In timber structures, beams which are loaded perpendicular to grain by connections along the span may fail by splitting, resulting in the collapse of the beam. In the past, empirical, semi-empirical and fracture mechanical models have been developed aiming at predicting the splitting failure load. Recent experiments with two and three connections show the empirical models to be very non-conservative in predict- ing the effect of multiple connections. Fracture mechanical models that are better equipped to estimate the failure load but the result dependents on the assumed crack growth direction. Test results with three connections cannot be explained by any of the models yet. Some structural design codes provisions are unsuitable and can result in very non-conservative predictions while othes result in a conservative estimation of the splitting failure load.