Abstract: The South China Sea （SCS） is located at the intersection of the Eurasian, Pacific, and India-Australia plates. It is the largest marginal sea in a series of marginal seas in the Northwest Pacific Ocean. Many models have been proposed for the opening of the SCS, such as the extrusion model driven by the collision of the Indian plate and the Eurasian plate, and the slab pull model related to the subduction of the proto-SCS. This study aims to constrain the models of opening the SCS through the anisotropic structure of the central basin of the SCS and its surroundings. Based on the seismic data recorded by ten ocean bottom seismometers recovered from two passive seismic experiments conducted by Tongji University in the central basin of the SCS in 2012 and 2014, three different shear wave splitting methods are used to obtain the XKS splitting results of the central basin for two global earthquakes and the S phase splitting results provided by 20 regional earthquakes surrounding the SCS. The SKS splitting results demonstrate the presence of strong anisotropy with the NE fast direction in the central basin of the SCS, which may be related to mantle flow along the ocean ridge during seafloor expansion and the mantle flow dragged by the subduction of the proto-SCS plate. Strong anisotropy is also observed in the upper mantle surrounding SCS, and the anisotropy observed in different azimuths is different. The fast directions obtained are consistent with previous SKS-splitting results, GPS, and plate motions, and importantly correspond well to the regional tectonics or mantle convection models. The anisotropic results are consistent with the expected results of the extrusion model driven by the collision of the Indian-Eurasian plate and the slab pull of proto-SCS. The anisotropy results are inconsistent with the ideal upwelling driven model of the mantle plume. Unfortunately, due to the limited splitting observations in the central basin, the anisotropic results cannot confirm or falsify the “Atlantic-type” seafloor spreading model, the backarc spreading model, or the plate-edge rifting model. To verify the above models, further observations are needed.