BOWLDEKS OF GLACIAL GKAVELS. 337 



fact that we ha,ve no proof that the whole length of au open canyon would 

 be simultaneously an area of deposition. On the contrary, the analogies 

 all favor the hypothesis that as the ice retreated northward the more north- 

 ern portion of superficial channels would be areas of denudation, their 

 drift being swept southward and deposited nearer the margin of the ice- 

 sheet. If so, we would have at each place the ice all melted in the bottoms 

 of the channels before sediments began to be deposited, and this would 

 result in all the bowlders first being freed from ice and left on the bed of 

 the canyon, to be afterwards covered with finer drift. This would well 

 account for the long reaches of osars and osar-plains containing few or no 

 large bowlders on the surface. 



In case of a subglacial river, the enlargement proceeded from below 

 upward and laterally. The bowlders contained in the ice forming the roof 

 of the vault would from time to time drop into the channel as it became 

 enlarged and they were released from the ice. If the bowlders were high 

 ujD in the ice, they would be last to fall. Yet if at this place the velocity 

 were such as to sweep all the finer matter from the channel, these bowlders 

 might be left on the bed of the subglacial stream. Then as the ice became 

 thinner a time might come when fine matter would be deposited at this 

 place, and now be found overlying the large bowlders. So also the local 

 land slopes must be considered, i. e., whether the place of observation was 

 on up or down slopes. 



Thxis the large bowlders do not form a crucial test between the sub- 

 glacial and superglacial streams. Yet we are warranted in affirming that 

 the presence of very large quantities of fine sedimentary matter overlying 

 the till bowlders is consistent with the hypothesis of a superficial stream, 

 and the presence of a large number of rolled bowlders in the upper parts 

 of the glacial gravels can be considered as probable evidence of a subglacial 

 stream. The fewer the number of such bowlders in the one case and the 

 greater the number in the other, the greater becomes the degree of proba- 

 bility. And the matter is still further complicated by the great difi^erence 

 in the size of the bowlders furnished by the different kinds of rock. In 

 slate I'egions there might be found only one bowlder to a square rod, while 

 in granite regions there might be ten or twenty. A superficial channel 

 would show very difi^ereiit deposits in the two cases, yet they would be 

 formed in the same manner; and so of a subglacial stream. 



MON XXXIV 2'2 



