292 W. H. WARREN AND S. H. BARRACLOUGH. 
by increase of temperature, but it is not possible’ to deduce from 
the curves the exact relationship between the two. The behaviour 
of a metal such as copper, under gradually increasing stress, is 
well illustrated by Fig. 9, which is the autographic stress-strain 
record of test 18. In this test the load was first increased up to 
a point, a little past the elastic limit, at which marked plastic 
deformation, corresponding to permanent shortening of length and 
increase of area, was evident. The load was then removed and 
re-applied, but no further change in the test piece took place till 
the maximum load previously applied was reached, the previous 
increase of area being sufficient to reduce the stress per unit area 
to a value lower than was necessary to produce plastic flow. 
10. The elastic limit in compression.—Two compressive tests, 
Nos. 53 and 54, were made on cylindrical test pieces approximately 
five inches long and an inch in diameter, with Marten’s mirror 
apparatus, to determine the position of the true elastic limit and 
the shape of the stress-strain curve at low loads. The observations 
as taken are given in Table IV., and the corresponding curves 
are plotted on Fig. 7, B and C. 
On examining curve B, the position of the points at the lower 
loads would indicate that the stress-strain line was curved con- 
tinuously, and that there was therefore no definitely marked elastic 
limit. To check this, test 54 was made, in which the load was 
reduced at different points of the test, and observations were taken 
as to whether the test piece perfectly recovered itself. As will 
be seen in the table, the recovery up to a load of 1500 ibs. was 
perfect ; beyond that load the matter is doubtful, as the point “c” 
at a load of 2000 Ibs. appears to show a slight permanent set 
(which may however, be due to an error of observation), whereas 
the point ‘“c” at 3000 Ibs. shows an apparent perfect recovery 
_ (which again may be due toan error of observation). The uncer 
tainty is however confined within fairly narrow limits, so that the 
elastic limits as shown on curves Band C are probably not subject 
to a large error. 
