We studied the effects of replacement therapy in two groups of patients with symptomatic hyponatremia. Thirty-three patients, who were studied prospectively, had no evidence of cerebral demyelinating lesions. Their hyponatremia (mean serum sodium concentration [+/- SE], 108 +/- 1 mmol per liter) was increased to 126 +/- 1 mmol per liter with hypertonic saline (856 mM) delivered at a rate of 1.3 +/- 0.2 mmol per liter per hour. The serum sodium concentration did not rise to normal or hypernatremic levels in the first 48 hours of therapy, and none of these patients had a respiratory arrest or other hypoxic episode. Twelve patients, evaluated retrospectively, had evidence of cerebral demyelinating lesions at autopsy or on computerized axial tomography. The rate of correction of hyponatremia (1 +/- 0.2 mmol per liter per hour) was similar to the rate in the patients in Group I. However, at least one of four characteristics was present: an increase in serum sodium to normal or hypernatremic levels in the first 48 hours, a change in the serum sodium concentration of more than 25 mmol per liter in the first 48 hours, a hypoxic-anoxic episode, and an elevation of serum sodium to hypernatremic levels in patients with hepatic encephalopathy. Although these four features were associated with demyelination, our observations suggest that this complication does not depend on the rate of correction of hyponatremia.