Slashdot Mirror
Monkeys Show Language Recognition
mmmscience writes "The cotton-top tamarin monkeys can apparently tell the difference between suffixes and prefixes. They will turn to face the direction of recorded words when they hear the nonsense syllables "bi-shoy" change to "shoy-bi." The lead author, Ansgar Endress, suggests that this is just like how human infants learn language, by tracking the beginning and ends of words."
A draft of the actual article is at:
http://adendress.googlepages.com/endress-affixation.pdf
The experiment did not proceed as you indicated (I'm not criticizing you, I had to go to find the draft to determine this). The monkeys were presented with a "familiarization" stage that consisted of ~30 minutes of "words" where "shoy" was either always a prefix or always a suffix (depending on condition) to one set A of stem syllables, then were presented with a "test" stage where they heard "shoy" sometimes as a prefix and sometimes as a suffix on a different set B of stem syllables. They found that monkeys who had heard "shoy" as a prefix in the familiarization stage looked at the speaker longer after hearing test items that had "shoy" as a suffix (as compared to test items that had "shoy" as a prefix), and that those who had heard "shoy" as a suffix in familiarization looked at the speaker longer after hearing test items with "shoy" as a prefix.
They do seem to have shown that the monkeys can do some sort of abstraction when performing this "shoy-first or shoy-last" sequence analysis. None of the test items ever appeared in the familiarization stage (since the stem syllables of familiarization were different from those of test), so they aren't simply indicating whether they've heard that particular sound file or not. It's also interesting that they could do this in the face of (some) talker variation (due to sex and other factors), as more than one talker was used to produce stimulus materials.
I'm not sure if they can really make any claims about how humans learn language though. Aside from how unnatural the stimulus materials are (each syllable of the two-syllable words was producedy by a different talker), their conclusions are that... kids pay attention to the order of when they hear things? We already knew that and more from e.g. Saffran et al. (1996). I'd like to see them do some variation of that artificial language study with their monkeys and see if the monkeys will do two levels of distributional analysis (word segmentation and morpheme segmentation)
And yes, I Am A Linguist.
Read the original article instead of what gets reported, which might clear up a few things. You can find it at http://adendress.googlepages.com/endress-affixation.pdf
In language, sequence onsets and sequence ends are extremely important. For example, in many languages, you have prefixes or suffixes, but infixes (e.g., fun-fucking-tastic) are exceedingly rare. Likewise, stress is always located relative to the first syllable of some unit or relative to the last syllable. (Stress is the difference in pronunciation between 'record' used as a noun, and 'record' used as a verb.) And you find much more abstract regularities like this.
There might be a simple reason for this pattern: it's easier to track the first and the last position than any other position. For example, when you hear the sequence XNVSUCYPL, you know that X came first, L came last, but probably not that S was in the fourth position, although you might know that it was in the first half of the string. The same is true for pretty much any animal that has been tested: it's easier to track edge positions than middle positions.
If the observation that sequence edges are important in language has anything to do with the observation that sequence edges are particularly easy to track for memory mechanisms, then there is one crucial prediction: nonhuman animals (who can track edge positions) should learn open-ended ordering regularities based on the first and the last position. That is, they should match regular expressions like /^shoy.+/ and /.+shoy$/ - even when, and that's the crucial part, they have never heard the items before. So they have to generalize the regularity to arbitrary, novel strings - as long as shoy comes first or last.
The results show that the monkeys can learn such open-ended ordering regularities. And while that's obviously not all it takes to build a language, ordering regularities are *one* crucial aspect, as "John kicks Mary" isn't the same thing as "Mary kicks John". So the basic ability to learn such ordering relations is present in cotton-top tamarins, but they obviously don't use it for anything linguistic. Conversely, we might find all these edge-based regularities in language because we inherited (memory) mechanisms from our common ancestors that are particularly good at tracking stuff in edge positions, and humans might have recycled these mechanisms for linguistic purposes. In sum, this is not a piece about the evolution of language, but rather a piece about the evolution of a very specific aspect of what gets used in language. But because it's so pedestrian you can actually test it experimentally.
And yes, I'm one of the authors.
Oh, we understand "the look" alright. It means "I'm angry, possibly not even at you, and it's likely you have no idea who did what wrong to make me this way, but you will pay. If I'm particularly clever I might even make you believe it's your fault." The lack of further information and the subsequent worry on your part is an intended effect.
Rampant carbon sequestration destroyed the Dinosaurs' tropical paradise. I'm here to help repair the damage.