We report results of measurement of particle motion in a foam-rubber model of normal faulting and compare the results with similar results for a strike-slip geometry. Standard computer modeling of strong ground motion from normal fault earthquakes has used dislocation theory, in which slip along the shallow part of the fault is prescribed by assuming particular time functions for fault slip. Unfortunately, in the case of normal faults, there are essentially no near-fault data from large earthquakes to constrain the modeling. In an extensional faulting regime the static normal and shear stresses along the fault must approach zero at the surface, and thus the upper few kilometers of the fault have inherently less stored strain energy than the maximum possible for strike-slip faults. In addition there are dynamic effects from geometry and drop in fault-normal stress which affect the fault motion. Physical models of faulting, such as foam-rubber modeling, are guaranteed to obey static and dynamic mechanical laws and thus can be used to gain insight into the physical processes involved. In this study we compare surface accelerations from normal fault and strike-slip geometries. The data show surface accelerations near the normal fault trace that are systematically lower, by an average factor of about 0.1, compared to the accelerations at the side sensors, which represent strike-slip motion. These results suggest that kinematic modeling of ground motion using classical dislocation techniques is inappropriate on the shallow part of the fault. The results of this study are qualitatively similar to those obtained for numerical models, such as a dynamic lattice model <Shi et al., 1997> and a finite element model <Oglesby et al., 1998>, lending support to the probability that similar effects take place in the real Earth. ¿ 1999 American Geophysical Union