Animal reintroductions and translocations are potentially important interventions to save species from extinction, but most are unsuccessful Mortality due to predation is a principal cause of failure. Animals that have been isolated from predators, either throughout their lifetime or over evolutionary time, may no longer express appropriate antipredator behavior. For this reason, conservation biologists are beginning to include antipredator training in pre-release preparation procedures. We describe the evolutionary and ontogenetic circumstances under which antipredator behavior may degenerate or be lost, and we use principles from learning theory to predict which elements can be enhanced or recovered by training. The empirical literature demonstrates that training can improve antipredator skills, but the effectiveness of such interventions is influenced by a number of constraints. We predict that it will be easier to teach animals to cope with predators if they have experienced ontogenetic isolation than if they have undergone evolutionary isolation. Similarly, animals should learn more easily if they have been evolutionarily isolated from some rather than all predators. Training to a novel predator may be more successful if a species has effective responses to similar predators. In contrast, it may be difficult to teach proper avoidance behavior, or to introduce specialized predator-specific responses, if appropriate motor patterns are not already present. We conclude that pre-release training has the potential to enhance the expression of preexisting antipredator behavior. Potential training techniques involve classical conditioning procedures in which animals learn that model predators are predictors of aversive events. However, wildlife managers should be aware that problems, such as the emergence of inappropriate responses, may arise during such training.