Previous research failed to uncover a replicable structure of dimensions or subtypes underlying the symptoms of depression. One reason might be that research failed to separate co-variation between symptoms due to overall depression severity vs. due to specific symptom profiles.

The study tested the hypothesis that a replicable dimensional structure of depression would be uncovered when depression severity is eliminated from symptom scores. Additionally, the study explored differences in the dimensional structure in general population vs. depressed people-only samples.

The cohort study on substance use risk factors (C-SURF), a large cohort of young Swiss men, and young men from the national health and nutrition survey in the US (NHANES 2009-2012) were analyzed. DSM-IV symptoms of depression were assessed via the Major Depressive Inventory (WHO-MDI) in C-SURF and via the Patient Health Questionnaire 9 (PHQ-9) in NHANES. Dimensionality was examined using principal component analysis in full samples vs. samples of participants with a current depressive episode for raw vs. severity-adjusted symptom scores.

When using severity-adjusted symptom scores, correlations between depressive symptoms largely disappeared and there were no replicable dimensions. When using raw scores in the full samples, one single dimension of depression consistently emerged. When using raw scores in depressed participants, only rudiments of dimensions were found across samples.

It is unlikely that there are stable dimensions underlying the DSM-IV symptoms of depression. The set of symptoms capture the disorder in the general population, but the disorder's manifestation is highly individual.