Sunday, January 10, 2010

More Russian Illusions Than I

Colin Phillips gave a nice plenary talk at the LSA this afternoon on the role grammatical illusions can play in studying the online processing of sentences (was it just me, or did his English accent seem more pronounced than usual? Was this a social register effect or am I off my rocker?).  He drew a really nice parallel with optical illusions and the value they have added to the study of vision.  The point is that there are some sentences that seem perfectly grammatical at first, but upon reflection, are completely incoherent. For example:
  • More people have been to Russia than I have.
Most native speakers of English will read this sentence and be perfectly happy, but re-read it a few times. Do you see the incoherence? It's incoherent because ...
it's comparing apples to oranges. In the more people have Xed than Yed construction, both X and Y should be events that "people" have participated in (e.g, more people have watched Avatar than read Moby Dick). Be careful not to force an interpretation. Yes, I (and Colin) understand that you can find an interpretation of this sentence that kinda makes sense, but that's not grammar. Take the following sentence:
  • the Wallace ball Gromit threw.
Now, most of us can kinda make some sense out of this if we try, sure. But that's not the point. The point is that this is clearly an ungrammatical sentence in the English language. The same is true of the Russia sentence above (well, its ungrammaticality is less clear, but its ungrammatical nonetheless). Colin's point is that sentences like the Russian sentence can give us valuable insight into the online process of parsing sentences. His other point seemed to be that we have at least two mechanisms for processing a sentence. I'll have to dig into this one deeper to explain it, but he has a paper in press detailing these findings and a pre-print is available right now HERE:
  • Grammatical illusions and selective fallibility in real-time language comprehension. Colin Phillips, Matt Wagers, & Ellen Lau. 26pp. June 2009. To appear in Language and Linguistics Compass. pdf.
From that paper:
...speakers build richly structured representations as they process a sentence, but that they have different ways of navigating these representations to form linguistic dependencies. The representations can be navigated using either structural information or using structure‐insensitive retrieval cues. In order to explain why structural constraints dominate in some situations but are at least temporarily overridden in others, one does not need to assume architectural priority for structural information. Rather, structural constraints may impact linguistic dependency formation most strongly in situations where relevant structural information is available in advance of potentially interfering material in the bottom‐up input.


Bridget said...
This comment has been removed by the author.
Forrest said...

I have to admit to a bit of confusion. It's plainly clear why "More people have been to Russia than I have" is incoherent, and this jumped out at me on my first reading. ( Although in fairness, your title and intro about grammatical illusions may have primed me a bit to be skeptical. )

I'm not as convinced that it's ungrammatical, and I wonder if you could spend a moment explaining that? Your blog post details why the sentence doesn't make sense, but making sense and obeying the rules of English grammar aren't the same things.

"Colorless green ideas sleep furiously."

From what I've read, that sentence was written to show why people can't be Markov chaining devices. Nonetheless, I think it provides a good counter-example here. It's perfectly grammatical, and I think most people construct a single, unambiguous parse tree ( NP, VP ). But of course it's meaningless gibberish.

Putting the Linguistics into Kaggle Competitions

In the spirit of Dr. Emily Bender’s NAACL blog post Putting the Linguistics in Computational Linguistics , I want to apply some of her thou...