Abstract: The direct method of fuzzy pattern recognition has been applied to the synthetic recognition of medium-short-term (several months) earthquake precursors of different kinds observed at different stations.The basic technique of this method is to construct a suitable membership function for each single precursor at each single station and then the comprehensive membership function for several precursors of different kinds or for the same precursor observed at different stations.By using this method,the precursors observed in the Tokai area and its vicinity of Japan are studied with emphasis. They are the radon content in ground water,groundwater level,tilt,tide and volume strain. The abundant observational data from 3 tilt stations,2 tide stations,2 stations for observation of radon content in ground water,2 stations for observation of water level and 30 stations for observation of volume strain are used. Two possible precursor patterns for medium-short-term earthquake prediction in the Tokai area were proposed. One is applied to the synthetic analysis of several precursors of different kinds and at different stations,the other is applied to the synthetic analysis of one-kind of precursor (volume strain) observed at different stations.For comparison,the result of synthetic recognition of medium-short-term precursors of radon content in ground water at many stations before the great M = 7.8 Tangshan earthquake of 1976 and its M = 6.9 strong aftershock observed in the Beijing-Tianjin-Tangshan area and its vicinity of China is introduced. The precursor pattern obtained corresponds to the second precursor pattern for the Tokai area of Japan.