This procedure is best noted from fig. 30. The connected series of three U-tubes on the rack on the table is joined on one end by well-fitting rubber connections to the tube leading from the mercurial manometer and on the other end to the rubber tube A leading to the gas-meter. On lowering the mercury reservoir E, the mercury is drained out of the tube D and air passes through both arms of the tube and then through the three U-tubes. In the first of these it is deprived of moisture, and in the last two of carbon dioxide. The air then enters the meter, where it is measured and leaves the meter through the tube B, saturated with water-vapor at the room temperature. To remove this water-vapor the air is passed through a tower filled with pumice-stone drenched with sulphuric acid. It leaves the tower through the tube C and enters the ventilating air-pipe on its way to the calorimeter. The method of manipulation is very simple. After connecting the U-tubes the pet-cock connecting the tube C with the pipe is opened, the mercury reservoir E is lowered, and air is allowed to pass through until the meter registers 10 liters. By raising the reservoir E the air supply is shut off, and after closing the stop-cock at C the tubes are disconnected, a second set is put in place, and the operation repeated. The U-tubes are of a size having a total length of the glass portion equal to 270 millimeters and an internal diameter of 16 millimeters. They permit the passage of 3 liters of air per minute through them without a noticeable escape of water-vapor or carbon dioxide. The U-tubes filled with pumice-stone and sulphuric acid weigh 90 grams. They are always weighed on the balance with a counterpoise, but no attempt is made to weigh them closer than to 0.5 milligram. GAS-METER. The gas-meter is made by the Dansk Maalerfabrik in Copenhagen, and is of the type used by Bohr in many of his investigations. It has the advantage of showing the water-level, and the volume may be read directly. The dial is graduated so as to be read within 50 cubic centimeters. The Elster meter formerly used for this purpose was much smaller than the meter of the Dansk Maalerfabrik we are now using. The volume of water was much smaller and consequently the temperature fluctuations much more rapid. While the residual analyses for which the meter is used are of value in interpolating the results for the long experiments, and consequently errors in the meter would be more or less constant, affecting all results alike, we have nevertheless carefully calibrated the meter by means of the method of admitting oxygen from a weighed cylinder.[23] The test showed that the meter measured 1.4 per cent too much, and consequently this correction must be applied to all measurements made with it. CALCULATION OF RESULTS.