CALCULATION OF RESULTS. 91 



ACCURACY OF CALCULATIONS OK THE RESIDUAL AMOUNT ov OXYGEN. 



In calculating the volume of oxygen remaining in the apparatus at 

 the end of each experimental period according to the method here de- 

 scribed, it is seen that the sum of all the errors in the determinations 

 of carbon dioxide and water, as well as the errors in the calculations of 

 the different volumes in the apparatus and of the nitrogen admitted, 

 affect directly the calculation of the amount of ox)'gen present. This 

 is a serious defect in this method of calculation. With the present 

 arrangements for sampling and analyzing the air, however, it is be- 

 lieved that the values obtained in a residual analysis represent very 

 correctly the actual amounts of water vapor and carbon dioxide in the 

 sample. Similarly it is probably true that the amount of nitrogen ad- 

 mitted to the chamber is known with sufficient accuracy. The main 

 source of error therefore lies in the calculations of the different vol- 

 umes in the apparatus, and of the factors affecting these calculations 

 that of temperature is open to the most serious criticism. 



THERMAL GRADIENT INSIDE THE CHAMBER. 



Inasmuch as the computation of the true volume of the gas inside 

 the respiration chamber depends in large measure upon a correct 

 knowledge of the average temperature of the mass of gas, it is neces- 

 sary to consider in detail the accuracy of its temperature measurements. 

 While the electrical resistance thermometers, both for the air and for 

 the copper wall, undoubtedly give a very accurate measure of the fluctua- 

 tions in temperature of their environment, it nevertheless remains a fact 

 that the inside of the respiration chamber contains a mass of air which 

 is subjected in different parts to widely varying temperatures. In the 

 case of the alcohol check experiments we have a thermal gradient extend- 

 ing from the high temperature of the alcohol flame down to the tempera- 

 ture of the incoming water of the heat-absorbing system. In this case 

 great heat is concentrated at one point, while the cooling area, i. e., the 

 area of the absorbers, is quite extensive. In the case of experiments 

 with man we have a body temperature, which on the surface is not far 

 from 33, affecting a relatively large area, and a cooling area similar 

 to that during the alcohol check experiments. In severe work experi- 

 ments we have the body more or less exposed and the temperature of 

 the absorbing system cooled nearly to zero. It is therefore difficult 

 to see how the electrical resistance thermometers, distributed as shown 

 in figure 33, can in any way assume accurately the average temperature 

 of the air in the whole chamber. During rest experiments and alcohol 



