94 
MR. F. E. SMITH ON THE ABSOLUTE MEASUREMENTS OF A 
1,000,000. The mean difference from the mean irrespective of sign is ±0'8 mm. 
These differences are typical of those obtained in our experiments and from them we 
conclude that the thermo-electric effects introduce no error in our final result. 
(B) Elimination of Error due to Electrostatic Effects. —We have already described 
the connections of the circuit to earth and the reversals made in order to eliminate 
any electrostatic effect on the galvanometer, but we thought it desirable to make 
tests from time to time of the efficiency of our arrangements. In these experiments 
the coils were removed from the circuit and an equivalent resistance substituted, 
which resistance was placed in such a position that the current through it produced 
no magnetic field in the neighbourhood of the discs. The standard resistance 
remained in the circuit, but the leads connecting it to the galvanometer were 
connected to one and not both of the potential terminals. Observations of the change 
of deflection were made as before at intervals of 15 seconds. The mean difference 
from the mean of such a set of measurements was always negligibly small and 
indicated that no error due to electrostatic effects could influence our results. 
(C) Constancy of the Speed. —The records of speed are remarkably good and show 
the variations to be very small. Each record enables the mean speed, during the 
measurement of resistance associated with it, to be calculated with great accuracy; 
but a fluctuation in speed of a few parts in 100,000 lasting for a comparatively short 
time cannot be detected. The method adopted to control the speed is described in 
Section 10, and it is evident that the speed must oscillate in value, the magnitude of 
the oscillations depending on the cause of the variations, the sensitiveness of the 
bridge, the sluggish movement of the galvanometer, and the ability of the 
operator. Now the galvanometer in the bridge circuit is sluggish in its movements 
and unsuited to detect changes in speed lasting for a few seconds only, but the 
Cambridge galvanometer used in the Lorenz circuit has a short period (5 '2 seconds), 
and the changes in its deflection during a measurement of resistance show that 
variations in speed lasting for a few seconds only are of common occurrence. 
An idea of the magnitude of these sudden changes of speed is afforded by a com¬ 
parison of the two graphs (A) and (B), fig. 24. Graph (A) shows the variation in 
the thermo-electric effects, while graph (B) shows the variation in the thermo-electric 
effects plus the variations in speed. The differences plotted in graph (B) are those 
taken during a measurement of a resistance. These two graphs are typical of the 
results generally obtained. In all our experiments the combined variation of 
thermo-electric effects and changes of speed were of about twice the magnitude of the 
changes in thermo-electric effects alone, and it appears reasonable to conclude that 
small oscillations in the speed were frequent. The magnitude of these oscillations 
appears to be about 5 parts in 100,000. However, such fluctuations had practically 
no effect on the final results. 
(D) Effect of the Leads. —It was possible that the current in the leads to and from 
the coils might produce a magnetic field of sufficient intensity to affect the results. 
