524 The Rev. T. E. Robinson's Experimental Researches on the 



when ithas been in part destroyed, it is not restored entirely, even by a set of 

 five in which L is four or five times its maximum. The three bracketed pairs 

 are examples of this.* The relations to i deserve attention : from i^ = 742 to 



1^ = 98, — r is nearly constant, but below this a new order of dependence 



arises, in which \ enters. Thus -f =98 adds only 5-96 to \= 11-78 ; if the 

 current be reversed, instead of reversing the polarity and giving — 1774 (the 

 sign — when applied to L meaning that what had been a north pole is now a 

 south), i continues positive =+ 9"04, 2'74 less than \ ; and when the reverse 

 current is withdrawn, it is found that no permanent change in the condition of 

 the magnet has been made, for \ is unaltered. With i^=196, L is STIS, a 

 much greater proportional increase ; with — i^ = 201 it is still positive, = + 2'48, 

 but the molecular state has been permanently altered, for \ is only 5'72, and 

 the reapplication of -f = + 201 does not raise it above 8'64 with an L = 26*40. 

 The same happens with the ± applications of i^=287 and 331, except that 

 with the last L becomes negative, while \ retains the small positive value 

 + 0"54. The two negative values, — 377 and — 482, give \ negative, though far 

 less than the values obtained by the same when positive, and the latter at once 

 reproduces the full positive \. I regret that I did not ascertain what negative 

 ^ would be required to give \ its full negative value. All this confirms the 

 inferences which I formerly made from the observations with the hollow mag- 

 net, that X is a function both of -^ and \. 



The iron parts of the magnet were then removed, and the steel keeper and 

 base substituted. This alters the conditions, not merely because their induc- 

 tion is less than that of the iron ones, but also because they too become perma- 

 nent magnets, and add the effect of this to that of the cylinders. In the first 

 instance the temperature correction was sought, 1^ = 1282'92. 



* The brackets in the column of \^ connect those which were used in computing — . Those 



were selected in which X might be supposed to retain its normal value, as in them only the mole- 

 cular condition of the steel can be considered definite. 



