1870.] 



and Horizontal Force at Kew Observatory, 



237 



On the whole, therefore, we have good evidence of a behaviour at Kew 

 analogous to that at Toronto. 



11. The probable error of a single monthly determination of the dip, de- 

 rived from the seventy- two monthly determinations given in Table I., and 

 after the application of the correction for secular change and annual varia- 

 tion, as derived from the results of these observations, has been made, is 

 + 0'*96. There is, however, reason to believe that this probable error is 

 increased to some extent by periods of disturbance, some of them of cons 

 derable duration, which exhibit themselves when the residual errors have 

 been deduced after the method indicated above. In order to test this, I 

 have formed a series of seventy-two yearly values of the dip corresponding 

 in epoch to the various monthly values of Table I. 



These yearly values will, of course, average the semiannual inequality, 

 while each yearly value may possibly be supposed to be affected to some 

 extent with the same sort of disturbance which affects the monthly value 

 to which it corresponds. Were both affected in precisely the same way 

 by these disturbances, the differences between the monthly and yearly 

 values would only be occasioned by the semiannual inequality and by 

 errors of observation. It is, however, too much to expect that all effects 

 of disturbance will be eliminated from the differences by this method ; 

 nevertheless we may naturally expect that they will be reduced in amount. 



12. Grouping these differences together in six monthly periods, we ob- 

 tain the following results corresponding to those given in Table II : — 



Table V. 



Date. 



Observed minus Calculated. 



April to September. 



October to March. 



Julv 1, 1863 



January 1, 1864 



July 1, 1864 



July 1, 1865 



January 1, 1866 



July 1, 1866 



January 1, 1867 



July 1, 1867 



January 1, 1868 



July 1, 1868 



January 1, 1869 



— 0"20 



-o-i 3 



-+-CV26 



-071 

 -0-95 



— 0-14 



-0-05 

 -0-15 

 4-0-32 

 4-0-68 



4-o-55 

 4-0-30 



Mean 



-0-31 



4-0-28 



It will be seen from this Table that the irregularities of the two last 

 columns of Table II. are very much reduced by this process, while the 

 result remains nearly the same. 



The probable error of a single observation is also reduced, and becomes 

 (when the correction for annual variation is applied to the individual dif- 



