In order to subdivide experiments into periods as short as 1 or 2 hours, it is necessary to deflect the air-current at the end of each period from one set of purifiers to the other, in order to weigh the set used and to measure the quantity of carbon dioxide and water-vapor absorbed. The conditions under which these changes from one system to another are made, and which call for an absolutely gas-tight closure, have been discussed in detail elsewhere.[21] It is sufficient to state here that the very large majority of mechanical valves will not serve the purpose, since it is necessary to have a pressure of some 40 millimeters of mercury on one side of the valve at the entrance to the absorber system and on the other side atmospheric pressure. A valve with an internal diameter of not less than 25 millimeters must be used, and to secure a tight closure of this large area and permit frequent opening and shutting is difficult. After experimenting with a large number of valves, a valve of special construction employing a mechanical seal ultimately bathed in mercury was used for the earlier apparatus. The possibility of contamination of the air-current by mercury vapor was duly considered and pointed out in a description of this apparatus. It was not until two years later that difficulties began to be experienced and a number of men were severely poisoned while inside the chamber. A discussion of this point has been presented elsewhere.[22] At that time mercury valves were used both at the entrance and exit ends of the absorber system, although as a matter of fact, when the air leaves the last absorber and returns to the respiration chamber, the pressure is but a little above that of the atmosphere. Consequently, mechanical valves were substituted for mercurial valves at the exit and the toxic symptoms disappeared. In constructing the new calorimeters it seemed to be desirable to avoid all use of mercury, if possible. We were fortunate in finding a mechanical valve which suited this condition perfectly. These valves, which are very well constructed, have never failed to show complete tightness under all possible tests and are used at the exit and entrance end of the absorber system. Their workmanship is of the first order, and the valve is somewhat higher in price than ordinary mechanical valves. They have been in use on the apparatus for a year now and have invariably proved to be absolutely tight. They are easy to obtain and are much easier to manipulate and much less cumbersome than the mercury valves formerly used. COUPLINGS.