The use of a generative and evaluative computerized prompting framework to improve the teaching of writing skills was studied with 164 children in grades 6, 7, and 8 with high or low writing skills, who were randomly assigned to control or treatment groups. The treatment group received computer prompts, think sheet scaffolds, and expert modeling of writing prompts. Dimensional and holistic assessment of written products, internalization of writing control strategies as measured by the Index of Writing Awareness (IWA) (an instrument developed to measure cognition about writing), open-ended questions, and a prompt sort task failed to show any advantage from treatment. Students' holistic scores were, however, significantly correlated with their awareness of writing skills. Keystroke analysis indicated that the availability of prompts seemed to lower the number of changes students made during the writing process. Differences in holistic performance, IWA performance, metacognitive guidance in writing, surface revisions, and repositioning within papers between grades 6 and 7 were not found between grades 7 and 8, suggesting that a developmental change had occurred between grades 6 and 7, and that procedural assistance may be most appropriate in grade 6. Nine tables present data from the study. A 52-item list of references and 7 appendices, including a proposed model of composition and the study instruments, are included. (SLD)