Establishing the presence of an epidemic is traditionally a first step in any outbreak investigation. For two reasons, however, this has not been a fruitful approach for suicide cluster investigations. First, the data necessary to statistically verify an excess number of suicidal incidents are often lacking or of poor quality. Second, and more important, when a community perceives that it is experiencing a suicide cluster, it is not immediately relevant whether the cluster is statistically significant. The perception of suicide clustering, and the highly charged emotional atmosphere associated with that perception, may dramatically heighten the potentially "contagious" effect of suicide. That the perception of clustering may itself be a risk factor for suicide distinguishes suicide clusters from all other clusters of fatal disease or illness. A community response plan should, therefore, be implemented to identify and refer persons who may be at high risk of suicide, regardless of whether the community-identified suicide cluster is statistically significant. Statistical techniques may be useful at several stages in the investigation and control of apparent suicide clusters, but statistical verification of a community-identified suicide cluster is not appropriate as a starting point for response to the cluster.