The fracture of blends of nylon 6 and maleated ethylene-propylene rubber was examined by both the Izod impact test and a single-edge notch three-point bend (SEN3PB) instrumented Dynatup test. The effects of EPR-g-MA content, ligament length, method of fracture surface measurement, sample thickness and fracture position in the molded bar on the fracture behavior were investigated. The data were analyzed by plotting the specific fracture energy (U/A) as a function of ligament length. The blends containing a high portion of EPR-g-MA in the rubber phase were found to be super tough over the whole range of ligament lengths and under all test conditions. However, a ductile-to-brittle transition was observed with ligament length for marginally tough blends which contained a low content of EPR-g-MA in the rubber phase and had a ductile-brittle temperature near or above room temperature; the specimens with short ligament length fractured in a ductile manner, while the specimens with long ligaments showed brittle fracture. The transition ligament lengths were found to be dependent on the rubber particle size. The dual mode of fracture was rationalized by equations for ductile yielding and brittle crack propagation; values of yield stress and critical intensity factor were estimated from these model equations. The dissipative energy density, u(d), was more sensitive to rubber particle size, sample thickness and location in the molded bar than the limiting specific fracture energy, u(o). There is a good correlation between the standard Dynatup impact strength and the parameter lid for the gate end specimens.