T he mechanism of lithium-induced diabetes insipidus was investigated in 96 patients and in a rat model. Polydipsia was reported by 40% and polyuria (more than 3 liter/day) by 12% of patients receiving lithium. Maximum concentrating ability after dehydration and vasopressin was markedly impaired in 10 polyuric patients and was reduced in 7 of 10 nonpolyuric patients studied before and during lithium therapy. Severe polyuria (more than 6 liter/day) was unresponsive to trials of vasopressin and chlorpropamide, but improved on chlorothiazide. Rats receiving lithium (3-4 meq/kg/day) developed massive polyuria that was resistant to vasopressin, in comparison to rats with comparable polyuria induced by drinking glucose. Analysis of renal tissue in rats with lithium polyuria showed progressive increase in the concentration of lithium from cortex to papilla with a 2.9-fold corticopapillary gradient for lithium. The normal corticopapillary gradient for sodium was not reduced by lithium treatment. The polyuria was not interrupted by brief intravenous doses of vasopressin (5-10 mU/kg) or dibutyryl cyclic AMP (10-15 mg/kg) capable of reversing water diuresis in normal and hypothalamic diabetes insipidus rats (Brattleboro strain). The present studies suggest that nephrogenic diabetes insipidus is a common finding after lithium treatment and results in part from interference with the mediation of vasopressin at a step distal to the formation of 3′,5′ cyclic AMP.