Computer programs for automatic sleep scoring of human and animals EEG records are widely used in many branches of physiological research. They are particularly useful during the sleep research because the traditional methods requiring human expert scoring of long records are very laborious and time-consuming. The aim of this work was to investigate the dependency of accuracy of automatic sleep scoring on different EEG parameters for EEG-recording quality assessment. We find statistically significant dependency of accuracy of automatic and expert scoring on several spectral characteristics. This dependency can be used as objective quality assessment method of EEG recording for sleep scoring and it makes possible to assess accuracy of automatic sleep scoring a priori.