According to the energy criterion for fracture, crack growth occurs when the crack driving force or the energy release rate is at least equal to the crack resistance which is de®ned as the energy consumed in crack propagation. It is well known that, for a cracked metal structure under the plane-strain loading condition, the crack resistance is independent of the crack length. In the case of plane stress, the crack resistance varies with the crack increment. Although a considerable amount of work has been performed to characterize the resistance curve for isotropic materials [1±3], the application of the resistance curve to composites has only recently been performed. Ratwani and Deo [4] applied the resistance curve approach to characterize the delamination growth of AS4=3501-6 composites under mode I loading. They used double-cantilever-beam (DCB) specimens and obtained the delamination growth resistance curve over a wide range of delamination extensions. They found that the resistance increased monotonically with increasing delamination extension, and the delamination growth resistance curve was path independent. Mai and Hakeem [5] noted a resistance-type fracture behaviour of cellulose ®bre cements. They used grooved DCB specimens and observed that the resistance increased monotonically with increasing crack length to a maximum and then dropped to a steady-state value. Foote et al. [6] developed a theoretical model for the crack growth resistance (KR) curve for a strain-softening material with a power-law stress± displacement relationship. They applied the model to a cellulose±asbestos-®bre-reinforced motar and obtained comparable results with the experimental data. Currently, the delamination resistance curve plays an important role in characterizing the delamination growth for unidirectional laminated composites particularly under mode I loading. However, the delamination resistance behaviour of graphite±epoxy composites is not fully characterized at present. In the present work, delamination resistance curves of unidirectional graphite±epoxy laminated composites were constructed as a function of delamination increment, and the effect of initial delamination length on the resistance curve was investigated. DCB specimens made by using graphite±epoxy prepregs (CU125NS) manufactured by Hankuk Fiber Group were used for the resistance curve tests. The DCB specimen was 08 unidirectional and 24 plies thick (3.8 mm) in which delamination was located at the midplane. Initial delaminations applied were 50 and 90 mm, and the specimen length was 230 mm. The edge of the specimen was painted with a white typewriter correction uid and was marked regularly in order to measure the delamination increments. The delamination length was measured using the travelling microscope. Aluminium blocks were adhesively bonded to the ends of the specimen arms to provide the loading points. The experiments were conducted under room-temperature dry environmental conditions, and a displacement-controlled loading of rate 3.5 mm miny1 was applied. After the initial delamination increased by a certain amount, the applied displacement was reduced and then increased to cause delamination growth. This process was repeated at least ten times until the delamination increased to a length of 170 mm. For each specimen, curves of load, P, against displacement, a, during slow delamination growth were obtained where the delamination length was marked against each loading and unloading line. A typical loading±unloading P±a diagram of initial delamination length 50 mm is shown in Fig. 1. As shown in the ®gure, the fracture behaviour is linear elastic. However, it was found that unloading curves do not return to zero displacement although the amount of residual displacement is very small. The reason is that ®bre bridging occurred behind the delamination tip during delamination growth, which prevents the delamination surfaces from closing up