336 
BULLETIN OF THE UNITED STATES FISH COMMISSION. 
errors of this kind were always unnecessary, as the cathetouieter in this respect, as 
in all others, has iiroved to be a most perfect instrument. 
The comparisons of special forms of thermometers require slight modilicatious of 
the processes that have been described. 
Alcohol thermometers are much slower in changing temperature than mercurial 
thermometers, both on account of the less conductivity of the liquid and because of 
the adhesion of liquid to the walls of the stem. Time must be allowed for the liquid 
to drain down when observations are made under conditions of falling temperature. 
Maximum thermometers are not as a rule compared below freezing, and require 
a special apparatus for whirling them while packed in ice. The tenn>erature of the 
water baths should be allowed or caused to gradually increase while comparing maxi- 
mum thermometers. 
The upsetting deep sea thermometers, owing to the peculiar construction of the 
bulb, require prolonged exposure in ice; about a half hour is allowed, to make sure 
that the bulb is at a true freezing temperature. The bunches of these thermometers 
contain only six instruments, three on a side. The bunch is upset while still packed 
with ice, which is then partially removed for reading. At other temperatures only 
one large vessel of water is used. After about live minutes exposure to a, stationary 
temperature, the water being thoroughly stirred, the substandard is read, the bunch 
miset in the Avater, and readings made and recorded. 
REDUCTION OP OBSERVATIONS. 
Each substandard, after comparison with the air thermometer, is iirovided with 
a table of normal corrections, but these for convenience are comimted on the sujiposition 
that the correction at the freezing point is 0.0°, for the reason that, as already pointed 
out under “A third source of error,” on page 331, the correction at the freezing point 
is apt to change from time to time, especially after the substandard has been used at 
some high temperature, say 130° F., or more. When any doubt exists as to what 
corrections to apply to the indications of the substandard, it should be tested at the 
freezing point. Suppose we find the thermometer reads 0.23 of a degree too high, 
the true correction at the freezing point is therefore —0.23°, and this amount must 
be algebraically added to each correction given in the table of normals. 
The recorded readings of the substandard being corrected by the apiilication of 
the corrections found in the above manner, we have the true temperature of the water 
according to the air-thermometer scale during the several comparisons. The correction 
of a thermometer at the temperature of one of these observations is the quantity 
which must be algebraically added to its reading in order to make the reading indicate 
the true temperature of the water, that is, to make the reading agree with the corrected 
reading of the substandard. 
Thermometers require recomparison from time to time, but only at the freezing 
l)oint. Whatever change in the correction is found necessary at this point must also 
be made in each of the other corrections found in the original comparisons. This is 
explained in the remarks about the corrections to be used with the substandard. 
When a greater or less portion of the mercurial column in the stem of the ther- 
mometer is exposed to a different temperature than the bulb, the indications are 
slightly in error and a correction should be applied. This is sometimes necessary in 
the use of long substandards. It is to be avoided as far as possible. 
