A study examined the effect of topic-specific background knowledge on the general quality and local coherence of student writing. Ninety-seven tenth grade students were assigned two writing tasks on specific topics at two points during a semester. J. Langer's measure of topic-specific knowledge was administered prior to each writing task. Each student's writing sample was then scored on five separate measures: (1) overall quality, (2) coherence, (3) syntactic complexity, (4) audience, and (5) function. The data from the pretest and writing evaluations clearly suggested a strong and consistent relationship between topic-specific background knowledge and the quality of student writing. More interesting, however, was the evidence that different kinds of knowledge were predictive of success in different writing tasks. The analysis of interactions between background knowledge and audience and function categories suggested that when students had only fragmentary knowledge about a topic, they used tactics to avoid engaging in writing activities that required them to say more than they knew. The findings suggest that for many tasks, writers may need the same kinds of highly organized information that readers do, but for others they may also require an abundance of loosely related informational "bits" for use in their writing. This distinction may arise from differences inherent in each of the two activities: writers not only need to generate, present, and relate intricately linked ideas, but also need a large body of loosely associated information that can be used to elaborate, embellish, and enliven the presentation. (HTH)