. 12 
over 4 per cent too large. This excess of water may have been due 
to either or all of several causes. There may have been a small leak 
whereby air from the room was drawn into the ventilating air current 
at some point beyond the "freezer'' a for the incoming air; but care- 
ful examination failed to reveal any noticeable leak inside the chamber, 
though later experiments implied that there may have been a leak in 
the "valve box," through which the air passed on its way from the 
chamber to the meter pump. It is possible that some of the moisture 
condensed on the outside of the valve box may have found its way 
into the air current. This would increase the amount of water found 
by analysis of the outgoing air. but would not materially affect the 
determination of either carbon dioxid or heat. There may have been 
evaporation of water which had condensed on the absorbers and walls 
of the chamber before the test began; however, there could not have 
been sufficient water present to account for all of the excess. 
Taking everything into consideration we do not feel warranted in 
assuming any definite error in the determination of water in metab- 
olism experiments No-. 25-34. Such error, if it existed, would not 
affect the nitrogen and carbon balance at all and would hardly affect 
the heat balance appreciably. From the figures for test experiments 
Nos. 16 and IT it will be observed that the excess of water found over 
the theoretical amounted to not far from 1 gram per hour. If we 
assume in experiments Xos. 25-31 a leak in the air current sufficient 
to account for about 1 gram of water per hour or 21 grams per day, 
the values given in column e of Tables 91, 106, 122. and 138 would be 
too large by a corresponding amount, and the figures in columns f and 
g of these same tables would be 11 calories too large. This amount 
would make a maximum error of 0.5 per cent in the heat in the rest 
experiments (Nos. 25, 26, and 28) and about 0.3 per cent in the work 
experiments (Xos. 29, 31, 32, and 31). 
The results of test experiments Xos. 1-9 have already been pub- 
lished. 1 ' The results of tests Xos. 10-17 are summarized in Table 1, 
together with the mean results of tests Xos. 1-9. The proportion of 
absolute ethyl alcohol (by weight) in the commercial alcohol burned in 
the lamp in the different test experiments was as follows: Test Xo. 
10, 91.03 per cent; tests Xos. 11-11. 90.85 per cent: and tests Xos. 
15-17, 90.57 per cent. 
a U. S. Dept. Agr., Office of Experiment Stations Bui. 63. p. 30. 
b U. S. Dept. Agr., Office of Experiment Stations Buls. 63, p. 50; 69, p. 13. 
