3384 H. A. Rowland—Studies on Magnetic Distribution. 
demagnetized. It does not seem to have been noticed that 
what we ordinarily consider as the magnetic distribution is not 
directly measured in this way; and indeed, to get correct 
results, the magnetization should have been reversed, seeing 
that a large portion of the magnetization will not disappear on 
taking away the magnetizing-force where the bar is long. The 
quantity which is directly measured is the surface-integral of 
the temporary magnetic induction across the section of the bar, 
while the magnetic surface-density is proportional to the sur- 
face-integral of magnetic induction along a given portion of the 
bar. In other words, the quantity measured is Q instead of 
aL We can, however, derive one from the other very easily. 
The sixth and last method is that which I used first in 1870, 
and by which most of my experiments have been performed. 
This consists in sliding a small coil of wire, which just fits the 
bar and is also very narrow, along the bar inch by inch, and 
noting the induced current over each inch by the deflection 
o vanometer-needle. This measures Qe, except for some 
corrections which I now wish to note. In the first case, to 
give exact results, the lines of force should pass out perpendic- 
ular to the bar, or the coil must be very small. But even 
rent, so that the distribution remains unchanged. Hence 1t 
seems to me that this method is the only one capable of giving 
exact results directly. 
The coils of wire which I used consisted of from twenty t? 
one hundred turns of fine wire wound on thin paper tubes 
which just fitted the bar and extended considerably beyond the 
coils. ‘The width of the coils was mostly from ‘1 to ‘25 of an 
inch wide and from ‘1 to 2 inch thick, A moasnrs heing Inid 
by the side of tae given bir an ler exneriinent, the coil was 
moved from one divisiva of the rule tv tue next very quickly. 
