Separating truth from error: A closer look at the effect of grouped versus intermixed questionnaire format

Abstract

A topic of ongoing debate in survey research is whether items measuring different constructs should be grouped by construct or intermixed. Previous researchers have failed to reach a consensus on whether grouping or intermixing measurement items leads to better construct validity. The inconsistency of their findings has to a large extent resulted from a tendency to confound true variance and systematic error variance. Drawing on measurement and cognitive theories, we propose that for established measures with satisfactory psychometric properties, grouped items capture less unsystematic and systematic error variance and more true variance than their intermixed counterparts. This argument is supported by the findings of a field experiment with 853 students in 142 project teams, who reported their teams’ relationship and task conflict on grouped and intermixed items. The theoretical and practical implications of these findings for researchers who use survey instruments are discussed.