DICKINSON: COMBUSTION CALORIMETRY 435 



is known, to be used to check serious errors in either the calori- 

 metric system or the electrical calibrating system. 



Observations have been made with two different calorimeters 

 built especially for the purpose and each calibrated by the above 

 method several times independently. Both calibrations and com- 

 bustions cover a period of more than three years during which 

 time hundreds of observations have been made with different 

 electrical equipment, and samples of material obtained from dif- 

 ferent sources and purified at different times and in different ways. 



Determinations of the heat of combustion of napthalene gave 

 9622 =*= 2 (20°) calories per gram weighed in air, with a maxi- 

 mum deviation from the mean of about 5 in 10,000 for groups of 

 observations upon the same samples and about the same maxi- 

 mum deviation of different groups of observations from the mean 

 of all, regardless of the sample. 



Determinations of the heat of combustion of benzoic acid 

 gave 6329 ± 1 (20°) calories per gram weighed in air, with a 

 maxmimn deviation of about 1 in 1000 for the earlier experi- 

 ments and 5 in 10,000 for the later ones. Observations taken 

 on samples, some by no means pure, from different sources, 

 show a maximum deviation of 15 in 10,000 and a mean deviation 

 of 7 in 10,000. 



Determinations of the heat of combustion of sucrose, fewer in 

 number, gave 3949 ± 2 (20°) calories per gram weighed in air. 

 The later observations show a maximum deviation of a little less 

 than 1 in 1000 and a mean deviation of about 3 in 10,000, though 

 the earlier ones show a maximum deviation of 15 in 10,000. 



It appears that, of the three materials included in this investi- 

 gation, benzoic acid is the most desirable as a combustion stand- 

 ard, as indicated by the agreement between the results of different 

 observers. Napthalene has been found very reliable and con- 

 venient, although it requires care in handlirg, since a gram 

 bricquet will lose more than 1 mgm per hour by sublimation.. 

 An accuracy of 3 parts in 10,000 is attainable. Sucrose seems 

 not to be so well adapted for use as a combustion standard as 

 is benzoic acid, because of its lower heat of combustion, its fre- 

 quent failure to ignite, and the lower precision of the results, 

 obtained. 



