Do you understand less when you read on a screen?
Skim a long report on your laptop with a meeting in ten minutes, and you'll probably take in a little less than you would from the same pages printed. Read a novel at your own pace, and the screen makes almost no measurable difference. Two large reviews of the research show that pattern. The gap is small, about a fifth to a quarter of a standard deviation, and it sits mostly in texts that explain or argue, read against the clock.
The two reviews
The larger of the two reviews is by Delgado, Vargas, Ackerman and Salmerón (2018) in Educational Research Review. They gathered studies from 2000 to 2017 that gave people the same text on paper or on a digital device: 38 compared two groups, 16 tested the same people on both, and together they covered 171,055 participants. Paper came out ahead in both kinds of design, with an effect size (Hedges' g) of 0.21. The gap grew when reading was timed rather than self-paced. It held for informational texts and for mixes of informational and narrative ones, but not for studies that used narrative texts alone. And it was larger in the later studies than in the earlier ones.
Virginia Clinton's 2019 review in the Journal of Research in Reading was stricter. She kept only experiments from 2008 to 2018 that assigned readers to paper or screen at random: 33 studies with 2,799 people. The pooled effect was close to Delgado's (0.25 in favour of paper), and the split by genre was sharper. Expository texts gave 0.32; narrative texts gave 0.04, which is next to nothing. Screen readers weren't faster, either: across 14 studies, reading time came out about the same.
A novel on an e-ink reader
An e-ink screen behaves differently from a monitor in other ways (the post on e-ink and your eyes covers the light), so one experiment with a real e-reader is worth a close look, and it fits Clinton's narrative figure. Mangen, Olivier and Velay (2019) had 50 young adults read a 28-page Elizabeth George mystery of about 10,800 words, half on a Kindle DX and half in a printed pocket book laid out from the same PDF. The two groups took the same time (about 58 minutes), recognised words and sentences from the text equally well, remembered the characters and places equally well, and felt equally absorbed in the story.
Where they parted was time and order. Asked when things happened, the Kindle readers got 44% right against 57% for the print readers, and they made more mistakes when asked to put 14 events of the plot in sequence. The authors suggest that the thickness of a book and the shift of pages from one hand to the other help a reader place events, while an e-reader gives mostly visual cues. They call the work largely exploratory. Their set-up also took away some of the help an e-reader normally gives: page numbers were removed from both versions and the Kindle's progress bar was hidden, and only two of the 50 readers were experienced Kindle users.
Why the clock matters
The timed reading in Delgado's review is where the screen loses most, and two experiments explain some of that. Ackerman and Goldsmith (2011) fixed the study time in one experiment and let people choose it in a second. With the time fixed, people did as well on screen as on paper. Once they chose how long to study, the screen group scored lower, predicted their own results less accurately and regulated their time more erratically. The authors put the difference down to metacognition, how people judged and paced their own study, rather than to how well they processed the text. Clinton later pooled this kind of self-prediction (calibration) across 11 studies and found it better on paper too.
Delgado and Salmerón then tested the clock directly in a study published in 2020. Undergraduates at a Spanish university read a 3,010-word magazine article, either in print or as a PDF on a 17 inch desktop monitor, with free time or with 16 minutes 30 seconds. With free time, the two groups scored about the same on a 14-question test (67% in print, 68% on screen). Under the limit, print readers slipped to 64% while screen readers fell to 50%. Calibration didn't differ between the groups this time, but attention did: probes during the reading showed that print readers' minds wandered less when they were hurried, and screen readers' minds didn't.