Maps of the strain accumulation rate in the San Francisco Bay area have been constructed from trilateration observations extending from about 1972 until the Loma Prieta earthquake in late 1989. The observations were corrected to remove offsets imposed by shallow fault creep and by four M ∼6 earthquakes that occurred in the Bay area during that time interval. The Bay area was divided into 32 contiguous polygons, and the uniform (in both space and time) strain rates that best explain the changes in the corrected (earthquake and shallow fault creep offsets removed) distances within each polygon were calculated. In a coordinate system with the 1 axis directed N58°E and the 2 axis N32°W (perpendicular and parallel to the local tangent to the small circle drawn about the Pacific-Sierra Nevada pole of rotation) the averages of these 32 strain rates (each weighted by the area of the polygon) are ε11=9.2±7.4, ε12=−160.7±4.6, and ε22=8.2±6.2 nanostrain/yr, where extension is reckoned positive and quoted uncertainties are standard deviations. As expected from the Pacific-Sierra Nevada relative plate motion, the overall strain rate is predominantly right-lateral shear across a vertical plane striking N32°W. The net increase in the 12,225 km2 area of the trilateration network is only 212±110 m2/yr, which arises from almost equal extensions in the N32°W and N58°E directions. Within the network the strain rates vary from polygon to polygon. Significant areal dilatation rates are observed in almost ⅓ of the individual polygons and the N32°W extension rates tend to be negative to the west of the Hayward-Rodgers Creek fault trend and positive east of it. The N58E extension rate is positive in 22 out of 32 polygons, a proportion that is significantly larger than would be expected by chance if the N58°E extension rate were zero or negative. The pre-1989 strain accumulation across the eventual site of the Loma Prieta rupture involves fault normal contraction as well as right-lateral shear, consistent with the rupture mechanism.