14

the predator and calling may continue for some tens of seconds or even minutes after the predator has disappeared (Fig. 2 in Evans et al., 1993a). Long bouts of ground alarm calling are associated with increased motor activity, with the bird typically walking up and down vigorously in an erect posture (Table I).

In summary, both acoustic structure and the time-course of signalling suggest that aerial alarm calls are designed to allow the sender to remain cryptic, while ground alarm calls likely make the sender more conspicuous (Evans et al., 1993a). There are also reliable differences in the effects of an audience on these two call types. Aerial alarm calling is potentiated by the presence of conspecifics, regardless of age and sex (Karakashian et al., 1988), but there is no such audience effect on the production of ground alarm calls (Table I). Isolated males confronted with a terrestrial predator call at a rate indistinguishable from that of males with social companions (Evans and Marler, in prep).

It seems unlikely that considerations of call meaning will be sufficient to explain the signal characteristics and other behaviour summarized above. Certainly the continued production of ground alarm calls, long after a predator stimulus had disappeared, grossly exceeds the level of signalling required to warn social companions. This may be an example of tonic communication (Schleidt, 1973; Owings et al., 1986), in which signals function to maintain a state of vigilance appropriate to a predator that, although no longer visible, may still be nearby - as is typically the case with slow-moving carnivores. It is also possible that the overall pattern of results, incorporating differences in signal structure, time-course, and sensitivity to social context, reflects signalling to different potential receivers. Aerial alarm calls are ideally structured to alert flock members foraging nearby. The structure of ground alarm calls and the duration of calling are both consistent with the idea that although these signals are clearly salient to companions, they may also be designed to deter potential predators (e.g., Klump and Shalter, 1984; Caro, 1986a,b; Hasson, 1991; Caro et al., 1995). Assessing this idea will require quantitative estimates of the active space and localizability of alarm calls, not only for conspecifics but also for representative predators(e.g., Klump and Shalter, 1984). Observational studies of wild or feral populations will be necessary to determine whether animals engaging in ground alarm calling and associated conspicuous movements are indeed less likely to be attacked. Work of this kind has the potential to explain aspects of referential signal design that have so far been neglected.

IX.

THE ROLE OF CONTEXT

IXa. Signal Production: Audience Effects
Classical treatments of animal signalling behaviour tend to assume that call production is essentially reflexive (e.g., Lyons, 1972). That is, that when a sufficient stimulus (e.g., a predator model or a food item) is presented, then the appropriate call, together with other responses (anti-predator behaviour or feeding) will necessarily be evoked. Recent work on a taxonomically-diverse array of species demonstrates that such simple models are inadequate. Ground squirrels (Sherman, 1977; Owings et al., 1986), marmots (Blumstein et al., in press), vervet monkeys (Cheney and Seyfarth, 1985), downy woodpeckers (Sullivan, 1985), and chickens (Marler et al., 1986b; Karakashian et al., 1988; Evans and Marler, 1991, 1992, 1994) all modulate their vocal behaviour according to social context.

In some systems at least, these 'audience effects' appear to be specific to vocal behaviour. Chickens respond with anti-predator behaviour when presented with simple, hawk-shaped models of the kind used in the classic experiments of Lorenz and Tinbergen (Tinbergen, 1948), or computer-generated animations of aerial predators (Evans and Marler, 1992; Evans et al., 1993a,b), and this has allowed exploration of the effects of social context under controlled conditions. Alarm calls are produced by cocks when there are conspecifics present, but not