556 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



theless, to avoid possible trouble from a possible minute leakage of 

 steam, it was determined to seal at the bottom the large glass tubes 

 which served as a shield for the junctions. 



At the same time a change was made in the manner of connecting 

 the small glass tubes at their tops. The couples were removed tem- 

 porarily from these tubes; the splints and rubber tube connectors 

 shown in Figure 2 were abolished ; and the two small glass tubes for each 

 couple were now joined directly to each other in one piece, as in Figure 3. 



Then the couples were replaced in 

 the tubes. In this change little or 

 no care was taken to distinguish be- 

 tween the two ends of each couple, 

 and it appears that D 3 was turned 

 about in its replacement, so that for 

 this couple " parallel" afterward cor- 

 responded to the same orientation 

 of the couple to which " crossed " 

 had previously corresponded ; for the 

 (p — c) of this couple was consistently 

 + after the change, whereas it had 

 usually been — before. This change 

 was made between the observations 

 of September 27 and those of Sep- 

 tember 28. A comparative summary 

 for September 26 and September 28 

 is here given, the interval of tem- 

 perature on both days being approxi- 



On the whole, the results of the overhauling given to the apparatus in 

 the change from the arrangement of Figure 2 to that of Figure 3 were 

 unimportant so far as harmony of performance is concerned. 



On September 27, still under the impression that the differences 

 of behavior of the different couples and the (p — c) values of the indi- 

 vidual couples might be accounted for by the gradients of temperature 

 in the bent parts of the german-silver wire above the pots, we tried 



