There is abundant evidence for substantial long-term prophylactic efficacy of lithium in bipolar manic-depressive disorders. Interruption of such treatment carries an extraordinarily high risk of recurrence within several months, even after several years of stability. Even a sharp reduction in dose may carry some risk. Gradual discontinuation of lithium was accompanied by markedly reduced risk of early recurrence. There is suggestive evidence that the phenomenon of high risk of recurrence after abrupt interruption of maintenance treatment may occur with other disorders and treatments, including neuroleptics in schizophrenia and possibly antidepressants in recurrent depression. The phenomenon of discontinuation-associated iatrogenic risk of early recurrence of major psychiatric illness has clear clinical implications. These include the need to evaluate safer methods of interrupting long-term maintenance treatment, particularly when clinical indications for rapid cessation are compelling and gradual discontinuation is not feasible. Questions also arise concerning interpretation of existing experimental studies of maintenance treatments that require interruption of treatment, reduction of dose, or crossover to a placebo, as well as the ethical and scientifically unambiguous design of future studies of this kind.