When a quiz stops being just a quiz
A professor finishes explaining a difficult concept. The slides were clear. The examples seemed to land. Students nodded.
There are ten minutes left.
The familiar instinct is to explain the idea once more. Perhaps another example will make it stick.
But there is another option.
Close the notes. Put away the slides. Ask students to explain what they remember.
That small change matters because a quiz can be a learning event, not simply a grade.
We often treat teaching and testing as separate stages. First, students learn. Later, we test whether the learning survived. But research on retrieval practice suggests that distinction is too simple. The act of trying to bring information back from memory can itself contribute to learning.
Henry L. Roediger III and Jeffrey D. Karpicke demonstrated this in their influential 2006 paper, Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention.
Across two experiments, students studied prose passages. Some then restudied the material, while others took immediate free-recall tests without feedback. The researchers tested students again after intervals of five minutes, two days or one week.
The short-term result is important.
When the final test came only five minutes later, repeated studying produced better recall than repeated testing.
But after the longer delays, the pattern reversed: prior testing produced greater retention than additional studying.
There was another revealing result. Repeated studying increased students' confidence in their ability to remember the material, even though prior testing led to better performance on the delayed tests.
In plain English, seeing something again can make it feel familiar. Trying to retrieve it requires something different.
The question is no longer merely measuring learning.
It can become part of the learning.
Retrieval changes what the learner has to do
Imagine two students preparing for an economics course.
The first rereads a page explaining opportunity cost three times. On each pass, the paragraph becomes easier to follow. The terminology looks familiar. By the third reading, the student may feel reasonably confident about the concept.
The second student reads the explanation, closes the material and receives a prompt:
Explain opportunity cost in your own words, then give an example that was not used in class.
Now the answer is no longer visible.
The student has to bring relevant knowledge back to mind and organize it into an explanation.
That difference gets close to the heart of retrieval practice.
Roediger and Karpicke's research does not mean studying is unnecessary. Students obviously need something to learn before they can retrieve it. Nor does the study establish that every quiz or question automatically improves learning.
Its finding is more precise: under the conditions they studied, retrieving previously learned material produced stronger retention over longer delays than spending comparable opportunities restudying it.
That distinction deserves attention in learning science in higher education.
University teaching understandably contains a great deal of exposure. Students encounter knowledge through lectures, readings, demonstrations, videos, slides, seminars and examples.
When students appear uncertain, the natural response is often more explanation.
Sometimes that is exactly what they need.
But sometimes the problem may not be a lack of exposure. Students have seen the concept several times. What they have had less opportunity to do is retrieve and use it when the answer is no longer sitting in front of them.
A carefully designed question changes that condition.
Instead of asking students to recognize an explanation, it asks them to produce one.
And that creates a different kind of learning opportunity.
The practical implication: make retrieval part of teaching
This does not mean turning every university course into a continuous sequence of exams.
Consider an introductory biology class.
At the beginning of a session, the professor displays one prompt:
Without looking at your notes, explain why increasing temperature does not indefinitely increase the rate of an enzyme catalyzed reaction.
Students have two minutes to write an answer.
They then compare their reasoning with a partner before the instructor revisits the concept.
No additional high stakes exam has been created.
But something important has changed.
Students are no longer simply hearing the explanation again. They are trying to retrieve relevant knowledge and organize it into a response.
Current higher-education teaching guidance points in the same practical direction.
In a March 2026 article for Times Higher Education Campus, Katy Burgess of Cardiff University describes retrieval practice as an active learning strategy in which students try to remember what they have learned. She argues for embedding retrieval into university teaching rather than treating it only as something that happens during formal assessment.
The practical implication is therefore less dramatic than “give students more tests.”
It is:
Create more purposeful opportunities for students to recall, explain and use important ideas.
Those opportunities can be low stakes. A retrieval prompt might appear at the beginning of class, after a reading, before introducing the next concept or several days after an important idea was first taught.
And the form of the question should follow the learning goal.
If students need to remember terminology, straightforward recall may be appropriate. If the course expects students to reason, explain or apply knowledge, then retrieval activities can ask them to perform those kinds of intellectual work.
There is an important limitation.
Roediger and Karpicke's 2006 experiments examined the learning and recall of prose passages under controlled conditions. That study alone cannot tell faculty exactly how retrieval practice will work in every discipline, for every learner, or for complex outcomes such as mathematical proof, clinical judgment, creative work or extended argument.
Retrieval practice is better understood as a useful learning principle than as a universal teaching recipe.
Faculty still have to decide what knowledge matters, what students should be able to do with that knowledge, when retrieval is appropriate and what evidence counts as meaningful understanding.
Why this matters now
Generative AI makes the distinction between having access to an answer and being able to explain an idea increasingly important.
Students can now obtain polished explanations of many topics almost instantly. Those tools may themselves be useful for learning. But access to an explanation is not the same thing as retrieving relevant knowledge, connecting ideas, applying a concept or explaining why an answer makes sense.
That gives an old finding from learning science a very current relevance.
The goal does not have to be more testing.
It can be better questioning.
A thoughtfully designed course might move repeatedly between exposure and retrieval: encounter an important idea, work with it, put the source away, bring the idea back to mind, explain or apply it, receive feedback and return to it later.
Seen this way, questions are not simply instruments for producing grades.
They are part of the architecture of learning.
For university leaders, this also offers a useful way to frame faculty conversations about assessment in an AI-rich environment. The discussion does not have to begin only with detection, policing or technology.
It can begin with a more fundamental question:
What intellectual work should students be able to perform when the answer is not already in front of them?
At Socratic Metric AI, this principle informs our interest in supporting student explanation and evidence of reasoning. Technology can create opportunities for students to articulate how they reached an answer and respond to questions about their thinking.
But technology should not decide, on its own, what genuine understanding looks like.
Faculty retain that judgment. They determine the learning goals, interpret the evidence and decide what counts as meaningful understanding within their discipline.
A good lecture can introduce an idea beautifully.
A good question asks the learner to bring that idea back.
For your next class, consider replacing just a few minutes of additional explanation with one carefully chosen, low-stakes retrieval question.
Then pay attention not only to whether students get the answer right, but to what they are able to bring forward and explain.
Sources & Further Reading
Roediger, H. L., III, & Karpicke, J. D. (2006). “Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention.” Psychological Science, 17(3), 249–255.
https://doi.org/10.1111/j.1467-9280.2006.01693.x
Burgess, K. (2026). “Ten Tips for Embedding Retrieval Practice in University Teaching.” Times Higher Education Campus, March 13, 2026. Cardiff University.
https://www.timeshighereducation.com/campus/ten-tips-embedding-retrieval-practice-university-teaching