We are now getting the air-pressure through the pipe in both ends of the tube alike, and do not move the pellet either way. To make it move we must do something more, and open one end of the tube, and close that fork of the air-pipe, and thus get all the pressure on one side of the pellet. Remove one finger from the end of the tube, and pinch the fork of the air-tube that is on that side. The pellet will now move toward that end of the tube which is open. Reverse the process, and it can be pushed back again with air-pressure to the other end, and so on indefinitely. Let us improve the process. We will close each end of the tube permanently, and insert four cocks in the tube and forked pipe. We have here two tubes inserted at each end of the large tube, and in each of these is a cock. We have each cock connected by a rod to the lever set on a pin in the middle of the tube. We must have these cocks so arranged that when the lever is moved (say) to the right, A. is opened and B. is closed, and D. is opened and C. is closed. Now if the air-pressure is constant through the forked air-tube, and the cock E. is open, if the top of the lever is moved to the right, the pellet will be pushed to the left in the large tube. If the lever is moved to the left, and the two cocks that were open are closed, and the two that were closed are opened again, the pellet will be sent back to the other end of the tube. This movement of the pellet in the tube will occur as often as the lever is moved and there is any air-pressure in the forked tube. There is a _supply_-cock, opened and an _escape_-cock closed, and an escape-cock _opened_ and a supply-cock _closed_, at each end of the tube, _every time the lever is moved_. [Illustration: Fig. 2] We are using air instead of steam, and the movement of these four cocks all at the same time, and the result of moving them, is precisely that of the slide-valve of a steam-engine. The diagrams of this slide-valve would be difficult to understand. The action of the cocks can be more readily understood, and the result, and even much of the action, is precisely the same. But to make the arrangement entirely efficient we must go a little further into the construction of a steam-engine. The pellet in the tube has no connection with the outside, and we can get nothing from it. So we give it a stem, thus: and when we do so we change it into a piston and its rod. Where it passes through the stopper at the end of the tube it must pass air- (or steam-) tight. Then as we push the piston back and forth we have a movement that we can attach to machinery at the end of the rod, and get a result from. We also move the cocks, or valves, automatically by the movement of the rod. [Illustration: Fig. 3]