128 
PROFESSOR HUGH L. CALLERDAR ON 
accidental errors, but so far as it goes, it tends to bring the observations into bettei 
agreement. 
O 
Table X.—Correction for Variation of Gradient in Flow-Tube, Calorimeter C. 
Number of 
Experiment. 
Values of Flow. 
Heat-Loss li per 1°. 
Specific Heat in Joules. 
Q'. 
QA 
Table. 
Reduced. 
Table. 
Corrected. 
III. 
•665 
•399 
•04914 
•04S59 
4-1803 
4-1813 
IV. 
•501 
•258 
•05123 
•04989 
4-1771 
4-1791 
V. 
•660 
•392 
•04937 
•04859 
4-1795 
4-1805 
VI. 
•590 
■ 375 
•04965 
•04882 
4-1790 
4-1802 
The general effect of the correction here is to raise the values by about 1 in 4000 
for the larger pairs of flows. The discrepancy between Experiments III. and IV. is 
mainly chie, as the values of D show, to accidental errors of the small flows in 
opposite directions. The mean value 4H803 agrees very well, allowing for the 
difference of temperature, with the value 4 - 1810 deduced from observations I. and II. 
with calorimeter D at 28°'0 C., which would give 4'1805 at 29'TO C. 
(37.) Variation of the Gradient-Correction with Temperature. 
The importance of this correction arises chiefly from the fact that the value of the 
heat-loss increases with rise of temperature for any calorimeter. As a result , the 
correction, which depends on the square of the heat-loss, is considerably larger at 
higher temperatures, and thus affects the curve of variation of the specific heat as 
well as the absolute values. If the correction were constant, it would affect only the 
absolute values, which would matter little owing to the uncertainty of the electrical 
units. 
As an illustration of the greater importance of this correction at higher 
temperatures, the last four observations (Barnes, Series 8, Table XVIII.) are given in 
the next table, reduced and corrected in a similar manner. Although these observa¬ 
tions do not include any exceptionally small values of the flow, the correction at the 
higher points nearly reaches 1 in 1000 owing to the larger value of the heat-loss. 
