Our project concerns the development of a system which integrates situation driven execution with constructivist learning. We begin with reactive planning as embodied in the RAP planning and execution architecture [Firby, 1989]. We describe how a system originally designed for unstructured constructivist learning, Drescher’s Schema learning mechanism [Drescher, 1991], can be modified to support similar goal directed reactive behavior. In particular, we have adapted the system to pursue explicitly defined goals and developed a set of macros for specifying composite schema and actions. Using the Truckworld simulation system [Firby, 1989] as a testbed we were able to verify that the modified Schema mechanism is able to perform much like the RAP system. Further, the mechanism learned information relevant to the systems explicit goals.