A high-resolution tomographic study of the Long Valley region began in 1989 with the installation of a special fan array of eight three-component borehole (80--160 m depths) receivers on the northwestern rim of the caldera to provide the data necessary for a joint inversion for hypocenters and the three-dimensional P and S wave velocity structure in and below the caldera. The experiment specifically targeted the subsurface location of the previously inferred magma chamber beneath the Resurgent Dome. Additional coverage was provided by existing U. S. Geological Survey and University of Nevada, Reno seismographic networks. We employed the progressive inversion scheme of Thurber (1983), with cubic spline interpolation after Michelini and McEvilly (1991). The final model is based on data from 280 well-distributed local earthquakes. Some 6900 arrival times were used, including about 700 S wave times from the three-component stations. The resulting velocity structure reflects the known geology, defining the low-velocity caldera fill to 2 km depth, contrasting markedly with the surrounding higher-velocity Sierran block and highland terrane. No isolated distinct low-velocity anomalies are revealed beneath the caldera floor, although a diffuse zone of reduced velocity persists to a depth of about 8 km. These lower velocities may be related to hydrothermal alteration and/or extensive fractures. The Vp/Vs structure contains significant lateral variation within and beneath the caldera to about 8 km depth, and these variations can be related to accepted models of the active geothermal system. The lack of a significant S wave velocity anomaly, along with the normal or low values of the Vp/Vs ratio, argues against the presence of a sizable and distinct magma body at shallow to midcrustal depth beneath Long Valley caldera, although a very low percentage of partial melt cannot be precluded. ¿ American Geophysical Union 1993