Mixed-mode bending tests were performed on fifty-three GFRP-adhesive-GFRP specimens. In each specimen the pre-crack was located near the GFRP adherend with fibres normal to the GFRP-adhesive interface, so as to encourage crack propagation along this interface rather than through the GFRP. The other GFRP adherend contained fibres parallel to the interface, leading to asymmetry of GFRP material (but not of specimen geometry) with respect to the adhesive layer in each specimen. This switching, across the adhesive layer, between normal and parallel fibres in the adherends, reflects a T-joint detail considered for a wave energy device. All tests used a video gauge system to measure crack propagation. Some tests also employed an acoustic emission (AE) system. Comparisons between the AE results and the video gauge-based crack propagation data strongly suggest that the AE energy vs time curve is a much clearer indication of crack initiation than is the widely used AE hit count vs time curve. Increase in Mode II ratio during the tests led to fewer but larger post-peak load drops. Williams's approach, based on beam theory for asymmetric sections with correction factors for large displacements and shear deformation included, led to fracture toughness (G(c)) values in the range 0.3-2.8 kJ/m(2), with a trend of increasing Gc values as pure Mode II is approached. The ASTM approach, developed only for symmetric specimens, gave much lower Gc values. (C) 2017 Elsevier Ltd. All rights reserved.