The adhesive fracture energy, G(c) , of metallic joints, bonded with a rubber-toughened epoxy adhesive, has been measured using monotonically-loaded tests. Such tests have been conducted in various relative humidities and in water, at 21 degrees C. Two surface pretreatments have been employed for the substrates prior to bonding: a simple grit-blast and degrease ('GBD') pretreatment or a grit-blast, degrease and silane primer ('GBS') pretreatment. The joints were formed using metallic substrates which were either (a) aluminium-alloy substrates, (b) steel substrates, or (c) 'dissimilar' substrates (i.e. one substrate being aluminium-alloy with the other one being steel). For both test environments, when G(c) was plotted against the crack velocity, three regions of fracture behaviour could be distinguished. At low rates of displacement the crack grew in a stable manner, visually along the interface, and relatively low crack velocities could be readily measured. This was termed 'Region I' and here the value of G(c) measured in the aqueous environment was relatively low compared to that measured in a relatively dry environment of 55% relative humidity. On the other hand, at relatively high rates of displacement the crack always grew in a stick-slip manner mainly cohesively in the adhesive layer at approximately 20 km/min. This was termed 'Region III', and here the value of G(c) was relatively high and independent of the environmental test conditions employed. In this region the crack was considered to grow faster than the water molecules were able to reach the crack tip, which explains the independence of G(c) upon the test environment. In between 'Region I' and 'Region III', a transition region was observed which was designated as 'Region II'. The major effect of the 'GBS' pretreatment, compared to the 'GBD' pretreatment, was to increase the value of G(c) both in 'Regions I and III', although the presence of the silane primer had the far greater effect in 'Region I'.