The May 17, 1976 Gazli earthquake (mb = 6.2, Ms = 7.0) has both seismological and engineering importance. Waveform modeling of long-period IDA Rayleigh waves, and WWSSN long- and short-period body waves indicates a thrust mechanism with strike N40°E, dip 54°SE, and rake of 78°. The moment is estimated to be 1.6 × 1026 dyne-cm; the fault area, 150 km2; the average dislocation, 3.3 meters; and the average stress drop, 200 bars. Strong directivity effects due to a propagating rupture are present in both the long- and short-period teleseismic P waves. The data are not adequately explained by a simple point source, and require a three-dimensional, finite-source model. The relative amplitudes of P and pP at both long and short periods are used to deduce that the rupture started at a depth of 15 km and propagated almost entirely unilaterally upward, with an average rupture velocity of 2.0 to 2.5 km/sec. Simultaneous deconvolution of long- and short-period P waves reveals a complex, azimuthally dependent, source-time function. Finite-source models further suggest that a major part of the faulting was concentrated at depth, below 7 km. The strong-ground motion within 10 km of the fault is characterized by unusually high-amplitude (1300 cm/sec2), high-frequency (10 Hz) accelerations and low-amplitude (10 cm), low-frequency (0.8 Hz) displacements. Synthetic Wood-Anderson records computed from the accelerograms give a local magnitude, ML, of 6.35. The duration and character of the strong-ground motion is consistent with a rupture propagating upward in the general direction of the station.