terranean Sea, the ice would rapidly melt away in the water and in the warm, moist atmosphere, and therefore have no tendency to erode a lake basin.
The Lake of Lugano, with its curious radiating arms, is said to be another difficulty. But each of these arms is the outlet of a valley or series of valleys, which were no doubt reduced to nearly level plains by subaërial denudation before the ice began its work. The basin of these valleys comprises about two hundred square miles and the watershed to the north is moderately high; but there can be no doubt that a large overflow from the Como Glacier poured into it; and the difficulty seems to me to be purely imaginary if we simply recognize the fact that an essential preliminary to lake erosion is a pre-existing nearly level valley bottom.
Another difficulty is said to be the frequent presence of islands in the lakes; but here again the answer is easy. The islands, always ground down to roches moutonnées, were craggy hills in the pre-existing valleys, and such hills existed because they had for ages resisted the subaërial denudation which had hollowed out the valleys. The same characters of density or toughness that enabled them to resist ordinary denudation, enabled them also, to some extent, to resist destruction by ice erosion; just as the character of the rocks which enabled ordinary denudation to bring them down to a nearly level surface in the valley bottom, also facilitated the ice erosion which converted the level valley floor into a rock basin and, after the ice left it, into a lake.
Every writer brings forward the well-known fact that the ends of glaciers pass over beds of gravel or moraine matter, without destroying or even disturbing it. But there is no reason why they should do more than compress such beds of loose material and roughly level their surfaces. It is the old delusion of a glacier acting like a scoop or plow that leads to the idea that if it can erode rock slowly it must altogether demolish gravel or bowlder clay. But if we turn to the description I have given of how a glacier erodes a rock basin and apply this to its passage over a bed of gravel or bowlder clay, we shall see that in the latter case the erosion would be much more difficult, because each ice-imbedded stone or rock would press into the yielding material, which would close up instantly behind it under pressure of the ice and thus leave no result. Where the subglacial water accumulated, channels would be cut in the gravel or clay, but elsewhere there would probably be no erosion at all. Some writers maintain that the lakes were all filled up with alluvium previous to the Glacial epoch, and that the ice cleared out this incoherent matter; but it is almost certain that no such clearance would have taken place, because the glacier would pass over such a sur-
