In the 15-month period (September 1999 to December 2000) after the 1999 Chi-Chi, Taiwan, earthquake (Mw 7.6) about 80 campaign-surveyed GPS stations in central Taiwan were occupied up to seven times. Furthermore, seven continuous GPS stations were put into operation in the epicentral region mostly within three weeks after the main shock. GPS data from these near-field stations and some pre-existing far-field stations were utilized to study the postseismic deformation following the Chi-Chi earthquake. The postseismic displacements of the GPS stations were approximated by a combination of an exponential transient decay function with a relaxation time of 86 days and a postseismic linear rate change. Stations on the hanging wall displayed west-northwesterly horizontal displacements of up to 252 mm and uplift as large as 229 mm. Postseismic displacements are much larger to the south near the epicenter of Chi-Chi main shock than in the north where the maximum coseismic displacement occurred. Stations on the footwall show only small displacements. The postseismic GPS data were inverted to infer the deeper fault geometry and afterslip distribution based on a four-segment fault model with its shallow part constrained by coseismic fault geometry deduced from GPS and seismological studies. The optimal model requires the lower fault segment to be a horizontal plane at a depth of 10.4 km, consistent with a model based on GPS data taken during the first three months after the main shock and other geological or geophysical studies. The maximum afterslip (459 mm) occurs in the hypocentral region and in the northern part of the decollement. Slip of about 430 mm also occurs in the southern part of the decollement. The afterslip moment inferred from 15 months of GPS data is 4.7 ¿ 1019 N m, approximately 2.3 times the seismic moment released by aftershocks. This implies that a major part of the postseismic deformation is aseismic.