Accelerometric records of the Chi-Chi earthquake from sites on the hanging wall exhibit larger acceleration than those from the footwall. Based on ground accelerations recorded at 79 near-field stations (10 hanging-wall stations and 69 footwall stations, respectively) and precise mapping of fault-rupture traces, the hanging-wall/footwall effects of the Chi-Chi earthquake have been fully studied. We show that the hanging-wall effects cannot be simply accounted for by a proper choice of distance metric. The closest distance to the rupture plane (Drup) is then selected to develop an empirical ground-motion model by using the data collected during the mainshock of the Chi-Chi earthquake that struck Taiwan. With the exception of some sites immediately next to the rupture traces (Drup ≤ 5 km), the acceleration residuals between this empirical model and the recorded data at the footwall stations are close to zero for stations in the distance range from 5 to 50 km. On the other hand, the average acceleration amplification on the hanging wall is equal to the natural logarithmic values of 0.64 ± 0.4 for all hanging-wall sites within 20 km of Drup. The hanging-wall/footwall effects have also been evaluated for several response spectral periods. It is observed that both the horizontal and vertical components of spectral acceleration are apparently amplified for sites on the hanging wall at a distance from 5 to 20 km for spectral periods 0.02 to 0.5 sec, whereas the vertical component has less amplification than the horizontal in all the spectral periods considered. The horizontal component of spectral acceleration at the hanging-wall sites also shows a larger value for the long-period motion, relative to the footwall, for periods larger than 1.0 sec. The hanging-wall effects are relatively constant, at low frequencies, as the distance extends to about 20 km. This observation cannot be explained by the simplified empirical model. Rather, it suggests that waves trapped in the hanging-wall wedge may have been involved.