24 JOURNAL OF THE WASHINGTON ACADEMY OF SCIENCES VOL. 13, NO. 2 
that measured. If the resistance for the first 10 seconds of 3 minutes 
differs by 1 per mille from the later, measured value, the discrepancy is 
appreciable in work of 0.1 per mille precision. Changes greater than 
this are possible, even in constantan wire, if the attempt is made to 
run at anything like the highest safe temperature. Although difficult 
to follow during a regular run, the change can easily be measured, and 
should be, by special experiments where nothing else is attempted, 
but subsequent variations in it will of course cause error. Unlike 
the lag effect already discussed, it has no tendency to cancel itself, 
since the reverse change in resistance comes only after the current has 
ceased to flow. To keep the heater coil running at a low temperature 
diminishes this error, and also that from loss of heat along the leads. 
I, B: DEALING WITH UNSTEADY BATTERIES 
Unsteady batteries are scarcely compatible with the very highest 
precision; with them the best results are obtained by some departure 
from standard procedure. (1) Very frequent, uninterrupted measure- 
ments of current or voltage tend to diminish the error from irregularly 
varying voltage, but demand either two observers with complete sets 
of apparatus, or else reliance upon the constancy of values for coil 
resistance, correlated with current and calorimeter temperature, de- 
termined by separate special experiments. (2) The voltameter, 
giving directly the product of current and time, handles perfectly the 
most sudden and irregular fluctuations® of current, and still leaves the 
observer’s time free for other things. This time can be employed for 
measuring the coil resistance by a Wheatstone bridge, using the heating 
current itself as the bridge current.° This involves a loss of simplicity 
in the electric circuit, but is of advantage if it is desired to shunt part of 
the current by the heater, in order to give the voltameter the relatively 
large current which promotes accuracy with it; the Wheatstone bridge 
measurements can also be used as measurements of the shunt ratio. 
A further disadvantage is the relatively long time, 20 minutes at the 
very least, needed for high precision with the voltameter. 
5 There is evidently an error from very large fluctuations, owing to the fact that 
the heat is as the square of the current, but this error is usually quite negligible, and 
if not, can be avoided by a little coarse voltage regulation. The voltameter result is 
also sufficient when accompanied only by accurate measurements of voltage. In this 
case the voltameter makes the time measurement unnecessary, but does not deal well 
with voltage fluctuations. 
6 This has been done; the reference is not at hand. The coil resistance will vary 
very much less than the current, hence is far more easily measured when the current 
is unsteady. 
