points were wanted.

As to surface-slope its measurement--from nearly 600 trials--was found
to be such a delicate operation that the result would be of doubtful
utility. This would affect the application of all formulas into which it
entered. The water surface was ascertained, on the average of its
oscillations, to be sensibly level across, not convex, as supposed by
some writers. There were 565 sets of vertical velocity measurements
combined into forty-six series. The forty-six average curves were all
very flat and convex down stream--except near an irregular bank--and
were approximately parabolas with horizontal axes; the data determined
the parameters only very roughly; the maximum velocity line was usually
below the service, and sank in a rectangular channel, from the center
outward down to about mid-depth near the banks. Its depression seemed
not to depend on the depth, slope, velocity, or wind; probably the air
itself, being a continuous source of surface retardation, would
permanently depress the maximum velocity, while wind failed to effect
this, owing to its short duration. On any vertical the mid-depth
velocity was greater than the mean, and the bed velocity was the least.
The details showed that the mid-depth velocity was nearly as variable
from instant to instant as any other, instead of being nearly constant,
as suggested by the Mississippi experimenters.

The measurement of the mean velocity past a vertical was thought to be
of fundamental importance. Loaded rods seemed by far the best for both
accuracy and convenience in depths under 15 ft. They should be immersed
only 0.94 of the full depth. The chief objection to their use,
that--from not dipping into the slack water near the bed--they moved too
quickly, was thus for the first time removed. A double float with two
similar sub-floats at depths of 0.211 and 0.789 of the full depth would