An efficient mathematical method is presented for computing the near- fault strong ground motions in a layered half-space, giving explicit consideration to the static offset due to surface faulting. In addition, the combined effects of and on the near-fault ground motions are investigated. First, after checking the fault integration in the representation theorem, it is found that when an observation point is close to the fault plane, Green's functions exhibit near singularities, which consist of extremely sharp peaks in a narrow band close to the observation point. Therefore, direct numerical integration becomes quite onerous for computing near-fault ground motions, because the dynamic Green's functions must then be distributed very densely in order to evaluate accurately the effects of the near singularities. Instead, a new form of the representation theorem is introduced, which exploits the property that the dynamic Green's functions can be approximated by the corresponding static Green's functions in the vicinity of the singularities. The modified theorem, which involves the device of adding and subtracting the static Green's functions from the dynamic ones, is the sum of two fault integrations. The first integration involves the difference of the dynamic and the corresponding static Green's functions, while the second contains only the static Green's functions. This formulation requires much less CPU time than the original one when near-fault ground motions are considered, because the near singularities of the dynamic Green's functions in the first integration are completely eliminated by subtracting the static Green's functions. While the second integration does require a densely distributed set of points to capture the near-singular behavior of the static Green's function, it needs to be performed only once, as it is valid for all frequencies. Subtraction of the static Green's functions from the dynamic functions has the added benefit of making the integration over the wavenumber in the determination of the Green's functions much more efficient, especially for surface faulting. This is because the difference of the dynamic and static integrands converges rapidly to zero with increasing wave- numbers, whereas the original integrands diverge in the case of a source point on the free surface. The proposed methodology is used to investigate the two most important effects in near-fault ground motions, fling step (e.g., Abrahamson, 2001) and rupture direc- tivity (e.g., Somerville et al., 1997), by paying special attention to the contribution of static and dynamic Green's functions. It is found that the fling effects stem mainly from the second integral in the modified representation theorem, which involves the static Green's function. The fling effects are dominant in the slip direction only in the vicinity of the surface fault and are negligible for buried faults, because the static Green's function attenuates rapidly with distance from the fault, r, as the order of (1/r 2 ). Also, more importantly, when an observation point is located above a buried fault, the medium has to remain continuous, and thus cannot fling. By contrast, the directivity effects stem mainly from the first integral, which involves the dynamic Green's function, and attenuate much more slowly than the fling, on the order from 1/r to . The directivity effects are dominant in the fault-normal direction, especially