The Effectiveness of Direct Instruction Curricula: A Meta-Analysis of a Half Century of Research
Stockard, J., Wood, T. W., Coughlin, C., & Rasplica Khoury, C. · 2018
grade Cmeta-analysisdeveloper-ledmixednumbers spot-checked
Sample
328 studies / 413 designs / 3,999 effects
Population
Preschool-college (majority grade 3 and under), high-poverty over-represented; reading/math/language/spelling.
Design
Largest Direct Instruction meta. ~20% of designs used random selection OR matching; the rest quasi-experiment/single-subject. ~75% of effects on normed/published tests, ~2/3 teacher-delivered in real classrooms. COI: authors are NIFDI (National Institute for Direct Instruction) staff — flag symmetrically with Adams & Engelmann.
Key findings
Strongest quantitative case for explicit/DI: all subjects ~d 0.5, larger with more exposure, maintained over <6-month follow-ups. Unusually credible on measurement (~75% normed tests, real classrooms). But mostly QUASI-experimental (~20% randomized-or-matched) and NIFDI-authored, so discount for allegiance; the 0.83 single-subject figure is design-inflated.
Genetic confound
The ability-measure d=0.34 is proximal-test overlap with taught content, not durable g. DI raises domain achievement, not intelligence.
Replication notes
Numbers verified against full text. ~0.3 SD below the proponent Adams & Engelmann 1996 (0.87) — consistent with allegiance inflation. Mostly quasi-experimental; NIFDI COI.
DOI / URL
10.3102/0034654317751919
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Total academic achievement | d | 0.54 [0.49,0.59] | mixed | mostly immediate; classroom | business-as-usual | end-of-treatment | domain-skill |
| By subject | d | reading 0.51 / math 0.55 / language 0.54 / spelling 0.66 | mixed | mostly immediate | business-as-usual | end-of-treatment | domain-skill |
| Ability/IQ-type measures | d | 0.34 (proximal/crystallized overlap, NOT g movement) | standardized | immediate | business-as-usual | end-of-treatment | domain-skill |
| Single-subject designs (design-inflated, analyzed separately) | d | 0.83 | researcher-designed | immediate | none | end-of-treatment | domain-skill |
Cited by
- Explicit instruction vs discovery/inquiry — how much guidance?moderate supportconf: mediumgc: low
- Explicit vs reform/constructivist math instructionmoderate supportconf: highgc: low