Reading on Screen vs Paper: What the Research Shows

Is Reading on Paper Better Than on a Screen?

For informational text, yes, slightly. Across 54 studies and 171,055 readers, people understood the same text a little better on the printed page than on a screen, a gap researchers call the screen inferiority effect; it didn't appear in studies that used only stories. Newer evidence narrows down when it appears: the gap is clear when the text has to be scrolled, and it shrinks to small, unreliable differences when it doesn't.

Those 54 studies come from a 2018 meta-analysis by Pablo Delgado, Cristina Vargas, Rakefet Ackerman, and Ladislao Salmerón in Educational Research Review, with a title that gave away the conclusion: "Don't throw away your printed books." They pooled studies from 2000 to 2017 and asked a simple question. When people read the same text on paper and on a screen, who understands it better?

Paper won. The advantage was modest (Hedges' g = −0.21, where the minus sign means paper did better; about a fifth of a standard deviation), and both kinds of study design produced the same estimate, though individual studies varied widely. Three moderators matter more than the headline.

First, the advantage held for informational text, the kind you read to learn something: textbooks, reports, news analysis, documentation. In studies that used only narrative text, it didn't appear. Second, the screen penalty grew under time pressure and narrowed with self-paced reading. Third, and most surprising, the paper advantage grew over the publication years in the dataset. The studies from the 2010s showed a bigger gap than the studies from the early 2000s.

That third finding undercuts the most common dismissal of this research: that screen inferiority is a generational artifact that will fade as digital natives take over. In Delgado's data the trend ran the other way, and later meta-analyses found no change over time in either direction (Schwabe et al., 2022; Salmerón et al., 2024). Whatever the gap is, it doesn't seem to be fading as readers get more practice with screens.

A year later, Virginia Clinton published "Reading from paper compared to screens: A systematic review and meta-analysis" in the Journal of Research in Reading. She pooled 33 randomized experiments with 2,799 participants and found the same direction: a paper advantage for comprehension (g = −0.25), concentrated in expository text (g = −0.32), with no difference for narrative text (g = −0.04) and no reliable difference in reading time. She also measured something most studies skip, and it may be the most practically important finding in this literature: screen readers were worse at judging their own comprehension. We'll come back to that.

Then, in January 2026, the same researcher, now publishing as Virginia Clinton-Lisell, and Christine Litzinger released a network meta-analysis that compared devices with each other as well as with paper, and split the results by whether readers had to scroll. Its results get their own section below.

Here's the evidence base in one table.

A note on signs: for comprehension, the 2018, 2019, and 2024 meta-analyses report a paper advantage as a negative number, while the 2026 network meta-analysis reports it as a positive one. This article keeps each source's sign, so read the direction from the words.

These are the studies this article leans on, not the whole literature. At least a dozen meta-analyses have compared comprehension on paper and on screens since 2018, including several the table leaves out (among them Kong et al., 2018; Öztop & Nayci, 2021; and Li & Yan, 2024). Most found a paper advantage, but not all. The exceptions looked only at stories or at young children's e-books (Savva et al., 2021; Schwabe et al., 2022; Liang et al., 2025), or found no overall difference (Fontaine et al., 2021, with health-professions students; Li & Yan, 2024, across 37 studies). That's where the nuances start.

---

The Screen Inferiority Effect: Five Nuances the Headlines Skip

If you stopped reading at the table above, you'd walk away with "paper is better," which is what most coverage says. That's roughly true and substantially incomplete. Five qualifications change what the finding means for you.

Narrative reading is mostly fine on screens. In 2022, Annika Schwabe and colleagues published a meta-analysis in Media Psychology focused exclusively on narrative texts: novels, short stories, anything with a plot. Across 32 studies and 2,239 participants, they found no significant effect of the reading medium, no difference between computers, e-readers, and tablets, and a positive effect when digital stories added multimedia or interactive functions. One plausible reason is that stories are easier and more intrinsically engaging, so readers process them deeply regardless of the surface.

One wrinkle: the 2024 handheld-device meta-analysis by Salmerón and colleagues found that genre didn't significantly change the result on tablets and e-readers (if anything, stories showed the larger gap: g = −0.21 versus −0.08 for informational text). The authors called that the most unexpected of their text-related findings. "Fiction is safe on screens" is the rule of the broader literature, not a guarantee for every device.

The effect is small. The pooled paper advantage in Delgado et al. is real but modest, on the order of a fifth of a standard deviation. That's meaningful at population scale and across a semester of studying. It is not "reading on screens is useless." A small average penalty can be offset by how you read, which is the argument of the second half of this article.

The researcher behind the 2019 meta-analysis is more careful than the coverage. In a 2019 Hechinger Report column, Jill Barshay explained that students in Clinton's underlying studies could only use the digital text the way they'd use paper. "My findings weren't fair to screens because the screens couldn't offer everything they could," Clinton said. "They were really just a shiny piece of paper."

Barshay also reported that Clinton tells students who prefer paper to order the printed book: "It's enough of a benefit that it's worth the paper and ink and the cost of the book." Clinton's 2026 network meta-analysis sharpened both points at once. The advantage is real, and it depends heavily on format.

For kids, design matters more than medium. Furenes, Kucirkova, and Bus's 2021 meta-analysis in Review of Educational Research covered 39 studies with 1,812 children aged 1 to 8. Digital books that differed from print only by being digital produced lower story comprehension, and an adult reading a print book with a child was more effective than digital enhancements a child used alone.

But digital books with story-congruent enhancements outperformed paper, while an embedded dictionary had no effect or a negative one on story comprehension (it did help vocabulary). The medium isn't destiny; the design and the behavior on top of it do real work.

High stakes can close the gap. Emely Hoch and colleagues tested the effect where it matters most to students: real end-of-module university exams. In their study, published online in 2025 in Learning and Instruction, 1,592 students took their exams on paper and 658 took them as e-exams on tablets. Scores didn't differ, and neither did how accurately students estimated their own scores. What differed was effort: e-exam takers reported investing less and rated the exam as less difficult. When the stakes force deep processing, the screen penalty on performance can vanish, even though the screen still feels easier.

So the honest summary isn't "paper good, screens bad." It's narrower and more useful: screens carry a small comprehension penalty for informational reading, the penalty is clearest when you have to scroll or rush, and it shrinks with paginated layouts, good design, and engaged behavior.

The more useful question now is why, because each mechanism points to its own fix.

---

Why Screens Underperform: The Shallowing Hypothesis

The leading explanation isn't about pixels, glare, or eye strain. Modern screens are typographically excellent. The explanation is about mindset.

Researchers call it the shallowing hypothesis: years of using screens for quick, fragmented, reward-driven interaction (messages, feeds, notifications, skimming) train an approach to screen text that defaults to shallow processing. When you pick up a printed book, the context signals "sustained attention." When you open the same content in a browser tab, the surrounding context, and your own history with that context, signals "extract the gist and move on." The medium cues the mindset, and the mindset determines the depth of processing.

If the hypothesis is right, it explains several of the findings. Time pressure makes screens worse because pressure pushes you toward your default processing style, and on screens the default is shallow. Narrative texts escape the penalty because a story pulls you into deep processing whether you intended it or not. It would also explain why the gap grew across the studies in Delgado's dataset. Not every prediction has held up, though: in the 2024 handheld analysis, tablets, which invite distraction, did no worse than single-purpose e-readers.

Rakefet Ackerman and Morris Goldsmith found a version of this in 2011, in the Journal of Experimental Psychology: Applied. When study time was fixed by the experimenter, screen and paper readers' test scores didn't differ. When readers controlled their own study time, screen readers scored lower. The authors suggested the main difference was metacognitive rather than cognitive: on screen, people predicted their own performance less accurately and regulated their study time more erratically. The screen didn't reduce their ability. It changed their behavior.

That distinction matters enormously, and it's the optimistic core of this article. If screens degraded comprehension through some intrinsic property of light-emitting surfaces, you'd be stuck. A mindset and a behavior pattern, though, can be deliberately overridden. The shallowing hypothesis says the problem is that screens invite skimming. The fix is to read on screens in ways that refuse the invitation. We cover what the attention research does and doesn't show in the attention span crisis, and how to cultivate the opposite mode in deep reading.

Two more specific mechanisms sit underneath the general mindset problem, scrolling and miscalibration, and both are directly fixable.

---

Scrolling vs Paging: Where the Gap Shows Up

When you read a printed book, every sentence has a fixed physical address. That claim you half-remember is on the bottom left, about a third of the way in, just after the chart. This sounds trivial. It isn't. Readers build a spatial map of a text and lean on it for memory and comprehension, the way you remember where things are in your kitchen without trying.

Scrolling demolishes the map. In a scrolling window, text has no stable location: the paragraph at the top of your screen is now in the middle, then gone. Your brain can't anchor content to place, so the work shifts onto working memory.

Christopher Sanchez and Jennifer Wiley tested this directly in a 2009 Human Factors paper titled "To Scroll or Not to Scroll." Readers studied a complex scientific text either in a scrolling interface or in discrete pages. Scrolling reduced understanding, and the damage concentrated in readers with lower working memory capacity. That pattern is exactly what you'd predict if scrolling forces working memory to compensate for the lost spatial scaffold: readers with capacity to spare absorb the tax, readers without it pay full price.

For years, though, scrolling was a plausible suspect with thin evidence behind it. Delgado's 2018 meta-analysis found a larger screen penalty in studies that used scrollable texts, but the difference wasn't statistically significant. Clinton-Lisell and Litzinger's 2026 network meta-analysis, published in Education and Information Technologies, gave the question its biggest test yet. They pooled 56 studies and 79 effect sizes on expository text, mostly with college students reading science texts in English, compared paper with four kinds of device (computers, tablets, e-readers, and smartphones), and split the comparisons by whether readers had to scroll.

When scrolling was necessary, paper beat computers and tablets by g = 0.35 and 0.38, well above the classic estimate of about 0.21. When no scrolling was needed, because the text fit on the screen or was turned page by page, the differences between paper and any device were small and unreliable. (None of the studies had readers scroll on an e-reader.) Much of what we call the screen penalty may be, specifically, a scroll penalty plus a mindset penalty, and you can reduce both.

Two caveats. On handheld devices alone, Salmerón's 2024 meta-analysis found no significant difference between paging and scrolling, though only five of its effect sizes involved scrolling and they leaned the same way (g = −0.15 for scrolling versus −0.11 for paging). And in the 2026 analysis, scrolling wasn't assigned at random: whether readers scrolled was coded from each study's procedure, so other differences between those studies ride along with it. The split shows where the gap appears, not that scrolling alone causes it, and the authors present scrolling as a possible key reason rather than a proven one.

Practical consequence number one of this entire article: when comprehension matters, page, don't scroll. Use reader modes that paginate, e-readers in page mode, PDF viewers in single-page view. Give your spatial memory something to hold.

---

Phone, Tablet, E-Reader, or Laptop: Does the Device Matter?

Somewhat, but less than the way the device makes you read. Across all conditions, the 2026 network meta-analysis ranked paper first and tablets second, with e-readers and computers tied and smartphones last. Here's how far behind paper each device came out overall.

Two cautions keep this from turning into a buying guide. First, few studies tested the newer devices: only six compared e-readers with paper, and three compared smartphones. Second, the decisive split in the 2026 data was scrolling versus no scrolling rather than one device versus another. When nothing had to be scrolled, e-readers and smartphones showed the same small, unreliable gap behind paper (g = 0.12 each).

Salmerón and colleagues' 2024 handheld analysis points the same way. Comparing studies that used tablets with studies that used e-readers, they found no significant difference (g = −0.10 for tablets and −0.16 for e-readers, against print). For fiction, Schwabe's meta-analysis found no difference between computers, e-readers, and tablets either.

The 2024 handheld analysis found one more pattern that matters for students. In studies that compared separate groups of readers, the print advantage was concentrated in undergraduates (g = −0.23), with no significant effect for primary and secondary school students (g = 0.06, a slight, non-significant lean toward screens). The authors offer two possible reasons: university students have had more years of shallow screen habits, and they tend to face harder, longer texts.

They also warn against reading that as "children are fine." The studies that tested the same readers on both media showed no difference by educational stage, and the only two studies with large, representative samples, of fourth and eighth graders, both found a significant print advantage.

What about phones? The smallest screen means the most scrolling, and the device most associated with feeds and messages may cue the shallowest mindset. Switching that same text to a paginated view, or to a tablet or e-reader if you have one, is the cheapest fix available.

---

The Overconfidence Trap

This mechanism is the hardest to catch, because it's invisible from the inside.

Clinton's 2019 review didn't just measure comprehension. It also measured calibration: how accurately readers judge their own understanding. Across 11 studies with 698 participants, readers' judgments were better calibrated on paper than on screens (g = 0.20 in paper's favor). Clinton read the gap as overconfidence: screen readers may "think they are understanding the text better than they are," as she put it in the paper (quoted by Education Week).

Think about what that does to self-regulated reading. Your decisions about when to stop studying, when to re-read, and when you're ready for the exam all run on your internal sense of "I've got this." If the screen inflates that sense, you stop too early, skip the re-read, and walk in confident and underprepared. The comprehension penalty is small. The calibration penalty makes you blind to it, so you never compensate.

Ackerman and Goldsmith found that on screen, people predicted their own performance less accurately and regulated their study time more erratically. One plausible reason is that they lean on a quick, fluent feeling of "this is easy" rather than on effortful self-monitoring. Fluency is a notoriously bad signal of learning, a trap we unpack in the illusion of competence. Text can feel smooth and familiar while leaving nothing behind, and screens, with their skim-trained mindset, amplify exactly that illusion.

The 2025 exam study is the exception: with a grade on the line, tablet takers judged their own scores as accurately as paper takers. One reasonable reading is that high stakes can override the screen's pull toward shallow processing. Ordinary studying rarely supplies that pressure, so you have to supply it yourself.

The fix for miscalibration has been known to learning science for a century: stop trusting the feeling and test yourself. Close the tab and write three sentences about what you just read. If you can't, you didn't understand it, no matter how smooth it felt. This is the testing effect, and it does double duty here, both strengthening memory and showing you what you actually know. Our guide on how to remember what you read goes deep on the retrieval side.

---

How to Read Better on Screens

Now the practical half. The mechanisms above point to five countermeasures, each mapped to a specific cause.

1. Page instead of scrolling. This attacks the scrolling cost directly, and it's the fix the 2026 network meta-analysis speaks to most clearly. Use your e-reader's page mode. Use reader view in your browser. Open PDFs in single-page view rather than continuous scroll. For long web articles, keyboard paging (the space bar jumps a full screen) approximates pagination. Sanchez and Wiley's low working-memory readers, the people most hurt by scrolling, stand to gain the most.

2. Put serious reading on the right screen. When you have a choice, read demanding text on a tablet or an e-reader rather than a phone or a laptop window you have to scroll. Moving a long article to a bigger, paginated surface removes most of the scrolling without printing anything.

3. Remove the clock. Time pressure was one of the three moderators in the Delgado meta-analysis: the screen penalty was larger when reading was time-constrained (the 2024 handheld analysis didn't find this on tablets and e-readers). You can't always remove deadlines, but you can stop importing artificial ones. Don't read important material in the ten minutes before a meeting, and don't set a "finish by" timer for dense text. Self-paced reading strips out a condition under which screens do the most damage, and it's the heart of slow reading as a deliberate practice.

4. Calibrate with self-testing. This attacks overconfidence. After each section of anything that matters, look away and summarize it in a sentence or two. The point isn't the summary. It's the moment of failure when you discover you can't produce one, which is exactly the information the screen's fluency illusion was hiding from you. Readers who self-test can re-read the right sections instead of the comfortable ones.

5. Highlight and annotate, actively. This one aims at the shallow screen mindset. You cannot skim and annotate at the same time. Deciding what deserves a highlight forces you to evaluate, compare, and prioritize, which is generative processing, the same family of operations behind the testing effect and elaborative learning.

Highlighting on its own does little, though. Dunlosky and colleagues' 2013 review of study techniques rated it low utility because it doesn't consistently boost performance. And when Gal Ben-Yehudah and Yoram Eshet-Alkalai tested it as a fix for the screen gap in 2018, telling readers to highlight a college-level expository text, it improved answers to inferential questions on paper but didn't improve comprehension on screen. So treat highlights as a way to mark what you'll come back to and test yourself on, not as a fix on their own. We've written a full breakdown of when highlighting works and when it doesn't in the science of highlighting.

Where screens can pull ahead is what happens after the first read, because paper can't do what digital annotation can. A highlight in a printed book is trapped on the page. A highlight made with Glasp's web highlighter is captured and searchable, which leaves you a record to come back to; trying to recall each point before you reread the highlight is what turns that review into retrieval practice. Using multiple highlight colors adds another layer of forced judgment: deciding whether a passage is evidence, a counterargument, or a question to chase requires categorizing it, and categorizing is processing.

Reading a study or report as a PDF? Glasp's PDF highlighter puts those highlights in the same searchable library. And if your long-form reading happens on a Kindle, syncing your Kindle highlights into that library means your most paper-like screen reading feeds the same review loop.

Most of these cost nothing beyond a change in how you move through the text.

---

The Screen Reading Protocol

Here's the whole thing as a protocol you can run today on any text that matters. It adds maybe 15 percent to your reading time and targets every mechanism in the research.

Steps 4 through 7 are the active-engagement core. The reason to run them digitally is leverage: on paper, step 7 means flipping through old books hoping to stumble on your margin notes. In a tool built for it, your highlights from articles, PDFs, and books sit in one searchable place, so the review loop actually happens. That loop, not the highlight itself, is where long-term retention comes from.

One honest caveat: no study has tested this exact protocol end to end. The research supports most of its pieces separately: paging over scrolling, self-paced over pressured, retrieval over re-reading. Highlighting is the exception, which is why step 4 only counts together with steps 5 to 7.

---

When to Deliberately Choose Paper

Reading better on screens doesn't mean reading everything on screens. The evidence supports keeping paper in the rotation for specific jobs, and choosing deliberately beats defaulting either way.

Choose paper for long, dense, high-stakes expository reading. A textbook chapter you'll be examined on, a contract, a 60-page report you must genuinely understand. Expository text is where the paper advantage is most consistent, and one meta-analysis found it only for longer texts (Li & Yan, 2024). It's also where miscalibrated confidence costs the most. As Clinton put it to The Hechinger Report: "Sometimes you should print it out, especially if it's long."

Choose paper when the only digital version scrolls. If a long document only exists as a scrolling web page and you'd have to read it on a phone, you're reading scrolled text, the condition with the clearest paper advantage in the 2026 data. Print it, or move it to a paginated view first.

Choose paper when you're too depleted to run the protocol. The screen countermeasures cost effort. At 11 p.m., after a full day, you won't page, self-test, and annotate. Paper's affordances do some of that regulation for you for free: fixed spatial layout, no notifications, a physical sense of progress.

Screens are fine, sometimes better, for fiction. Schwabe's meta-analysis found no penalty for stories. Read novels wherever you enjoy reading novels. Enjoyment drives volume, and volume drives reading skill more than medium ever will.

Screens win when the reading feeds a system. If what you read needs to connect to what you've read before, get retrieved months later, or get shared and discussed, digital's advantages (search, sync, accumulation, sharing) outweigh a small comprehension penalty you're already countering with the protocol. A slightly worse first pass that enters a review loop beats a slightly better first pass that evaporates.

A useful rule of thumb: paper optimizes the single reading session, screens optimize the reading system. Most knowledge work lives or dies by the system.

---

Frequently Asked Questions

Is reading on a screen really worse than paper?

For informational, study-style text, yes, by a small margin. Most meta-analyses published since 2018 found a paper advantage, anchored by Delgado et al. (2018) with 54 studies and 171,055 participants; the ones that didn't mostly looked at stories or young children's e-books, or found no overall difference. The newest, Clinton-Lisell and Litzinger (2026), found the gap is clear when you have to scroll and small and unreliable when you don't. For narrative text, Schwabe et al. (2022) found no screen penalty.

What is the screen inferiority effect?

The screen inferiority effect is the finding that people tend to understand a text slightly less when they read it on a screen than when they read the same text on paper. It is best documented for informational text, it is larger under time pressure and when the text must be scrolled, and it comes with a second cost: screen readers judge their own understanding less accurately. The average gap is about a fifth of a standard deviation, so it's a tendency rather than a law, and reading behavior can offset it.

When is the advantage of paper largest?

When the text is informational, when it has to be scrolled, and when you're reading against the clock. Delgado et al. (2018) found the paper advantage larger under time pressure and absent in narrative-only studies, though the 2024 handheld meta-analysis by Salmerón and colleagues didn't find the time-pressure effect on tablets and e-readers. Clinton-Lisell and Litzinger (2026) found it reliable for computers and tablets when scrolling was needed (g = 0.35 and 0.38). In the Salmerón analysis, studies that compared separate groups of readers found it larger for university students than for school students, but the authors caution that this doesn't mean children are unaffected.

Why do I remember less when reading on my phone?

Three likely mechanisms. Your phone may cue a skim-trained mindset built by years of feeds and messages (the shallowing hypothesis). A small screen means constant scrolling, which robs you of the spatial map paper gives you for free, and the 2026 network meta-analysis found paper's advantage was clearest when scrolling was needed. And screens make it harder to judge what you've understood (Clinton, 2019), so you tend to stop before you've understood as much as you think. In everyday use, notifications pile on top.

Are e-readers better than phones for reading comprehension?

Possibly, but the evidence is thin. No study in the 2026 network meta-analysis compared e-readers with phones directly. In its overall ranking, e-readers tied with computers, below tablets and paper and above smartphones, and when no scrolling was needed, e-readers and smartphones showed the same small, unreliable gap behind paper (g = 0.12 each). Salmerón et al. (2024) found no significant difference between studies using tablets and studies using e-readers, both slightly behind print. Whatever the device, page mode avoids scrolling, the condition where most of the gap appeared in the 2026 data.

Does scrolling really hurt reading comprehension?

It appears to. Sanchez and Wiley (2009) found that a scrolling format reduced understanding of complex web text, especially for readers with lower working memory capacity. The 2026 network meta-analysis of 56 studies found paper reliably ahead of computers and tablets when scrolling was necessary (Hedges' g = 0.35 and 0.38) and only small, unreliable differences when it wasn't (g = 0.03 to 0.12). Scrolling wasn't randomly assigned in those studies, so treat it as the leading suspect rather than a proven cause. Paging, reader mode, and single-page PDF view are the cheapest fixes.

Will the screen gap disappear as digital natives grow up?

There's no sign of it so far. Delgado et al. (2018) found the paper advantage grew across publication years from 2000 to 2017, and later meta-analyses (Schwabe et al., 2022; Salmerón et al., 2024) found no change over time. What the gap does track is format and behavior: the 2026 network meta-analysis found only small, unreliable differences when readers didn't have to scroll. Treat the gap as real, modest, and dependent on how you read rather than on when you were born.

---

Conclusion

The science is clear enough to act on and nuanced enough to deserve better than the usual headline. Paper holds a real, modest comprehension advantage for informational reading, there's no sign of the gap fading, and screen readers compound the problem by misjudging what they've understood. But fiction mostly escapes the penalty, and the newest device comparison found that when readers didn't have to scroll, paper's lead shrank to small, statistically unreliable differences.

So don't print everything. Page instead of scrolling, put serious reading on your best screen, drop the time pressure, test yourself instead of trusting the feeling of fluency, and highlight selectively, then try to recall what you marked. Save paper for the long, dense, high-stakes reads where its affordances do the regulating for you.

If you want to start today, install Glasp's web highlighter and run the protocol on the next article that actually matters to you: paginate, highlight only what surprises you, and write a two-sentence summary at the end. A week later, try to recall the main points before you open your highlights, then check what you still remember. That one loop, repeated, is how you make what you read on screens stick.