Several studies have shown that behavior modification procedures aimed at a single behavior may have effects on other, nonmanipulated behaviors. A young retarded boy engaged in excessive conversation with his preschool teacher. The teacher then began to ignore his initiated conversation during free-play periods, and it decreased. In addition, (1) social behavior relative to children increased, and (2) use of girls' toys decreased during free play. Also, (3) appropriate behavior at group academics declined, while (4) disruptions rose. In a second study, the teacher alternated conditions of praise and ignoring for talking with children. Talking with children varied accordingly. In addition, use of girls' toys and group academics disruptions rose during the ignoring condition. Appropriate behaviors dropped. Lastly, a timeout procedure was used to eliminate the undesirable side effects of disruptions and of use of girls' toys. Apparently, a response class may have member behaviors that covary directly and/or inversely. Some covariations may be socially desirable, others undesirable. The appearance of undesirable "side effects" can be controlled using behavior modification techniques.