Anaphora has been the subject of a number of very interesting and productive accounts. The comprehension and interpretation of anaphora is extremely important for Nature Language Processing. Indirect anaphora, which permeates verbal communication, refers to the relationship between an anaphoric expression and its intended referent which is not explicitly expressed in prior context, but can be inferred via the specific information provided in the prior linguistic context, without coreference with another expression. As a semantic concept, indirect anaphora has been researched in the literature and significant insights have been gained through various approaches. However, not a single model is capable of handling all the cases of indirect anaphora interpretation due to the fact that each model approaches the problem either from a different perspective or with a different goal. The frame theory really plays an important role to discourse understanding, but it processes from the static perspective. Humans cognitive process of indirect anaphora resolution, which is a vital part of discourse analysis, is concerned with dynamic work. Based on the previous studies on indirect anaphora and especially the recent cognitive research in anaphora study, the present paper, reviewing the different models of anaphoric resolution and trying to provide a new perspective for indirect anaphora research, establishes an integrated cognitive solution to indirect anaphora within conceptual integration network model and frame theory, which reveals great significance of human cognition in discourse analysis. Key words: indirect anaphora, cognitive linguistics, discourse analysis, frames, conceptual blending