In the autoshaping preparation subjects are exposed to magazine training (US-only trials) prior to the conditioning phase in which a stimulus (conditioned stimulus, CS) predicts the delivery of a response independent reinforcer (unconditioned stimulus, US). Two experiments examined the hypothesis that irrespective of the number of US-only trials administered the magazine training and autoshaping contexts interact to determine conditioning, as measured by contact responses to the CS. The contexts employed were houselight on (light, L) and houselight off (dark, D). In Experiment I pigeons were exposed to 1, 20, 100, or 900 US-only trials in a D, or L, context prior to autoshaping in the D, or L. The results indicated that first, autoshaping in the L was superior to autoshaping in the D. Second, irrespective of the autoshaping context performance was better following magazine training in the different context. Third, the function relating performance to the number of US-only trials was an inverted U if magazine training occurred in the D and biphasic if it occurred in the L, irrespective of the autoshaping context. In Experiment II pigeons were exposed to 900 US-only trials in a D, or L, context. Prior to autoshaping in the D, or L, they were exposed to either the magazine training, or a novel, context; this constituted extinction of the US-only context. The results demonstrated that when magazine training and autoshaping occur in the D extinction in the magazine training context results in superior performance relative to extinction in a novel context. However, extinction in a novel context results in better performance, relative to extinction of the magazine training context, if magazine training and autoshaping proceed in the L. In summary, conditioning in the autoshaping paradigm is determined by the magazine training and autoshaping contexts and their interaction. The development of conditioning is therefore dependent on both the associative value of the CS and the background stimuli.