This paper presents the results of a two year follow-up of a community sample of boys who initially were in grades one, four, or seven (labeled the youngest, middle, and oldest samples, respectively). Initiation in offending was most marked for the youngest sample, escala* Several of the measures used in this study were derived from our earlier work at the Oregon Social Learning Center. We are much indebted to the input from the staff there. In addition, several measures derive from other investigators in the Program on the Causes and Correlates of Delinquency (Terence P. Thornberry, Alan J. Lizotte, Margaret Farnworth, Marvin D. Krohn, and Sung Joon Jang at Albany, NY, and David Huizinga, Finn-Aage Esbensen, and Anne Wylie Weiher at Boulder, CO). Joyce Thompson's help is acknowledged in commenting on an early draft of the paper. Bruce Giroux and Phen Wong effectively assisted in the preparation of the findings. A version of the paper was presented by the second author at the meeting of the American Society of Criminology, Baltimore, Maryland, November 1990. The paper was prepared under Grant No. 86-JN-CX-0009 from the Office of Juvenile Justice and Delinquency Prevention, Office of Justice Programs, United States Department of Justice. Points of view or opinions in this document are those of the authors and do not necessarily represent the official position or policies of the United States Department of Justice. ** RolfLoeber is an Associate Professor of Psychiatry, Psychology, and Epidemiology at the Western Psychiatric Institute and Clinic, School of Medicine, University of Pittsburgh, Pittsburgh, PA. Magda Stouthamer-Loeber is an Assistant Professor of Psychiatry and Psychology at the Western Psychiatric Institute and Clinic, School of Medicine, University of Pittsburgh, Pittsburgh, PA. Welmoet Van Kammen is a Program Coordinator at the Western Psychiatric Institute and Clinic, School of Medicine, University of Pittsburgh, Pittsburgh, PA. David P. Farrington is a Reader in Psychological Criminology at the Institute of Criminology, University of Cambridge, Cambridge, England. CORRELATES OF JUVENILE OFFENDING tion in the seriousness of offending was prominent for the middle and oldest samples, while de-escalation was most prevalent in the oldest sample. The strength of association between the initial and later seriousness of offending appeared to increase with age. We classified offenders according to their pattern of seriousness of offending over time (called a dynamic classification of offenders). Many variables correlated with this measure, showing a covariation with both increases and decreases in the seriousness of offending over time. We noted major shifts in the correlates of offense seriousness between the three age samples-physical aggression and social withdrawal decreasing in strength, while school problem behaviors, peer deviance, and boys' positive attitude to deviancy increased in magnitude. Several factors were associated with the early initiation of offending (before age twelve), including social withdrawal and depression, positive attitude to problem behavior, association with deviant peers, and family problems. In contrast, the later onset of offending (between ages thirteen and fourteen), among other factors, was associated with low school motivation. Correlates of escalation were found for the two older samples but not for the youngest sample, and were particularly prominent in the area of school functioning, disruptive behaviors, positive attitude to deviant behavior, and some aspects of family functioning. Several variables were associated with desistance in offending, including low social withdrawal, low disruptive behavior, and positive motivational and attitudinal factors. The correlates of initiation were distinct from the processes explaining escalation, but were similar to the correlates of desistance. Finally, the paper discusses the relevance of the findings for preventive interventions.