The problem-solving performance of groups of high-achieving (HA) and lowachieving (LA) Year 11 students was compared during solution of geometry problems using a think-aloud procedure. Students also undertook knowledge assessment tasks involving theorem recall and a prompted recall task to provide estimates of knowledge available for access during the solution attempts. Detailed analysis of problem-solving protocols indicated that HA students not only accessed a greater body of geometrical knowledge but also used that knowledge more effectively. On the more difficult problems, these HA students generated more information, used that generated information to access further relevant knowledge, and showed more frequent management of their processing behavior. The nature of the difference in generative activity of a student from each of the two groups is illustrated.