KENDALL: THE GLACIER LAKES OF CLEVELAND. 11 
The fall of these overflow valleys is usually very small near 
the head and steepens rapidly down stream, a feature which is 
rarely observed in normal valleys. The small fall in the upper 
parts of these channels usually results in the accumulation of 
peat, and this often produces a drainage out at the top end which 
obscures the characteristics of the intake. 
The transverse sections of the overflow valleys are very 
characteristic ; they invariably exhibit exceedingly steep sides, 
and, where unmodified by subsequent stream-action, possess 
broad, flat floors. These features indicate rapid cutting by a large 
stream. 
In a large valley of normal drainage the sinuosities of the 
valley bear little or no direct relation to the meanderings of the 
stream. But in these overflov^' channels the glacier stream fitted 
the valley so closely that the valley walls conform to the well- 
known relation of the banks of a stream to its bends, a steep 
bank on the outside of the bend, with a gentler slope on the inside. 
These features are well shown in the upper portion of Newton 
Dale. 
Both the intake and the outfall of an overflow valley may 
end quite abruptly on a steep hillside, each end of the gorge 
being approximately at the level of the lake from which and to 
which the waters flowed. Subsequent cutting by a small stream 
may, however, render these points obscure. 
A remarkable and striking effect of fluctuations in the edge 
of the ice, both in advances and retreats, is the production of 
deserted oxbows. These features are peculiar to marginal 
overflows, never occurring in direct overflows. When the 
advance of a lobe of ice causes an obstruction of a marginal 
outflow at one point, then an " in-and-out " crescentic valley 
is carved in the opposing hillside to connect the two portions 
of the valley. This new portion may be cut to such a depth as 
to become the permanent channel ; the part obstructed b}' ice 
will therefore never come into use again, and on the withdrawal 
of the ice may appear as a high-level loop many feet above the 
functional channel. The more frequent case, however, is that 
in which the new channel has not been cut so deeply as the old 
one. In these instances, the withdrawal of the ice has reopened 
