We use a two-dimensional (2-D) finite element code to investigate how mechanical properties and boundary conditions influence progressive strain localization during extension of a 2-D elastoplastic layer. The deforming medium is modeled using a strain softening Von Mises rheology, with a Gaussian heterogeneity in yield strength distributed randomly in space. Specifically, we examine the effects of changing (1) the thickness of the deforming layer, (2) the basal boundary condition, (3) the range of strengths in the deforming layer, (4) the strength loss on failure, and (5) the total accommodated strain. Discrete zones of plastic shear strain are observed to nucleate and grow progressively during each experiment. In order to quantify and thus discriminate between the deformation patterns produced under differing conditions, we calculate the size-frequency distributions of the shear band populations. Size is defined as the total plastic strain represented by each discrete structure. A wide range of size-frequency distributions is observed, including both power law and exponential as well as distributions that show breaks in scaling with a transition from power law (at small sizes) to exponential (at large sizes). We show that these variations in population statistics are directly reflecting a change in strain localization in space and time caused by changing model parameters. Furthermore, by making an analogy between the model shear bands and tectonic faults, we provide insights into key features of fault size-frequency distributions that have been measured in continental and oceanic extensional settings and also observed in analogue experiments.