378 Prof. E. Goldstein on 
actually observed form (shown by the fall lines). Similar 
remarks apply to rectangles &c. 
It may be asked in how far all these observations serve to 
explain the genesis of the canal rays proper. In this con- 
nexion it is only necessary to remark that a straight slit in a 
plane parallel plate represents nothing else than a double 
rectangle or a hollow parallelepiped, and that a plate with a 
number of slits forms a system of such double rectangles. 
Each double rectangle acts like an independent cathode of 
the same shape, and thus there arises a number of luminous 
bands similar to those shown for the case of a single double 
rectangle in fig. 4. If the plate containing the slit completely 
fills the opening of the tube, only a single luminous band can 
be produced — viz., that which issues from the edges of the 
slit on the side towards the anode. 
If the perforations of the cathode consist of holes and not 
of slits, each hole acts like a small hollow cylinder. In the 
case of a single hollow cylindrical cathode there is produced 
a rosy feebly divergent luminous cone. 
In the small double rectangles of plates provided with 
slits, and, mutatis mutandis, in the short hollow cylinders of 
cathodes perforated with holes, the canal rays pass both 
tangentially along, and at a moderate inclination to, the inner 
walls, from the front to the back surface, and into free space. 
The fact that the rays do not exclusively follow tangential 
paths explains why even slits a few millimetres in width 
appear to be completely filled with canal rays, and why each 
luminous band when viewed in a direction parallel to the 
shorter side always appears divergent: those rays which are 
inclined to the walls of the slit and which proceed from 
opposite walls crossing each other while still in the slit and 
then passing out in a divergent form. In order to explain 
details, especially the relatively feeble divergence of the 
luminous bands, it must be borne in mind that just as the 
tangential rays are subject to the attraction of the adjacent 
sides, so also the inclined rays are attracted by the surface of 
the emitting cathode, and become concave towards it. The 
rays therefore emerge not along their original directions, but 
along the tangents to the last elements of the curves which 
they follow between the walls of the slit. 
The fact, on the other hand, that with decreasing pressure 
of the gas the boundaries of the bands proceeding from the 
slits become more and more strongly convergent when viewed 
broadside-on, may be explained by taking into account the 
relation between the canal rays and the first layer. The 
canal rays start on the front walls of the narrow double 
