Context plays an important role in the recognition of objects, allowing the general content of a scene to influence identification of individual parts. An autonomous learning system is presented that examines processes involved in the formation of context between multiple co-occurring objects, under the task of identifying abstract objects in a scene. Learning is performed using a form of Learning Classifier System, that builds representations of features autonomously under reinforcement. The feature identification system is used in combination with an associative network, used for finding co-occurrence relationships for establishing context. Experiments show the influence of the associative network to resolve ambiguous observations through the use of context. This approach involves the interaction of a reinforcement system, analogous to dopaminergic processes, with an associative system, based on associative Hebbian learning processes, and demonstrates the ability of a recurrent associative network for establishing context relationships.