regression toward the mean
the tendency for extremely high or extremely low scores to become more moderate (i.e., closer to the mean) upon retesting over time. In experimental studies, this tendency threatens internal validity in that shifts of scores may occur for reasons unrelated to study manipulations or treatments. For example, regardless of the interventions a researcher is investigating to improve mathematics performance (e.g., extra study sessions, providing positive or negative reinforcement), low scoring students will tend to perform slightly better on the next math exam, whereas high scoring students will tend to perform slightly worse. Random assignment to treatment and control conditions may be used to minimize the influence of regression toward the mean on experimental results. Also called regression effect. See regression artifact.