Peels made from 10 geoslices beneath a riverbank at Washington's Hunting Island, 45 km inland from the Pacific coast, aid in identifying sand that liquefied during prehistoric earthquakes of estimated magnitude 8-9 at the Cascadia subduction zone. Each slice was obtained by driving sheetpile and a shutter plate to depths of 6-8 m. The resulting sample, as long as 8 m, had a trapezoidal cross section 42-55 cm by 8 cm. The slicing created few artifacts other than bending and smearing at slice edges. Each slice is dominated by well-stratified sand and mud deposited by the tidal Columbia River. Nearly 90% of the sand is distinctly laminated. The sand contains mud beds as thick as 0.5 m and at least 20 m long, and it is capped by a mud bed that contains a buried soil that marks the 1700 Cascadia earthquake of estimated magnitude 9. Every slice intersected sills and dikes of fluidized sand, and many slices show folds and faults as well. Sills, which outnumber dikes, mostly follow and locally invade the undersides of mud beds. The mud beds probably impeded diffuse upward flow of water expelled from liquefied sand. Trapped beneath mud beds, this water flowed laterally, destroyed bedding by entraining (fluidizing) sand, and locally scoured the overlying mud. Horizontal zones of folded sand extend at least 10 or 20 m, and some contain low-angle faults. Many of the folds probably formed while sand was weakened by liquefaction. The low-angle faults may mark the soles of river-bottom slumps or lateral spreads. As many as four great Cascadia earthquakes in the past 2000 yr contributed to the intrusions, folds, and faults. This subsurface evidence for fluid escape and deformation casts doubt on maximum accelerations that were previously inferred from local absence of liquefaction features at the ground surface along the Columbia River. The geosliced evidence for liquefaction abounds not only beneath banks riddled with dikes but also beneath banks in which dikes are absent. Such dike-free banks of the Columbia River, if interpreted without study of postdepositional structures in deposits beneath them, provide insufficient basis for setting upper bounds on the strength of shaking from great Cascadia earthquakes. Online material: Data from outcrop surveys, vibracores, and penetrometer tests; tabular summary of depositional and postdepositional features in geoslices.