April 03, 2005

Elephant talk

Some recent observations of African elephants apparently learning to imitate sounds were noted by Henry Fountain in the NYT 3/29/2005:

In Kenya, a 10-year-old elephant named Mlaika seems to think she's a truck. At least she has been heard imitating the low rumble that trucks make on a nearby highway.

Mlaika's mimicry is described in the journal Nature, along with a report of an African elephant that lived in a Swiss zoo with Asian elephants and learned to imitate the chirping that only the Asian species makes.

The two findings show for the first time that elephants - like primates, birds, bats and some marine mammals - are capable of vocal learning. The discovery has important implications for understanding how elephants communicate.

This is important for reasons that go far beyond communication among elephants. Vocal learning seems to be rare among animals, to a degree that is surprising to most people. Imitating a sound is easy and natural for us, and so it's natural to assume that any intelligent animal who can hear and can vocalize shoud also be able to do it. However, the fact seems to be that this ability is quite rare: Eric Jarvis at Duke University discovered not too long ago that hummingbirds have it, and that was big news at the time. This page on his lab's web site expresses the now-standard view that

Vocal learning, the substrate of human language, is a very rare trait. It is known to be present in only 6 groups of animals: 3 groups of birds (parrots, songbirds, and hummingbirds) and 3 groups of mammals (bats, cetaceans[whales/dolphins], and humans). All other groups of animals are thought to produce genetically innate vocalizations. To understand this concept, it is important to distinguish vocal learning from auditory learning. Auditory learning is the ability to make sound associations, such as a dog learning how to respond to the sound "sit". All vertebrates have auditory learning. Vocal learning is the ability to imitate sounds that you hear, such as a human or a parrot imitating the sound "sit". Currently only vocal learners have been found to have forebrain regions dedicated to vocal learning and production of these learned vocalizations. Vocal non-learners only have been found to have non-forebrain vocal regions responsible for the production of innate vocalizations. [emphasis added]

Thus the statement in the NYT article that "primates, birds, bats and some marine mammals" are capable of vocal learning has false implications. Since it says "some marine mammals", but leaves "primates, birds, bats" unmodified, most readers will think that all primates and birds have the ability, whereas just three groups of birds (parrots, songbirds and hummingbirds) and one species of primates (humans) were previously known to have it.

The source of the new information about elephants is a paper in last week's issue of Nature: Joyce H. Poole, Peter L. Tyack, Angela S. Stoeger-Horwath & Stephanie Watwood, "Animal behaviour: Elephants are capable of vocal learning". Nature 434, 455-456 (24 March 2005). Here's the abstract:

There are a few mammalian species that can modify their vocalizations in response to auditory experience— for example, some marine mammals use vocal imitation for reproductive advertisement, as birds sometimes do. Here we describe two examples of vocal imitation by African savannah elephants, Loxodonta africana, a terrestrial mammal that lives in a complex fission–fusion society. Our findings favour a role for vocal imitation that has already been proposed for primates, birds, bats and marine mammals: it is a useful form of acoustic communication that helps to maintain individual-specific bonds within changing social groupings.

On Nature's "Supplementary Information" site for this paper, you may be able to listen to .wav files of Mlaika imitating a truck, Calimero producing a chirp-like call, and an adult female Asian elephant producing a chirp call. (I say "may" because I'm not sure which parts of the site require a subscription -- Nature in general is not an Open Access publication.)

I suppose that this Asian-elephant "chirp call" is what Kipling called "the 'hoot-toot' of a wild elephant" in Toomai of the Elephants:

At last the elephants began to lie down one after another, as is their custom, till only Kala Nag at the right of the line was left standing up; and he rocked slowly from side to side, his ears put forward to listen to the night wind as it blew very slowly across the hills. The air was full of all the night noises that, taken together, make one big silence—the click of one bamboo-stem against the other, the rustle of something alive in the undergrowth, the scratch and squawk of a half-waked bird (birds are awake in the night much more often than we imagine), and the fall of water ever so far away. Little Toomai slept for some time, and when he waked it was brilliant moonlight, and Kala Nag was still standing up with his ears cocked. Little Toomai turned, rustling in the fodder, and watched the curve of his big back against half the stars in heaven; and while he watched he heard, so far away that it sounded no more than a pinhole of noise pricked through the stillness, the ‘hoot-toot’ of a wild elephant.

All the elephants in the lines jumped up as if they had been shot, and their grunts at last waked the sleeping mahouts, and they came out and drove in the picket-pegs with big mallets, and tightened this rope and knotted that till all was quiet. One new elephant had nearly grubbed up his picket, and Big Toomai took off Kala Nag’s leg-chain and shackled that elephant fore-foot to hind-foot, but slipped a loop of grass-string round Kala Nag’s leg, and told him to remember that he was tied fast. He knew that he and his father and his grandfather had done the very same thing hundreds of times before. Kala Nag did not answer to the order by gurgling, as he usually did. He stood still, looking out across the moonlight, his head a little raised, and his ears spread like fans, up to the great folds of the Garo hills.

Of course none of this is not really "elephant talk", although Kipling assumes in his usual anthropomorphic way that elephants can communicate complex ideas:

Kala Nag, which means Black Snake, had served the Indian Government in every way that an elephant could serve it for forty-seven years, and as he was fully twenty years old when he was caught, that makes him nearly seventy—a ripe age for an elephant. He remembered pushing, with a big leather pad on his forehead, at a gun stuck in deep mud, and that was before the Afghan War of 1842, and he had not then come to his full strength. His mother, Radha Pyari,—Radha the darling,—who had been caught in the same drive with Kala Nag, told him, before his little milk-tusks had dropped out, that elephants who were afraid always got hurt; and Kala Nag knew that that advice was good, for the first time that he saw a shell burst he backed, screaming, into a stand of piled rifles, and the bayonets pricked him in all his softest places. [emphasis added]

It's very unlikely that elephants can communicate at anything like that level of complexity and abstraction. Still, vocal learning is felt to be one piece of the biological substrate needed for (spoken) language to develop.

I suspect that vocal learning is somewhat commoner among animals than scientists now recognize, so that hummingbirds and african elephants are not the last species who will be found to have it. I've seen someone who taught a Yorkshire terrier to imitate slowly rising pitch contours, and have myself sung along with a mutt who seemed to imitate motifs from George Jones and Mozart. It never occurred to me to submit a paper to Nature -- perhaps I should have done so!.

And I wonder, could the famous hybrid whale song of the North Pacific be the result of confusing adult role models rather than cross-species breeding? As I understand it, the species apparently involved are not among those that have been thought to exhibit vocal learning, but I think that this has simply been assumed on the basis of the stereotyped nature of those of their vocalizations that have so far been identified and studied.

Anyhow, all this raises again the question that I asked in an earlier post: "The mechanical substrate for language seems to have been lying around, ready for use, for hundreds of millions of years. Why didn't evolution pick up on the possibilities in a serious way until so very recently?"

Posted by Mark Liberman at April 3, 2005 08:26 AM