ITS ERRORS AND CORRECTIONS, ETC. 475 
the corrected result, which séz// varies with temperature. Where the scale unit 
coefficient can be accurately determined, this consideration has no value; but 
where we have to deal with observations already made, it is one of much im- 
portance. 
20. Temperature Coefficient.—Several facts, such as those already indicated, 
relatively to the effects of temperature on the parts of the balance magnetometer, 
induced me, in 1842, soon after I received charge of Sir THomMas BrisBANE’s 
Observatory at Makerstoun, to seek some other method of obtaining the tempera- 
ture coefficient in the terms of the scale divisions. This I sought first by arti- 
ficially heating and cooling the room in which the instruments were placed ; but 
the effects of currents of air thus engendered were too considerable to render the 
method satisfactory. I then sought to determine it by comparisons of the read- 
ings at the same hours of different days, when the temperatures were different ; 
I finally employed successfully the daily means, at first from incomplete diurnal 
series of observations in 1842-43. This method, by the formula 
AX q 
os amid +N, + FALy 
gave the result accurately corrected in scale divisions, from which the laws of 
variation could be determined even when ignorant of the value of £.* 
21. When this method was first used by me, I selected periods free from large 
disturbances, and combined the daily means in different periods.t In determining 
the coefficient for the colonial and other observatories, I have thought it prefer- 
able (to avoid any doubt that might arise from such selection as to the character 
of the final results) to obtain the coefficients from the comparisons of all the daily 
means without exception, from the commencement to the end of the periods for 
which observations have been published. In general, each daily mean of scale 
divisions and of temperature has been compared with the succeeding daily 
mean only ; but where the variation of mean temperature from day to day has been 
very small, comparisons with the second, and sometimes with the third day follow- 
ing, have been taken. The following are the discussions for the different bifilars :— 
* I first proposed to Professor James D. Forzzs (in a letter dated 21st November 1842) to 
determine the temperature coefficient from the usual hourly observations. (See Trans, Roy. Soc. 
Md, Vol, XVI. p. 74.) 
P.S.—Since this paper was written, I have received the third volume of the Toronto Observations 
(London 1857), where I am glad to perceive that General Sazinz has followed the principle described 
in the text, for the determination of the temperature coefficient. The mode he has adopted is how- 
ever far from accurate, including as it does irregular annual secular and instrumental changes, with 
too few results to render plus and minus errors equally probable; as a consequence, the coefficient 
obtained by him is still erroneous (1°63 sc. div., whereas I find 1°36 sc, div.), General SaBine 
notices my result for the Makerstoun bifilar (Adjustments, &c., part v., Toronto Observations, 
vol. iii.), but he omits to remark that I had pointed out, twelve years before, the necessity of applying 
the process to all bifilars, and that I had applied it to the very magnet (the Toronto one) under his 
discussion. See Trans, Roy. Soc. Ed., Vol, XVI. p. 77 (20), p. 102 (11), and curves in No. 5, 
Plate IV. 
t See the Introduction to the Makerstoun Observations for 1844 and 1845, where the process 
is described. 
VOL. XXII. PART III. 6G 
