A program for the training of perceptual speed and attention with a duration of four weeks was developed and given to thirty brain damaged patients. In order to evaluate a possible generalization effect of the training all subjects were tested three times, i.e. at the beginning of the training, after four and after eight weeks with a series of 16 standardized psychometric tests. These comprised tasks which closely resembled the training procedures as well as tasks testing psychomotor functions in a broader sense and even tasks of reasoning and other intellectual functions. To test the different effects of training, spontaneous recovery and test repetition an experimental design with two groups of patients and two control groups was necessary. Only the groups of patients showed a significant and marked improvement in all tests which could neither be explained by spontaneous recovery nor by test repetition. The training effect proved to be stable after four weeks. The most marked improvement, however, was seen with those tasks most closely resembling the training procedures; in no case the performance level of the control groups was reached.