23

This vocalization is a non-specific signal reflecting bird A's internal state (i.e., high levels of arousal), which is usually associated with the appearance of predators. This contingency is responsible for the call evoking an escape response from bird B.

The alternative cognitive account could be outlined as follows: bird A has a theory of mind and wishes bird B to believe that a predator is approaching. To create this belief in bird B, bird A intentionally produces a highly-specific referential alarm call. This evokes a mental representation of an approaching predator in bird B, which, in turn, mediates the choice of escape behaviour. Bird B elects to flee to cover, which is the most appropriate response for avoiding an aerial predator.

Both of these models describe signalling that is deceptive in the functional sense, and if we were concerned solely with the fitness consequences of behaviour, we would have no need to choose between them. Only the cognitive account would constitute deceptive behaviour in the vernacular sense. Note that discriminating between these alternatives depends heavily upon understanding the type of information encoded in the signal. I do not believe that a convincing case can be made for deception in the cognitive sense if the signal does not meet the criteria for functional reference, because it would then be simply an accurate commentary on the internal state of the sender. Other components of the cognitive account may be less important and intermediate positions are surely possible. For example, we might choose to define cognitive deception in such a way that call production need not be intentional. A decision on this point will depend upon a more detailed understanding of the nature of audience effects and of whether or not they reflect volitional signalling. We may similarly be able to dispense with the requirement that birds have a 'theory of mind', or perhaps be able to take a neutral position on this issue, because it is likely to be intractable to experimental investigation.

The issue of call meaning, however, remains central. In many recent accounts of deceptive signalling, it is not clear whether the authors intend the functional or the cognitive sense, although reviews often assume the latter (e.g., Ristau, 1991; Griffin, 1992; Gould and Gould, 1994). There are typically not sufficient data to make a case for referential signalling, so the functional interpretation might be preferred on grounds of parsimony. For example, in Møller's analysis of alarm calling by great tits (Møller, 1988), and Munn's description of alarm calling in shrikes (Munn, 1986a,b), we do not have the necessary information about production and perception to establish call meaning. This is not a serious deficiency in either account, because the authors are principally concerned with the functional consequences of signalling.

XIc. Some Logical Difficulties In The Interpretation Of Apparent Deception.
Descriptions of food calling in chickens (Gyger and Marler, 1988) provide an intriguing potential case of active deception. Cocks were reported to food call while holding inedible objects, and to do so preferentially when hens were distant and hence less likely to detect the prevarication. Laboratory studies analysing the conditions of food call production have demonstrated that, at least under controlled conditions, this vocalization is elicited specifically by food (Evans and Marler 1994). Playback experiments have revealed that food calls are sufficient to elicit anticipatory feeding movements. Food calls thus have properties that satisfy both production and perception criteria for functional reference. Are we then required to conclude that selective calling by chickens provide an unambiguous example of deception in the cognitive sense?

There are both technical and logical deficiencies in the Gyger and Marler (1988) analysis that suggest such an interpretation would be premature. The data were obtained from opportunistic recordings of five cocks maintained in large pens. Statistical analyses were then performed with degrees of freedom determined by the total number of calling bouts (108) rather than the number of subjects (5). This is an example of the 'pooling fallacy' (Machlis et.al., 1985;