The primary objective is to test whether the polarity of recurrence (depressive vs manic or mixed), is related to lithium levels. A total of 86 euthymic bipolar patients (DSM-IV) on lithium monotherapy were prospectively followed up for 2.5 years with regular monitoring of both lithium levels and psychopathology. The last lithium level during the free interval that preceded worsening of affective symptoms was related to polarity of symptoms. To account for effects of major confounders, results were corroborated by multivariate analysis. An intervention for manic or mixed symptomatology was required in 27 patients, whereas 22 patients were treated for depressive symptoms. Average lithium levels preceding reappearance of manic or mixed symptomatology were lower than levels preceding reappearance of depressive symptoms (0.53+/-0.13 vs 0.66+/-0.21 mmol/l, P=0.01). This result was confirmed using logistic regression analyses with type of index episode, diagnostic subtype and residual manic and depressive symptoms as covariates. The results indicate that manic or mixed recurrences might rather occur at lower lithium levels, whereas the depressive pole prevails in the higher range. If substantiated by further studies, this finding might indicate that higher lithium levels are needed to prevent manic episodes than to prevent depressive episodes.