Let's talk about a familiar feeling—that moment when you've done just enough to feel productive, but still wonder if it's really enough. You want that “that'll do for today” moment without pushing so far that you become fatigued and slip into overlearning. Finding the study sweet spot is a delicate balance between feeling accomplished (happy today) and achieving real retention (happy tomorrow; see Figure 1).
Students know they should study more—but how much is enough to achieve mastery without going overboard? That's true Memory Lab comes in. Our new mastery user interface implements a feature called practice until crown to help students learn less long per session and memorize more of the material in the long-term (see Figure 2).
Figure 2. The new mastery user interface.
But first, let's backtrack a bit. Why are we talking about crowns and mastery? And what in the world does it have to do with learning efficiency?
🤫 Cramming versus spacing – Quick win versus long-term victory
Here's the situation: Let's say you're preparing for an exam and want to cram (guilty as charged, we've all been there). Don't tell us learning scientists that you have never crammed. Here's a graph of when university students started studying for the test using our app (Figure 3).
As mentioned in our previous post, cramming just before the test is not irrational behavior. Since the information hasn't had much time to decay, you're more likely to recall it a short while later. Cramming works enough to pass the test the next day, but try to recall those facts a week later and… where did all that information go?
Satisfying learning gains from cramming quickly fade away
Research has long established that leaving time between study sessions yields better long-term retention than cramming (Krug et al., 1990). For instance, four 30-minute study sessions spaced over a week enhance retention more effectively than a single 2-hour session (Cepeda et al., 2008).
In 2021, Sense, van der Velde and van Rijn performed a study investigating the effectiveness of our spaced learning algorithm. As expected, the closer you get to the exam, the more people start learning (see Figure 3). Even though there are exceptions that manage to get a decent grade with a few days of study, there's a significant correlation between the number of unique days of study and the exam grade (see Figure 4).
Furthermore, cumulative usage time (hrs) is a lot less predictive of study performance than the number of unique sessions. “Okay, mister blog writer: 'How long should you study per session then?'”
Worry not about how long you should study, but about increasing your learning sessions. Spaced repetition — a term that might make your inner procrastinator groan — is the king when it comes to long-term retention. Memory Lab has optimized spacing learning items within study sessions, determining the ideal intervals between fact repetitions to achieve superior learning outcomes compared to non-adaptive methods (van Rijn et al., 2009; Wilschut et al., 2024). So, should you then learn as long as you can on multiple days to train your brain optimally? (Spoiler alert: not really.) Learning is like a balanced workout regimen for your brain, and let's be honest, not everyone likes jogging around the block when you can sprint to the end (aka cramming).
🥱 The science of crowning – Stop study fatigue, start spacing
We spent a lot of time thinking about when learners should stop practicing. If you practice too little, you won't retain anything. Practice too long, and you've hit diminishing returns—the Dark Playground of study fatigue, where you're doing the work, but not seeing much benefit.
Previously, learners determined their own study duration. Our data showed that users often stopped before their timer ran out, mainly when they achieved a crown. Students frequently requested a feature to end sessions automatically upon achieving a crown. As one student eloquently put it in our survey: “bro, why a timer.”
So, we decided to turn our learners into kings and queens of their own study sessions (Figure 5). Now, you practice until crown, meaning until you reach a level of retention where we predict you'll be able to recall what you've learned with over 50% accuracy the next day. And then, you stop. No more, no less, you're “Done for today” (Figure 6).
🤓 Why “overlearning” isn't always a good thing
Overlearning sounds like a productive idea on paper. Diminishing returns set in faster than you'd think. As Rawson & Dunlosky (2022) showed, overlearning is not efficient for long-term memorization. Subsequent spaced learning sessions negated the initial benefits of what we crudely call “cramming”. Our practice until crown system is aimed at finding the optimal amount of productive learning time, that sweet spot—where you're learning enough but not overexerting effort with marginal results.
Marginal gains after extended effort
Our system is – and continuously will be – fine-tuned with data from our NOLAI project. The project is aimed at figuring out when learners should stop learning. Turns out, pushing past that magic point where recall probability exceeds 50% the next day gives progressively smaller benefits (see Figure 7). In fact, it's like eating more food when you're already full—in the short term you feel happy, but long-term, this calorie overflow cannot be used efficiently. In this metaphor our model's memory activation—a measure for a memory's strength, indicates the amount of calories. You can make the memory more active, but at some point you're full for today. Only tomorrow can you eat more calories that can be efficiently used by your body. Similarly, you can repeat a study item more often today, but at some point your long-term memory activation for that item will not increase significantly anymore.
Why should you only know 50% of the knowledge tomorrow?
Our data from the NOLAI project for two mastery criteria (50% vs 80%) shows that learning until a higher threshold (that is, an expected accuracy tomorrow of 80%), does yield better results than the lower threshold (see Figure 8; Van den Broek , 2024). It also showed that (on average) our model's activation parameter accurately represents the probability of recall tomorrow.
However, for each individual fact, the memory accuracy gain decreased each study repetition (see Figure 7). Therefore the project will investigate more mastery criteria, eg 40% and 60%, which will give us a better sense of the marginal gain in retention associated with investing more study effort: the study sweet spot.
Crowns, progress bars, and healthy study habits
So how does this fit into daily life? When you log into MemoryLab, your progress is no longer measured by endless hours of study. Instead, it's visualized through a simple progress bar broken into three parts — each one moving you closer to that final gold crown (Figure 9).
You stop when you reach a crown today, and come back tomorrow for another go. It gives you a little victory every day. By limiting practice to this threshold, we're gently guiding learners into the art of spaced repetition, by crowning them kings and queens each day (Figure 10).
Even though three significant achievements are a great starting point for a monarch, their rule won't be memorable if they stop working after those achievements. Likewise, after gaining a gold crown learners should continue practicing or applying the material to keep their memories fresh.
Learning isn't just about cramming the most into your brain at once—it's about creating long-term retention. At Memory Lab, we've designed our practice until crown feature to be the middle ground between cramming and fatigue. With this we want our learners to leave the study session not feeling exhausted, but empowered, and (maybe, just maybe) having learned enough about spacing to make their inner procrastinators moan not groan.
References
Cepeda, NJ, Vul, E., Rohrer, D., Wixted, JT, & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095-1102.
Krug, D., Davis, TB, & Glover, J. A. (1990). Massed versus distributed repeated reading: A case of forgetting helping recall? Journal of Educational Psychology, 82(2), 366–371. https://doi.org/10.1037/0022-0663.82.2.366
Rawson, K. A., & Dunlosky, J. (2022). Successive Relearning: An Underexplored but Potent Technique for Obtaining and Maintaining Knowledge. Current Directions in Psychological Science, 31(4), 362–368. https://doi.org/10.1177/09637214221100484
Sense, F., van der Velde, M., & van Rijn, H. (2021). Predicting University Students' Exam Performance Using a Model-Based Adaptive Fact-Learning System. Journal of Learning Analytics, 8(3), 155-169. https://doi.org/10.18608/jla.2021.6590
Van den Broek, GSE (2024). Assessing Mastery During Adaptive Retrieval Practice: An Alternative for Traditional Knowledge Testing? TeaP 2024. Link
Passing the test: Improving learning gains by balancing spacing and testing effects. In Proceedings of the 9th international conference of cognitive modeling (Vol. 2, No. 1, pp. 7-6).
Wilschut, T., Sense, F., & van Rijn, H. (2024). Speaking to remember: model-based adaptive vocabulary learning using automatic speech recognition. Computer Speech & Language, 84.
