Comparison of histories of great earthquakes and accompanying tsunamis at eight coastal sites suggests plate-boundary ruptures of varying length, implying great earthquakes of variable magnitude at the Cascadia subduction zone. Inference of rupture length relies on degree of overlap on radiocarbon age ranges for earthquakes and tsunamis, and relative amounts of coseismic subsidence and heights of tsunamis. Written records of a tsunami in Japan provide the most conclusive evidence for rupture of much of the plate boundary during the earthquake of 26 January 1700. Cascadia stratigraphic evidence dating from about 1600??cal yr B.P., similar to that for the 1700 earthquake, implies a similarly long rupture with substantial subsidence and a high tsunami. Correlations are consistent with other long ruptures about 1350??cal yr B.P., 2500??cal yr B.P., 3400??cal yr B.P., 3800??cal yr B.P., 4400??cal yr B.P., and 4900??cal yr B.P. A rupture about 700-1100??cal yr B.P. was limited to the northern and central parts of the subduction zone, and a northern rupture about 2900??cal yr B.P. may have been similarly limited. Times of probable short ruptures in southern Cascadia include about 1100??cal yr B.P., 1700??cal yr B.P., 3200??cal yr B.P., 4200??cal yr B.P., 4600??cal yr B.P., and 4700??cal yr B.P. Rupture patterns suggest that the plate boundary in northern Cascadia usually breaks in long ruptures during the greatest earthquakes. Ruptures in southernmost Cascadia vary in length and recurrence intervals more than ruptures in northern Cascadia.