Two experiments were conducted to investigate the relationships between individual difference variables and discovery vs. non-discovery instructional treatments in learning of an imaginary science. All subjects were able to learn the material, but presentation of examples and rules (non-discovery) led to more rapid learning. There was no difference between the two treatments on transfer to a higher-order task, although such a difference would have been predicted by the learning-by-discovery hypothesis. However, an interaction between treatment and reasoning ability combined with a floor effect on the number of examples to criterion measure leaves open the possibility that discovery learning may still show advantages, if only for individuals high in reasoning ability. (RH)