Retrieval Practice vs. Worked Examples: When to Use Which
[cover image by u_fg0tkeqgiy on Pixabay]
by Cindy Nebel
We talk about a lot of different strategies on our blog that all have been shown effective for student learning. One question we sometimes get in talks and workshops is how to incorporate everything, or how to know when to do what. Today I’m reviewing a study that directly compared two effective learning strategies (retrieval practice and worked examples) on their usefulness for generalizing knowledge (transfer).
First let’s talk about some of the background research on these topics. A very typical set up for research on retrieval practice involves having participants read a passage with interesting facts about a topic, maybe an animal. Then some of that material (or half the participants) is quizzed and, after some kind of delay, participants take a final test that usually includes the same material that was on the initial quiz and maybe some new items. Worked examples research looks a bit different. In worked examples studies, participants see multiple, different worked examples usually for the same type of problem and this is compared with different control conditions.
Retrieval practice and worked examples have sometimes been directly compared to each other, but this issue of variability in testing materials has not been directly addressed. That is, retrieval practice usually uses repeated items whereas worked examples usually uses varied items during learning.
The Present Study
In this study (1), participants were randomly assigned to either receive repeated items or varied items but everyone participated in one round of retrieval practice and one round of worked examples (with the order mixed up).
Image by AngelRound on Pixabay
Now the fun part… to make sure that participants didn’t have any background knowledge in the area, they were exposed to a new alien language and alien math where they had to figure out the rules. For example, in English you can add -ed to the end of a verb to make it past tense and maybe in the alien language you add -fge. And in human mathematics a “+” symbol means to add the digits together, but in alien math a diamond symbol might mean to reverse the digits and subtract them. Ok, maybe that isn’t as fun as it sounds.
Experiment 1
Participants were first taught the new rules and given a simple example of the rule. Then they practiced with 10 items for each of the 5 rules they were taught, but remember that the 10 items were either all different or all the same depending on which condition they were in. For worked examples, they saw the problem and an explanation of how to arrive at the answer. For retrieval practice, they answered multiple choice questions and were given immediate feedback for the correct answer. After a short filler task, they were given 5 new items to solve for each of the 5 rules they were taught.
In every condition, retrieval practice outperformed worked examples. This was perhaps unsurprising because participants were taught the material first and then they were asked to retrieve. This is exactly what retrieval practice is designed to do.
Experiment 2
The researchers decided to try the same thing, but this time they removed the initial instruction. Now, participants wouldn’t just be recalling the rule, they would have to learn the rule from either the worked examples or retrieval practice. This time, worked examples DID outperform retrieval practice, but only when the retrieval practice was with repeated items (instead of varied). Participants were better able to figure out the rules and solve novel problems when they were walked through the rules in worked examples or when they got to attempt multiple problems with the same rule in retrieval practice.
Across both experiments the researchers also looked at how much time it took to complete the learning task. Retrieval practice took longer and so did varied problems, meaning the longest time was spent on varied retrieval practice.
Application
So, what do we make of all this? How can it be applied to the classroom or student self-directed learning?
What this research fundamentally shows is that the most appropriate strategy to use at a given time depends on the learning context and the educational goals.
Overall performance was highest when students were given instruction first and then retrieval practice, providing an opportunity for them to recall the rule they had been taught, so arguably in our classes we might want to aim for short lessons followed by retrieval practice to reinforce what was taught. However, that’s not always possible nor does life always look like this. For example, my medical students technically get taught information first, but are required to retain that information over the course of 9 months or more before being assessed on board exams. They will need to relearn the material before starting. What should they do in their self-directed learning? These data indicate that it might be most efficient for them to engage in worked examples rather than re-engaging in content review or jumping straight to retrieval practice. The same might be true of your students - when reviewing knowledge from pre-requisite courses, it may be most efficient to have them walk through worked examples instead of trying to recall the old material.
The other conclusion from this study is that when using direct instruction it is likely best to use retrieval practice and if using inquiry learning, it is important to make sure students are seeing a wide variety of examples if you want them to develop schema. But direct instruction is likely a more efficient route to learning for most novice learners.
Limitations
This is only one study and it was designed to make some very distinct comparisons. In the situations I described above, it’s unclear whether prior knowledge (even if very old) would change these results. It’s unclear exactly what the efficiency advantage is. And it’s unclear how the density of the material itself might matter. Learning an alien language is (mostly) different than learning human physiology and likely differs significantly from understanding the historical context of the Vietnam War.
So my biggest takeaway from this study is that there are no hard and fast answers when you’re dealing in the messy world of human learning. What will work in one situation will differ from what works best in another. As an educator, you should try different approaches to see what works best for your situation until you develop the expertise to recognize and anticipate the best learning strategies for your learning context - which might be a little like learning alien math.
Reference:
1) Cao, M., & Carvalho, P. F. (2026). Striking the Balance: How Variability Shapes Retrieval Practice and Worked Examples for Transfer Learning. Educational Psychology Review, 38(1), 71.
