the Canal-Ray G 
roup 
Ml 
between the plates (fig. 5). A precisely similar appearance 
is presented by the radial arms of the stars formed when 
using the various double polygons mentioned above, 
when the line of sight is parallel to the planes of the 
polygons. 
A more exact knowledge of the path followed by 
the rays in the space between the plates may be derived 
from the fact that the rays themselves map out this 
path on the plates. If a double square has been used 
for some time as cathode in a gas at suitable pressure, 
then on stopping the discharge the two opposed plates 
show a pattern which at first sight appears compli- 
cated, but which on closer inspection is seen to consist 
simply of the superposition of four bands, each arising 
from a side of the square (fig. 6 a). The bands have 
a boundary which is concave outwards, and therefore 
represent a system of rays which gradually narrows 
as it approaches the opposite side of the square. The 
rosy arm of the cross shown in fig. 2, which issues 
from the opposite side, is the direct continuation of 
the concave band. While the discharge is still in 
progress it is possible to observe directly that the rosy bands 
between the plates have the same direction and dimensions 
as the traces left on the plates when the discharge is arrested. 
Fit:. 5. 
O 
Fig. 6 a. 
E3 
Fig. 6 b. 
Fig. 7. 
On the plates of a double rectangle there appears a pattern 
which is the result of the superposition of two pairs of con- 
cave bands, one pair of which is shown in fig. 6 b. Corre- 
sponding patterns are obtained with cathode plates of other 
shapes. The patterns result partly on account of differences 
in the reflecting power connected with the formation of thin 
layers of oxide, and partly from shallow grooves produced in 
the originally plane cathode surface. 
From the concave boundary of the traces and of the pencils 
themselves, it may be inferred that the rays proceeding from 
any side are subject to an attraction due to the neighbouring 
sides. We may, for example, imagine that in the case of 
the square the rays are originally convergent along straight 
lines, somewhat after the fashion indicated by the dotted 
lines in fig. 7, and that it is in consequence of the attraction 
exerted on them by the adjacent sides that they assume the 
