How Much Spacing, for How Long, and for What Kind of Learning?

The broader evidence asks whether the classroom result is a local success or part of a reliable pattern. A 2025 systematic review and meta-analysis focused on applied classroom research found that distributed practice generally outperformed massed practice, but the size of the advantage varied widely across studies. The useful judgment is therefore not “use seven days” or “space everything,” but how to read an average effect alongside the retention goal, the learning task, the amount of practice, and the uncertainty in the evidence.

What the classroom average really means

The review began with more than 3,000 records and applied criteria designed to keep the comparison close to everyday education. Participants had to be enrolled in primary, secondary, or tertiary education. The learning material had to be relevant to a classroom or curriculum, and the distributed condition had to include interstudy intervals of at least one day. The comparison condition had to spend an equal amount of time studying in a massed arrangement. The review looked for retention or transfer outcomes, using the longest common retention interval when more than one assessment was available.

Twenty-five studies met the review criteria. Outcome data could be extracted for 22 studies, producing 31 effect sizes and more than 3,000 learners. The researchers calculated a standardized mean difference for each comparison: the mean outcome in the distributed condition minus the mean outcome in the massed condition, divided by a pooled standard deviation. A positive value therefore favors distributed practice. Their main random-effects analysis found d = 0.54, with a 95% confidence interval from 0.31 to 0.77 and p < 0.001.

In practical terms, the average retention outcome in the distributed conditions was a little more than half a standard deviation higher than in the massed conditions represented in the analysis. That is meaningful evidence in favor of spacing curriculum-relevant material. It is not the same as saying that every student will remember 54% more, that every spaced plan will produce the same gain, or that a learner can convert the effect size directly into a predicted test score. A standardized difference describes separation between group averages on their study outcomes; it does not specify the result for one person, one topic, or one timetable.

The applied estimate was smaller than the d = 0.85 reported by an earlier review that combined laboratory and applied studies and included more than 150,000 unique participants. That contrast is informative rather than embarrassing. Laboratory tasks often simplify the material, shorten the timescale, and reduce the number of competing demands. A classroom learner may be studying a complex concept, switching among subjects, fitting practice around lessons, and being assessed on more than verbatim recall. The smaller applied estimate is consistent with the possibility that real-world complexity changes the size of the advantage. It also shows why evidence from tightly controlled tasks should not be presented as if it were already a classroom prescription.

The average must be read with its spread. In the main analysis, more than 92% of the variance between study estimates was attributed to systematic differences rather than chance, reported as I² = 92.13%, with Q(30) = 230.33 and p < 0.001. This is very high heterogeneity. It means the studies did not all estimate one stable effect under interchangeable conditions. Their subjects, learners, schedules, practice activities, comparison conditions, and outcome measures differed in ways that the average does not fully explain.

The distribution of study results also showed asymmetry in a funnel plot, and Egger’s test was significant.