In a previous randomized placebo-controlled clinical trial, we observed significant visual field enlargements induced by computer-based restitution training in patients with cerebral lesions (Kasten et al., Nature med., 4, 1998, 1083-87). Now we asked the question whether this effect is stable after training was discontinued? Here we report data of a follow-up study after a training-free interval (mean 23.5 +/- 2.3 months after end of therapy). 16 patients of the original restitution group and 6 patients of the placebo group were re-examined. On average, in high resolution computer campimetry (stimulus detection: PeriMa, form recognition: PeriForm, color perception: PeriColor) as well as in conventional automatic perimetry (TAP-2000) both groups showed no significant decline in the number of correctly detected stimuli after training was discontinued. However, cluster analysis revealed three different types of patients, who showed either increase (Type-I), decrease (Type-II) or stability (Type-III) in performance. We propose that many patients learn to use the regained visual capacities not only in the setting of a computer training but also in every day life, while other patients do not use the areas of restored vision and show a decrease of visual functions after the end of training. The Type-I group does not need continuous training, while the Type-II group may benefit from phases of refreshment exercises.