218 J. A. POLLOCK. 
more uniform effect, a series of sparks may be taken for 
each observation, a method which has been used by many 
experimenters. Asis well known, however, only a limited 
number of sparks can pass between the knobs before the 
character changes owing to the deterioration of the dis- 
charge surfaces. In the present case after trial, it was on 
the whole considered best to discharge the condenser once 
for each observation. 
The condenser was charged from the secondary of a small 
induction coil. The primary circuit of the coil was closed 
and opened by two keys worked by a heavy pendulum, the 
interruption taking place very quickly. The circuit was 
opened in air, the gap being short circuited by the condenser 
of the coil. A variable resistance was inserted in the 
primary circuit and adjusted so that when the spark knobs 
were set, the difference of potential established on working 
the coil was just sufficient to break down the dielectric. 
If the spark gap is watched through a lens, as in these 
experiments for each discharge, it is seen that successive 
sparks jump across the gap from different points of the 
knobs. Their character may be either round, long or 
irregular, either large or small, single or double, scarcely 
ever quite the same for two sparks together. The demag- 
netisation of the detector, has been found to so greatly 
depend on the path of the spark and on its character that 
it was useless to retain any observations except in those 
cases where the sparks took place across the centre of the 
gap and where the character was as far as could be judged 
the same for all. Hundreds of observations have been 
rejected on account of irregularity of the spark. On the 
other hand, hundreds of discharges where the sparks looked 
perfect, have given results utterly non accordant. 
For the earlier of these experiments the sparks took place 
between aluminium spheres 1 cm. indiameter. Thespheres © 
