A meta-analytic review of the distribution of practice effect: now you see it, now you don't

63 studies, 112 effect sizes, N=8,980

Adults in laboratory and organisational training settings; child and clinical samples explicitly excluded

63 studies, 112 effect sizes, N=8,980. Hedges g corrected to d. Tasks clustered into 4 groups by rated complexity and mental/physical demand. Moderators: task cluster, intertrial interval, 9-item rigor checklist, acquisition vs retention. Authors flag that 'retention' required only 24h separation and just 3 of 16 retention effect sizes exceeded 24h.

Distributed beats massed overall (d=0.46) but this is not a general law. The effect concentrates in simple psychomotor tasks (0.97) and collapses toward zero as complexity rises (0.11, 0.07); complexity correlates -.25 with effect size. Low-rigor studies yield a 3x larger effect (1.22 vs 0.40). Counterintuitively short rests helped most (<1 min d=0.71) and >1 day rests least (0.16), opposite to the verbal spacing literature. Acquisition and retention effects were statistically indistinguishable (0.45 vs 0.51) — but 'retention' meant 24h in 13 of 16 cases, so this meta cannot speak to durable learning.

None; no aptitude-treatment interaction analysis, though differential 'response to schedule' is exactly what heritable trainability would predict.

Substantially deflates Lee & Genovese (1988), which reported d=0.96 acquisition and 0.53 retention on simple motor tasks only; the discrepancy is explained by task-type composition.