May 09, 2006

Dolphin naming?

The big animal communication story of the day is is the idea that dolphins have -- and use -- "names". The background of the story, known since 1991 or so, is that infant bottlenose dophins develop "signature whistles", and as they grow up, learn to recognize the signatures of their relatives and friends. This general sort of thing is not unique to dophins. If you saw the hit movie March of the Penguins, you'll recall that "speaker recognition" was pictured as crucial to several aspects of penguin family life, for example allowing returning females to find their mates in the crowd. Many other species have similar abilities, which is not surprising given that identification of familiar individuals is crucial for most types of social organization.

What's new here, apparently, is how identity is coded. In many cases, animal "caller ID" is coded somewhat in the way that human speaker identity is, as a holistic amalgam of the effects of vocal tract size and shape, low-level motor habits, and so on. In this experiment, the "signature whistles" were "synthesized", so that the researchers were able to control which aspects of the signatures were preserved and which were omitted. However, it's a long way from the recognition of synthetic signature-whistles to the conclusion that dophins have and use "names". It would be exciting to find out that such dolphins do exhibit naming behavior, such as dolphin A calling for dolphin B by using B's "signature", but I haven't found any indication in the media coverage that the PNAS article contains any evidence of such things. (There are plenty of assertions that dolphins do name one another, some attributed to Vincent Janik, the lead author of the forthcoming PNAS article that is causing this media buzz, but it's hard to know what to make of this).

I wrote that the experiment's use of synthetic whistles is "apparently" what's new because the article, due to appear online in the Proceedings of the National Academy of Sciences this week, and presumably in the journal itself next week, isn't available yet. I'll post a more careful evalutation once the article is out. Meanwhile, experience suggests that it's a bad idea to spend too much time trying to construe what the MSM has to say about it. In particular, I'd be very careful about using words like "name" and "naming" to describe these experiments (even if the researchers themselves are quoted as doing so).

Here's what The Times says about the experiments themselves (under the headline "Dolphins 'know each other's names'"):

In the study some of the Sarasota Bay animals were corralled in a net. The researchers then played synthetic versions of the signature whistles of other dolphins through underwater loudspeakers to see if they would evoke a response in the captive animals. The use of synthetic whistles ruled out the possibility that the animals might simply be recognising the sound of each other’s voices.

They found that dolphins responded strongly to the whistles of their relatives and associates while generally ignoring those of dolphins to whom they had no link.

Janik said: “Bottlenose dolphins are the only animals other than humans to have been shown to transmit identity information independent of the caller’s voice.”

In my opinion, it's hard to know how to evaluate this statement without knowing exactly how the whistles were synthesized. When you talk on a cell phone, for example, your voice is digitized, analyzed, compressed and re-synthesized at the other end for the person you're talking to, who recognizes your voice as well as your words despite this treatment. So another way to express the results of these experiments might be to say that the experimenters' method of analysis and re-synthesis has preserved some of the features that dolphins use to distinguish between familiar and unfamiliar voices. And there's nothing in the description of the experiments to suggest that they addressed the question of whether dolphins use others' signatures vocatively or referentially.

Let me close by saying that I hope my skepticism is unwarranted in this case. It would be exciting to learn that dolphins' signature whistles are (say) like the songs of some birds, which are individualized by learned and practiced sequences of particular types of "syllables", and not like the individual barks of baboons, which (as I understand it) are individualized by their relatively automatic encoding of vocal tract size and shape, vocal-cord vibration patterns, and so on. It would be stupendous to learn that dolphins use such identifying calls in some of the ways that humans use names, e.g. to call for others by name, to refer to absent third parties by name, and so on. All of that, and more, is stated or implied in Michael Hopkin's story on the subject in, under the headline "Dolphins play the name game".

Unfortunately, the smart money says that I'll be disappointed, at least in terms of the evidence presented in the forthcoming PNAS paper. The media reliably overinterpret science stories that push their buttons, and nothing pushes people's buttons like talking animals.

Posted by Mark Liberman at May 9, 2006 06:25 AM