Anaphora is a common phenomenon in discourses and a significant area for research in the use of natural language processing. In this paper we have concentrated on context within the discourse and trying to resolve anaphora with the rule-based approach. In applications of NLP like machine translation, summarization, question-answering systems, sentiment analysis, and others, anaphora resolution is essential. In this article, we make the case that many common adverbial phrases that are typically thought to indicate a discourse relation between syntactically related units within discourse structure really add relational meaning anaphorically. So many works have been done on anaphora resolution for simple sentences but very less amount of work has been done on complex English discourses. We have developed manual dataset of 100 complex English discourses to resolution of an anaphora from discourses with appropriate rules and algorithm which we have developed on the basis of grammar. The system achieved 94% accuracy.