This chapter aims to attempt model the process by which humans improve their ability to solve problems using diagrams. It provides an established model for human perceptual memory, and extended it with an ability to learn and manipulate plans for drawing schematic diagrams. This model is applied to the learning of planning information in a specific problem-solving domain: the construction of a diagrammatic representation for electric circuits. When confronted with novel problems, the new model will attempt to match the problem with a previous entry in its memory. If such a match exists, any equivalence link will suggest a possible solution to the current problem based on what was learnt earlier. Perceptual similarities ensure that each chunk may be generalised to a range of future examples, showing how our novel version of CHREST learns to improve its problem-solving ability.