The classification of depression is well established in major diagnostic systems and operational definitions now make the diagnosis of depression reasonably reliable. However, classification and diagnosis continue to be based on clinical presentation and course and are not currently informed by aetiological or pathophysiological considerations. It is still unclear, for example, whether or not categories such as melancholic depression represent distinct subforms of illness or whether a dimensional classification based on severity can capture clinical presentation adequately and more economically. Despite these caveats, there is some evidence that distinctions between various subforms of depression can predict treatment response and this can be useful for clinical practice as well as guiding research strategies.