Systematic research has demonstrated that many students express essentially the same incorrect ideas at the end of a university physics course as they did at th e beginning. These findings have motivated many physics instructors to implemen t small-group collaboratve work, hands-on activities and other interactive teach ing methods as part of an effort to improve student learning. Often a similar s equence of activities is followed: student difficulties are first elicited, then confronted, and finally resolved. In many cases significant gains in student understanding have been achieved. However, the process of designing research-bas ed instruction is less straightforward than this summary might suggest. Example s will be used to show how students' spontaneous reasoning patterns can suggest effective instructional strategies.