SIPS 2026 DC

Psychological science relies heavily on scale scores for diagnosis, prediction, and evaluating intervention effectiveness. Yet scale scores often aggregate items that reflect multiple, and sometimes weakly related, constructs. Drawing on validation work with the Adolescent Attachment Anxiety and Avoidance Inventory (AAAAI), this talk examines what is gained and lost when item-level information is reduced to a scale score. Scale scores are contrasted with the shared variance of underlying latent factor structure, illustrating how theoretically distinct dimensions can be “smushed” into one number. At the same time, different populations (such as clinical and non-clinical) distributions overlap, such that a proposed cutoff value trades misclassification between groups. This multi-dataset analysis invites closer scrutiny of routine scale score practices by using a concrete example to show how assumptions involving latent homogeneity and cutoff values are embedded in a single scale score.