Key points are not available for this paper at this time.
23 US Navy officer raters evaluated the performance of subordinate officers using (a) behaviorally anchored scales, (b) scales containing the same dimensions and definitions but without behavioral anchors, and (c) a series of scales involving trait-oriented dimensions, also without anchors. Comparisons of the formats' psychometric properties indicated that the behaviorally anchored scale format was somewhat superior to the other 2 on 3 of 4 dependent variables (involving estimates of leniency, halo, interrater agreement, and degree of differentiation among ratees). However, the magnitudes of the differences due to formats were small, in no case exceeding more than 5% of the variance on the dependent variable.
Borman et al. (Wed,) studied this question.