Blocked vs random practice: the retention trade

The effect that punishes tidy practice

Hitting the same club at the same target until the bucket is empty is blocked practice. Shuffling club, target and distance so no two consecutive balls are the same shot is random practice. The difference has a name, contextual interference, borrowed from William Battig's verbal learning work in 1966.

Shea and Morgan carried it into movement in 1979 with the barrier knock-down task, three arm patterns run against a stimulus light. One group practised all the repetitions of a pattern before starting the next; the other got the three shuffled. During practice the blocked group was the better of the two. On the test afterwards, the shuffled group beat them.

That reversal is the whole subject. How well you strike it inside a session and how much of the change survives until Saturday are two different measurements, and this is the clearest case in sport of them pointing opposite ways.

What the putting green measured

34.7 vs 45.9 cmRandom against blocked putting error, 72 hours after the last session

Forty novice golfers were split into four groups of ten for a 2024 experiment in the European Journal of Sport Science. Each hit 10 blocks of 12 putts a day for three days, 360 in all, at a 10.8 cm hole from 1.22, 2.44 and 3.66 metres. The blocked group worked one distance per day; the random group had the three shuffled.

Seventy-two hours after the last session, error was measured as deviation from the middle of the target. Random averaged 34.72 cm, blocked 45.85 cm, and an untrained control 67.71 cm. The gap between random and blocked came out at a Cohen's d of 1.18, and a second retention test with mixed distances repeated it, 33.56 against 46.23.

A transfer test followed from four metres, a distance nobody had practised. Random averaged 45.48 cm and blocked 50.27 cm, and that difference was not significant. The advantage that was plain at trained distances thinned out at an untrained one.

Where the golf studies stop agreeing

3 of 6Golf practice-scheduling trials with statistical power under 0.6

Set that beside the rest of the golf literature and it stops looking settled. Brady tested the idea on real golf skills in 1997, giving two groups of 18 novice golfers four foundational skills in a naturalistic setting under low and high interference. A t test found no significant difference between the groups.

The putting trials are no tidier. Chua and colleagues gave 36 novices 60 putts and found nothing at retention, with random ahead only on transfer. Goodwin and Meeuwsen ran 30 novices through 198 trials and again found nothing at retention, nor at the two closer transfer distances; the varied groups pulled clear only at 6.23 metres.

A 2024 systematic review in Frontiers in Sports and Active Living gathered 52 randomised trials on learning golf. Six covered practice scheduling, and their post-hoc statistical power averaged 0.70, three of the six below 0.6. The reviewers add the caveat that matters most: nearly all of it was simple putting, and nearly all of it was novices.

The number that should slow the coaching down

0.23 (p = 0.24)Random-practice retention benefit in applied settings, 54-study meta-analysis

The largest pooled estimate is a 2024 meta-analysis in Scientific Reports covering 54 studies of delayed retention. Overall, random beat blocked by a standardised mean difference of 0.63. Split by setting, the picture splits with it. In laboratories the figure was 0.92. In applied settings it was 0.23, with a confidence interval from minus 0.16 to 0.62 and a p value of 0.24.

That is not a significant result. Dropping three outlying outcomes took the applied figure to minus 0.01, fractionally favouring blocked practice, and heterogeneity reached 89 per cent. A companion meta-analysis on transfer, in Frontiers in Psychology, put the applied estimate at 0.37 with a p value of 0.06.

So the effect is real and reliable on a laboratory bench. Out where balls are struck it has never been shown cleanly. The coaching version says random practice is simply better; the measured version says better for adults, in laboratories, on tasks chosen because they were easy to score.

One bucket that uses both

Porter and Magill built the schedule worth copying in 2010. Sixty novices hit 81 putts across three distances. One group took 27 consecutive putts from a spot before moving on; a second had them shuffled; a third started blocked, moved through a serial order, and finished random. That third group beat both of the others at retention, and blocked and random did not differ at all.

The order that follows is unglamorous. Open blocked, one club at one target, while the change still feels foreign. Move to serial, rotating a small set of shots in an order you know in advance. Spend the rest of the bucket random: choose club and target before each ball, never repeat a shot, and play your full routine.

Hold it lightly. Every participant above was a novice and nearly every task a putt, so nothing has shown the schedule survives contact with a driver on a course. The Challenge Point idea underneath it, which Guadagnoli and Lee set out in 2004, holds that difficulty must be matched to the player rather than maximised. That is an argument for beginning blocked, not for skipping it.