Problem solving is one of the most important goals of any science course. However it is notoriously difficult to improve students’ problem solving abilities, and many students never develop competence. This is particularly true for open-ended or case-based problems – which are also more difficult to assess. We use a number of methods including a suite of software tools and inventories that allow us to assess both student problem solving strategy, student ability, and metacognitive activity as they change over time. Using these tools we can predict how a student will perform on subsequent problems with a 90% probability. Now that we have a set of fairly robust assessment materials, we have begun to develop and investigate intervention methods designed to improve student problem solving strategies and abilities. These methods include collaborative grouping, metacognitive strategies, laboratory projects, and concept maps. The effects of these interventions will be discussed, with regard to student ability, developmental level, and gender.