Films such as Natural Born Killers (Oliver Stone, 1994) and Crash (David Cronenberg, 1996) have become famous for their alleged associations with real violence. Politicians, the media and campaign groups apparently believe that these movies are a drag which produces serious side effects. Look at any discussion of these movies in the press and you will find a raging debate on violence in the mass media and violence in our society. There is a basis to this dominant discourse that can be understood by reference to social theories of risk. It is my thesis that risk theories can help us to understand the dynamic of the current debate on media violence. Risk analysis is concerned with invisible risks that are harmful to individual and global environments. The companies who manufacture products that are perceived to have risks refuse to claim responsibility for their actions and, instead, talk of 'acceptable risk levels'. Campaign groups, such as Greenpeace, argue that anecdotal evidence suggests there are no acceptable levels: they reveal the full extent of the real and potential side effects produced by these large industries. Pro-censorship groups and self-appointed moral watchdogs have utilized this dominant discourse to engineer a political debate that they hope will lead to the regulation of media violence. The entertainment industry is presented as a manufacturer of risks. They produce products, for example violent movies, which are perceived to contain side effects that are harmful to individuals and to social environments. The industry may talk of 'acceptable levels', but anecdotal evidence, such as the James Bulger case, suggests that media violence can lead to real violence in our society. Anti-violence campaign groups undertake their own research which claims to measure the side effects of television violence. The entertainment industry is asked to self-regulate, and legislation is called for. My argument is that the debate on media violence has become polarized. Independent researchers need to change the terms of reference in order to alter the dominant discourse surrounding risk theories and the mass media. What is more, researchers must recognize that the effects of watching films or television cannot be measured in the same way the effects of car pollution are measured. Watching films or television programmes is a complex and dynamic process that does not lend itself to scientific measurements. Thus, new methods of research must be utilized in order to break the circularity of the debate on violence and the mass media. One such method is to understand viewer response to risk, and perceptions of the dangers and rewards associated with risk-taking behaviour. John Adams' (1995), 'risk thermostat hypothesis' indicates that people have a propensity to take risks which varies from one individual to another. Recent qualitative research in audience response to media violence also suggests that that is the case. New research methods can seek to understand individual perceptions of risk-taking, and the cultural construction of risk, in relation to the mass media.