Phonemic Awareness Instruction Helps Children Learn to Read: Evidence From the NRP's Meta-Analysis
Ehri, L. C., Nunes, S. R., Willows, D. M., et al. · 2001
grade Cmeta-analysisindependentunreplicated
Sample
52 studies / 96 comparisons
Population
Preschool through grade 6; normal, at-risk, and reading-disabled readers.
Design
Mix of RCT and QED; heavy researcher-designed proximal measures. Reports end-of-treatment and delayed follow-up.
Key findings
PA training moves the trained skill hugely (d=0.86) but reading transfer is moderate (0.53) and comprehension weakest (~0.34). The proximal PA gain vastly exceeds the distal reading gain, and the NRP's own moderator analysis already pointed to 'PA needs letters.'
Genetic confound
Experimental base; proximal researcher measures inflate the PA-skill effect ~2x.
Replication notes
Canonical PA meta; NRP concluded PA is most effective taught WITH letters, focused on 1-2 skills, in small groups.
DOI / URL
10.1598/RRQ.36.3.2
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Phonemic awareness (the trained skill) | d | 0.86 (large) | researcher-designed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Reading | d | 0.53 | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Spelling | d | 0.59 (but ns for disabled readers) | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Reading comprehension | d | ~0.34 | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Moderator: PA taught WITH letters vs without | d | with letters larger; 1-2 skills > many; small group best | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
Cited by
- Phonemic awareness training (and whether it needs letters)moderate supportconf: mediumgc: low