Fig. 1: Testing a one-dimensional hypothesis.

a Depressive and manic surveys were collected bimonthly in individuals with BP. Mood is highly variable and only a limited amount of data is available on each individual, so patient-level hypotheses cannot be formally tested from visual inspection of mood course alone. b Manic scores were plotted against concurrent depression scores. If mood is one-dimensional, concurrent manic and depressive symptoms for all patients would have been plotted near a one-dimensional curve and depressive symptoms would never arise with manic symptoms, indicated by negative correlation between the two symptom types. c Average risk for mania (ASRM score ≥ 6) across individuals was as high as 15% when depressed (PHQ9 score ≥ 10), close to the 23% risk for mania when not depressed. d Rank correlation was measured in 178 patients to evaluate the degree to which manic symptoms were negatively correlated with depressive symptoms. Persons 1–3 were chosen to illustrate the range of rank correlations from negative to near-zero to positive