In two studies, the process of insight as used in problem solving and the training of insight were examined. In Experiment 1, subjects either practiced solving insight problems or studied insight problems before attempting to solve an insight problem where they were able to ask limited questions of the experimenter. In Experiment 2, subjects were either trained on identifying common assumptions or given practice solving insight problems. Solution rates and verbal protocols for four test problems were collected. Neither study provided evidence of the re-organization of subjects' problem knowledge (as modeled by the Pathfinder algorithm) during problem solution. Though type of training did influence the rate and types of questions asked when solving the test problems, it did not facilitate the actual solution of those problems. The definition of insight was discussed as the identification of relevant and irrelevant problem features rather than as a change in problem representation.