88 Transactions of the Boyal Society of South Africa. 
strength of the current used to work the electromagnet of the coincidence 
apparatus. 
In the Sterneck apparatus a flash is produced each time the current 
is made or broken by the clock : there are thus two independent series of 
flashes, either of which may be used to determine the coincidence period. 
Unfortunately, before the influence of current variation was thought of, 
no record was kept as to whether the make or break was used. The 
influence of current variation is much greater on make than on break, 
but is still perceptible on the latter, as certain special observations 
showed, k set of 10 coincidences was observed with the aid of current 
from two accumulators, which remained practically steady at 9^ (arbitrary 
units) ; the battery was then exchanged for three leclanches, which gave 
a current of 1^ at first, ^[falling to 6 ; with this another set was observed 
25 coincidences later than the first. The accumulators were then restored 
and a third set taken, 50 coincidences later than the first. When obser- 
vations were made by attraction of electromagnet (" make ") the results 
were : — 
Mean interval for 50 coincidences = 1549'31s. whence c = 30*9862 
„ first 25 ,, - 776-22 „ c = 31-0488 
The first value of c is the true one, the second is lengthened because the 
current was weakened and the electromagnet pulled over more slowly 
in the second half of the observations than in the first. The real delay 
in the action of the magnet is given by 776*22 - "^^^.^ = 1'67 ; this is 
about coincidence, and indicates a delay of period of pendulum, 
or 4^ sec. 
When the recovery of the spring ("break") was used, the result is 
much less inaccurate, although when the current is weak the spring 
recovers a little quicker. Similar observations gave — 
Mean interval for 50 coincidences = 1549-17s. whence c = 30-9894 
,, first 25 „ = 774-57 „ c = 30-9828 
i.e., the recovery of the magnet is accelerated by — 
154947 
— 2 ~ 774-57 = -16 or L coincidence, or by sec. 
The effect of these errors would be to make the calculated period of 
the pendulum too short by 70 units and too long by 7 units respectively. 
It seems probable that there may be some change in length of the 
pendulums also. No. 23 gives results that are notably more irregular 
than the other two, though no defect in it could be discovered. 
