Evaluation is a Confound in Personality Measurment: Evidence and Implications

Evaluation is the tendency to endorse items based on their valence, irrespective of their actual descriptive content. Subsequently, this dimension may cluster items of opposite content as long as their perceived desirability is similar. We propose that it is beneficial to isolate evaluative variance from descriptive variance, and that doing so may improve convergent, discriminant, predictive, and, ultimately, construct validity. Study 1 demonstrated that once evaluation was isolated from personality pathology markers, the remaining factors evidenced negative traits at both ends of the poles. Furthermore, the non-evaluative factors displayed more nuanced and sensible patterns of overlap with DSM-III personality disorders. Study 2 demonstrated that once evaluation was isolated from emotion and affective descriptors, items tapping positive affect and approach temperament clustered with items tapping anger. In study 3, we developed a novel approach to measure personality pathology free from evaluation. Study 4a demonstrated that once evaluation was controlled, personality traits evidenced more sensible patterns of overlap with naturalistic behavior. Study 4b demonstrated that once evaluation was isolated, non-evaluative personality traits evidenced more sensible patterns of overlap with self-reported interests and engagement in activities. Study 5 examined whether evaluation suppressed the relation between specific genes and self-reported personality traits. This hypothesis was not supported; before and after isolating evaluation, personality traits accounted for less than one percent of the variance. We conclude by arguing that traits that are unrelated to evaluation better approximate behavioral, biological, and evolutionary data and theory.

Note: Abstract extracted from PDF text