As a part of a Copenhagen homicide project, 52 released homicide offenders were followed during a mean period of 9 years. Pessimistic expectations were confirmed as post-release rates of criminality, psychiatric admission, and suicide were all very high. By logistic regression analyses, young age by release, and long stay in prison were found to increase the risk of further criminality, and previous psychiatric admission was found predictive of future admission. The problems in using logistic regression analysis on variables--all of whom may be viewed as "parallel" indicators of a common tendency to adverse behaviors--are discussed. In conclusion, the results have been used for a tentative division of homicide according to the offender-victim relationship into intra-family homicide and extra-family homicide. In intra-family homicide the most important dimension seems to be the psychic state of the offender, in extra-family homicide more impact should be attributed to the social dimension.