riment was repeated numberless times, and invariably
with the same result. The sand contained about 40% of voids. The
deduction from this experiment is that the pressure of water is against
rather than through sand and that any excess of voids occurring adjacent
to a face against which there is pressure of water will be filled with
sand, excepting in so far, of course, as the normal existing voids allow
the pressure of the water to be transmitted through them.

[Illustration: PLATE XXVII, FIG. 1.--EXPERIMENT SHOWING PROPERTIES
OF SAND.]

[Illustration: PLATE XXVII, FIG. 2.--SAND PUSHED UP FROM BOWL BY
WATER PRESSURE THROUGH GOOSE-NECK.]

If, then, the covering of sand over a structure is sufficiently heavy to
allow arching action to be set up, the structure against which the
pressure is applied must be relieved of much of the pressure of water
against the area of sand not constituted as voids acting outside of the
arching area. This is confirmed by the two following experiments:

_Experiment No. 5._--The same apparatus was used here as in Experiment
No. 2, Fig. 7, except that the inside bucket had a solid bottom. The
inside and outside buckets were filled with water and the point was
noted at which the weight would balance the inside bucket at a point
some 3 in. off the bottom of the outside bucket. This point was
measured, and the bottom of the larger bucket was covered over with sand
so that in setting solidly in the sand the inside bucket would occupy
the same relative position as it did in the water. The same weight was
then applied and would not begin to lift the inner bucket. For instance,
in the first part of the experiment the weight stood at 12 in. from the
pivot, while in the next step the weight, standing at the end of the
bar, had no effect, and considerable external pressure had to be exerted
before the bucket could be lifted. Immediately after it was relieved,
however, the weight at 12 in. would hold it clear of the sand. No
attempt was made to work the bucket into the sand; the sand was leveled
up and the bucket was seated on it, turned once or twice to insure
contact, and then allowed to stand for some time before making the
experiment. No attempt was made to establish the relationship between
sands of varying voids, the general fact only being established, by a
sufficient number of experiments, that the weight required to lift the
bucket was more than double in sand having 40% of voids than that
req