Many approaches for locating earthquakes attempt to solve a nonlinear inverse system in a one-dimensional velocity model. To ensure precise earthquake locations in geologically complex areas, however, a three-dimensional velocity model has to be considered. In this study, the authors have applied an alternative approximate method based on the dynamic ray theory with a 3-D velocity model to relocate 374 events from 1980 to 1989 recorded by the Taiwan Telemetered Seismographic Network (TTSN). The solutions to locating earthquakes using 1-D and 3-D velocity models are compared. These results show that the root mean square (RMS) of travel time residuals for the 3-D model are much lower than those for the 1-D velocity model. This discrepancies between the 1-D and 3-D solutions reflect the importance of lateral velocity variations in locating earthquakes in the Taiwan area. These discrepancies are larger in focal depths than in epicentral determinations. From the examination of seismicity along the Chaochou fault with both 1-D and 3-D velocity models, the results here show that the hypocenter determinations from a 1-D model any even lead to an absurd interpretation in seismicity. The presence of the lateral heterogeity if seismic velocity in Taiwan makes the necessity of locating earthquakes with a three-dimensional velocity model. The increases in computation time on using a 3-D velocity model to locate earthquakes can be minimized by applying the dynamic ray tracing technique due to its fast computation capacity and high level of accuracy.