July 01, 2006


I just finished reading Temple Grandin's recent book Animals in Translation, one of the two best books I've read this year (the other was Jared Diamond's Collapse). She includes some speculations among her observations about how animals and autistic people perceive and interact with the world, and most of these speculations are stimulating, intriguing, and even plausible -- including some of her comments on animals and language, for instance about Alex the grey parrot (I confess, at the risk of being drummed out of Language Log Plaza, that I am both partial to parrots and inclined to be gullible, and also that I would be absolutely thrilled if I could be convinced that some non-human species has the equivalent of human language; still, even allowing for my biases, Alex is one spectacular bird). But she lost me when she talked about Con [sic] Slobodchikoff's claims about prairie dogs' language.

I'm not the first Language Logger to wonder about Slobodchikoff's claims: see Mark Liberman on the subject back in 2004. As Mark observed then, "it looks very much like the pattern familiar from Seyfarth and Cheney's classic work on vervet alarm calls, with additional results on the encoding of more abstract size and shape information in call variation, and especially a focus on the use of `variation in the internal structure of a vocalization to define possible information structures', as Placer & Slobodchikoff put it in their 2004 paper." Like Mark, I haven't seen the Slobodchikoff paper that Grandin cites -- only page 1, which is available on a link from Slobodchikoff's website. Page 1 consists mainly of background information on prairie dogs and their social organization. The whole paper, C.N. Slobodchikoff's "Cognition and Communication in Prairie Dogs" (2002) is just eight pages long, so the amount of argumentation for properties equivalent to those of human language has to be very limited. Here's part of Grandin's report on it:

Using sonograms to analyze the distress calls of Gunnison's prairie dog, he [Slobodchikoff] has found that prairie dog colonies have a communication system that includes nouns, verbs, and adjectives. They can tell one another what kind of predator is approaching -- man, hawk, coyote, dog (noun) -- and they can tell each other how fast it's moving (verb). They can also say whether a human is carrying a gun or not.

Something's wrong here, and it's not just the fact that rapidity of movement is a lot more likely to be expressed by an adverb (at least from an English speaker's perspective) than a verb. The main problem is that concepts like "noun" and "verb" have to be defined syntactically, according to their functions in sentences and discourse: the old "name of a person, place, or thing" definition of "noun", for instance, just doesn't work well in real life. A phrase like "destroy Carthage" is semantically very similar to "the destruction of Carthage", but the former is a verb phrase and the latter is a noun phrase. The on-line Oxford English Dictionary defines "noun" as a word "capable of functioning as the subject and direct object in a sentence, and as the object of a preposition", and a verb as "that part of speech by which an assertion is made, or which serves to connect a subject with a predicate"; both of these are syntactically-based definitions.

The point is that unless the prairie dogs have syntactic structure in their communication system, no linguist is going to accept a claim that they have nouns, verbs, or any other part of speech, or that their system approaches the level of human language. And nobody is likely to be able to prove, in an eight-page article, that the prairie dogs have syntactic structure. According to Grandin, Slobodchikoff does in fact claim that the prairie dogs have syntactic structure -- namely, that they use "transformational rules to create their calls". She explains transformational rules as follows:

In human language, a transformational rule allows you to turn words into sentences that make sense....The prairie dogs seem to have a transformational rule based on speed. Depending on how fast a predator is moving, they speed up their calls or slow them down.
But this interpretation has nothing at all to do with the transformational rules linguists talk about, or with syntactic structure. So if this is what Slobodchikoff has in mind, his understanding of what human language is like is shaky. None of this means, of course, that prairie dogs don't have an impressively elaborate system of alarm calls. But it is far, far from human language in its expressive and structural properties, again assuming that Grandin is reporting Slobodchikoff's brief article accurately.

Posted by Sally Thomason at July 1, 2006 11:05 PM