May 28, 2006

Monkey words

To balance our occasional complaints about foolish and misleading science reporting, I'd like to commend an article by Nicholas Wade in the NYT ("Nigerian Monkeys Drop Hints on Language Origin", 5/23/2006), on recent research by Kate Arnold and Klaus Zuberbühler ("Language evolution: Semantic combinations in primate calls", Nature 441, 303, 18 May 2006).

Wade's first two paragraphs describe the new research, with an appropriately nuanced claim about its importance:

Researchers taping calls of the putty-nosed monkey in the forests of Nigeria may have come a small step closer to understanding the origins of human language.

The researchers have heard the monkeys string two alarm calls into a combined sound with a different meaning, as if forming a word, Kate Arnold and Klaus Zuberbühler report in the current issue of Nature.

In the third paragraph, Wade sets the stage in an informative and sensible way:

Monkeys are known to have specific alarm calls for different predators. Vervet monkeys have one call for eagles, another for snakes and a third for leopards. But this seems a far cry from language because the vervets do not combine the calls into anything resembling words or sentences.

(This is a reference to the work of Dorothy Cheney and Robert Seyfarth, explained at length in their wonderful book How Monkeys See the World.) Wade then describes the new facts as Arnold and Zuberbühler reported them :

The putty-nosed monkeys have a "pyow" call meaning there are leopards about and a hacklike sound to warn of the crowned eagle. The "pyow" calls attention to a leopard on the ground.

When hearing the "hack" sound, a monkey tends to freeze because movement would betray its position to an eagle.

Dr. Arnold and Dr. Zuberbühler, zoologists at the University of St. Andrews in Scotland, noticed that adult male monkeys in each troupe were combining the "pyow" and "hack" calls.

Playing back a "pyow-hack" call to see how the monkeys interpreted it, the zoologists found it made the troop leave the area.

This lays out what is new in the research: the animals' response to playback of the combined calls. Monitoring response to call playbacks is a technique pioneered by Cheney and Seyfarth with vervet alarm calls, so this is a new application of an old technique. And what does it mean that putty-nosed monkeys use (and respond to) combinations of two different alarm calls? Wade explains:

Researchers studying monkeys and apes have learned that they possess all the basic apparatus needed to make and analyze sounds. But the nonhuman primates did not seem to possess either of the two combinatorial features of language, those of combining discrete sounds into compound words, and of stringing words together under rules of syntax.

Dr. Zuberbühler said that he and Dr. Arnold had not observed anything resembling syntax, but the putty-nosed monkeys, Cercopithecus nictitans, "combined two types of utterances according to a rule and the combination takes on a novel meaning," a procedure perhaps analogous to forming a word from two sounds.

Notice first that Wade correctly reports Zuberbühler's evaluation that this has no apparent connection to the development of syntax, but may have something to do with phonology (which was also David Beaver's more vividly expressed suggestion). Wade (and Zuberbühler) are also appropriately tentative about the connection to phonology.

Logically, there are a number of possible sources for the phonological "duality of patterning" which is robustly observed in all human languages, but exists (at best) in embryonic or allusive forms in non-human animals. This system for making meaningful messages out of intrinsically meaningless parts -- a digital form of message coding -- might arise from assigning meanings to vocal displays that were originally purely formal (like the songs of birds or whales); or it might arise from combining meaningful bits of behavior (like alarm calls) into communicative structures with new and (and least partly) arbitrary associations. I believe that Klaus Zuberbühler's idea is that the putty-nosed monkeys may be taking a first small step along the second path.

Wade ends with a skeptical quote from Marc Hauser:

Marc Hauser, an expert on animal communication at Harvard, said that the observation was very interesting but that stricter criteria should be applied before assuming the combination of alarm calls was similar to the way people combined sounds into words.

"Because there is no evidence that the calls are words or even wordlike, the connection to language is tenuous." he said.

This might be a bit too skeptical, at least as a way to end the piece; but it's an appropriate corrective for the range of spectacular over-interpretations elsewhere in the media. These include headlines like "Monkeys use 'Sentences', Study Suggests" [National Geographic], "African monkey can 'talk in sentences'" [The Independent], "Monkeys Found Using Primitive Linguistic Grammar" [Fox News]; and leads like "Monkeys are able to string together a simple 'sentence', according to research that offers the first evidence that animals may be capable of a key feature of language" [Mark Henderson in], or "Researchers said Nigeria's putty-nosed monkey sometimes communicates with a combination of sounds different from others, offering, they say, the first proof that animals may be able to talk" [UPI].

Another good feature of Wade's story: it gives a link (in the margin under the heading "Related") to the original paper in Nature.

Recently, the major journals Science and Nature have been vying with one another in giving prominent display to papers about animal communication research. This in a fascinating topic, in my opinion, and I think that the published research is individually and collectively valuable (even if I sometimes disagree with the interpretation). Animal communication stories evoke deep resonances in our culture, and so these stories generally also make into news media of all sorts, mostly in bizarrely misconstrued forms.

[ One of the consequences of recent attention from Science and Nature is that the animal communication research featured in the media is not always the most interesting stuff. For example, it's (logically though not journalistically) odd to play up the recent Nature article on putty-nosed monkeys, while ignoring Zuberbühler's (in my opinion even more interesting) 2002 paper on cross-species communication between Diana monkeys and Campbell's monkeys in Côte d'Ivoire ("A syntactic rule in forest monkey communication", Animal Behaviour, Vol. 63 no. 2 , Feb. 2002, pp. 293-299).

Here's the abstract from that paper, for those without subscription access:

Syntactic rules allow a speaker to combine signals with existing meanings to create an infinite number of new meanings. Even though combinatory rules have also been found in some animal communication systems, they have never been clearly linked to concurrent changes in meaning. The present field experiment indicates that wild Diana monkeys, Cercopithecus diana, may comprehend the semantic changes caused by a combinatory rule present in the natural communication of another primate, the Campbell's monkey, C. campbelli. Campbell's males give acoustically distinct alarm calls to leopards, Panthera pardus, and crowned-hawk eagles, Stephanoaetus coronatus, and Diana monkeys respond to these calls with their own corresponding alarm calls. However, in less dangerous situations, Campbell's males emit a pair of low, resounding 'boom' calls before their alarm calls. Playbacks of boom-introduced Campbell's alarm calls no longer elicited alarm calls in Diana monkeys, indicating that the booms have affected the semantic specificity of the subsequent alarm calls. When the booms preceded the alarm calls of Diana monkeys, however, they were no longer effective as semantic modifiers, indicating that they are meaningful only in conjunction with Campbell's alarm calls. I discuss the implications of these findings for the evolution of syntactic abilities.

Anyway, given the need to follow Nature's lead in assigning relative importance, I think that Wade's 5/23/2006 NYT story is a model of how to approach such topics in a responsible way. ]

If someone were to treat this research in a longer feature, there are some other kinds of background that would be worth bringing out. As I discussed at greater in an earlier Language Log post ("Cotton-top tamarins on the road to phonology as well as syntax", 2/9/2004), a number of other animal communication systems are said to

exhibit what Charles Hockett called "duality of patterning": larger patterns made up of well-defined combinations of recurrent, well-defined smaller units.

In that post, I linked to a page (now alas at a different URL, so I've updated the link) offering spectrograms and audio clips of the vocalizations of cotton-top tamarins, including some vocalization type that seem to be combinations of sound classes used independently in other circumstances.

These include repeated sequences called "pulsed vocalizations". In most cases, the "pulsed" form seems just to be an intensified form of the simple call, like the multiple dog barks or whines that David Beaver cites. Thus the "Type F Chirp" is said to be used "During intergroup antiphonal calling of Normal Long Calls. To audible outgroup vocalizations." The corresponding pulsed form, the "Type F Chirp Trill", is "Same as for Type F Chirp. Tilling [sic] indicative of a higer state of arousal than Type F Chirp alone."

But sometimes , there is a suggestion of a different process. Thus the "type D chirp", which is glossed as a "post-food" call used "when an animal actually possesses food or object", has a pulsed form called "hooked chatter", which is used "as infants approach". Perhaps this is because the adults are saying "hey kid, I got something for you here!" And perhaps the "hooked chatter" then acquires a new set of associated meanings, associated with welcoming infants rather than with signaling the possession of food, by the sort of process that Charles Darwin described (in "The Expression of the Emotions in Man and Animals") for the development of meaningful displays (such as snarls) from fragments of gradually-decontextualized behavior.

An apparent typo in another entry raises expectations of greater interest. The "Type E Chirp" is a "general alarm" associated "[t]o sudden visual and acoustic stimuli. To sudden leaping movement by group members if animal startled." In contrast, the "Type A Chirp" is said to be used "During mobbing behavior, to sudden animated stimuli. By some groups to preferred foods. Rarely given to acoustical stimuli."

The "Type E chirp Chatter" is a pulsed form whose usage the cited page describes as "Same as for Type A Chirp only when animal is more highly aroused." That would be neat -- replication shifts the meaning from the expected intensified form of Type E to an intensified form of Type A? That's the kind of unexpected little irregularity that happens in human morphology all the time. Alas, I'm afraid it's just a typo.

There are also some "combination calls", such as the "Type F Chirp + Whistle", said to be used "By individuals less confident than when giving Normal Long Calls. As response to Combination or Normal Long Calls or non-group Type F Chirps. In isolation." The associated sound files suggest that this is the same as a "Normal Long Call" with a "Type F Chirp" substituted for the first of two (or more) rising "whistles". This is interesting and suggestive.

[Full disclosure: Klaus Zuberbühler got his PhD in 1998 from the University of Pennsylvania, where I teach, and I had a very high opinion of him while he was a student, which has been maintained by his subsequent work.]

Posted by Mark Liberman at May 28, 2006 08:54 AM