The state of stress and its implications for shear on fault planes during fluid injection are crucial issues for the HDR (Hot Dry Rock) or EGS (Enhanced or Engineered Geothermal System) concept. This is especially true for hydraulic stimulation experiments, aimed at enhancing the connectivity of a borehole to the natural fracture network, since they tend to induce the shearing of fractures, which is controlled by the local stress regime. During the 2000 and 2003 stimulation tests at Soultz-sous-Forets, France, about 10,000 microearthquakes were located with a surface seismological network. Hundreds of double-couple (DC) focal mechanisms were automatically determined from first-motion polarities using the FPFIT program [Reasenberg, P.A., Oppenheimer, D., 1985. FPFIT, FPPLOT and FPPAGE: Fortran computer programs for calculating and displaying earthquake fault-plane solutions. US Geological Survey Open-File Report 85-739, 25 pp.]. The majority of these mechanisms indicate normal-faulting movement with a more or less pronounced strike-slip component. Some quasi-pure strike-slip events also occurred, especially in the deeper part of the stimulated rock volume, at more than 5 km depth. Although we found a double-couple solution for all events, we tried to observe and quantify the proportion of the non-double-couple (NDC) component in the seismic moment tensor for several microseisms from the 2003 data. The study shows that the NDC is higher for the events in the vicinity of the injection well than for the events far from the well. We used the method of Rivera and Cisternas [Rivera, L., Cisternas, A., 1990. Stress tensor and fault-plane solutions for a population of earthquakes. Bull. Seismol. Soc. Am. 80, 600-614.] to perform the inversion of the deviatoric part of the stress tensor from P-wave polarities. This method was applied to different datasets from the 2000 test, taken from the shallower and deeper parts of the stimulated region. The results show a stable, horizontal, NE-SW-oriented trend of the minor horizontal stress, but a rotation of the major stress from a sub-vertical direction (top of the stimulated region) to a sub-horizontal one (bottom of the stimulated region). This implies a change from a normal-faulting to a strike-slip regime, in agreement with our fault-plane solutions. Finally, we applied the stress components to the nodal planes of several events and were able to determine their fault plane and obtain a 3D image of the fracture network, based on real data. (c) 2006 Published by Elsevier Ltd on behalf of CNR.