Menu ▼

Proven learning methods

Learning exists to enable an organism to adapt itself to an unpredictable environment in which it’s going to live. We don’t know what kind of world we will be born in and how it will change over time, so we need tools that help us change with it.

A few learning methods have consistently outperformed the rest. They also have scientific evidence to back them up. Those methods are:

  • Retrieval: Forcing yourself to retrieve something from memory (recall) makes you remember it better than reviewing it (recognition).
  • Variation: Learning different topics or skills in the same learning session, and actively learning from two or more examples.
  • Spacing effects: A topic or skill is studied and practiced more than once at different times.

Retrieval, variation, and spacing complement each other and should be combined for the best effect. Let’s go into more detail about each method.

Method 1: Retrieval (recall over recognition)

Forcing yourself to retrieve something from memory (recall) makes you remember it better than just reviewing it (recognition).

If you’re asked a question, you will either know the answer or not—the feedback to you is clear. If you’re also presented with four possible answers to the question, you might deduce the correct answer by eliminating incorrect ones or recognizing it from previous learning sessions. That approach doesn’t test your memory and retrieval strength but your deduction skills.

If you highlighted something in a book, you likely just noted that this information might be important and where to find it. When you re-read the study material in subsequent learning sessions, it’s too easy to deceive yourself that you learned it. The illusion of knowledge has been tested and verified in many contexts and always comes down to human overconfidence. But where does that confidence come from? One of the scientists exploring the problem explains:

We think the source of the illusion is that people fail to distinguish what they know from what others know. We’re constantly depending on other people, and the actual processing that goes on is distributed among people in our community. It’s as if the sense of understanding is contagious. When other people understand, you feel like you understand.

I will add, “When it’s written down clearly by others, you feel like you can recall it.” You probably can’t.

Always choose recall over recognition as a learning method. Retrieving something from memory feels harder than recognizing it on a page or a screen, but long-term performance when using retrieval is always better.

Here are some specific ways to benefit from retrieval practice:

  • For flashcards, favor withholding the correct answer over showing multiple answers.
  • Frequent quizzing helps retain knowledge for longer, but you can quiz yourself or others even before the learning session because that will prime and divert the attention to essential parts of the lesson.
  • When you read, hear, or watch something you want to remember, write it down in your own words. If you had to take verbatim notes during a lecture or meeting, take a new piece of paper or open a new document, and try to recall what is most important without looking at the first set of notes.
  • Instead of re-reading your notes, seek someone to ask you questions from the notes. Or, as you’re writing your initial notes, immediately write questions you’ll want to answer in the future.
  • Teaching, explaining, and presenting from memory is a form of retrieval practice. You must deeply understand the material to be ready to explain it to different audiences and field questions on the spot. One great way to eliminate jargon and reveal gaps in understanding is explain like I’m five, a technique where you imagine you need to explain a concept to a five-year-old. Most of my talks and courses were sparked by a desire to master a particular subject. This essay too.

Method 2: Variation (interleaving and transfer)

Interleaved practice, or interleaving, is a way of learning where different topics or skills are learned at the same time. This is in contrast to focusing on one skill or topic at a time, or blocked practice, which entails little-to-no variation.

Interestingly, interleaving works best when things to be learned are similar and likely to interfere with one another. A classic cognitive example from the scientific literature is learning to distinguish painting styles from different eras or painters. Instead of focusing on one at a time, combining them and trying to differentiate when they are similar will be hard but will lead to better long-term results in the ability to distinguish them. It’s the same for learning any motor skill.

Interleaving is more challenging during the initial acquisition because it feels harder to learn several related things at once and because the observed performance improves slower. However, if you persevere, the long-term retention is significantly higher than if you would have used blocked practice.

Beyond interleaving, there are additional benefits of variation in learning. When you learn something in only one way or place or from only one example, it’s impossible to disentangle the thing to be learned and the context. For instance, one replicated study demonstrated that when students learn something in one physical location and then take a delayed test in another location, their recall of learned material suffers. However, if they switch locations for each learning session, the recall in a new test location remains high.

Another benefit of variation is creating useful abstractions. When you have multiple examples, you can highlight similarities and differences, find the underlying principle, and then apply it in new domains. This behavior is called transfer learning, one of the main ways we quickly adapt to new situations. In studies that explored this phenomenon, people who were given:

  • only one example were least likely to create a useful abstraction that could be applied to solve a novel but similar problem
  • only one example and the relevant abstraction together weren’t much more successful
  • two examples were much more likely to create a useful abstraction and apply it to other domains
  • two examples and the relevant abstraction together were most successful in solving a novel problem

If you wonder if learning only an abstraction would be the most efficient way, the answer is “no.” Multiple examples serve as real-world applications of the abstraction, and the best results for transfer come only when they are paired together.

Unlike interleaving, where similar training material added additional difficulty and better long-term retention, it seems that dissimilar training examples make it more difficult to extract a useful abstraction but allow it to be used more flexibly once acquired.

The general rule of applying variation in learning—if you’re optimizing for long-term retention—is to always acquire and use knowledge or skills in two or more contexts or to learn something from two or more examples.

Here are some specific ways to do that:

  • If you’re practicing a move in sports—and depending on the sport—practice the move with different players or opponents, on various surfaces, under diverse weather conditions, and with different equipment and clothing. Also, try to combine learning different moves in one learning session and over a short period instead of focusing on only one move at a time.
  • Most schools are already set up in a way where multiple subjects are taught at the same time. While it feels like juggling too many things at once, it turns out to be beneficial long term.
  • You can also add variation within the same subject. For example, if you’re learning a foreign language, practicing how to apply multiple tenses (past, present, future) during one learning session should yield better long-term results than focusing on only one.
  • Reading several publications and media about the same or similar topic helps highlight similarities and differences between different authors, where they agree and disagree. Getting to that point of view is impossible from only one source.
  • Related to the point above, I take extensive notes about things I read, encounter in the world, or contemplate. I review them later and connect them to something similar or new I find. I can’t do it by relying on memory alone.

Method 3: Spacing effects

Spacing effects or temporally distributed practice is a way of learning where certain information or skills are studied and practiced more than once at different times. That is opposed to massed practice, where everything is crammed into a single longer session. In other words, if you want to learn a particular topic, scheduling two one-hour sessions a week apart will produce better results than one two-hour session. The distributed practice has been studied for more than a century. Its effects are undeniable and often underutilized by students and learners.

The biggest challenge in applying the method is knowing the optimal distribution of practice sessions because too many variables go into figuring it out:

  • Your current knowledge of the topic and adjacent topics.
  • How much time are you able to devote to one practice session? For how long can you focus and engage with the learning material?
  • When do you need to know the topic? In other words, are you studying for a test in a month, in a year, or are you trying to learn something that you want to remember forever?
  • Are you learning a motor skill or improving your knowledge?

In one experiment, researchers wanted to understand the optimal spacing when someone wants to learn not-well-known facts about the world (imagine history and geography facts). The researchers set up a study with an initial learning session, a subsequent repeated learning session, and the final test sometime later. The researchers varied the period between the two study sessions and the period between the initial study session and the test. The experiment results showed that the farther the test is in the future, the closer the second study session should be to the initial one as a ratio to the total duration. For example:

  • If your test is in two months, you will perform best if you restudy in about three weeks (~35% of two months).
  • If your test is in a year, you will perform best if you restudy in five weeks (~9% of one year).

“Great!” you say, “I can apply that myself.” But will the findings hold for learning Spanish? What if you already know Italian, which is similar to Spanish? Will the results stay valid for learning how to paint? Or basketball? And what if you are willing to practice twice a week until the test? It’s unclear how the findings for optimal spacing are transferable to different skills and topics. Future research will likely help.

Not everything is lost. Distributed practice is fantastic, and you will perform better by using almost any repeated schedule instead of doing only one massed practice. Here are some specific ways to benefit from spacing effects:

  • Use a specialized service for a skill you want to learn if it includes spacing effects. For example, Duolingo uses billions of data points to optimize the best time for you to refresh your vocabulary. The service adapts to your needs, starting from good baselines and recommendations that would be impossible to get only from your information.
  • Use a generic tool that spreads learning material over time, increasing the gaps between learning sessions. The one I use often is Anki flashcards.
  • After your learning session (or several related sessions), make a reminder to test yourself at a future time; a recurring calendar entry works just fine. When reminded, evaluate how well or badly you performed. If it was good, do a quick restudy and double the time until the next test reminder; if it was bad, plan for a restudy session soon.
  • In general, spread learning over a longer period with many shorter learning sessions in-between. Remember 7Ps: Proper Planning and Preparation Prevents Piss Poor Performance.

Desirable difficulty

Scientist Robert Bjork talks about the desirable difficulty in his work as a good indicator of the effectiveness of a learning method. The three methods described above are unintuitive in a way that learners find them more challenging and feel that they are not progressing as fast as they could, but later test results undeniably demonstrate that they perform better than methods that feel easier.

If you want to learn something long-term, it will be demanding along the way, often so much so that you will question your sanity and competence.

Learning for the test

There is a small caveat to what I wrote in this essay. Blocked and massed practices usually deliver better test results immediately after the learning session. These methods could lead to the “learning for the test” approach that I too often applied to my early schooling:

  • Go through one narrow topic days before the test
  • Pass the test
  • Never revisit the topic
  • Completely forget it

My goal then was to survive school, not learn for the long term.

That being said, many people will find themselves in a situation where they must pass an unnecessary-but-bureaucratic exam, so it’s good to be aware of what outcomes will be driven by different learning tools.

Learning myths and methods to avoid

There is no scientific basis for the following:

  • Individual learning styles: visual, auditory, kinaesthetic. Some learning modes are better suited for certain types of lectures or topics, but all learners benefit from that mode similarly.
  • Multitasking produces good learning results.
  • Being “left- or right-brained” affects the ability to learn.
  • Brain training games.
  • Statements that humans use only 10% of their brain capacity and that we can somehow increase the percentage.

References and additional resources

I have been collecting notes and references about learning methods for many years. I stumbled upon Acquiring an Accurate Mental Model of Human Learning: Towards an Owner’s Manual article in the second half of 2022. It was published just a year earlier and is a fantastic collection and analysis of existing scientific work about learning. It’s so good and extensive that it almost prevented me from writing this essay. The article is 68 pages long. The last 25 pages contain references to studies and scientific literature, which is why I decided not to cite many references here. If learning methods interest you and you want to dig deeper, this is probably the best starting point: [PDF file] Pan, Steven & Bjork, Robert. (2021). Acquiring an Accurate Mental Model of Human Learning: Towards an Owner’s Manual

The quote about the source of our illusion of understanding is by Steven Sloman, professor of cognitive, linguistic, and psychological sciences at Brown University, on Freakonomics Radio episode How to Change Your Mind.

The research about creating useful abstractions is Gick, M. L., & Holyoak, K. J. (1983). Schema induction and analogical transfer. Cognitive Psychology, 15(1), 1–38. https://doi.org/10.1016/0010-0285(83)90002-6

The research mentioned in the distributed practice is Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095-1102. https://doi.org/10.1111/j.1467-9280.2008.02209.x

About the student model and forgetting curves from Duolingo blog in How we learn how you learn.

The Biggest Myth In Education is a great video from Veritasium on YouTube in which the myth about individual learning styles is debunked. All relevant research papers are linked in the description of the video.

Back to top ▲