The Evidence on Teaching

The Shuffling of Mathematics Problems Improves Learning

Rohrer, D., & Taylor, K. · 2007

grade Crctindependentreplicated
Sample
Small lab-style RCT, college students (~n=24)
Population
College students; artificial math (volumes of unfamiliar solids); lab-like tutorial.
Design
Origin study for math interleaving. Blocked vs mixed practice; unannounced test 1 week later. Proximal aligned content.
Key findings
Origin study: large interleaving benefit (d=1.34) on a 1-week test, with the signature crossover (blocked looks better during practice, mixed wins at test). Small lab sample, artificial content, proximal test.
Genetic confound
Domain-skill outcome.
Replication notes
Very large lab effect (1.34); far above later classroom estimates (0.79-0.83) — illustrates lab-to-classroom shrinkage.
DOI / URL
10.1007/s11251-007-9015-8

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Interleaved vs blocked, 1-week testd1.34 (~63% vs ~20% accuracy per secondary reports)researcher-designeddelayed 1wk; lab-likeactive-alternativeend-of-treatmentdomain-skill
Practice-phase crossoverdirectionblocked higher during practice, mixed higher at delayed test (metacognitive illusion)researcher-designedimmediate vs delayedactive-alternativeuncleardomain-skill

Cited by

The Shuffling of Mathematics Problems Improves Learning · The Evidence on Teaching