The application of Griffith＇s energy balance argument to cracks at adhesive interfaces is studied. Adhesive interfaces are generally brittle, representing the simplest form of fracture mechanics geometry because cracks are constrained to travel along the interface, giving a defined crack path which eases analysis. Experimentally, such cracks may be propagated along the interface between optically smooth rubber pieces, and measured through the transparent material. The development of adhesive fracture test-pieces since Griffith＇s time reveals difficulties in his reasoning, and allows improved understanding of the energy balance method. The most important conclusion is that stress does not normally enter the cracking criterion. It is demonstrated experimentally that stress may remain constant while the crack criterion changes. The strength of an adhesive interface is shown to be a meaningless parameter; instead, the work of adhesion, or adhesive energy, which is the work of adhesion together with energy losses, should be used to define the behaviour of cracks at interfaces.