28 PHILOSOPHICAL SOCIETY OF WASHINGTON. 
minus the reading of its depressed freezing point. This, which 
should be exactly one hundred degrees, rarely isso. When the fun- 
damental distance is taken in this way it is invariable with age. It 
is the same forty years after the thermometer is made as four hours 
after, provided the thermometer is kept at ordinary temperatures. 
The fundamental distance is not invariable when the raised freez- 
ing point is used in forming it instead of the depressed freezing 
point. In this case there is a constant diminution with age as the 
raised freezing point rises more rapidly than the boiling point. 
There is not uniformity of practice in the matter of forming the fun- 
damental distance, but it is greatly to be desired. 
Heating a thermometer to 350° causes a permanent increase in 
its fundamental distance between depressed freezing point and boil- 
ing, varying from 0.°4 to 0.°9. An increase of 0.°4 in the funda- 
mental distance corresponds to a decrease of sc part in the coefii- 
cient of expansion of the glass. 
The total correction of a mercurial normal thermometer for 
errors in its construction is composed of three parts: 
1st. The correction for erroneous fundamental distance. For 
any temperature this is a proportional part of the differ- 
ence between the fundamental distance and 100°. 
2nd. The calibration correction. This is the correction to the 
scale marks considered as subdividing the capacity of the 
tube from 0° to 100° into one hundred equal parts. It 
involves variations in the bore of the tube as well as irreg- 
ularities in the placing of the marks. 
3rd. The correction at freezing point. This is the amount the 
thermometer reads in melting ice above or below 0°. At 
any time it is the observed reading of the thermometer in 
melting ice immediately after exposure to the temperature 
measured. Sometimes it is impossible and it is almost 
always inconvenient to observe the freezing point of the 
thermometer immediately after observing a temperature. 
In such a case the position of the depressed freezing point 
for that temperature must be computed from the law of 
the variation of the freezing point. It is always prefera- 
ble however when the highest accuracy is required to ac- 
tually observe the freezing point. . 
When a thermometer is put in ice this is what happens: The 
column falls rapidly at first, then more slowly. Presently it be- 
