Underlining a page with a pencil used to mean one thing to me: a student cramming for an exam. Lately I notice the same gesture in a different context altogether — readers, well past school age, marking up novels and essay collections the way they once marked up textbooks. Not to memorise anything for a test. To slow themselves down long enough to notice what a sentence is actually doing before their eyes slide on to the next one.
That small, slightly self-conscious habit sits at the centre of something larger. The readers returning to slow, immersive reading are not doing it because they have decided screens are bad or that novels are morally superior to feeds. Most of them still read plenty online, still scroll, still skim. What they describe, when you ask, is closer to fatigue with their own reading habits than rejection of any device. They skimmed everything for years, got efficient at it, and found the efficiency was costing them something they wanted back.
To understand why that cost is real rather than sentimental, it helps to look at where the skimming habit came from in the first place, because it was not chosen so much as trained.
The skimming habit was built by the page, not the reader
In 2006, the Nielsen Norman Group ran an eye-tracking study of how people actually look at web pages, and found a consistent pattern: readers scan across the top of a block of text, drop down and scan a shorter line further in, then run their eyes down the left margin looking for anything that stands out. Researchers nicknamed it the F-shaped pattern, and a follow-up study eleven years later found it just as strong on phones as on desktops, and mirrored in right-to-left scripts. The follow-up research was blunt about the cause: users are not reading for pleasure or study, they are hunting for what is relevant, and most visitors read only about a quarter of the text on any page they land on.
That is a sensible way to use a browser, and it turns into a problem only when it is the only reading posture a person has left. A habit built for finding an answer fast does not switch off cleanly when the material in front of you is a novel or a long essay that asks to be followed rather than mined. I have edited enough manuscripts to recognise the readers who have lost that switch: they tell me they finish books without being able to say what happened in them, or that they reread the same paragraph three times because nothing from it stuck the first time.
What the comprehension research actually shows
It would be easy to dismiss this as anecdote, so it is worth looking at what happens when the same text is compared across formats under controlled conditions. In a 2013 study, the reading researcher Anne Mangen and colleagues at the University of Stavanger had 72 Norwegian tenth-graders read identical texts, one group in print, one group as PDFs on screen, then tested them on comprehension. The print readers scored significantly higher, particularly on questions that required tracking where information sat in the text and how events related to each other in sequence. Mangen’s explanation is not that screens are inferior objects, but that paper gives readers a stable physical and spatial sense of the text — its length, its shape, where they are within it — and that this scaffolding supports memory in ways scrolling does not.
The literacy scholar Maryanne Wolf made a related argument in 2018, describing a shift toward skim reading as our default mode even when we sit down with a book. In a widely discussed essay for the Guardian, Wolf warned that the habits built by scanning screens do not stay contained there; they migrate into how we approach every text, including ones that reward patience. Her worry was specific: sequencing and memory for detail, the very things Mangen’s students struggled with on screen, are also what deep reading depends on for building understanding that goes beyond the gist.
The return to paper is not nostalgia
Something in that argument seems to be landing, because the market is responding. Unit sales of print books rose for a second consecutive year in 2025, according to Circana BookScan data reported by Publishers Weekly, settling above pre-pandemic levels even if short of the 2021 peak. That is a modest swing, and it says nothing on its own about how carefully anyone is reading, but it fits what I hear from readers who have gone looking for print again after years of reading almost everything on a phone or tablet: they describe it less as taste and more as a corrective measure, a way of removing the option to skim because the object in their hands does not scroll, ping, or reward a quick scan of the first two lines.
None of this amounts to rejecting technology. The readers I am describing still bank, message, and read news on screens without apparent conflict. What they have done is separate one activity — sustained reading — from the rest of their digital habits and given it a different set of physical conditions.
What this means for how things get written
This matters for anyone who writes, not only for those who read. A great deal of advice for online writing, quite reasonably, is built around the F-pattern: front-load the point, break up the prose, put the important word first in every line. That advice serves a reader who has decided in advance to skim, and serves badly a reader who has chosen to stay with a piece of writing. A sentence that delays its payoff, a paragraph that asks you to hold two ideas in mind before it resolves them, a structure that rewards attention rather than accommodating its absence — these are not indulgences but the things that separate writing meant to be read from writing meant to be scanned.
Editing for one audience and editing for the other are different jobs. The scannable version wants clarity at the surface: short sentences, familiar phrasing, the conclusion up top. The version written for a reader determined to slow down can carry more weight in its syntax, because that reader has agreed to do some of the work of following it. Knowing which reader you are writing for changes almost every choice on the page, from where a sentence breaks to how much a paragraph is allowed to ask of the person reading it.
Slow reading is a skill that atrophies, not a mood to summon
I taught mathematics for long enough to know that concentration is a capacity that weakens with disuse and returns with practice, not with willpower applied in a single sitting. The readers rebuilding their patience for long, difficult, or simply unhurried texts are doing a kind of training: choosing books long enough that skimming stops working as a strategy, putting the phone in another room not out of virtue but because the habit needs the interruption removed, accepting that the first twenty pages of anything demanding will feel like friction before they feel like immersion.
That pencil, underlining a paragraph in a novel long after school has ended, is less about the notes it leaves behind than about the equipment it provides for a harder task: staying with one sentence until it has said what it came to say, rather than moving on before it finishes.
