Influence of structural size on axial stress-lateral strain curves due to shear localization is investigated and a new method for determining the axial stress-lateral strain curves is proposed according to the axial stress-strain curves obtained by tests for different sizes. Using the analytical solution of the axial stress-axial strain curve based on gradient-dependent plasticity

linear regressions on the experimental axial stress-axial strain curves in strain-softening regime are carried out to calculate the number of shear band for different widths. After the number is determined

the axial stress-lateral strain curves for different sizes can be obtained according to the analytical solution of axial stress-lateral strain curve depending on the number of shear band and width of specimen. It is found that the ratio of the number of shear band to the width of specimen is a key parameter for describing the axial stress-lateral strain curves. If the ratio remains a constant

then the curves are size-independent. Otherwise

the post-peak curves are dependent on the width of specimen. For a single shear band appears in specimens with different widths

specimen with larger width leads to steeper curves and even exhibits snap-back behavior. The method proposed can also be used to analyze the problem of size effect of axial stress-circumferential strain curves. The reason for the size effect is shear localization

but the case that localization can not cause the size effect is not also impossible.