625 



in successive horizontal lines, and the hours of the day in vertical 

 columns. The " means " of the entries in each vertical column indi- 

 cate the mean direction of the magnet at the different hours of the 

 month to which the table belongs, and have received the name of 

 " First Normals." On inspecting any such monthly table, it is at 

 once seen that a considerable portion of the entries in the several 

 columns differ considerably from their respective means or first nor- 

 mals, and must be regarded as " disturbed observations." The laws 

 of their relative frequency, and amount of disturbance, in different 

 years, months and hours, are then sought out, by separating for that 

 purpose a sufficient body of the most disturbed observations, com- 

 puting the amount of departure in each case from the normal of the 

 same month and hour, and arranging the amounts in annual, monthly, 

 and hourly tables. In making these computations, the first normals 

 require to be themselves corrected, by the omission in each vertical 

 column of the entries noted as disturbed, and by taking fresh means, 

 representing the normals of each month and hour after this omission, 

 and therefore uninfluenced by the larger disturbances. These new 

 means have received the name of " Final Normals," and may be de- 

 fined as being the mean directions of the magnet in every month and 

 every hour, after the omission from the record of every entry which 

 differed from the mean a certain amount either in excess or in defect. 

 In this process there is nothing indefinite ; and nothing arbitrary 

 save the assignment of the particular amount of difference from the 

 normal which shall be held to constitute the measure of a large dis- 

 turbance, and which, for distinction sake, we may call " the separating 

 value" It must be an amount which will separate a sufficient body 

 of disturbed observations to permit their laws to be satisfactorily 

 ascertained ; but in other respects its precise value is of minor sig- 

 nificancy ; and the limits within which a selection may be made, 

 without materially affecting the results, are usually by no means 

 narrow ; for it has been found experimentally on several occasions, 

 that the Ratios by which the periodical variations of disturbance in 

 different years, months and hours are characterized and expressed, do 

 not undergo any material change by even considerable differences in 

 the amount of the separating value. The separating value must ne- 

 cessarily be larger at some stations than at others, because the abso- 

 lute magnitude of the disturbance- variation itself is very different in 



