Thirty eight patients with ischaemic and non-ischaemic central retinal vein occlusion were evaluated for the effect of isovolaemic haemodilution. They were allocated at random to a haemodilution group (19 patients, panretinal photocoagulation and isovolaemic haemodilution) and a control group (19 patients, panretinal photocoagulation). Haematocrit was lowered in steps to 30 to 35% in the haemodilution group by repeated exchanges of whole blood for plasma and dextran (MW 40 000) and kept at this level for a period of six weeks. The haemodilution did not lead to serious complications. Three months after starting the treatment eight of 19 patients with haemodilution showed a better visual acuity, whereas only one of 19 control patients had improved. Seven of 17 patients with haemodilution, but only one of 17 control patients, retained a better visual acuity after one year. In the haemodilution group there were fewer patients with macular fibrosis and more with only minor foveal changes. The haemodilution seems to be more effective in patients with ischaemic than with non-ischaemic central retinal vein occlusion. It is concluded that isovolaemic haemodilution improves the visual outcome of patients with central retinal vein occlusion, probably mediated by enhanced retinal blood flow.