Identification of mechanical properties is essential to conduct reliable numerical simulations but remains a challenging task. Usually, a separate test is required to determine each material property and there are no standardised identification procedures for some parameters (e.g. those governing the cracking behaviour of brittle materials, represented often by softening in continuum constitutive models). The Virtual Fields Method (VFM) is an inverse identification strategy which can determine material parameters under uncertain load and boundary conditions which induce heterogenous stress states. It departs from the weak form of the equilibrium equation of an arbitrarily shaped solid and identifies the constitutive parameters by means of measured strains. Direct identification of properties is possible for linear stress-strain relationships, while iterative algorithms need to be implemented for nonlinear constitutive models. This paper is a preliminary investigation on the use of VFM to identify the nonlinear parameters of a simple elasto-plastic model with linear softening. A tensile loading test on a bar is numerically simulated and the identification uses the force and strain data as input. The results demonstrate the need to regularise the material properties depending on the size of the measurement strain interpolation window to obtain accurate identification. The choice of virtual fields (manual or sensitivity-based) is also explored. Findings highlight the feasibility of identifying the parameters governing the softening behaviour of brittle materials. This carries practical significance for the mechanical identification of masonry units and mortar, which are constructed of brittle materials.