A Study of the Diet and Metabolism of Eskimos. 23 



20 % t-he error becomes 2 "/o, and if we assume on the other hand 

 that the ventilation is constant while the sampling after 1 hour be- 

 comes so much slower that the period lasts 3 hours instead of two 

 this will have no influence whatever if the respiratory exchange re- 

 mains constant and an influence of 3 — 4 per cent only if the exchange 

 rises or falls 20 per cent just at the moment when the sampling be- 

 comes slower. In the Greenland experiments the water wheel ac- 

 tuating the meter worked with such regularity that errors from this 

 source are in all cases negligible, while the sampling was never abso- 

 lutely uniform, the rate of outflow of mercury through a capillary 

 point being generally slightly slower each time it is used and some- 

 times considerably slower. The largest alteration observed in the 

 time of outflow of 100 cc. of mercury through the same capillary is 

 however from 2 hours to 3 hours and the maximum error from this 

 source may be taken as 4 per cent. Errors arising from this source 

 of 1 — 2 "/o may be of rather frequent occurrence. 



The meter was very carefully calibrated and its error cannot 

 exceed ±^ 0.3 %. 



As the temperature of the meter was read only at the beginning 

 and end of each period the assumed mean temperature may be incor- 

 rect. The largest error possible in one or two cases is 1° corresponding 

 to an error in the measured volume of 0,5 "/o- ^^ almost all cases 

 the error cannot exceed 0.2° and is therefore negligible. 



Summary of errors in respiration experiments. 



Influence of errors ou 

 Or. and CO, 



Calibration of meter + 0.3 7o systematic ] equal 



Diffusion of CO2 and 0, — 0.5 to — 2 «/0 probably — 1 7o \ probably 



systematic i about equal 



Sampling errors to + 4 7o generally 1 to 2% accidental ] equal 



Temperature errors to + 1 % gener, negligible accidental ] equal 



Analytical errors to + 3«/o gener. 1 to 2% accidental ] Larger on O.^ 



4. Other determinations. 



The food was weighed before each meal for each of the subjects 

 separately and a sample representing ^/^ of the weight eaten was 

 secured each time. As the food was not reduced to homogeneity, 

 the samples could not of course be very representative and it is ob- 

 vious when the results of the analyses are compared that large devia- 

 tions must exist in certain cases. 



The food samples for each subject and each day were placed in 

 à 500 ce. glass jar and were dried for 24 hours in a drying oven at about 

 100 — 110° in order to remove most of the water and to preserve the 

 contents. The jars were closed air tight while still hot. 



