The Evidence on Teaching

The Equivalence of Learning Paths in Early Science Instruction: Direct Instruction and Discovery Learning

Klahr, D., & Nigam, M. · 2004

grade Brctindependentmixednumbers spot-checked
Sample
112 3rd-4th graders
Population
Elementary science (control-of-variables strategy); in-school, researcher-run.
Design
RCT: direct instruction vs pure (unassisted, no-feedback) discovery. Day-1 mastery, ~1-week transfer task (judging others' science-fair posters), blind-scored.
Key findings
Direct instruction produced ~3x more masters of the control-of-variables strategy than pure discovery. Transfer to judging posters depended only on whether a child mastered CVS, not how they learned it — so DI produced more transfer in absolute terms by producing more masters. Note the 'discovery' arm was pure/unassisted (the weakest comparison); this is not a test of guided inquiry.
Genetic confound
Shifted a specific domain skill (CVS) and its near transfer; no g claim.
Replication notes
Boundary per verifier: the poster outcome is MASTERY-GATED NEAR transfer, not far transfer — DI wins by producing more masters, not by out-transferring equally-skilled discoverers. Contested long-term by Dean & Kuhn 2007.
DOI / URL
10.1111/j.0956-7976.2004.00737.x

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
CVS mastery (proximal)% masteringDI 77% vs discovery 23%researcher-designedimmediate; in-schoolactive-alternativeend-of-treatmentdomain-skill
Transfer to judging others' postersddepends on MASTERY not learning path (masters d=0.83 over non-masters; DI-masters = discovery-masters)researcher-designeddelayed ~1wk; blindactive-alternativeend-of-treatmentnear-transfer

Cited by