In the tests of the engine working in the frame, both while mounted on the car and also when suspended from the springs, a great amount of delay was caused from the fact that the ball-bearings on the transmission and propeller shafts frequently went to pieces. There were two reasons for this: In the first place, although carefully selected balls were used, defective ones were continually encountered. Even a slight defect in a single ball resulted in its breaking under the rather severe test to which they were subjected, and, as is well known, the breaking of one ball in a ball-bearing usually results in the destruction of the whole bearing, especially if the races are light. The second cause was that [p253] the whole aerodrome had been originally designed with the expectation of using a maximum of 24 horse-power, and as no margin had been left to provide for possible increases in the size of the bearings, there was no room to permit them to be increased without almost completely reconstructing portions of the transverse frame. While in the end it would have been cheaper to have reconstructed these portions in order to put in larger bearings, yet, as is always the case in experimental work of this kind, small changes which seem to hold out hope of overcoming difficulties are usually followed, rather than reconstructions which can be seen to involve considerable expense and delay. After a number of minor changes had been made in the bearings, they were finally able to stand up fairly well under the severe strain to which they were subjected when the engine developed its full power, and no further changes were made in them; a defective race being, however, replaced by a new one as occasion demanded. These tests demonstrated very clearly that at speeds of approximately 1000 revolutions per minute ball-bearings which are subjected to considerable loads should be calculated with a considerable margin of safety, as the yielding of the frame, which must necessarily be far from rigid, causes more or less error in the alignment of the shafts and bearings, and this introduces considerably increased strains on the bearings. In the early tests before the bearings were strengthened, the balls in some of the races were on a few occasions ground to a very fine powder before it was discovered that they had failed. Such a result, it will be understood, could and did occur in the course of a very minute length of time.