We report three experiments, which investigate the role of hypotheses for metalogical puzzles, in which some persons are always telling the truth and other persons are always lying. In Experiment 1, the subjects received an accurate or an inaccurate single hypothesis or no hypothesis. Accurate hypotheses did lead to an increase of correct conclusions, while inaccurate hypotheses did lead to a decrease of correct conclusions. This is a failure to replicate the results of Byrne and Handley (1995): In their experiment with a separate control-group, inaccurate hypotheses did not lead to a decrease of correct conclusions. In Experiment 2, subjects received reformulated single hypotheses, in order to avoid unconditional acceptance of the given hypotheses by the subjects. However, the same results were obtained as in Experiment 1. Therefore, subjects seem to try to verify the given hypotheses. In Experiment 3, subjects were given double hypotheses, in order to avoid verification. As a consequence, there was no difference between problems with an accurate, an inaccurate and no hypothesis. The results are discussed in terms of strategies in reasoning.