Charades does not reveal a universal sentence structure

« previous post | next post »

Here's an article in yesterday's New Scientist: "Charades reveals a universal sentence structure." The ever eagle-eyed Ben Zimmer thrust it under our noses as we hung around the LL water-cooler this morning. My interest was piqued. It would be much easier to learn about language by playing charades than by using the extraordinarily laborious standard method, i.e. studying language.

The article reports on a new paper in the prestigious journal PNAS, "The natural order of events: How speakers of different languages represent events nonverbally", by Susan Goldin-Meadow, Wing Chee So, Aslı Özyürek, and Carolyn Mylander. I've taken a look at the PNAS paper, but for now I just want to give you my immediate reactions to the New Scientist article. I'll follow up with some comments on the PNAS paper later.

In the study the New Scientist article reports on, subjects with various native languages look at pictures involving characters doing things to other characters, and then they mime what they saw. And the main finding the New Scientist article reports is that:

"Regardless of the order used in their native spoken language, most of the volunteers communicated with a subject-object-verb construction."

Cute! But now it gets a little weird:

"Goldin-Meadow argues that this kind of sentence syntax might therefore be etched into our brains. Languages that veer away from this form, such as English, must have been influenced by cultural forces."

I'm thinking about this as I take the executive elevator from the water cooler to my basement office in LL plaza. Actually, it's the first time I've been in the executive elevator, since I don't have a key of my own. Its a very nice plush elevator, what with all that mahogany, and the velvet covered benches. I could get used to it. But as usual, I digress. Here's what I'm thinking:

  • Uncontroversial observation number 1) Plays tend to start with a list of characters. And when we tell a story, we tend to introduce characters before connecting them. And when we describe pictures verbally or in writing, we tend to describe the objects and characters and then how they relate. ("There's a bear and there's a rabbit and the bear is hugging the rabbit")
  • Uncontroversial observation number 2) We tend to give priority to characters that act rather than characters that are acted on. And in casual conversation, stories, and descriptions, we tend to mention more important things before less important things.

So now it turns out that people tend to mime Subjects, then Objects, then Verbs.  Of course it is not appropriate to describe the mimes as Subjects, Objects, or Verbs, since they're not part of a language, but let's ignore that. The SOV order observed could perhaps be explained without any reference to syntax: all the characters before the action or relation because of (1); and agents before patients because of (2).

"[T]his kind of sentence syntax might therefore be etched into our brains." says the article. Yeah, or maybe people mime in a similar way to the way they tend to present information in discourses.  Maybe the mime study highlights one edge of a broad and interesting discourse level phenomenon, not a syntactic phenomenon at all.

In fact, it's not clear that the phenomenon is specific even to communication: the data reported would be compatible with perceptual or cognitive biases that had nothing to do with communication at all.  For example (and though I'm certainly not advocating this explanation since I have no evidence for it) subjects could have a perceptual bias that allowed them to identify actors first, then things acted on, and then actions, combined with a general disposition to report things in the order in which they were identified.

From the combination of talk of "universal sentence structure" and "brain etching", it seems that the reporter takes Goldin-Meadow to be arguing for an innate language module. Well, the fun charade ordering phenomenon is in need of an explanation, but I don't see any evidence whatsoever for brain etching. Mind you,  I do sometimes wonder whether INNATE LANGUAGE MODULE is etched onto the brains of some linguists. 

Linguists hypothesize that there is a general bias towards SOV word order in the early evolution of languages. And this is one of the most impressive pieces of evidence that is cited in favor of a strongly innatist position. However, the presence of a bias for certain orderings in discourses about pairs of individuals, as seen in the new study, might provide the beginnings of an explanation for whatever tendency there is for languages to crystalize with SOV structures. And that would be an explanation that does not involve reference to a language module at all.

There's perhaps a tiny nugget of some deep sociological truth here. It seems that faced with a neat new observation, some people, at the very least a New Scientist reporter, feel that we have new evidence in favor of the innateness of language. But I'm a cynic who's never been terribly impressed by strongly innatist arguments. So perhaps it's no surprise that when I hear about the new data, I come to exactly the opposite conclusion: the new data shows that yet another argument for strong innatism is hollow.

Here I am, back down in the basement, outside my dark little office. And you seem to have followed me down here. Just you, me, and an already rather crumpled and scrofulous-looking page torn hastily from Zimmer's brand new New Scientist while he looked the other way. And this, an impressive bunch of keys that Quince absent-mindedly left by the water-cooler. Do you suppose that one of these keys might fit the lock to that heavy oak door over there, the portal from which I so often see one or other of the senior writers emerging, dusty old bottle in hand?

(Comments off until I've sobered up and reported on the PNAS article.)

UPDATE: follow-up post is here.


Comments are closed.