282 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



:b. 



T^j6 



and reading nearly alike at all temperatures, each graduated to 0°.2 C. 

 The steel bars b and h\ and the copper wires soldered to them, extend 



through the rubber stoppers rj 

 and Tn, so that the ends are 

 exposed to the streams of water 

 for a distance of about 1.3 cm. 



To prevent disturbance from 



*~ chemical action upon the steel 

 and copper, the parts exposed 

 to the water had been dipped 

 in a solution of copal in ether. 

 With this protection of the junc- 

 ' tions it appeared from tests 



made at various temperatures 

 that there was no chemical ac- 

 tion of sufficient magnitude to 

 affect materially the results of 

 the experiments made to deter- 

 mine their thermo-electric be- 

 havior. The tests for chemical 

 action were made by exposing 

 both ends of one bar to water 

 at the same temperature, some- 

 times with the bar in situ, as in 

 Figure G, and sometimes not. 

 The difference of temperature used in the thermo-electric tests was 

 commonly about 4° C. The thermometers were read, by means of a 

 telescope, to one tenth of one division, that is, to 0°.02. To eliminate 

 disaoreements of the thermometers, and various other possible sources 

 of error, the streams were regularly interchanged, other conditions 

 remaining the same. Thus, if a set of observations had been made 

 with Ti reading 28° and T2 reading 24°, another would immediately 

 be made with 7\ reading about 24° and T^ about 28°. 



The following course of reasoning will show how this method of 

 using the thermometers tended to eliminate their errors from the 

 result. Let us suppose that 7\ is used at 24° and 7T, at 28°, then 

 7\ at 28° and T^ at 24°, and that afterward 7^ is used at 66° and 

 7; at 70°, then T, at 70° and T. at 66°. Let [^Joq be the error of 

 Ti at 20°, etc. Then we have in the various cases supposed : — 



