Wide stream-cut trenches, now filled with lakes, swamps, and underfit streams, transect the drumlin plain of the Superior Lobe in a subparallel pattern trending southwest, oblique to the modern regional slope and drainage. They are pictured as the products of high-velocity streams in subglacial tunnels, driven by the great hydrostatic pressure resulting from the thick mass of still-active ice. The water for such flow cannot have come from the glacier surface, because the cold upper part of the ice could not permit its penetration. It must have come rather from the base, through melting by the geothermal flux or by the frictional heat of basal ice flow. The collecting basin for such water may have extended to the center of the ice sheet in the Hudson Bay area, and the water may have been stored for thousands of years until it could break through the frozen toe of the Superior Lobe and "catastrophically" cut the tunnel valleys during its exit. As the ice thinned to stagnation and the hydrostatic head was lost, the subglacial streams changed their habit from erosional to depositional, forming small eskers along the trenches. After extensive wastage of the Superior Lobe at the end of the St. Croix phase, the ice readvanced to or slightly beyond the rim of the Lake Superior Basin at least three times. The later readvances, which involved the overriding of proglacial lake beds and the deposition of red clayey till, were not closely synchronized with advances of adjacent ice lobes, and the pollen sequence for northeastern Minnesota shows no pattern of climatic reversals that matches ice-lobe fluctuations. Accordingly, the hypothesis is presented that nonclimatic factors may have controlled the most recent minor advances of the Superior Lobe. Subglacial meltwater may have built up beneath the then-restricted Superior Lobe and behind the dam afforded by the cold ice of the thin toe until the reduction in basal friction permitted rapid glacial flow (a surge) through the dam.