To compare the effect of diet, cirrhotic patients with ascites were randomised into two treatment groups: (1) a low sodium diet (21 mmol) per day or (2) an unrestricted sodium intake. Both groups received effective doses of diuretics (spironolactone or, if necessary, spironolactone and furosemide. One hundred and forty patients from 12 liver units were included according to well defined criteria (group 1: 76; group 2: 64). After an initial four to seven day period of bed rest and salt restriction (21 mmol sodium pd), randomisation was done in each centre. We saw no significant difference between the two groups with respect to clinical and biochemical data; mortality or withdrawal (definitive or temporary) because of biochemical disturbances (group 1: 34%, group 2: 22%); the time for complete disappearance of ascites was significantly shorter (p = 0.014) for the salt restricted patients actuarial survival (curves plotted up to the 120th day) was not statistically different (p = 0.18), but division into subgroups using various prognostic factors showed that survival was significantly better for salt restricted patients without previous gastrointestinal bleeding (p = 0.02); hospitalisation time and costs were identical in both groups. We conclude that the advantage of a normal salt diet was not shown in this study.