In two experiments the effect of knowledge of the seriousness of an event's outcome on the estimated strength of its causes was studied. In experiment 1 participants (n = 87) were shown a video of a homeless man being evicted from a police station by a police officer, during which the evictee fell. One group of participants were told that the homeless man subsequently died; the other group was told that he was uninjured. Participants who thought the homeless man had died more often blamed the policeman for the man's fall. Estimations of the force of the push were not directly related to the outcome of the event, but rather to the attribution of blame, with those who blamed the officer giving higher estimates of the force of the push. In experiment 2, before watching a revised version of the video, the participants (n = 88) were assigned to one of four groups. Groups A and B were told that the homeless man had died; groups C and D were told that he had survived. In an effort to counter interference of blame attribution, groups A and C were told that the officer was to blame for the man's fall, and groups B and D that the officer was not to blame. The manipulation of guilt was not successful, but this time a significant relation between event outcome and push force was found, with those who thought the homeless man had died giving higher estimates of the force of the push than those who thought that the man had survived. As in experiment 1, those who thought the homeless man had died more often blamed the policeman and those who blamed the policeman again gave higher estimates of the force of the push. It appears then that the more tragic the outcome of a violent incident, the more blame witnesses tend to attribute to the perpetrator, and the more they tend to overestimate the amount of violence involved.