Previous research has indicated that the way of learning and the sequence of study influence how we learn and represent categories. However, most studies have focused on classification learning and it has been rarely studied how learning sequence influences inference learning. The current study attempted to address this issue. Participants learned four categories by classification or inference in both blocked and interleaved sequence. Then participants completed a transfer task and a feature prediction task. Results showed that classification learners encoded characteristic features and formed similarity-based representations in the blocked study, whereas in the interleaved study, they encoded deterministic features and formed rule-based representations. In contrast, for inference learners, the blocked and interleaved study changed their learning and representation in the same direction. In both sequences, inference learners encoded deterministic features and formed rule-based representations. These results suggest that different mechanisms are likely to be involved for inference and classification learning.