Changes from the expected life course can arise from at least three different reasons in longitudinal work: developmental novelty, errors in diagnosis, and errors in expectation. This longitudinal study examined the social development of 220 fourth graders and 475 seventh graders over a period of 10 years with over 98 percent of the original subjects providing returns in the last year. Embedded in the original sample were students identified as being at high risk by being nominated by teachers and principals for aggressive behavior. These high risk subjects were matched with control subjects. It was expected that it would be possible to identify factors that led to changes in trajectory for the students, for good and for ill. At the conclusion of the study the subjects were 17 to 21 years of age. Although changes in behavior were expected, stability was found. The simple procedure of asking teachers and principals to identify the most aggressive students provided robust prediction for later adaptation beyond the school, even for measures of life and death. However, most of the true-positive high risk subjects showed major deflections from trajectory, even though it was rare these changes were maintained. When changes endured it tended to be the result of multiple changes in the circumstances of the child's life. (ABL)