6 PROCEEDINGS OF THE AMERICAN ACADEMY. 



sisted ill accidental traces of water adhering to glass or mercury, which 

 had the effect of unequally diluting the electrolyte in the cell, and thus 

 causing concentration currents. If the apparatus is dried and the mer- 

 curous chloride is carefully washed with the appropriate solution before 

 being added, the ease with which one may obtain consistent results is 

 quite remarkable. The water used in these experiments, except in the 

 first crude ones, was ordinary distilled water which had been redistilled 

 from alkaline permanganate in a platinum still. 



The firsL systematic experiments had for their aim the determination 

 of the probable magnitude of the errors of observation due to accidental 

 causes. 



The apparatus consisted of two test tubes mounted in a convenient rack, 

 in such a way that the liquid contents could be connected by a syphon. 

 This tube could be filled by suction through a third arm above. Each 

 test tube contained a few cubic centimeters of mercury, and above this 

 mert-ury was jilaced a layer of finely divided mercurous chloride, sur- 

 rounded by the electrolyte which filled the rest of the apparatus. If the 

 layer was left undisturbed, its thickness was found to be unessential. 



The two test tubes were kept at different temperatures by means of 

 suitable baths of constant temperature, a steam jacket being employed for 

 100°, and pure cracked ice for 0°. At first the readings of the external 

 thermometers were verified by means of very small accurate Anschiitz 

 thermometers placed in the test tubes themselves ; but after many such 

 verifications, which always resulted satisfactorily, this practice was dis- 

 continued, and the readings of the external thermometers alone were 

 taken. 



The differences of potential between the two layers of mercury were 

 measured by the well known Ostwald-Poggendorff method,* within per- 

 haps 0.0002 volt. The final reading was not taken until the temperature 

 conditions had thoroughly adjusted themselves, an operation requiring 

 from half an hour to many hours, according to circumstances. 



In the first place a series of determinations of the effect of dilution 

 upon the temperature coefficient of the electrode containing supernatant 

 hydrochloric acid was made. The stronger solutions gave very uncer- 

 tain results, and hence cannot be considered accurately determined; and 

 correspondingly, in these the calomel was partially decomposed with 

 separation of finely divided mercury. The reasons for these irregulari- 

 ties were better understood later. 



* Hand- uiid Hilfsbuch pliys. cheni. Mess., pp. 251-255. 



