My own earliest models employed a light wooden frame with two propellers, which were each driven by a strand of twisted rubber.[13] In later forms, the rubber was enclosed and the end strains taken up by the thinnest tin-plate tubes, or better still, paper tubes strengthened by shellac. Little was known to me at that time as to the proper proportions between wing surface, weight and power; and while I at first sought to infer the relation between wing surface and weight from that of soaring birds, where it varies from 1/2 to 1 sq. ft. of wing surface to the pound, yet the ratio was successively increased in the earlier models, until it became 4 sq. ft. to 1 pound. It may be well to add, however, that the still later experiments with the steam-driven models, in which the supporting surface was approximately 2 sq. ft. to the pound, proved that the lack of ability of these early rubber-driven models to properly sustain themselves even with 4 sq. ft. of wing surface to the pound, was largely due to the fact that the wings themselves had not been stiff enough to prevent their being warped by the air pressure generated by their forward motion. During the years I presently describe, these tentative constructions were [p010] renewed at intervals without any satisfactory result, though it became clear from repeated failures, that the motive power at command would not suffice, even for a few seconds’ flight for models of sufficient size to enable a real study to be made of the conditions necessary for successful flight. In these earliest experiments everything had to be learned about the relative position of the center of gravity, and what I have called the center of pressure. In regard to the latter term, it might at first seem that since the upward pressure of the air is treated as concentrated at one point of the supporting surface, as the weight is at the center of gravity, this point should be always in the same position for the same supporting surface. This relation, however, is never constant. How paradoxical seems the statement that, if ‹ab› be such a supporting surface in the form of a plane of uniform thickness and weight, suspended at ‹c› (‹ac› being somewhat greater than ‹cb›) and subjected to the pressure of a wind in the direction of the arrow, the pressure on the lesser arm ‹cb› will overpower that on the greater arm ‹ac›! We now know, however, that this must be so, and why, but as it was not known to the writer till determined by experiments published later in “Experiments in Aerodynamics,” all this was worked out by trial in the models. [Illustration: FIG. 1. Diagram of suspended plane showing position of C. P.] It was also early seen that the surface of support could be advantageously divided into two, with one behind the other, or one over the other, and this was often, though not always, done in the models.