The fracture behaviour of an Araldite-F epoxy and its rubber-modified systems was evaluated using compact tension specimens pre-cracked by three methods, namely, razor blade pressing, razor blade tapping and fatigue pre-cracking. The results show that the razor blade tapping method produces a lowest critical stress intensity factor, K-lc, while the razor blade pressing produces an abnormally high K-lc, being about five times higher than the former for the pure epoxy. Transmission polarized optical microscopy reveals that the crack tip produced by razor blade pressing in the pure epoxy specimen was completely surrounded by a plastic deformation zone with compressive residual stress, but the crack tip produced by razor blade tapping was free of residual stress and plastic deformation. It was found that the sensitivity of the fracture toughness value to the pre-cracking methods decreases after the pure epoxy was modified by 10% core-shell rubber or 10% liquid rubber.