Abstract: In this paper,discussions are made on the widely applied intensity attenuation models of seismic hazard analysis: the point-source model introduced by Cornell,C. A. and the fault-rupture model given by Der Kiureghian and A. H-S. Ang. Comparisons with practical intensity data show that there is systematic deviation which isn't very large in the results given by point-source model for regions where isoseismals are of elongated shapes. To reduce the above deviation,the concept of rupture length is introduced into the fault-rupture model,but the attenuation is quite different from practical dam. Comparisons with isoseismals of actual earthquakes and computation show that the results given by the fault-rupture model overestimate the attenuation in high-intensity area and underestimate it in low-intensity area along the directions of fault or the major axis of the isoseismals. The reason causing such systematic deviation in the fault-rupture model is analysed,and the rationality of the empirical ellipse model is discussed,with illustration by examples of seismic hazard analysis.