Abstract: In 1973 the American Scholz C H et al. published a paper titled “Earthquake Prediction: A Physical Basis” in the journal of Science. However, world studies on earthquake prediction in recent several decades have not authenticated the validity that many kinds of precursors may appear before earthquakes. In 2010 Scholz once more published in Science a paper titled “The Prediction Puzzle”, in which he negated the idea on the physical basis for earthquake prediction. In this paper, I discuss why the physical basis they proposed is unscientific.
   The earthquake prediction studies should be focused on the prediction of large earthquakes. A large earthquake sharply differs from a small one at that the former, after the fracture has initiated, has a long time dynamic rupture propagating process, resulting in a large size quake. For a large earthquake, when fault fracture has initiated, we can not know how long the duration of the dynamic rupture process will be, and how large the final fault size will be. This is just the point why predicting the occurrence of a large earthquake is such a difficult task. Now we have to work out a new physical basis for large earthquake prediction. The author of this paper tentatively propose that, mainly based on geodetic survey and seismicity study, we should find out a few locked sections of active faults as possible locations of coming large earthquakes.

Key words: Earthquake prediction, Physical basis, Dynamic rupture process, Fault locked section