Brain water content was measured in tissue samples taken at operation from 19 patients with intrinsic cerebral tumours imaged preoperatively by magnetic resonance. A high correlation (r = 0.94, p less than 0.0001) between white matter water content and the longitudinal relaxation time (T1) enabled water content to be estimated from T1 to within 4%. 11 patients received dexamethasone and improved clinically but their T1, and thus brain water content, was unchanged an average of 6 days later. Intravenous infusion of 20% mannitol in 11 patients significantly reduced T1 in oedematous white matter and tumour within 15 min of administration, and by 30 min the T1 of oedematous white matter had fallen to a mean of 32.4 (SEM 7.1) ms, corresponding to a reduction in water content of 1.4 (0.3)%.