Center for Adaptive Behavior and Cognition
Max-Planck-Institut für Bildungsforschung
Lentzeallee 94, 14195 Berlin, Germany
noble@mpib-berlin.mpg.de

Human language may or may not be continuous with animal communication. If there is a continuity, and language is an elaborated version of an earlier animal signalling system, then clearly language-evolution theorists should learn as much as they can about how animal communication evolves. If there is no continuity, and human language turns out to be an adaptation with a unique genesis and structure, then theories of language evolution will still be constrained by general results from evolutionary theory, such as Hamilton's (1964) findings on altruistic behaviour towards kin, or the possibility that the cost of a signal can ensure its reliability (Zahavi, 1975; Grafen, 1990). So ideas on animal communication will have at least some relevance to theories of language evolution.

However, animal communication is a broad subject. Animals influence each other's behaviour in many different ways, and pinning down exactly which kinds of influence that we wish to call communication can be troublesome. It is uncontroversial to say that vervet monkeys are communicating when one gives a leopard alarm and the others scramble for the safety of the trees. But is a camouflaged insect signalling to its predators? By running away, is an antelope signalling to a cheetah? In both cases the answer is yes under certain definitions of communication that have been adopted in the biological literature. Intuitions differ about how such borderline cases should be treated; mimicry and deception are two other notable problem areas.

This definitional problem is despite the fact that in ordinary language we have a clear idea of what we mean by communication, or at least an archetypal image: a sender imparts information to a receiver via some sort of signalling channel. For example, one person says truthfully to another, "It's raining outside." This has been dubbed the conduit metaphor (Reddy, 1979; Lakoff & Johnson, 1980).

The ordinary language definition has not been sufficient for most biologists who have looked at communication. I will consider three types of definition from the biology literature: those phrased in terms of behavioural influence, those dealing with information transmission, and those invoking the intent to communicate. All three of these categories will ultimately be rejected as inadequate.

An example of the first type of definition is Krebs and Dawkins (1984), who defined a "signal" as an action or structure which increases the fitness of an individual by altering the behaviour of other organisms. The main problem with such a definition is over-inclusiveness: one animal directly causing a behaviour in another, for instance by attacking it, would count as communication. So might the incidental transfer of information: it might well be to the advantage of bird A that bird B should notice that it (A) has found food, perhaps because the two are related. However, that would mean that A's behaviour in simply observing food and approaching it counts as a signal to B; this is certainly a counter-intuitive result.

Hurd (1997) provides a good example of communication defined in informational terms: "Information is said to be received whenever an agent changes it's [sic] expectations about the consequences of an action, and communication has occurred whenever the action of one animal transmits information to another." One problem is that this is still over-inclusive, because incidental information transfer qualifies as communication. A deeper difficulty for such a definition is the problem of error: suppose we have a working hypothesis that a certain bark is used by vervet monkeys to mean "leopard approaching!". One day we observe the bark being made when a hyena approaches. We are then faced with a problem of indeterminate content, which we cannot evade by stipulating that the bark means leopard and that all other uses of it are mistakes, because that would be begging the question.

Grice (1969) put forward the case – later developed by Bennett (1976) and Dennett (1987) – that considering intentionality allows us to pick out a special kind of communication that is genuinely worthy of the name. These authors rely on the intuition that there is a difference worth marking between a situation in which causal automatons exchange signals, and a communication system in which participants really mean what they say. Their argument is that real communication can be roughly equated with human speech acts, and must involve, at a minimum, third-order intentionality. Two problems exist for this definition: first, even if it could be demonstrated that an apparently communicative system involved mere zero-order tropisms, we may well still want to classify the system as communicative. Second, a definition of communication in terms of higher-order intentionality is founded on the dubious premise that a particular animal either unambiguously does or unambiguously does not possess such intentional capabilities.

After consideration of the alternatives, I will defend an evolutionary-functional definition of communication based on the work of Millikan (1984, 1993). This definition stipulates that true communication, or "proper signalling", involves the production of a signal and the performance of a response that both have a history of selection in this context. We should expect to find proper signals when it is evolutionarily stable for two animals to coordinate their behaviour in an interaction, that is, when there is a mutual benefit in transmitting information. However, this does not mean that proper signals will only evolve in nakedly cooperative situations. Processes such as the handicap principle show that communication can be evolutionarily stable despite an apparent conflict of interests: poor-quality signallers at a handicap equilibrium do not honestly signal their low quality for any mystical reason, but because the excessive costs of exaggeration make it in their interests to do so. It is (apparently) the function of the peacock's tail to signal male quality just as much as it is the function of the bee dance to indicate the location of nectar.

The practical consequences of this definition will be explored. The most important implication is that those sections of the biological literature on animal communication that do not deal with cases of proper signalling will probably not be relevant to work on the evolution of human language.

 Conference site: http://www.infres.enst.fr/confs/evolang/