26

information for companions to select appropriate responses (Macedonia and Polak, 1989; Macedonia, 1990; Macedonia and Yount, 1991; Pereira and Macedonia, 1991). Ruffed lemurs react to approaching predators with a graded series of calls which, like the alarm calls of ground squirrels and marmots, have only a probabilistic association with predator class. Consistent with these studies of call usage, playback experiments with ruffed lemurs do not reveal qualitatively different responses to different types of alarm call.

Ring-tailed lemurs share with vervet monkeys the traits of living in an open habitat and of spending a great deal of time on the ground. The transition from an arboreal to a largely terrestrial existence probably produced a substantial increase in predation risk. This factor is, however, clearly not sufficient to account for the characteristics of the ring-tailed lemur alarm call system, because there are many other species of small mammals (e.g., sciurid rodents) that are also exposed to frequent attacks from terrestrial predators but have rather less specific alarm calls. It has been suggested that one selection pressure for the evolution of referential alarm calls may have been the use of qualitatively-distinct and incompatible strategies for avoiding different classes of predator (Macedonia and Evans, 1993). Squirrels avoid both coyotes and hawks by running toward a burrow, and their alarm call system seems to provide information that allows receivers to assess the time available for such a response. Vervets and ring-tailed lemurs have the shared characteristic that responses to avian predators involves movement into dense cover, while the safest refuge from ground predators is the outermost branches of trees (a location where vulnerability to attack from raptors is actually increased). Such ecological factors may have played a role in the evolution of alarm call systems that designate predator class (Macedonia and Evans, 1993).

XIIb. Habitat Characteristics And Predator Recognition
Differences in ecology have also been implicated as one determinant of signal specificity. Studies of alarm call production under naturalistic conditions suggest that chickens respond to a fairly broad array of overhead objects (e.g., Gyger et al., 1987). Subsequent laboratory studies revealed that the size and speed of a simulated predator (Evans et al., 1993b) play an important role and that there is also a degree of sensitivity to shape and spatial location. Alarm call production in chickens is nevertheless substantially less highly-specific than that of some other birds. For example, studies of alarm calling by lapwings (Walters, 1990) demonstrate that both South American and African species make subtle discriminations between raptors that are visually very similar. Lapwings inhabit open terrain and this probably affords them the opportunity to examine approaching raptors for some time before producing an alarm. In contrast, red junglefowl, which are the ancestral species for domesticated chickens (Fumihito et al., 1994), live in forest and dense brush where visibility is typically limited (Collias and Collias, 1967) and the time available for responding to potential avian predators will consequently be brief. This ecological constraint may have selected for a simple rule of thumb for predator recognition which, although it entails some loss in accuracy, facilitates rapid response. Other reports describing the anti-predator behaviour of forest-dwelling birds also suggest a fairly high frequency of false alarms (Trail, 1987).

Although the evidence remains fragmentary, the results so far available are consistent with the idea that the specificity of alarm calls has been shaped by habitat characteristics. It is possible that the range of events evoking alarm calls is the product of a trade-off between the rate of Type I errors (i.e., calling when the approaching bird is not dangerous) and Type II errors (i.e., failing to respond to an approaching predator) over evolutionary time (Evans et al., 1993b). Birds living in open habitats may have been subject to selection for low Type I error rates, as the cost of the frequent false alarms that would otherwise be produced in an environment where aerial objects are visible much of the time would be prohibitive. This process would give rise to relatively specific alarm calls. Species inhabiting habitats where visibility is restricted and response times must be short may have been selected to reduce Type