A study investigated the effects of instruction in metadiscourse on composition students' writing skills. Subjects were students in two 100-level college composition classes. A Control Class (CC) was taught using a process approach, and the Experimental Class (EC) had direct teaching of metadiscourse. The CC students worked on the propositional content of their essays while the EC students concentrated on the pragmatic functions of metadiscourse. Posttests written by EC students were significantly better than those of the CC, although pretest results did not differ. Sources of this difference are discussed, including changes in metadiscourse markers, tone of the essays, and topical progression. It is argued that the EC students not only used metadiscourse markers more effectively, but also wrote with more attention to audience needs, thereby making global changes that improved their papers. Contains 13 references. (Author/MSE)