Forty fifth and sixth graders, matched on sex and measures of test anxiety, defensiveness, and IQ, were divided into two groups, each of which solved Porteus maze tasks and a marble puzzle, with and without memory support, respectively. An anxiety-by-memory support interaction occurred in the number of errors made prior to solving the marble puzzle, as predicted, but did not occur in the Porteus maze task. It was suggested that anxiety's interference with short-term memory could be offset by a variety of external aids, such as diagrams or notational systems, which problem solvers could be taught to use. (Author)