The Evidence on Teaching

A Randomized Controlled Trial of Interleaved Mathematics Practice

Rohrer, D., Dedrick, R. F., Hartwig, M. K., & Cheung, C.-N. · 2020

grade Arctindependentreplicatednumbers spot-checked
Sample
787 7th-graders, 54 classes, multiple schools
Population
7th-grade, multiple real public schools, authentic algebra curriculum, own untrained teachers.
Design
PREREGISTERED cluster-randomized trial. ~4-month practice, unannounced test ~1 month later. Test used a BLOCKED format (biased against interleaving). Meets WWC without reservations; replicates Rohrer 2015.
Key findings
Preregistered cluster RCT, large diverse sample, untrained teachers, delayed test, meets WWC without reservations, replicates Rohrer 2015: large effect (d=0.83), positive for all 15 teachers. Key caveats: the test is proximal researcher-designed (same problem kinds; standardized equivalent likely ~0.4), and the effect confounds discrimination + embedded SPACING + retrieval (the per-kind gradient tracks spacing dose).
Genetic confound
Domain-skill outcome; no g. Larger for novices.
Replication notes
Numbers verified. The strongest evidence in the interleaving literature; but the outcome is a proximal researcher-designed test (standardized equivalent likely ~0.4).
DOI / URL
10.1037/edu0000367

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Interleaved vs blocked, unannounced test ~1 month laterd0.83 [0.68,0.97]; 60.7% vs 37.6%researcher-designeddelayed 1mo+; classroomactive-alternativeunder-1yrdomain-skill
Consistency across 15 teachersd rangeall positive, 0.23 to 1.48researcher-designeddelayed 1mo+; classroomactive-alternativeunder-1yrdomain-skill
Per-problem-kind (tracks spacing dose)dA 0.86 / B 0.63 / C 0.40 / D 0.34researcher-designeddelayed 1mo+; classroomactive-alternativeunder-1yrdomain-skill

Cited by