The Item Context Effect: Do Previously Answered Items Influence Psychometric Properties of Self-Report Personality Scales?

Individual ratings of people, political parties, and objects are contingent on comparisons with standards raters may not be aware of. In this vein, striking evidence from attitude research suggests that previously answered questions affect responses to subsequent attitude questions such that previous questions provide a frame of reference. It stands to reason, however, that previously responded-to items are less influential in self-report assessments of personality because personality is self-referential and assumed to be more stable and consistent than attitudes. Across four experiments (ntotal = 4,895), we tested whether univariate, bivariate, and multivariate properties of the Big Five Inventory-10 and HEX-ACO-18 personality scales differ as a function of the type of items a person has responded to already (i.e., buffer items, items with the same or opposite keying directions, respectively, or no additional items in a control group). Across groups, little variation occurred in the responses and subscale-specific inter-item correlations, and measurement invariance did not hold across conditions. These findings militate against what we refer to as the item context effect for personality assessment and suggest that researchers need not worry too much about the order in which they administer their target measures. Further steps for advancement and refinement are presented.