ITS ERRORS AND CORRECTIONS, ETC. 479 
TABLE VII.—Sums or DIFFERENCES OF BIFILAR DAILY MEANS at St HELENA 
FOR BACH Monta, 1842-48. 













Neen Pern. All. Above 0°5 Fahr. 
Months, 1842-43. perature 
of Bifilar. At. ZAR. ZAR. qV 
oo 4 Se. div. Se. div. | Sc. div. 
January, . 67:6 22°50 46°37 22:54 1-41 
February, . 69-4 21:92 33°65 19:38 Wilt? 
larch, ~. -. 69°8 26°49 47-11 32°38 1:54 
ys a 68-7 20-91 1:22 8°32 0:59 
May,  *. 66:1 23°82 17:63 13:98 0:86 
ounce, % . 62:5 25°45 21:04 14:09 0:76 
Daly 60°8 25:08 ,—14:31 6°80 0°36 
pugust, . . . 60-6 24°63 13-79 15°63 0:88 
September, . 60°7 24°89 26°72 18°81 1:05 
October, 61:3 21°36 | 38°67 16°46 1:29 
November, . 63°3 23°01 29-17 15°65 1:06 
December, . 65:2 27°57 33°02 26-06 1:13 





The resulting values are :— 


From all, 1842, q’= Lies = 0-980 scale divisions. 
1843, g’= aoe T-066...,, 
Both years, q’= ee 1-02 
From all above 0°°5 Fahr, 1842, g’= He == 046.0... 
1843, ¢’= ae = 0-980 ... 
Both years, diasarcn = 1-018 se, div.; probable error = 0-071. 
27. An examination of the values of ¢’ for the different months seems to show 
a variation with the season; the variation is shown also in the results for the 
separate years. If we discuss the quantities in the last column of Table VII. by 
the method of least squares with reference to the mean temperatures of the bifilar, 
we find the probable variation of g’ for 1° Fahr. to be 0:0371 sc. div.; or if ¢ be 
the temperature of the bifilar, that we may put, 
q' = 0°84 + (t—60) 0-0371. 
Therefore at the mean temperature of bifilar 65° 
q/=1-02 se. div., with a probable error of 0:06 sc. div. 
The coefficient for the temperature variation of gq’ is very large, probably three 
times greater than has ever been found by water experiments for any magnet: 
VOL. XXII. PART III. . 6H 
