A source model for the 1975 Pocatello Valley earthquake (mb 6.1) in eastern Idaho is determined by fitting observed long- and short-period teleseismic P waves with synthetic seismograms. The source parameters are constrained by results from a study of the aftershock sequence by Arabasz et al. and by a focal mechanism determined from P-wave first motions. The source model for the synthetic seismogram calculations is a three-dimensional finite difference calculation of bilateral, constant effective stress faulting done by Day et al., which is scaled to fit the long- and short-period data. The source fitting the short-period data has markedly lower amplitude (a factor of 5 to 10) and higher corner frequency (0.5 versus 0.1 Hz) than that fitting the long-period data. This leads to the conclusion that the faulting began with a relatively small region of high stress drop and high rupture velocity, then propagated into a region of lower stress drop and lower rupture velocity that adds to the long-period radiation while being a weak source of high-frequency waves. A model approximating this behavior fits the data in both frequency bands. The inferred stress drop for the initial faulting is about 50 bars, which is an order of magnitude smaller than that inferred with the same methods for the 1971 San Fernando earthquake, an event with similar dimensions and focal mechanism. The Pocatello Valley earthquake is also compared to the superficially similar Yellowstone Park event (June 30, 1975), which is found to be much more complex as a source of short-period P waves, suggesting that this event was characterized by highly variable stress drop.