Background Psychiatry has long sought to develop biological diagnostic subtypes based on symptomatic differences. This effort assumes that symptoms reflect, with good fidelity, underlying etiological processes. We address this question for major depression (MD). Methods We examine, in twins from a population‐based registry, similarity in symptom endorsement in individuals meeting criteria for last‐year MD at separate interview waves and in concordant twin pairs. Among individuals with MD, we explore the impact of genetic‐temperamental and child adversity risk factors on individual reported symptoms. Aggregated criteria do not separate insomnia from hypersomnia, weight gain from loss, etc. while disaggregated criteria do. Results In twins with MD at two different waves, the mean tetrachoric correlations (±SEM) for aggregated and disaggregated DSM‐IV A criteria were, respectively, +0.31 ± 0.06 and +0.34 ± 0.03. In monozygotic (MZ) and dizygotic (DZ) twin pairs concordant for last‐year MD, the mean tetrachoric correlations for aggregated and disaggregated criteria were, respectively, +0.33 ± 0.07 and +0.43 ± 0.04, and +0.05 ± 0.08 and +0.07 ± 0.04. In individuals meeting MD criteria, neuroticism predicted the most MD symptoms (10), followed by childhood sexual abuse (8), low parental warmth (6), and genetic risk (4). Conclusions The correlations for individual depressive symptoms over multiple episodes and within MZ twins concordant for MD are modest suggesting the important role of transient influences. The multidetermination of individual symptoms was further evidenced by their prediction by personality and exposure to early life adversities. The multiple factors influencing symptomatic presentation in MD may contribute to our difficulties in isolating clinical depressive subtypes with distinct pathophysiologies.