Determination of the Earth's Horizontal Magnetic Force. 383 



concluding that the larger mean /x possessed by group A as compared 

 to groups B, C, or D, is due to the possession of a larger volume by the 

 first-mentioned group. The pre-eminence of the mean fx of group E is, 

 however, explicable only by supposing that the magnets of that group 

 possess a considerably larger permeability than the others ; and this 

 larger permeability, as we see from Table II, is a property not of one or 

 two magnets, but of the whole group. 



On the whole, as Table II shows, /x is much less variable in the 

 several groups than q or q. 



The greatest recorded value of viz., 11*89, occurs in group D. I 

 feel, however, some doubts respecting this and a second large value, 

 9*72, found for a second magnet of the same group. The two magnets 

 for which these values are recorded were tested in 1864 ; and one of 

 them, when retested many years afterwards, had a /x less than 7. If we 

 excluded these two magnets, we should find for group D a mean /x much 

 the same as that of group C. 



§ 1 3. Coefficient P. — In about half the cases P was determined from 

 only one observation, and under such circumstances the probable error 

 is considerable. Still I do not think that the mean values given in 

 Table I can be much in error. It will be noticed that no magnet of 

 group B gave a positive P, and that no magnet of groups E or F 

 gave a negative P. In groups A, C, and D, positive and negative values 

 occur in fairly similar proportions ; but investigation showed a great 

 predominance of negative values amongst the older magnets of group 

 C, while in group D no positive value appeared in the six oldest mag- 

 nets. As pointed out in § 5, P depends on the deflected or mirror 

 magnet as well as on the collimator magnet, and I am inclined to ascribe 

 the interesting difference between older and newer unifilars mainly to 

 change in the pattern of the mirror magnets (see § 46 later). 



As appears from either Table I or Table II, P when negative is 

 usually numerically smaller than when positive. When an observer 

 employs a unifilar strange to him the probable error in P is doubtless 

 larger than when the instrument is one to which he is thoroughly 

 accustomed, and it would not be unreasonable to attribute to this 

 cause some of the large values recorded in Table II. It happens, how- 

 ever, that three of the largest values of P in group G were each based 

 on three independent observations. The results of the individual 

 experiments were as follows, the numbering being purely arbitrary : — 



Unifilar. 



i. 



ii. 



iii. 





f + 2415 



-1439 



+ 1031 



Values of 10 5 x P/r 2 at 30 cm. 



■I + 2870 



-1917 



+ 872 



L + 2303 



-1107 



+ 904 



