The Coalinga earthquake sequence of 1983 provided a unique opportunity to perform a three-dimensional velocity and hypocenter inversion in an area of complex three-dimensional structure dominated by folding and blind thrusts. Additionally, since other varied geological and geophysical studies have been completed in this area, the three-dimensional inversion solution could be compared to other interpretations. Inversion of 7696 P and 1511 S first arrivals from earthquakes and 696 P first arrivals from refraction shots produced a three-dimensional velocity model with grid spacing of 1–2 km in the hypocentral area. The overall shape and location of velocity features correspond well to the mapped surface geology. The three-dimensional inversion yields details of folds where the resolution is good and the general shape where resolution is lower. The amounts of structural relief inferred for the local folds are similar to values inferred from geologic data. The three-dimensional velocity solution has several distinctive features. There is a linear high-velocity body (6.1–6.5 km/s), about 25 km long, from 6 to 8 km depth, that may represent a fragment of Coast Range ophiolite. A shallow low-velocity zone (LVZ), which extends for 20 km along the fold axis at about 6 km depth and correlates with LVZs observed in both the refraction and the reflection data, may indicate high pore pressure caused by lateral compressive strain. Deeper LVZs occur within inferred Franciscan material and are characterized by horizontal or southwest dipping zones of varied thickness, 4–8 km wide and 5–10 km long. These LVZs may represent multiple thrust faults. Their locations and geometry are consistent with thrust faults inferred with seismic reflection data. The three-dimensional velocity solution compares well to prior two-dimensional seismic reflection and refraction models and observed gravity. The shape of the inferred sedimentary section agrees well with the reflections from the Cenozoic strata. Compared to the refraction model, the three-dimensional solution has similar velocities and similar locations of velocity features but is more detailed in the hypocentral zone where it uses more data. The gravity computed from the three-dimensional velocities is similar to the observed gravity in both shape and amplitude. Both a simple one-dimensional initial model and a complex initial model derived from the refraction interpretation were tried. A simple starting model gave the best results. The S velocity solution has different resolution than the P velocity solution because it uses a different set of stations, and it has lower resolution because it uses fewer arrival times. While the general patterns of velocity variation are similar for both V_p and V_s, the V_s solution tends to have more smearing of velocity features and can have somewhat different locations of velocity features.