We are advancing rapidly in our understanding of how new subduction zones form, because geologists and geodynamicists are working together to find convincing examples and develop increasingly realistic numerical models. Results indicate that it is the density and strength of the oceanic lithosphere that determines whether or not SI is likely, along with the length of any lithospheric weakness to be exploited. Because of the strength of the oceanic lithosphere, some styles of subduction initiation (SI) are easy, for example, transform collapse and polarity reversal, which take advantage of pre-existing weaknesses. SI can also occur around the margin of a sufficiently large, hot, and long-lived plumehead, for example as happened in Late Cretaceous time on the W and S margins of the Caribbean plumehead. Other styles of SI are difficult to accomplish, for example transference and passive margin collapse, because these require rupturing strong oceanic lithosphere. These insights are used in this presentation to challenge existing paradigm for how the modern Cascadia subduction zone formed in Eocene time. This is generally thought to have occurred when an oceanic plateau (Siletzia) was partially subducted in a pre-existing subduction zone, whereupon a new convergent margin formed west of Siletzia. If this occurred, it would be a convincing example of transference, requiring rethinking of the likelihood of this process. I propose instead that the Cascadia subduction zone formed by lithospheric collapse along the margins of the paleo-Yellowstone plumehead in Eocene time. Geologic constraints from Siletzia are integrated with evidence from crustal structure, uplift-related sedimentation, and magmatic provinces and core complexes along the NE margin of the Eocene plumehead. The plumehead was ~1200 km in diameter, extending from the craton in the east to the Pacific in the west. A key mystery is why – if the Yellowstone plume began ~55 Ma – is there no evidence for plume activity for ~30 m.y. until the Columbia River flood basalts erupted? I propose that the descending slab disrupted the plume both in terms of flow (downwelling slab disrupted upwelling plume) and thermally (slab cooled upwelling hot mantle), and that it took ~30 m.y. for regions of upwelling and downwelling to separate enough to allow plume activity to resume.