xlvii INTRODUCTION TO THE MAKERSTOUN OBSERVATIONS, 1843. 
ture experiments it is 4°9 micrometer divisions ; the mean of the two series in 1844 
gives g’ = 7:91, while the temperature experiments give it 3°8 microm. diy. 
65. The following Table contains the results deduced in three cases when re- 
spect has been paid to the sign of ¢, — ¢,,,3 the last column contains the value of g’, 
— 24 (y) 
obtained by giving equal values to the quantities Za (0 and — Za (0 
TaBLE 17.—Results of Comparisons, regard being paid to the sign of ¢,—t,,,. 
by > bose bo < fosr 
za) |-2a(@) S| —2a@| 24) 
Mean 
Value of 
1843. 5 Mic. Div. Mie. Diy. of Mic. Div. 
June 1—July 22 4003-4 335238 8-37 19518-9 | 163340-7 
1844. 
May 9—May 24 5404-3 37559-9 6:95 3011-1 29061-8 
Aug. 3—Sept. 18 6409-0 52022-7 8-12 16295-4 | 134950-6 
For all the periods || 15816-7 123106-4 7-783 38825-4 | 327353-1 
66. The three final results obtained are g’ = 7-832, q’ = 7-903, and q/ = 8:11; 
the adopted value of g’ = 7-9 micrometer divisions. 
67. Early in the investigation of this subject it occurred to me that it might 
not be desirable to eliminate all the effects of change of temperature on the posi- 
tion of the needle, as the actual daily or diurnal variations of the earth’s magnetism 
might depend to some extent on the same cause. The results, obtained in many 
ways which it is not necessary to repeat wholly here, shewed that, at least for periods 
of a month, no such connexion exists, or that it is inappreciable. To such an extent 
had the supposition been refuted, that it was ultimately totally forgotten by me, 
and, of course, it was also forgotten that others might entertain a similar suspicion ; 
the fact that it has been proposed as an objection to the method will render it de- 
sirable that I should adduce distinct evidence of its futility. 
68. As it will scarcely be supposed by any one that the earth’s magnetism 
varies immediately with changes of the aérial temperature, the results obtained from 
the comparison of one day with the next following may be first considered. As far 
as the results, Table 15, go, the comparisons of one day with the next following 
give a considerably larger value of q/ than that adopted, and, therefore, one differ- 
ing still more from the value obtained by the usual method; this at least tends to 
prove that the result adopted is much better than that by the other method. In the 
following, which also prove the same fact, one-sixth of the comparisons are of the 
readings on days with those on the second days following, namely, Saturdays with 
