care, and with certain costly precautions, (including
precisely graded wooden floors,) which could hardly be expected in
private work.

9 The tile has been said, by great authorities, to be broken by
contraction, under some idea that the clay envelops the tile and
presses it when it contracts. That is nonsense. The contraction
would liberate the tile. Drive a stake into wet clay; and when the
clay is dry, observe whether it clasps the stake tighter or has
released it, and you will no longer have any doubt whether expansion
or contraction breaks the tile. Shrink is a better word than
contract.

10 Taking the difference of friction into consideration, 1-1/4 inch
pipes have fully twice the discharging capacity of 1-inch pipes.

11 No. 5 was one inch in diameter; No. 4, about 1-1/3 inches.

12 If the springs, when running at their greatest volume, be found to
require more than 1-1/4-inch tiles, due allowance must be made for
the increase.

13 Owing to the irregularity of the ground, and the necessity for
placing some of the drains at narrower intervals, the total length
of tile exceeds by nearly 50 per cent. what would be required if it
had a uniform slope, and required no collecting drains. It is much
greater than will be required in any ordinary case, as a very
irregular surface has been adopted here for purposes of
illustration.

14 The stakes used may be 18 inches long, and driven one-half of their
length into the ground. They should have one side sufficiently
smooth to be distinctly marked with red chalk.

15 The depth of 4.13, in Fig. 21, as well as the other depths at the
points at which the grade changes, happen to be those found by the
computation, as hereafter described, and they are used here for
illustration.

16 The figures in this table, as well as in the next preceding one, are
adopted for the published profile of drain _C_, Fig. 21, to avoid
confusion. In ordinary cases, the points which are fixed as the
basis of the computation are given in round numbers;--for instance,
the depth at _C3_ would be assumed to be 4.10 or 4.20, instead of
4.13. The fractions given in the table, and in Fig. 21, arise from
the fact that the decimals are not absolutely correct, being carried
out only for two figures.