360 Proceedings of the Royal Society of Edinburgh. [Sess. 
line portion to obtain true values of deflection. The author has little doubt 
that an error due to slackness would occur even in riveted plates ; for no 
matter how the plates were mechanically fixed to the retaining walls, there 
would probably be that minute give due to the limitations that are involved 
in making absolute and uniform contact between two surfaces, which, 
being magnified at the centre, especially with large plates, would produce 
a considerable zero error. 
3. From B onwards. At a pressure varying from two to six times that 
which first produces permanent set, the elastic line AB begins to bend 
downwards, and there occurs the portion BC. The author believes this 
declension to be due solely to the arching of the plate ; that above a certain 
deflection the plate does not follow the flat-plate laws. In fact, from a flat 
it becomes an arched surface, and the resistance to bending increases. 
For the smaller size plates the zero error is of a different nature. The 
initial want of flatness becomes negligible, and the error is almost solely 
due to give of the fixed edges. This initial slackness is not generally taken 
out by a few pounds pressure, but is only gradually removed. Hence 
readings for deflection for the first 20 or 30 lbs. pressure cannot be relied 
on, as they include not only the actual deflection reading, but the reading 
due to slackness as well. Therefore if values are taken from this portion 
of the curve they will be too high. If the graphs for plates E and F 
above be examined, it will be seen that the straight line has been drawn 
through the further points. These points are not on the falling portion of 
the curve, for with the smaller plates, especially the 4-in. size, it was found 
practically impossible to reach this region. 
Turning once more to the results given, it will be seen that the 
