A three-component seismometer was emplaced in a borehole in Long Valley caldera for a period of 2 months to record local earthquakes to the south, mostly in the Sierra Nevada. The seismic records were plotted in depth versus time sections to produce ''vertical hypocentral profiles'' for both the vertical and horizontal components. The data collected and displayed in this way provide good resolution of the crustal structure at depth and avoid the attenuation and complications introduced by te near-surface caldera fill. The record sections were then modeled with two-dimensional ray tracing to match the observed travel times. Although nonreversal of the profiles and uncertainties in hypocentral locations introduce ambiguities, the final interpretation reveals two separate low-velocity bodies which we have tentatively identified as magma beneath the resurgent dome in Long Valley. The first of these is a small (2 km by 6 km) body with a depth to top of approximately 3.7 km beneath the surface at the southern end of the resurgent dome. The other body appears larger and lies at a depth of 5.5 km beneath the northern end of the resurgent dome.