206 Proceedings of Royal Society of EdinhiorgJi. [sess. 
junction from the less carefully annealed iron to the more carefully 
annealed, and from the hard drawn to the annealed. The wire, 
which was simply made yellow-hot in air by a dynamo current, was 
found to have its line nearly parallel to that of the annealed, and 
presented no discontinuity as in the case of the wire which had 
occluded the gas. The effect of the dynamo current was smaller 
than that of less careful annealing. 
In the experiments with iron wires occluding these various 
gases, the peculiarity was sometimes remarkably well defined, 
but sometimes imperfectly so at high temperatures. In the 
latter cases the heating and cooling curves also intersected at high 
temperatures. The change of colour of the surface of hydro- 
genised iron by heating had suggested to me the importance of the 
condition of the surface of the affected wires. In the earlier experi- 
ments, the ends of the wires where the junctions were made were 
rubbed with sandpaper to make the contact surer. Later, I left 
them untouched, and the peculiarity was always well defined and 
took place at low temperatures. Thus it seems clear that the effect 
of occlusion is largely confined to the surface. 
Invariably the cooling curve of the charged iron lay above the 
heating curve, while in annealed iron and iron in which gas was not 
occluded, the curve of cooling was the lower. 
In the case of CO-iron (a, PI. I.), the electro-motive force be- 
tween thin iron and CO-iron rose steadily with the temperature to 
about 180°, and then suddenly decreased as the temperature rose a 
few degrees ; hut as the heating continued it began again to rise 
steadily, giving a curve entirely different from the earlier one. 
Thus it is evident that carbonic oxide suddenly escapes at about 
180°, the wire regaining a stable condition at about 190°. During 
cooling, the curve runs without break between these two tempera- 
tures. The cooling curve lies above the second part of the heating 
curve, but below the first part. Possibly the end of the wire 
partially recovers some portion of occluded gas from neighbouring 
cooler parts of the wire. Another annealed wire, cut from the same 
piece, hut not made to occlude gas, gave a heating curve running 
parallel to the former curve, but passing between its first and second 
parts, and on cooling gave a curve parallel and close to its own 
heating curve. From this indirect comparison, it maybe concluded 
