In Fig. 36, _a_, _a_ is the body of the apparatus, partly of gunmetal and partly of wood. It is provided with a steel shaft to which the screw _h_, is attached, and also with a cylindrical pulley for taking the belt. The thrust of the screw pushes the shaft inwards and records the lift at _o_ (Fig. 32). The corners of the aeroplane _g_, _g_, are attached by wires to the steel plate _e_. _b_, _b_, is a four-arm spider for holding the ends of the parallel bars _c_, _c_, and _d_, _d_, show vertical steel bars to which all devices to be tested are attached. In testing aeroplanes, weights may be placed at _e_, sufficient to balance the lifting effect, and then by adding the weight to the upward pull of the aeroplane, the true lift of the aeroplane is obtained. It is also possible to attach an aeroplane at _e_, that is, the machine was made to test superposed aeroplanes if required. In these experiments, I naturally assumed that the best position for a screw was at the rear and in the path of the greatest resistance, but as some experimenters with navigable balloons placed the screw in front in order to pull the apparatus along instead of to push it, I made experiments to see what the relative difference might be. In order to do this, I wound a large amount of rope one-half inch in diameter around the whole apparatus forward of the screw, converting it into an irregular mass well calculated to offer atmospheric resistance. Upon starting the engine, I was rather surprised to see how little retardation these ropes gave to the apparatus. It appeared to me that nearly all of the energy consumed in driving the ropes through the air was recovered by the screw. I then removed the right-hand screw and replaced it by a left-hand screw of the same pitch and dimensions (Fig. 37_a_). I then found that the blast of the screw blowing against the tangle of ropes greatly retarded the travel; in fact, with the same number of revolutions per minute, the velocity fell off 60 per cent. I think that these experiments ought to show that there is but one place for the screw, and that is at the stern, and in the direct path of the greatest atmospheric resistance. [Illustration: Fig. 37_a_.--Right- and left-hand four-blade screws used in my experiments for ascertaining the comparative efficiency between screws placed in front and in the rear of the machine.] [Illustration: Fig. 38.--Apparatus for indicating the force and velocity of the wind direct without any timing, counting, or mathematical calculations.]