It is very common to see someone in libraries photograph another person’s notes, download a beautiful mind map from the internet, and feel like they have already studied. But the problem is that what fixes an idea in memory is not its packaging, but having built it yourself. Most students keep doing the opposite, out of laziness or because nobody has explained to them the difference between feeling productive and learning.
Self-explanation consists of stopping at each step of a piece of material and justifying why it makes sense, aloud or in writing, reasoning about the content.
Elaborative interrogation is its close cousin: you ask yourself “why is this true?” or “what would happen if I changed this condition?” and look for the answer before checking the book.
Well-built mind maps are diagrams that connect concepts in hierarchies and relationships; badly built ones are decoration with arrows. All three techniques share one principle: they force you to generate information, not copy it. That act of generation makes the difference.
The foundational study on self-explanation was published by Chi and her team in 1989 [1]. They worked with physics students who solved problems using worked examples. They divided the participants into two groups: those who tended to self-explain spontaneously and those who did not. The former solved later problems better, even when they started with less prior knowledge. It wasn’t talent: it was what they did with the material. Those who explained to themselves why each step was valid and why a given formula was used built a network of reasons; those who only read were left with loose steps. A 2018 meta-analysis of 64 studies and more than 6,000 participants [2] confirms that inducing self-explanation has a moderate positive effect on learning, with an effect size of g ≈ 0.55.
A 2025 meta-analysis [14], which analyzed 56 studies in digital learning environments, confirms these findings with a similar effect (g = 0.46) and adds an important nuance: self-explanation is especially powerful when students set their own learning pace, rather than when the system imposes the timing. It isn’t magic, but it is a real effect, replicated in very different contexts.
The 2013 meta-analysis by Dunlosky and his team [3] reviewed hundreds of studies on ten study techniques. Self-explanation and elaborative interrogation have moderate utility: they work, but not always, and they depend on how they are applied. Highlighting and rereading, the students’ favorites, have low utility. Retrieval practice (closing the book and trying to recall) appears as the most powerful technique, with effects that often double retention compared to rereading. The authors themselves point out that most students use the least effective techniques because they are easier and give an illusion of fluency. Fluency is not learning.
Karpicke and Blunt published a now-classic experiment in 2011 [4]: they compared students who built concept maps while studying with others who practiced active retrieval. One week later, the retrieval group outperformed the concept-map group on retention and comprehension tests. The maps gave a sense of mastery that did not match the results. The students who made maps felt more confident about their learning than those who practiced retrieval, but performed worse. Confidence not only fails to predict learning: sometimes it masks it. An earlier meta-analysis by Nesbit and Adesope [5] had found that concept maps have a moderate effect, but with a crucial condition: the benefit appears mainly when the students build them themselves, not when they are handed ready-made ones. Copying someone else’s map is like watching someone lift weights: it doesn’t make you stronger.
When do these techniques really work? When they force you to produce answers, make mistakes, and correct them. Self-explanation works best with material that has an internal logic (mechanics, programming, physiology) and worse with lists of vocabulary or dates. Renkl tested this in Germany with secondary school students learning mathematics from worked examples [6]. Those who self-explained learned more, but only if their explanations focused on the underlying principles and not on repeating the steps. Explaining “first you calculate the net force, then the acceleration” is of little use; explaining “why the net force determines the acceleration according to Newton’s second law” leaves a mark. Elaborative interrogation has a similar profile: it works with expository texts rich in causal relationships, such as history or biology, but not with arbitrary material. Pressley and his colleagues demonstrated this as early as 1987 [7]: students who asked themselves “why is it so?” about facts in a text learned more than those who only read, but the effect disappeared if the material lacked a clear causal structure. There is no universal technique. There is a universal question, “what does this have to do with what I already know?”, and each technique answers it in its own way.
A 2025 study adds an uncomfortable nuance [8]. Schindler and Richter tried to replicate the generation effect with a trivial generation task: reordering sentences that were already given, without producing any content of their own. Many activities that seem active (moving cards around, rearranging other people’s outlines, coloring maps) generate nothing new in your head. They are sophisticated forms of copying. A 2025 study using pupillometry [15] (measuring pupil diameter as an indicator of mental effort) supports this reading: it confirms that the benefit of generating information lies not in the activity itself, but in the real cognitive effort it demands. If the task is trivial, no matter how much you move pieces around, your pupil doesn’t dilate and your memory doesn’t improve.
Handwritten Notes vs Laptop
The case of handwritten notes versus laptop is the perfect example of how a seemingly trivial practice hides a generative mechanism. Mueller and Oppenheimer published a study in 2014 [9] that went viral: students who took notes by hand answered conceptual questions better than those who used a laptop, even though the latter wrote more words. The explanation: the keyboard encourages verbatim transcription, while the hand forces you to paraphrase, select, and reorganize. Paraphrasing is a form of self-explanation. A 2019 replication by Morehead, Dunlosky and Rawson [10] did not find the same advantage for handwritten notes. The discrepancy has a plausible explanation: in the replication, participants could review their notes before the exam, which reduced the advantage of deep elaboration during note-taking. The critical variable is not the medium, but what you do with it. If you use the laptop to transcribe like a stenographer, you learn little. If you use it to summarize, paraphrase, and ask yourself questions, you learn just as well or better. A 2024 meta-analysis by Flanigan and her team [16] comes to settle the debate: after reviewing 24 international studies, it concludes that handwritten notes are associated with better academic performance (Hedges’ g = 0.248), even though fewer words are written than with a keyboard. Paper retains an aggregate advantage, precisely because it makes verbatim transcription harder and encourages paraphrasing. A 2021 Japanese study with functional magnetic resonance imaging [11] provides an interesting neurobiological finding: in a sample of university students and young adults, taking notes on paper activated the hippocampus and prefrontal cortex more than doing so on a tablet or phone. These are regions associated with memory and deep encoding. It doesn’t prove causation, but it fits the behavioral data: paper encourages slower, more elaborate processing.
Curiously, it has also been shown that retrieval practice produces benefits even without feedback. Roediger and Karpicke demonstrated this in 2006 [12] with university students: one group read a text four times, another read it once and then tried to recall it three times without looking. A week later, the retrieval group remembered 61% of the material, compared to 40% for the rereading group. That is 52% more. And during the study, those who practiced retrieval felt less sure of what they knew. Difficulty is not a flaw of the method; it is the method. A 2025 study (Corral and Carpenter) [17] extends this finding: retrieval practice improves not only verbatim retention but also the ability to transfer those concepts to new problems, provided there are enough rounds of practice and measurement takes place after several days. In addition, a 2024 meta-analysis (Bego and collaborators) [18] applied to nine introductory science courses found mixed results: in some courses (such as certain calculus or chemistry subjects) spaced retrieval improved performance, but in others no benefit was observed. The aggregate analysis showed a significant positive effect that invites further exploration of this low-cost method, but with the warning that it is not a universal magic bullet. Bjork and Bjork [13] called this “desirable difficulties”: conditions that make learning slower and clumsier in the moment, but multiply long-term retention. Copying other people’s notes, highlighting, rereading: all of that is fluid, comfortable, and almost useless. That said, a 2024 theoretical review (Pyke, Lunau and Javadi) [19] qualifies this concept: not every difficulty is desirable. They propose a model that integrates desirable difficulties with cognitive load theory; extra difficulty helps when the material is of low complexity, but it can be counterproductive and overload the novice learner when faced with very dense material with high interactivity between its elements. The key is to match the difficulty to the learner’s level.
Why does generating your own connections work better than copying someone else’s? I have two theories.
The first: personalization creates idiosyncratic retrieval cues. When you explain something in your own words, you associate the concept with experiences, mistakes, and examples that only you have. Those associations act as access routes in memory. If you copy someone else’s map, you use their routes, not yours, and in an exam you have no access to their mind.
The second: the act of generating information produces a cognitive commitment effect. Your brain values more what has cost it effort to produce. This has been observed in error-based learning paradigms: making a mistake and then correcting it produces better memory than getting it right the first time, as long as there is feedback. Generating connections, even wrong ones, leaves a deeper trace than receiving the correct answer prepackaged. These are reasonable conjectures, not established facts.
On the other hand, we can think that if generation is the mechanism, then any tool that takes it away from you, including automatic AI summaries, could be working against you. Not because AI is bad, but because it steals your sweat. You can use a machine-generated summary to get your bearings, but if all you do is read it, you are in the same territory as the person photographing notes.
Practical Guide
First, replace rereading with retrieval. Close the book and write down everything you remember. Then check, correct, and repeat. It is uncomfortable and it works.
Second, when you study a worked example, cover it up and ask yourself why each step was taken. Don’t move on to the next one until you can justify it in your own words.
Third, if you are going to make a mind map, build it yourself from scratch, without looking at the source, and then compare it with the material. The map is not a product to keep: it is a retrieval exercise disguised as a diagram.
Fourth, take notes with any tool, but forbid yourself from transcribing verbatim. Paraphrase, summarize, jot down questions and tentative answers. If you use a laptop, turn off autocorrect and, if necessary, type without looking at the screen.
Fifth, use elaborative interrogation as a closing ritual: at the end of each session, write three “why” questions about what you studied and answer them without consulting anything.
Keep in mind that students systematically overestimate the effectiveness of passive techniques. In the Dunlosky meta-analysis [3], rereading and highlighting were the most popular techniques despite having the worst evidence. The authors ask themselves why, and the answer is as simple as it is unsettling: fluency feels good, and real learning feels clumsy. If you finish a study session feeling that you understood everything without effort, you probably haven’t learned much. If you finish exhausted, with doubts and the feeling that you know nothing, you are on the right track.
Could an artificial intelligence tool be designed that, instead of giving you the summary, asks you questions so that you generate the answers yourself? Current technology prioritizes fluency and comfort, which are exactly the enemies of learning. But there is no physical law that prevents building an assistant that makes you sweat.
The next time someone offers you their perfect notes or a pastel-colored mind map, thank them and turn it down (but not solved exams from previous years, that is important material). Because they are handing you a shortcut that leads nowhere. Learning is generating. And generating is, by definition, something only you can do.
References
[1] Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13(2), 145–182.
Label: Reliable.
Methodology: Observational study with verbal protocols. Physics students solved problems while thinking aloud; the frequency and quality of their self-explanations were coded and correlated with later performance. Small sample (n=10), but the effect has been replicated in multiple contexts.
[2] Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., & Winne, P. H. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30(3), 703–725.
Label: Reliable.
Methodology: Meta-analysis of 64 experimental and quasi-experimental studies with more than 6,000 participants. Calculated overall effect sizes and moderators such as age, domain, and type of self-explanation instruction.
[3] Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4–58.
Label: Reliable.
Methodology: Systematic review and meta-analysis of more than 700 studies on ten learning techniques. The authors rated each technique according to the available evidence and its applicability in educational contexts.
[4] Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science, 331(6018), 772–775.
Label: Reliable.
Methodology: Randomized controlled experiment with 200 university students. One group studied with retrieval practice (writing what they remembered) and another with concept maps. Retention was measured one week later through recall and comprehension tests.
[5] Nesbit, J. C., & Adesope, O. O. (2006). Learning with concept and knowledge maps: A meta-analysis. Review of Educational Research, 76(3), 413–448.
Label: Reliable.
Methodology: Meta-analysis of 55 studies comparing learning with concept maps versus other techniques. Analyzed variables such as who built the map, the type of material, and the educational level.
[6] Renkl, A. (1997). Learning from worked-out examples: A study on individual differences. Cognitive Science, 21(1), 1–29.
Label: Reliable.
Methodology: Study with German secondary school students who learned mathematics from worked examples. Their self-explanations were recorded and the analysis examined which types of explanations best predicted success on new problems.
[7] Pressley, M., McDaniel, M. A., Turnure, J. E., Wood, E., & Ahmad, M. (1987). Generation and precision of elaboration: Effects on intentional and incidental learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13(2), 291–300.
Label: Reliable.
Methodology: Series of controlled experiments with university students. They manipulated whether participants generated their own elaborations or received those of others, and measured free recall and incidental memory.
[8] Schindler, J., & Richter, T. (2025). Does text generation improve learning from expository text? A conceptual replication attempt. Cognitive Research: Principles and Implications, 10, Article 34.
Label: Recent.
Methodology: Seven experiments with independent samples comparing a trivial generation task (reordering given sentences) with a reading condition. No generation effect on memory was found, and in some experiments a disadvantage was even observed. The study questions the interpretation that any generative activity improves recall.
[9] Mueller, P. A., & Oppenheimer, D. M. (2014). The pen is mightier than the keyboard: Advantages of longhand over laptop note taking. Psychological Science, 25(6), 1159–1168.
Label: With reservations (finding with mixed replication).
Methodology: Three experiments with university students randomly assigned to take notes by hand or with a laptop during a lecture. Memory for facts and conceptual understanding were assessed half an hour later.
[10] Morehead, K., Dunlosky, J., & Rawson, K. A. (2019). How much mightier is the pen than the keyboard for note-taking? A replication and extension of Mueller and Oppenheimer (2014). Educational Psychology Review, 31(3), 753–780.
Label: With reservations (replication that does not reproduce the original effect).
Methodology: Replication with a larger sample (n=342) and a design that allowed participants to review their notes before the exam. It found no significant differences between handwriting and laptop, which suggests that the advantage depends on the assessment conditions and on whether review is allowed.
[11] Umejima, K., Ibaraki, T., Yamazaki, T., & Sakai, K. L. (2021). Paper notebooks vs. mobile devices: Brain activation differences during memory retrieval. Frontiers in Behavioral Neuroscience, 15, 634158.
Label: Reliable.
Methodology: Neuroimaging (fMRI) study with 48 Japanese participants (university students and young adults) who completed a memory task after taking notes on paper, tablet, or mobile phone. Brain activation during retrieval was compared. The paper group showed greater activation in the hippocampus and prefrontal cortex.
[12] Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
Label: Reliable.
Methodology: Two experiments with university students who read texts and were then assigned to rereading or to repeated recall tests. Retention was measured at 5 minutes and at one week, with and without feedback. At one week, the retrieval group recalled 61% versus 40% for the rereading group.
[13] Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher et al. (Eds.), Psychology and the Real World: Essays Illustrating Fundamental Contributions to Society (pp. 56–64). Worth Publishers.
Label: Theoretical framework.
Methodology: Theoretical chapter synthesizing decades of research on memory and learning. It presents no new empirical data, but articulates the concept of “desirable difficulties,” widely supported by experimental studies.
[14] Tan, L.-P., Gong, S.-Y., Wang, Y.-J., et al. (2025). Enhancing Academic Performance through Self-Explanation in Digital Learning Environments (DLEs): A Three-Level Meta-Analysis. Educational Psychology Review, 37(1), Article 20.
Label: Recent.
Methodology: Three-level meta-analysis with 204 effect sizes extracted from 56 studies, focused on digital environments. Confirms the effect of self-explanation and finds that it is more effective at learner-paced settings and for conceptual knowledge.
[15] Grudzien, A. M., & Unsworth, N. (2025). Investigating a mental effort explanation of the generation effect using pupillometry. Memory & Cognition, 54, 980–996.
Label: Recent.
Methodology: Series of experiments using pupillometry (measuring pupil diameter as an indicator of mental effort) to investigate the generation effect. Confirms that the benefit of generation is tied to real cognitive effort and not to mere generative activity.
[16] Flanigan, A. E., et al. (2024). Typed Versus Handwritten Lecture Notes and College Student Achievement: A Meta-Analysis. Educational Psychology Review, 36(3), Article 78.
Label: Recent.
Methodology: Meta-analysis of 24 international studies comparing the academic performance of students who take notes by hand versus those who use a laptop. Finds a significant advantage for handwritten notes (Hedges’ g = 0.248).
[17] Corral, D., & Carpenter, S. K. (2025). Effects of retrieval practice on retention and application of complex educational concepts. Learning and Instruction, 100, 102219.
Label: Recent.
Methodology: Three experiments with more than 700 participants that evaluated the effect of retrieval practice not only on retention but also on the transfer of complex concepts to new contexts, with measurements after one week.
[18] Bego, C. R., Lyle, K. B., Ralston, P. A. S., et al. (2024). Single-Paper Meta-Analyses of the Effects of Spaced Retrieval Practice in Nine Introductory STEM Courses: Is the Glass Half Full or Half Empty? International Journal of STEM Education, 11, Article 9.
Label: Recent.
Methodology: Aggregate-data meta-analysis of nine introductory science (STEM) courses in which spaced retrieval practice was implemented. Results were mixed across courses, although the aggregate analysis showed a significant positive effect, suggesting that it is a promising but not uniformly effective method.
[19] Pyke, W., Lunau, J., & Javadi, A.-H. (2024). Does difficulty moderate learning? A comparative analysis of the desirable difficulties framework and cognitive load theory. Quarterly Journal of Experimental Psychology, 78(10), 2181–2195.
Label: Recent.
Methodology: Theoretical review and integrative model proposal comparing the desirable difficulties framework with cognitive load theory. Proposes that difficulty must be adjusted to the complexity of the material and the learner’s level of expertise in order to be beneficial.
Note on references:
The “With reservations” labels on [9] and [10] reflect a legitimate debate in the literature about the advantage of handwritten notes. The practical conclusion of the post is based on the convergence of independent studies on the role of paraphrasing and generation, not on the dispute over the specific medium. The new references [14] to [19] are studies published between 2024 and 2025; all have been verified against the primary source, with the sole exception of the reference