The question of indicators has been a longstanding concern for Futuribles, since the robustness of an analysis of a social phenomenon or trend can vary appreciably depending on the data and indicators employed. This has frequently been highlighted with regard to employment figures, for example, the question being whether the rate of unemployment or of employment is the more informative. In this column, the indicators issue arises with regard to the measurement of violence across the world. Marc-Antoine Pérouse de Montclos demonstrates here that, contrary to an image frequently conveyed by the media and originating in the major international organizations, violence hasn’t necessarily been rising over the long term. Once again, it all depends on what the indicators include: the civil and/or military victims of conflict, direct and/or indirect casualties, related to the number of countries (though this has changed over time), to a population that is itself rising, the definition of the notion of armed conflict and protagonists concerned, etc. This column enables us to revise downward the idea that violence was increasing in the world over the long term, to stress the complexity of the measurement of violent phenomena on an international scale and, in that context, the difficulty of research on international conflict and criminality.