The May 23 1995 Sorachi-Chubu earthquake (M_JMA=5.7) occurred at 18:01:29 JST in northwest Hokkaido at the western end of the Okhotsk plate. This is the largest of the earthquakes which occurred during the last 160 years in the shallow seismic zone with a length of about 250km, which can be considered to be a plate boundary between the Okhotsk and Eurasian plates in the Late Mesozoic. This is also the first significant shallow event in this inland region recorded by a modern digital broadband seismic network, thus providing a first opportunity to examine the focal mechanism of a felt earthquake occurring along this old collision zone. Double-couple mechanism solutions were obtained from three kinds of data: P-wave first motions, broadband waveforms recorded by STS 1 and STS2 seismographs, and acceleration seismograms recorded by strong motion seismographs (JMA Model-87). The source parameters obtained by centroid moment tensor (CMT) inversion are: the centroid depth=7km; (strike, dip, rake)=(187.7°, 23.2°, 115.4°; 339.3°, 69.2°, 79.6°); seismic moment M_o=8.6×10¹⁷N·m(M_W=5.8). The waveform inversion of the strong motion seismograms for a single total source indicates: centroid depth=5km; (strike, dip, rake)=(192.6°, 44.2°, 110.9°; 344.6°, 49.3°, 70.9°); seismic moment M_o=3.0×10¹⁷N·m(M_w=5.5). The seismic moment is about one third as large as that obtained by the CMT inversion. The discrepancy between the two methods may be caused by the large later phases generated within the travel paths used for the CMT inversion. We made the waveform inversion excluding these phases for the strong motion seismograms. The long period CMT solution gives an information on the integrated view of the overall faulting regardless of a higher frequency information on the crustal structure as well as on the fault geometry during the earthquake. As far the fault plane solution, P-wave first motions gives the similar result as those inferred from both of the waveform inversions. As supplementary evidence, the disaster distribution, including tumbled tomb stones, suggests thrust faulting. The principal horizontal stresses obtained are also consistent with those predicted from the active faults ascertained by the geographical surveys. They suggest a reverse fault in which the western hanging wall moves upward. The results of the present study reflect the plate tectonic driving forces. They may be localized at reactivated sites of prior deformation, or at sites of stress concentration resulting from sutured structure, such as along the margin of the Okhotsk plate in the Late Mesozoic. It is reasonable to suspect that part of the present Okhotsk plate movement towards the Eurasian plate is absorbed in the weak zone created by the Late Mesozoic tectonic processes.