The effect of affect in performance appraisal
The present investigation included three studies of the effect of liking upon the differential accuracy of performance ratings of instructors. Likableness and performance level were manipulated within vignettes. Liking was manipulated with trait terms that were found in a pretest to have little implication for performance on the rating dimensions. Instructor performance level was manipulated with incidents of teacher behaviors.
The first study investigated the effects of ratee likableness (liked vs. disliked vs. neutral) and ratee performance (high vs. low) as well as the effects of rater sex, rater selective attention ability (high vs. low), and the memory demand of the rating task (memory vs. no memory) on the differential accuracy of ratings. A total of 288 subjects, 144 males and 144 females, rated the performance of four instructor vignettes.
The differential accuracy of ratings was analyzed as a 3x2x2x2x2 between-subjects design. Differential accuracy was found to be significantly influenced by the performance level of raters, the selective attention of raters, and the memory demand of the rating task. The factors of likableness, ratee performance, and rater sex jointly influenced differential accuracy.
The second study investigated the effect of order of format presentation (before vs. after ratee observation) as well as ratee likableness and performance on rating accuracy. A 3x2x2 (liking x performance x format) between-subject ANOVA on the differential accuracy of ratings provided by 144 male and female subjects revealed the three factors to have a joint influence-on rating accuracy. Ratings were more accurate in the format before than in the format after condition with liking and performance having an interactive effect only in the after condition.
Study 3 investigated the operation of schemata, discounting, and recall bias as well as liking and performance on rating accuracy. A regression analysis revealed the recall bias measure to be the only significant predictor of rating accuracy.
It was concluded that the pattern of results across the three studies did not offer direct support for either an integrality or schema conceptualization of affect. Consideration of schematic processing offered a potential explanation of the recall bias finding. Applied implications of the results are discussed.