88 ROYAL SOCIETY OF CANADA 
If the electrode C were given either a positive or a negative charge 
it sometimes happened that no leak occurred. This result was chiefly 
noticeable when the terminals D and À were connected by a wire and 
both acted simultaneously as cathodes with B as the anode. Also 
throughout the experiments, sudden and violent deflections were fre- 
quently obtained which seemed to indicate strong inductive action ac- 
companied by sparking within the cylinder. 
With this form of tube the results were exceedingly unsatisfactory. 
Owing to gas being given off by the metal it was difficult to maintain 
the same degree of exhaustion for any length of time. The character 
of the discharge also seemed to be affected by the state of the electrodes, 
and it was consequently almost impossible to reproduce all the con- 
ditions necessary to make a proper comparison of results obtained at 
different times. 
Besides, the Faraday cylinder D was not free from defects. Schus- 
ter” has shown that when a vigorous discharge is passed through one 
compartment of a vacuum tube the gas in an adjoining one is also 
thrown into a sensitive state. It was just possible then that, though 
the opening C' served to equalize the pressures, it also afforded, by diffu- 
sion, a means of communicating to the gas inside the cylinder the con- 
ductivity impressed upon that outside by the discharge from the induc- 
tion coil. 
Again, in this form of tube the electrode necessarily passed through 
the glass and could not therefore be completely surrounded by the 
cylinder. 
In order to overcome any disturbing influence arising from these 
defects, and to localize more definitely the effect investigated, a new 
form of tube was devised which could readily be taken apart and which 
did not require even a small opening in the Faraday cylinder. 
2. ELECTRICAL CONDUCTION ALONG THE SURFACE OF GLASS. 
This form of tube which is shown in Fig. II, was divided into two 
compartments which could be separately exhausted. The metal cylin- 
der D in this case formed part of a metal socket H into which the glass 
parts of the tube were fastened. Air tight joints were made by in- 
serting rubber washers 6 between two bevelled brass ones, and then 
compressing the whole by a threaded brass piece which slid over the 
glass and could be screwed into the socket. 
The upper end of the cylinder was made air tight by placing lead 
washers above and below the plate d and then screwing a brass cap 


1 Schuster, Proc. Roy. Soc., 42, p. 371, 1887. 
