568 MR EDMUND J. MILLS’S 
depends upon adjusting the thread so as to have only a slight error, and then 
immediately correcting for this error. It is carried out as illustrated in the 
accompanying section of a thermometer. Suppose the first position of the 
thread to have been, as shown, 0—26°9 on the scale. By sucking a little air 
from the right hand india-rubber tube, or by blowing a little air into the left 
hand tube, the thread is now caused to occupy, say 29°6—54'9 on the scale ; 
that is, it has been placed 2:7 divisions too 
much to the right. The corrected position is 
therefore 26°9—522; and the lengths cor- 
responding to the two positions are 26-9 and 
25°3 successively. In practice the misplacement never need exceed ‘2 or ‘3 
millimetre, a quantity that can be subtracted without introducing any sensible 
inaccuracy. It is advisable, in order to neutralise constant error, to make 
one entire calibration to the right, and one to the left; the mean of these 
should be adopted. 
The degree value of the calibration unit is, of course, ascertained by 
determining how many times it is contained in a scale whose extreme points 
are the melting-point of ice and the boiling-point of water. LaApuacr’s standard 
atmospheric pressure* for 100° C. is 
Ce FS | (eS 80 
oon aon! ee) Ao | 
Fig. 1. 
{760 + 0001492H + 1:946 cos 2\} millimetres ; 
H being the height in metres above sea-level, and \ the latitude. In my own 
experiments, which were performed in London, the temperature corresponding 
to 760 millimetres was found, by the aid of this expression, to be 100°°012. 
On account of the secular diminution in the capacity of the bulb (which 
diminution is much greater at zero than at 100°) the value of the calibration 
unit may undergo a measurable diminution. The following is a comparison of 
five values of 7, taken at a considerable interval of time apart, in the expression 
New unit=z2 old unit. 
Thermometer. 2 3 6 454 C. 
2. ‘99945 ‘99789 1:00013 ‘99986 99997. 
Thus in all the above cases, excepting one, there has been a diminution in the 
value of the unit; but the increase for thermometer 6 lies well within a fair 
allowance tor experimental error.t 
* Mitier, Phil, Trans, (1856), p. 775. 
t A series of measurements of changes in the scale value of the 0° - 100° interval was made by 
Prerre, Ann. Ch. Phys., 1842 (444). 
