The screen is costing you less than you think, and something stranger than you think. Across 54 studies and 171,055 people, paper beats screens on comprehension by an effect size of 0.21 — small, real, and nowhere near the reason a chapter has already evaporated by Thursday. What the screen reliably takes is your judgement of how well you understood. On paper and on screen people perform the same when someone else controls the clock. Let them decide for themselves when they’ve got it, and the screen readers stop too early, feeling fine.
Which means the fix isn’t a better way to read. It’s a check that arrives after you’ve stopped feeling clever:
- Decide before you start what you’d have to be able to say afterwards. One sentence. What is this chapter’s claim, and what is it arguing against?
- Read normally. Don’t print it out, don’t fight the device, don’t buy a highlighter system. The medium is not where you’re losing this.
- Close it, wait, then say the chapter’s argument out loud from memory. Not right away — later that evening, or the next morning. The delay is the active ingredient, and so is the closing.
- Notice the gap between how well you thought that would go and how it went. That gap is the actual thing the screen gave you. Repeat the attempt a few days later.
Everything below is the evidence for that, the part of the standard advice that measurably doesn’t work, and what the Kindle’s own AI features can and can’t do about it.
Do you remember less reading on a screen than on paper?
A little, yes — and the honest version of that answer is much less dramatic than the headlines.
The largest pooled estimate comes from Pablo Delgado, Cristina Vargas, Rakefet Ackerman and Ladislao Salmerón, who analysed studies published between 2000 and 2017 comparing comparable texts on paper and on digital devices: 38 between-participant and 16 within-participant designs, 171,055 participants in total. Both design types produced the same result — Hedges’ g = −0.21 in favour of paper (Educational Research Review 25, 2018, 23–38 — record). An effect of that size means paper helps, and it also means the medium is a minor character.
On the devices most people actually read books on, it’s smaller still. Salmerón, Altamura, Delgado, Karagiorgi and Vargas pooled 49 studies of handheld devices — smartphones, tablets and e-readers — against paper, covering more than 160,000 participants in total: g = −0.113 in the between-participant comparisons and g = −0.103 in the within-participant ones (Journal of Educational Psychology 116(2), 2024, 153–172 — summary). That is roughly half the disadvantage found for reading on a computer.
Two details from the larger meta-analysis are worth more than the headline number:
| Moderator | What the data showed |
|---|---|
| Text type | Paper’s advantage held for informational texts and mixed sets, and was absent in studies using narrative texts only |
| Time frame | Paper’s advantage was larger under time pressure than when reading was self-paced |
| Publication year | Paper’s advantage has increased across the years studied, not shrunk |
The first row is the one that should get your attention if you read to learn things. The effect lives in exactly the kind of text you’re uploading and doesn’t show up in the novel you read on the same device for fun. The third row quietly kills the comforting story that we’re all adapting to screens and this will sort itself out. Between 2000 and 2017, it got bigger.
Then what does the screen actually cost you?
Not comprehension, on the evidence. Calibration.
In 2011 Rakefet Ackerman and Morris Goldsmith ran the experiment that still frames this whole literature. Participants studied expository texts of 1,000 to 1,200 words on screen or on paper, predicted how well they would do, then took a multiple-choice test. The results split cleanly by who controlled the clock:
- Under a fixed study time, test performance did not differ between the two media.
- When participants regulated their own study time, screen performance was worse.
- On screen, people showed a consistent bias toward overconfidence — predicting they’d done better than they had — in both time conditions.
Their conclusion, in their own framing, is that the primary differences between screen and paper “are not cognitive but rather metacognitive”: less accurate prediction of performance, and more erratic regulation of study time on screen (Journal of Experimental Psychology: Applied 17(1), 2011, 18–32 — record).
Sit with what that means for the way you actually read. Nobody sets your study time. There is no test on Friday. You decide when you’ve understood a chapter, and then you turn the page — and that decision is precisely the one the screen degrades. Reading for your own curiosity is the most exposed case, not the least, because self-regulation is the only mechanism you have.
This is a different failure from the one audio imposes. Listening costs you the inferences, because you can’t stop and go back — that argument, and what to do about it, is in how to remember what you listen to in audiobooks. The screen leaves your comprehension roughly intact and takes away your ability to tell how much of it you have. And if the more basic question is why anything you read disappears at all, regardless of medium, that’s a different diagnosis.
Why the usual digital-reading advice doesn’t fix it
Pool the advice from the pages that rank for this question — a research-software vendor’s blog, a university study-skills blog, a test-prep tutor, a summary app, a company that sells a brain-sensing headband — and you get the same seven moves every time. Kill your notifications. Scroll to the end first to fake the spatial feel of paper. Set a purpose. Highlight. Summarise in your own words. Take a break to reflect. Review after 2, 7 and 30 days.
Some of that is genuinely useful, and the useful parts survive below. But look at the shape of the list: every single item happens while the text is in front of you. They are all input-side fixes for what the research says is an output-side problem. None of them makes you find out whether you actually have it.
There’s a second problem. Highlighting, summarisation and rereading are three of the ten techniques Dunlosky, Rawson, Marsh, Nathan and Willingham assessed in their review of learning strategies, and all three landed in the low utility band, while practice testing and distributed practice were the two rated high utility (Psychological Science in the Public Interest 14(1), 2013, 4–58 — abstract). The most-recommended advice for reading on screens is drawn disproportionately from the bottom of that list. (The same review is why the case for dropping the notebook entirely is stronger than it sounds.)
And “review after 2, 7 and 30 days” only counts if the review is a retrieval attempt. Rereading your highlights on day 7 is re-exposure. It feels like consolidation, and its main measured effect is to make you more confident — which, on a screen, is the thing you already have too much of.
The one thing that closed the gap in a lab
In 2014 Avital Lauterman and Rakefet Ackerman went looking for something that would remove screen inferiority rather than describe it. Participants read six texts, on screen or on paper. Two things worked: gaining experience with the task, and being required to generate keywords summarising the essence of the text after a delay.
The keyword requirement eliminated screen inferiority even for the first text studied, and the paper reports that experience and in-depth processing attenuate overconfidence (Computers in Human Behavior 35, 2014, 455–463 — record).
Read that intervention closely, because it is not “take better notes”. It is: the text is gone, some time has passed, and you have to produce its substance out of your own head. That is a retrieval attempt with a delay in front of it — and it is the exact move that none of the standard advice asks for.
One inconvenient result sits in the same study, and it cuts against the tidy version of this argument: after practising the keyword method, screen inferiority re-emerged among the participants who preferred studying on paper. The authors’ own conclusion is that how well a learning-enhancing method works depends on the study context and the learner’s preferences. So this is not a universal switch. If you genuinely think better on paper, that preference is real data about you, and printing the hard chapter is a reasonable thing to do.
What to do while you read on a screen
Short list, because this half matters less than the field thinks.
- Set the question first. One sentence, before you start, about what you should be able to say when you finish. It gives the next forty minutes something to attach to, and it gives you something specific to fail at later, which is the point.
- Put the device in one mode. The measured advantage of paper shrinks when reading is self-paced and unhurried; the screen’s worst condition is a hurried, interrupted, half-there one. A Kindle in aeroplane mode is a different instrument from a browser tab with eleven siblings.
- Highlight if you like, but demote it. Marking a passage is a bookmark, not a memory. Its real job is to let you find the thing again, and it does that well.
- Don’t print it out. For most informational reading the effect is too small to be worth the ritual, and the thing that actually closed the gap in the lab wasn’t paper.
What to do after you close it
This is where the recovery happens, and it’s four minutes.
- Close the device. Not minimise. The check is worthless with the text within reach, because recognising a sentence you’re looking at is not remembering it.
- Wait. An hour is enough; that evening is better. Lauterman and Ackerman’s intervention had a delay built into it, and delay is what separates a memory test from an echo.
- Say the chapter’s argument out loud, from memory. The claim and what it’s arguing against — not the anecdote you liked. Sixty seconds. Out loud, because typing lets you trail off and speaking doesn’t.
- Then check yourself, and pay attention to the surprise. Open the chapter and find the two things you didn’t say. The gap between “that went fine” and what actually came out is the overconfidence the screen handed you, made visible. It’s uncomfortable exactly once per chapter, and it’s the whole mechanism.
- Come back a few days later and do step 3 again. Same closed-book attempt, no rereading first. For how to space those attempts sensibly across a whole book, the interval question has its own answer, and there’s a by-hand chapter schedule if you want to run this yourself without any software.
That’s the entire method. Its only real problem is step 5: you have to remember to come back, about material you have just been shown you don’t know as well as you thought, with nobody asking you to.
Does the device matter — Kindle, phone, or laptop?
Somewhat, and less than device arguments suggest. Handheld devices show about half the disadvantage of computers in the pooled data above. Two honest caveats on that number, though:
- The “handheld” category in that meta-analysis pools smartphones, tablets and e-readers together. There is no clean e-ink-only figure in it, so nobody can honestly tell you that a Kindle Paperwhite specifically scores −0.11 against paper.
- The same analysis found the effect was stronger among undergraduates than among primary and secondary students, and stronger when people were assessed individually rather than in groups. Adults reading alone are not the group the effect is mildest in.
The practical read: an e-reader in a single-purpose state is a better environment than a phone that also contains your messages, and neither difference is large enough to be your retention strategy. What you do after you close it dominates what you closed.
What Kindle’s AI features, Readwise and the reading apps do — and don’t
The gap here isn’t a lack of tools. It’s that almost every tool aimed at this problem is built to show you the material again, which is the move the calibration research says inflates confidence.
Amazon’s “Ask this Book” launched on the Kindle iOS app in the US on 15 December 2025, enabled for thousands of best-selling English-language Kindle titles, with Kindle devices and Android following in 2026. You highlight a passage and ask about plot, characters or themes, and it answers only up to your current position in the book (9to5Mac). If you’re stuck inside a Kindle bestseller, this is a better-placed tool than we are: it’s already in the book, it’s from the company that sold it to you, and it costs you nothing extra. Its limits are structural rather than qualitative — it works on the enabled Kindle catalogue, not the PDF a colleague emailed you, and it answers your questions rather than asking you any.
Amazon’s “Recaps”, announced on 2 April 2025 for US customers across thousands of best-selling English-language eBooks and extended to web browsers by July 2026, is described by Amazon as providing “quick refreshers for books you’ve already read while preserving the joy of discovery, much like the ‘Previously on…’ segment before a show” (About Amazon). For picking up book four of a series after eight months, that is straightforwardly the right tool. It is also the cleanest possible example of the distinction this article turns on: a recap is re-exposure. You read it, everything feels familiar, and familiarity is the signal the screen already over-reports.
Readwise gathers the highlights you made on your Kindle and elsewhere and sends you a daily digest email of a selection of them. That’s a real improvement on highlights rotting in a file. It also inherits a ceiling: it resurfaces the passages you already chose, in the author’s words, for you to recognise. It says nothing about the rest of the chapter — the part you read without marking, which is exactly the part your own judgement quietly dropped.
And the tools that do ask questions — Loxie’s daily drills, Headway’s quizzes — mostly ask them about their own pre-built overviews and summaries, not about the book you chose.
What’s missing from all of it is a tool that takes a document you picked, on the screen you’re already using, and asks you to produce its content from memory after a delay. That absence is roughly why we built ours.
How Study Junkie replaces the judgement a screen degrades
Here’s the admission that belongs at the top of this section: Study Junkie is a screen. Reading a chapter in our app is screen reading. We don’t fix the medium, and anyone selling you a screen that fixes screens is selling you something.
What we replace is the thing the medium actually degrades — your own verdict on whether you understood it.
You upload the book, the PDF or the document (about two to three minutes for a 500-page book) and it becomes a structured course with generated chapters, with the full text preserved rather than summarised. You read it, or listen to the narration at 0.75× to 2× if your eyes are done. When you’re stuck, the tutor chat answers from your source rather than from the internet at large. So far, so ordinary: that’s all consumption, and consumption is the part you were already fine at.
Then the two things that map onto the research:
- Assignments generated from your chapter, with feedback on what you write. You produce the substance in your own words and something reads it back to you. That is the keyword intervention with a marker attached — the external verdict standing in for the internal one that reads high on a screen.
- Scheduled recall that comes back later, from the same source. Not a recap. Questions, after a delay, about material you didn’t choose to highlight — the parts of the chapter your own selection would have quietly dropped.
Neither is magic and neither is unique in isolation; several tools grade a written answer now. What’s uncommon is running both over the reader’s own book, preserved whole, on a schedule nobody has to maintain. And it removes the one failure point in the free method above: you don’t have to remember to come back.
It works in eight languages, and you can upload in one and study in another. Signing up gives you 1,500 free Study Credits with no card, and after that it’s one-time credit packs at $10, $25 or $50 that never expire — no subscription, because there’s no term to your curiosity.
If the four moves at the top of this page are enough on their own, use those; they cost nothing and they’re most of the benefit. If step 5 is where you always fall off, start a course from something you’re already reading and let the schedule do the remembering.
Frequently asked questions
Do you really remember less reading on a screen than on paper? A little. The pooled estimate across 54 studies and 171,055 participants is g = −0.21 in favour of paper (Educational Research Review 25, 2018), and about half that for handheld devices across 49 studies (Journal of Experimental Psychology 116(2), 2024). It’s a real but small effect, and it shows up in informational texts rather than narrative ones.
Is a Kindle better than a phone for remembering what you read? Probably, mostly for reasons that aren’t about the display. Handheld devices show roughly half the disadvantage that computers do, but that category pools smartphones, tablets and e-readers together, so there’s no honest e-ink-specific number to quote. The bigger difference is that a single-purpose device isn’t also delivering your notifications.
Should I print things out to remember them better? For most reading, no. The effect is small, printing is a real cost in time and paper, and the intervention that actually erased screen inferiority in a lab wasn’t printing — it was being made to produce the text’s essence from memory after a delay. The exception is worth naming: in that same study the screen disadvantage came back for people who preferred paper. If that’s you, print the hard chapter.
Does highlighting on a Kindle help you remember? Highlighting is a bookmark. It’s excellent for finding a passage again and weak as a memory technique — it was rated low utility in Dunlosky’s 2013 review. The related question of whether notes help at all has a more interesting answer than yes or no, and it’s here.
Why do I forget what I read on my phone specifically? Usually regulation rather than eyesight: interrupted attention, a hurried pace, and the overconfidence that lets you swipe on before the idea has landed. If nothing you read sticks regardless of device, that’s a different problem with a different fix — start here.
Does reading fiction on a screen have the same problem? On the available evidence, no. The paper advantage in the 2018 meta-analysis held for informational and mixed text sets and was absent in the studies that used narrative texts only. Read novels on whatever you like.