RESEARCHES IN THERMOMETRY. 569 
It is clear that, for all accurate work, the degree value of the calibration 
unit should from time to time be re-determined. Where this cannot be done, 
as for instance in geographical expeditions (in which thermometers are used 
to determine heights, by boiling-point observations), the results are to a certain 
extent untrustworthy. | 
II. THe Exposure CoRRECTION. 
If the indicating portion of a thermometer be only partially immersed in a 
medium whose temperature it is intended to exhibit, there will be a + correc- 
tion, according as the temperature of the air is below or above that of the 
medium. This is generally determined by placing the bulb of a subordinate 
thermometer half way up the exposed or free portion of the indicating column, 
and then calculating out the value of y in 
y =-0001545(T—2N. 
In this expression y is the correction ; ‘0001545 is the difference between the 
coefficients of cubical expansion of mercury and glass ; T, ¢ are the readings of 
the principal and subordinate thermometers in centigrade degrees ; and N is 
the total number of scale degrees ex- 
posed.* ReEGNAULT, who devised this 
formula, did not verify it by experiment. 
I have therefore instituted a series of Sereen 
trials in which the numerical coefficient 
above specified is regarded as a quantity 
a to be determined. 
The method of operation consisted 
essentially in placing a long and delicate 
thermometer vertically in a current of 
steam, the thermometer being held in 
the same position throughout, but vary- 
ing in the depth of its immersion in the 
medium. The apparatus, of which a 
drawing is given below, consists of a thin 
glass flask holding at a about 1°4 litres 
of boiling water, to which a trace of zinc 
dust is added. From this proceeds a 
piece of hard glass tubing, 1°4 centimetres in diameter, encased in thick 
india-rubber tubing. Through the top of the glass tube passes the long 
Screen 
* Mémoires de V Acad. xxi, p. 225, 
