I have analyzed records from 20 strong-motion instruments that recorded the 1989 Loma Prieta earthquake (ML = 6.9) to determine the spatial and temporal distribution of slip using a tomographic back-projection technique. I find that the earthquake ruptured bilaterally at approximately 80% of the local shear-wave velocity over a distance of slightly more than 30 km from 13 km northwest to 20 km southeast of the hypocenter. Slip within the rupture zone was highly variable both in magnitude and direction. Slip at the hypocenter and immediately up dip was relatively low, about 1 m. There were two main areas of concentrated slip, one centered 7 km northwest of the hypocenter at a depth of 14 km and the other centered 6 km southeast of the hypocenter at a depth of 12 km. Peak slip amplitude on these high-slip patches exceeded 4.5 m. A surprising aspect of the rupture model is that rake varies from being predominately strike-slip to the southeast of the hypocenter to being predominately reverse slip to the northwest. Despite this variation in rake, the average rake determined from the strong-motion data is in agreement with that determined teleseismically. There is a correlation between areas of high slip and areas of low aftershock activity. A simple model to explain this observation is that the areas of high slip are also areas of high strength, which slip only during infrequent large mainshocks. Other areas of the fault are weaker and may slip both seismically and aseismically. Most of the slip in the Loma Prieta earthquake occurred between 9 and 16 km depth on a structure that dips to the southwest and runs underneath the surface trace of the San Andreas fault at a depth of 8 to 10 km. This observation complicates the assessment of seismic hazard on the southern Santa Cruz Mountains segment of the San Andreas fault.