Specifically, the study investigated what determines into which of three categories a person's most serious (as measured by sentence length) conviction during a followup period will fall. The categories are no conviction, conviction for a misdemeanor, and conviction for a felony. This dependent variable was analyzed using the logit model, based on a sample of men released from the North Carolina prison system. Findings show that a person is more likely to be convicted of some crime (misdemeanor or felony) if he is an alcoholic or a user of hard drugs, if he is white, the longer his activities are followed, the more convictions he had before entering prison, and the younger he is upon release. Having more rule violations when incarcerated or being incarcerated for a misdemeanor increases the probability of conviction for a misdemeanor relative to either no conviction or to conviction for a felony. The younger a person when first arrested, the greater is the probability of conviction for a felony relative to no conviction or a misdemeanor. Those who participate in the prisoner work release program or are supervised when released are less likely to be convicted for a felony compared to a misdemeanor. Predicted probabilities of various types of convictions show the strong influence of personal characteristics on the seriousness of postrelease criminal activity as well as the type of predictions possible with the estimated logit model. This model can enrich criminal justice evaluations by permitting the investigation of postrelease criminal activity in its full complexity (frequency and seriousness) rather than merely examining whether or not a person returns to some form of criminal activity. Tabular data, 20 notes, and 18 bibliographic listings are provided. (Author summary modified)