Inversions of seismic body waves indicate that most large earthquakes involve spatial and temporal heterogeneity of the moment release, but seismic body wave usually lack sufficient bandwidth to constrain the long-period component of the radiation. Analysis of long-period surface waves and free oscillations can constrain the overall source duration and moment; however, most procedures assume simple trapezoidal or boxcar source-time histories, inconsistent with the body wave complexity. We find that source-time function complexity can affect long-period surface waves sufficiently to impact estimates of the source moment tensor, rupture duration and centroid depth. We present a procedure to objectively determine the long-period component of complex source-time function in which we directly utilize results from body wave analysis. The method is applied to two great normal fault events of June 22, 1977 (Tonga, Mw=8.2) and August 19, 1977 (Indonesia, MW=8.5). Standard long period analysis procedures underestimate the total slip duration in both cases. The overall source process times of both earthquakes are longer than 120 s. ¿ American Geophysical Union 1989