TESTING THE PREDATION HYPOTHESIS FOR VERTEBRATE SOCIALITY - PROSPECTSAND PITFALLS

Primates are in some ways excellent subjects for studying the impact o f predation on prey. They are generally easy to watch and identify as individuals, so that long-term tracking of both death rates and anti-p redator behaviors is possible, as amply shown by many of the studies i n this volume. On the flip side, their low predation rates and large g roup sizes require very large total sample sizes for statistically pow erful tests of the direct effects of sociality on predation rates. To study the indirect effects of predation on primate behavior requires d efining the intrinsic predation risk they experience, that is the expe cted rate of predation they would suffer under standardized levels of anti-predator. behavior (possibly none - see Hill & Dunbar, this volum e). This abstract variable can be assessed qualitatively across differ ent conditions by reference to modeling or common sense, or quantitati vely by analyzing the hunting success of the predator independent of t he prey's behavior (Cowlishaw, 1997). Great care must be taken in inte rpreting the behavioral responses of animals to different levels of pr edation risk when a given behavior can, serve multiple functions, such as is the case with vigilance. Furthermore, most anti-predator behavi ors carry fitness costs, not only from the lost opportunity to perform other fitness-enhancing activities, but even in terms of predation it self - apparently some primate species, benefit from living in small g rbups which are very difficult for predators to detect instead of usin g a large-group early-warning defense as postulated in many theoretica l models. Such costs will limit the extent to which primates are able to reduce their intrinsic predation risk (Fig. 1). Although comparison s of predation rates or anti-predator behaviors across species or popu lations may be very revealing, there are some potential problems to co ntend with. First is the widely-recognized problem of analyzing phylog enetically-structured data (cf: Hill & Dunbar, this volume), which req uires robust and detailed phylogenies and requires a long list of assu mptions to make the results interpretable statistically (e.g, see Garl and et al., 1992). Second, when performing any type of multiple regres sion to tease apart the confounding effects of correlated variables on the dependent variable, the exact results may depend on the particula r set of independent variables examined. Thus, conclusions from such a nalyses should always be treated as tentative. Finally, predicted ecol ogical responses of prey group size to changing predator density may m imic expected evolutionary changes. Thus, tests of evolutionary predic tions with comparative ecological data need to be sensitive to the pos sibility that the observed differences may not be caused by evolutiona ry responses and hence may not qualify as adaptations.