The Evidence on Teaching

The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory

Kluger, A. N., & DeNisi, A. · 1996

grade Cmeta-analysisindependentreplicatednumbers spot-checked
Sample
131 papers / 607 effect sizes / 23,663 observations
Population
Mostly lab/adult performance tasks; limited K-12.
Design
Foundational feedback meta with double-coded moderators. Mostly immediate researcher-designed task performance; no standardized-achievement or long-term data.
Key findings
The field's foundational corrective: feedback helps on average (d=0.41) but MORE THAN A THIRD of effects are negative — the only major method with a stable backfire rate. The 30-year-stable moderator structure: cues carrying TASK information help (correct solution, progress-over-time); cues directing attention to the SELF hurt (praise 0.09, discouragement -0.14, ego threat 0.08). More frequent feedback showed no benefit.
Genetic confound
Experimental within-task designs; the self-esteem-threat moderation implies interaction with stable person traits.
Replication notes
Headline 0.41 and >38%-negative verified from full text; moderator sub-tables consistent with widely-cited values but not line-verified. Moderator structure replicated by Wisniewski 2020.
DOI / URL
10.1037/0033-2909.119.2.254

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Overall performance from feedback interventionsd0.41; >38% of 607 effects NEGATIVEresearcher-designedimmediatebusiness-as-usualend-of-treatmentdomain-skill
Correct-solution information present vs absentd0.43 vs 0.25researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill
Change-from-prior-trials (velocity) infod0.55 vs 0.28researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill
Praise cues present vs absentd0.09 vs 0.34researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill
Discouraging feedbackd-0.14 vs 0.33researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill
High threat to self-esteem (top vs bottom quartile)d0.08 vs 0.47researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill
Very complex tasks (top vs bottom quartile)d0.03 vs 0.55researcher-designedimmediateactive-alternativeend-of-treatmentdomain-skill

Cited by