The primary versus secondary distinction is often used as a way of subtyping depression. Its applicability to bipolar disorders has been unclear. This report examines the relative primacy of the bipolar versus unipolar distinction as compared to the primary versus secondary distinction in a sample of 955 patients in the NIMH Collaborative Study of the Psychobiology of Depression. These patients are divided into nine groups of the basis of whether they are bipolar I, bipolar II, or unipolar, and whether they are primary, 'pure' secondary, or 'complicated' secondary (i.e. bipolar I primary, bipolar I pure secondary, bipolar I complicated, etc.). Three sets of variables are used to determine the predictive validity of these various subtypes: data concerning age of onset and phenomenology of current episode, outcome, and familial prevalence. In general, data from these three sets of validators suggest that the bipolar distinction takes precedence over the primary versus secondary distinction. Within bipolars, there is little value in further subtyping into primary versus secondary.