ABSTRACT
In some ways, risk is an ideal subject for research. It can be defined in terms of probability, so it invites statistical analyses of large-scale studies. Most of these studies have been done in the USA. They ask questions about the relationship between mental disorder and violence, and about the ability of psychiatrists to predict violence. In other respects, research on risk has been problematic. Science should proceed from observation to the design of experiments, but this is rarely possible when dealing with serious risk, so we are often stuck at the level of observation. We cannot easily proceed to testing hypotheses. Furthermore, even at the descriptive level, the large statistical studies have serious limitations, both theoretical and practical.
