We inferred minutely the geometry of the upper surface of the seismic Philippine Sea (PHS) slab beneath the region from Ise Bay to western Shikoku, southwest Japan, based on hypocentral distribution. Despite the importance of the slab geometry for the subduction zone seismotectonics as well as for the assessment and forecasting of destructive interplate and slab earthquakes, the shape of the PHS slab has been obscure and controversial. We selected well-determined hypocenters during the period from Oct. 1997 through Dec. 2002 from the integrated hypocenter database prepared by the Japan Meteorological Agency. At first, we composed depth data set of the slab upper surface from vertical cross sections of hypocentral distribution along longitude and latitude of every 0.1 degree, and drew first-trial isodepth contours of the upper surface of the PHS slab assuming that the upper boundary of slab earthquake distribution roughly coincides with the slab upper surface. Then, in many areas we made vertical cross sections of hypocenter distribution in many directions including inferred dip-direction and strike of the slab, and revised isodepth contours. The main results are as follows: (1) In the region to the east of Lake Biwa, northwestward gently-dipping PHS slab reaches around the depth of 35km, in which disastrous slab earthquakes presumably took place. (2) To the northeast it continues to the slab beneath the central Tokai district, but to the southwest it is obscure whether it continues to the slab beneath the eastern Kii Peninsula or it is separated from the latter. (3) The slab beneath the middle part of the Kii Peninsula may be torn into two parts, with the southwestern part underlying the northeastern part. (4) There are double seismic zones beneath the southern Kii Peninsula. (5) In the west part of the Kii Peninsula no slab earthquake occurs in deeper part, whereas just in this area a remarkable shallow crustal swarm earthquake activity exists. (6) Beneath Shikoku the slab is being subducted with gentle dip angle and forms a broad ridge. The slab looks continuous from Shikoku to Kyusyu. Recently the PHS slab geometry has been investigated by seismic refraction/reflection profilings and receiver function analyses. Those results are shallower than our result in eastern Shikoku suggesting that slab earthquakes in this region occur in the oceanic mantle. We relocated slab earthquakes in this region using a velocity structure model consistent with the result of seismic profiling and found that the shallower main activity lays in the slab mantle of the assumed model with deeper remarkable events beneath it. In order to clarify the true slab geometry it is essentially important to know in which part of the slab, oceanic crust or mantle, slab earthquakes are actually taking place in each region.