Saturday, August 13, 2011

Stanford professors and grammatical illusions

I started reading The Origins of Political Order by Stanford professor Francis Fukuyama when I stumbled across this sentence:
  • But it is clear that the political job of finding the right regulatory mechanisms to tame capitalism's volatility have not yet been found.
I read it a second time. Then a third. It just didn't seem right. The verb "have" is wrong, right? It should be "has", I think. But it was difficult to see why until I broke the sentence down to its essential elements. The linguistic dependencies are obscured by multiple embedded PPs, in this case, making it difficult to get the agreement right.

Distilling a complex sentence down to its essential elements is a technique I taught to students for years because it helps us see what the sentence "means" in a basic sense, and all we need is our intuition. No linguistic theory or technical terms are required. One way to do begin is to simply replace noun phrases with variables like X and Y:
  • But it is clear that the X of finding the right Y to tame Z have not yet been found.
Already we see that "have" sounds more clearly wrong. Next we can delete the introductory phrase "But..."
  • The X of finding the right Y to tame Z have not yet been found.
Now we have distilled the sentence to its basic elements. To figure out why "have" sounds wrong, we can delete all prepositional phrases, as they are not essential to the basic meaning of the sentence:
  • The X have not yet been found.
Finally, we need only put the original X phrase back in:
  • The political job have not yet been found.
Using just our intuition, two things should be obvious:
  1. The subject "political job" is singular and requires the verb to be "has" (syntax).
  2. It is odd to speak about finding a political job (when it clearly does not mean job in the sense of getting paid to do something). Rather, this is referring to political will (semantics).
A little analysis of the original and we can guess that Fukuyama was probably trying to construct a linguistic dependency between "have" and "regulatory mechanisms." But this dependency is ungrammatical in the strict sense that "regulatory mechanisms" is not the subject of the sentence. Yet the sentence is not obviously ungrammatical (like "Bob feet have."). I taught 18 year old public school freshmen about this, but we can see that this kind of thing gets the best of even highly educated, accomplished Stanford professors. It is a function of how our brains construct and parse sentences.

This sentence presents what Colin Phillips has called a grammatical illusion. He explains them thusly:

Research on the online implementation of grammatical constraints reveals a strikingly uneven profile. The parser shows impressive accuracy in the application of some rather complex constraints, but makes many errors in the implementation of some relatively simple constraints.

His plenary talk at LSA 2010 provided a variety of examples (many involving "have", curiously enough). Here's a pre-print PDF of his work on the topic:

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.

NLPers: How would you characterize your linguistics background?

That was the poll question my hero Professor Emily Bender posed on Twitter March 30th. 573 tweets later, a truly epic thread had been cre...