A study examined whether writing modeled from children's picture books would improve reading comprehension of fourth and fifth graders as much as traditional skills instruction. Subjects, 69 children reading at least one year below grade level from six Chapter 1 Colorado schools, were pretested for reading comprehension levels. Subjects were then randomly assigned to one of three ten-week treatment groups using one of the following methods: (1) writing (without word processors) modeled after children's picture books that were read aloud by the Chapter 1 teacher and followed by oral language prewriting activities; (2) writing modeled after children's picture books that were read aloud and followed by oral language activities and revision and rewriting on microcomputer word processors; and (3) using a sequenced skills series instructional program, "Pegasus Pace," in current use by Chapter 1 teachers. Comparison of posttest and pretest scores indicated that although writing modeled after children's picture books resulted in significant comprehension gains compared to the skills control group, the addition of word processors to the writing method resulted in a significant loss of comprehension for that treatment group. Supporting the notion that writing improves reading ability, the results suggest that teachers need to provide children with many regular writing activities, while limiting the use of commercial programs and worksheets. (Tables of data and a sample lesson plan from the treatment groups are included.) (HTH)