In order to find that the outcome of vitrectomy could be predicted in diabetics with very low visual acuity (finger counting at 3 m--approximately 20/400--or less), flash electroretinograms (ERG) and visual evoked cortical potentials (VECP) were simultaneously recorded from about 300 patients within one month before the operation. From this series two groups were formed, based on the outcome of surgery. One group (30 patients) had improved visual acuity, with finger counting at 0.1 to 1.0 m (20/200-20/20). The other group (29 patients) was unimproved by surgery. The groups were matched with regard to age, duration of diabetes, preoperative visual acuity and occurrence of cataract and previous panretinal photocoagulation. All patients had dense vitreal opacities, precluding inspection of the ocular fundus. Only patients without surgical and postoperative complications were included. There was a statistically significant (P less than 0.05) difference between the groups in ERG amplitude, but not in VECP latency or amplitude. Based on ERG amplitude, the outcome of vitrectomy could be predicted correctly in 36 of 59 patients (61%) and in 35 of 56 patients (63%) based on VECP latency. The VECP amplitude or the combination of ERG and VECP parameters did not improve the prediction rate. Thus, preoperative flash ERG and VECP can not safely predict the outcome of vitrectomy in diabetics with very low visual acuity.