February 09, 2004

Cotton-top tamarins: on the road to phonology as well as syntax?

Fitch and Hauser's recent investigation of monkeys' perception of sound patterns featured cotton-top tamarins, whose own vocalizations are interesting. Specifically, they seem to exhibit what Charles Hockett called "duality of patterning": larger patterns made up of well-defined combinations of recurrent, well-defined smaller units. Hockett's model was phonology: words are not arbitrary equivalence-classes of vocal noises, but instead spell out their claims on sound in sequences of phonemes, whose phonetic interpretation is independent of the identity of the words involved. (A more modern story would represent the sound of words with phonological features arranged in syllables and similar structures -- but the principle is the same).

This idea was generalized by Michael Studdert-Kennedy and Louis Goldstein in their paper “Launching language: The gestural origin of discrete infinity” (In Language Evolution, edited by Morten H. Christiansen and Simon Kirby,  New York: Oxford University Press, 2003):

"...the only route to unbounded diversity of form and function is through a combinatorial hierarchy in which discrete elements, drawn from a finite set, are repeatedly permuted and combined to yield larger units higher in the hierarchy and more diverse in structure and function than their constituents. The particulate units in physical chemistry include atoms, ions, and molecules, in biological inheritance, chemical radicals, genes and proteins, in language, gestures (as will be argued below), segments, syllables, words and phrases."

Another way to think about this is that such systems are essentially digital:  a (relatively) small number of qualitatively distinct things can be combined in a limited number of ways to create a much larger number of distinct things. This is a useful property for a communications system, if you want to be able to send messages reliably over noisy channels. Many descriptions of animal vocalizations lack this property: instead, we find a limited number of qualitatively distinct vocal displays (such as alarm calls or isolation calls), with no further (digital) substructure, and no combinatoric properties other than what follows from being emitted at various different times as individually appropriate.

But cotton-top tamarins are said to have at least a two-level hierarchy. This page claims that "Cotton-top tamarins have an extensive vocal repertoire which is derived from the variation of two basic elements and the sequential combination of those elements (Cleveland and Snowdon 1982)." Looking at the spectrograms and listening to the recordings, I can't figure out how to cash in this generalization as a model of the specific calls shown, but I'll keep at it.

There are quite a few animals whose communicative displays have at least the beginnings of this sort of digital structure. Bird song is a well-known and well-studied case, where some terrific science is being done these days. It's less well known that cephalopods have complex communications systems that exhibit clear duality of patterning. Their systems involve body patterns (of color, reflectance and skin texture) and posture (especially of tentacles). Here's a picture of the "chromatic components in Loligo vulgaris reynaudii that are used to build up body patterns" (from Hanlon et al. 1994):

Sepia officinalis (the cuttlefish) has even more "chromatic components" (35), and apparently cephalopods combine such components in many different ways, sequentially as well as synchronously. The displays to which researchers assign meanings are combinations of these components, for instance the "lateral display" of Loligo plei involves (among other things) "arms compressed dorsoventrally, tentacles extended/dark, mid-ventral ridge extended, lateral flame markings, dark eye ring, arm spots, dark stitchwork fins, iridescent arms." As the costume designer William Ivey Long said about another kind of visual display, "those are just words. The effect is, of course, insane." At least if you're down with the semiotics of loliginid courtship displays.

I'm sure that monkeys have more to say to one another than squid do, even if they don't come up to Geoff Pullum's standards for conversational partners. But still, it's interesting that among non-human primates, some 400 million years after cephalopods evolved, the complexity of communication systems does not seem to have increased at all, at least as measured by the size of the set of basic communicative "atoms" and the richness of their modes of combination. In fact, I don't know of any non-human primates that are said to have as many qualititatively distinct elements of communicative displays as squid and cuttlefish do.

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? There are several ideas -- new and old -- for an answer to this question, which I'll survey in a later post. One theory is provided by von Humbolt: "The articulated sound, the foundation and essence of all speech, is extorted by man from his physical organs through an impulse of his soul; and the animal would be able to do likewise, if it were animated by the same urge." But there are other stories about this as well, all interesting, none yet shown to be true.

Posted by Mark Liberman at February 9, 2004 12:30 AM