emphasise the necessity for uniformity in the mode of working. Whether a
process yields proportional results, or not, will be seen from a series
of standardisings. Having obtained these, the results should be arranged
as in the table, placing the quantities of metal used in the order of
weight in the first column, the volumes measured in the second, and the
standards calculated in the third. If the results are proportional,
these standards will vary more or less, according to the delicacy of the
process, but there will be no apparent order in the variation. The
average of the standards should then be taken.

+-------------+---------------+----------+
| Weight. | Volume found. | Standard |
+-------------+---------------+----------+
| 0.2160 gram | 72.9 c.c. | 0.2963 |
| 0.2185 " | 73.9 " | 0.2957 |
| 0.2365 " | 79.9 " | 0.2959 |
| 0.2440 " | 82.3 " | 0.2964 |
| 0.2555 " | 85.9 " | 0.2974 |
+-------------+---------------+----------+

Any inclination that may be felt for obtaining an appearance of greater
accuracy by ignoring the last result must be resisted. For, although it
would make no practical difference whether the mean standard is taken as
0.2961 or 0.2963, it is well not to ignore the possibility that an error
of 0.4 c.c. may arise. A result should only be ignored when the cause of
its variation is known.

In this series the results are proportional, but the range of weights
(0.216-0.2555 gram) is small. All processes yield fairly proportional
results if the quantities vary within narrow limits.