Microearthquake clusters form distinct, planar patterns within five study regions of a geothermal reservoir undergoing hydraulic fracturing at Fenton Hill, New Mexico. The patterns define individual, slipping joint surfaces of dimension 40--120 m, containing 80--150 events each. Sharp, straight edges truncate the clusters; we interpret these as marking intersections with aseismic joints. Each edge orientation is consistent with an intersection between the active joint and a plane oriented parallel to one of the other clusters we identify. Therefore it appears that cluster shapes constrain the geometry of seismic and aseismic joints; both could be important components of the fluid-flow network. The distribution of inferred slip plane orientations is consistent with but fails to provide sufficient constraint to differentiate conclusively between two, very different, stress field estimates, one measured using pressurization and wellbore breakouts, the other using focal mechanisms of the largest microearthquakes. An impermeable joint model, requiring pore pressure in excess of the normal stress on a joint before slip can occur, was inconsistent with many of the inferred slip plane orientations. The high-quality locations were possible because events from the same cluster generated nearly similar waveforms, permitting the precise determination of relative arrival times. Standard deviations of arrival-time residuals fall between 0.1 and 1.1 ms for these clusters. Major axes and aspect ratios of the 90% confidence ellipsoids range from 6 to 28 m and 1.5 to 8, respectively. Small events dominate the seismic energy release and thoroughly populate the identified, active joints, allowing the hypocenters to reflect details of the joint structure. To further investigate the reservoir structure, we applied a source-array, slant-stack technique to waveforms from the well-located clusters, yielding directions that scattered energy left each cluster. By studying paths of scattered waves we expected to pinpoint impedance contrasts that might have indicated concentrations of fluid-filled joints. However, results show that scattered energy in the S wave coda left the source region in the same direction as the direct S wave. Direct waves may have excited borehole tube waves that became trapped in the vicinity of the geophone tool, overwhelming any energy scattered from the reservoir.¿ 1997 American Geophysical Union