Elastoplastic toughness testing of pure aluminium replicated foam is conducted following a J-procedure adapted from the ASTM E1820-08a standard using disc-shaped compact tension specimens. Tests cover a wide range of foam relative density V-m (10-24%). Resulting data show pronounced R-curve behaviour, computed J values increasing steadily beyond the crack blunting line before reaching a plateau, corresponding to a "steady-state" J value. Fractography reveals that the crack propagates via the rupture of struts normal to the crack plane, this being accompanied by limited plastic deformation of the struts near the crack plane, the extent of which, made visible by slip markings along the strut surface, increases in extent with V-m. Intact and fractured struts coexist over a significant portion of the crack plane, indicating that the strut rupture is stochastic. Measured crack initiation and steady-state propagation J values vary roughly as V-m(3); this is a stronger dependence than has been observed in commercial (closed-cell) metal foams. A simple model is proposed to describe the initiation toughness of these open-cell foams, which are characterized by a greater matrix ductility than other metal foams characterized in the literature. The model is based on an estimate of the crack tip opening displacement at the instant when a strut aligned in the loading direction fails by ductile rupture just ahead of the crack tip. The model accurately predicts the scaling law observed here between the toughness and the relative density of the foams, and provides a relatively good predictor of the absolute value of fracture toughness up to a relative density near 20%, while at higher densities measured toughness values exceed predictions. (C) 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.