Feedback frequency and mental practice in motor learning

Verdict

This cluster is a case study in the replication crisis reaching motor learning. Of four popular levers, one survives at half size, one is clean-but-unreplicated, and two collapse.

What the evidence shows

Source	Design	Grade	Key effect
Winstein & Schmidt 1990	experiment	C	50% faded KR → 35% less error at 24h retention, F(1,56)=6.24, d≈0.67
McKay 2022 (feedback frequency)	meta	B	No significant effect at acquisition, immediate or delayed retention; no crossover
McKay 2022 (self-controlled)	meta-reanalysis	B	g=0.44 → 0.107 after bias correction (k=52, N=2,061); "not distinguishable from zero"
Driskell 1994	meta	C	Mental practice d=0.527; decays with retention interval
Tóth 2020	replication meta	C	Replicates at r=0.131 (d≈0.26) — half size
Jiménez-Díaz 2023	meta	C	Physical beats mental at acquisition (ES −1.16) and transfer (0.50); combined ≈ physical alone

The guidance hypothesis — that frequent augmented feedback breeds dependence, so faded feedback should hurt acquisition but help retention — has a beautiful founding demonstration and a disappointing literature. Winstein & Schmidt's result is real and specific (the benefit came from fading, not merely from reduced uniform frequency, and appeared only at 24h, not 5min). But the meta-analysis of the whole literature finds no significant effect at any phase and no acquisition-to-retention crossover — the hypothesis's defining prediction. Prediction intervals span more than a standard deviation in both directions; the underlying studies are severely underpowered.

Self-controlled feedback is the cleanest replication-crisis signature in this database: a naive g=0.44 falls to g=0.107 after selection-bias correction, with a p-curve lacking evidential value. A 2025 meta-analysis reports retention 0.63 / transfer 0.68 — but found significant Egger's bias and did not correct for it, so it simply reproduces the naive estimate. The related OPTIMAL-theory motivational pillars (enhanced expectancies, autonomy support) are weak-to-null in preregistered work across ~95 studies.

Mental practice survives, at roughly half its advertised size, and is a supplement: physical practice beats mental practice head-to-head, and mental+physical is about equal to physical alone at retention. In children the imagery evidence is grade D — 19 of 22 studies positive but only 2 poolable, only 2 with any retention test, GRADE rated LOW/VERY LOW.

Hereditarian-lens assessment

Risk: low — randomised feedback and practice-mode manipulations. These methods target the rate at which a specific skill is acquired; none tested underlying capacity, and none claims to.

Boundaries & what critics say

The founding study did not generalise. Treat faded feedback as a reasonable default with weak meta-analytic backing, not an established law.
Bias correction is decisive here. Any motor-learning claim in the motivational family should be checked for whether the meta corrected for small-study effects; several do not.
Children are under-evidenced across all four levers.

Practical guidance

Fade feedback as competence grows rather than cueing every trial — low cost, plausible benefit.
Don't buy self-controlled-feedback or OPTIMAL-theory packages; the effect is statistically indistinguishable from zero.
Mental practice supplements, never substitutes — most useful for cognitively-loaded tasks and more experienced performers, and it decays with delay.

Open questions

Why the faded-feedback demonstration fails to reproduce (task simplicity? fading schedule specifics? power?) is unresolved.
No adequately powered child-focused trial exists for any of these.