The geometry and recurrence times of large earthquakes associated with the Cascadia Subduction Zone (CSZ) were discussed and debated at a March 28-29, 2006 Pacific Northwest workshop for the USGS National Seismic Hazard Maps. The CSZ is modeled from Cape Mendocino in California to Vancouver Island in British Columbia. We include the same geometry and weighting scheme as was used in the 2002 model (Frankel and others, 2002) based on thermal constraints (Fig. 1; Fluck and others, 1997 and a reexamination by Wang et al., 2003, Fig. 11, eastern edge of intermediate shading). This scheme includes four possibilities for the lower (eastern) limit of seismic rupture: the base of elastic zone (weight 0.1), the base of transition zone (weight 0.2), the midpoint of the transition zone (weight 0.2), and a model with a long north-south segment at 123.8? W in the southern and central portions of the CSZ, with a dogleg to the northwest in the northern portion of the zone (weight 0.5). The latter model was derived from the approximate average longitude of the contour of the 30 km depth of the CSZ as modeled by Fluck et al. (1997). A global study of the maximum depth of thrust earthquakes on subduction zones by Tichelaar and Ruff (1993) indicated maximum depths of about 40 km for most of the subduction zones studied, although the Mexican subduction zone had a maximum depth of about 25 km (R. LaForge, pers. comm., 2006). The recent inversion of GPS data by McCaffrey et al. (2007) shows a significant amount of coupling (a coupling factor of 0.2-0.3) as far east as 123.8? West in some portions of the CSZ. Both of these lines of evidence lend support to the model with a north-south segment at 123.8? W.