562 Rev. P. J. Kirkby on the 
wire fused into the top of the tube. The anode B is attached 
to an iron rod, which, sliding in the fixed steel guides P, Q, 
and bent round the bottom, terminates at E. Thus by 
raising or lowering E, the movable anode could be ad- 
justed to any distance up to about 30 cms. The positive 
end of the battery Z was in metallic connexion with the 
mercury, and so with the anode ; and the negative end was 
connected through a galvanometer G and the variable re- 
sistance R to the cathode A. Z was usually of about 1500 
volts, G was a high resistance and dead-beat voltmeter, and 
R several 100,000 ohms. 
With such an apparatus R could be varied and D deter- 
mined by adjusting the anode till the galvanometer recorded 
C. And then we may. as stated above, regard X as a linear 
function of D, C and p being constant. 
Now X and R are connected by the ohmic equation 
X = Z-C(R + G), 
if Z denotes the E.M.F. of the battery and G the resistance 
of the galvanometer, together with any other invariable 
resistance in the circuit including that of the battery. 
Therefore R is a linear function of D, and the points whose 
coordinates are D, R will determine a straight line, the slope 
of which multiplied by C is the electric force in the positive 
column, since 
v dX _ p rfR 
The slope of the D — R line was determined throughout 
the earlier experiments by finding the values of D corre- 
sponding to several different values of R. Thus a number 
of points were plotted, which, if the discharge had been 
steady, lay upon or close to a straight line. If they did not, 
I often found that some lay very near one line and others 
fell close to a line parallel and clcse to the former. That 
showed in rather an interesting way that the discharge had 
shifted from one to the other of two slightly different states 
or positions without, however, affecting the magnitude of the 
force in the positive column. Sometimes three such parallel 
lines could be recognized; but of course, if there were more 
than two such shifts, the corresponding parallel lines cannot 
easily be detected, and the force Y can only be estimated, 
and with considerable error, by drawing the mean locus of 
the points. Owing to this experience and in order to obviate 
such shifts, I ultimately determined Y, when possible, in a 
single discharge by diminishing R twice by the same amount 
