The aim of this project is to dramatically improve the quality of immersive social virtual environments (IVEs). By 'quality' we refer to the response of participants to virtual social situations, in particular the extent to which they respond realistically to what they perceive. By 'response' we mean at every measurable level, ranging from non-conscious physiological processes (such as changes in electrodermal activity or heart rate variability) through to overt behavioural, emotional and cognitive responses / including what they report in interviews about their subjective state of mind. By social IVEs we specifically refer to applications where one or more human participants interact with virtual humanoid characters (avatars) in a socially defined context. Specifically, our objectives are to improve the visual appearance of interactive characters and their interactive behaviour especially so that their behaviour responds appropriately to the behaviour of the participants. Third, social IVEs will be constructed that are well-studied in the social psychology literature, and which are of great societal importance, referred to as bystander behaviour in violent emergencies such as in street violence. Finally, these virtual social situations will be used in a series of experimental studies in order to test whether indeed these objectives do improve the quality of response of participants within these social IVEs.One of our goals is to exploit our research in socially useful applications, and thereby also contribute to the growing body of research that uses VEs as a laboratory for social psychological research. In particular we consider the research program of Levine and colleagues at Lancaster on bystander behaviour in violent emergencies. This research program revisits the classic 'bystander effect' in social psychology. The bystander effect suggests that the more witnesses there are to an emergency, the less likely an individual bystander is to intervene. This phenomenon was identified as a consequence of the apparent inaction of 38 witnesses to the brutal rape and murder of Kitty Genovese in New York in 1964. The bystander effect is one of the most robust and reproduced effects in social psychology. However, it lacks practical utility, since for ethical and practical reasons it is not possible to study it scientifically under controlled conditions. In this project we aim to study the bystander effect in the context of virtual environments, where other work has shown that people do tend to respond realistically to virtual social situations.