The Evidence on Teaching

When Problem Solving Followed by Instruction Works: Evidence for Productive Failure

Sinha, T., & Kapur, M. · 2021

grade Cmeta-analysisdeveloper-involvedreplicated
Sample
53 studies / 166 comparisons (largely RCTs)
Population
Mostly secondary and undergraduate; math/science conceptual learning.
Design
Meta of Productive Failure / Problem-Solving-then-Instruction (PS-I) vs Instruction-then-Problem-Solving (I-PS). The missing counter-body in the explicit-instruction debate.
Key findings
Problem-solving BEFORE instruction (productive failure) beats instruction-first on conceptual understanding and transfer (g=0.36, up to 0.58) without harming procedural knowledge — and OLDER learners benefit more from exploring first. This refutes the strong 'explicit is always the default' claim: for conceptual/transfer goals with older learners, exploration-first can win.
Genetic confound
Domain conceptual/transfer outcomes; no g. Interaction is with age/prior knowledge.
Replication notes
Converges with Schwartz invention/preparation-for-future-learning studies; the key evidence that a well-designed exploration-FIRST sequence can beat instruction-first.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
PS-I (explore first) vs instruction-firstg0.36 [0.20,0.51] on conceptual understanding + transfer; up to 0.58 at high PF fidelitymixedmostly immediate/shortactive-alternativeend-of-treatmentdomain-skill
Age moderationdirectionOLDER learners (gr 6-college) benefit MORE from exploration-first; younger (gr 2-5) and domain-general skills favor instruction-firstmixedmixedactive-alternativeuncleardomain-skill

Cited by