For an intelligent agent to be truly autonomous, it must be able to adapt its representation to the requirements of its task as it interacts with the world. Most current approaches to on-line feature extraction are ad hoc; in contrast, this paper presents an algorithm that bases judgments of state compatibility and state-space abstraction on principled criteria derived from the psychological principle of cognitive economy. The algorithm incorporates an active form of Q-learning, and partitions continuous state-spaces by merging and splitting Voronoi regions. The experiments illustrate a new methodology for testing and comparing representations by means of learning curves. Results from the puck-on-a-hill task demonstrate the algorithm's ability to learn effective representations, superior to those produced by some other, well-known, methods.