Since Kraepelin grouped affective disorders under the title of 'manic-depressive insanity', there has been controversy over whether the bipolar and unipolar entities within this are distinct affective disorders or whether they are merely two ends of an affective continuum. In order to bring some clarity and goal-posts to this argument, we define the criteria that must be fulfilled by diseases in order to be considered as part of a spectrum. We analyse bipolar disorder and major depressive disorder with respect to these criteria and find the model fits in many respects but fails to account for either the poor correlation in severity of manic and depressive symptoms or for the apparent discontinuity in the distribution of familial mania. A one-dimensional spectrum is thus too simple and a two-dimensional approach is required; this also fits much better with our current understanding of the genetic picture.