568 PROCEEDINGS OP THE AMERICAN ACADEMY. 



The mean of the numerical values of (n — s) is 0.29, about 1.4 per cent 

 of the deflection, for December 16, and 0.17, about 0.8 per cent of the 

 deflection, for December 20. 



On December 22 and 23 calibration tests were made with the tem- 

 perature interval 100° to 140°, approximately. Both of the pots used 

 on these days contained water, and the reversal of temperature conditions 

 was effected, as on December 16 and 20, by having first one and then 

 the other of the pots the hotter. The usual summary is as follows: 



The mean of the numerical values of (n — s) is 0.105, about 0.5 per cent 

 of the deflection, for December 22; and 0.17, about 0.9 per cent of the 

 deflection, for December 23. Nearly all the (n — s) values of these two 

 days have the minus sign, whereas on December 16 and 20 the plus 

 sign prevailed. But the fact is that on the two earlier days the 

 " crossed " arrangement existed, whereas on the two later days the 

 " parallel " arrangement was used. 



Finally, on December 28, a test was made with the interval 140° to 

 180°, appproximately, one pot now containing water and the other 

 naphthalin, as in the early experiments. The result was : 



A 3 (p_C) B 3 (p-C) C 8 (p-C) D 3 (p_c) 



20.30 (0 + 6) 20.19 (0 +) 19.96( +) 20.01^, 



The mean of the numerical values of (p — c) is here 0.11, about 0.5 

 per cent of the deflection. 



The differential percentages, (n — s) or (p — c), found on these last 

 five days of the calibration tests of couples A 3 , B 3 , C 3 , and D 3 , have a 

 mean of 0.8 per cent. The temperature interval on each of these days 

 was not far from 40°. The corresponding mean for previous tests of 

 the same couples made with a like interval of temperature is 3 per cent. 

 It is plain, then, that the policy of keeping the middle parts of the 



