A large number of landslides were triggered by the 5.12 Wenchuan earthquake and the 4.20 Lushan earthquake. The statistical results of field surveys indicate that the overall distribution pattern of the scale of earthquake-induced rockmass collapse and landslide changes with the earthquake intensity. This statistics-based result needs to be confirmed by performing laboratory physical experiments. Based on the framework of self-organized criticality (SOC) theory, shaking table tests of sandpile model under seismic excitations was conducted. The results show that for peak ground acceleration (PGA) in the range of 0.075g-0.125g, the relation between the amount and cumulative frequency of sand follows a negative power law; for PGA between 0.15g and 0.25g, the relation obeys a lognormal distribution; for PGA between 0.35g and 0.45g, the relation turns to obey a normal distribution. Data from the cellular automata numerical simulation demonstrate that, as the earthquake intensity increases, the dynamic behaviors of sandpile model exhibit a strong power-law first, then a weak power-law, and finally a normal distribution. It is suggested here that the above-revealed distribution laws may also apply to other areas. The new recognition will provide a scientific basis for the prediction of landslides triggered by earthquake.