Measured the relative contribution of multiple variables toward the pr
ediction of recidivism in 129 male juvenile delinquents committed to a
state training school. We identified those classes of variables (as o
pposed to individual variables) that best predict repeat offending. Us
ing factor analysis, we classified the variables empirically and ident
ified four factors: Institutional Adjustment antisocial Behavior, inte
llectual Achievement, and Psychological Distress. With stepwise discri
minant function analysis (using the factor scores for each subject), w
e determined the equation that best delineated group membership, one t
hat correctly classified 73.4% of recidivists and 65.6% of nonrecidivi
sts. Results indicate that, in attempting to predict recidivism, the i
nvestigation of classes of variables rather than individual measures c
an provide a meaningful and effective approach to the problem.