The Evidence on Teaching

The Environment in Raising Early Intelligence: A Meta-Analysis of the Fadeout Effect

Protzko, J. · 2015

grade Bmeta-analysisindependentunreplicatednumbers spot-checked
Sample
39 RCTs, 7,584 participants
Population
Children in RCTs that successfully raised IQ before kindergarten, then tracked IQ afterwards.
Design
RCT-ONLY growth-curve mixed-effects meta-analysis. Its key contribution is directly testing two rival mechanisms: reciprocal (controls catch up) versus unidirectional (the treatment group loses its gain).
Key findings
The most mechanistically precise fadeout study. IQ fadeout happens because the TREATED group loses its gain (b=-1.09), not because controls catch up (b=-0.21, ns). A striking illustrative case: LCPUFA supplementation raised IQ 0.44 SD at age ~4 with no environmental change at all, and had fully faded by age 7 — so fadeout is not merely 'returning to a disadvantaged environment'.
Counterfactual
Randomised control groups across the included trials.
Genetic confound
RCT base. Explicitly rules out genetic set-point, practice effects, lost knowledge, selective attrition and publication bias as sole explanations.
Replication notes
The mechanism it identifies is IQ-SPECIFIC. For achievement outcomes the mechanism is often the opposite — control-group catch-up (Bitler/Hoynes/Domina; the Building Blocks case in Bailey 2017). Record fadeout mechanism by outcome class, not globally.
DOI / URL
10.1016/j.intell.2015.10.006

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
IQ trajectory, CONTROL group after treatment endsb-0.206 [-1.149, 0.737] — stable, nsstandardizedgrade-3noneover-2yrg
IQ trajectory, EXPERIMENTAL group after treatment endsb-1.09 [-2.149, -0.026] — significant DECLINEstandardizedgrade-3noneover-2yrg
Intervention duration as a moderatormoderatorno significant effect on magnitude or decay ratestandardizedgrade-3noneover-2yrg
Initial effect size vs decay rater-0.298 (ns) — larger gains do not fade fasterstandardizedgrade-3noneover-2yrg

Cited by