The Evidence on Teaching

A Meta-Analysis of Writing Instruction for Students in the Elementary Grades.

Graham S, McKeown D, Kiuhara S, Harris KR · 2012

grade Cmeta-analysisindependentreplicated
Sample
115 documents containing the statistics needed to compute an effect size; 13 treatments each tested in 4+ studies
Population
Elementary-grade students (Journal of Educational Psychology 104(4), 879-896)
Design
Companion meta-analysis to Graham & Perin (2007), applying the same procedures to the elementary grades. Same measure-type caveat: the outcomes are overwhelmingly researcher-scored writing quality and production measures, so METHODOLOGY's ~2x inflation for treatment-aligned measures applies. A published correction exists (Graham et al., 2012, Journal of Educational Psychology, doi 10.1037/a0029939).
Key findings
Average weighted effect sizes: strategy instruction 1.02; creativity/imagery instruction 0.70; text structure instruction 0.59; adding self-regulation to strategy instruction 0.50; transcription skills (handwriting, spelling, typing) 0.55; prewriting activities 0.54; peer assistance 0.89; setting product goals 0.76; assessing writing 0.42; word processing 0.47; extra writing 0.30; comprehensive writing programmes 0.42. GRAMMAR INSTRUCTION WAS THE ONLY ONE OF THE 13 TREATMENTS THAT DID NOT PRODUCE A STATISTICALLY SIGNIFICANT EFFECT. That replicates the pattern in the adolescent meta-analysis - grammar is the single dissenting entry in an otherwise uniformly positive table - at a younger age band, and it does so without the negative sign, so the elementary reading is "nothing" rather than "harm".
Genetic confound
Low for direction (experimental and quasi-experimental comparisons), medium for magnitude - the quasi-experimental fraction admits selection into treatment classrooms, and quality is scored by raters who are usually not blind to condition.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Writing quality, grammar instruction, elementary gradesaverage weighted ESnot statistically significant - the only one of 13 treatments to failresearcher-designedpost-treatmentunclearend-of-treatmentdomain-skill
Writing quality, strategy instruction, elementary gradesaverage weighted ES1.02researcher-designedpost-treatmentunclearend-of-treatmentdomain-skill
Writing quality, adding self-regulation to strategy instructionaverage weighted ES0.5researcher-designedpost-treatmentunclearend-of-treatmentdomain-skill
Writing quality, transcription skills instruction (handwriting/spelling/typing)average weighted ES0.55researcher-designedpost-treatmentunclearend-of-treatmentdomain-skill
Writing quality, extra writingaverage weighted ES0.3researcher-designedpost-treatmentunclearend-of-treatmentdomain-skill

Cited by