b 



Magnetic Declination. 



339 



days in each month the magnet proceeds slowly and regularly from one known singular point to another, the 

 deviations from this regularity being altogether of the second order with respect to the principal motion. 

 2d, On other days, the regularity of the first class is suddenly interrupted, for a short period, by exciu-sions 

 of considerable magnitude, after which the previous regularity is continued. 3d, The last class of motions is 

 wholly irregular for large portions of the day, and consists of alternate deviations to the east and west of the 

 mean position within short periods of time. The fact that there are certain days in each month during which 

 the motion of the magnet seems to obey a simple and continuous law, while the disturbances which destroy 

 the regularity on other days are so obviously discontinuous or intermittent, occurring very seldom in some 

 months, points out a method which, if not wholly free from objection, has the advantage of simplicity, and 

 offers, it appears to me, a very near approximation to the truth. This method consists simply in the selection 

 of the days which comprehend the first class of motions. 



Having carefully examined the nature of the diurnal variation for each day, I at first selected in each 

 month the 10 days on which the effect of the intermittent disturbances appeared to me least. In the same 

 manner, the 5 days with the smallest apparent u-regularity in each month were selected. As the mode in 

 which these selections were made was to some extent arbitrary, I desired Mr Welsh to make similar selections 

 independently. Upon comparison, it was found that there were 17 days in 120, and 13 days in 60, for which 

 we differed, and that the difference in almost every case occurred on days of which the preferability was very 

 doubtful. It will be very obvious, however, from the coincidences of the results for the 5 days and 10 

 days selected, that a difference in choice of a few nearly equally good days, is altogether immaterial. The 

 following Table contains a list of the days selected, which differ only in three instances in each class from the 

 selections at first made by myself, and employed in forming the Tables ; the days adopted from Mr Welsh's 

 selection in preference to my o^vn, were, March 1 , for March 27 ; November 5, for November 8 ; December 

 25, for December 23, in the 10-day class: and April 12, for April 10; August 28, for August 21 ; Decem- 

 ber 7, for December 3, in the 5-day class. 



TABLE VII. — List of the Ten Days and Five Days in each Month of 1844, which have been 



selected as being the least disturbed. 



The * indicates that the day has been selected as one of the 5 days of least disturbance. 



The following Tables contain the hourly means for each month, as deduced from the 10 days and from 

 the 5 days of most regularity in each month. 



