The language of elephants


Hot upon the heels of the prairie dog story comes another about animal language. This time it’s elephants.  And two different teams of researchers.  There is one in San Diego Zoo, reported on by the BBC:

And another at Cornell University with field workers in Africa, reported on at some length by CBS news:

Essentially, they are claiming the same thing; that elephants make low frequency vocalisations which humans can’t hear, and which convey meanings, such as (in the zoo research) ‘I am about to give birth’. The CBS report has more detail and so I’ll concentrate on that.

As a piece of journalism, it has the high level of anthropomorphism you would expect, and you have to cut through all this to try to see what exactly the Elephant Listening Project at Cornell University is up to. As far as I can make out, they take extensive recordings of all elephant vocalisations, both audible and inaudible to humans, while observing very closely what the elephants are doing at the time. By matching sound with behaviour, they say they hope to compile an elephant ‘dictionary’. They have a few entries already: glad to see you, hit the road, I’m upset, everything is OK, here I am!

Convinced that there are more words to be deciphered, the Cornell team pours over the mass of sound and video recordings delivered to them by Andrea Turkelo, who spends the best part of every day observing a large group of elephants at a nature reserve in Central Africa. The Cornell team use computers to generate spectrograms of each vocalisation so that they get a visual representation to analyse.

“And what does this visualization tell us?” asks the reporter, looking at the tight mass of steeply falling and rising lines. Peter Wrege replies, “It tells us that there’s incredible complexity. Many of their calls are actually similar in some ways to human speech.”

To a linguist, this is immensely interesting and I would love to know more. But we don’t get any more. The reporter asks what practical value this research has, (which is a fair question) and Peter Wrege veers off to talk about how threatened the elephants are by poaching, and how important it is to be able to keep track of where they are so that they can be protected. Very true. Elephants deserve whatever protection we can give from people who want to kill them for their ivory, and it’s great that a research project which tracks them to match their calls to their behaviour can have this handy spin-off effect.

But I really do want to know what this ‘incredible complexity’ is in their vocalisations and in what ways it resembles human speech. Because on the face of it, it does not appear to resemble human speech at all. Human speech (as opposed to other human vocalisations  such as screaming, laughing, giggling, crying, tutting, shushing and whooping) is made up of discrete bits of sound (phonemes) which are ordered in different ways to give different words.  In English, for example, the phonemes /æ/ /p/ and /t/ can be arranged to give three words: apt, pat and tap. Add an /s/ to the mix and you can have spat and taps and also sapped. (Remember that phonemes are sounds, not letters of the alphabet.) It is on this brilliant design that the complexity of human speech rests. If you can articulate a set of phonemes, let’s say at least 20, you can combine them in different ways to make an impressive number of words. Combine these words in rule-governed  ways and you have sentences, and no end to what you can speak about.

I would doubt the Cornell team are looking for phonemes in the elephant vocalisations, or for grammatical rules.  Actually, I don’t even think they are looking for words. It’s not really an elephant dictionary they are compiling. It’s more a phrasebook. And it’s a phrasebook that is a bit vague about meaning. Such and such a call with a particular spectrogram contour, combined with engaging in a particular behaviour, ‘means’ the elephant is, say, angry. We can put language into that by claiming the elephant is ‘saying,’  Stay away from me or else’, or another such phrase with the same general meaning, such as, ‘I may well attack you if you fail to remove yourself from my immediate vicinity.’  There is certainly structural complexity in the human language translation. But was there any in the original elephant signal? Even if the spectrogram looks complex, with many spiking rises and falls, it might be complex in the way a human scream of arghhhhhaaaaghhhaaaa!!!!!!  is complex on a spectrogram. But the rises and falls of a scream don’t convey a precise meaning. If there is anything semantically complex to be teased out of these elephant vocalisations, it has to be consistent across calls and across individuals. Then it gets interesting; then it gets similar to human language.

The CBS report did not give me enough to make a judgment about what the Cornell team are doing. I shall have to keep an eye out for their publications and hold fire till I can read what they themselves claim for their data. All I can say at this point is that to call this discovery the ‘secret language’ of elephants (which is the kind of simplification the press loves) is somewhat premature. 


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s