Latent risk situations are always present in society. General information an these risk situations is supposed to be received differently by different groups of people in the population. In the aftermath of specific accidents different groups presumably have need of specific information about how to act to survive, to avoid injuries, to fmd more information, to obtain facts about the accidents etc. As targets for information these different groups could be defined in different ways. The conventional way is to divide the population according to demographic variables, such as age, sex, occupation etc. Another way would be to structure the population according to dependent Variables measured in different studies. They may concern risk perception, emotional reactions, specific technical knowledge of the accidents, and belief in the information sources. One procedure for forming such groupings of people into homogeneous clusters would be by statistical clustering methods on dependent variables. Examples of such clustering procedures are presented and discussed. Data are from a Norwegian study on the perception of radiation from nuclear accidents and other radiation sources. Speculations are made on different risk information strategies. Elements of a research programme are proposed.