Why Do Judgments on Different Person-Descriptive Attributes Correlate With One Another? A Conceptual Analysis With Relevance for Most Psychometric Research

Abstract

Patterns of correlations among judgments of targets on different attributes are the basis for common psychometric procedures such as factor analysis and network modeling. The outcomes of such analyses may shape the images (i.e., theories) that we as scientists have of the phenomena that we study. However, key conceptual issues tend to be overlooked in these analyses, which is especially problematic when the items are descriptions expressed in the natural language. A correlation between judgments on two such attributes may reflect the influences of (a) a common substantive cause, (b) substantive target characteristics on another, (c) semantic redundancy, (d) the perceivers' attitudes toward the targets, (e) the perceivers' formal response styles, or (f) any mixture of these. We present a conceptual framework integrating all of these mechanisms and use it to connect formerly unrelated strands of theorizing with one another. A lack of awareness regarding the complexity involved may compromise the validity of interpretations of psychometric analyses. We also review the effectiveness of a broad range of solutions that have been proposed for dealing with the various influences, and provide recommendations for future research.