236 
Proceedings of the Royal Society of Edinburgh. [Sess. 
proportional to a power of the distortion. The latter postulate was 
adopted in the paper of 1898, and is sufficient in cases in which one definite 
power suits over the whole range of the observed oscillations, large and 
small alike. But, in those cases [ e.g . iron slightly at very small oscillations 
{Trans. R.S.E., 1898) and other metals very distinctly under certain con- 
ditions (Ritchie, Proc. R.S.E., vol. xxxi. p. 424, and vol. xxxiii. p. 183)] in 
which one power suits with larger oscillations and a different power suits 
with smaller oscillations, the change being very sharply marked, it seems 
to be necessary to assume that a new type of groups becomes active only 
when the distortion exceeds a definite limit. 
It might be expected that a new type of this kind would first become 
effective at the surface of the wire, since the shear is greatest there, and 
that its effect would gradually become evident in more and more central 
regions as the oscillations were made greater. In this case there could 
not be a sudden change in the value of n at a definite range of oscillation. 
The change should take place more or less gradually, unless indeed the 
dissipation of energy became very greatly increased whenever the change 
did begin. In the experimental observations the suddenness of the change 
in the dissipation is made evident by a quite sudden alteration in the drop 
of the amplitude per swing, when the decreasing oscillations pass the 
critical stage. A way of avoiding the assumption of so great a volume 
density of dissipated energy in the mere surface layers as would be needed 
to account for the suddenness of the change can be found in the assump- 
tion that the action, though starting at the surface, actually spreads in- 
wards to regions in which the stress is not sufficient to start it — -just as 
the complete disintegration of a Rupert’s drop follows a surface rupture, 
or as the tremors caused by one landslip precipitate other slips. 
Another experiment, carried out by Mr Ritchie {Proc., vol. xxxiii. p. 193), 
seems to indicate that the proper view of the action is that which regards 
the energy loss when averaged over all types of contributing groups as 
being proportional to a power of the distortion. In that experiment 
zinc was deposited electrolytically on a core of copper wire, and the 
value of n was found to pass continuously from the value characteristic 
of copper alone towards that characteristic of zinc alone, as the thickness 
of the zinc increased. In this case the zinc molecular groups and the 
copper molecular groups were entirely separate, except at the surface of 
junction. It might therefore be thought that in general the energy 
dissipated should be representable as a sum of two terms, one involving 
that power of the distortion characteristic of the one metal, the other 
involving the power characteristic of the other metal, while, at small 
