252 Proceedings of the Royal Society of Edinburgh. [Ses's. 
In the course of the carrying out of the tests the specimen was very 
frequently removed, in order that the zero might be tested, a proceeding 
which was found to be very necessary, as the small changes in the position 
of the zero which took place in the course of a test were sometimes found 
to be large enough to make the readings for sensitive state quite unreliable, 
and a large number of readings had frequently to be taken before agree- 
ment could be obtained. 
Before going on to the discussion of the results obtained for the different 
specimens, it may be worth while pointing out how very close is the 
analogy between the effect of strain and the effect of thermal treatment. 
Fig. 3 (i.) shows a hysteresis curve for Specimen I. (0’15 per cent. C) taken 
after the specimen had been demagnetised with a load of 10 kilos on, and the 
load had been removed. As will be seen, the hysteresis curve is not closed, 
as it would have been had the specimen been demagnetised just previous 
to being tested, and the value of the intensity of magnetisation corre- 
sponding to the maximum field employed is distinctly less after even one 
reversal of the field. Indeed, the first reversal removes about 50 per cent, 
of the total improvement produced by the alteration of stress. If we 
compare with this curve that shown in fig. 3 (ii.), a reproduction from a 
paper by Gray and Ross on the effect of thermal treatment in inducing 
