This study examined how the students' expectation of context sharing with audience causes effects on their writing arguments. Fourth-grade students (N=30) from elementary school in Japan were assigned to do two writing argument tasks. In each task, students made an argument and persuaded different audiences (transfer student and old friend) at random. The contents of written arguments were categorized into "claims", "evidence", and "reasoning". Moreover, difficulty of each audience's persuasion and its reasons were asked in questionnaire. The analysis suggested followings. (1) Students generated more "reasoning" in transfer student condition than old friend. (2) Students evaluated persuasion of transfer student was more difficult than old friend. Students judged that the transfer students do not share much context with them, so they tried to persuade by giving more information. The students read the degree of context sharing and changed the contents of writing with changing audiences.