Long-period strain seismogram recorded at Pasadena is used to determine the focal process of the 1960 Chilean earthquake. Synthetic seismograms computed for various fault models are matched with the observed strain seismogram to determine the fault parameters. A low-angle (∼ 10°) thrust model with rupture length of 800 km and rupture velocity of 3.5 km/sec is consistent with the observed Rayleigh/Love wave ratio and the radiation asymmetry. A seismic moment of 2.7 · 10 30 dyn · cm is obtained for the main shock. This value, together with the estimated fault area of 1.6 · 10 5 km 2, gives an average dislocation of 24 m. The strain seismogram clearly shows unusually long-period (300-600 sec) wave arriving at the P time of a large foreshock which occurred about 15 minutes before the main shock, suggesting a large slow deformation in the epicentral area prior to the major failure. A simple dislocation model shows that a dislocation of 30 m, having a time constant of 300-600 sec, over a fault plane of 800 × 200 km 2 is required to explain this precursory displacement. The entire focal process may be envisaged in terms of a large-scale deformation which started rather gradually and eventually triggered the foreshocks and the "main" shock. This mechanism may explain the large premonitory deformations documented, but not recorded instrumentally, for several Japanese earthquakes. The moments of the main shock and the precursor add to 6 · 10 30 dyn · cm which is large enough to affect the earth's polar motion.