Abstract :
In performance appraisal, the halo-accuracy paradox describes the surprising result that rater accuracy can be positively correlated with the halo rating error. Fisicaro (1988) provided an explanation for this unlikely relationship by proposing an inverse V function as the relationship between accuracy and invalid halo in which maximum accuracy is located at zero invalid halo. This paper develops the model by proposing that maximum accuracy does not have to be at zero invalid (Hypothesis 1). As the cognitive difficulty of a rating task increases, a negative monotonie relationship between maximum achievable accuracy and associated value of absolute invalid halo is specified (Hypothesis 2). The hypotheses were tested in two different experimental situations. Results from both studies supported Hypothesis 1 but, whilst a distinct pattern between accuracy and absolute invalid halo was noted, only a weak version of Hypothesis 2 could be supported. The evidence from this paper demonstrates that the halo-accuracy paradox is not an artefact as some recent reviewers have proposed (Balzer & Sulsky, 1992; Murphy & Balzer, 1989; Murphy & Cleveland, 1991).
