ABSTRACT
This paper reports a study of learning phenomena occurring during an analogy. The study reproduces a situation where an earlier example is used to solve a current one, but abstract knowledge generated over the earlier example is insufficient to solve it. In this situation, the earlier example is re-analyzed and schematic knowledge is generated that allows the solution of the current example. We show that reanalysis is input-driven, i.e. that knowledge contained in the generated schema is of functional nature and at a level of abstraction that reflects the demands of the current problem. We further show that during reanalysis the knowledge about the kind of elements and the structure of the earlier problem is changed in response to the current problem.
