WASHINGTON, D.C.—If you take the lessons of your middle school English teacher on faith, you probably think there’s a right way and a wrong way to construct a sentence. But linguists have long acknowledged that some grammar falls into a “gray zone,” a middle ground in which sentences are neither 100% right nor 100% wrong. Now, a new study shows that the linguists who map out the structure of grammar—syntacticians—rarely use this gray zone in their own studies. It also suggests a wide gap between their black-and white views and those of ordinary people.
“Grammar is not this binary thing,” says Jana Häussler, a linguist at the University of Wuppertal in Germany and one of the study’s authors. She adds that many of her colleagues still judge grammar using old binary models, when they should be coming up with systems that build in gradience—or the gray zone—as a possibility.
To find out just how many syntacticians use gradience, the researchers looked at 89 grammar papers published in the highly cited journal Linguistic Inquiry. Most of the authors seemed to give the gray zone its due: They used more than three categories to judge sentence grammar: “completely acceptable,” “completely unacceptable,” and at least one “intermediate” category. Problem was, they almost never used the intermediate categories for describing sample sentences. Instead, 94% of their sample sentences (2619 in all) made it into the “completely acceptable” or “completely unacceptable” categories, the researchers reported here this month at the Linguistic Society of America meeting.
Researchers also wanted to see how these expert conclusions tallied with the intuitions of ordinary people. In theory, syntacticians base their models on what sounds natural and right to native speakers. The judgments of regular people define the rules; syntacticians are supposed to describe and explain them. So the team took 100 “black-and-white” sentences from the studies and ran them by 65 native English speakers, recruited from the online labor-sourcing platform Amazon Mechanical Turk. Their answers didn’t square with those of the linguists. On a scale of 1 to 7, participants ranked 40% of the black-and-white sentences between 3 and 5, putting them squarely in the “gray zone.”
The results could affect everything from research into how the human brain processes language to building speech recognition software. By ignoring the gray zone, say the researchers, syntacticians are failing to describe how language really works.
The project also fits into a larger effort to integrate different fields of linguistics, says Sali Tagliamonte, a sociolinguist at the University of Toronto in Canada, who was not involved in the project. Sociolinguists study how people use language in the real world, so they are more apt to see grammar as “gradient,” existing on a sliding scale.
“This poster is an example of a new way of thinking,” Tagliamonte says. “They’re saying, ‘Hey you guys doing theory, we’re going to look at your own sentences, and we’re going to show you that it’s not as clear-cut as you think.’”
But some say the ratings don’t necessarily reflect an underlying gradience in grammar. “If you give people the opportunity to give gradient ratings, they’ll give gradient ratings,” says Colin Phillips, a neurolinguist at the University of Maryland, College Park, who has published in Linguistic Inquiry. People will even give gradient ratings for whether a number is even or odd.
Phillips acknowledges that grammar probably has gray areas, but doubts whether the new study measured them specifically. Using sentence ratings to size up grammar’s gray zone is, in his view, like trying to use Yelp ratings to judge whether the chili tastes good at a restaurant. People base their ratings on too many other factors.
“You’ve got a combination of a linguistic system, a memory system, and people’s judgments of what makes sense in the world, and then you mush those all together and you get something gradient,” he says. “An enduring question is: Why? That’s hard to answer.”