An experiment to determine the effects of conclusive evidence on weakly-based beliefs was presented. The subjects explained a hypothetical relationship between two variables, and it was found that the subjects' opinions were systematically influenced as a result of their explanations. The subjects' judgments of conclusive evidence were not influenced by their earlier explanations, however. When the subjects' beliefs were measured after viewing the evidence, the following results emerged. The subjects' stated opinions were not affected by their earlier explanations, but were affected by the evidence that the subjects had viewed. However, when the subjects had to predict the outcome of an experiment designed to discover the true relationship between the variables of interest, there was a clear effect of the explanation manipulation and no effect of the actual evidence. In addition, a policy-capturing measure revealed a relationship between theory biases and the subjects' use of information.