Saturday, December 12, 2009

Could I get some binding indices, please?

This is actually a rather old news article, but I've only just now got around to blogging about it. The lede read "In an emotional interview, Whitney Houston opened up to talk-show host Oprah Winfrey about the pivotal role her mother played in getting the singer back on track." Wait, what? Whose mother? What singer? It's pretty unreasonable to expect that someone wouldn't be able to parse this sentence, but it still strikes me as very odd. First, there's the inevitable ambiguity in "her mother". This is resolved by pragmatics, since we assume that since Whitney Houston is the topic of the passage, "her mother" naturally refers to Houston's mother. However, syntactically this could just as easily be Winfrey's mother. (This is why I maintain we need a proximate/obviative distinction in English: "Whitney Houstonwa opened up to talk-show host Oprah Winfreyi about the pivotal role omotherwa played in getting oma singerwa back on track". When I mix Blackfoot and English I like to call it Blinglish.)

But the most serious "hey wait a minute" moment for me is the use of "the singer" toward the end of the sentence to refer to Whitney Houston. I would think that strictly speaking, this sentence should be ungrammatical by pretty much any version of the binding theory you adopt (classic GB, R&R, or Ken Safir's FTIP). This has to to with reference, which I think is captured nicely by Ken Safir's version of binding. Once we've established the context with an R-expression (referring expression, i.e., any noun phrase that's not a pronoun or an anaphor), we need to use the most dependent form for each successive instance of a coreferent NP. "Her mother" is fine, but it would be weird to say "Whitney Houston's mother"; that's why we have pronouncs. And "the singer" just seems really odd to me. I would expect "in getting her back on track". I assume they reverted back to an R-expression for Houston because it had been so long since the initial mention, but it's still very marked for me, perhaps even ungrammatical.

This is because you can't have an R-expression coindexed with a previous R-expression. If we say "Whitney Houston picked up the singer's clothes at the dry cleaner's", we want to ask "Wait, whose clothes did she pick up?" This violates whatever theory of binding you subscribe to, unequivocally. Luckily, actual language use is much more fluid, and clearly that was an acceptable sentence to someone, again, probably because of the distance between the two expressions, but I don't have to like it.

7 comments:

NW said...

At what point do we draw the line between a rule that has been violated, and a rule that is invalid? For R-Expression binding, there are many examples of them being bound, and referring to antecedents. Should that binding principle be modified or left as is? It makes me very uncomfortable and annoyed when people say, "well, I know this SHOULD be ungrammatical because theory tells me so." Theory concerning coreference is only true insofar as it describes what is actually occurring, not an idealized state of affairs.

We have the technology to look at these relationships over large corpora, but going sentence by sentence seems very prone to error and over-generalization.

[rant over]

Ryan Denzer-King said...

An intriguing question. I am mostly of your opinion: that it is the data which should come first, theory second. But I think we have to remember that any theory begins as a simple generalization that has exceptions. As time goes on we refine the theory to cover more and more instances. I also think part of the problem is the general idea that a good theory will explain all of the data all of the time. I reject the model of the human brain as a computer that spits out the same result every time, given the same input.

Could you provide some examples of bound R-expressions? I usually find them odder than many people do. Apparently "His manager told John to pick up lunch" is totally fine for most people, but it's definitely weird for me. I think Ken Safir's version of binding gets us closest to the truth. If nothing else we can fudge what "most dependent form allowable" means. If we include pragmatic factors in that, I'm sure we could account for all the data.

NW said...

Alright. Personally, I seem to accept many sentences that others reject. Like your "His manager told John to pick up lunch" is fine for me. Oh, here's a funny one: "you can call me John" so bound to an ACC DP pronoun. "I told him, John, you must...", "...and him? John left." "I said to myself, John, you must..." but maybe that would count as a parenthetical. "I called him 'John'". ''Call Him John the Careless'' from Washington Post.

I can find a fair number, but maybe they are examples of a boundary condition over which binding can occur. I will have to consider these a bit, because the classic example of R-Expressions being free is something like "Sue_a saw Sue_a in the mirror" (although that's not too bad), or "Sally_a kicked Sally's_a dog", where there's a simple repeated name, which has also been found to be bad from a processing standpoint.

NW said...

I forgot to mention that I also agree with you about the brain. It is definitely much more dynamic than we give the brain credit for being in much of theoretical linguistics.




[I slipped another example in there]

NW said...

It seems there's an initiative out of UCLA looking into exceptions to Binding Principle C:

http://sites.google.com/site/bindingucla/exceptionstoprinciplec

The plot thickens.

meagan louie said...

Sorry, I can't resist a Blackfoot-Geek-Out:

You'd think that Blackfoot would be great for distinguishing among a bunch of different third persons, but I think the Blackfoot would also be ambiguous, since obviation is not marked on possessive pro-clitics.

So it would be o-mother-yi as opposed to o-mother-wa, since the possessed nominal "her mother" as a whole would be further obviative (Whitney = prox, Oprah = obv, Whitney's mom = further obviative), but the possessive o- isn't marked for obviation, so should be able to refer to either Whitney or Oprah...

And the theme markers for furtherobviative acting on obviative, I think, are the same for furtherobviative acting on proximate...

As for Binding:

I NEVER get any of those stupid binding judgements that I am supposedly supposed to get, as a native-speaker of English... "John wonders who likes which pictures of himself" is just fine for me...

Ryan Denzer-King said...

Quite right; I hadn't thought about the fact that there's no difference between the proximate and obviative pro-clitics.