The theory seems simple enough. If you’ve got, say, eight hours of instructional time to teach a specific topic, you’d be best off teaching the topic in smaller chunks rather than one eight hour block.
In the image below, taken from a slide I used recently to explain the point, the top scenario, with all eight hours delivered long before the content will be applied in the real world, is the least effective. The second scenario, affectionately known as “cramming,” is not ideal, either.
The bottom two scenarios are better. One requirement: sessions after the first need to have review from the previous sessions built in. This is a bit of a disadvantage because it means during your eight hours the review eats into your time, so theoretically you cover less. But this more than made up for by deeper learning.
The mechanics of this advantage aren’t necessarily well understood. I like the theory that suggests that while reviewing the material, learners realize they aren’t remembering as much as they think they should, which causes them to rethink their learning strategies. Perhaps metacognition wins the day.