A new model to predict the failure load of adhesively bonded joints is proposed. An extensive finite element analysis and experimental study of single and double lap joints with aluminium and composite adherends is first undertaken to validate the model. Emphasis is then placed on understanding the behaviour and predicting the failure load ofbondedjoints with thin postbuckling composite skins. The design of structural adhesively bonded joints is complicated by the presence of singularities at the ends of the joint and the lack of suitable failure criteria. Literature reviews indicate that bonded joint failure typically occurs after a damage zone at the end of the joint reaches a critical size. For bonded joints suffering cohesive failure the damage zones form in the adhesive. For bonded composite joints suffering inter/intralaminar failure the damage zones form in the adherends. Based on these findings a damage zone model using a critical damage zone size and predominantly strain based failure criteria is proposed to predict the failure load of adhesively bonded joints. The damage zone model correctly predicts the joint failure locus and appears, up to a point, to be insensitive to finite element analysis mesh refinement. Results from experimental testing of various graphite/epoxy composite and aluminum lap joints are obtained and compared with finite element analysis. Good correlation of experimental and numerically predicted joint failure loads are achieved. While the damage zone model can be classed as a scientific approach which explains the physical process of failure, a modified damage zone model is also developed which allows the model to be implemented in a practical engineering analysis environment. A number of failure criteria are evaluated in conjunction with the modified damage zone model. Using the knowledge and experience gained from the research into the analysis of adhesively bonded joints, failure of skin-to-rib/stiffener joints in postbuckling structures is then investigated both experimentally and numerically. Good correlation of numerically predicted joint behaviour and experimental test results is obtained for both direct compressive buckling and shear buckling skins. This correlation shows that the postbuckling behaviour of stiffened panels can be accurately modelled using the finite element method. Bonded joint failure predictions for these joints are made with the damage zone model to validate it for more complex structures. The results from this investigation show that bonded joints that form hard points in postbuckling structures can induce premature failure. It is concluded that the damage zone model can be successfully applied across a broad range of aluminium and composite joint configurations and loading conditions. To implement the damage zone model for this broad range of joints, an adhesive peel strain failure criterion is shown to be consistently accurate. Because this criterion is applied to the adhesive, detailed modelling of composite plies is not required and therefore it can be implemented relatively simply using either finite element analysis or basic onedimensional shear lag algorithms.