A new look at moderator variables of agreement: The role of target standing

Date of Completion

January 1998


Psychology, Social|Psychology, Personality




In three studies the moderator variable approach to agreement was examined. Four potential moderators were considered: behavioral consistency, observability, confirmability, and social desirability. Study 1 explored how the moderators vary by target standing--the extent to which a target possesses a trait--for different traits. Using 108 perceivers, the moderators were rated for 400 hypothetical targets depicted for 80 traits and 5 target standings (e.g., extremely honest or not very deceitful). Analyses revealed that the moderators varied as a complex function of target standing and that function varied by trait. These results imply problems in the measurement of moderators in the previous literature. A new method for measuring moderators based on target standing was proposed. Using this new method, Studies 2 and 3 examined the moderators at low and at high acquaintance by both within-trait and between-trait analyses. In Study 2 (80 perceivers and 32 targets) and in Study 3 (160 perceivers and 40 targets), each target was judged by 20 perceivers on 20 trait dimensions. Agreement was measured by dissensus indexed by standard deviation of ratings, and by consensus estimated by target variance in the Social Relations Model (SRM). Study 3 also examined potential differences between hypothetical and real targets in judging the moderators. Analyses revealed that target extremity had a strong effect on agreement and effects of the moderators were generally stronger at high than at low acquaintance. There was evidence that both consistency and social desirability moderate dissensus, regardless of the levels of acquaintance. Observability at the between-trait level had a reliable effect on consensus, but only at low acquaintance. It was also found that dissensus is not the opposite of consensus, but is formed by a combination of the two variances of the SRM, perceiver and relationship. Further analyses revealed that the moderators had stronger effect on the perceiver than on the relationship variance at low acquaintance and vice versa at high acquaintance. There was strong, converging evidence that conventional analyses are inappropriate for tests of the moderators, and that the new method based on hypothetical targets is valid and appropriate. ^