The Evidence on Teaching

Does Self-Control Training Improve Self-Control? A Meta-Analysis

Friese M, Frankenbach J, Job V, Loschelder DD · 2017

grade Bmeta-analysisindependentmixed
Sample
33 studies yielding 158 effect sizes, published and unpublished, mostly adult and laboratory samples
Population
Predominantly adults; self-control training experiments.
Design
The meta-analysis that applies the corrections this literature usually avoids - trim-and-fill and PET-PEESE - alongside moderator tests for control-group type and author allegiance.
Key findings
The uncorrected effect is g = 0.30 (95% CI 0.17 to 0.42). Bias-corrected it falls to a range of 0.13 to 0.24. Effects were larger for stamina than strength paradigms, larger against INACTIVE than active control groups, larger for males, and - the finding this archive weights - larger WHEN PROPONENTS OF THE STRENGTH MODEL OF SELF-CONTROL WERE AUTHORS OR CO-AUTHORS, an explicit allegiance effect. The authors note that the mechanisms underlying the effect are poorly understood.
Genetic confound
Low - randomised training experiments.
Replication notes
Mixed at best. The effect survives bias correction only at 0.13 to 0.24, shrinks against active controls, and varies with who wrote the paper.
DOI / URL
10.1177/1745691617697076

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Self-control training, uncorrectedHedges g0.30 (95% CI 0.17 to 0.42)mixedpost-trainingunclearend-of-treatmentnon-cognitive
Self-control training, publication-bias correctedHedges g0.13 to 0.24 across correction methodsmixedpost-trainingunclearend-of-treatmentnon-cognitive
Author allegiance moderatordirectionlarger effects when strength-model proponents were authorsmixedpost-trainingunclearend-of-treatmentnon-cognitive
Control-group type moderatordirectionlarger against inactive than active controlsmixedpost-trainingactive-alternativeend-of-treatmentnon-cognitive

Cited by