88 
On the Errors of Thermometers. 
[March, 
By farther bisections, other points may, likewise, be found ; but as, on both 
centigrade and Fahrenheit, these would correspond to fractional parts of degrees, the 
following mode may be adopted : divide one of the distance portions before found, 
suppose the lower, into 5 equal parts by compass, the upper division will then be the 
approximate 20— let it be assumed for the present as correct ; then separate, as before, 
a thread about 40° long, and with this bisect the tube between 20 and 100— What- 
ever error might be at the 20 will thus be reduced one half. In like manner, bisect 
between the 60 and 100, and we get the 80 point with only $ of the error at the 
approximate 20 ; the error is thus reduced to a quantity almost insensible. The 80 
may, therefore, be assumed to be correct. In the same manner get 40 by bisecting 
between 0 and 80, and then bisect each of these portions, and we have correctly 20, 
40, 60 and 80. In the same manner we may, if we choose, get the tens and fives. 
Each of these points is obtained as correctly as the original freezing and boiling 
points, and thus the scale is graduated according to the capacity of the tube, inde* 
pendant of any partial irregularities. In this way I have graduated different thermo- 
meters, whose indications agreed perfectly through their whole extent. The same 
method is applicable, with equal facility, to ascertain the position of the degrees 
above the boiling point, and below freezing. If the operator have a little dexterity, 
he will do the whole in a very short time. In the course of a single day a good 
workman may thus ascertain the 25, 50 and 75 points, on a very great number of 
thermometers, each independent of any thing but the extreme points of its own scale, 
and with less trouble than by a careful comparison with a register which, very pro- 
bably, may be incorrect. I have found thermometers graduated from a register 
having scales sensibly unequable, though the bore of the tube, ascertained by this me- 
thod, was to all sense perfectly so, and others having equable scales with unequable 
bores. The thermometers which I examined in this way, were made by Cary, New- 
man, and Jones ; in every one of which I found serious discrepancies : so that I am 
inclined to believe, that the registers from which they are in the habit of graduating 
their instruments, would be found defective, if examined in this manner. 
I have been thus particular on this subject, for it is obvious, that whatever be the 
laws of expansion, &c. it is quite impossible that they can be correctly ascertained, if 
fhe instruments employed be essentially defective. I hope, therefore, that experi- 
menters will see the necessity of using more correct instruments than heretofore, and 
that workmen will see the necessity of paying more attention to the graduation of 
their instruments, laying aside the present system of assuming ■, without sufficient 
proof, one instrument to be correct, and then graduating others from it. 
The separated thread in the stem is very easily obtained of any required length, 
or finally joined to the mercury in the bulb. For example, if we want to increase 
the length of the detached thread, we have merely to cool the bulb, and then bring 
the detached portion in the stem into contact with that from the bulb ; then on 
heating the bulb the point of junction will not alter its position, while the mercury 
rises ; and when it is of the desired length above the point of junction, it will sepa- 
rate at that point on reversing the instrument. The detached portion may be lessen- 
ed by the reverse process, by first causing it to run to the upper end of the tube, 
and while in that position, cause the mercury from the bulb to join it; then, bycool- 
ing the bulb, the detached portion may be reduced, or exhausted, according to 
pleasure. A little practice will explain these operations much better than any 
veibal description, and I desire to be as brief as consists with perspicuity. 
