SUBGLACIAL AND ENGLACIAL CHANNELS. 315 



then returns to its old position in the latest crevasse. Probably all subgla- 

 cial tunnels wander, but the transverse ones more than the longitudinal. 

 A transverse channel could be stationary in two ways: by the melting of 

 the ice as fast as it advanced, or by becoming filled with sufficient gravel 

 to force the ice to flow over it. The stratified osars date from a time when 

 the channels were approximately stationary, due probably to both the above- 

 cited conditions, aided perhaps by a sluggish ice movement. 



The same causes which enlarge crevices into tunnels maintain the. tun- 

 nels against a slow inward movement of the ice. An instance is seen on 

 the Malespina glacier, where the siibsidence of the roof of a glacier river 

 has resulted in the formation of scarps of depression on the upper siirface 

 of the glacier. That glacial rivers do not succeed in eroding so broad can- 

 yons and tunnels in the ice as they would in rock is due partly to the 

 gradual collapse of the walls and partly to the fact that the glacier is ever 

 being renewed. 



While we postulate some inward flow, the assumption must be so held 

 as to allow the formation of crevasses, which we can not account for if we 

 assume very much fluency or plasticity. 



In case of a decaying ice-sheet having its ne'v^ and higher ice unbroken, 

 the thinning of the ice over the hills would from time to time cause the 

 appearance of crevasses at places before free from them. NeAv subglacial 

 tunnels would soon be formed, if surface waters flowed into them. Thus, 

 in Maine, as the ndve retreated northward, there would be a corresponding 

 advance of the subglacial rivers, so far as they serve to carry off superficial 

 waters. The advance would take place into a i-egion previously drained by 

 superficial streams. The thinning of the ice would cause a multitude of 

 crevasses to appear in new places, but many of these would be of no sig- 

 nificance. To use a biological phrase, there would be a natural selection 

 of the crevasses, only those intersecting the established superficial streams 

 having the power to determine the courses of the larger subglacial rivers to 

 that place. In this manner the superficial drainage systems of the ice-sheet 

 may have had an important influence in determining the number and coui'ses 

 of the subglacial rivers. 



It is difficult now to ascertain the causes of the dividing of the ice- 

 sheet into superficial drainage systems or how far it was determined by the 

 underlying hills. There may be parts of the slush zone that are so flat 



