Then, again, the flight of shells was not nearly so certain as that of solid shot. The effects of eccentricity, which in the case of solid shot had always militated against accurate shooting, were in the case of shells considerably enhanced. The varying thickness of the shell, the lack of homogeneity of the metal, the presence of the protruding fuze, all tended to produce eccentricity and give a bias. The centre of gravity of a shell was seldom at its centre of figure; and this eccentricity was the cause of deviations in flight, in range and direction, which made the trajectory of a shell not easily predictable. Savants and artillerists, both here and in other countries, discussed for years these deviations, and on the relationship between range and eccentricity numbers of trials were made and theories were propounded. Which is the more strange, seeing that Robins had placed on record an almost complete solution. Briefly, the effect of eccentricity may be explained as follows. Just as a stick held vertically by a thread receives, when struck at a point in it other than the centre of percussion, a tendency to motion not only of translation but also of rotation round that centre of percussion; so a spherical shell whose centre of gravity lies away from its centre of figure receives, from the pressure of the powder gases acting at its centre of figure, a rotary motion about its centre of gravity in addition to a motion along the bore. If the centre of gravity lies below the centre of figure this rotary motion is in such a direction that, as the shell approaches the muzzle, points on its upper surface are moving towards the muzzle, points on the lower part are moving inwards. And this rotation, maintained during flight, has the effect--as was demonstrated by Robins with the musket ball--of giving the sphere a vertical deviation in a downward direction; i.e. of reducing its range. It follows, then, that an artificial increase of range could be obtained by placing the sphere with its centre of gravity _above_ the centre of figure? This is precisely what was done; and by many a measured eccentricity was considered a desideratum, as giving a higher range than could be obtained without it. With such a system, however, the deviations still remained large and flights still more irregular. And the best opinion held that the most satisfactory solution lay in reducing the errors of flight as far as possible by the use of perfectly concentric shells. This ideal was difficult of attainment. Sir Howard Douglas has described at length experiments with shells the axis of whose eccentricity was found by floating them in mercury: experiments which revealed that not one shell in a hundred of those supplied was perfectly balanced. For this reason misgiving was felt as to the effectiveness of shell fire when carried out at considerable ranges against solid shot, and efforts were continuously made to correct all shell before issue.