76 Proceedings of the Royal Society 
bend the disc while it is rotating, we find that the bent part does 
not rotate with the disc, and that the disc only slowly regains its 
original flat form. If we load the outside of the disc with a row 
of flattened pellets of shot, we increase the resistance or rigidity 
of the disc while in motion, and if the weight is such that it just 
balances the elasticity of the paper, then the bend will remain in 
the same place for a very long time while the disc is rotating 
rapidly. The disc may even be bent till the circumference touches 
the centre, and while the bend keeps its place the chain of shot is 
passing through many planes, and the tension at the different 
points just balances the centrifugal force. 
Before proceeding to experiment with the horizontal chain, I 
must refer to the disturbing forces in action causing the chain to 
change its form while in motion. When looking at these endless 
chains in motion, the most marked effect of this motion is to cause 
a curious reverse curve just after the chain has turned at the lowest 
point of its path and has begun to ascend. This reverse curve was 
supposed to be produced by friction from the great tension produced 
by the centrifugal force; but that it is not really so, is easily 
proved by taking two precisely similar chains and oiling one and 
passing the other through a flame to remove all grease. The only 
difference between the two chains now is that the friction in the one 
is greater than in the other. If we hang these two chains over two 
pulleys of the same diameter on the same shaft so as to drive both 
chains at the same velocity, we find that the oiled chain has the 
reverse curve well marked, while the friction in the other chain 
causes the loop to open out and take up a curve approaching a circle 
and shows no reverse curve, and when both chains are compelled to 
have the same curvature at bottom, the reverse curve is much the 
least where the friction is greatest. The reverse curve seems to be 
due to the change of motion which takes place in the links when 
moving in a path of varying curvature. For instance,, when the 
links are descending along the flat part of the curve, their motion is 
almost simply one of translation, whereas when passing round the 
curves they have a motion of rotation as well as a motion of transla- 
tion, the result of which is, that the links resist this rotation at the 
entrance of the curve, and thus flatten out the curve on that side, 
aud after the rotation has been communicated to them, they tend to 
