A fuzzy stratifying model was suggested by the way of fuzzy relation. It contains a score function to measure the hierarchy sense for giving responses. The extension concept from discrete to fuzzy sense was clearly investigated on the construction of this model. In applications,the CSMS team introduced the concept of hierarchy or level of understandings. A method was sought by which test items could be grouped to form a 'type' and then the groups ordered to form a chain in such a way that success on a harder group automatically entailed success on all easier groups. Items had to satisfy specific criteria before they were considered to form a group. In this paper, we developed a method to describe the association in fuzzy sense. We also found some fuzzy ranking ways. Finally we added some constraints on it and solved the problem of hierarchy concepts. Some related properties were proved also.