The wedge test is of considerable use for evaluating adhesion between two bonded rigid substrates. In its (usual) static form, release of elastic strain energy is equated to effective adhesion energy during crack growth. However, the test is usually treated as two-dimensional. In fact, it is really three-dimensional due to anticlastic bending effects of the bent beam(s) during crack propagation. We studied a composite material/epoxy/aluminium alloy system and observed a curved crack front during propagation. This leads to doubt as to the value of crack length to be inserted in the adhesion energy formula. In addition, by using the highly sensitive technique of speckle interferometry, it was possible to study anticlastic bending effects in a quantitative manner. Far from the crack front, agreement between theory and experimental is good, yet work remains to be done to understand the zone near the fracture zone.