MemoryLab evolved from scientific research into memory and is closely linked to learning technology, research and innovation. New insights from science are used to improve our learning system. In addition, research by MemoryLab contributes to fundamental knowledge about learning and memory. Here, you can read more about how and why our system works: the science behind MemoryLab.
Decades of scientific research into learning, memory and forgetting tell us that testing yourself is the most effective way to learn information. To ensure optimal learning, you want to regularly repeat the facts you need to learn. MemoryLab measures how fast and how accurate your answers are during learning and determines the optimal repeat schedule for the learner based on that information. Studies show that students or learners using MemoryLab achieve 10 percent higher grades compared to non-adaptive learning methods (van Rijn et al., 2009; Wilschut et al., 2021).
MemoryLab is based on research in the field of learning and memory. Decades of this research has demonstrated two reliable effects: the testing effect and the spacing effect.
The testing effect is the finding that testing yourself is one of the most effective ways of learning. Testing yourself has been found to be more effective than reading and note-taking, for example (Rummer et al., 2017). When you successfully recall an item from memory, the memory representation for that item becomes stronger (Roediger & Karpicke, 2006). Practicing to recall an item during studying will also make it easier to recall that item on a test.
A popular method of self-testing is through the use of written or printed flashcards. Today, many students use digital learning systems such as MemoryLab to learn through retrieval practice. The MemoryLab system will present questions one at a time from a list of facts you want to learn. When a fact is first presented, you see both the question and the answer. All other times you only see the question, so you have to try to come up with the answer yourself (Figure 1).
Figure 1. Learning vocabulary with MemoryLab. Actively trying to answer a question is a proven effective learning strategy.
The spacing effect describes that you learn better when learning is spread over several learning moments (Figure 2a). It is now known that cramming out all the material just before a test or exam is not the best way to effectively memorize information. In the short term, this learning approach can work, making it a tempting and common strategy. However, if you want to make sure you remember all items, or if you want to remember information for a longer period of time, it is a good idea to use a number of separate learning sessions. Studying for half an hour four times in a week will ensure that you will remember information better and longer than if you try to learn all the material the night before the test (Cepeda et al., 2006).
Figure 2a. Learning over multiple sessions is more effective than learning all the materials in one long session.
Interestingly, this spacing effect does not only apply to the time between learning sessions, but are also effective within the learning session when you spread the repetition of facts (van Rijn et al, 2009). You will remember facts better if you repeat them back and forth, rather than learning them one by one (Figure 2b). Learning is therefore most effective if you always leave some time between the repetition of a fact. In that time you can learn other facts. If you wait too long to repeat a fact, you will forget it and have to start over. Therefore, the optimal point to repeat a fact is when it is almost forgotten, and it takes effort for you to retrieve it. The aim of MemoryLab is specifically to achieve this. The program estimates which item you are closest to forgetting and will present that item to you next.
Figure 2b. The spacing effect also works for facts within a session. Spreading facts during a learning session is more effective than learning the facts one by one.
Estimating which item a learner will likely forget soon is not simple. After all, not everyone will forget an item after the same amount of time. How quickly items are forgotten differs from person to person and it also depends on the difficulty of the item itself.
How does the MemoryLab learning algorithm estimate when a student is going to forget a fact? If you don't know the answer to a practice question, you have obviously forgotten (part of) the answer. However, knowing the answer does not always mean that you will remember the answer forever. When you are asked what the capital of England is, you will probably quickly answer: 'London'. When asked what the capital of Norway is, you may be able to answer it ('Oslo'), but it will probably take longer to come up with that answer. This is because this fact is not very active in the memory of most students. Therefore, it takes more time to think of the answer. In other words, the time it takes to give a correct answer can be used to estimate how well a student already knows a fact. MemoryLab therefore analyzes the reaction time for each answer (see Figure 3).
Figure 3. A visual representation of the memory model used by MemoryLab. The red line shows the activation of a fact in memory. When the activation gets too low (below the dotted line) the fact is forgotten. Each black dot represents a repetition of the fact. After the first repetition of the offense, the activation decreases quickly, but as the offense is repeated more often, the activation decreases less quickly. MemoryLab can estimate the activation and the rate at which the activation decreases for each fact. To do this, MemoryLab uses response times. For a correct and fast answer, the memory activation is high: the student knows the answer well. For a slow answer, the activation is lower: the student does not know it very well yet.
Based on the response time and accuracy of your answers, MemoryLab calculates a rate of forgetting for every item: a number that predicts how quickly you will forget an item (Sense et al., 2016). If this rate of forgetting becomes very high for an item, the program will make it a priority to show you the item again; if the number is very low, the program won’t show the item for a while. With every response you give, the rate of forgetting is updated. This way, the program adapts to your learning pace.
In 2009 van Rijn, van Maanen and van Woudenberg found that studying with Memorylab is more effective than studying with a traditional flashcard method. In their study, students learned French words for fifteen minutes with MemoryLab or with a widely used flashcard system. The flashcard system divided the French words into groups of 5 facts and presented them one by one to the student. If the student had translated all the words correctly, he continued with a new group of 5 words. A test followed one day after the learning session. On this test, the students who learned with MemoryLab scored significantly higher than students who learned with the flashcard method (Figure 4).
Figure 4. Grades on a French word test after a 15-minute learning session the previous day. 1 = Flash card, 4 = Memory Lab. Algorithm 2 and 3 represent two other learning algorithms.
More recently, a study by Wilschut and colleagues (2021) compared MemoryLab to a flashcard method called the Leitner method. In the Leitner method, items to be learned are grouped according to how well the learner knows them. All items start in one group. If a learner answers an item correctly, the item moves to the next group, and so on. For items in every next group, the learner waits a longer time before practicing them again. Using either the Leitner flashcard method or MemoryLab, participants studied a set of English words. After a break, they were tested on the words they had studied. Participants who studied with MemoryLab were approximately 10% more likely to give a correct answer during the study session, and approximately 8% more likely to give a correct answer on the test compared to participants who had learned with the Leitner flashcard method (Figure 5).
Figure 5. Accuracy during the learning session (pink solid) and during the test (purple striped), for students who learned with the Leitner flashcard system (LS) and for students who learned with the MemoryLab system (RT-S).
These results confirm that MemoryLab results in a higher learning efficiency than traditional flashcard learning methods: students learn more information in the same study time.
Van Rijn, H., van Maanen, L., & van Woudenberg, M. (2009, July). 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., van der Velde, M., Sense, F., Fountas, Z., & van Rijn, H. (2021, June). Translating a typing-based adaptive learning model to speech-based L2 vocabulary learning. In Proceedings of the 29th ACM conference on user modeling, Adaptation and Personalization (pp. 245-250).