We used the ratio of spectral characteristics of the EEG alpha and beta frequencies recorded in 10 apparently healthy subjects (university students) to control a computer cursor in the graphic interface in three scenarios with the use of neurofeedback. Our results showed that the scenario which uses the power of alpha- and beta-2-frequencies provided the highest accuracy and the best speed control. The parameters of beta-1-frequency were found to be less effective since an increase or a decrease of their power could result from an increase or a decrease of the power of alpha- and beta-2-frequencies. However, the subjects got the skill of cursor control with efficiency of 81%, gradually growing during learning, after a relatively short period of time (5 trainings per 2 weeks).