22 JOURNAL OF THE WASHINGTON ACADEMY OF SCIENCES VOL. 13, No. 2 
accuracy, but can also be corrected for with ease. m — L should usually 
be less than half a minute. With the large values of K characteristic 
of some aneroid calorimeters, closer approximations would be needed. 
In using this method the value of L, the lag represented by BD in 
fig. 2, should be known rather accurately, since this affects directly 
the time assigned to the higher temperature, and so is in effect multi- 
plied by KA@ in getting the thermal leakage. Hence for 0.03 per 
mille precision, that is, for security in 0.1 per mille precision, L, if 
K is 0.003, should be known to 0.01 minute, like any other lag affect- 
ing the whole calorimeter. <A very satisfactory way to determine it is 
this: Make a series of blank heatings, observing only thermal head 
or calorimeter temperature, as frequently as, say, every 15 seconds. 
Then, as in any regular calorimetric determination, find, from the 
calorimeter temperature, the total heat supplied, and also find the 
thermal leakage up to the time of each 15-second reading, which is not 
nearly so troublesome as it may sound. Then, knowing the total 
heat, the fact that it was supplied uniformly, and the thermal leakages, 
compute the calorimeter temperature corresponding to the heat actu- 
ally in the calorimeter at each reading. Comparing these with the ob- 
served temperatures gives a series of values for the lag, whose average 
is comparatively unaffected by the fluctuations of the rapidly rising 
temperature, and of course has the thermal leakage eliminated. ‘The 
same data give a check of correction (6) for the apparatus and condi- 
tions employed. The total leakage is usually so small that a prede- 
termined value of K is quite good enough, hence no rating periods 
(“after periods”) are needed. In a number of tests of this procedure 
the check of (6) was seldom out by more than 1/3 of (6) which, in this 
case, was about the uncertainty of a single reading, and corresponded 
to 0.25 second during the rapid rise of temperature. The main 
probable cause for the discrepancies observed was the fluctuation 
of the rising temperature as read. This seems to show that where’ 
the lag is constant, the method of calculation here given, corrected by 
means of (6), will probably give more accurate results in actual cali- 
brations than observations of ¢. Readings once a minute, however, 
though less accurate, will, with reasonably good stirring, be sufficient 
for observing the thermal head; this is why it can be observed along 
with the electric energy. 
Such readings may be preferable or even necessary if the blank 
heatings have shown the lag to be variable. 
Blank heatings will also enable an empirical expression, a substitute 
for (6), to be found in cases where the lag is large, and where, conse- 
