Two experiments were designed to assess the effects of different order of presented instances on learning of scientific concepts. In Experiment I two kinds of reading materials described as common features of metal were constructed. The text given to Dk group explained those features by referring to instances in order of ‘copper→ calcium’. For Kd group the arrangement of ‘calcium→copper’ was used. As a result, Dk group showed better performance than Kd group in the recall-application test for common features. In Experiment II two kinds of reading materials describing a universal property of livestock were constructed. In the text given to Bk group, the property was explained by mentioning instances in order of ‘swine→silkworm’. For Kb group the arrangement of ‘silkworm→swine’ was chosen. As a result, Bk group showed higher score than Kb group in the recall-application test for a universal property. The obtained results from the two experiments were discussed on the basis of “heteroformulation theory” proposed by Fushimi, Y.(1990, 1991).