A Test of the Measurement Assumptions Underlying Behaviorally Anchored Rating Scales

Abstract

Foundational and outcome assumptions of behaviorally anchored rating scales (BARS) are reviewed. Two studies were then conducted to examine two untested foundational assumptions of BARS. The first study examined the assumption that the critical incident methodology produces behavioral specific incidents with which to anchor BARS. Qualitative and quantitative analyses were conducted. Contrary to this assumption about BARS, results indicated that there was considerable variation in the specificity of the behavioral anchors. The second study examined the assumption that the interpretation of critical incidents is invariant across nonstimulus bound developmental and the actual stimulus bound rating contexts. The assumption was not satisfied. Results indicated that the evaluation of critical incidents was significantly different across rating contexts. Implications are discussed.