The Evidence on Teaching

Reduced Frequency of Knowledge of Results Enhances Motor Skill Learning

Winstein, C. J., & Schmidt, R. A. · 1990

grade Cquasi-experimentindependentfailed
Population
UCLA kinesiology undergraduates (Exp1 n=136; Exp2 n=58; Exp3 n=46); healthy young adults
Design
Three lab experiments, quasi-random assignment, researcher-designed arm-movement waveform-matching task scored as RMS error. Exp1: 8 groups n=17 (uniform reduced KR). Exp2: n=29/group, 50% FADED vs 100% KR, 5-min + 24-h no-KR retention. Exp3: n=23/group, 24-h retention WITH KR. Canonical guidance-hypothesis demo; no preregistration, no transfer test.
Key findings
Founding evidence for the guidance hypothesis: 100% KR flatters practice performance but breeds dependence; 50% faded KR gave 35% less error at 24-h no-KR retention and also beat 100% on a delayed KR retention test. Crucially the benefit was absent at 5-min retention and absent in Exp1 where reduced frequency was uniform, not faded — the effect depends on SCHEDULE, not frequency alone. Small n (~29/group), one artificial lab task, one lab.
Genetic confound
Randomized, so not confounded; but individual differences in feedback dependence are treated as error variance — no aptitude-treatment interaction modeled.
Replication notes
Cited as canonical, but the literature it spawned fails to confirm it in aggregate (McKay et al. 2022 frequency meta-analysis).

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Delayed 24-h no-KR retention (Exp 2)RMS error, 50% faded KR vs 100% KR50% group 35% lower error; F(1,56)=6.24, p<.01 (d≈0.67, computed from F)researcher-designedretention; delayed 24h; labactive-alternativeend-of-treatmentdomain-skill
Immediate 5-min no-KR retention (Exp 2)RMS error50% < 100% in same direction but NOT significant — benefit only emerges on the delayed testresearcher-designedretention; immediate; labactive-alternativeend-of-treatmentdomain-skill
Acquisition performance (Exp 1)RMS error during practice blocksReduced-KR groups performed WORSE than 100% KR, especially early practice (the paradox)researcher-designedacquisition; during practice; labactive-alternativeend-of-treatmentdomain-skill
Delayed 24-h retention WITH KR present (Exp 3)RMS error50% group lower; F(1,44)=6.56, p<.02 (d≈0.77, computed) — not just test-specificityresearcher-designedretention; delayed 24h; labactive-alternativeend-of-treatmentdomain-skill
Uniform (non-faded) reduced frequency (Exp 1)RMS error across 4 retention conditionsReduced ≈ 100% (no reliable benefit); advantage appeared only with a FADED scheduleresearcher-designedretention; delayed; labactive-alternativeend-of-treatmentdomain-skill

Cited by