On 6 April 1992 at 13:55 UT, an earthquake of Ms = 6.8 occurred along the Revere-Dellwood-Wilson (RDW) fault, in the triple-junction region at the northern end of the Cascadia subduction zone. This was the first significant event in this region recorded by modern digital broadband seismic networks, thus providing the first opportunity to examine the rupture process of a major earthquake along this young oceanic transform fault. In this article, an empirical Green's Function technique is applied to regional and teleseismic surface waves to estimate the rupture directivity, the extent of rupture, and the slip distribution along the Revere-Dellwood-Wilson fault associated with this earthquake. The 20-sec low-pass-filtered relative source time functions (RSTF's) are single pulses with an azimuthal variation in the pulse width. This suggests that the rupture propagated to the NW (315° ± 20°), along the Revere-Dellwood-Wilson fault (striking 326°). Higher frequency RSTF's reveal two discrete subevents. The clear azimuthal variation in the time separation of these subevents requires that relative to the first subevent, the second is located 13 to 20 km in the direction 345° ± 20°. Using the RSTF at HRV (perpendicular to the rupture direction), a total rupture length of 35 km is estimated, with the bulk of the slip concentrated in a 20-km-long segment of the RDW fault to the northwest of the epicenter. Two peaks are observed in the estimated slip distribution, with maximum values of 1.8 and 1.1 m, respectively. The rupture model derived from this analysis is similar to that obtained from the analysis of body waves and is consistent with the results of aftershock studies. The latter indicate a paucity of aftershock activity (and low moment release) in the 20-km-long segment of the RDW fault to the NW of the epicenter. A distinct peak in aftershock activity 30 to 40 km to the NW of the epicenter likely represents the termination of rupture. The good agreement between the results of this study, the rupture model estimated from body-wave analysis, and the aftershock distribution bode well for the application of the empirical Green's function method using surface waves. It suggests that this method could be applied to large, historic earthquakes in this region, for which regional and teleseismic surface waves are often the most reliable data set.