Retrieval practice, spaced practice, and interleaving have become some of the better-known insights from cognitive psychology. They have found their way into education. A lot of research exists on all three. But… a considerable part of that research was conducted in lab settings or in higher education. So what happens in a classroom with real curriculum content? For instance, consider a class of 25 twelve-year-olds who all bring different prior knowledge.

A new meta-analysis by Sterre Ruitenburg and colleagues examined what effects remain when we look only at research conducted in real primary and secondary school classrooms. To answer this question, the researchers used fairly strict criteria. Studies had to be experimental, involve students aged 6 to 18 in mainstream education, and use material that was part of or directly related to the students’ curriculum. Instruction, practice and testing also had to take place in a school setting. In the end, the researchers found 53 studies, including 7,303 students and 233 effect sizes.

Good news: all three approaches hold up. For retrieval practice, the researchers found an average effect of g = 0.37. For spaced practice, it was g = 0.20, and for interleaving g = 0.40. These effects are not huge. In fact, for retrieval practice and spaced practice, they are smaller than those found in several previous meta-analyses. But that isn’t all that strange, as this is often the case.

But this does not mean that you can simply say: use retrieval practice, spacing and interleaving, and everything will be fine (see also my earlier reservations about the term ‘effective teaching’). For retrieval practice, for example, the effect depended on the type of knowledge. It also depended on how the researchers tested the students afterwards, the match between practice and testing, and the learning outcome.

The researchers found the clearest effect for factual knowledge (g = 0.53), but no significant effect for conceptual or procedural knowledge. Retrieval practice also benefited retention, but not comprehension or transfer. Still, we shouldn’t conclude too much from this. Some comparisons included only a few studies, and several moderators were correlated. The researchers themselves therefore warn against strong conclusions. It would be too easy, for example, to say that retrieval practice does not work for comprehension or transfer. We simply don’t have enough good research in real school classrooms yet.

The researchers offer a possible explanation. Retrieval practice mainly helps students consolidate what they have learned in memory. But if we want students to understand complex knowledge or apply it in a new situation, they also need to engage in the cognitive processes required for understanding and application. Just retrieving information from memory may not be enough. Generative activities and retrieval practice may therefore complement each other.

Spaced practice tells a similarly nuanced story. It had a relatively small effect size (g = 0.20), but it still outperformed massed practice. When comparing distributed practice with massed practice, the researchers found an effect size of g = 0.39. But when they compared different forms of spaced practice, they found no significant differences between them. So spacing seems to be important, but more spacing is not always better.

Timing seems to be important as well. The researchers found that the interval between practice sessions in relation to the time before the final test mattered. This corresponds with previous research suggesting that there can be too little or too much time between practice sessions.

And finally, interleaving. The researchers found an average effect of g = 0.40, although here they could draw on only ten studies involving 1,902 students. Again, the results varied considerably across studies.

There was one interesting finding about feedback. In this meta-analysis, interleaving had a significant positive effect when there was no feedback or delayed feedback, but not when there was immediate feedback. We should be particularly careful with the large effect for delayed feedback because it was based on only five effect sizes. The researchers acknowledge that their explanation is speculative. It may well be that immediate feedback reduces the need for students to discriminate between different types of problems being interleaved.

None of this means that retrieval practice, spacing, or interleaving suddenly do not work. On the contrary. The researchers found positive average effects for all three strategies, even when restricting their analysis to authentic classroom situations. However, a learning strategy is not a switch you simply turn on.

Yes, they can work, but not just like that.