ON COMPARISON AND REDUCTION OF MAGNETIC OBSERVATIONS. 211 



twelve sets, each of twenty-five days, had been selected at random, the 

 mean diurnal variation got out from either set of 150 days would in all 

 probability have appeared almost truly cyclic, and the mean diurnal 

 vai'iations got out from the twelve sets of twenty-five days might have 

 been expected, if not truly cyclic, to show about equally many increments 

 and decrements of the element considered in the twenty-four hours. 

 How far the actual result difiers from this ideal is shown by the 

 following table. 



The ' total increment ' means the algebraic sum of excesses of the 

 values of the element considered at the second midnights of the twenty- 

 five quiet days in a month over the values at the first midnights ; in 

 other woi'ds, it equals twenty- five times the mean increment during a 

 'quiet ' day in the month specified. The five years supply of course five 

 Januarys, ifcc, or sixty individual months in all. 



Table II. 



According to Table II. if every day behaved like a ' quiet ' day there 

 would be an annual increase of 35' in westerly declination, instead of the 

 actually existing decrease of 7', and an increase of 1205x10"'' C.G.S. 

 units in H.F., instead of the actually observed 20 x 10"''. 



Taking the figures as they stand, the evidence in favour of a general 

 tendency to a non-cyclic variation in the direction of increasing westerly 

 declination during ' quiet ' days is perhaps not conclusi^'e. Of the indivi- 

 dual months' records twenty-four show a decrease, and the balance the other 

 way might be pure chance. The figures relating to the horizontal force, 

 however, unless ascribable to error, prove conclusively that during the five 

 years in question the force increased at a wholly abnormal rate on ' quiet ' 

 days. It is difiicult to imagine any cause of error -.vorking so uniformly 

 throughout the year. Supposing uncomj^ensated temperature effect to 

 exist, for instance, the phenomenon should tend one way during one 

 season, the opposite during another. The interval between succepsive 



r2 



