The objective of our work is to invert the intensities of preinstrumental destructive earthquakes to retrieve their principal source characteristics. This is relevant in countries (as Italy and Greece) where a large and high-quality data bank of earthquake intensities during the past few centuries exists. We have previously validated our technique presented in this journal for the ML 5.9 1987 Whittier Narrows earthquake, and here we demonstrate that our algorithm works on three other earthquakes with known sources. Three more validations of the algorithm are presented here by using the site intensities observed by the U.S. Geological Survey for the M 6.4 1983 Coalinga, the MS 7.1 1989 Loma Prieta, and the Mw 6.7 1994 Northridge earthquakes. Our simplified KF formula simulates the body-wave radiation from a linear source, and 11 source parameters are retrieved: the three nucleation coordinates, the fault-plane solution, the seismic moment, the rupture velocities, the along- strike and antistrike rupture lengths, and the shear-wave velocity in the half-space. To find the minima on the hypersurface of the residuals in the multiparameter model space, we use a genetic process with niching (Niching Genetic Algorithm) because we have already shown that the problem is bimodal for pure dip-slip mechanisms. The objective function of the nonlinear inversion is the sum of the squared residuals (calculated-minus-observed intensity at all sites). The three tests presented here were successful because the results agreed with the sources already known from instrumental measurements. A very accurate solution was found for Loma Prieta, notwithstanding the complexity of its source. In the two other cases, sources close to the reference ones were retrieved. The quality of our source determinations are obviously lower than those of instrumentally based models, but would be highly noteworthy for earthquakes without seismological recordings. The potential of the “KF inversions” seems particularly promising for European countries where seismicity rates are not very high, and where a lot of information on earthquake damage is available for preinstrumental earthquakes. The present series of validations raises our hope that further information on historic seismic events can be obtained from intensities, thus increasing the knowledge of seismotectonics and seismic hazard.