336 Chronometers. 
vanced towards it by less than one fourth part of the quantity, shew- 
ing that the variation in the rate of going arose chiefly from a dimi- 
nution of the elastic force of the spring, as seen in the following Ta- 
BLE— 
{ Thermom. Rate per hour. Remarks. Experiment. 
s 
32° + 5.74 Length of steel spring, 11.04 in. 1 
66 — 1.80 Do. 2 
100 — 10.30 Do. 3 
100 — 7.0 Spring shortened, .01 in. 4 
. Having shortened the spring for the fourth experiment, by .01 inch, 
if the difference of rate had depended on its length, it would have 
shewn the original rate of +-5*.74 at 100°; instead of this, the rate 
was —7®.0, differing only 3°.3 from the former at the same tempera- 
ture. To the decrease in length of .01 inch, therefore, is due the 
loss of 3°.3, while to the loss of elasticity is due the difference be- 
tween 3%.3 and 16%.04, the whole difference of rate for 689°, i. e. 
+ 58.74, at 32° and —10s.30 at 100°. 
The fact having been proved, that glass does not lose its strength 
heat in the same ratio as metals, and being now acquainted with 
the extent of its loss, which was 40s for 68° of Fahr. we next had 
to construct a balance suitable to correct so small an error, and our 
previous experiments having pointed out to us that a metallic bal- 
ance was unfit for such a purpose, we continued to employ the glass 
disk 4, as shewn in the annexed diagram— 
to which was attached the glass balance-spring a. ‘To the disk we 
applied in an horizontal position, two lamin: composed of platina, 
