5 Effective Tips for Those Who Say ‘I Can’t Remember the Books I Read’: What Does the Research Say? (2026)

Summary. Finishing a book and remembering almost nothing of it a few weeks later is a common and normal experience. Ebbinghaus’s forgetting curve was replicated in 2015 by Murre and Dros: the “savings” gained when relearning the same list fell from 0.472 after 20 minutes to 0.168 after 6 days. The good news: in Roediger and Karpicke’s (2006) experiment, students who closed the text and tried to recall it remembered 56 percent of it a week later, while those who reread the text remembered 42 percent. Rowland’s (2014) meta-analysis, which pooled 159 effect sizes, put this “testing effect” at an average of g = 0.50. CEO move: tie every book to a single decision and record that decision in writing. Student move: close the book, write down what you remember, and repeat the same thing two days and one week later.

You can find the table that ranks the techniques one by one by strength of evidence in the evidence ranking of 12 study techniques, the details of review intervals in spaced repetition beyond flashcards, why beliefs like “I’m a visual learner” don’t hold up in the learning styles myth, and how to carry what you read into your next round of learning in second-order learning.

Why was this article rewritten?

This page was first published in 2021 and offered five tips: taking notes on the book itself, copying quotes into a separate notebook, marking striking pages with sticky notes, reading in different places, and listening to music while reading. The text cited no sources; claims such as “our mind remembers through its perception of place” and “music makes remembering easier” were not grounded in research. In the 2026 update the title and search intent were kept, but the content was rewritten from scratch:

• Each of the old five tips was checked against primary sources that were opened and read. The results are in the “old tips” table below.
• Unsourced generalizations were removed. In place of the music and place tips, we wrote what the meta-analyses on these topics actually found.
• A measurable post-reading protocol and a checklist were added.
• An official dataset (U.S. Bureau of Labor Statistics, American Time Use Survey 2025) was used for reading time.

How fast do we actually forget what we read?

The most cited source on forgetting is Hermann Ebbinghaus’s work from 1880 and 1885. Ebbinghaus learned lists of nonsense syllables on himself and measured how much time he saved when relearning the same list at set intervals. This measure is called “savings”: 0 means you worked as hard as learning the list from scratch; 1 means you forgot nothing.

Jaap Murre and Joeri Dros of the University of Amsterdam replicated this experiment in PLOS ONE in 2015. A single participant learned lists over roughly 70 hours and relearned them at intervals ranging from 20 minutes to 31 days. The authors conclude that the curve did replicate and probably shows an upward jump at the 24-hour point.

Table 1. Ebbinghaus’s original data and the 2015 replication: savings ratio in relearning

Time elapsed	Ebbinghaus (original)	Dros (2015 replication)
20 minutes	0.582	0.472
1 hour	0.442	0.373
9 hours	0.358	0.276
1 day	0.337	0.317
2 days	0.278	0.230
6 days	0.254	0.168
31 days	0.211	0.041

Source: Murre and Dros (2015), PLOS ONE, Table 3. Figures as published; the Mack and Seitz columns of the table are not shown here. The authors use 19 minutes, 63 minutes and 8.75 hours for Ebbinghaus’s first three intervals.

This table corrects two common misreadings. First, the savings ratio is not “percent remembered.” A savings of 0.337 at 1 day does not mean you forgot two thirds of what you read; it shows that relearning took roughly a third less time than the first learning. Second, these data come from nonsense syllables, not from a meaningful book; the figures should not be read directly as a book’s retention rate. The authors also note that they could not spread the lists for the 31-day condition across the experiment period, so the 0.041 value in the replication should be read with caution.

Still, the lesson is clear: if you do nothing, the loss is fastest in the first hours and days. That is why all of the tips below are about “what you do after reading.”

The “we remember 10 percent of what we read” claim

The “learning pyramid” often seen in articles on reading and memory (ratios like we remember 10 percent of what we read and 20 percent of what we hear) is not a research finding. Letrud and Hernes (2018), in an extensive search of digital libraries, showed that versions of the pyramid have circulated in educational debate for more than 160 years and that the models did not originate from empirical research. According to the authors, the conception of memory these models assume departs markedly from the standard picture of human memory. For this reason, these ratios are not used in this article.

Tip 1: Close the book and write down what you remember

The most common thing people who can’t retain what they read do is reread the important section. Research shows that trying to recall (retrieval practice, the testing effect) is more durable.

In the experiment Roediger and Karpicke published in Psychological Science in 2006, 120 undergraduates at Washington University read short texts on scientific topics. One group reread the text; the other closed it and wrote down everything they remembered, without feedback. On the initial test, students recalled roughly 70 percent of the text. The results reversed depending on when the measurement was taken.

Table 2. Roediger and Karpicke (2006), Experiment 1: proportion of idea units recalled

Timing of final test	Reread group	Recall group	Effect size (d)
After 5 minutes	81%	75%	0.52 (favoring rereading)
After 2 days	54%	68%	0.95 (favoring recall)
After 1 week	42%	56%	0.83 (favoring recall)

Source: Roediger and Karpicke (2006), Psychological Science 17(3), Experiment 1 results. Figures as published.

The table shows the mechanism behind the “I can’t remember what I read” complaint. Rereading feels good right afterwards, because 5 minutes later you really do remember more. A week later the picture reverses. The authors stress one more detail: a week later, the recall group remembered as much as the reread group remembered after two days.

CEOtudent analysis: a different reading of the same data

Measure	Reread group	Recall group
Loss from 5 minutes to 1 week (points)	39	19
Relative loss versus the 5-minute level	48.1%	25.3%
Recall gap at 1 week (recall minus rereading)		+14 points

CEOtudent analysis, recalculated from the Roediger and Karpicke (2006) Experiment 1 figures (calc.py). Relative loss = (rate at 5 minutes minus rate at 1 week) / rate at 5 minutes.

A single experiment is not enough, so we need to look at meta-analyses. Rowland (2014) pooled 159 effect sizes from 61 studies reported between 1975-2013 and found an average testing effect of g = 0.50 (95% confidence interval 0.42-0.58). Adesope, Trevisan and Sundararajan (2017) likewise concluded that practice tests are more beneficial for learning than restudying and all other comparison conditions.

Rowland’s subanalyses also tell you how to do it in practice. We have collected them in Table 3 below, together with reading-time data.

Tip 2: Recall again days later, not on the same day

The value of retrieval practice grows when it is combined with spacing. The meta-analysis by Cepeda and colleagues (2006) examined 839 assessments of spaced practice from 317 experiments in 184 articles. The main finding: the review interval that gives the best results gets longer the longer you want to retain the information. If you have an exam tomorrow, a short interval is enough; if you want to use a book’s ideas a year from now, your reviews should also be spread over days and weeks.

Rowland’s data point the same way: when the final test came at least 1 day later, the testing effect was g = 0.69; when it came in less than 1 day, g = 0.41. In other words, the benefit of retrieval practice grows in exactly the time window in which books slip from our minds.

In practice, this means a short recall when you finish a chapter, a second recall 1-2 days later, and a third recall a week later. We explained in detailed tables how to apply intervals to skills beyond flashcards in the spaced repetition article; we do not repeat that table here.

Tip 3: Explain it to yourself instead of underlining

The old article’s first three tips (taking notes in the book, copying quotes into a notebook, marking pages) were really variations of the same thing: marking the text. Research is clear on this.

Dunlosky, Rawson, Marsh, Nathan and Willingham (2013) reviewed 10 learning techniques in Psychological Science in the Public Interest. Only two techniques received a high utility rating: practice testing and distributed practice. Elaborative interrogation, self-explanation and interleaved practice received a moderate utility rating. Summarization, underlining/highlighting, the keyword mnemonic, imagery for text and rereading received a low utility rating. According to the authors, most students say they reread and highlight, but these techniques do not consistently boost performance.

Underlining does no harm; the problem is that on its own it is not enough. For the line to work, a second step is needed: explaining in your own words why the sentence is true and what it connects to.

There is evidence for this second step too. The meta-analysis reported in Bisra’s doctoral thesis at Simon Fraser University found an average of g = 0.55, across 69 effect sizes from 64 research reports, when students were prompted to self-explain while studying or solving problems. Self-explanation has a simple formula: after each important section, answer the questions “Why is this so?” and “What that I already know can I connect this to?”

A note on summarization: Dunlosky and colleagues placed summarization in the low utility group because its benefit was limited to some students and some tasks. That is why our recommendation is not “open the book and write a summary” but, as in Tip 1, close the book and write down what you remember. From the outside the two look the same, but the second is a retrieval exercise.

Tip 4: Read the text you want to remember on paper and without rushing

On the screen-versus-paper question, two meta-analyses point the same way:

• Delgado, Vargas, Ackerman and Salmerón (2018) reviewed studies from 2000-2017, 38 between-participant and 16 within-participant designs, with 171,055 participants in total. Paper showed a comprehension advantage over digital reading (g = -0.21; a negative value favors paper). Paper’s advantage grew when reading under time pressure, and did not appear in studies that used only narrative texts. The advantage did not shrink over the years; it grew.
• Clinton (2019) pooled 33 randomized studies from 2008-2018 (2,799 participants). Reading from screens lowered reading performance compared with paper (g = -0.25). The effect was clear for informational texts (g = -0.32) and almost absent for narrative texts (g = -0.04). No reliable difference in reading time was found (g = 0.08). Readers judged their own performance more accurately on paper (g = 0.20).

The last finding is an important part of the “I can’t remember it” problem. When reading on a screen, it is easier to misjudge how much you have understood. The practical upshot: reading a novel on your phone is no great loss; if you want to retain a business book, a report or a contract, read it on paper and without time pressure. If you have to read on a screen, the recall step in Tip 1 becomes even more important.

Tip 5: Tie every book to a decision

This tip is the CEOtudent lens itself. A CEO reads a report not for the sake of having read it, but for the decision taken after it. The same applies to books: giving a written answer to the question “What will I do differently after this book?” combines two evidence-backed techniques in a single step. To write the answer you have to recall the book (retrieval practice), and you have to connect the idea to your own work (elaborative interrogation and self-explanation).

Honesty is needed here: the “decision journal” practice itself is not a directly tested method. It is the evidence-backed techniques above adapted to a work context, in other words a CEOtudent editorial framework. You can find the data-driven version of the idea of carrying what you read into your next round of learning in the second-order learning article.

Write a single line for each book in your decision journal: the book’s title, one idea, one decision you will take with that idea, and when you will check the decision. A week later, without looking at that line, try to recall three ideas from the book. Compare what you recalled with the line; this comparison corresponds to the feedback condition, which showed the numerically largest testing effect in Rowland’s meta-analysis.

What happened to the old tips: place and music

Place. The old article said that reading in different places makes remembering easier. Smith and Vela’s (2001) meta-analysis found that the effect of environmental context on memory is reliable; in other words, remembering can be a little easier in the place where you learned. But the same meta-analysis showed that focusing on non-contextual cues while learning, and mentally reinstating the context while recalling, reduce this effect. This finding suggests that for a reader who thinks about the meaning of the text, place is a weak lever. Reading in a café can be enjoyable; but what makes a book stick is not the place but the recall you do after reading.

Music. The old article said that music makes remembering easier and that classical music can liven up a dull book. The evidence is mixed. Kämpfe, Sedlmeier and Renkewitz (2011), in their meta-analysis on adults, found that background music disrupts the reading process and has small negative effects on memory. A more recent meta-analysis (de la Mora Velasco and colleagues, 2023), with 71 effect sizes from 47 studies, found a small positive average effect on learning (d = 0.314); in studies where music was played before the learning assessment, the effect rose to a moderate size. Kämpfe and colleagues also reported that music has a positive effect on emotional reactions. Reading the two results together, the reasonable inference is this: music can be used to lift your mood before you start reading, but playing it in the background while reading a difficult text is not a reliable memory aid.

CEOtudent analysis: the old 5 tips against the evidence

2021 tip	What the evidence says	Verdict
Taking notes on the book, underlining	Highlighting received a low utility rating (Dunlosky et al. 2013)	Not enough on its own; combine it with self-explanation
Copying quotes into a notebook	Copying is not recall; closing the book and writing is recall (Roediger and Karpicke 2006)	Closed-book free recall instead of quotes
Marking pages with sticky notes	A marking method on its own; highlighting and rereading received low utility ratings (Dunlosky et al. 2013)	Write a question on the marked page, recall the answer later
Reading in different places	Context effect is reliable but weakens with meaning-focused processing (Smith and Vela 2001)	Harmless but not a priority
Listening to music while reading	Can disrupt the reading process (Kämpfe et al. 2011); there is also a finding of a small positive effect on learning (2023)	Yes before reading, no in the background with a difficult text

CEOtudent editorial assessment; the evidence in each row rests on the opened primary sources in the Sources section.

How much time do we actually spend reading?

To judge the cost of the recall protocol, you need to know how much time goes to reading. One of the most detailed public, regularly updated sources is the U.S. Bureau of Labor Statistics (BLS) American Time Use Survey. The 2025 results, published on June 25, 2026, give average hours per day for the population aged 15 and over. This is U.S. data; we use it here to give a sense of scale.

Table 3. Reading and the testing effect: published figures

Indicator	Value	Source
Reading, average for ages 15 and over (hours/day)	0.27	BLS ATUS 2025, Table 11A
Reading, full-time workers (hours/day)	0.17	BLS ATUS 2025, Table 11A
Reading, bachelor’s degree and higher, ages 25 and over (hours/day)	0.39	BLS ATUS 2025, Table 11A
Reading, ages 75 and over (hours/day)	0.62	BLS ATUS 2025, Table 11A
Watching TV, average for ages 15 and over (hours/day)	2.61	BLS ATUS 2025, Table 11A
Watching TV, full-time workers (hours/day)	1.92	BLS ATUS 2025, Table 11A
Testing effect, full dataset	g = 0.50	Rowland 2014
Testing effect, initial test success at most 50 percent, no feedback	g = 0.03 (not reliable)	Rowland 2014
Testing effect, initial test success above 75 percent	g = 0.56	Rowland 2014
Testing effect, studies with feedback	g = 0.73	Rowland 2014
Testing effect, published and unpublished studies	g = 0.58 and g = 0.25	Rowland 2014

Sources: U.S. Bureau of Labor Statistics, American Time Use Survey 2025 Results, Table 11A; Rowland (2014), Psychological Bulletin. Figures as published. BLS notes that because of the federal government shutdown from October 1-November 12, 2025, no time diaries are available for the period September 30-November 11, 2025, and that the estimates were reweighted.

Converting hours into minutes makes the table more concrete: the average person reads about 16.2 minutes a day, a full-time worker about 10.2 minutes, and a bachelor’s degree holder or higher about 23.4 minutes. On an average day the time spent on TV is about 9.7 times the time spent reading; for full-time workers this ratio is about 11.3. (CEOtudent analysis, calc.py.)

The lower part of the table says how the protocol should be built. In Rowland’s data, when success on the first recall attempt stayed at 50 percent or below and there was no feedback, the testing effect was not reliably different from zero. This suggests that putting off your first recall until weeks later, and managing to remember very little without feedback, will not do much good. The first recall should come early, and where you cannot remember, you should go back to the book and check. The smaller effect in unpublished studies is also a good reminder not to overstate the effect.

Practical tool: a 15-minute post-reading recall protocol

For someone who reads an average of 10-16 minutes a day, adding hours to every book is not realistic. The protocol below fits the findings above into about 15 minutes for a chapter or a short book. The times are editorial assumptions, not measurements.

CEOtudent editorial framework: post-reading recall protocol

Step	When	What you do	Time (assumption)	Finding it rests on
1. Close and write	As soon as the chapter ends	Close the book and write down the main ideas you remember as bullet points	3 minutes	Testing effect (Roediger and Karpicke 2006; Rowland 2014)
2. Check	Right afterwards	Open the book and correct missing and wrong points	2 minutes	Largest effect in studies with feedback (Rowland 2014)
3. Explain	Same day	For the most important idea, answer “why is it true?” and “what does it connect to?”	2 minutes	Self-explanation (Bisra; Dunlosky et al. 2013)
4. Second recall	1-2 days later	Rewrite the main ideas without looking at your notes, then check	3 minutes	Spaced practice (Cepeda et al. 2006)
5. Decision and third recall	1 week later	Recall three ideas and write a single line in your decision journal	5 minutes	Larger testing effect at least 1 day later (Rowland 2014)
Total			15 minutes

The times are editorial assumptions (calc.py). For comparison: according to BLS 2025 data, 15 minutes equals about 1.5 days of a full-time worker’s average reading time.

Post-reading checklist

• [ ] After finishing the chapter, I closed the book and wrote down at least three ideas.
• [ ] I compared what I wrote with the book and corrected what was missing.
• [ ] I explained the most important idea to myself with the question “why?”
• [ ] 1-2 days later, I tried to recall it again without looking at my notes.
• [ ] A week later, I tied the book to a single decision and wrote it in my decision journal.
• [ ] If I am reading this book for work, I read it on paper and without time pressure.
• [ ] I wrote a question next to the sentences I underlined instead of just underlining them.

What this does not mean

• You don’t have to remember every book. A novel read for pleasure does not need a protocol. The protocol is worth it when you want to use a book’s ideas.
• Rereading is not forbidden. The figures show that rereading works in the short term. The problem is that at the end of the measured period, a week later, the advantage reverses.
• The screen is not the enemy. The effects are small and almost zero for narrative texts. The effect is an average; it is not the same for everyone.
• A single figure is not a personal prediction. Meta-analytic averages come from different materials, ages and tests. Rowland’s data show high heterogeneity among effect sizes.
• An AI summary does not replace recall. In retrieval practice, the effect comes from you trying to bring the information back. Use the assistant only to compare the recall note you wrote with the book.

Frequently asked questions

Why can’t I remember the books I read?
Forgetting is how memory normally works; never retrieving the information after reading speeds up that loss. Ebbinghaus’s forgetting curve and its 2015 replication show that without review the loss is rapid in the first hours and days. Closing the book after reading and trying to recall it reduces this loss markedly.

Is it better to read a book twice, or read it once and try to recall it?
In the long run, trying to recall it. In Roediger and Karpicke’s experiment, the reread group was ahead after 5 minutes (81 percent versus 75 percent) and fell behind after a week (42 percent versus 56 percent).

Does underlining work?
On its own, only to a limited extent. Dunlosky and colleagues placed highlighting and underlining in the low utility group. Writing a question next to the line and answering it later without looking at the book turns marking into a retrieval exercise.

Will I remember an e-book or a printed book better?
For informational texts, the printed book gives a small advantage: Delgado and colleagues found g = -0.21, Clinton g = -0.25 (a negative value favors paper). No clear difference was found for novels.

Does listening to music while reading improve recall?
The evidence is mixed. A 2011 meta-analysis found that background music disrupts reading, while a 2023 meta-analysis found a small positive average effect on learning. Turning the music off while reading a difficult text is the safer choice.

How far apart should I review what I read?
It depends on how long you want to retain the information. According to the meta-analysis by Cepeda and colleagues, the best review interval gets longer as the target retention period gets longer. A simple start: the same day, 1-2 days later, and a week later.

Is “we remember 10 percent of what we read” true?
No. These ratios come from “learning pyramid” models and, according to Letrud and Hernes’s (2018) historical review, did not originate from empirical research.

Sources

1. Roediger, H. L., III and Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3), 249-255.
2. Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin, 140(6), 1432-1463.
3. Adesope, O. O., Trevisan, D. A. and Sundararajan, N. (2017). Rethinking the Use of Tests: A Meta-Analysis of Practice Testing. Review of Educational Research, 87(3). (Abstract only, ERIC record EJ1141817.)
4. Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T. and Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. (Abstract only.)
5. Murre, J. M. J. and Dros, J. (2015). Replication and Analysis of Ebbinghaus’ Forgetting Curve. PLOS ONE, 10(7), e0120644. (Ebbinghaus’s original data taken from Table 3 of this article.)
6. Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J. and 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. (Abstract only.)
7. Bisra, K. Inducing self-explanation: A meta-analysis and experiment. Doctoral thesis, Simon Fraser University. (Thesis abstract only, SFU Summit repository.)
8. Delgado, P., Vargas, C., Ackerman, R. and Salmerón, L. (2018). Don’t throw away your printed books: A meta-analysis on the effects of reading media on reading comprehension. Educational Research Review, 25, 23-38. (Author copy, University of Valencia.)
9. Clinton, V. (2019). Reading from paper compared to screens: A systematic review and meta-analysis. Journal of Research in Reading, 42(2), 288-325. (Abstract only.)
10. Smith, S. M. and Vela, E. (2001). Environmental context-dependent memory: A review and meta-analysis. Psychonomic Bulletin and Review, 8(2), 203-220. (Abstract only.)
11. Kämpfe, J., Sedlmeier, P. and Renkewitz, F. (2011; online 2010). The impact of background music on adult listeners: A meta-analysis. Psychology of Music, 39(4), 424-448. (Abstract only.)
12. de la Mora Velasco, E., Chen, Y., Hirumi, A. and Bai, H. (2023). The impact of background music on learners: A systematic review and meta-analysis. Psychology of Music, 51(6), 1598-1626. (Abstract only.)
13. Letrud, K. and Hernes, S. (2018). Excavating the origins of the learning pyramid myths. Cogent Education, 5(1). (Abstract only.)
14. U.S. Bureau of Labor Statistics (2026). American Time Use Survey: 2025 Results (USDL-26-1022, June 25, 2026) and Table 11A: Time spent in leisure and sports activities for the civilian population by selected characteristics, 2025 annual averages.

Note: The tables “CEOtudent analysis: a different reading of the same data,” “CEOtudent analysis: the old 5 tips against the evidence” and “CEOtudent editorial framework: post-reading recall protocol” are CEOtudent analyses; all derived figures were recalculated with calc.py. The times in the protocol are editorial assumptions. The unsourced claims in the 2021 version (“our mind remembers through its perception of place,” “music makes remembering easier,” “classical music can liven up a dull book”) were removed and replaced with meta-analytic findings. Learning pyramid ratios of the “we remember 10 percent of what we read” kind are mentioned only to debunk them. Because the full text of the frequently cited journal article on self-explanation could not be opened, the effect size was taken from the author’s thesis abstract. The “no feedback” effect value for Rowland (2014) circulating in secondary sources was not used, because the PDF table could not be read reliably from the text and the value does not appear in the narrative section.

---

This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.

This post is also available in: