Error in Calorimeiric Work. 329 



III. 15*3 cal. as the increase of difference in two experi- 

 ments due to the increased difficulty of manipulation when 

 two thermometers are used instead of one. 



IV. Difference in two experiments when two thermometers 

 are used, comparing the results obtained with the one instru- 

 ment in the one experiment with those of the other instrument 

 in the other experiment, 53*02 cal. (from 76 experiments). 



Y. Difference in the results obtained from the two different 

 thermometers in the same experiment, 33*95 cal. (from 142 

 experiments), 



Now, subtracting II. from IV*, we get 



VI. That portion of the difference in the results given by 

 two thermometers in any single experiment which is due to 

 the instruments not being absolutely concordant, 11*12 cal. 



(This amount will represent the sum of the errors in the 

 calibration and standardization of the instruments. As it is 

 dependent on two readings of each instrument, the amount 

 attributable to each reading will be 3*7 cal., corresponding to 

 •014 millim., or 0*0005(5° C.) 



Subtracting this amount from V. gives 



VII. Difference due to the error inherent in reading the 

 two thermometers, 22*83 cal., and therefore a difference due 

 to error in reading each thermometer of 11*42 cal. 



Subtracting these two quantities from I. and II. respec- 

 tively, and halving the results, we obtain : — 



A. In a single experiment where one thermometer is used : — 



(1) Error of reading ±5*71 cal.* 



(2) Error due to manipulation ... ±7*59 „ 



Total ±13*3 



V 



B. In a single experiment where two thermometers are 

 used: — 



(1) Error of reading ±11*42 cal. 



(2) Error due to manipulation ... ± 9*53 



->/• 



Total +20*95 „ 



These duplicate experiments being performed at exactly 

 the same temperature, no difference will appear between them 

 due to errors of calibration or standardization ; but in regard- 

 ing the absolute value of the numbers deduced from them, this 

 source of error must also be taken into account. The amount 

 of this error has already been shown to be 11*12 cal. when the 

 results of two instruments are compared (VI.) ; the total 



* Corresponding to -021 millim. of the mercurial column. 



