A Quantitative Synthesis of Research on Writing Approaches in Grades 2 to 12. Best Evidence Encyclopedia (BEE)

14 studies of 12 programmes met the inclusion standards; 12 randomised, 2 matched

Students in Grades 2-12 (roughly ages 7-18)

The single most important methodological contrast in this subject. Where the Graham- family meta-analyses admit any experimental or quasi-experimental study with a reliably scored writing-quality measure, the Best Evidence Encyclopedia standards require randomised or well-matched control groups, ADEQUATE SAMPLE SIZE AND DURATION, and - decisively - "measures independent of the program developers, researchers, and teachers". Applying those filters to the same literature leaves 14 studies of 12 programmes. The resulting effect sizes are therefore the closest thing this subject has to an independent-measure estimate, and they are roughly a quarter to a fifth the size of the researcher-scored headline numbers.

Average effect on writing achievement across all categories ES = +0.18. By category: programmes focused on the writing process +0.17; programmes using cooperative learning +0.16; programmes focused on interactions between reading and writing +0.19. All three categories are positive, consistent, and small. Set this against Graham & Perin's +0.82 for strategy instruction, Graham et al.'s +1.02 in the elementary grades, and Koster et al.'s +0.96, all on researcher-scored quality measures. The size of the gap is larger than METHODOLOGY's nominal 2x measure-type inflation, which is itself the finding: the writing literature is unusually dependent on treatment-aligned scoring. The authors also note how thin the qualifying evidence base is - fourteen studies for an entire school subject across eleven year groups.

Low. Twelve of the fourteen qualifying studies are randomised; the two matched designs are the residual exposure.