A retrospective study of treatment and outcome is described in 59 consecutive referrals to a catchment psychogeriatric service meeting ICD9 criteria for manic depressive illness or depressive neurosis. Of 22 who failed to respond to tricyclic antidepressants, nine (of whom four had also failed to respond to ECT) were treated by lithium augmentation. Systematic comparisons between lithium-treated subjects, tricyclic responders and others failing to respond to tricyclics revealed no significant demographic differences. Lithium augmentation was successful in 6 / 9 subjects. Two 'lithium failures' were treated with tranylcypromine to good effect. At follow-up (median six months, range 3-20 months) 7 / 9 subjects in the lithium-treated group were well. This was similar to the follow-up status in tricyclic responders and significantly better than outcome in the other tricyclic non-responders. Lithium augmentation appears to be a relatively well-tolerated treatment manoeuvre in refractory depression in old age, with treatment response similar to that reported in younger subjects, and may be of particular use where ECT has failed. K E Y wom-Lithium, depression, geriatric psychiatry.