Origin of Eskcrs. — Crosby. 33 
is well nigh inconceivable, and contrary, as previously noted, 
to all our observations on existing glacial streams. The fact 
that, in general, eskers trend toward cols or depressions in 
water partings calls for no special comment, cince it is entire- 
ly consistent with both hypotheses. 
Relations of Eskcrs to tJie Ground Moraine. — Eskers, as 
well as their terminal plains, normally overlie the ground mo- 
raine or till, the only important exception being when, locally, 
the till i'S wanting and they are superposed directly upon the 
underlying bed-rock ; and they are rarely, if ever, covered 
by till, or even sprinkled with bowlders, except such as might 
readily be supposed to slide or fall into a superglacial channel 
from the bordering slopes of ice. This relation is, in every 
particular, strictly normal for superglacial fluvial deposits ; 
but the subglacial fluvial deposits could not possibly escape 
being covered by till and angular bowlders on the melting of 
the ice, except on the supposition that the englacial drift was 
so strictly limited to the basal portion of the ice during the 
esker-forming period as hardly to warrant its classification 
as englacial, which would make the maximvmi elevation of 
drift in the Pleistocene ice-sheet distinctly less than in the 
Malaspina glacier and many of the Greenland glaciers. If the 
normal esker is of subglacial origin, then the englacial drift 
was indeed scanty and confined to very low levels in the ice , 
and the subglacial stream was deprived of one important and 
necessary source of detritus for aggrading its bed and build- 
ing its terminal plain. The only alternative, apparently, is 
to suppose that the subglacial stream always held tenaciously 
to its course until finally, through the general process of abla- 
tion, superficial and basal, its tunnel became roofless at all 
points. This would mean, for one thing, that every esker was 
formed in part, or at the last, in earth-bottomed canyons open 
to the sky, which is impossible wherever the grade rises south- 
ward, as, practicall}', it does for nearly every true esker in 
some part of its course. The subglacial esker must remain 
under cover, under a roof of ice thick enough to hold the sub- 
glacial stream to a channel which, regardless of the ground 
topography, crossed, directly or obli(|uely, ridges hundreds of 
feet in bight, until it is finished ; and then, as a necessary cor- 
ollary, the stream which made it must completely abandon its 
