In much of the experimental literature on social learning, the ecological validity of the social context is often sacrificed in the interests of greater control over the material to which the participants are exposed. It is by no means a criticism to draw attention to the tight control involved in these designs, and indeed, such approaches have p proven to be highly successful in recent years, resulting in important advances in both the nonhuman and human social-learning literature.For example, using such methods, Horner and Whiten (2005) were able to show intriguing differences between chimpanzees (Pan troglodytes) and human children in terms of the elements of a demonstration that were copied. Following a demonstration by an adult human experimenter, the chimpanzees omitted actions that app peared not to be causally related to the goal (of extracting a reward from the experimental apparatus). In contrast, young children, 3-4 years of age, copied all the elements of the demonstration, irrespective of how relevant to the goal they appeared to be. Whiten, Horner, Litchfield, and Marshall-Pescini (2004) reviewed numerous studies of social learning in nonhuman primates in which similar methods were used. Children's language learning has also b been studied using similar dyadic experimenter-subject designs. Experiments have shown that young children can readily incorporate novel nonsense words into their vocabulary and use these appropriately, after hearing such words spoken by an adult experimenter (for a review, see Tomasello, 2000). The effects of social influences on adult humans have also been studied in a similar fashion, with manipulations achieved through the actions of experimental confederates. Experiments by Sherif (1936) and Asch (1955) showed that participants asked to make a simple perceptual judgment would alter their responses if these were not consistent with responses given by actors posing as fellow participants in the same experiment. In a more recent study following the same basic logic, Eriksson and Coultas (2009) found that participants' responses to opinion questionnaire items could be influenced by the apparent responses of previous participants (which were actually faked, in order to directly manipulate the magnitude of the apparent majority).As these studies illustrate, controlled manipulations of modeled behaviors can reveal much about what can be transmitted. And yet, when we look at real cultures and traditions-that is, those that occur spontaneously in natural populations-there is an inevitable shift of focus, due to the kind of data involved. Different issues arise as the important questions, and a rather different perspective tends to be taken, with populations, rather than individuals, representing the principal units of study.For example, research on traditions in nonhuman primates began with reports of novel behaviors spreading through populations. Early reports of this kind were given for Japanese macaques, Macaca fuscata , concerning a number of new behaviors, including food wa...