We investigate the spatiotemporal pattern of crustal anisotropy in the source area of the 2004 Niigata earthquake (M 6.8) that occurred in the northern segment of the Niigata–Kobe tectonic zone, central Japan, by measuring shear‐wave splitting parameters from waveform data of local earthquakes. Our results show that the fast polarization directions in the upper crust have spatial variations across the region of the earthquake that are likely caused by both structural and stress field effects. The northwest–southeast direction near the northeastern end of the source zone (beneath station N.NGOH) and the east–west direction to the southwest (beneath station N.KWNH) are consistent with the spatial variation of the orientation of the maximum compression of the local stress field. Fast polarization directions at other stations tend to align in the directions of active faults and folds and thus are considered to be structure induced. These spatial patterns were unaffected by the earthquake. However, at two stations (N.NGOH and N.KWNH) we observe an increase in both the average and scatter of the normalized delay times () during the aftershock period. In addition, two stations (HIROKA and N.YNTH) that are located in the strike‐normal direction east of the source area show an increase in the average of the normalized and a rotation of up to 90° of the fast direction immediately after the mainshock. We also notice that stations located very close to the source fault (DP.YMK and DP.OJK) show larger average delay times compared with stations farther away (HIROKA and N.YNTH) during the postseismic stage. To explain the temporal changes in the strength of the anisotropy, we speculate that spatiotemporal variations in microcrack development in and around the source area could be caused by static stress changes due to tectonic deformation and the earthquake rupture.