60 J. A. EWING AND J. G. MACGREGOR ON THE 
rock g. The spot of light moves (say) first to the right, but immediately passes 
to the left of zero, the deflection to that side gradually increasing if the rocker 
is held down. This after-deflection to the left takes place although the rocker 
g was down for the smallest possible time only. ‘The first motion to the right 
shows that R is too great ; the next, to the left, shows that owing to the polari- 
sation the apparent resistance in the tube has come to exceed R. But the first 
indication is the one which is to be received ; so diminish R, reverse f, and give 
the tube some time to depolarise itself through cc’. Then rock g again. Sup- 
pose this time there is no motion whatever to the right, but one to the left only. 
This shows that R has now been made too small. In this way, by successive 
trials, a very exact value of R will be obtained, the object being to adjust R so 
that the motion to the left shall just, and only just be preceded by an exces- 
sively slight motion to the right. When this state of things is nearly arrived 
at, the motion to the right becomes a mere trembling of the spot of light, requir- 
ing great care to observe it at all. 
We found it almost impossible to get rid of local currents in the tube, which 
seemed to be due either to impurities on the platinum plates, or to the presence 
of moisture at the junction of two metals in the binding-screws or rockers. 
They were so considerable as at first to threaten to spoil the method of testing 
completely. Numerous anomalous results which we got in our earlier experi- 
ments were traceable to their action. This difficulty, which at first seemed very 
formidable, was easily overcome in the following way. Whenever we noticed 
that there was a local current in the tube (which we could tell by removing 
g altogether, and observing whether there was any deflection when e and e’ were 
connected), we set the commutator f, so that when g was rocked the residual 
polarisation in the tube would be in the opposite direction to that of the local 
current. By closing the battery circuit for a sufficient number of seconds, a 
polarisation could be induced considerably greater than the local current; then, 
as that gradually died away, there was a particular instant at which polarisation 
was exactly equal and opposite to the local current ; or, in other words, for an 
instant the tube was in an absolutely neutral condition. If, then, the battery 
circuit were closed just at this instant, the test would be a fair one ; and in order 
to effect this, it was only necessary that we should know the exact instant when 
the tube reached the neutral state. This could be ascertained by keeping a 
small rocker, which connected ¢ and é’ only, in a constant state of oscillation, 
and noticing when the spot of light came to rest at zero on the scale. Of 
course, during this time the left hand part of g, which connected ¢ and ¢’, had to 
be removed, else the galvanometer could not give any indication of the electrical 
state of the tube. 
These various improvements on the original mode of testing were effected 
gradually, and as the want of them was felt. It took us about two months to 
