It is readily seen that the test piece having the I-beam section weighed less than the hollow square in the first and second tests, and had a deflection of less than half. This I-beam section, however, did not show up so well when a longer piece was tested, for as soon as the length was made appreciably greater than a metre it began to twist, the twisting becoming more and more serious the [p198] greater the length, until with a piece 11 feet long, the full length of a cross-rib, the twisting was so serious as to make the rib practically useless. It was at first thought that this twisting might be overcome by making the webs slightly wider, and it would to a certain extent, but in looking ahead and planning how the cross-ribs were to be fastened to the main ribs, the I-beam section was seen to present so many difficulties that it was thought hardly worth while to spend time on further experiments with it. This decision was made all the more imperative by foreseeing the difficulty of bending the I-beam section to the curve which the cross-ribs were to have. In fact it had been found by experience that while many different forms of ribs could be bent to the proper curve by steaming and clamping them over a form and then drying them out while still clamped to the form, yet the grain of the wood varied so in different ribs, that of a dozen steamed and bent over the same form it was seldom that as many as three would have approximately the same amount of curvature when removed from the form after drying. If, however, the curve was formed in the ribs by making them in two parts, which were glued together and clamped up on the form while the glue dried, practically any number could be made which would have the same curvature when thinned down to the proper thickness of wall. It was recognized at all times that the gluing together of the ribs not only entailed extra work, but introduced an element of uncertainty unless some kind of a varnish for the ribs could be found which would prevent any possibility of the glue becoming soft from moisture in the atmosphere or from the wings actually coming down into the water when the aerodrome was tried in flight. A search was therefore made for a varnish that was water-proof. A large number of different varnishes were tried, and one was finally found which, after repeated tests, seemed to be thoroughly good. Several test ribs were given three coats of this varnish, and were then kept immersed in water for 24 hours without the glue showing any signs of softening. It was therefore decided to follow the plan of gluing the ribs together and protecting them with three coats of this varnish, which seemed to possess the remarkable properties of being not only impervious to water, but also unaffected by the application of concentrated ammonia or of gasoline, either of which produces immediate softening when applied to ordinary varnishes.