Thick Roxana Silt (middle Wisconsinan) in central and southwestern Illinois traditionally has been interpreted as loess derived from valley-train deposits in the ancient Mississippi River valley. Winters et al. (H. A. Winters, J. J. Alford, and R. L. Rieck, Quaternary Research 29, 25-35, 1988) recently suggested that the Roxana was not directly related to glacial activity, but was derived from sediment produced by increased shoreline and spillway erosion associated with a fluctuating ancestral Lake Michigan. Because (1) paleoenvironmental and paleohydrologic conditions inferred in the hypothesis are unlikely for a loess depositional system and (2) loess did not accumulate during late Wisconsinan deglaciation under conditions similar to those hypothesized, we suggest the hypothesis should be rejected. Roxana distribution suggests the major source was drainage from the upper Mississippi River valley, and variations in loess thickness in Illinois can be explained by consideration of valley width, depth, orientation, and postdepositional erosion. Tills in the headwaters region of the ancient Mississippi drainage system in Minnesota and Wisconsin occur in the appropriate stratigraphic position and have colors and mineralogic compositions that suggest they could be the parent till of the Roxana. We believe a valley-train source for thick Roxana is most probable and urge continued consideration of middle Wisconsinan glaciation in the upper Great Lakes area. ?? 1989.