Abstract: This paper is written as requested by the Chinese scientific journal during the author's visit to China in 1983.The author believes that seismic stresses are quite small in magnitude and therefore, their effects must be subtle. Sensitive instruments will be required to detect them. For such disciplines used in many countries for earthquake prediction, as piezomagnetic effect, velocity ratio, strain, tilt, geodetic surveying, electrical resistivity, radon and foreshocks, their nature and prospectiveness are discussed and suggestions made. It is noted that although the argument of Brune1(1979) that earthquakes may be unpredictable in principle is forceful, the author is not so pessimistic. The earth's crust may be very inhomogeneous and its behaviour fickle, but that only demands that we have a massive redundancy of data. Several methods and numerous instruments and observations must be used simultaneously. More attention should be given to (1) seismic anisotropy and attenuation, (2) multicoloured laser-ranging, (3) improvements in the signal/noise ratio in tectonomagnetism, (4) use of alternative and more extensive arrays of detector electrodes in resistivity measurements and (5) a careful re-examination of the statistics of foreshock and aftershock sequences.