Accurate definitions of phenomena are essential to any scientific enterprise. A definition of violence should be fully capable of accounting for the exclusion of behaviors such as accidents and self-defense, and the inclusion of behaviors such as child abuse, sexual offenses, and manslaughter. Violence research has produced numerous and sometimes conflicting definitions of violence that can be organized into 4 general camps: the exemplars approach, the social psychology approach, the public health approach, and the animal research approach. Each approach has strengths and limitations, but to fully distinguish violence from other behaviors requires incorporating elements from all of them. A comprehensive definition of violence includes 4 essential elements: behavior that is (a) intentional, (b) unwanted, (c) nonessential, and (d) harmful. More sophisticated recognition of some elements is needed. For example, shortened telomeres—a known consequence of child abuse—is a far more serious harm than a scratch or bruise that will fully heal in a few days. Many problems in the field are due at least in part to insufficient attention to definitions, such as minimization of sexual violence, bullying, and other behaviors that do not map onto prototypical exemplars. More precise definitions of violence can improve surveillance, promote more accurate identification of causes and consequences, enhance evaluation of treatment outcomes, and guide development of prevention programs, among other benefits.