198 
He lias fouiul by experiiiieiit tliat wlieii Z is I cm., ami F^ is 
151 volts pel' cm., then (in amperes) 
C-<S-0;3 X ]()-" 200 (area of electrode - L>00 s(|. cm.) 
= 4-01 X 10-i« 
^-3-94 X 10-1'^ 
Therefore, 
0- 
and by substitution in the ei] nation it can l)e found that 
a/6? = 44:U {loc r?t, p. :)'M)). 
Now, in a similar experiment, with far smaller currents, we find 
that when /5 is 1 cm. and t\^ is 150 volts per cm. 
C- 4-2:3x10-^^^ 
c = 3-90x 10-^^^ 
Therefore 
C-G 
^^—- ^-I-Ix 10^' 
and substitution in the equation gives a value of a/e about 5,000 
times greater than Retschinsky's, or about 2 x 10'. 
In the second formula (Stark's) the values of a are more 
correct, because the current values used are taken, one from 
the lower part of the curve, and one from the saturation 
values, so that their difference dejjends less on the effects of 
initial recombination. Retschinsky draws attention to these 
anomalies in his results, but ascribes them to absorption of 
ions by the electrodes. He argues that in a shallow ionisation 
chamber tins effect must be greater than in a deeper one ; 
and so he accounts for the lack of saturation in the small 
chamber, a lack which is excessive if attempt is made to 
explain it as wholly due to general recombination. But we 
think that a more reasonable explanation is to be found in 
the hyjiothesis and results described in this paper, in connec- 
tion with which Retschinsky's results fall naturally into place. 
Retschinsky points out that McClung obtained smaller values 
of a than he himself obtained, and he suggests that, since 
McClung's apparatus consisted of a series of shallow ionisa- 
tion chambers, the absorption effects would be considerable. 
On the hypothesis of this paper McClung's method is so far 
the more reliable, that it avoids the complications due to tlie 
special phenomena which occur in connection with the origi- 
nal forming of the ions, and deals only with a state of things 
in which ions have been formed and are distributed at ran- 
dom through the gas. All methods in which ions are formed 
whilst the potential gradient is in existence must be more or 
