156 



r* 



When uncorrected When corrected 

 for annual variation, for annual variation. 



From the 115 Results in Table I. . . 1-50 1-49 



From the 96 Results in Table II. 1-44 1'39 



From the 71 Results in Table III. + 1'5/. . . + 1-46 



+ 1-50 1-45 



The probable error thus obtained represents all the diversities 

 ascribable to the employment of different instruments (all of the one 

 construction), to the supposed peculiarities of different observers, 

 to the occasional presence of magnetic disturbance (for which no 

 correction has been attempted), and to differences due to different 

 hours of observation ; in addition to what may be more strictly 

 viewed as " observational errors." It may thus serve in some mea- 

 sure as a guide to those engaged in similar researches, as to the degree 

 of accuracy which is attainable in such experimental inquiries, when 

 proper care is taken in the procurement of a reliable inclinometer 

 and in its manipulation. 



For the purpose of comparing the probable error thus obtained 

 with inclinometers of the later English pattern with that of the 

 instruments of earlier construction, four of the latter were selected, 

 viz. a 9-inch circle by Robinson, a 9-inch circle by Barrow, and two 

 6-inch circles by Robinson, all in good order. Each circle was 

 furnished with two needles of the same length as the diameter of the 

 circle, and read by a lens in lieu of verniers and microscopes. Table IV. 

 contains the particulars of 20 determinations made with these instru- 

 ments in 1860 by Messrs. Stewart and Chambers. Their mean 

 result is 68 20''04 reduced to the epoch 1860.5, and corrected for 

 annual variation. The mean result of the 71 determinations at the 

 same epoch in Table III. is 68 1 9''8. There is therefore no notable 

 difference in the mean results obtained by the two classes of instru- 

 ments ; but there is a considerable difference in the probable error ; 

 as from the 20 determinations in Table IV. we obtain +3' f 65 as the 

 probable error of a single determination with the instruments of the 

 earlier pattern, whilst + l'*5 has been shown to be the probable error 

 when inclinometers of the more recent pattern were employed. 



