s.r 
, “4 
ba > 
204 RICHARD THRELFALL. 
but the observed pull was always over two pounds, and in some 
experiments about three pounds. On guiding the bars so that no 
wedge-shaped gap appeared, the traction could be got down to 
about one pound eight ounces. I do not think it is possible to 
get much closer than this, for if proper arrangements are made to 
absolutely insure a really true and rigid separation, friction would 
inevitably come in to introduce errors. My results at higher 
inductions were so similar to Bosanquet’s that they are not worth 
reproducing. 
7. With regard to the reduced formula not applying to the case 
of non-magnetic gaps of sensible dimensions parallel to the lines 
of induction—as when Bosanquet separated the bars by wood and 
paper—the explanation is obvious. The lines of induction no 
longer leave the surfaces normally and the conditions postulated 
by the formula are not in existence. 
8. Resulting position of the theory. When the bars are in 
contact, the stress theory and what I will call the magnetic fluid 
theory, lead to the same results, which is true certainly within 
about five per cent. and may be exactly true. In any case 
measuring tractions is not the way to get accuracy, though I have 
no doubt that rather better results could be got by going into the 
matter more elaborately than was done either by Bosanquet or 
myself. In what follows I shall suppose that the theory is true, 
and that the real cause of magnetic forces is to be sought in some 
condition of the ether mechanism which receives a sufficient 
mean definition from the induction diagram. 
9. The effect of varying the kind of iron employed should be 
the same as varying the induction density, at least in so far as the 
phenomenon can be considered to depend on permeability. I 
used induction densities of from 2,000 to 18,000 but could not 
detect any effect, when the cause of error referred to above was 
eliminated. I also used all kinds of iron, from annealed Swedish 
iron to ordinary cast iron. I varied the lengths of the bars from 
60 cm. to 6 cm., and the diameter from about 2 cm. to about °6 
em. In no case could I detect any deviation from the predicted 
