474 MR JOHN ALLAN BROUN ON THE BIFILAR MAGNETOMER, 
The mean scale reading having changed about 9 sc. div. from the 7th to the 
20th, a correction has been applied after the 7th, for the change supposed regular 
at the rate of 0°67 sc. div. a-day ; the corrected quantities are given in the third 
column. A correction has been allowed for a shock to the instrument on the 
11th, depending on the change at the hour of the shock ; and the change 10th to 
11th is on this account probably inaccurate. This series is, on the whole, con- 
sistent, an increase of temperature being accompanied by a diminution of scale 
reading (or vice versa) on every day with one exception ; the final result is+ 1-0° 
Fahr., equivalent to —2°5 sc. div. This also satisfies the hypothesis. 
18. Perhaps the most curious fact connected with this series of observations 
is, that though the daily mean scale reading increased, torsion retrograde, with 
an increase of temperature, and torsion direct, with a diminution of temperature, 
there was in both cases a strongly and regularly marked hourly variation, whose 
direction was not changed by the direction of the torsion, and whose amount, as 
related to the temperature, was much greater than that obtained from the daily 
means. I shall have to consider this fact at another time: it does not, however, 
depend upon the bifilar suspension. Whatever conclusion we may arrive at with 
reference to the relation of these facts to the direction of torsion, it appears to me 
they are sufficient to prove that any trustworthy temperature coefficient must 
include their causes. 
19. There still remains one considerable objection to the usual method of 
obtaining the temperature coefficient of the bifilar magnetometer; namely, that 
before we can correct the observations we must be certain that we have obtained 
the true value of the scale unit coefficient. Should that be at all inaccurate (as it 
frequently has been when derived from the angle of torsion only), the corrections, 
even by a true temperature coefficient, will be inaccurate. Thus, if , be the scale 
reading from the zero, & the unit, and g the temperature, coefficients, the reduced 
result is 
AX 
x k+t,q: : : : ; : Gy 
where ¢, is the number of degrees Fahr. from the zero. If 7, be the part of m, due 
to change of the earth’s magnetism, and 7, the part due to temperature, 
AX 
x =nk +n, k + tog. ‘ : : ‘ ; (17) 
the last two quantities should destroy each other if & and g be the true coefii- 
cients. If, however, /.>=4+« be the true unit coefficient, then 
nok, = —t%) 
ime mem SS i, toe sole Ugg ag Raaer 
Substituting this value in equation Gris 
AX Cea 
x =n, kK—e@ is t e A 3 : : : (19) 
)) 

