became small, the effectiveness of the device would decrease.
A number of rather difficult experiments were performed last year to test this point. The chambers, C, Fig. 3, used in air, were suspended by a spring from a 3-inch pipe-cap, A. The mass of the movable system was increased by lead collars, K; and a strip of smoked glass, G, was provided to record the, displacements produced by firing. This is, of course, a very condensed description of the apparatus.
In order to approximate as closely as possible to the condition of firing into a tank of infinite extent, a special pipe system was constructed (by autogeneous welding), Fig. 4. The system shown in Fig. 3 was introduced at the top of the upright pipe, P, of Fig. 4. It will be seen that, on firing, the expelled gases passed around the circular part of the 3-inch pipe system until their velocity had been reduced to a small value by friction. It was found that the rebound of the gases was rarely sufficient to tear very thin tissue paper at T, Fig. 3. By means of an entirely different form of tank, which permitted actual measurements of the rebound to be made, it was demonstrated that such rebound was negligable. All of the experiments in vacuo showed that the recoil was practically the same as that at atmospheric pressure, down to a pressure of 0.5 mm. The recoil was therefore the result of an actual jet of gas, and was not due to reaction against the air.It will be evident that a heavy steel chamber, such as described, could not compete against the ordinary rocket, even with the high velocities which I attained. If, however, successive charges were fired from the same chamber, much as in