12 PROCEEDINGS OF THE AMERICAN ACADEMY. 



On each of these two curves three points were chosen, and the value 

 of 02 — ©1, -jj,, and k computed with the following results (Table III): 



Difference. 



1.3 

 1.6 



1.7 

 1.9 



These results show that the lag is not proportional to the rate of 

 cooling, but that the value of k increases as the value of dQ/dT 

 decreases, and that this change in k is nearly linear. This increase in 

 the value of k cannot be accounted for by the fact that the rate of 

 cooling to the point at which the value of k was computed had not 

 been constant but had been decreasing, so that in each case the lag was 

 more than it would have been had the rate of cooling been constant 

 and of the value at the point in question. The lag would be greater 

 in proportion for the higher points as the decrease in the rate of cooling 

 is greater for the portion of the curves before those points. 



The true cause for this variation in the value for k does not concern 

 us here. The fact is that the second assumption is not correct, and, 

 unless the heating and cooling curves have the same slope, the formula 

 does not give the correct value for e. Moreover, a comparison of the 

 results obtained from Curves 1' and 2 shows that even under these last 

 named conditions the formula is not trustworthy. Curve l' gives a lag 

 of 5.25 degrees centigrade for a rate of cooling of 0.373 degrees per 

 second, while Curve 2 gives a lag of 5.88 degrees centigrade for a rate 

 of cooling of only 0.268 degrees per second. 



Contact lag is variable. — These curves were taken with the same 

 quartz protection tube and the same thermo-couples, and the discrep- 

 ancy is not due to variation in the thermo-couples, as they were tested 

 after each curve and found to have remained constant. The variation 

 was apparently due to change in the contact lag, and this variation is 



