The centroid moment tensor (CMT) waveform inversion method of Dziewonski et al. (1981) is modified to analyze long period seismic waves generated by ''single force'' events such as the gigantic landslides associated with the 1980 eruption of Mount St. Helens. We refer to the method as centroid single force (CSF) inversion. As the result of the inversion we obtain the spatio-temporal centroid and three components of a vector we call the CSF vector, which is the seismic point source representation of a landslide and whose direction is the opposite of the direction of sliding. The scaler magnitude of the CSF vector measures the overall size of the landslide and is given by MCSF=MD, where M is the mass of the sliding object and D is the sliding distance. We apply CSF inversion to the long period surface waves generated by the 1980 Mount St. Helens event, the 1975 Kalapana earthquake, and the 1974 Mantaro landslide, using seismograms from the GDSN, IDA, and HGLP networks. The CSF solution for the St. Helens event is remarkably consistent with the actual geological observations and proves the efficacy of CSF inversion. The CSF solution for the Kalapana event does not fit the overall waveforms better than the CMT solution does. However, because the CMT solution does not explain the observed Love wave radiation pattern, it may be necessary to invoke a combination of both types of mechanism for this event. Although the geometry of the CSF solution (the direction of the force) for the Mantaro event is consistent with the actual landslide, MCSF is about 5 times smaller than that expected from the mass and travel distance of the landslide estimated by geological observations. This discrepancy may suggest either the relatively aseismic nature of this landslide or that the total volume of the slide was overestimated. By analogy to single station CMT inversion, single station CSF inversion also appears to be stable and useful. ¿ American Geophysical Union 1989