554 PROCEEDINGS OF THE AMERICAN ACADEMY. 



2. That the couple C 2 is sufficiently exceptional in its behavior to 

 demand especial attention. This couple gave the lowest mean deflec- 

 tion for the 100°-218° interval and also for the 140°-180° interval ; for 

 the 100°-136° interval it was still exceptional, but it gave here the 

 largest mean deflection. On the whole, too, the difference between 

 " parallel " and " crossed " was larger with this couple than with any 

 other. 



All attempts to account for such variations and peculiarities by causes 

 lying outside the german-silver wire were unsuccessful. Looking for 

 possible causes of difference in the german-silver wires we naturally gave 

 much thought to the rather sharp bends which, with the arrangement 

 shown in Figure 1, inevitably occurred just above the tops of the glass 

 tubes through which these wires ran. It was thought that these bends, 

 by introducing some change of quality in the wires at a place where the 

 gradient of temperature is large, might account for the differences noted. 

 Tests made by heating or cooling the various wires at or near the bends 

 in question did not, it is true, fully justify this theory ; but in order to 

 give it a fuller trial it was determined to make a new set of four couples 

 and to support the german-silver wire of each couple in such a way that 

 it would be subject to no violent or greatly variable bends in its course 

 from one heating pot to the other. 



Couples A S) B z , C & , and D 3 . 



The four new couples, made from the same materials as the others and 

 in substantially the same way, will be called A 3 , B 3 , C 3 , and D 3 , respec- 

 tively. The new method of mounting is indicated by Figure 2, in which 

 S is a wooden splint to which are lashed the horizontal ends of the 

 small glass tubes containing the german-silver wire, these ends being 

 connected by a short rubber tube. Corks fitting snugly on the outer 

 glass tubes were lashed to the upper parts of the smaller tubes, so that 

 when the splint S was raised the whole system of protecting tubes was 

 raised with it. The greater part of each small tube above the larger glass 

 tubes was exposed freely to the air of the room. The upper part of the 

 larger tubes was surrounded by loose asbestos, which, as before, lay on 

 the covers of the pots to a depth of about 3 cm. 



The result of this attempt to improve the conditions of experiment 

 was disappointing. With the new arrangement and the new set of 

 couples the differences between the behavior of different couples and the 

 differences between " parallel " and " crossed " with the same couple were 

 greater in general than we had found with the old set of couples, A 2 , 



