Dr. Hooker has described some parallel terraces, very analogous in their aspect to those of Glen Roy, as existing in the higher valleys of the Himalaya, of which his pencil has given us several graphic illustrations. He believes these Indian shelves to have originated on the borders of glacier-lakes, the barriers of which were usually formed by the ice and moraines of lateral or tributary glaciers, which descended into and crossed the main valley, as we have supposed in the case of Glen Roy; but others he ascribes to the terminal moraine of the principal glacier itself, which had retreated during a series of milder seasons, so as to leave an interval between the ice and the terminal moraine. This interspace caused by the melting of ice becomes filled with water and forms a lake, the drainage of which usually takes place by percolation through the porous parts of the moraine, and not by a stream overflowing that barrier. Such a glacier-lake Dr. Hooker actually found in existence near the head of the Yangma valley in the Himalaya. It was moreover partially bounded by recently formed marginal terraces or parallel roads, implying changes of level in the barrier of ice and moraine matter.* (* Hooker, "Himalayan Journal" volume 1 page 242; 2 pages 119, 121, 166. I have also profited by the author's personal explanations.) It has been sometimes objected to the hypothesis of glacier-lakes, as applied to the case of Glen Roy, that the shelves must have taken a very long period for their formation. Such a lapse of time, it is said, might be consistent with the theory of pauses or stationary periods in the rise of the land during an intermittent upward movement, but it is hardly compatible with the idea of so precarious and fluctuating a barrier as a mass of ice. But the reader will have seen that the permanency of level in such glacier-lakes has no necessary connection with minor changes in the height of the supposed dam of ice. If a glacier descending from higher mountains through a tributary glen enters the main valley in which there happens to be no glacier, the river is arrested in its course and a lake is formed. The dam may be constantly repaired and may vary in height several hundreds of feet without affecting the level of the lake, so long as the surplus waters escape over a col or parting ridge of rock. The height at which the waters remain stationary is determined solely by the elevation of the col, and not by the barrier of ice, provided the barrier is higher than the col. But if we embrace the theory of glacier-lakes, we must be prepared to assume not only that the sea had nothing to do with the original formation of the "parallel roads," but that it has never, since the disappearance of the lakes, risen in any one of the glens up to the level of the lowest shelf, which is about 850 feet high; for in that case the remarkable persistency and integrity of the roads and deltas, before described, must have been impaired.