Charades does not reveal a universal sentence structure II
« previous post | next post »
A couple of days ago I reported on an article in last week's New Scientist, "Charades reveals a universal sentence structure." The New Scientist article reports on some neat experiments in an article in PNAS involving how people represent events non-linguistically, e.g. when miming. The main result, as the New Scientist reporter saw it, is that people mime in the order Subject, Object, then Verb, regardless of the word order of their native language, and that this provides evidence that this word order is "etched into our brains".
The PNAS article is "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. Unfortunately you or your institution needs a subscription (or $10) to see it. Fortunately, I've read the PNAS article on your behalf. And here I say "fortunately" only in the sense that I might have saved you money, and not with the intention of discouraging anyone from reading the original article: it's a clearly written and thought provoking scientific paper presenting a couple of clever little studies which garnered some neat results. (Have I mentioned before that there's no good reason why every clearly written and thought provoking scientific paper presenting a couple of clever little studies which garnered some neat results is not free for everyone?)
So anyway, as I say, I looked at the PNAS article, and, well, I dunno. I'm glad New Scientist covered the story, and they got the main results factually correct, which is a good start, but it still looks to me like misleading reporting in New Scientist, though far from being the most egregious example we've seen here at Language Log.
I obviously wasn't privy to private exchanges between the reporter and the lead author, Goldin-Meadow, but nowhere in the PNAS article do the authors say anything that makes it plausible, as the New Scientist article asserts, that "Goldin-Meadow argues that this kind of sentence syntax might therefore be etched into our brains." In fact, to my mind, the PNAS article suggests the opposite. But before we see what the PNAS article suggests, let's go back in time.
Over the last few years, the sign language researcher Carol Padden and colleagues have been investigating the emergence of a new sign language, Al-Sayyid Bedouin Sign Language, or ABSL, which is now only about 3 generations old. The focus of the research has been on word-order: ABSL seems to have emerged with SOV word-order, despite the fact that the local languages (Hebrew, Israeli Sign Language, colloquial Arabic) have other word-orders. This was reported in New Scientist back in 2005, here, and the reporter (a different New Scientist reporter) spoke to Goldin-Meadow about what Padden's results implied:
Could the SOV word order reflect some biologically innate preference? Psychologist Susan Goldin-Meadow at the University of Chicago, who has been studying 14 home signers in the US, Taiwan, Turkey and Spain, says she has also found a preference for a particular word order among home signers that cannot be down to the spoken language in the environment. "Home sign isn't quite handed down. It's invented anew each time."
And each time it is invented, it seems to be created in much the same way. As in ABSL, home signers also develop a stable word order, Goldin-Meadow says, which is almost always OV (home signers tend to omit the subject).
She has also done related experiments in which she has asked hearing native English speakers to put some pictures in order. Each picture corresponds to a subject, an object and a verb. Goldin-Meadow says they prefer to order them as SOV.
Yet whether this preference is a reflection of an innate linguistic trait, or merely of a broader cognitive one, remains to be proved. "To answer that question, you have to look at things beyond language," she says. "You have to show that it's specific to language." (Michael Erard, A Language is Born, New Scientist, October 22, 2005)
Excellent. We have a hypothesis, that SOV is an innate linguistic trait, and we have a suggestion from Goldin-Meadow about how to test it: look at non-linguistic representations, and if the same order is found in non-linguistic representations, then there's no reason to think that SOV is a specifically linguistic trait.
Fast-forward to the present. Goldin-Meadow has now finished doing the work. And she's found that the same order is indeed used in non-linguistic representations. Ergo we've no evidence at all that the SOV preference is "an innate linguistic trait." No evidence of syntax etched into the brain. If only the new New Scientist reporter had started his working day by reading what the old New Scientist reporter had written.
The discussion in the Goldin-Meadow PNAS article is completely in tune with her comments in the earlier New Scientist article (and, incidentally, with my own immediate reactions, as reported earlier), and out of tune with the way she is reported in the new New Scientist article. The PNAS article suggests that a basic order of Actor-Patient-Action (thus ArPA order, where "Patient" is one way linguists describe things that are acted upon) is cognitively natural, independently of language. That is, to the extent that anything is "etched into our brains" it's not sentence syntax, but a way of thinking about events. The PNAS article then says:
We suggest that, initially, a developing language co-opts the ArPA order used in nonverbal representations and uses it as a default pattern, thus displaying SOV order, which may have the virtue of semantic clarity.
Personally, and I say this as someone who works primarily on semantics, I don't see any evidence in the article or anywhere else that this order aids semantic clarity. (To measure clarity we'd presumably want data on how reliably, and perhaps how quickly, messages with that order were correctly understood. There's no such data in the article.) But putting this aside, the basic message of the Goldin-Meadow discussion is clear: to the extent that SOV word order tends to emerge in new languages, that can be explained without postulating that such a word order is hard-wired into our brains as a default syntax. As Goldin-Meadow et al say (commenting on the more general issue of whether our language influences our thought):
Indeed, the influence may well go in the other direction; the ordering seen in our nonverbal tasks may shape language in its emerging stages.
I can't help but wonder whether the polemical brain etching position adopted in that one paragraph of the recent New Scientist piece may result from the unfortunate *success* of linguistic education. It is possible that some relatively nativist linguists (Pinker, even) may have practiced their etching on the blank slate of the journalist's mind, perhaps years before he wrote the piece.
On balance, and looking at the two New Scientist articles together, I'd give New Scientist an A-. Generous, perhaps, but if I had any influence over them, I'd want to give encouragement: the authors/editors have picked on interesting topics, and they've by and large got the facts right, but they're occasionally off-beam with their interpretations, and they'd do better to quote scientists directly wherever possible (as in the first New Scientist piece), rather than providing indirect, and probably incorrect, pseudo-quotations of the form "X argues that…." More generally, I'm curious: when someone writes "X argues that", I wonder how often X would agree?
Janice Huth Byer said,
July 6, 2008 @ 5:21 pm
You're right to encourage good ideas, Mr. Beaver. Plus, human nature being what it, the authors/editors for New Scientist may, as a consequence, be more receptive to your essential points and good advice not to put words in researchers' mouths. For this, they may well be grateful, because it's almost certain, in time, to incur complaints.
John Cowan said,
July 6, 2008 @ 5:40 pm
Of course there is a good reason: it would infringe on entrenched privilege. This is the "good reason" why most obvious reforms don't happen.
Mary Blockey said,
July 6, 2008 @ 5:57 pm
The 2005 article was written by Michael one-r Erard.
John Roth said,
July 6, 2008 @ 8:56 pm
Science Daily appears to have it right.
http://www.sciencedaily.com/releases/2008/06/080630173943.htm
John Roth
Nathan Myers said,
July 6, 2008 @ 11:01 pm
Gives new meaning to "would you like to come up and see my etchings?"
James A. Crippen said,
July 7, 2008 @ 1:33 am
Any time I see “X argues that…” I instantly suspect the writer to have essentialized the position of X such that X would likely disagree with the description of their position. I suspect this largely because it’s exactly what “X argues that…” means in my own writing. That’s unfair to other writers, but unless the “…” are a direct quotation it’s probably true that the summarizer has gotten the argument wrong. After all, if X’s argument was so simple that it could be summed up in a single statement, why wouldn’t X have done so?
As for SOV, the really interesting thing is that SVO is probably about as common as SOV. That probably can’t be answered by argument from nonlinguistic behavior, but I’d definitely enjoy seeing someone try. This isn’t my field, so I’m curious if there are any good arguments for why SVO should be so common if SOV can be taken to be the default from nonlinguistic ArPA.
Jon Rain said,
July 7, 2008 @ 3:34 pm
"…any good arguments for why SVO should be so common if SOV can be taken to be the default from nonlinguistic ArPA."
Maybe its better?
Rubrick said,
July 7, 2008 @ 5:10 pm
I'm surprised no one has commented on the title of the article. Whatever the study may or may not have shown, it has nothing to do with Charades. Any study equating actual Charades play with sentence structure would arrive at some strange conclusions indeed. For example, I'm pretty sure humans don't have an underlying tendency to begin sentences with whichever syllable has an easily-mimed homophone.
john riemann soong said,
July 7, 2008 @ 7:02 pm
Personally I see the idea of innate preferences for word order as belonging to a "broader cognitive trait" as more exciting. Because for one, that seems to be the most natural explanation for any "universal grammar" — language universals develop out of the constraints of the architecture of the brain. It also means that many linguistic principles ultimately have some connection with other cognitive phenomena, and we can do comparative studies.
john riemann soong said,
July 7, 2008 @ 8:49 pm
"For example, I'm pretty sure humans don't have an underlying tendency to begin sentences with whichever syllable has an easily-mimed homophone."
But rebus-type logic — as it has been noted by Victor Mair in an essay on Chinese characters at pinyin.info — was key to the invention of writing.
D Jagannathan said,
July 7, 2008 @ 10:56 pm
"As for SOV, the really interesting thing is that SVO is probably about as common as SOV. That probably can’t be answered by argument from nonlinguistic behavior, but I’d definitely enjoy seeing someone try."
From a typological standpoint, pinning down the percentages is actually a pretty difficult problem because the traditional 3-letter paradigm is a poor descriptive tool. First, S and O are typically used very vaguely to try to encompass a lot of variation in argument-marking strategies. In particular, marked topicalization and focus really throw a wrench in the works, but I doubt there are very many paradigmatic languages in this respect. Next, rigid word-order and flexible word-order languages are usually lumped together when a dominant order can be determined for the latter cases. Finally, few of the sentences in corpora of spoken languages have the requisite two full-blown NPs with neutral emphasis/focus/topic in transitive, main clauses, so it's a theoretical, not a statistical choice to assign this kind of sentence as representing the basic word order of a language.
Of course these word orders correlate with other syntactic patterns, so they can be useful typological features. In any case, subject-first (SV) does seem to be a dominant pattern, but as far as I know, there's no convincing theoretical story for this aside from a vague intuition that this is the grammaticalized consequence of a general tendency for subjects to be topics and a general tendency for topics to be fronted (a point which is itself disputed).
Darekun said,
July 7, 2008 @ 11:00 pm
When I heard of this article, I immediately thought of the rebus-type miming I've seen English-speakers do – which is generally SVO. However, for me at least there's a clear intuitive difference between rebus-type miming and non-rebus-type miming, and I can easily see borrowing syntax from a native language only for rebus-type.
On the commonality of SVO, I'm reminded of some of my first forays into conlanging; verbs being more distinguishable from nouns than subjects are from objects, SVO's basic structure is easier to parse. However, this was a naive judgment made by a native speaker of English, so English's famed fondness for verbs may be the only reason I found verbs so distinguishable. SVO also describes the order of a head-initial language that doesn't distinguish verbs from nouns, which would imply an alternate development order of verb distinction and sentence order.
In any case, if SOV were natural, we'd expect most of those SVO languages to have SOV ancestors – and a quick Googling indicates Proto-Indo-European was SOV.
To me, the concept of an innate language syntax implies an innate syntax that extends beyond language, and I find it hard to imagine an innate syntax that only applies to language…
Jason Orendorff said,
July 8, 2008 @ 5:17 pm
Heh. Right after reading this, I came across a spectacularly ArPA comic strip over at Penny Arcade. Maybe SOV just has better dramatic qualities. :)
john riemann soong said,
July 8, 2008 @ 6:32 pm
"To me, the concept of an innate language syntax implies an innate syntax that extends beyond language, and I find it hard to imagine an innate syntax that only applies to language…"
Indeed!
Let's review the hypothesis: "that SOV is an innate linguistic trait".
But the key word of this hypothesis is "innate," not "linguistic trait". The latter has already been supported by the mention of SOV. What is SOV if not a linguistic trait?
Moreover, I get reminded of what a previous Language Log author said about what constitutes a good hypothesis. It was mentioned that one of the traits of a good hypothesis was if it could connect seemingly disparate elements together, because it would be far less likely that any non-null result occurred purely by chance. So what if the original hypothesis had been true — that the pattern was solely found in linguistics? What light would that shed, and what good would it do for science. But if you connected it to a pattern outside of linguistics, haven't you just made a very useful connection?
Alex Drummond said,
July 11, 2008 @ 5:10 pm
"So what if the original hypothesis had been true — that the pattern was solely found in linguistics? What light would that shed, and what good would it do for science."
The good would be in discovering something true, obviously. Take this argument to its limits and we'll be trying to connect quantum physics to discourse analysis.
Aoede said,
October 23, 2008 @ 11:27 am
Actually, you can get the article free from Goldin-Meadow's personal website.
marciano guerrero said,
August 17, 2010 @ 11:04 am
Doesn't imaging (miming) takes place in a different part of the brain outside the language centers?