and Voit, if the weight of the excretions and the loss in
body-weight are taken into consideration, the difference between the
weight of the excretions and the loss in body-weight should be the
weight of the oxygen absorbed. With this apparatus we are able to
determine the water-vapor, the carbon-dioxide excretion, and the weight
of the urine and feces when passed. If there is an accurate
determination of the body-weight from hour to hour, this should give the
data for computing exactly the oxygen consumption. Moreover, we have the
direct determination of oxygen with which the indirect method can be
compared.

In the earlier apparatus this comparison was by no means as satisfactory
as was desired. The balance there used was sensitive only to 2 grams,
the experiments were long (24 hours or more), and it seemed to be
absolutely impossible, even by exerting the utmost precaution, to secure
the body-weight of the subject each day with exactly the same clothing
and accessories. Furthermore, where there is a constant change in
body-weight amounting to 0.5 gram or more per minute, it is obvious that
the weighing should be done at exactly the same moment from day to day.
It is seen, therefore, that the comparison with the direct oxygen
determination is in reality an investigation by itself, involving the
most accurate measurements and the most painstaking development of
routine.

With the hope of contributing materially to our knowledge regarding the
indirect determination of oxygen, the special form of balance shown in
fig. 9 was installed above the chair calorimeter. This balance is
extremely sensitive. With a dead load of 100 kilograms in each pan it
has shown a sensitiveness of 0.1 gram, but in order to have the
apparatus absolutely air-tight for the oxygen and carbon-dioxide
determination, the rod on which the weighing-chair is suspended must
pass through an air-tight closure. For this closure we have used a thin
rubber membrane, weighing about 1.34 grams, one end of which is tied to
a hard-rubber tube ascending from the chair to the top of the
calorimeter, the other end being tied to the suspension rod. In playing
up and down this rod takes up a varying weight of the rubber diaphragm,
depending upon the position which it assumes, and therefore the
sensitiveness noted by the balance with a dead load and swinging freely
is greater than that under conditions of actual use. Preliminary tests
with the balance lead us to beli