The performance of two popular migration methods--the Kirchhoff integral and reverse-time migrations--is evaluated through applications to imaging complex structures using prestack shot records. The migration results from the Marmousi model data demonstrate that reverse-time migration is more accurate than Kirchhoff migration in imaging the steeply dipping faults. However, the improved accuracy of reverse-time migration requires higher computational costs. In the application to the Alberta Foothills data where a good estimate of the velocity model is available, however, both the Kirchhoff and the reverse-time migration methods produce almost identical results. This implies that in the real world of exploration seismology, it will be relatively difficult to identify which method performs better because we never know the exact answer of the subsurface.