We are now ready to consider the probability that loess accumulated mainly during the retreat of the ice. Such a retreat exposed a zone of drift to the outflowing glacial winds. Most glacial hypotheses, such as that of uplift, or depleted carbon dioxide, call for a gradual retreat of the ice scarcely faster than the vegetation could advance into the abandoned area. Under the solar-cyclonic hypothesis, on the other hand, the climatic changes may have been sudden and hence the retreat of the ice may have been much more rapid than the advance of vegetation. Now wind-blown materials are derived from places where vegetation is scanty. Scanty vegetation on good soil, it is true, is usually due to aridity, but may also result because the time since the soil was exposed to the air has not been long enough for the soil to be sufficiently weathered to support vegetation. Even when weathering has had full opportunity, as when sand bars, mud flats, and flood plains are exposed, vegetation takes root only slowly. Moreover, storms and violent winds may prevent the spread of vegetation, as is seen on sandy beaches even in distinctly humid regions like New Jersey and Denmark. Thus it appears that unless the retreat of the ice were as slow as the advance of vegetation, a barren area of more or less width must have bordered the retreating ice and formed an ideal source of loess. Several other lines of evidence seemingly support the conclusion that the loess was formed during the retreat of the ice. For example, Shimek, who has made almost a lifelong study of the Iowan loess, emphasizes the fact that there is often an accumulation of stones and pebbles at its base. This suggests that the underlying till was eroded before the loess was deposited upon it. The first reaction of most students is to assume that of course this was due to running water. That is possible in many cases, but by no means in all. So widespread a sheet of gravel could not be deposited by streams without destroying the irregular basins and hollows of which we have seen evidence where the loess lies on glacial deposits. On the other hand, the wind is competent to produce a similar gravel pavement without disturbing the old topography. "Desert pavements" are a notable feature in most deserts. On the edges of an ice sheet, as Hobbs has made us realize, the commonest winds are outward. They often attain a velocity of eighty miles an hour in Antarctica and Greenland. Such winds, however, usually decline rapidly in velocity only a few score miles from the ice. Thus their effect would be to produce rapid erosion of the freshly bared surface near the retreating ice. The pebbles would be left behind as a pavement, while sand and then loess would be deposited farther from the ice where the winds were weaker and where vegetation was beginning to take root. Such a decrease in wind velocity may explain the occasional vertical gradation from gravel