176 A. TANAKADATE. 



Since the results of observations are subject to various sources 

 of errors, and the empirical expressions of mean isomagnetics can 

 never be exact representatives of magnetic state of the whole 

 country, the disturbing forces calculated as above are attended 

 with greater uncertainty than the magnetic elements themselves. 

 Hence much discretion is required in drawing any inference from 

 them. 



We have seen that the mean probable errors of a single 



y 



observation are ± GM6, ± 5.'47, ± 73. 2, for declination, dip and 

 horizontal intensity, while those of the calculated values depend 

 upon the co-ordinates of stations, amounting in extreme cases to 

 more than double the above magnitudes. As already remarked^ 

 however, those probable errors are rather due to the disturbing 

 forces themselves than to errors of observations ; it would seem 

 that they can be more relied upon than is indicated by the sole 

 assumption of promiscuous occurrance of erroi's in the applied 

 method of least squares. The directions of those disturbing forces 

 are still more uncertain than their intensities, especially in places 

 where they are small, in the extreme case of which they become 

 altogether indeterminate. 



Distribution of these forces in different regions, when the 

 stations are taken in sufficient number to represent the main 

 characteristics, can not fail to be of interest for the physics of the 

 earth's crust. As they now stand, it is difficult to co-ordinate 

 them with anything like satisfaction. The rapidity with which 

 they vary from place to place, as is seen in the neighbourhood of 

 Huzi and Asama (see Map 10), shows us what a rough approxi- 

 mation we come to by simple interpolation or inspection. Crude 

 as they are, they may be better than nothing, and when studied 

 in connection with collateral facts and interpreted with proper 



