In this paper, an uncertain reasoning approach is put forward based on the cognitive structure: a quadruple cognitive structure is adopted to represent the uncertain knowledge, and a two way cognitive inferential structure is used to handle the uncertainty of the inferential rules. The approach presnted in this paper can satisfy the exact probability reasoning; contain the uncertainty of the information, effectively avoid the unknown for the relationship among the rules, and the computing complexity of its briefest inference is linear to the number of inferential nodes.