194 
:^ Turns. 
3(X» volts 
(Metal over Ra) 
300 volts 
47 
47 
52 
50 
27 30 31 / ,. .«,- 
99 00 03 1'^^^^" = ^^'' 
-Nett leak = 416 = 
=13 (say) 
10 volts 
(Metal over Ra) 
10 volts 
47 
47 
51 
50 
41 4'> 1 
^i^-;Meaa^38 
26 31 31 1 ., ..^^ 
23 28 27 i ^^16^" = '^'^'^ 
-Xett leak ^340^ 
= I's 
300 volts 
, (Metal over Ra) 
3( M» A-olts 
47 
47 
52 
50 
45 42(^^—48 
06 07 04 ( ,. 
83 83 83r^^^^» = 445 
-Netl leak = 396 = 
= 13 
A repetition of 
Ig/ 13= 1-947, and 
.'. Mean value of l3=406. 
Value of I', =340. 
Hence, I^/i; = 1 -208, and yi'^-MOS 
the experiment, in different order, ^nve 
Ve/Tg=l'897. These agree well with each other, for we find 
from the first set that IJ'^ I',!, = 1 -013, and from the second 
that the same fraction = 1-025. 
The fraction I^T'e is the ratio of the saturation current 
in a chamber about 6 mm. wide to the current when the 
potential gradient is about 35 volts jjer cm., and IgT's i^^ the 
ratio when the chamber is 3 mm. wide, all other conditions 
being exactly the same. It ought, perhaps, to be mentioned 
that the current for the chamber of double width was not 
quite twice that for the other, because the widening was 
effected by raising the top plate, and so adding to the cham- 
ber a layer of air which was about 3 mm. further away from 
the radium than the original layer. As a little heap of 
radium bromide was used, the curve was of the form shown 
in Plate xviii., '^ Philosoj)hical Magazine^,'' December, 1904, so 
that ionisation decreased as distance from the radium in- 
creased. These results show clearly the existence of at least 
one effect which we should expect to find as a result of initial 
recombination. 
Again, we ought to find that variation in current 
strength, caused by altering the power of the ionising ap-ent, 
makes little difference to the form of the curve when the cur- 
rent is small. We have made several experiments in this 
direction also. In Fig. 2, curves A and B show the results 
