The Evidence on Teaching

Does Discovery-Based Instruction Enhance Learning? A Meta-Analysis

Alfieri, L., Brooks, P. J., Aldrich, N. J., & Tenenbaum, H. R. · 2011

grade Cmeta-analysisindependentreplicatednumbers spot-checked
Sample
164 studies; 580 + 360 comparisons
Population
Mixed ages, mixed domains; lab-leaning; setting not systematically coded.
Design
Two meta-analyses: (1) pure/unassisted discovery vs explicit instruction; (2) enhanced/guided discovery vs other instruction (incl. explicit).
Key findings
The result that dissolves the debate: PURE discovery loses to explicit instruction (d=-0.38), but ENHANCED/GUIDED discovery (scaffolds, elicited self-explanation, worked examples, feedback) BEATS both explicit and pure discovery (d=+0.30; guided subtype 0.50). The live variable is amount/quality of GUIDANCE, not directness vs discovery.
Genetic confound
Domain-learning outcomes; no g. The expertise interaction is prior-knowledge, not ability.
Replication notes
The pivotal 'false dichotomy' result; converges with Furtak 2012, Lazonder 2016.
DOI / URL
10.1037/a0021017

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Pure/unassisted discovery vs explicit instructiond-0.38 [-0.44,-0.31] favoring EXPLICITmixedmostly immediateactive-alternativeend-of-treatmentdomain-skill
Enhanced/GUIDED discovery vs other instruction (incl. explicit)d+0.30 [0.23,0.36] favoring GUIDED discoverymixedmostly immediateactive-alternativeend-of-treatmentdomain-skill
Guided-discovery subtyped0.50mixedmostly immediateactive-alternativeend-of-treatmentdomain-skill

Cited by

Does Discovery-Based Instruction Enhance Learning? A Meta-Analysis · The Evidence on Teaching