Psycho Babble

Please Excuse My Dear Aunt Sentence

By Jessica Love | February 23, 2012


In language, as in politics, the importance of constituency cannot be overstated.

Consider the following sentence: Put the book on the shelf on the table. The sentence is temporarily ambiguous: until a listener encounters the second on, he or she is likely to think that the phrase on the shelf is where we wish the book to go, not where it is. That the book on the shelf should be interpreted as a single constituent is not immediately clear. So, being skilled speakers with a keen desire to be understood, we might pause longer before the second on than the first (e.g., say something along the lines of Put the book on the shelf … on the table). Or, we might pronounce the words preceding that second on more quickly than the words after it. Either of these strategies would give our listeners a fighting chance of interpreting on the shelf correctly the first time around. (Seeing a book already resting on a shelf probably helps too.)

Last week, I discussed sentences like I berated the student at the podium. Here, the phrase at the podium can either reference the student or the act of berating. A speaker’s delivery might help clarify whether the student at the podium should be understood as a single constituent. But what about written language, particularly in situations in which traditional punctuation is of no help? Might it not be useful to have a new sort of punctuation whose explicit purpose is to clarify constituency?

A precedent, interestingly enough, exists. Many mathematical equations would also be ambiguous without a set “order of operations” to clarify whether five plus 15 divided by five equals four or eight, e.g., (5+15)/5 versus 5+15/5. Could we do something similar for ambiguous sentences, e.g., I {berated {the student at the podium}} versus I {{berated {the student} at the podium}?

Some recent research suggests that we already do intuit parallels between mathematical and verbal constituencies. Psychologist Christoph Sheepers, along with colleagues at University of Glasgow and University of Edinburgh, asked participants to complete sentence fragments like The tourist guide mentioned the bells of the church that _____. The sentences were written so that they could be completed in two ways: participants could either interpret the clause beginning with that as modifying bells, in which case they would write something like were really loud or had cracks in them, or they could interpret the clause as modifying “church,” in which case they would write something like was really old or had cracks in it. (Because one interpretation involved modifying a plural noun like bells and the other involved modifying a singular noun like church, researchers could use subject-verb agreement to tell which noun was being modified.)

Before completing the sentence, however, participants had to solve one of three types of mathematical equations. For the first type of equation, one like 9-5+15/5, the order of operations dictated that the final two numbers be treated as a single constituent. For the second type, equations like 9-(5+15)/5, two earlier numbers formed this constituent. For the third type, like 5+15, constituency didn’t matter, as there was no ambiguity.

The researchers reasoned that seeing an equation in which the final numbers formed a constituency would encourage participants to modify the final noun phrase (e.g., church in the example above), whereas seeing an equation in which two earlier numbers formed a constituency would encourage participants to modify the earlier noun phrase (e.g., bells). Seeing an unambiguous equation shouldn’t affect participants one way or another. And this is exactly what the researchers found.

The results have no practical benefit. We will not be prefacing our ambiguous sentences with mathematical equations, nor will we pepper them with order-of-operations-inspired symbols anytime soon. But the fact that constituency preferences in one domain seem to “prime” constituency preferences in another domain is nonetheless important. We already know that language is special; no other aspect of cognition approaches its power, complexity, and learnability. But what’s left is defining the parameters of this specialness—to discover what it does and does not share with music, numeracy, logical reasoning, and everything else that makes us human.

Permission required for reprinting, reproducing, or other uses.

Comments powered by Disqus