Wednesday, September 26, 2007

Wouldn't it be Interesting......?

I for one would really like to see the HCF Hypothesis 3 proven wrong. That would be an interesting bit of happenstance. Perhaps we're so conditioned to believe that humans are the only ones lucky enough to indulge in linguistic behaviors that we're oblivious to what's actually going on in nature. Much like Jane Goodall's professor who was adamant that only humans solved problems with tools. So hence my inquiry about recursion. Are we sure birds lack it? I haven't studied this enough, so I don't know much of the facts, but I'd be happy to view them. On the other hand, I know it's probably difficult to deny that recursion obtains in language use (except maybe for the Piraha), but HCF did leave the door open to such a negation when claiming that that capacity may be a characteristic of other cognitive systems, such as navigation and social interpretation. Given this, might there not be minimally analogues of such a capability in other species? Other navigators or other beings that interact with others like themselves? I personally won't be so quick to write off birdsong to finite-state output, simply in deference to a theory, which is simply and only a theory. But maybe this is because I'd just be tickled to see it dashed in the name of science.


Asad Sayeed said...

Well, then, you'd still be left with the question: how is it that you're arguing this and not a bird?

Tim Hawes said...

Sarah, a letter in Nature April 2006 by Gentner et al. makes the claim that birds (European starlings) do indeed possess the capability to distinguish CFG from FSG. The grammars are composed of an alphabet of distinct motifs (8 rattles and 8 warbles; rattles and warbles being the classes A and B). They show that starlings are able to distinguish AnBn (n<=2 clearly n<=4 slightly less so) from A*B*, (AB)n, and wwr | w is in {A,B}* i.e. a string followed by its reverse. They point out that the conditioning was "slow" compared to other song-recognition tasks. However, as far as I know the training is still on the order of days or hours and not weeks or months (depends on how long a training block is….blocks are 100 trials; and it took between 94 and 562 blocks for acquisition).

They do their best to eliminate as many FSG solutions to AnBn n<=4 as they can, including the possibility that birds are classifying on the presence or absence of specific transitions. The one possibility that they clearly didn't rule out (and it may not be possible to), is that the birds are learning a FSG approximation of a CFG, which is possible when your n is finite. They do acknowledge that they have not eliminated this possibility (though they suggest that this is “contrived”) and (rightly) point out that studies in humans run into the same problems in proving that a higher level grammar is in fact being used instead of a FSG approximation.

Up until a few days ago, I was very skeptical of these results. However when I became aware that tamarins appear to fail at this task, I was slightly less than very skeptical about these results given the other things tamarins can do. As far as I can tell though, they only clearly demonstrate that the birds can distinguish A2B2 (or even give them AnBn n<=4) and (AB)n and that the birds actually recognize these strings as “grammatical” and recognize other strings of A’s and B’s as ungrammatical. My problem was (and still is) that you really don’t need that many states or transitions in your FSG to recognize AnBn n<= 4. n<= 4, 3, 2 or whatever takes AnBn which is context free and makes it finite state; and I don’t consider the FSG used to accomplish this contrived. That being said, their point that you have the same problems in conclusively proving that humans use a higher level grammar is well taken. However, in that case, I think the higher you move the upper bound on n, the more contrived things get.

In the spirit of full disclosure, I should say that my college roommate and close friend is a student of Gentner’s at UCSD, however, he started in the lab after this study was published. If anyone wants a copy I can either mail it to you or get it into the locker.

Tim Hawes said...


I talked to my friend, he said a really good bird can do up to 1000 trials a day (10 blocks) but you are looking at more like 4-5 blocks/day so about 18 – 140 days (best/worst case) given those averages.