At low temperatures air is able to hold only a negligible amount of water-vapour, and this current, coming in contact with the extremely cold surface of the ice, is sucked dry, and its moisture added to the ice-sheet. Probably there is little true snowfall, but the condensation takes place chiefly close to the surface, forming a frozen mist resembling the “ice-mist” of Siberia. Even if the central land is not high enough to reach into the upper current at its normal level, the surface outflow of cold air will draw the current down to the level of the ice. This will warm it by compression, but the ice-surface is so cold that such warming makes little difference in the end. This process of condensation ensures that after the ice reaches a certain thickness it becomes independent of topography, and in fact the centre of the Scandinavian ice-sheet lay not along the mountain axis, but some distance to the east of it. It is probably only on the edges of the ice-sheet, and especially in areas of considerable local relief, that snowfall of the ordinary type takes place, associated with moist winds blowing in the front section of depressions which skirt the ice-edge. But when conditions are favourable this source of supply is sufficient to enable these local ice-sheets to maintain an independent life, merely fusing with the edges of the larger sheet where they meet. Examples of such local centres in Europe were the Irish and Scottish glaciers, and at a later stage the Lofoten glaciers of the west of Norway, and in America the Cordilleran glaciers of Columbia. Penck and Brückner have demonstrated that in the Alps the increase of glaciation was due to a fall of temperature and not to an increase of snowfall. The argument is threefold: firstly, the lowering of the snow-line was uniform over the whole Alpine area, instead of being irregular as it would be if it depended on variations of snowfall; secondly, the area and depth of the parent snow-fields which fed the glaciers remained unchanged, hence the increased length of the glaciers must have been due to decreased melting below the snow-line, i.e. to lower temperatures; thirdly, the upper limit of tree-growth in Europe sank by about the same amount as the snow-line. The same conclusion holds for the great Scandinavian and North American ice-sheets, the extension of which was undoubtedly due to a great fall of temperature. In the case of the Alps the interesting point has come to light that the fall of temperature, though due in part to increased elevation, is mainly accounted for by the presence of the Scandinavian ice-sheet, which extended its influence for many miles beyond the actual limits of glaciation, so that its waxings and wanings are faithfully reproduced in those of the Alpine glaciers, even to the details of the final retreat after the last maximum.