396 procp:edings : boston society natural history. 
must favor the aggrading of the floor of the tunnel by the coarser 
detritus and a corresponding elevation of the roof by corrasion and 
melting. If the stream is sufficiently long-lived, this process may 
continue until the deposit attains the level of its terminal alluvial 
cone or delta. The breadth of the deposit is sharply limited by the 
stability of the ice arch; and if we should assume two hundred feet 
as the maximum breadth of ice tunnels, portions of most important 
eskers would demand some other explanation, and in many cases a 
breadth of 500 or 1000 feet would prove inadequate. In fact, the 
subglacialist finds it convenient in such cases to suppose, either that 
the stream rose through crevasses, perhaps in consequence of the 
clogging of the tunnel, and became superglacial for a longer or 
shorter distance ; or that the tunnel became, by excessive ablation, 
locally roofless, or open to the sky, and the resulting canyon was 
widened by the recession or melting back of the walls. As in the 
case of superglacial streams, the water is supposed to be diverted 
after a time, and the deposit left in the abandoned tunnel gradually 
develops the steep lateral slopes and other formal features of eskers 
during the slow melting of the retaining walls and arch of ice. 
Finally, effective crevassing throughout the zone of esker forma¬ 
tion and a failure of englacial drift to rise to any considerable eleva¬ 
tion in the ice are always postulated as important, if not absolutely 
essential, elements of the subglacial hypothesis. 
Review of Evidence. 
Direction or trend of eshers. — That.the typical esker tends to 
conform closely in trend with the movement, and especially with 
the latest movement, of the ice sheet, as recorded in striae, drumlins, 
and bowlder trains, is undoubtedlv true. The movement of the ice 
sheet must have been, in general, normal to its margin, or, in other 
words, in the direction of its steepest surface slope, and therefore in 
the direction which superglacial streams would necessarily follow. 
Eskers depart from this ideal trend no more than might reasonably 
be expected in the case of a superglacial stream, considering that as 
the ice sheet becomes thin, its surface contours must begin to show 
the influence of the underlying topography ; and considering further 
the inevitable inequalities in the distribution of the superglacial 
