An assessment is made of the J-integral test procedure for initial crack growth in an open-cell aluminium alloy foam by combining finite element (FE) simulations with experiment. It is found experimentally that a zone of randomly failed struts develops ahead of the primary crack tip, and is comparable in size to that of the plastic zone. Hence, a crack tip J-field is absent at the initiation of crack growth from the primary crack tip. This implies that the measured J_IC value and the J versus crack extension Da curve cannot be treated as material properties despite the fact that the specimen size meets the usual criteria for J validity. The toughness tests were performed on a single-edge notched bend specimen, and crack extension was measured by the direct current potential drop method, by digital image correlation and by X-ray computed tomography. The crack growth resistance of the foam is associated with two distinct zones of plastic dissipation: (i) a bulk plastic zone emanating from the crack tip (containing a cluster of randomly failed struts), and (ii) a crack bridging zone behind the advancing crack tip. The applicability of a cohesive zone model to predict the fracture response is explored for the observed case of large scale bridging. To do so, FE simulations are performed by replacing the discrete lattice of the open-cell metallic foam by a compressible, elastic-plastic hardening solid while the fracture process zone in the foam is represented by a cohesive zone, as characterised by a tensile traction versus separation law. A detailed comparison of the cohesive zone model with experimental observations reveals that it is possible to capture the load versus displacement response but not the details of the fracture process zone using a single set of process zone parameters.