Recognizing grammar (or door chime changes, or anything)

« previous post | next post »

It has been two weeks now, and so far no one here at Language Log Plaza has commented on the BBC News story entitled "Monkeys recognize bad grammar." I suppose people are assuming that I cover the Stupid Animal Communication Stories desk. And often I have. But I have been procrastinating, because I am getting tired of being the animal grammar killjoy. People are beginning to think I hate monkeys and dogs and parrots and dolphins and such (my previous posts include this one and this one and this one and this one and this one and this one and this one and this one and probably others).

The little animals in question (it's cottontop tamarins again) are cute. I don't have anything against them, or against the experiments on them being done by people like Marc Hauser. In the present case, the team was led by Ansgar Endress. And here is the evidence for these little creatures' ability to "recognize bad grammar". It's quite simple, and I don't think it's going to get them jobs as copy editors.

It turns out that if you expose a cottontop tamarin to a recording of someone saying shoy-bi, shoy-la, shoy-ro, etc., over and over again for a whole day, they get used to it; and then the next day if you do it some more, but then interrupt the monotony with the sound of someone saying bi-shoy, or la-shoy, they look at the sound source. It's novel to them now, and the novelty surprises them a little bit, so they look at the loudspeaker. That's it.

Clever monkeys: they know the sound sequences of the current ambient environment, and they're alert to what's new in the auditory environment. But really: recognizing bad grammar? I suppose Harvard behavioral scientists have to make their papers sound potentially relevant to cognition, and BBC science reporters have to make stories sound potentially interesting, and a headline like "Monkeys recognize changes in their auditory environment" would not set the world (or the World Service) on fire. But that seems to be what we have here. The researchers have just dressed up the description of the recorded sounds in linguistic terminology ("prefixation", "suffixation"). But it's not clear to me that a monkey's ability to notice the difference between shoybi and bishoy is any more linguistically interesting than an ability to notice the difference between thump-splash and splash-thump, or to notice that your door chime has just gone DONG DING instead of DING DONG or that you just changed the radio station.


  1. Spectre-7 said,

    July 22, 2009 @ 2:13 am

    …or that you just changed the radio station.

    Which raises a very important question: can cottontop tamarins recognize different musical genres? Mmmmm, I can smell the research grant already.

  2. JAK said,

    July 22, 2009 @ 2:19 am

    Exactly my thoughts when I read the BBC article in question. And yes, I have been wondering when someone at my favorite plaza would comment on it. :-)

    I must confess that I have not read the full paper, however, reading through the article I got a feeling that the authors (not sure if its the article's or the paper's) were trying to suggest that since the response to change in the auditory environment (as you had put it) is similar to those of humans, they were assuming that similar neurological circuits should be involved in processing the information, suggesting either that primates have the ability to parse basic grammar or that processing sounds like the ones in the experiment involves some other basic circuitry based on memory which I guess somehow changes our view of how we acquire and process language.

    Personally, I thought it was bit of a leap in imagination to get to that conclusion, but again, as I said I didn't read the full paper. I would be interested in knowing if anyone else got the same feeling.

  3. Sili said,

    July 22, 2009 @ 2:27 am

    I'll gladly hate monkeys on your behalf.

    But when is someone gonna take Ben Goldacre to task for linguification in his book?

  4. Ceiswyn said,

    July 22, 2009 @ 3:26 am

    Yeah… when I was reading that story I was expecting them to at least have some simple 'grammatical rules' about when sounds could be used as prefixes or suffixes, rather than 'this sound always comes at the start'. Because frankly, forget monkeys; my *cats* could 'recognise bad grammar' by this definition.

    (Wonder if I could get a research grant to check? :) )

  5. Alex Case said,

    July 22, 2009 @ 5:50 am

    Well debunked. Is anyone going to do the "we can teach Japanese adults /l/ and /r/ with motherese on a computer and that apparently proves that small kids are better at learning foreign languages" story?

  6. Jan said,

    July 22, 2009 @ 5:58 am

    They also used different voices for the syllables: A man saying "shoy" and a woman adding "-la" or so. I guess that was done to amplify the contrast between "stem" and "affix", but makes the task even less linguistic and more akin to the "ding-dong" thing, I think.

    [(myl) Good point. See "Hi Lo Hi Lo, It's off to formal language theory we go", 1/17/2004, for a discussion of the role that this technique played in an earlier experiment on Tamarins. While I normally hesitate to say this sort of thing about colleagues, I find it hard to see this stimulus-design technique as anything other than a public-relations stunt. ]

  7. Mark Liberman said,

    July 22, 2009 @ 6:04 am

    Geoff makes just the right point about this news item (and the associated experiment), in just the right way.

    When the news broke, with news organizations re-writing the press release in unison on July 8, I started a post on the topic, collecting the original article (Ansgar D. Endress, Donal Cahill, Stefanie Block, Jeffrey Watumull and Marc D. Hauser, "Evidence of an evolutionary precursor to human language affixation in a non-human primate", Biology Letters, Published online before print July 8, 2009), and a sample of headlines and links:

    Victoria Gill, "Monkeys recognize 'bad grammar'", BBC News, 7/8/2009; Matt Kaplan, "Monkeys Recognize Poor Grammar", National Geographic, 7/8/2009; Jennifer Viegas, "Monkeys display verbal skills", ABC Science Online, 7/8/2009; Allison Bond, "Clever Monkeys Can Recognize Basic Grammar", Discover, 7/8/2009; Meg Marquardt, "Monkeys show language recognition", 7/8/2009, Examiner; Catherine Brahic, "Monkeys have a memory for grammar", New Scientist, 7/8/2009; "Monkeys understand unspoken complexities of language", redOrbit; 7/8/2009; "Even Monkeys have their grammar in place", Thaindian, 7/8/2009; Tudor Vieru, "Tamarins Can Discover Bad Syllable Order in Words", Softpedia, 7/8/2009; "Clever Monkeys Can Recognize Basic Grammar", Discover Magazine; …

    I even looked into the prior literature on habituation to temporal-sequence stimuli.

    But I couldn't get myself interested enough to knock out a post on it all — the original article didn't strike me as very important, and the fact that the media were misled into being so misleading about it is just one more example among thousands. So I'm really grateful for Geoff's deft handling of the topic.

    I'll just add one ironic note: the whole point of the original article was that "components of our linguistic competence are shared with other animals, having evolved for non-linguistic functions" — i.e. what the Tamarins were doing is precisely not what the authors think is really "grammar" proper. (See "JP versus FHC+CHF versus PJ versus HCF", 8/25/2005, for a discussion of the linguistic arguments on this point.)

    Endress et al. could have contributed to this discussion in a more forceful and pointed way by using clearly non-linguistic stimuli, like buzzers or colored lights, and by using less cute and anthropomorphic lab animals, like rats; but I predict that this would not have made the news so widely. On the other hand, what all the news articles wrote about this experiment was the opposite of the point that the authors wanted to make within the field.

  8. Richard said,

    July 22, 2009 @ 7:41 am

    In an unrelated (and entirely not stupid) news story: since this is, I think, his first post on LL since the news was announced formally, I hope GKP won't object to us congratulating him here on his election to the British Academy.

  9. James Wimberley said,

    July 22, 2009 @ 7:45 am

    The item does however provide Geoff with a readymade label for the PR-driven scientists and the gullible journalists they feed: cottontops.

  10. Alex said,

    July 22, 2009 @ 7:54 am

    To answer the very first question in the thread: apparently they don't like music much:

  11. Joshua said,

    July 22, 2009 @ 11:21 am

    I haven't read the article (and probably won't), but it occurs to me that there's a more obvious and illustrative way to demonstrate what they're claiming they saw in this experiment.

    Specifically, they're testing with a series of affix-suffix pairs, right? So what if they got the tamarins to memorise a subset of the affix-suffix pairs, then tested them on a different one? So, they train the tamarins on "shoy-bi" and "shoy-la", then test them on "shoy-ro". If they know grammar, that should be novel to the monkeys but less novel than the "bi-shoy" combinations.

  12. Robert Coren said,

    July 22, 2009 @ 11:36 am

    I don't think it's going to get them jobs as copy editors.

    Not that anyone would particularly notice, these days, if all the copy editors in the world were replaced by cottontop tamarins.

  13. Joshua said,

    July 22, 2009 @ 11:47 am

    Although, now that I think about my last comment, there's still no way to distinguish between reacting to novel grammar and reacting to novel sounds. Phooey.

  14. Ellen said,

    July 22, 2009 @ 1:32 pm

    I find the idea that the tamarins are noting and reacting to anything grammartical odd, and surprising that someone would think that, because, frankly, I see no grammar evident in the pairs of syllables.

  15. Mark P said,

    July 22, 2009 @ 2:14 pm

    If I have my dog before me in the context of giving a command (that is, he knows I am going to say something or give him food in exchange for behavior) he will look at me with greater attention if I say something different from but similar to "sit." Thus all that's really necessary to be able to say that my dog can recognize grammar is to redefine "grammar" as "sound order" or "sound pattern." It seems to me that many great scientific discoveries can be made by redefining things. For example, let's define the speed of light as 66 miles per hour.

  16. MikeyC said,

    July 23, 2009 @ 3:07 am

    So if you say "sid" to your dog, he doesn't sit? Does that mean he recognises minimal pairs?

  17. MBM said,

    July 23, 2009 @ 6:19 am

    Great to see somebody commenting on the news story. I was suspicious when I read it and now I know why.

  18. Mark P said,

    July 23, 2009 @ 8:26 am

    MikeyC, I don't think he can distinguish between "d" and "t". He would probably react if I said "eat". It may be that he isn't recognizing the newly redefined grammar, but instead is wondering where I got my strange accent.

  19. Achim said,

    July 23, 2009 @ 8:46 am

    Maybe the experimenters should have used human infants as controls. By way of sucking rate changes as a result of changing sound patterns news writers could see a "proof" that cottontop tamarins have same grammatical abilities as toddlers. Imagine these headlines ;-)

  20. Simon Cauchi said,

    July 23, 2009 @ 6:08 pm

    Once when our cat miaowed in a commanding fashion, our dog obediently sat.

  21. Faldone said,

    July 24, 2009 @ 6:43 am

    Did the tamarins-have-grammar conceit come from the study or was it generated by the journalists?

    [(myl) The abstract of the Endress et al. paper:

    Human language, and grammatical competence in particular, relies on a set of computational operations that, in its entirety, is not observed in other animals. Such uniqueness leaves open the possibility that components of our linguistic competence are shared with other animals, having evolved for non-linguistic functions. Here, we explore this problem from a comparative perspective, asking whether cotton-top tamarin monkeys (Saguinus oedipus) can spontaneously (no training) acquire an affixation rule that shares important properties with our inflectional morphology (e.g. the rule that adds –ed to create the past tense, as in the transformation of walk into walk-ed). Using playback experiments, we show that tamarins discriminate between bisyllabic items that start with a specific ‘prefix’ syllable and those that end with the same syllable as a ‘suffix’. These results suggest that some of the computational mechanisms subserving affixation in a diversity of languages are shared with other animals, relying on basic perceptual or memory primitives that evolved for non-linguistic functions.

    Ipse dixit.]

  22. Mark P said,

    July 24, 2009 @ 8:02 am

    So, that's one small step for researchers, one giant leap for journalists.

  23. Ansgar & Marc said,

    July 24, 2009 @ 12:14 pm

    We are glad that Mark at least bothered to read the actual article before commenting, and noticed that the different media outlets concluded the opposite of what we wrote in the original article, and that there is no silly claim of the sort that the monkeys have grammar. Nevertheless, we would like to make a few points.

    First, we have an actual empirical data point about some aspects of language evolution, rather than mere speculation. It is debatable what it is worth, but at least it goes beyond mere speculation. While it will certainly not answer how language at large evolved, speculating about this questions hasn’t led anywhere anyhow, so trying to track down the origins of very simple and specific aspects of language is the only way to go if you want to get any traction at all. And maybe even this approach will lead us nowhere at all.

    Second, the monkeys managed to generalize regularities based on the first and the last position to entirely novel items, probably due to species-general memory mechanisms that track positions in sequences. The first and the last position are important in many linguistic regularities, as any book about typology will show. Besides, these regularities are just the kinds of regularities that have been discussed most extensively in language acquisition (e.g., in the past tense debate, where the regular form also involves putting stuff into the edges), so we would guess that it is a relevant data point that monkeys can learn such regularities too: putting stuff into edges is definitely not something that is specific to language. We further note that this is a capacity that is unlikely to be specific to tamarins, other monkeys or even nonhuman primates more generally; we suppose it is a very basic capacity tapping general mechanisms of serial ordering and memory.

    Third, far from being a confound, using human speech for monkeys is crucial. It's not part of their vocal repertoire, and it has no meaning to them. Nonetheless, it causes them to show an orienting response. Hence, they can learn ordering regularities for rather arbitrary yet biological stimuli that are clearly not part of their own vocal repertoire. Whether they would perform similarly with bells and whistles is an open question, but does not undermine the results presented nor the conclusions drawn as they did not ride on species-specific sounds, or the possibility that tamarins consider the stimuli as speech.

    Ansgar & Marc

RSS feed for comments on this post