1909-10.] Method of investigating Certain Systems of Stress. 43 
though how it could he greatly improved in any simple manner is not easy 
to see. The readiness with which the departure of a line from straightness 
is detected depends somewhat upon the scale of the diagram : enlargements 
of the photographs to nearly three times full size were used for this 
purpose. Considerable irregularities were found in the positions of the 
individual points ; but when a mean curve was drawn, the corresponding 
ordinates of each of the four branches, measured from the transverse centre 
line of the diagram, were found to agree fairly well. The ordinates of the 
curves plotted are the means of the four separate branches obtained in this 
way. The inner curves pass upwards or downwards from a point of zero 
curvature on the edge of the notch, and when near the middle of the face 
they bend over sharply, and all four branches meet at the centre of the 
face. The exact form of the curves near the centre point is difficult to 
determine. The outer curves start from the same or from another point 
of zero curvature — according to the shape of the notch — and, receding from 
the transverse centre line, pass to infinity along the longitudinal centre 
line of the face. 
When these dotted curves are transferred to the photograph of a bar, 
the outer ones are found to form the boundary, as well as could be expected, 
between the area from which the scale has dropped completely and those 
parts which show well-developed Liiders’ lines. Since the stress lines 
(stream lines) are sensibly straight in the region between the two sets of 
dotted curves, the deformations within this area may be either normal 
to, or inclined to, the direction of stress. In the area enclosed between the 
inner sets of dotted curves the deformation lines are normal to the stress 
lines, and beyond the outer dotted curves the deformation lines are inclined 
at 50° to the stress lines. Thus there would be, in all probability, more 
loss of the oxide scale between the two sets of curves than elsewhere. The 
difference at the outer curve is noticeable on all the bars, but that at the 
inner one is not, although it can be seen in fig. 4. The lack of deformation 
at the parts of the outer dotted curves near the longitudinal edges of each 
bar is due to the small stress there. Traces of the boundary indicated by 
the inner dotted curves are found sometimes upon broken bars. Fig. 70- 
shows an example, somewhat spoilt by a surface flaw near the middle. 
The curved outlines of the fracture are of the same general form as 
the inner dotted lines of fig. 7b, obtained in the manner described 
above, but they are by no means identical. It is difficult to reproduce 
exactly the shape of such a bar upon the flow template. Moreover, it 
must not be forgotten that fracture is not instantaneous, but proceeds 
gradually, and that the stress system alters during the process ; and 
