It is common practice in the analysis of structural frames, to either assume that the joints are pinned or rigid. In fact the real behaviour of a joint is neither pinned nor rigid, and lies somewhere between the pinned and rigid assumption. This is referred to as the semi-rigid behaviour. Semi-rigidity not only refers to the rotational behaviour of the joint as commonly studied, but also in axial and shear actions. The moment distribution between pinned and rigid analysis differs substantially and therefore a more accurate method of modelling the semi-rigid joint is necessary to predict the overall structure response. The level of semi-rigid behaviour varies in different joints due to the material, construction and type of connector. The degree of semi-rigidity can be determined through physical tests. The type of joint for this study is the Metal Plate Connector (MPC) for timber trusses, 6 chosen connector used in residential trusses. An extensive test program was carried out in this study. Four different types of joints of a Queen truss were tested. In addition, the effect of combined loads on the joint characteristics was investigated. The loading arrangement in the tests allowed independent control of the bending moment and axial load. A novel approach is adopted to measure displacement, using high-resolution digital photogrammetry and specially developed software. The data produced gave details of timber movement in cartesian co-ordinates and measurement of plate deformation. From these tests, semi-rigid bending moment and axial stiffness values were determined for use in the theoretical study. An attempt to measure shear stiffness is also presented. Further tests were carried out on full-scale trusses under two different load conditions. The theoretical work comprises two approaches to truss modelling. The first is an automated structural analysis program, which accounts for non-linear semirigid joint characteristics derived from the joint tests using the Foschi power function. The effects of stability and geometrical non-linearity are also implemented into the analysis. The second approach calculates truss response using Finite elements where 2-D planar elements were used to calculate the response of the truss. Parameters for the connection strength are derived from the joint tests. Moment stiffness and axial stiffness values of the connections were determined. Combined load tests showed that there is indeed a measurable effect on joint stiffness and capacity due to combined loads, some of which actually contribute to the stiffness, but also some which are detrimental. There is good correlation between the truss test results and the FE model using semi-rigid joints. However, results of the simpler non-linear frame analysis, did not compare so well, but nevertheless exhibited fundamental characteristics of the truss.