not some deconstructed semantic subpart of it which isn't present in the real syntax. (See, e.g., Jason Merchant's work.)

Another approach to ellipsis says that the speaker doesn't bother constructing the syntax for the elided bit; rather, it's just not present at all, and the hearer reconstructs it from the semantic interpretation of the antecedent clause. (see, e.g., Ray Jackendoff and Peter Culicover's work). Under that approach, the bizarre feeling might show that miss is not decomposable, semantically; although the lexical meaning crucially involves a negative implication of some kind, the negative part of the meaning can't be prised apart from the rest of the meaning.

Either way, the bizarre feeling Sridhar and I get from (1) shows that in some important way, watch the Daily Show is not a subpart of miss the Daily Show.

1Actually, the imagined negated verb probably shouldn't be 'watch' here -- it'd be safer to posit a more general negated verb like 'not experience' with this reading of 'miss'. That would cover things like 'I missed the dance', where you don't mean 'I didn't watch the dance', but probably 'I didn't attend the dance'; 'I didn't experience the dance' would cover it), or 'I missed what you just said', where you don't mean 'I didn't watch what you just said', but rather, 'I didn't hear what you just said' -- 'I didn't experence what you just said' would cover that too. Nonetheless I'm going to go with watch rather than experience just coz the problem is already complicated enough; it doesn't change anything about the reasoning undertaken below, though.

2I represent the filled-in-bit as 'I watch it' rather than 'I watch a Daily Show' because using the latter to elucidate the filled-in-bit wouldn't give the necessary bound-variable reading of the object of 'watch' in the elided bit; 'it', however, does. Just consider it a slight fictionalization that allows me to gloss over a discussion of why an overt repeated 'a Daily Show' wouldn't give right bound variable reading and the mechanics of how some syntacticians think an invisible, elided 'a Daily Show' can give such a reading.

3 For another LL discussion of ellipsis phenomena see this post of David Beaver's.

4Note that the n't in the though-clause is not the same as the n't in the main clause. It's crucial to leave that main clause negation behind, otherwise you end up with the bad interpretation in (5), where it's my not watching it, i.e. my missing it, that starts several days after the initial air date.)