Response-Shift Bias

Abstract

Evaluations of experimental interventions which employ self-report measures are subject to an instrumentation-related source of contamination known as response-shift bias. The difficulty arises when the experimental intervention changes the subject's evaluation standard with regard to the dimension measured with the self-report instrument. In such cases even the true experimental designs (Designs 4, 5, and 6; Campbell and Stanlev, 1963) can provide internally invalid results. Retrospective pretest ratings are recommended as one way in which response-shift bias might be attenuated. Research demonstrating re sponse-shift effects and the superiority of retrospective ratings over tradittonal self-report pretest ratings in providing a measure of change is reviewed. Finally, the current status of retrospection in psychological research is reviewed, and issues are considered for future research needed to identify the unique strengths and limitations of retrospective approaches.