glacier known as _moulins_, or wells, which may of course become perfectly dry, if any accident turns the rivulet aside or dries up its source. The most common cause of the intermittence of such a waterfall is the formation of a crevasse higher up, across the watercourse which supplied it, and which now begins another excavation. These wells are often very profound. I have lowered a line for more than seven hundred feet in one of them before striking bottom; and one is by no means sure even then of having sounded the whole depth, for it may often happen that the water meets with some obstacle which prevents its direct descent, and, turning aside, continues its deeper course at a different angle. Such a well may be like a crooked shaft in a mine, changing its direction from time to time. I found this to be the case in one into which I caused myself to be lowered in order to examine the internal structure of the glacier. For some time my descent was straight and direct, but at a depth of about fifty feet there was a landing-place, as it were, from which the opening continued its farther course at quite a different angle. It is within these cylindrical openings in the ice that those accumulations of sand collect which form the pyramids described above. One may often trace the gradual formation of these wells, because, as they require certain similar conditions, they are very apt to be found in various stages of completion along the same track where these conditions occur. Fissures, for instance, will often be produced along the same line, because, as the mass of the glacier moves on, its upper portions, as they advance, come successively in contact with inequalities of the bottom, in consequence of which the ice is strained beyond its power of resistance and cracks across. Rivulets are also likely to be renewed summer after summer over the same track, because certain conditions of the surface of the glacier, to which I have not yet alluded, and which favor the more rapid melting of the ice, remain unchanged year after year. Of course, the wells do not remain stationary any more than any other feature of the glacier. They move on with the advancing mass of ice, and we consequently find the older ones considerably lower down than the more recent ones. In ascending such a track as I have described, along which fissures and rivulets are likely to occur, we may meet first with a sand-pyramid; at a certain distance above that there may be a circular opening filled to its brim with the sand which has just reached the surface of the ice; a little above may be an open well with the rivulet still pouring into it; or higher up, we may meet an open fissure with the two semicircles opposite each other on the margins, but not yet united, as they will be presently by the closing of the fissure; or we may find near by another fissure, the edges of which are just beginning to wear in consequence of the action of the water.