THE MECHANICAL EQUIVALENT OF HEAT. 
385 
Poggendorff’s method was used ; a change of one plug indicated a change of E/6000, 
each cell was balanced against the standard ones in the manner adopted by Messrs. 
Glazebrook and Skinner, and in no case could any alteration whatever be detected. 
This point having been satisfactorily established, we considered it unnecessary to 
repeat the observations in a similar manner this year, but contented ourselves with 
occasional comparison of the cells used with others which we kept as standards. 
(2) The second difficulty was not surmounted so satisfactorily. A.s before mentioned 
(Introduction) the laboratory was maintained in cold weather at a constant tempera¬ 
ture of about 16° C., by an automatic regulator. However, during the months of 
August and September, 1892, while the nights were cool, the temperature during the 
day was extremely high, and occasionally rose in the laboratory to over 20° C. The 
Clark cells were contained in a tank of water, but their temperature at times changed 
by 3° or 4° C.; and under such circumstances the difficulty was to know wliat tem¬ 
perature to take. 
Messrs. Glazebrook and Skinner have shown that the E.M.F. does not change 
with the thermometer, but lags considerably. A difference between the “virtual” 
and “real” temperature of 1° C. would cause a difference of 1 in 1000 in our 
corrected times, and we believe that many of the irregularities in our results, 
especially in experiments before August 15, are attributable to this cause. On 
that date we got into working order a regulator which much diminished this source 
of error. A large bulb of mercury was immersed in the tank, and was connected by 
a narrow tube with a trap of the ordinary gas regulator pattern, through which 
cold tap-water, instead of gas, was kept flowing. When the mercury expanded, the 
water ceased to flow through the regulator, and diverged by a bye-way into the 
tank; when the mercury contracted, the cold water passed into a waste pipe, instead 
of into the tank. The pressure of the water was kept constant by a small supply 
cistern, with an overflow pipe, which was suspended by a string, and the head of 
water could thus be regulated. This apparatus worked most satisfactorily—a change 
of ^° C. turning the inflowing water from the tank to the waste. Unfortunately 
the weather became so warm that even the tap-water at that end of the laboratory 
rose to about 16° C. The changes, after the introduction of the regulator, were, 
however, much diminished, and rarely amounted to 0°'5 C. As regards the errors 
thus introduced, although they seriously affect individual experiments, they must 
tend to eliminate when a large number of observations are taken. 
The Clark cells used for the 1892 experiments were constructed during January 
and February of that year. Particulars of their construction are given in an 
Appendix to Messrs. Glazebrook and Skinner’s paper; therefore it is unnecessary 
to give here more than a brief summary. 
, Although the method of preparing the solutions for different batches of six was 
varied, the E.M.F. of all the cells is practically the same. In all cases, the platinum 
wires were amalgamated by being dipped in mercury when red hot. The mercury, 
MDCCCXCIII.—A. 3 D 
