SCTBGLACIAL AND ENGLACIAL CHANNELS. 313 



into one another but separated by a solid slab of ice. When a new cre- 

 vasse forms, it becomes filled with water, but it is narrow, and melting bj- 

 convection currents is very slow. Under a pressure of thousands of feet tlie 

 water searches out every point of weakness. It acts by its pressure to rup- 

 ture the ice, also to penetrate between the ice and the underlying- rock, and 

 also by its superior weight to raise bodily the ice in contact with it. The 

 last can not be done without fracturing ice of great thickness, and this the 

 flotation is not able to do. The line of contact between the ice and the rock 

 is that of weakness, since the adhesion of the ice and rock is less than the 

 cohesion of the ice, and probably of the ground moraine, where there is 

 one. If the ice has been held above the rock by a film of basal water, or 

 there is a basal furrow in the bottom of the ice, the water immediately 

 penetrates between the ice and the rock, and soon enlarges the smallest 

 chink to the capacity of the stream. Moreover, the ice must flow down 

 into each scratch of the rock or the trickle will begin and all the rest 

 follow. Whether ice held under great pressure in fair contact at all points 

 with the rock could prevent the passage of the waters is a matter of con- 

 jecture. It is possible that continued pressure might cause a minute flow 

 of the ice, so as slowly to raise in arch form the central parts of the block 

 forming the dam, and thiis permit the water to escape. Only the minutest 

 opening would be required to initiate the flow, and the melting would do 

 the rest. 



It is known that ice can flow over deeply buried ridges without being 

 crevassed at the surface. If the . motion continues while the thickness 

 diminishes, the time will come when the ridge will cause an increasing 

 bulging of the ice surface, and finally crevasses. In many cases of retreat- 

 ing glaciers surface waters are seen to pour down crevasses that would not 

 exist when there was considerably deeper ice, and in these cases the waters 

 must have established subglacial or englacial channels for themselves not 

 very long ago. At the moulin, where the water in the new crevasse is 

 separated from a large tunnel by at most only a few feet of ice, it is not so 

 wonderful that it finds a passage. The difficulty is to show how a channel 

 is for the first time established beneath or within the ice, often underneath 

 long reaches of ice unbroken at the surface. It is constantly being done 

 on the glacier longitudinally, yet the large Marjelen See can not keep open 

 a permanent channel transverse to the ice flow. We seem to be driven to 



