The literature on the effect of exposure to media violence (including exposure to violent pornography) on aggressive behavior is critically reviewed. Evidence and theoretical arguments regarding short-term and long-term effects are discussed. Three points are emphasized: 1. Exposure to violence in laboratory and field experiments is as likely to affect nonaggressive antisocial behavior as it does aggressive behavior. The pattern is consistent with a sponsor effect rather than a modeling effect: an experimenter who shows violent films creates a permissive atmosphere; 2. the message that is learned from the media about when it is legitimate to use violence is not much different from the message learned from other sources, with the exception that illegitimate violence is more likely to be punished in media presentations; 3. the fact that violent criminals tend to be versatile—they commit nonviolent crimes as well—is inconsistent with explanations that emphasize proviolence socialization (from the media or other sources). I conclude that exposure to television violence probably does have a small effect on violent behavior for some viewers, possibly because the media directs viewer's attention to novel forms of violent behavior that they would not otherwise consider.