corks. The corks fitted loosely, and a vertical groove, filed in each, allowed any gas which might be evolved to escape. The junction tube had a length, from B to C, of 12.6 centims., and a diameter of about 1.2 centims. If we add one-half the diameter as a correction for each end we get 13.8 centims. as the effective length of the tube, which must be divided into the total difference of potential between the electrodes to give the potential gradient.
The advantages of such an apparatus are shortly as follows:—
1. Any gas evolved at once escapes without interfering with the experiment.
2. Since the liberation of ions and decomposition of salt occur only at the electrodes, it is (in general) only near them that alterations of concentration occur. Since they are surrounded by a large mass of liquid, these alterations will not appreciably change the resistance of the tube as a whole, and will be entirely without effect on the resistance of the column of liquid in the narrow tube, which is the important point.
3. Changes in density at the electrodes will not appreciably accelerate or retard the velocity of the junction. Take the case of copper chloride when the current is passing down as a fairly typical example. The average current was in C.G.S. units. The quantity of electricity which passed in one hour was, therefore, , and the weight of copper deposited,
grm.
Supposing that all this copper comes from the kathode vessel, and that when deposited it occupies no volume (both of which suppositions would increase the disturbing effect we are considering), half this weight of solution would pass from anode to kathode to bring the liquids in the two tubes to their original levels. The area of cross-section of the tube is 0.430 sq. centim. (see p. 347). Therefore the passage of .0023 cub. centim. of solution from anode to kathode in one hour would mean a velocity of .0054 centim. per hour. The actual velocity observed in this case, due to the ionic velocities, was 1.6 centims. per hour. The correction for differences in density is thus utterly negligible.
4. A more important objection was pointed out by Professor J. J. Thomson. With salt solutions of equivalent strength there is, in general, a difference of specific resistance, although this is small in most cases. At the junction there will thus be a discontinuity of potential gradient, and, therefore, a distribution of electricity, which will be acted on by the electric force and introduce disturbances at the boundary. By a little consideration we shall see that this is a non-reversible effect, and should be eliminated by reversing the current and taking the mean of observations in each direction. Suppose that the current is passing in the direction which gives a distribution of positive electricity over the boundary. This will tend to make the
