The study focused on 633 cleared burglaries committed by 244 different offenders in the Finnish Metropolitan Area between 1990 and 2001. Each case file was examined for 85 crime-scene behaviors that pertained to the items stolen, search methods, type of house, means of access, and types of tools used. Offender characteristics were divided into those that remained stable from one offense to another and those that might vary between two offenses. In addition to offender characteristics coded from the case files, the criminal histories of the offenders up to the day of the first burglary included in the study were searched. Logistic regression was used in the attempt to predict offender characteristics from crime-scene behaviors. Component scores calculated on the basis of the principal-components analysis of the crime-scene behaviors were used as predictors. The type of burglar characteristics predicted by three independent researchers were as follows: visited the target earlier; lives in the city in which the crime was committed; committed the crime alone; had a "fence" prior to the burglary; was unemployed or in retirement/sickness pension; and had a traffic violation. Statistically significant predictors of almost all offender characteristics were identified, suggesting that predictive models could be used in police burglary investigations to narrow the number of suspects. 5 tables and 57 references