Neuroticism's prospective association with common mental disorders (CMDs) has fueled the assumption that neuroticism is an independent etiologically informative risk factor. This vulnerability model postulates that neuroticism sets in motion processes that lead to CMDs. However, four other models seek to explain the association, including the spectrum model (manifestations of the same process), common cause model (shared determinants), state and scar models (CMD episode adds temporary/permanent neuroticism). To examine their validity we reviewed literature on confounding, operational overlap, stability and change, determinants, and treatment effects. None of the models is able to account for (virtually) all findings. The state and scar model cannot explain the prospective association. The spectrum model has some relevance, especially for internalizing disorders. Common causes are most important but the vulnerability model cannot be excluded although confounding of the prospective association by baseline symptoms and psychiatric history is substantial. In fact, some of the findings, such as interactions with stress and the small decay of neuroticism's effect over time, are consistent with the vulnerability model. We describe research designs that discriminate the remaining models and plea for deconstruction of neuroticism. Neuroticism is etiologically not informative yet but useful as an efficient marker of non-specified general risk.