OF Til E MAG \ KT I (' FO RC E. 



The monthly means arc contained in the following table: — 



13 



Correction for progressive change in the readings. — The observations having been 

 referred to a uniform temperature, still require a correction for the effect of the 

 progressive change during each month before Peirce's criterion can be applied for 

 the purpose of separating the disturbances. We have seen that the mean monthly 

 value of this change due to loss of magnetism of the bar and to change in the 

 horizontal force itself, was 17.6 scale divisions; on the average, therefore, a correc- 

 tion must be applied to the observations on the first and last day of each month of 

 + 8.8 and — 8.8 scale divisions, and in proportion for the intermediate days. At. 

 Toronto, also, the progressive change in some months was so great as to present a 

 practical difficulty by its interference with the proper comparability of the observa- 

 tions, and in these cases new means at shorter intervals than a month were taken. 



1 The actual mean of 17 days was 203 ; to reduce this to the mean of 27 days, 19 scale divisions 

 were subtracted, resulting from an interpolation between January 1st and January 12th ; the mean 

 of 7 days preceding and following the gap was made use of. 



" Owing to causes already explained, the means of May and June differ so much as to affect the 

 continuity of the series ; the same is to be said of the differences between June and July, 1840, and 

 between December, 1840, and January, 1841 ; the corresponding differences between the same months 

 in the other four years furnish us with the means of correcting the scries for the first year, as will be 

 seen hereafter; it also appeared advisable to omit the readings in June. 1840, altogether, the instru- 

 ment not having then been in stable adjustment. 



3 The numbers in table 11 have been slightly changed, to refer the mean of the hour of observation 

 to the mean resulting from observation of 12 hours a day. Comparing the mean at 14'' 22"' in each 

 month with the respective monthly means in the other four years, the above corrections became — 5, 

 — 5 and for January, February, and March. 



The bar between September and October, 1st:;, separates the means from the bi-hourly and the 

 hourly series. 



In the application of the reduction for temperature no attempt whatever has been made at inter- 

 polation in the magnetic series, but whenever a temperature reading was accidentally omitted, it has 

 been supplied by comparison with the observed temperature immediately preceding and following 

 No magnetic nailing can be supplied by interpolation, however short the interval, as long as the law 

 of the occurrence of the disturbances remains unknown 



