Timber joints can experience ductile failures as modeled by Johansen in the European Yield Model adapted in Eurocode 5, or they can fail in a brittle manner. In nailed or screwed joints where the fastener does not protrude through the timber, plug shear failure can occur where an entire plug defined by the perimeter of the joint is torn away from the timber. The brittle plug shear failure, which can occur in joints loaded in tension parallel to the grain, results in a lower joint resistance than the ductile failure modes. The aim of this study is to evaluate existing prediction formulas for plug shear failure in timber connections, compare them to test results and observations, and propose a new prediction formula. Test results from four different experimental studies are presented. Using hypothesis testing, a prediction formula for plug shear failure is proposed based on 92 experiments. The resistance of the tensile failure mode of plug shear failure is best modelled by the tensile resistance of the end face of the plug. The resistance of the shear failure mode of plug shear failure is best modelled by the shear resistance of the bottom area of the plug taking into account the volume effect on shear strength. The model currently in Annex A of Eurocode 5 is overestimating the plug shear resistance compared to the test results analysed in this research. To avoid plug shear failure, short and wide joints are preferred, minimising the number of fasteners in line with the load and grain direction