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



to be 97-98 greater, and no cause of this is evident except the one assigned. 

 To obtain some idea of the probable effect of it in a more direct way, I intro- 

 duced two additional joints into the circuit. Two discs of soft iron were 

 procured, 2' diameter, and O''2o thick, with central apertures equal to those 

 of the magnet. They were marked so that they could be placed always in 

 the same position on the polar surfaces, and were accurately fitted by the 

 scraping process both to them and the keeper. If L'" be determined when 

 they form part of the magnet, and L when they are removed, and the magnet 

 is brought to the same length by lowering the slide 0"25, -i^ remaining the 

 same, L — L'" is the loss caused by the second pair of joints. The helices F, F' 

 were used always on the slide; ^ was as near as possible to 549'08, and re- 

 duced to that by interpolation. 



Table XXV. 



The difference is considerable, and both it and the change of ratio are 

 nearly inversely as c. I regret that it did not occur to me to experiment with 

 different values of ^ in order to ascertain whether this effect changes with 

 pressure, but it probably does, and therefore the effect of a single pair of joints 

 will be still greater, and, as such, too considerable to be neglected in any attempt 

 to discover the law of induction in this way. If the supposition in (c) be cor-, 

 rect, such an interruption in the circuit may be considered equivalent to an 

 increased length of c, and half as much of z. I tried this by developing the L 

 of Table sviii. by Cauchy's process as a function of c and z* and thus ascer- 



L = log-' (3-13245) 



* This formula is — 



{ 1 - c X L-' (8-55692) + c' X £-' (7-07878) - c' x i-i (5-15437) | 



|l-(3 + .')'xi-'(8-14374) + (3 + 2)Hi:-'(6-99018)-(3+z)3xi-'(5-91248)|. 

 The variable in the third factor is 3 + z instead of 2, for a reason already assigned. The products 

 of c and z -were not used, the approximation being sufficient -without them, as the ten observations 

 of the Table are represented with a mean error of 1-59, and a maximum one of 2-91. 



