22 



August Krogh and Marie Krogh. 



Table of test experiments. 



Loss of CO, per hour 0.43. 



The surface of the respiration chamber was about 10 times that 

 of the box and the mean percentage of CO about 0.6 °/o. The actual 



loss per hour through identical walls should therefore be . ' 



1.84 liter, but as the walls were about 4 times thicker and of much 

 tighter material it cannot be put higher than 0.5 1. or 1 "/^ of the mean 

 production of CO2 per hour. 



It is probable that there is a similar systematic error in the oxygen 

 values but on this point we have no experimental evidence. If the 

 O2 diffusion is different from that of CO2 a systematic error in the 

 respiratory quotient must result, but such an error will obviously 

 be very small. 



The accidental errors in the test experiments are caused exclu- 

 sively by the analytical errors as both the air current and the sampling 

 were absolutely uniform. In the Greenland experiments the analyti- 

 cal errors were smaller absolutely, but in relation to the percentage 

 of CO2 they were of about the same magnitude. Errors of about 

 1 — 2 "/0 must therefore be expected but, besides these, slight errors 

 may arise from want of correspondence between the meter and the 

 sampling arrangement. The revolutions of the meter ought to have 

 governed the sampling after the manner adopted by Zuntz. We had 

 the sampling independent of the meter and could only obtain quite 

 accurate results if both were working at a uniform rate. The influence 

 of even large deviations is however surprisingly small. 



If we assume f. inst. that during an experimental period of 2 

 hours the ventilation is suddenly increased 20 per cent after 1 hour, 

 while the gas exchange and sampling remain constant, the resulting 

 error in the gas exchange for the whole period is less than 1 per cent. 

 If the gas exchange rises 20 % the error becomes nil and if it falls 



