the Loire in the midst of this tableland flowing due north to shed their waters ultimately into the Atlantic. Parallel with them, divided from the Loire only by the chain just described, distant from it from forty to fifty miles, is the Rhone, running in a precisely opposite direction, due south, to discharge into the Mediterranean.
Then, again, going west, we have the Lot, the Tarn, the Jonte, the Truyére, and the Dordogne, all in their early youth streaming from east to west, their sources, or those of some of them, twenty miles from the Allier, which is racing as hard as it can run to the north. The explanation of this last feature is easy. When the granite was upheaved it lifted a crust of Dolomite on its back like a huge shell, and in lifting split the shell in many places.
The rivers, rising in the granite of the Margeride, the Mont d'Aubrac, the Aigoual, the Monts de Lozère, and slipping off their impervious sides looking for outlets, found these fissures, took possession of them, and rushed down them on their way to the plains in the west.
The average depth of these chasms is from 1,300 to 1,500 feet, and their width at the bottom varies from 160 to 1,500 feet. Their rocky walls are carved by rain and frost into the most fantastic forms. At one time I held, with M. Martel, that these cañons were originally subterranean watercourses, and that the caverns formed by the underground waters became open valleys by the falling in of their roofs. But this idea is untenable, as I now see. The rivers descending from the granitic range must have found a passage, and they found it in the already cleft masses of the Gausses. They rise out-
