Covariation within behavior repertoires of problem children were examined. Two boys, referred for psychological help, were observed both at school and at home for about 3 yr. A coded observation system permitted scoring of 19 child‐behavior categories and six social‐environment categories. After a two‐month baseline, behavior categories were intercorrelated, demonstrating that each child showed a group of behaviors that covaried. These groupings were specific to the home and school settings. Contingency management procedures were then applied to each child's problem behaviors inonesetting. Next, a reversal phase was instituted, followed by resumption of the initial contingency management phase. These three phases lasted seven months, until the end of the children's public school terms. Results showed that the baseline group of covarying behaviors continued to covary over the three experimental phases. The children then entered a remedial education setting for three summer months, and then returned to schools and were observed in follow‐up for 2 yr. The baseline group of behaviors continued to covary during both phases. The behavior covariations could not be accounted for on the basis of temporal relationships between the behaviors and social environment categories. Although no behavior covariations extended across either child's home and school settings, contingency management procedures produced across‐settin